CLEAN ELECTRICITY

FROM PHOTOVOLTAICS
2nd Edition

p798hc_9781848167674_tp.indd 1 4/6/14 10:44 am

 SERIES â•›ON â•›PHOTOCONVERSION â•›OF â•›SOLAR â•›ENERGY
ISSN: 2044-7701

Series Editor: Mary D. Archer (Cambridge, UK)

Vol. 1: Clean Electricity from Photovoltaics

eds. Mary D. Archer & Robert Hill

Vol. 2: Molecular to Global Photosynthesis

eds. Mary D. Archer & Jim Barber

Vol. 3: Nanostructured and Photoelectrochemical Systems

for Solar Photon Conversion
eds. Mary D. Archer & Arthur J. Nozik

Vol. 4: Clean Electricity from Photovoltaics, 2nd Edition

eds. Mary D. Archer & Martin A. Green

Catherine - Clean Electricity from Photovoltaics.indd 1 12/9/2014 4:38:15 PM

 Series on Photoconversion of Solar Energy — Vol. 4

CLEAN ELECTRICITY
FROM PHOTOVOLTAICS
2nd Edition

Editors

Mary D Archer
Imperial College, UK

Martin A Green
University of New South Wales, Australia

Imperial College Press

ICP

p798hc_9781848167674_tp.indd 2 4/6/14 10:44 am

 Published by
Imperial College Press
57 Shelton Street
Covent Garden
London WC2H 9HE

Distributed by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE

Library of Congress Control Number: 2014950883

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.

Series on Photoconversion of Solar Energy Â�— Vol. 4

CLEAN â•›ELECTRICITY â•›FROM â•›PHOTOVOLTAICS
Second Edition
Copyright © 2015 by Imperial College Press
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy
is not required from the publisher.

ISBN 978-1-84816-767-4

Typeset by Stallion Press

Email: [email protected]

Printed in Singapore

Catherine - Clean Electricity from Photovoltaics.indd 2 12/9/2014 4:38:15 PM

September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page v

This volume is dedicated

to

Peter Theodore Landsberg

8 August 1922–14 February 2010

who saw so far beyond the detailed-balance limit

v
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page vii

CONTENTS

About the Authors xiii

Preface to the First Edition xxi

Preface to the Second Edition xxv

1. The Past and Present 1

M. D. Archer
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Milestones in the development of photovoltaic technology . . . 6
1.3 The crystalline silicon solar cell . . . . . . . . . . . . . . . . . 8
1.4 GaAs and III–V multijunction cells . . . . . . . . . . . . . . . 10
1.5 Concentrator photovoltaics . . . . . . . . . . . . . . . . . . . . 12
1.6 Inorganic thin-film cells . . . . . . . . . . . . . . . . . . . . . 13
1.7 Organic and hybrid technologies . . . . . . . . . . . . . . . . . 16
1.8 Overview of photovoltaic cell operation . . . . . . . . . . . . . 19
1.9 Other junction types . . . . . . . . . . . . . . . . . . . . . . . 29
1.10 Structure of this book . . . . . . . . . . . . . . . . . . . . . . . 33
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

2. Limits to Photovoltaic Energy Conversion Efficiency 41

M. A. Green
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.2 Photovoltaic converters: essential requirements . . . . . . . . . 44
2.3 Thermodynamic properties of sunlight . . . . . . . . . . . . . . 46
2.4 ‘Top-down’ thermodynamic efficiency limits . . . . . . . . . . 49
2.5 Single-cell efficiency limits . . . . . . . . . . . . . . . . . . . . 52
2.6 Multiple-junction devices . . . . . . . . . . . . . . . . . . . . . 66
2.7 Other high-efficiency options . . . . . . . . . . . . . . . . . . . 70
2.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

vii
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page viii

viii Contents

3. Crystalline Silicon Solar Cells 87

M. A. Green
3.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.2 Silicon cell development . . . . . . . . . . . . . . . . . . . . . 89
3.3 Substrate production . . . . . . . . . . . . . . . . . . . . . . . 105
3.4 Cell processing . . . . . . . . . . . . . . . . . . . . . . . . . . 113
3.5 Cell costs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
3.6 Opportunities for improvement . . . . . . . . . . . . . . . . . . 124
3.7 Silicon-supported thin films . . . . . . . . . . . . . . . . . . . 127
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

4. Thin-Film Solar Cells based on Amorphous

and Microcrystalline Silicon 139
C. Ballif, M. Despeisse and F.-J. Haug
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
4.2 Basic properties of amorphous and microcrystalline
silicon alloys . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.3 Thin-film silicon solar cells . . . . . . . . . . . . . . . . . . . . 156
4.4 Fabrication of device-grade amorphous
and microcrystalline silicon . . . . . . . . . . . . . . . . . . . 162
4.5 Light management . . . . . . . . . . . . . . . . . . . . . . . . 168
4.6 Advanced device architecture and record devices . . . . . . . . 173
4.7 Industrialisation and large-area production technology . . . . . 183
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

5. Polycrystalline Cadmium Telluride Photovoltaic Devices 209

T. A. Gessert and D. Bonnet
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
5.2 Brief history of CdTe PV devices . . . . . . . . . . . . . . . . . 210
5.3 Initial attempts towards commercial modules . . . . . . . . . . 211
5.4 Review of present commercial industry/device designs . . . . . 213
5.5 General CdTe material properties . . . . . . . . . . . . . . . . 214
5.6 Layer-specific process description for superstrate
CdTe devices . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page ix

Contents ix

5.7 Where is the junction? . . . . . . . . . . . . . . . . . . . . . . 234

5.8 Considerations for large-scale deployment . . . . . . . . . . . . 236
5.9 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

6. Cu(In,Ga)Se2 and Related Solar Cells 245

U. Rau and H. W. Schock
6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
6.2 Material properties . . . . . . . . . . . . . . . . . . . . . . . . 247
6.3 Cell and module technology . . . . . . . . . . . . . . . . . . . 253
6.4 Device physics . . . . . . . . . . . . . . . . . . . . . . . . . . 271
6.5 Wide-gap chalcopyrites . . . . . . . . . . . . . . . . . . . . . . 288
6.6 Kesterite (CZTS) solar cells . . . . . . . . . . . . . . . . . . . 294
6.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295

7. Super-High-Efficiency III–V Tandem and Multijunction Cells 307

M. Yamaguchi
7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
7.2 Principles of super-high-efficiency multijunction solar cells . . . 310
7.3 Epitaxial technologies for growing III–V compound cells . . . . 314
7.4 Monolithic vs. multi-terminal connection modes . . . . . . . . 316
7.5 Key issues for realising high-efficiency multijunction
solar cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
7.6 High-efficiency InGaP/GaAs/Ge three-junction solar cells
and their space applications . . . . . . . . . . . . . . . . . . . 324
7.7 Multijunction solar cells: recent results . . . . . . . . . . . . . 330
7.8 Future directions . . . . . . . . . . . . . . . . . . . . . . . . . 333
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334

8. Organic Photovoltaics 339

D. Credgington
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
8.2 Basic concepts in organic semiconductors . . . . . . . . . . . . 340
8.3 Overview of organic photovoltaic devices . . . . . . . . . . . . 350
8.4 Device physics of organic photovoltaics . . . . . . . . . . . . . 359
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page x

x Contents

8.5 Beyond the state of the art . . . . . . . . . . . . . . . . . . . . 386

8.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
Acknowledgement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 400

9. Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 413

M. Grätzel and J. R. Durrant
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
9.2 Historical background . . . . . . . . . . . . . . . . . . . . . . 414
9.3 Mode of function of dye-sensitised solar cells . . . . . . . . . . 415
9.4 DSSC research and development . . . . . . . . . . . . . . . . . 426
9.5 Solid-state mesoscopic cells based on molecular dyes or
perovskite pigments as sensitisers . . . . . . . . . . . . . . . . 440
9.6 Pilot production of modules, field tests and commercial
DSSC development . . . . . . . . . . . . . . . . . . . . . . . . 441
9.7 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444

10. Quantum Well Solar Cells 453

J. Nelson and N. Ekins-Daukes
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453
10.2 Device design, materials and technology . . . . . . . . . . . . . 454
10.3 Physics of QWs . . . . . . . . . . . . . . . . . . . . . . . . . . 459
10.4 Performance characteristics of QWSCs . . . . . . . . . . . . . 469
10.5 Developments in QWSC design and performance . . . . . . . . 478
10.6 Limits to efficiency . . . . . . . . . . . . . . . . . . . . . . . . 480
10.7 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . 481
10.8 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 485

11. Concentrator Systems 491

I. Luque-Heredia and A. Luque
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491
11.2 The early development of CPV . . . . . . . . . . . . . . . . . . 494
11.3 Concentrator solar cells . . . . . . . . . . . . . . . . . . . . . . 499
11.4 Optics for photovoltaic concentrators . . . . . . . . . . . . . . 508
11.5 Photovoltaic concentration modules . . . . . . . . . . . . . . . 523
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xi

Contents xi

11.6 Tracking systems for photovoltaic concentration . . . . . . . . . 531

11.7 High-concentration systems . . . . . . . . . . . . . . . . . . . 538
11.8 Rating and performance . . . . . . . . . . . . . . . . . . . . . 547
11.9 Cost considerations . . . . . . . . . . . . . . . . . . . . . . . . 556
11.10 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 563
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565

12. Photovoltaic Modules, Systems and Applications 571

N. M. Pearsall
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 571
12.2 Photovoltaic modules . . . . . . . . . . . . . . . . . . . . . . . 572
12.3 The photovoltaic array . . . . . . . . . . . . . . . . . . . . . . 582
12.4 The photovoltaic system . . . . . . . . . . . . . . . . . . . . . 587
12.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 609
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 610

13. The Photovoltaic Business: Manufacturers and Markets 613

A. Jäger-Waldau
13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 613
13.2 The photovoltaic market . . . . . . . . . . . . . . . . . . . . . 618
13.3 The photovoltaic industry . . . . . . . . . . . . . . . . . . . . . 642
13.4 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 658

Appendices 663

Index 671
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xiii

ABOUT THE AUTHORS

Mary Archer read chemistry at Oxford University and took her PhD from Imperial
College London, in 1968. From 1968 to 1972, she did post-doctoral work in elec-
trochemistry with Dr John Albery at Oxford, and she then spent four years at The
Royal Institution in London, working with Lord Porter (then Sir George Porter)
on photoelectrochemical methods of solar energy conversion. She taught chem-
istry at Cambridge University from 1976 to 1986. From 1991 to 1999, she was a
Visiting Professor in the Department of Biochemistry at Imperial College London,
and from 1999 to 2002, she held a Visiting Professorship at ICCEPT (Imperial
College Centre for Energy Policy and Technology). She is President of the UK
Solar Energy Society and the National Energy Foundation and a Companion of the
Energy Institute. She was awarded the Melchett Medal of the Energy Institute in
2002 and the Eva Philbin Award of the Institute of Chemistry of Ireland in 2007.
In 2012, she was appointed a Dame Commander of the British Empire for services
to the UK National Health Service.
Christophe Ballif received his MSc and PhD degrees in physics from the Fed-
eral Polytechnic School of Lusanne, Switzerland, in 1994 and 1998, respectively,
focusing on novel photovoltaic materials. Following post-doctoral research at the
National Renewable Energy Laboratory, Golden, USA, on CIGS and CdTe solar
cells, he moved to the Fraunhofer Institute for Solar Energy Systems, Freiburg,
Germany, where he focused on crystalline silicon photovoltaics until 2003. He
then joined the Swiss Federal Laboratories for Materials Testing and Research,
Thun, Switzerland, before becoming a full professor at the Institute of Micro-
engineering, University of Neuchâtel, Switzerland, in 2004. In 2009, the Institute
of Microengineering was transferred to EPFL. He is the Director of the Photo-
voltaics and Thin-Film Electronics Laboratory in the Institute, and since 2013 he
has also been the Director of the PV-Centre of the Swiss Centre for Electronics
and Microtechnology (CSEM), Neuchâtel. His research interests include thin-film
silicon, high-efficiency heterojunction crystalline cells, module technology, con-
tributing to technology transfer, and industrialisation of novel devices.
Dieter Bonnet was born in Stuttgart, Germany, in 1937 and obtained his PhD on
the photoelectric properties of organic materials at Frankfurt University in 1963.
In 1965, he joined Battelle Institute in Frankfurt, where in 1968 he started work on
thin-film solar cells based on II–VI compounds, including CdTe. In 1970, he made

xiii
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xiv

xiv About the Authors

the world’s first CdTe/CdS thin-film solar cell in the presently known configuration.
After a period working on other solar cell materials, he resumed work on CdTe
technology in 1990, and initiated the successful EU (European Union) projects
EUROCAD in 1992 and the CdTe interest group SOLARPACT in 2005. In 1993, he
co-founded the pioneer company ANTEC GmbH to manufacture CdTe modules.
In May 2001, the road in which ANTEC is located in Arnstadt, Germany, was
named Dr. Bonnet Weg in honour of his accomplishments, and he was awarded
the Becquerel Prize of the European Commission for outstanding achievements in
photovoltaics in 2006. He retired in 2001 and is now an independent consultant.

Dan Credgington received his MSci in Natural Sciences from the University of
Cambridge in 2004, and was awarded the Herchel Smith scholarship to study at Har-
vard University. In 2010, he obtained his PhD for work on the nanoscale microscopy
and lithography of conjugated molecules from University College, London, under
the supervision of Professor Franco Cacialli, and went on to conduct post-doctoral
research on organic solar cells with Professor James Durrant at Imperial College
London. He is currently a post-doctoral researcher in the group of Professor Sir
Richard Friend at the Cavendish Laboratory of the University of Cambridge, where
his interests lie in the study of recombination processes in organic light emitting
diodes and solar cells, and hybrid organic/inorganic technologies.

Matthieu Despeisse received his degree in electrical engineering from INSA,

Lyon, France, in 2002. He then joined CERN in Geneva, Switzerland, to work
on low-noise, low-power and radiation-hard microelectronic circuits and silicon
detectors. He obtained his PhD in 2006 for his work on amorphous silicon radi-
ation sensors vertically integrated onto integrated circuits. He then pursued a fel-
lowship from 2006 to 2008 at CERN, working on the design of a beam tracker
and the transfer of CERN technologies to medical imaging and biotechnologies.
He worked on fast Si 3D sensors, SiPM for PET-scan imaging, MCP-based sys-
tems for time-resolved spectroscopy, and high-speed readout electronics. He then
joined the PVLAB group of Professor Christophe Ballif, which was part of the
University of Neuchâtel, Switzerland, until 2009 and of the Federal Polytechnic
School of Lusanne since then. He worked first on the development of amorphous
and microcrystalline silicon and silicon alloys, thin-film silicon (TF-Si) single,
tandem and triple junction solar cells and metrology equipment. He then became
R&D manager of the PVLAB TF-Si activities, leading a research team of 15 peo-
ple providing innovation to enhance performance and lower the costs of TF-Si PV.
Since January 2013, he has headed research activities on crystalline silicon solar
cells developments at the CSEM PV-Centre, in Neuchâtel, with special focuses on
silicon heterojunction technology, passivating contacts and metallisation. He has
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xv

About the Authors xv

authored or co-authored more than 70 papers in peer-reviewed journals, and he is

inventor/co-inventor on 5 patents/patent applications.

James Durrant is Professor of Photochemistry in the Department of Chemistry at

Imperial College London. After completing his undergraduate studies in physics at
the University of Cambridge, he obtained a PhD in biochemistry at Imperial Col-
lege London, in 1991, studying the primary reactions of plant photosynthesis. After
post-doctoral positions and a BBSRC Advanced Fellowship, he joined the Chem-
istry Department at Imperial College in 1999, taking up his current post in 2005.
His additional responsibilities include being Deputy Director of Imperial’s Energy
Futures Lab 2009–2014 and, from 2013, Sêr Cymru Solar Professor, University
of Swansea. His interests are in photochemical approaches to solar energy conver-
sion, including both excitonic solar cells and solar-driven fuel synthesis (artificial
photosynthesis). He has published over 280 research papers and was awarded the
2012 Tilden Prize by the Royal Society of Chemistry.

Ned Ekins-Daukes is a Senior Lecturer in the Department of Physics of Impe-

rial College London. He is also director of the IC Energy Futures Centre for
Doctoral Training. He researches high efficiency, ‘third generation’ approaches
to photovoltaic energy conversion, in particular III–V multijunction solar cells,
intermediate-band and hot-carrier solar cells. He also works on modelling the
energy yield from III–V solar concentrator systems. He previously worked in Aus-
tralia as a lecturer at the School of Physics at the University of Sydney, and later as
a visiting research fellow at the ARC Photovoltaics Centre of Excellence, UNSW.
Prior to that, he was a JSPS research fellow at the Toyota Technological Institute,
Japan.

Timothy Gessert is a Principal Scientist in the National Center for Photovoltaics at

the National Renewable Energy Laboratory (NREL) in Golden, Colorado. He holds
degrees in physics from the University of Wisconsin-River Falls (BSc), Colorado
School of Mines (MSc), and University of Wales — College of Cardiff (PhD). He
joined NREL (then known as the Solar Energy Research Institute — SERI) in 1983
where he has worked on various aspects of photovoltaic and thermophotovoltaic
devices related to GaAs, InP and CdTe material systems. Activities have also
included the development of vacuum and photolithographic processes, transparent-
conducting oxides and related electrical contacts. His current research is directed
at understanding how choices in processes and device design affect the dominant
defects and ultimate performance/stability of thin-film photovoltaic devices. Apart
from these research activities at NREL, he is active in the American Vacuum
Society, the Materials Research Society and the IEEE. He also teaches specialised
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xvi

xvi About the Authors

courses on photovoltaics and vacuum technology for professional societies and

academic institutions.
Michael Grätzel is a Professor at the Federal Polytechnic School of Lusanne,
where he directs the Laboratory of Photonics and Interfaces. He pioneered studies
of mesoscopic materials and their use in energy conversion systems, in particular
photovoltaic cells and photoelectrochemical devices for the solar generation of
chemical fuels, as well as lithium ion batteries. He discovered a new type of solar
cell based on sensitised nanocrystalline oxide films. His most recent awards include
the Leonardo Da Vinci Medal of the European Academy of Science, the Marcel
Benoist Prize, the Albert Einstein World Award of Science, the Paul Karrer Gold
Medal, the Balzan Prize and the 2010 Millennium Technology Grand Prize. He
received a doctoral degree in Natural Science from the Technical University Berlin
and holds honorary doctorates from ten European and Asian Universities. He is a
member of the Swiss Chemical Society and a Fellow of the European Academy
of Science as well as the Royal Society of Chemistry (UK) and the Max Planck
Society. He is also an elected honorary member of the Société Vaudoise de Sciences
Naturelles and the Bulgarian Academy of Science. As the author of over 1000
publications and inventor of 50 patents, he is with some 120,000 citations and an
h–index of 162 one of the three most highly cited chemists in the world.
Martin Green is a Scientia Professor at the University of New South Wales,
Sydney, and the Director of the Australian Centre for Advanced Photovoltaics.
He was born in Brisbane and educated at the University of Queensland and then
McMaster University, Canada. His contributions to photovoltaics include heading
the team that has improved silicon solar cell performance by over 50% since the
1980s, and the commercialisation of several different solar cell technologies. Major
international awards include the IEEE William R. Cherry Award in 1990, the 1999
Australia Prize, the 2002 Right Livelihood Award, also known as the Alternative
Nobel Prize, the 2007 SolarWorld Einstein Award and the 2010 Eureka Prize for
Leadership. In 2012, he was appointed as a Member of the Order of Australia
in recognition of his contributions to photovoltaics and photovoltaics education.
He has been elected to Fellowship of the Australian Academy of Science, the
Australian Academy of Technological Sciences and Engineering, the Institute of
Electrical and Electronic Engineers and the Royal Society of London. He is the
author of four books on solar cells, several book chapters and numerous reports and
papers in the area of semiconductor properties, microelectronics and solar cells.
Franz-Josef Haug studied physics at the Universities of Ulm, Germany, and
Waikato, New Zealand. For his diploma he investigated nucleation and coales-
cence phenomena during the initial stages of chemical vapour epitaxy of silicon. He
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xvii

About the Authors xvii

undertook his PhD program at ETH Zürich, Switzerland, working on Cu(In,Ga)Se2

solar cells in the superstrate configuration. After graduation, he worked on plasma
processing of super-hard nano-composite coatings of TiN/SiNx before joining the
Jülich Research Centre in Germany, where he worked on surface modifications of
the transparent front electrode of thin-film silicon solar cells. In 2005, he joined
the Institute of Microengineering, initially affiliated to the University of Neuchâ-
tel, Switzerland, and since 2009 part of the EPFL. There he leads a research group
devoted to thin-film silicon solar cells on glass and flexible plastic substrates, inves-
tigating novel texturing techniques for light scattering and absorption enhancement.
He is author or co-author of some 70 peer-reviewed papers and 4 patents. His
research interests include solar cells, semiconductor physics, and optoelectronics.

Arnulf Jäger-Waldau received his Dr. rer. nat. from the Physics Department of
the University of Konstanz, Germany, in 1993. He has worked in the field of
material research for solar cells since 1987 and holds patents on semiconductor
material deposition for thin-film solar cells and solar module design. In 1994 and
1995 he worked as a post-doctoral JSPS fellow at Shinshu University, Nagano,
Japan, before joining the Hahn–Meitner Institute Berlin in 1996. Since 2001 he
has been a Scientific Officer and Senior Scientist at the Renewable Energy Unit,
Institute for Energy and Transport of the European Commission’s Joint Research
Centre, where he works on the assessment of renewable energy technologies, the
effectiveness of their implementation, their integration into energy infrastructures
and the role of renewable energy for climate change mitigation. Among other
roles, he has been the Technical Chairman of the European Photovoltaic Solar
Energy Conference (EUPVSEC) since 2011, and he was a Lead Author for Solar
Energy of the Special Report of the IPCC on Renewable Energy and Climate
Change Mitigation. He also serves as a member of the Executive Committee of
the European Materials Research Society, the Academic Advisory Board of the
Chinese Trina State Key Laboratory for Photovoltaics, the International Advisory
Board of the Warsaw University Photovoltaic Centre and the Scientific Advisory
Board of the Solar Research Centre of the Bulgarian Academy of Science, and he
is Vice-Chairman of the Academic Committee of the Asian Photovoltaic Industry
Association (APVIA).

Antonio Luque obtained his Doctor of Engineering degree from the Polytechnic
University of Madrid in 1967. Today he is also Doctor Honoris Causa of three other
Spanish universities. In 1969, he joined the university staff and founded its Semi-
conductor Laboratory. In 1979, this centre became the Institute of Solar Energy that
he leads at present. In 1981, he founded the company Isofotón to manufacture the
bifacial cells he had invented, and he chaired its board until 1990. He has written
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xviii

xviii About the Authors

some 300 papers and registered around 20 patents, of which some are in exploita-
tion. He has won several scientific awards, among which are the Spanish National
Prize for Technology in 1987 and 2003, the King Jaime I Prize for environmental
protection, the Becquerel Prize awarded by the European Commission for PV in
1992, the IEEE William Cherry Award in 1996 and the SolarWorld Einstein Award
in 2008. Among other distinctions, he is a member of the Royal Academy of Engi-
neering, and of the Russian Academy of Sciences. He has also been a member
of the Advisory Council for Science and Technology, which advises the Spanish
Prime Minister.
Ignacio Luque-Heredia is CEO of BSQ Solar, a high concentration photovoltaics
(HCPV) manufacturing company that he co-founded in 2009. He received his MSc
and PhD degrees in electrical engineering from Polytechnic University of Madrid
in 1995 and 2010 respectively. He was co-founder in 1995 of the company Inspira,
also operating in the CPV field, which was acquired by the Silicon Valley leading
CPV manufacturer Solfocus in 2007. From 2007 to 2009 he was CTO for Solfocus
Europe. Leading Inspira’s and BSQ Solar’s engineering, he has participated in 23
collaborative projects in the field of CPV, 11 of them funded by the European
Commission. These range from the EUCLIDES project in 1995, which resulted
in the biggest CPV plant of its time, to the most recent NGCPV project, on 3rd
generation photovoltaic devices and their integration in CPV technologies. He
holds four patents and has led the deployment of CPV pilot systems and large-
scale plants, as well as several technology transfer programs in the USA, China,
India, Japan, Australia, MENA, Brazil, Mexico and Europe. He is a member of
the Scientific Committee of the International Conference on CPV Systems, and
of Work Group 7 for the development of CPV standards in Technical Committee
82 for Solar Photovoltaic Energy Systems of the International Electrotechnical
Commission. He is also a Senior Member of the IEEE.
Jenny Nelson is a Professor of Physics at Imperial College London, where she
has researched novel types of solar cell since 1989. Her current research focuses
on photovoltaic energy conversion using molecular materials, characterisation of
the charge transport, charge separation and morphological properties of molecu-
lar semiconductors, the theory of charge transport in organic semiconductors and
modelling of photovoltaic device behaviour. She has published over 200 papers on
photovoltaic materials and devices, and a book on the physics of solar cells.
Nicola Pearsall is a Professor at Northumbria University, where she leads their
photovoltaic research activities. She holds a degree in physics from the University
of Manchester Institute of Science and Technology and obtained her PhD from
Cranfield Institute of Technology for her research on indium phosphide solar cells
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xix

About the Authors xix

for satellite applications. She has been involved in research in photovoltaics for 30
years, and her current interests relate to the performance and implementation of
photovoltaic systems. She is a member of the Steering Committee of the European
Photovoltaic Technology Platform and has contributed to the development of their
Strategic Research Agenda, as well as working with the Solar Europe Industry
Initiative.
Uwe Rau received his PhD in physics in 1991 from the University of Tübingen,
Germany, for his work on temporal and spatial structure formation in the low-
temperature electronic transport of bulk semiconductors. From 1991 to 1994, he
worked at the Max Planck Institute for Solid State Research, Stuttgart, on Schottky
contacts, semiconductor heterojunctions and silicon solar cells. From 1994 to 1997,
he worked at the University of Bayreuth, Germany, on electrical characterisation
and simulation of Si and CuInSe2 solar cells. In 1997, he joined the Institute for
Physical Electronics at the University of Stuttgart, where he became leader of
the Device Analysis Group. His research interests centre on transport phenomena,
especially electrical transport in solar cell heterojunction devices and interface and
bulk defects in semiconductors. He has authored or co-authored more than 100
scientific publications.
Hans-Werner Schock received his diploma in electrical engineering in 1974, and
doctoral degree in electrical engineering in 1986, from the University of Stuttgart’s
Faculty of Electrical Engineering. Since the early 1970s, he has worked on the
development of polycrystalline II–VI and I–III–VI2 compound semiconductor thin-
film solar cells, taking the development of chalcogenide solar cells from research
to pilot fabrication. A series of successful research projects on thin-film solar cells
under his guidance resulted in several production lines for thin film solar cells
in Europe. From 2004 to 2012, he was director of the Institute of Technology at
the Helmholtz Zentrum Berlin for Materials and Energy and he is an Honorary
Professor at the Technical University Berlin. He received the Becquerel Prize of
the European Commission in 2010 for his achievements in the development of
thin-film solar cells. At present he is a consultant in the field of photovoltaics.
He is the author or co-author of more than 300 contributions in books, scientific
journals and conference proceedings.
Masafumi Yamaguchi is a Professor of the Toyota Technological Institute. He
received his BS and PhD degrees from Hokkaido University, Japan, in 1968 and
1978, respectively. In 1968, he joined the NTT Electrical Communications Labo-
ratories, working on radiation damage to Si and III–V compounds, ZnSe blue-
light-emitting diodes and III–V compound solar cells. In 1983, he discovered
the superior radiation-resistance of InP materials and solar cells, thereby showing
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xx

xx About the Authors

the great potential of InP cells for space applications. His group also developed
high-efficiency InP, GaAs-on-Si, and AlGaAs/GaAs tandem cells by proposing
a double-hetero structure tunnel junction for realising a high performance and
stable multijunction cell interconnection in 1987. As Japanese team leader of the
EU–Japan Collaborative Research on Concentrator Photovoltaics, he contributed to
the attainment of InGaP/GaAs/InGaAs 3-junction cells with efficiencies of 44.4%
at 302 Suns of AM1.5D and 37.9% at AM1.5G. He is also the project leader
of the Next Generation High Performance Photovoltaics Research and Develop-
ment Project of the Japanese New Energy Development and Industrial Technology
Development Organisation, and the Research Supervisor in the Research Area
of the Creative Clean Energy Generation using Solar Energy programme of the
Japan Science and Technology Agency. He has published more than 300 origi-
nal papers and received numerous awards, such as the Becquerel Prize from the
European Commission in 2004 and the William Cherry Award from the IEEE
in 2008 for outstanding contributions to science and technology development of
high-efficiency photovoltaics.
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xxi

PREFACE TO THE FIRST EDITION

And there the unregulated sun
Slopes down to rest when day is done
And wakes a vague, unpunctual star …
Rupert Brooke, The Old Vicarage, Grantchester, May 1912

Since the dawn of history, man has been fascinated by the Sun, the provider of
the light and warmth that sustains life on Earth. In pre-industrial times, our major
sources of energy — wood, wind and water power — derived from solar energy.
The subsequent discovery and massive exploitation of fossil fuels laid down in
the Earth’s crust by early aeons of photosynthetic activity have conditioned the
developed world to be dependent on convenient, readily available energy. But we
are living on our energy capital. The Earth’s reserves of coal, oil and gas are
finite and likely to become resource-depleted in the course of this century. A sense
of living on borrowed time was therefore appropriate even before concerns about
global climate change, sustainability and energy security combined to raise interest
in renewable energy to its current encouraging level.
This book is the first in a series of four multi-authorial works on the pho-
toconversion of solar energy. It was created from my long-held conviction that,
despite slow starts and setbacks, solar energy — broadly defined to encompass
other renewable energy forms that derive from solar — will become the Earth’s
major energy source within this century. The Sun is a source of both radiant heat
and light, and techniques for using solar energy correspondingly divide into ther-
mal methods (solar power towers, water heaters and so on) and photoconversion
(sometimes called direct) methods. Photoconversion is the subject of this book
series. A photoconverter is a device that converts sunlight (or any other source of
light) into a useful form of energy, usually electrical power or a chemical fuel, in
a process that relies, not on a raised temperature, but on the selective excitation
of molecules or electrons in a light-absorbing material and their subsequent de-
excitation in a way that produces energy in a useful form. Volume I covers the most
developed of the man photoconversion devices, photovoltaic (PV) cells, which are
solid-state semiconductor devices that produce electrical power on illumination.
Volume II will cover the natural photoconversion system of photosynthesis, the
potential of biomass as an energy source and the global carbon budget. Volume

xxi
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xxii

xxii Preface to the First Edition

III will explore the less developed but exciting possibilities of synthesising arti-
ficial ‘molecule-based’ photoelectrochemical or photochemical photoconverters.
Finally, Volume IV will draw together the common themes of photoconversion and
provide some background material.
The series is intended mainly for senior undergraduates, graduate students and
scientists and technologists working on solar photoconversion. Chapters 1–12 of
this book deal with PV cell design, device physics and the main cell types — crys-
talline and amorphous silicon, cadmium telluride and copper indium diselenide —
as well as more advanced or less developed options such as quantum-well and
thermophotovoltaic cells. These chapters are mainly technical, requiring sound
knowledge of physics, chemistry or materials science for ready understanding.
Chapters 13–18 deal with PV systems, manufacturers, markets and economics and
are accessible without specialist knowledge.
A multi-authorial work owes its very existence to its authors, and my whole-
hearted thanks must go to the twenty-five distinguished individuals, all recognised
authorities in their own fields, who have contributed to this book and patiently
answered my queries during the editing stage. I have also been helped by dis-
cussions about PV with many friends and colleagues, and visits to installations
throughout the world: I have been up Swiss mountains, onto Japanese rooftops and
into the Arizona desert, and thoroughly enjoyed every minute. I am most grateful
to those who have read and commented on various parts of this book or provided
specialist information in advance of publication: Dennis Anderson, Jeffrey and
William Archer, Stephen Feldberg, Martin Green, Eric Lysen, Larry Kazmerski,
Bernard McNelis and Nicola Pearsall. I also warmly thank Alexandra Anghel,
Barrie Clark, Stuart Honan and my PA Jane Williams for editorial assistance, and
Ellen Haigh and John Navas of IC Press and Alan Pui of World Scientific Press for
guiding the book to publication.
For me the sad part of writing this preface is that I must do so in the first
person, for my co-editor Professor Robert Hill died suddenly on 26 November
1999. Bob was the most knowledgeable champion of photovoltaics in the UK, and
his premature death has deprived the British PV community of its cornerstone. He
had drafted his chapter with Nicky Pearsall some months before he died, and the
flow of emails delivering his astute editorial comments on other chapters continued
until the day before his death.
Bob believed unshakeably in the future of PV. Although he knew that system
costs will have to fall by another factor of 2–3 if PV is to become cost-competitive
in major new grid-accessible markets, there are good grounds for believing this is
possible. PV technology is still young, and significant further economies of scale
from larger manufacturing facilities, as well as further advances in the fundamental
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xxiii

Preface to the First Edition xxiii

science, can confidently be expected. The world’s first-generation televisions and

mobile telephones were at least as uncommon and expensive as PV is now.

The Old Vicarage, Grantchester Mary Archer

December 2000
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xxv

PREFACE TO THE SECOND EDITION

O Sonne, Königin der Welt,
Die unser dunkles Rund erhellt
In lichter Majestät;
Erhab’nes Wunder einer Hand,
Die jene Himmel ausgespannt
Und Sterne hingesät!
Johann Peter Uz (1720–1796), An die Sonne

The world of photovoltaics has advanced at a phenomenal pace since the first edition
of this book was published in 2001. Then we were able to report with modest pride
that 200 MWp of PV had been installed worldwide during 1999, taking global
cumulative installed capacity to just over 1 GWp. By the end of 2012, global
installed capacity had edged past the iconic figure of 100 GWp. The final figure for
global installations in 2013 is yet to be determined, but it will add around another
35 GWp, with plausible estimates for cumulative installed capacity by 2020 topping
300 GWp. PV is now, after hydro and wind, the third most important renewable
energy source in terms of installed global capacity.
This prodigious growth has impacted on all sectors of the PV industry. In
1999, much of the small quantity of pure silicon needed by the cell manufacturing
industry came from the moderate resistivity p-type waste material discarded by the
electronics industry. As the PV market grew, a tipping point was reached in 2006
when, for the first time, over half of the world’s supply of polysilicon was used for
PV production. Silicon feedstock for the PV industry is now made by dedicated
plant, and manufacturing capacity has swung sharply away from the USA and
Europe to the lower-cost economies of the Far East.
As the PV market has grown, so the price of cells and modules has dropped.
In their early years, photovoltaic cells were an expensive and exotic novelty for use
in space. In 1955, market leader Hoffman Electronics were offering 2% efficient
silicon cells at a price of $1500/Wp (about 9520/Wp in today’s money). At the
end of 2013, spot prices on the European market were nearly 20,000 times lower
at 0.5–0.7/Wp for silicon modules of efficiencies in the range 15–18%.

xxv
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xxvi

xxvi Preface to the Second Edition

Between the two editions of this book lie nearly 15 years of scientific discov-
ery, technological advance and market maturation, punctuated by global economic
and financial turmoil, international bickering about the need to mitigate climate
change, national anxieties about energy security, stability and independence, and
political and regulatory instability as governments reconsider their commitment to
renewable energy sources and carbon offset.
The architecture of this edition follows the lines of the first edition, with a
set of chapters about the main PV technologies following an introductory chapter.
Omitted from this edition are thermophotovoltaics, PV for space applications and
electricity storage, while 3-junction III–V concentrator cells of nearly 45% effi-
ciency, CZTS (copper–zinc–tin–sulphide/selenide), dye-sensitised and perovskite
solar cells are notable newcomers to the stage. A further addition is an overview
of the limits to photovoltaic conversion efficiency and how far they can be pushed
beyond single-junction detailed-balance constraints. The book concludes with a
look at PV systems, applications and markets.
We are deeply indebted to all our authors, without whose generosity with their
knowledge and patience with the editorial process this book could not have been
produced. We are particularly grateful to those who have stayed with us from the
first edition and undertaken the painstaking task of radically updating their chapters:
Dieter Bonnet, Antonio Luque, Jenny Nelson (writing about QW solar cells in
the first edition and OPV in the second), Nicola Pearsall, Hans-Werner Schock
and Masafumi Yamaguchi. We warmly welcome newcomer authors Christophe
Ballif, Dan Credgington, Matthieu Despeisse, James Durrant and Michael Grätzel
(who jointly contributed a chapter to Volume III of this series), Ned Ekins-Daukes,
Timothy Gessert, Franz-Josef Haug, Arnulf Jäger–Waldau, Ignacio Luque-Heredia
and Uwe Rau.
The structure of this book series has altered since the preface to the first
edition written: Volume III did not deal with solar fuels as originally planned, but
with photoelectrochemical and nanostructured devices for solar photon conversion;
solar fuels are now planned for the fourth and final volume.
Editing this book series has become a way of life for one of us (MA). She
thanks her family for their tolerance of this enduring preoccupation, her PA Carol
Burling for her skill at the electronic keyboard, James Archer, Richard Friend,
Larry Kazmerski, Bernard McNelis and colleagues at the UK’s annual PVSAT
conferences for much helpful advice and input. MG wishes to thank his partner,
Judy Green, for her ongoing support and Joyce Ho for her assistance on several
fronts during book preparation, as well as the many colleagues and students who
have helped make photovoltaics such a stimulating field of interest.
September 17, 2014 11:29 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-fm page xxvii

Preface to the Second Edition xxvii

We jointly acknowledge with gratitude the editorial assistance of Jacqueline

Downs, Leah James and Tom Stottor of Imperial College Press and Catherine Yeo
of World Scientific Publishing, who have seen the book through to publication.

January 2014 Mary D. Archer

Martin A. Green
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 1

CHAPTER 1

THE PAST AND PRESENT

MARY D. ARCHER
The Old Vicarage, Grantchester, Cambridge CB3 9ND, UK
[email protected]

Time present and time past

Are both perhaps present in time future.
T. S. Eliot, Burnt Norton, Four Quartets, 1935–1942.

1.1 Introduction
Photovoltaic (PV) cells generate electric power when illuminated by sunlight or
artificial light. They are by far the most highly developed of the man-made pho-
toconversion devices. Born of the space age in the 1950s, their earliest terrestrial
applications emerged in the 1970s and they have enjoyed rapid market expansion
since the turn of the millennium. PV technology has a number of advantages over
conventional methods of electricity generation. First and foremost, solar energy
is the world’s major renewable energy resource. PV power can be generated from
the Sun anywhere — in temperate or tropical locations, in urban or rural environ-
ments, in distributed or grid-feeding mode — where there is adequate light. As a
fuel-free distributed resource, PV could in the long run make a major contribution
to national energy security and carbon dioxide abatement. PV is uniquely scalable,
the only energy source that can supply power on a scale of milliwatts to megawatts
from an easily replicated modular technology with excellent economies of scale
in manufacture. A standard 156 × 156 mm crystalline silicon PV cell generates
about 4 peak watts1 (Wp) of DC power and typical PV 60-cell and 72-cell modules
about 240 Wp and 300 Wp respectively. The world’s largest PV generating facility
is currently the Agua Caliente Solar Project in Arizona, which had 251.3 MWp
installed as at November 2013.

1 The power output of a PV cell or module is rated in peak watts (Wp), meaning the power output
at 25◦ C under standard AM1.5 solar radiation of global irradiance 1 kW m−2 . PV ratings in ‘watts’
invariably mean peak watts. To convert from peak watt rating to 24-hour average power output in a
sunny location, divide by a factor of ∼5.

1
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 2

2 M. D. Archer

PV cells are made of thin semiconductor wafers or films. They contain small
amounts only of (usually non-toxic) materials and, when manufactured in volume,
have modest embedded energy, with energy payback times for silicon PV modules
in Southern Europe of about one year (Fraunhofer, 2012). They possess no moving
parts, generate no emissions, generally require no cooling except when used in
concentrator systems and are silent in operation. PV systems are reliable, easy to
use and long-lived if properly maintained (most commercial modules have lifetime
warranties of 25–30 years, though some balance-of-system components are less
reliable and long-lived than this). Carefully designed, PV arrays are not visually
intrusive, and can indeed add architectural merit to the aesthetic of a built structure.
The PV industry has advanced dramatically in the decade since the first edi-
tion of this book was published. Over the period 2000–2012, solar PV was the
fastest-growing renewable power technology worldwide, and its market growth
has outstripped even the most bullish of forecasts: cumulative global installed PV
capacity in 2000 was only 1.4 GWp, while by the end of 2012 it had just edged past
the milestone of 100 GWp installed (EPIA, 2013). Despite the difficult economic
situation, global PV module production in 2012 was 35.9 GWp (GTM Research,
2013). Future growth scenarios differ widely, but the average forecast is for some
260 GWp of cumulative installed global PV capacity by 2020, with demand shift-
ing from Europe to countries such as China, India and the USA. These projections
show there are huge opportunities for PV in the future.
A number of factors have driven this growth. Many countries, including most
in the OECD and several developing countries, have introduced tax or regulatory
policies that favour the renewables. Modest renewable set-asides (requirements
on major utilities to source some power from renewables) are in place in many
countries with liberalised electricity supply industries, and the right to supply power
to the grid has been extended to independent power producers (IPPs), sometimes
with incentives to source electricity from renewable sources. IPPs can site their
plant close to the consumer and avoid the costs of distribution, often as significant
as the costs of generation.
The 100,000 roofs programme in Germany, which ran from 1999 to 2004, and
other national roofs programmes stimulated demand for distributed PV systems.
Feed-in tariffs (FiTs), also pioneered by the German government, have stimulated
PV installations by guaranteeing an above-market price for a fixed term for PV-
generated power to be fed into the grid. These tariffs, which are now offered in
many countries, give assurance of return to investors. Later FiT rounds generally
provide less generous rates of return, which has helped to drive system prices down,
although abrupt tariff changes can produce boom/bust cycles and create investor
uncertainty.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 3

The Past and Present 3

In the USA, residential and commercial PV developments were stimulated

by a Treasury Program that ran from 2009 to the end of 2011, which provided an
upfront federal grant in lieu of recompense through the less certain mechanism
of an Investment Tax Credit. US utilities seeking to fulfil their RPS (Renewable
Portfolio Standard) obligations have also developed a number of grid-connected
power arrays, and the domestic market for PV in the USA doubled in 2011, and
again in 2012. The US Department of Energy’s SunShot initiative currently sup-
ports Research & Development in most PV technologies with the aim of making
solar electricity cost-competitive with other sources of energy by 2020. SUNPATH,
which stands for Scaling Up Nascent PV AT Home, is one of the initiatives under
SunShot, aimed at supporting innovative, high-tech companies to accelerate cost
reductions and commercialisation of their PV technologies.
The PV manufacturing industry has also undergone profound changes over the
last decade. Since 2000, global PV cell/module production has increased by almost
two orders of magnitude, with annual growth rates in the range 40–90, stalling only
in 2012 as a result of massive inventory overhang. Until 1998, USA PV manufac-
turers accounted for the greatest share of global shipments, but the manufacturing
base has switched to Asia, particularly to China, which accounted for 64% of
worldwide production in 2012, while Europe’s share dropped to 11% and Japan’s
and the USA’s to 5% and 3% respectively (GTM Research, 2013). Several years
of explosive growth in China’s production capacity led to significant overcapacity
(Yang, 2012), causing industry profits to slump, with major manufacturer Suntech
finding itself in severe difficulties (PV Magazine, 2013). According to PV Tech
(2012), 90% of China’s polysilicon raw material producers have stopped produc-
tion. The Chinese government has called for industry consolidation and sought to
stimulate the domestic market for PV in its latest 5-year solar energy development
plan by setting a target of 20 GWp installed by 2015 (China Briefing, 2012).
The oil-rich nations of the MENA (Middle East/North Africa) region have
enormous solar resources, and are beginning to prepare themselves for the post-
hydrocarbon era and increase their oil export revenues by developing their solar
resources for domestic use. In May 2012, Qatar Solar Technologies secured an
unprecedented investment of US$1 bn from a Qatari bank to build a polysilicon
production facility in Ras Laffan Industrial City. Saudi Arabia aims to meet a third
of its electricity with solar power by 2032, while Abu Dhabi has targeted 7% of its
power from renewable sources by 2020 and Dubai 5% by 2030 (Mahdi, 2012).
Wafer-based crystalline Si remains the dominant commercial technology, with
about 89% market share of module production in 2012, although its hegemony
was challenged in the period 2005–2009 by the temporary global shortage of Si
feedstock. This led to massive expansion of investment in non-silicon thin-film
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 4

4 M. D. Archer

PV capacity, particularly in CdTe. Thin-film’s market share of production grew

rapidly to 21% in 2009, but fell back to 11% in 2012 as new production facilities
for high-purity polysilicon were brought on line.
Supply of polysilicon outstripped demand in 2011 and 2012, and as a conse-
quence Si module prices dropped rapidly throughout 2011 from an average of about
$1.7 Wp−1 to a level approaching $1 Wp−1 later in the year. 2012 saw further price
reductions, to an average of $0.87 Wp−1 for Tier 1 (major manufacturers) of crys-
talline silicon (c-Si) modules (GTM Research, 2012). The c-Si production industry
is currently having to consolidate, with idled plant, sector exits and bankruptcies
producing a painful shake-out of unprofitable business. Even the venerable BP
Solar shut down in 2011 after 40 years in the business, saying it ‘simply can’t
make any money from solar’ (MIT, 2011).
The dominant thin-film materials are cadmium telluride (46% of total thin-
film production), amorphous Si or amorphous/microcrystalline Si (35%), and CIGS
(copper indium gallium selenide, 19%). Thin-film modules save on PV materials,
are highly suited for low-cost mass production and open up new applications such as
the provision of power for consumer electronics. The active layers of thin-film cells
are typically only a few microns thick, about 50 times thinner than standard c-Si
cells. Despite this, thin-film modules are not lighter per Wp than Si modules due to
their lower efficiency and the need for double glass encapsulation (because of their
greater moisture sensitivity). Nevertheless, the generally lower cost per Wp of thin-
film modules and their insensitivity to partial shading make them attractive where
area is not limited, though their lower efficiency puts them at a disadvantage where
it is. Another advantage over c-Si technology, in which wafers, cells and modules
must be produced in different steps, is that thin-film cells can be fabricated as series-
connected modules in an integrated manufacturing process. Continued efficiency
improvements remain important because higher efficiency modules incur lower
balance-of-system costs per unit output.
Figure 1.1 shows the full range of PV cell types with the world record efficien-
cies achieved by each over the past twenty years, as catalogued by Green and his
colleagues. These are research cell, not production cell, efficiencies, but they set
a benchmark for their commercial cousins, the efficiency of which also improves
year on year.
Grid parity — where the cost of PV power has dropped sufficiently that the
installed cost per watt is equal to the local cost of fossil fuel power — marks the
point that PV should be able to expand without subsidy. This has already been
reached or passed in sunny, remote regions and islands where the alternative is
to generate electricity using imported oil. Grid parity has also been achieved in
parts of Spain, and was expected to be reached by parts of Italy, Brazil, Chile and
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 5

The Past and Present 5

Figure 1.1 Highest confirmed efficiencies for (a) ≥1 cm2 area cells (b) ≥800 cm2 area
modules (200–400 cm2 for OPV); (c) concentrator cells and modules. Source: Green et al.
(2013).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 6

6 M. D. Archer

Australia during 2013 (Brittlebank, 2012). PV can also be cost-effective in peak

shaving before it has reached full grid parity.
International policy to mitigate climate change has also provided some stim-
ulus to PV. The Kyoto Protocol to the UN Framework Convention on Climate
Change, which came into force in 2005, set binding obligations on industrialised
countries to reduce greenhouse gas (GHG) emissions. This led to a number of
regional bloc and national programmes to stimulate the use of renewable energy
forms, PV among them, as well as a number of ‘flexible mechanisms’ by means of
which signatories could meet their GHG reduction targets by investing in emissions
reduction programmes overseas.
Enthusiasm for the Protocol has been at best patchy. The UK, law-abiding
as ever, passed the world’s first long-term legally binding framework to mitigate
climate change in its 2008 Climate Change Act. However, little PV was installed
until attractive FiTs were introduced in 2010. By the end of Q2 2013, cumulative
installed capacity had risen to 2.4 GWp, and the government forecasts 10 GWp
installed by 2020 (DECC, 2013).
The EU–152 embraced the Kyoto Protocol wholeheartedly and now has the
greatest installed PV capacity of any regional bloc, with a cumulative installed
capacity of over 66 GWp at the end of 2012. PV accounted for 56% (21.5 GWp) of
new net generating capacity in the EU in 2012, beating wind with 25% (9.4 GW) of
market share, into second place. PV now provides 2.6% of EU electricity demand
on average, and 5.2% of peak demand (EPIA, 2013).
The US signed, but never ratified, the Kyoto Protocol, and Canada withdrew
from it in 2011, arguing that a treaty that does not include the world’s two largest
GHG emitters — the US and China — would prove ineffectual at reducing emis-
sions. At the 2012 Doha Climate Change Conference of Kyoto signatories, little
progress was made on emissions targets: 37 countries agreed to a second round
of reductions, but a number of others, notably Japan and Russia, did not take on
new obligations. It is becoming clear that the Kyoto Protocol is not achieving its
original aim, and it was agreed in principle in Doha that it should be replaced by a
new treaty binding on all UN members whereby rich nations will compensate poor
nations for the damage caused by climate change; the 2013 UN Climate Change
Conference in Warsaw achieved little in this or other respects.

1.2 Milestones in the development of photovoltaic technology

The discovery of photovoltaism is commonly, if inaccurately, ascribed to Becquerel
(1839), who observed that photocurrents were produced on illuminating platinum

2 The 15 states that made up the European Community at the time of the Kyoto negotiations.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 7

The Past and Present 7

electrodes coated with silver chloride or silver bromide and immersed in aque-
ous solution.3 The observation by Smith (1873) of photoconductivity in solid
selenium led to the discovery of the photovoltaic effect in a purely solid-state
device by Adams and Day (1877), who observed photovoltages in a selenium rod
to which platinum contacts had been sealed, which they (incorrectly) ascribed
to light-induced recrystallisation of the selenium. The first practical photovoltaic
device — a light meter consisting of a thin layer of selenium sandwiched between
an iron base plate and a semi-transparent gold top layer made by Fritts (1883) —
was promoted by the German industrialist Werner von Siemens as demonstrating
‘for the first time, the direct conversion of the energy of light into electrical energy’
(Siemens, 1885). Photometers based on selenium photocells were commercialised
in Germany in the 1930s, and amorphous selenium photodetectors remain of inter-
est today for medical imaging applications.
The selenium photocell is an example of a barrier layer cell, so called because
it contains an electrical barrier that is highly resistive to current flow in one direc-
tion — a rectifying junction, in modern parlance. Two further barrier layer cells,
the thallous sulphide cell (Case, 1920) and the copper oxide cell (Kennard and
Dieterich, 1917; Grondahl and Geiger, 1927), were developed during the 1920s.
Garrison (1923) and later Grondahl (1933) reviewed work on Cu2O, Garrison being
the first to report the logarithmic dependence of open-circuit voltage on irradiance.
Fink and Fogle (1934) discovered the efficacy of antireflective coatings accidentally
when they found that beeswax, applied to protect cells during chemical etching,
enhanced the photocurrent. Schottky (1930) noted and explained the difference in
spectral response of ‘Vorderwandzelle’ (front-wall-illuminated cells) and ‘Hinter-
wandzelle’ (back-wall-illuminated cells) while experimenting with semitranspar-
ent metal contacts for the Cu2 O/Cu barrier cell.
Lange (1930), like Siemens, foresaw the possible application of the photo-
voltaic effect to energy conversion, pointing out that ‘using a more appropriate
semiconductor for the intermediate layer [of a barrier cell], it will be possible to
select the cell effectiveness for a specific spectral region … Use of a more appropri-
ate unipolar [contact] layer is expected to give a further increase in the efficiency of
the cell. It is then possible that efficiencies can be reached which allow direct con-
version of light into electrical energy.’ Sheldon and Geiger (1922), Geiger (1923)
and Bergmann and Hänsler (1936) observed and recorded photovoltaic effects with
a wide range of other minerals and compounds, Bergman reporting what may be
the earliest observation of semiconducting effects in organic dyestuffs. Reviews

3 Becquerel’s observation was strictly speaking a photoelectrochemical effect, but its basis — the rectify-
ing junction formed between two dissimilar electric conductors — is the same as that of the photovoltaic
effect in purely solid-state devices.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 8

8 M. D. Archer

by Winther (1928) and Crossley et al. (1967) and the book by Lange (1938) give
an account of these early devices.
The electrical barrier of barrier layer cells was originally thought to lodge
in an interfacial foreign layer of high resistivity such as an oxide, but Schottky
(1938), and independently Davydov (1939) and Mott (1939), showed that a third
phase was not necessarily involved. Rather, metal/semiconductor junctions could
in themselves be rectifying by virtue of the space-charge layer created in the semi-
conductor by electronic equilibration/charge redistribution when contact was made
with a metal of different work function. The first diffusion theory of p-n junction
rectification, which became the basis for Shockley’s theory, was published by
Davydov (1938).
Metal/semiconductor devices make inefficient solar converters because their
dark currents are relatively large and this diminishes the photovoltaic response.
Semiconductor/ semiconductor junctions are better in this regard. The first p-n
junctions to be reported were the germanium homojunctions of Lark-Horovitz’s
group at Purdue University (Benzer, 1946, 1947) and the lead sulphide quasi-
homojunction formed by pressing together two PbS wafers, one enriched with
lead and the other with sulphur (Sosnowski et al., 1947).

1.3 The crystalline silicon solar cell

It was Russell Ohl, a metallurgist working at Bell Telephone Laboratories in New
Jersey before World War II, who ushered in the modern era of photovoltaics by
his observation that crystallisation of a melt of commercial ‘high purity’ silicon
produced a ‘well-defined barrier having a high degree of photovoltaic response’
(Ohl, 1941). This barrier was in fact a p-n homojunction formed from the unequal
distribution of impurities as the Si crystal grew from the melt. From this discovery,
after a delay occasioned by the war, grew the seminal work of Chapin et al. (1954) at
Bell on the development of the diffused p-n junction single-crystal silicon cell. This
technology was commercialised by Hoffman Electronics, and a place on board the
1958 NASA Vanguard I satellite for Hoffman’s 9%-efficient c-Si cells pioneered
the way for a number of space applications.
Efficiencies improved rapidly. The so-called silicon ‘violet cell’ with its
improved short-wavelength response, developed at COMSAT in the early 1970s
(Lindmeyer and Alison, 1973), achieved AM1.5G efficiency of over 15%. Next
came the ‘black cell’ of Haynos et al. (1974), in which front-surface reflectance
was greatly reduced by a chemical etch of the Si surface, giving the cells a black
appearance after the application of antireflection coatings. This took terrestrial cell
efficiency above 17% for the first time. The 18% milestone was passed a decade
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 9

The Past and Present 9

later by the MINP (Metal-Insulator-NP junction) cell of Green et al. (1985). From
then on, Si cell technology progressed from MINP to PESC (the Passivated Emitter
Solar Cell), to rear contact cells, to PERC (Passivated Emitter and Rear Contacts)
cells, and ultimately to the PERL (Passivated Emitter, Rear Locally-diffused) cell,
which is 24.8% efficient in monocrystalline form (Zhao et al., 1998). Green (1999)
describes the development of silicon solar cell technology in more detail.
A small terrestrial market for c-Si modules started to appear for remote, small-
power applications and demonstration projects in the 1970s, stimulated by the oil
price hikes of that time (Green, 1993; Treble, 1998). Successive markets opened
up for silicon PV as costs reduced. In 1970, c-Si cells for use in space cost several
hundred dollars per watt. By the mid-1970s, the efforts of Elliot Berman and his
Solar Power Corporation (backed by Exxon) had reduced the cost of cells made
specifically for terrestrial applications to $20 Wp−1 . In round terms, the RAPS
(remote area power supplies) market opened up in the 1980s at module costs of
$10 Wp−1. Solar lighting in grid-remote locations opened up in the early 1990s
at $5 Wp−1 . Multicrystalline silicon (mc-Si) cells entered the PV market in 1981
and captured an increasing market share as module efficiencies improved from an
early 10–12% to ∼16–18% today.
Until the mid-2000s, sufficient high-quality material for c-Si cell manufacture
was available from the electronics industry as cut-price off-spec and waste polysil-
icon (small-grained, polycrystalline silicon of 9N purity).4 This was mostly made
from metallurgical-grade silicon (MG-Si) by the energy-intensive Siemens pro-
cess, which dates back to the 1940s. The main method for producing single-crystal
Si from polysilicon is the even more venerable Czochralski (CZ) method, in which
a rod of monocrystalline Si is grown by slowly withdrawing a seed crystal from a
polysilicon melt. The CZ method dates from the 1916 work of the Polish scientist
Jan Czochralski, co-founder of the German Metallurgical Society, and is also an
energy-intensive process. Cutting the CZ rod into thin wafers to make c-Si cells
incurs further losses and wastage of material, and fabricating cells from wafers is
a complex and multi-step process. Multicrystalline silicon (mc-Si) is commonly
made by the Bridgman method of directional crystallisation: silicon melt is poured
into a cast and solidified from the bottom up by extracting heat from the crucible
base, yielding blocks with columnar crystal growth that are then cross-sectioned
into large-grained wafers.
By 2005, the market for PV had grown to the point where Si PV manufac-
turers were competing against chip companies for electronic-grade material. This

4 The purity of silicon is often expressed as the total number N of nines in 99.99 … %.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 10

10 M. D. Archer

produced a bottleneck in the supply of polysilicon in the period 2006–2008, during

which its price rose to the point where lower-cost upgraded metallurgical silicon
(UMG-Si) was able to enter the PV supply chain. This had ∼6N purity, adequate
to make multicrystalline Si cells of reasonable performance. However, the global
recession of early 2009 caused the price of polysilicon to fall sharply to the point
where wafer and cell manufacturers found the higher cost of 9N polysilicon was
justified by the higher cell efficiencies they could achieve using it. Thus UMG-Si
fell from favour.
As already outlined in Section 1.1, overcapacity in the Si module market,
combined with stagnation of demand in Europe, has led to a sharp drop in c-
Si prices. At the time of writing (November 2013), business conditions remain
very difficult for manufacturers. Some full-service companies that have substantial
installation contracts are doing rather better.

1.4 GaAs and III–V multijunction cells

Gallium arsenide (GaAs) and related III–V materials have a fortunate combination
of properties which gives them several advantages as solar cell materials. Their
effective carrier masses are small and hence their carrier mobilities are high, and
their optical absorption is intense with sharp absorption edges because of their
direct-gap band structure. III–V materials can be doped both n- and p-type, and
lattice-matchable ternary (e.g. Alx Ga1−x As, Inx Ga1−x As) and quaternary (e.g.
Alx Ga1−x As y P1−y ) alloys can be made over wide composition ranges. This allows
the fabrication of multijunction (MJ) III–V cells of very high performance in
which the bandgap of each sub-cell is optimised by adjusting the alloy compo-
sition. Most MJ cells are monolithic devices consisting of two or three ‘current-
matched’ sub-cells stacked vertically with a tunnel junction series connection
between each.
A photovoltaic effect in a GaAs p-n homojunction was first reported by Welker
(1954), followed a year later by Gremmelmaier (1955), who obtained ∼1% effi-
ciency in a polycrystalline p-n homojunction solar cell. Jenny et al. (1956) of RCA
Laboratories followed up with a >6%-efficient monocrystalline GaAs p-n cell.
However, GaAs has the disadvantage of a high surface recombination velocity, so
most of the photogenerated carriers recombine in the top layer of a GaAs p-n homo-
junction cell before they can reach the junction. Adding a thin top window layer of
Al x Ga1−x As avoids this because carrier recombination at the Alx Ga1−x As/GaAs
interface is slow. The first lattice-matched AlGaAs/GaAs heterostructures to incor-
porate such a window layer were independently reported by Rupprecht et al. (1967)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 11

The Past and Present 11

at IBM’s Thomas Watson Research Center and Alferov et al. (1967) at the Ioffe
Institute of Leningrad (as it then was).5 This was quickly followed by the p-
AlGaAs/ p-GaAs/n-GaAs heteroface cell developed by Woodall and Hovel (1972)
at the Thomas Watson Research Center, which in 1977 achieved a record single-
junction efficiency for its time of 21.9%.
Tandem solar cells were proposed as far back as 1955 by Jackson (1955) and
later by Wolf (1960). However, efficient III–V tandem cells were not achieved
until the late 1980s because of difficulties in making high-performance stable tun-
nel junctions and the effects of oxygen-related defects in the AlGaAs window layer
(Ando et al., 1987). Yamaguchi et al. (1987) proposed the use of double-hetero
(DH) structure tunnel diodes as optically and electrically low-loss cell intercon-
nects. Olson et al. (1990) achieved an efficiency of 27.3% in a cell with a Ga0.5In0.5 P
homojunction grown epitaxially on a GaAs homojunction with a GaAs tunnel junc-
tion interconnect, and Japan Energy (Takamoto et al., 1997) broke the (1 Sun)
30% barrier with their InGaP/GaAs tandem cell, which had a DH tunnel junction
interconnect in which the InGaP layers were surrounded by high-bandgap AlInP
barriers.
The main drawback of III–V cells is their high cost. Aluminium is so reactive
in the vapour phase that it is not possible to prepare AlGaAs layers by conventional
chemical vapour deposition using elemental sources. Rather it is necessary to use
the slower and more expensive method of metal-organic chemical vapour epitaxy
(MOVPE), using sources such as trimethyl aluminium. Cell stacks are usually
grown epitaxially on substrates such as germanium. However, the high efficiency,
good radiation hardness of InP and InGaP (Yamaguchi et al., 1997) and low series
resistance of III–V MJ cells have allowed them to dominate two niche markets:
provision of PV power in space and use in concentrator PV systems (CPV). III–V
MJ concentrator cells have the highest conversion efficiencies in their class of any
PV technology. Boeing Spectrolab broke the 40% efficiency barrier in 2006 with
their 3-junction GaInP/GaInAs/Ge concentrator cell, which was 40.7% efficient
under 240 Suns (King et al., 2007). At the time of writing, the world record holder
for CPV is the Fraunhofer Institute’s 4-junction cell, which is 44.7% efficient
under 297 Suns (Fraunhofer, 2013). In theory, efficiencies of 50% or more are
possible with MJ concentrator cells with four or more junctions (Yamaguchi et al.,
2008).

5 Zhores Alferov won the 2000 Nobel Prize for Physics for his development of heterostructures and is
probably the only solid-state physicist to have an asteroid named after him.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 12

12 M. D. Archer

1.4.1 Quantum well solar cells

Quantum well solar cells (QWSCs) are relatively recent constructs, proposed by
Barnham and Duggan (1990) and discussed by Jenny Nelson and Ned Ekins-
Daukes in Chapter 10. They are based on III–V p-i -n or p-n homostructure cells
into the i layer of which several extremely thin layers of a second semiconductor
of smaller (bulk) bandgap are introduced, forming quantum wells in which ‘extra’
charge carriers are created by the additional light absorbed. The aim is that, com-
pared with the parent cell (i.e. the host cell without the QWs), the photocurrent
will be enhanced while the voltage is not reduced, so the efficiency is boosted over
that of the parent cell.
Photocurrent enhancement has been demonstrated in a range of QWSCs and
efficiency enhancement has been observed in materials whose bandgap is larger
than the optimum for solar energy conversion. In materials of bandgap close to the
optimum, experimental tests on QW cells of equivalent quality to homojunction
cells have not yet been possible. Achieving these efficiency improvements needs
carefully grown, defect-free structures. Successive layers are often grown in alter-
nately compressive and tensile stress to allow more QWs to be inserted. For a short
period in 2011, a strain-balanced GaAsP/InGaAs QWSC set a world record for
power conversion efficiency in a single-junction cell.
Because QWSCs are as costly to produce as high efficiency III–V homojunc-
tion cells, they are likely to find application only where III–Vs are preferred. At
the present time this means space, concentrator PV (where the drop in perfor-
mance with increasing temperature is less marked than for the parent cell) and
thermophotovoltaic systems.

1.5 Concentrator photovoltaics

In CPV systems, direct sunlight is concentrated by optical means on small, high-
performing PV cells. The cell area required for a given power output is reduced
by roughly the solar concentration ratio, so cell costs per output watt are reduced,
although the optics and Sun-tracking required add complexity and increase balance-
of-system costs. Concentrations of up to ×200 Suns can be achieved with linear
Fresnel lenses or parabolic mirrored troughs, and low-resistance Si cells are suitable
in this range. Higher concentrations of 250–1000 Suns are achievable with two-
axis solar tracking. III–V MJ cells come into their own at these high irradiances,
because their very high efficiency justifies their high cost.
CPV was developed at Sandia National Laboratories in the 1970s, but there
was virtually no commercial activity until the 1990s, when firms such as Amonix
and Entech were founded, and a number of CPV demonstration plants were built.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 13

The Past and Present 13

In the early 2000s, the combination of the relatively high cost of flat-plate c-Si
PV systems and the impressive improvements in III–V MJ cell efficiency stimu-
lated a number of other commercial CPV ventures. By mid-2011 the cumulative
global installed CPV capacity was 23 MWp, Spain accounting for 70% of this
total.
2012 has seen a more mixed picture. Some commentators predict contin-
ued market growth: Lux Research (2012) saw the market growing to 697 MWp
p.a. by 2017 and IMS Research (2012) predicted that cumulative installations
would reach 1.2 GWp by 2016. However, 2012 has proved to be as difficult a
market for CPV as it has for conventional PV, and commercial CPV ventures have
retrenched. Amonix, which announced a world record CPV module efficiency of
33.5% in October 2012, closed its manufacturing facility in North Las Vegas in
July. The development of III–V multijunction cells of even higher performance
may tilt the economics back in favour of CPV: King et al. (2012) calculate that
50% efficient MJ CPV cells would be the lowest cost option for solar electric-
ity generation in high direct normal irradiance regions. Other commentators are
forecasting that CPV will be unable to compete with the continued low price of
flat-plate PV.

1.6 Inorganic thin-film cells

1.6.1 The cadmium sulphide cell: p-Cu2 S/n-CdS
From the 1970s, when terrestrial applications of c-Si technology began to emerge,
there was a parallel effort to develop other semiconductors in order to make thin-
film (polycrystalline) devices of lower cost and better light-absorbing properties
than c-Si. The original motive for investigating thin-film cells was not, however,
lower cost but their better power-to-weight ratio for space applications. The first
commercially developed thin-film PV device was the cuprous sulphide/cadmium
sulphide ( p-Cu2 S/n-CdS) heterojunction cell, made in 6%-efficient single-crystal
form by Reynolds et al. (1954) at Wright Patterson Air Force Base in the USA,
and in thin-film form by Carlson (1956) at the Clevite Research Center in Cleve-
land, Ohio. This cell excited much interest because of its low intrinsic costs and
the simple way in which the junction is formed, by dipping CdS into a solution
containing cuprous ions. Clevite Corporation mounted a major development effort
on CdS PV in 1964, and several other companies followed suit. However, in spite
of promising results, reviewed by Hill and Meakin (1985), that included the first
thin-film cell of >10% efficiency, these cells suffered from poor stability aris-
ing from the high diffusivity of copper ions into the CdS layer, and there were
also serious problems in making ohmic contacts to Cu2 S. Thus this technology
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 14

14 M. D. Archer

fell from favour. CdS lives on, however, as the window layer of CdTe and CIGS
cells.6

1.6.2 Thin-film (amorphous and microcrystalline) silicon

The Japanese had effectively already delivered the coup de grâce to Cu2 S/CdS
technology by the early 1980s by commercialising small amorphous hydrogenated
silicon (a-Si:H) PV panels of modest but sufficient (3–4%) efficiency to power
small consumer goods such as watches and calculators. Amorphous silicon of
good quality (with sufficiently few mid-gap states to be dopable either n- or p-
type) had earlier been made by Spear and Le Comber (1975) in Dundee, Scotland.
Independently, David Carlson and Chris Wronski, then both at RCA Laboratories
in Princeton, New Jersey, made several square-centimetre sized n-i - p and p-i -n
cells of ∼2% efficiency (Carlson and Wronski, 1976), and smaller area MIS cells
of 5.5% efficiency. The p-i -n configuration was to be the forerunner of modern a-
Si:H photovoltaic technology. The Staebler–Wronski effect, which is the ∼10–30%
diminution of efficiency that occurs on the first prolonged exposure of a cell to light,
was discovered soon afterwards, in 1977. Unwelcome as this was, ways to reduce
its impact by using thin cells (in which the higher built-in field reduces this volume
recombination effect) have been developed.
Single-junction amorphous silicon modules now achieve stabilised efficien-
cies of 6–8%. However, most commercial product now has the so-called micro-
morph configuration, consisting of a two-junction (tandem) cell with an a-Si:H
p-i -n top cell and a microcrystalline (µc-Si:H) p-i -n bottom cell which utilises
the red and near-IR light that passes through the a-Si:H top cell. These cells, orig-
inally developed by Meier et al. (1994) at IMT Neuchatel with manufacturing
technology licensed to the Swiss company Oerlikon Solar in 2003, have achieved
stabilised module efficiencies of 11.9% (Bailat et al., 2010) at a reported cost of
EUR0.5 Wp−1 . Triple-junction cells (a-Si:H/a-SiGe:H/µc-Si:H) of 1.5% stabilised
efficiency have recently been reported (Söderström et al., 2012).

1.6.3 The cadmium telluride cell: n-CdS/p-CdTe

CdTe has long been familiar to the semiconductor industry from its use, in very
pure crystalline form, as a photoconductive high-energy particle detector. Although
it can be doped both n- and p-type, it is hard to make efficient p-n homojunction
CdTe solar cells because of the difficulty of forming a shallow junction with an

6 Cadmium sulphide also lives on in the paintings of impressionists such as Monet, whose favourite
yellow pigment it was.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 15

The Past and Present 15

active top layer in the face of the material’s high surface recombination velocities.
The way forward proved to be the n-CdS/ p-CdTe heterojunction cell, in which
CdTe forms the active, light-absorbing base layer and CdS the front window layer.
The n-CdS/ p-CdTe device structure combines good optical transparency with suf-
ficiently close lattice and thermal matching to form a ‘good’ (spike-free) junction,
albeit after a special activation process. Single-crystal n-CdS/ p-CdTe cells of up to
8% efficiency were made in the 1970s (Saraie et al., 1972; Yamaguchi et al., 1977;
Mitchell et al., 1977), and this good performance allied to ease of junction forma-
tion and tolerance to materials purity that caught the attention of industry. General
Electric was an early market leader in CdTe cell manufacture (Cusano, 1962, 1966).
BP Solar followed in the 1990s with a research programme at Sunbury-on-Thames
near London on electrodeposition of CdTe cells, and later module production in
the US, but axed this programme in late 2002. Dieter Bonnet, who co-authored the
account of CdTe PV in Chapter 5, was one of the pioneers of this field. With his
co-worker Rabenhorst, he was the first to report the all-thin-film CdS/CdTe hetero-
junction cell (Bonnet and Rabenhorst, 1972) and he founded ANTEC in Kelkheim,
Germany, which for some years fabricated 7%-efficient cells deposited by closed
space sublimation (CSS).
The CdTe market phenomenon of the late 2000s — First Solar of Tempe,
Arizona — was founded by entrepreneur Harold McMaster in 1990 and is now
owned by the investment arm of the family that owns Walmart. The company has
focussed aggressively on cost reduction, with impressive results. In 2009, First
Solar’s CdTe modules were the first to break the US$1 Wp−1 cost barrier (one PV
equivalent of the 4-minute mile) and the company was the top PV supplier of that
year. Since then, the falling price of c-Si modules has eroded the value proposition
for CdTe, and First Solar, like other PV manufacturers, has had to retrench.
Best research-cell efficiency plateaued at 16.7% for a decade but has now
reached 18.7% in a cell made with First Solar’s commercial-scale manufacturing
equipment and materials (First Solar, 2013), and module efficiencies of above
11% are routinely achieved (Semiconductor Today, 2011). Given that the CdTe
bandgap of 1.44 eV is almost ideally matched to the terrestrial solar spectrum, cell
efficiencies of >20% should be possible.
Cadmium is highly toxic, and tellurium is toxic if ingested. However, fears
about the safety of CdTe technology have been largely allayed by tests showing that
module encapsulation in glass prevents escape of CdTe to the environment, though
the safe disposal of panels will require attention. Te is currently classified as an
extremely rare element (1–5 ppb in the Earth’s crust) and this could cause supply
difficulties if CdTe PV were to become a mainstream power provider. However, it
has recently been discovered that some undersea ridges are rich in the element.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 16

16 M. D. Archer

1.6.4 Copper indium gallium diselenide (CIGS) cells

CIS/CIGS (copper indium selenide/copper indium gallium selenide) is another
promising thin-film PV technology. The very high optical absorptivity of CIGS is
an advantage, although its quaternary nature led to past difficulties in composition
control during manufacture. Like CdTe, CIS/CIGS can exhibit both n- and p-type
conductivity arising from intrinsic defects, but it is better used in the p-type form
in a heterojunction device with an n-CdS top window layer. 12%-efficient single-
crystal n-CdS/ p-CuInSe2 cells were made by Wagner et al. (1974) and Shay et al.
(1975), and thin-film cells of 4–5% efficiency quickly followed (Kazmerski, 1976).
By the end of the 1980s, commercialisation efforts by PV pioneer ARCO
through its subsidiary ARCO Solar had achieved thin-film CIS modules with areas
of up to 1 × 4 ft2 and ∼10% efficiency. Persistent problems with the process yield
were later overcome by control of sodium impurities in the CIS film and improved
junction fabrication processes. The work of the EuroCIS consortium in the early
1990s resulted in significant efficiency increases to ∼16%.
The current best laboratory cell efficiency is an impressive 20.8% (ZSW, 2013),
with sub-module efficiency not far behind at 17.8% (Solar Frontier, 2012). Like
other PV technologies, commercial CIGS has not been immune to the continued
declining prices of c-Si modules. The 2011 bankruptcy of California-based CIGS
start-up Solyndra, which had received substantial US federal support, became a
2012 presidential campaign issue. The Japanese company Solar Frontier, a 100%
subsidiary of Showa Shell Sekiyu, continues as the only vendor shipping CIGS
product in volume.

1.7 Organic and hybrid technologies

1.7.1 Organic photovoltaics
The observation of photoeffects in organic materials has a long history: the photo-
conductivity of crystalline anthracene was studied over a century ago by Pochet-
tino (1906). Photoconductivity in such organic solids arises because the conjugated
π-electronic orbitals of neighbouring molecules ovelap sufficiently to allow charge
transport by hopping. Light absorption in organic semiconductors creates short-
lived excitons, rather than ‘free’ carriers, as in broadband inorganic semiconduc-
tors. However, the presence of an electric field can assist exciton dissociation and
the drift-assisted separation of electron-hole pairs. In an organic photoconducting
device, the field is supplied by the external electric bias on the electrodes. In an
organic photovoltaic cell, the field is usually created at the heterojunction between
two different materials, one with donor, and the other with acceptor, properties.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 17

The Past and Present 17

Porphyrins, phthalocyanines and chlorophyll, all strongly-coloured π-conju-

gated compounds, were among the earliest organic semiconductors to be investi-
gated for photovoltaic activity. Kearns and Calvin (1958) reported a photovoltage
of 200 mV, but negligible power output, in a bilayer cell with magnesium phthalo-
cyanine as the electron donor and oxidised tetramethyl p-phenylene diamine as the
acceptor. Ghosh et al. (1974) found a maximum power conversion efficiency of
only 0.01% in an Al/MgPh/Au cell under monochromatic illumination at 690 nm.
Tang and Albrecht (1975a; 1975b) obtained an efficiency of 0.05% in a Cr/Chl-
a/Hg cell under monochromatic illumination at 745 nm.
The very poor efficiency of these early organic photovoltaic (OPV) cells was
due to a number of factors, high electrical resistance, poor materials control and
low internal quantum efficiency prominent among them. The thin (70 nm) copper
phthalocyanine/ perylene tetracarboxylic acid cell of Tang (1986) had a lower
resistance and achieved an improved power conversion efficiency of 1%.
Since these pioneer endeavours, steady and substantial improvement in OPV
efficiency has been achieved by better junction engineering and better materials
purity and control; at the time of writing, the best reported efficiency is 10.7%
in Heliatek’s tandem OPV cell (Heliatek, 2013). The development of bulk het-
erojunctions (also called dispersed heterojunctions) was a major step forward. In
bulk heterojunction (BHJ) cells, the donor and acceptor materials are intimately
mixed together while remaining two separate phases, so that excitons generated
anywhere in either material are within a diffusion length of an interface between
the two at which they are dissociated and mobile charge carriers are formed. The
development of high-quality dopable polymers for the OLED industry and other
optoelectronic applications has also been of major assistance to OPV. The first
small molecule dye/dye BHJ cell was reported by Hiramoto et al. (1991) and the
first polymer BHJ cell by Yu et al. (1994).
OPV has the advantages of cheapness, lightness and bandgap tunability. Their
disadvantage is their poor stability and lifetime. OPV companies, notably Konarka,
have been among those PV companies to have been overcome by falling c-Si prices
and competition from more durable and efficient thin-film products. However, there
should be a place even in today’s harsh PV market for off-grid small-power OPV
applications such as the Indigo ‘pay-as-you-go’ power offered by Eight19 (2012), a
company spun off from the Cavendish Laboratory of the University of Cambridge.
Spanggaard and Krebs (2004) give an excellent account of the history of OPV.

1.7.2 Dye-sensitised solar cells

The modern era of dye-sensitised solar cells (DSSC) began with the seminal report
by O’Regan and Grätzel (1991) of a 7%-efficient cell containing a ruthenium
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 18

18 M. D. Archer

dye-sensitised mesoporous TiO2 photoanode and an iodide/triiodide redox couple

in a liquid organic solvent. The efficiency of this cell greatly exceeded that of any
previous dye-sensitised photoelectrochemical cell, and the reason for this remained
unclear until kinetic studies revealed the main factor as the exceptionally slow rate
of back electron transfer from TiO2 into the oxidised dye or I− 3.
Since the discovery of the Grätzel cell, as it is sometimes known, DSSC effi-
ciencies have increased through improvements in all cell components, but par-
ticularly the redox couple and the dye. Cobalt2+/3+ -based redox couples have
increase the open-circuit voltage above that possible with I− /I− 3 , and donor-π-
acceptor dyes absorb much more strongly than the original ruthenium-based com-
plexes (Mishraet al., 2009). The current world-record DSSC cell is 12.3% efficient
under standard test conditions (STC), and this increases to 13.1% under low light
intensity (Yella et al., 2011). This good performance under low light, coupled with
the fact that the cell materials do not need to be rigorously purified, has stimu-
lated considerable commercial interest in BIPV (building-integrated photovoltaic)
and other applications of DSSC, applied as translucent films on glass. However,
liquid electrolytes are volatile and the I− / I−
3 couple is corrosive, so a solid-state
hole conductor generally replaces both in commercial product. Solid-state DSSC
efficiencies have improved over the year, currently to a record 7.2% in a cell
with an organic D-π-A sensitiser and p-doped spiro-MeOTAD as hole conductor
(Burschka et al., 2011). Hardin et al. (2012) describe improvements that could take
DSSC efficiencies towards 20%.

1.7.3 Perovskite solar cells

Hybrid organic/inorganic perovskite solar cells based on methylammonium lead
halides have made a dramatic debut onto the PV scene, demonstrating efficiencies
that have risen to over 15% in only four years of development. Back in the 1990s,
there was considerable interest in the electronic properties of solution-processable
organometallic perovskites in applications such as ferroelectric and superconduct-
ing materials. In particular, David Mitzi and colleagues at IBM examined some
electronic applications of these materials (Mitzi et al., 1994; Kagan et al., 1999).
However, it was the Tokyo-based group of Tsutomu Miyasaka (Kojima et al., 2009)
who made the discovery that the molecular dye sensitiser of the ‘classical’ DSSC
could be replaced by a nanocrystalline film of CH3 NH3 PbBr3 or CH3 NH3 PbI3 ,
yielding a cell of 3.8% efficiency in the case of the triiodide. These materials have
the perovskite structure and are direct-gap semiconductors, with a bandgap that
can be tuned through the visible and into the near IR by adjusting the mole frac-
tion x in CH3 NH3 Pb(I1−x Brx )3 . The charge carriers generated in them by light
absorption are quite long-lived and mobile in these materials, with both holes and
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 19

The Past and Present 19

electrons having diffusion lengths of ∼100 nm. Moreover, little energy is lost in
separating the charge carriers (as compared with the losses incurred in exciton dis-
sociation in OPV) so high open-circuit voltages are achieved. Finally, the materials
can be prepared in tailored compositions and absorption thresholds by low-cost,
low-temperature solution methods.
The first cells that were made were not stable because the perovskites were
attacked by the liquid electrolyte of the DSSC . However, it was soon discovered
that this could be substituted by a solid-state hole conductor, and then that this
could be omitted as the materials are ambipolar and carry both holes and electrons
(Etgar et al., 2012). Other features of the DSSC also proved expendable: Henry
Snaith’s group at Oxford showed that by using CH3 NH3 PbI2 Cl the mesostructured
TiO2 scaffold could be replaced by alumina (Lee et al., 2012), possibly because
of the order of magnitude increase in the exciton/charge carrier diffusion length in
CH3 NH3 PbI2 Cl compared with CH3 NH3 PbI3 (Stranks et al., 2013). It was also
found that the perovskite itself did not have to be nanostructured: a simple planar
heterojunction cell consisting of a few hundred nanometres of vapour-deposited
perovskite sandwiched between two charge-selective electrodes can have efficien-
cies of over 15% (Liu et al., 2013). Michael Grätzel’s group (Burshka et al., 2013),
working with a mesostructured TiO2 scaffold, showed that sequential deposition
of the perovskite, first laying down PbI2 and then converting it to perovskite by
exposure to CH3 NH3 I in solution, yields more consistently high-performance cells
than co-deposition.
Perovskite cells will certainly be an active area of research in the next few
years, with some commentators predicting that efficiencies of over 20% may be
possible (Park, 2013). However, many questions about stability, durability and
device repeatability remain to be addressed.

1.8 Overview of photovoltaic cell operation

This book aims to present an in-the-round approach to PV technology, touching
on aspects from the choice of semiconductor materials through system design to
market status. But PV cells themselves and how they work form its main subject
matter. By way of introduction to the specialist chapters that follow this one, I
therefore conclude this chapter with an account of the main PV cell types and the
basic principles of cell operation. Archer et al. (1996) give a more detailed account
of analytic solutions for p-n junction cells than is here presented.

1.8.1 The p-n homojunction cell

All PV cells, with the exception of dye-cells, work in essentially the same way.
They contain a junction between two different materials across which there is a
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 20

20 M. D. Archer

Figure 1.2 The essential features of a p-n homojunction Si solar cell.

‘built-in’ electric field. When the cell absorbs light, excitons (electron-hole pairs)
are created. In broadband (most inorganic) semiconductors, these excitons very
quickly dissociate into mobile electrons and holes, which flow in opposite direc-
tions across the junction. In narrow-band and organic semiconductors, excitons do
not dissociate spontaneously but are separated at the junction. In either case the flow
of absorbed photons is converted into a flow of DC power from the illuminated cell.
The crystalline silicon (c-Si) cell has a simple junction structure, and provides
a good model with which to explore the PV effect. Figure 1.2 shows the essential
features of these cells, which are typically square wafers of dimensions ∼15 cm ×
15 cm × 0.2 mm. The top (emitter) region is a 0.2–0.3 µm thick layer of n-type
silicon, and the base region is a 300 µm thick layer of p-type silicon.7 The work
function of the p material is greater than that of the n material, so the two layers
reach electronic equilibrium (in the dark) by the transfer of some electrons from
the n to the p side. The structure as a whole remains electrically neutral, but the
junction region contains an electric double layer, consisting of two space-charge
regions or depletion regions (DRs), as shown in Fig. 1.3. These DRs are typically
less than a micron thick, and the charges they contain are those of the ionised
dopants (P+ and B− in the case of c-Si). Beyond the base-layer DR in the c-Si cell
(and some other cells) lies a quasineutral region (QNR) — a region that contains
no space charge.
Figure 1.4 shows what happens in the illuminated c-Si cell. The absorption
of photons of energy greater than the bandgap energy of silicon promotes elec-
trons from the valence band to the conduction band, creating hole-electron pairs

7 c-Si cells are usually configured n-on- p because this best suits the properties of silicon, but some other
p-n cells are configured p-on-n. These cells are also quite thick, because c-Si absorbs light relatively
weakly. Most other cells are much thinner.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 21

The Past and Present 21

Figure 1.3 Cross section through a p-n homojunction cell, showing the electrical double
layer consisting of ionised dopant atoms (denoted + and –) in the junction region, the two
depletions regions (DRs) that contain equal and opposite quantities of junction charge, and
the base-layer quasineutral region (QNR).

Figure 1.4 Generation and movement of free carriers in a p-n junction cell.

throughout the illuminated part of the cell, which in c-Si cells extends well into
the base layer. In c-Si and most other semiconductors, these hole-electron pairs
quickly dissociate into ‘free’ carriers — mobile holes and electrons that move
independently of each other. Those free carriers that approach the junction come
under the influence of the built-in electric field, which sweeps electrons from the
p to the n side, and holes from the n to the p side.

1.8.2 Junction structure and dark current

The electric double layer at the p-n junction has an important effect on the semi-
conductor energy levels, as shown in Fig. 1.5. The separate (uncharged) phases
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 22

22 M. D. Archer

(a) (b)
Figure 1.5 Energy band structure of a p-n homojunction in the dark: (a) in uncharged
blocks of p-type and n-type semiconductors before contact, showing the conduction and
valence-band energies E c and E v , the forbidden gap E g and the Fermi levels E Fn and E Fn
(red dashed lines) in the n and p phases; (b) across the p-n homojunction after contact and
equilibration of the two phases, showing the electric double layer formed by transient charge
transfer, the depletion regions (DRs) and quasi-neutral regions (QNRs) and the common
Fermi level E F throughout the device.

(Fig. 1.5a) have the same conduction and valence band-edge energies E c and E v ,
separated by the forbidden gap E g , but different work functions  p and n , and
therefore different Fermi levels E F p and E Fn .8 In the equilibrated cell (Fig. 1.5b),
the Fermi level E F is the same throughout the device but the band-edge energies E v
and E c (in common with all the energy levels of the semiconductor) bend across
the junction in response to the local electric field. Inspection of Fig. 1.5 shows that
the equilibrium band-bending energy is qVob is related to the difference in the work
functions of the (separate, uncharged) materials by

q Vbo = n −  p (1.1)

Since the Fermi level in a doped semiconductor normally lies within the forbid-
den gap but near the majority-carrier band edge, qV ob is normally slightly smaller
than the bandgap energy E g . Figure 1.6 shows how the band bending is affected
and a current is caused to flow when a bias voltage V j is applied across the cell
in the dark. At equilibrium (Fig. 1.6a), no net current9 flows through any part of

8 The Fermi level is the energy for which the probability of a state being occupied by an electron is
exactly one-half. In an intrinsic (undoped) semiconductor, the Fermi level falls in the middle of the
forbidden gap. In a lightly doped semiconductor, the Fermi level remains within the forbidden gap but
is near the majority-carrier band edge. In a heavily doped semiconductor, the Fermi level lies within
the majority-carrier band.
9 All the currents given the symbol i in Figs. 1.6–1.8 are strictly speaking current densities.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 23

The Past and Present 23

(a) (b) (c)

Figure 1.6 Dark p-n homojunction cell in the dark (a) at equilibrium; (b) under forward
bias V j ; (c) under reverse bias V j , showing the generation and recombination currents as
dotted lines and the Fermi level E F and hole and electron quasi-Fermi levels E Fh and E Fe
across the junction region as red dashed lines.

the cell. However, small, balanced fluxes of electrons in the conduction band and
holes in the valence band pass each way across the junction. These are referred to
as generation and recombination currents. The (thermal) generation currents i h,gen
and i e,gen shown in Fig. 1.6a come from the minority carriers (electrons in the p
side and holes in the n side) generated throughout the device, albeit at a minuscule
rate, by thermal excitation. Those minority carriers that reach the junction with-
out recombining are swept across it in opposite directions by the strong electric
o
field. The recombination currents i h,rec o
and i e,rec also shown in Fig. 1.6a come
from majority carriers (holes in the p side and electrons in the n side) that flow
‘up’ the band-bending barrier (this is energetically unfavourable, but entropically
favourable because the carriers move from a region of high to low concentration).
At equilibrium, the generation and recombination currents in each band
exactly balance each other. The sum of the hole and electron thermal generation
currents is called the saturation current density i o of the junction.

i o = i h,gen + i e,gen = i h,rec

o
+ i e,rec
o
(1.2)

When a forward bias10 voltage V j is applied across the junction of the dark
cell, the barrier height is reduced to qVb = q(Vbo − V j ), as shown in Fig. 1.6b. This
does not affect the generation currents, but it strongly increases the recombination
currents. The net current across the junction, which is the difference between the
recombination current and the generation current, is called the dark current or

10 Forward biasing a junction means applying a voltage across the device that lowers the band-bending
barrier. Reverse biasing means applying a voltage in the opposite direction.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 24

24 M. D. Archer

junction current i j .

i j (V j ) = i h,rec (V j ) + i e,rec (V j ) − i h,gen − i e,gen

= i h,rec (V j ) + i e,rec (V j ) − i h,rec
o
− i e,rec
o
(1.3)

When a reverse bias (V j < 0) is applied, the barrier height is increased as

shown in Fig. 1.6c to qVb = q(Vbo + |V j |). The generation currents are unaffected,
but the recombination currents are now suppressed. Thus only the very small,
bias-independent saturation current passes.

i j (V j < 0) = −i o (1.4)

The dependence of the recombination currents i h,rec (V j ) and i e,rec (V j ) on V j

is determined by the dominant recombination mechanism of the carriers injected
into the junction. In most cells, the dark current–voltage characteristic conforms
well to the empirical diode equation

i j (V j ) = i o [exp(q V j /βkT ) − 1] (1.5)

where β is called the diode ideality factor. For an ideal junction, in which no
injected carriers recombine in the junction, β = 1. For a non-ideal junction, in
which some carriers do recombine in the junction, 1 < β < 2. For many cells,
Eq. (1.5) is better written as the two-diode equation

i j (V j ) = i o1 [exp(q V j /kT ) − 1] + i o2 [exp(q V j /2kT ) − 1] (1.6)

where the first term corresponds to carriers that move across the junction without
recombining, and the second to the carriers that recombine in mid-gap. Regardless
of the exact form of the diode equation, all PV cells behave as rectifiers in the dark,
showing highly non-linear current–voltage characteristics similar to that labelled
‘dark’ in Fig. 1.8. Junctions must show rectifying properties in the dark if they are
to show photovoltaic properties in the light.

1.8.3 The illuminated cell

When a PV cell is illuminated, a photocurrent and photovoltage are generated.
Figure 1.7 shows how this happens, again using the example of a p-n homojunc-
tion cell. Absorption of photons of energy greater than the bandgap energy of the
semiconductor creates excess minority carriers throughout the illuminated region
of the cell (the light intensity in the cell interior falls off exponentially with dis-
tance into the cell, but often it penetrates into the base layer). The photogenerated
minority carriers in the illuminated cell behave like the much smaller popula-
tion of thermally generated minority carriers in the dark cell. That is, they diffuse
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 25

The Past and Present 25

(a) (b)
Figure 1.7 Illuminated p-n homojunction cell (a) at open circuit, showing the photocurrents
as red dotted lines and the hole and electron quasi-Fermi levels E Fh and E Fe across the junction
region as red dashed lines; (b) at short circuit, assuming that the cell has no internal resistance
(i.e. that carrier mobility is infinite).

from the QNRs towards the junction, where they are swept across it by the strong
junction field. These fluxes of photogenerated minority carriers give rise to the
photogeneration currents i e, ph and i h, ph shown in Fig. 1.7a, consisting respec-
tively of photogenerated electrons drifting from the p to the n side of the junction
and photogenerated holes drifting the other way. The sum of the two is the overall
photocurrent i ph .

i ph = i h, ph + i e, ph (1.7)

The photocurrent is directly proportional to the absorbed photon flux but inde-
pendent of bias (provided that the junction field is always high enough to sweep
carriers across the junction). At open circuit (Fig. 1.7a), no current is drawn from the
cell and the photocurrent must be balanced by the recombination current. The junc-
tion self-biases in the forward direction by the open-circuit voltage Voc , at which
point the recombination (junction) current exactly opposes the photocurrent, i.e.

i ph − i j (Voc ) = 0 (1.8)

As shown in Fig. 1.7a, the hole and electron quasi-Fermi levels E Fh and E Fn
diverge across the junction and converge past the quasi-neutral regions, and qVoc is
the difference between the Fermi levels on the two far sides of the junction. Since
metal contacts always equilibrate with the local majority carrier Fermi level, Voc
is an observable output voltage.
Figure 1.7b shows what happens when the illuminated cell is short-circuited.
The cell delivers maximum current but at zero output voltage. Provided internal
resistance effects are negligible, the junction bias V j is also zero, so the band
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 26

26 M. D. Archer

bending is the same as in the dark junction at equilibrium.11 The short-circuit

current is given by

i sc = |i ph | − i o (1.9)

Under normal operating conditions, the band bending and junction current are
intermediate between the open-circuit and short-circuit cases, and the cell delivers
current i at output voltage V ≈ V j , where i is given by

i = i ph − i j (V j ) (1.10)

Provided the photocurrent i ph is bias-independent, the current–voltage char-

acteristics of the dark and illuminated cells will therefore show superposition. That
is, they will map onto each other, but are shifted with respect to each other by the
constant amount i ph , as shown in Fig. 1.8.
Superposition is an idealisation that is seldom accurately obeyed. It is not to be
expected where the photocurrent is bias-dependent, which can happen for a number
of reasons. In the amorphous silicon cell, for example, the field in the junction region
becomes weaker as the forward bias increases and this makes it more difficult to col-
lect the photogenerated carriers. Cells operating in the high-injection mode, where
the concentration of photogenerated minority carriers becomes comparable with

Figure 1.8 Current–voltage curves in the dark and the light for a cell that shows super-
position (i.e. one in which the photogenerated current is bias-independent), showing the
short-circuit and maximum-power currents i sc and i mp , the open-circuit and maximum-
power voltages Voc and Vmp , and the maximum power point (
).

11 If the cell has significant internal resistance, the output voltage V drops below the junction voltage
V j , and a small forward bias remains across the junction when the cell is short-circuited.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 27

The Past and Present 27

that of the majority carriers, do not show superposition because the majority-carrier
concentrations and fluxes are not then the same in the light and the dark. Cells with
significant internal series resistance or shunt conductance also depart from super-
position.

1.8.4 Cell current–voltage characteristics

The current–voltage characteristic of the illuminated cell is found by substituting
Eq. (1.5) into Eq. (1.10). Assuming superposition and negligible internal resistance
effects, the current–voltage characteristic is given by

i = i ph − i o [exp(q V /βkT ) − 1] (1.11)

The output power is the product iV, which is the area of a rectangle of sides i
and V inscribed in the i –V curve. The power is zero for both the open-circuit and
short-circuit conditions. The maximum-power condition is reached where the area
i mp Vmp (shaded in Fig. 1.8) is a maximum. The fill factor ηfill is a measure of the
squareness of the i –V curve and is defined as
i mp Vmp
ηfill = (1.12)
i sc Voc
In efficient cells, the fill factor is around 0.7–0.8. In poor cells, it can be 0.5 or
lower.
By setting i = 0, V = Voc in Eq. (1.11) and rearranging, the open-circuit
voltage of the illuminated cell is found as



βkT i ph βkT i ph
Voc = ln 1 + ≈ ln (1.13)
q io q io
For good performance, i ph and Voc must be as large as possible. The max-
imum value of i ph would be obtained if all photogenerated electron-hole pairs
were collected as photocurrent, and i ph can achieve over 90% of this limit if light
absorption and minority carrier collection are both highly efficient. If there are no
sources of voltage in the cell other than at the junction, the limiting value of Voc is
the built-in voltage Vbo , corresponding to complete flattening of the bands across
the junction.12 This would only happen under extremely intense illumination, and
1 Sun Voc values are usually no more than ∼0.75 Vbo ; for the best GaAs cells,
Voc ≈ 0.79E g . For a high open-circuit voltage, Vbo should be as large as possible

12 In lightly doped Si cells, V can exceed V o at the junction, with the bulk going into high injection
oc b
and an additional voltage across the rear back-surface field: SunPower cells effectively work in this
way.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 28

28 M. D. Archer

given the semiconductor bandgap, so the work function difference between the two
sides of the junction should be as large as possible.
Inspection of Eq. (1.13) shows that Voc increases as the saturation current i o
decreases. Interestingly, i o has no absolute minimum value. In thin cells with well-
passivated surfaces, i o can be driven down toward zero, and Voc towards its upper
limit of Vbo . In thicker cells in which volume recombination occurs, the lower limit
on i o is determined by the rate of radiative recombination of minority carriers.
Usually nonradiative recombination also occurs and this raises i o by several orders
of magnitude, lowering Voc accordingly.

1.8.5 Cell efficiency

The maximum-power solar conversion efficiency ηmp of a solar cell (often called
simply the cell efficiency) is defined as
i mp Vmp = ηfill i sc Voc
ηmp = (1.14)
IoS IoS
where IoS (watts per unit area) is the incident solar irradiance. Since i ph normally
increases in direct proportion to IoS , while Voc increases as ln i ph (Eq. (1.13)),
it follows that ηmp should increase logarithmically with irradiance, other factors
being equal. This is observed for solar concentrations of up to several hundred
Suns for some cell types, though ultimately the series resistance of the cell and the
increased operating temperature limits the efficiency increase obtainable by using
concentrated sunlight.
Most commercial PV cells have (1 Sun) efficiencies in the range ∼11–18%;
the best commercial cell is now at least 23% efficient. As Fig. 1.1 shows, the best
research cells have higher efficiencies, now up to ∼28.8% for a single-junction
device under AM1.5G Sun. Establishing the theoretical limits of cell efficiency is
of considerable practical importance. Since PV cells are direct conversion devices,
they are not subject to the Carnot limits that control the efficiency of heat engines.
Nevertheless, there are constraints on PV cell efficiency. The major constraint
comes from the poor match between the broadband spectral distribution of sunlight
and the single bandgap E g of a given semiconductor. Solar photons of energy
E < E g are not absorbed in the semiconductor (or if they are, they do not create
hole-electron pairs). Photons of energy E ≥ E g can be absorbed and create hole-
electron pairs, but their initial ‘excess’ energy (E − E g ) is very quickly lost by
thermalisation, that is, dissipation as heat via carrier–phonon collisions.
The bandgap of the photoactive semiconductor determines the upper bound on
both the open-circuit voltage Voc and the short-circuit current i sc . A large-bandgap
cell has a larger Voc than a small-bandgap one, but it absorbs fewer solar photons so
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 29

The Past and Present 29

it has a smaller i sc . The ‘detailed-balance’ AM1.5G limiting efficiency of an ‘ideal’

isotropic single-junction cell13 of optimal bandgap E g = 1.34 eV is 33.8%. In real
cells, ‘non-ideal’ loss mechanisms — for example, nonradiative recombination
of carriers in the cell interior or at junction defects or cell surfaces — lower the
efficiencies below the detailed-balance limit.
The route to high efficiency in a single-junction cell lies in eliminating these
non-ideal losses as far as possible. Higher efficiency can be achieved by stacking
two or three single-junction devices on top of each other so that each absorbs the
portion of the solar spectrum best suited to its bandgap, and the loss of energy from
carrier thermalisation is diminished. This is the approach taken in multijunction
cells. More ambitiously, if thermalisation losses could be avoided altogether, very
high efficiencies of over 80% could be achieved. Green (2000) proposed a number
of ‘third generation’ device designs, such as hot-carrier and thermophotonic cells,
that in principle avoid thermalisation, and instigated a programme to bring these
to ‘proof-of-concept’ level.

1.9 Other junction types

The p-n homojunction examined in the preceding section is not only the simplest
type of PV junction, but also the most common, being that found in crystalline
silicon cells, but there are others. Figure 1.9 shows how the conduction and valence
band-edge energies E c and E v , and the equilibrium Fermi level E F , vary across
the main junction types in solar cells.
Homojunctions (Fig. 1.9a) are p-n junctions formed by the creation of adjacent
p- and n-doped regions in the same semiconductor, of bandgap E g . Homojunction
cells have the advantage that the junction, which is typically about 0.2 m wide,
can be formed in an almost defect-free state within a single crystal (or crystallite
of a polycrystalline material). Outside the junction region, there is no electric
field to assist carrier collection, so the movement of carriers to the junction must
rely on diffusion. Good minority lifetimes are therefore necessary, particularly in
an indirect-gap material like silicon. The number of semiconductors that can be
satisfactorily doped both n- and p-type is limited, but they include Si, as well as
GaAs and InP.
p-i -n junctions (Fig. 1.9b) are junctions of extended width (∼0.5 µm) formed
by interposing an intrinsic (i ) layer of the undoped semiconductor between p- and

13 An ideal isotropic cell is one in which electrons and holes are thermalised to the band edge, the only
decay channel for excited states is radiative recombination, and light can enter the cell at all forward
angles.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 30

30 M. D. Archer

Figure 1.9 Common PV junction types. (a) p-n homojunction, formed within a single
semiconductor of bandgap E g ; (b) p-i-n junction, formed within a single semiconductor of
bandgap E g ; (c) anisotype P-n heterojunction formed between semiconductors of bandgaps
E g1 and E g2 , showing a valence-band spike E v and a conduction-band notch E c ; (d)
P + - p-n heteroface junction; (e) MS junction between a metal M and an n-type semicon-
ductor S; (f) MIS junction with a thin layer of an insulator I interposed between M and S; (g)
organic cell containing an organic co-polymer blend between a top transparent conducting
oxide (TCO) electrode and a metal electrode M. All these junctions are shown at equilibrium
in the dark, so the Fermi levels, shown by the red dashed lines, are the same throughout
each junction.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 31

The Past and Present 31

n-layers of the same semiconductor. The i -layer behaves like the dielectric in
a capacitor, effectively stretching the electric field of an ordinary p-n junction
across itself. The extended electric field throughout the i -layer, which is where
most of the light is absorbed, aids in the collection of photogenerated carriers by
adding a component of drift (migration in an electric field) to their normal diffusive
motion. This is the junction type in amorphous silicon (a-Si:H) cells, and it was
adopted because the high density of states and traps in this disordered material
causes carrier mobilities to be low.
Heterojunctions (Fig. 1.9c) are junctions formed between two chemically dif-
ferent semiconductors with different bandgaps. The larger bandgap material is
often denoted by writing its conductivity type as upper-case N or P, and the
smaller bandgap material by a lower-case letter. In the frontwall configuration,
the top layer is the main light-absorbing layer, and light enters through its front
surface. In the backwall configuration, the top layer is an optical window material
of wide bandgap through which light passes to enter the light-absorbing base layer
at the junction. Heterojunctions may be anisotype, meaning that the two semicon-
ductors have opposite conductivity types, or isotype, meaning that they have the
same conductivity type.
The advantage of this junction type is that it allows semiconductors which have
good light absorption and carrier lifetime properties, but can only be doped n- or
p-type, to be used in solar cells. Its disadvantages are first, that any significant lat-
tice mismatch between the two materials creates numerous junction defects, which
diminishes the photovoltage, and second, that the energy band mismatch between
the two materials creates notches or spikes in the junction band-edge profile. Fig-
ure 1.9c shows a poor junction with a pronounced valence-band spike, which would
seriously impede the collection of photogenerated holes from the n-semiconductor.
The two most successful heterojunctions for PV applications, n-CdS/ p-CdTe and
n-CdS/ p-CuInSe2 , are both N- p anisotype backwall devices with good lattice
matches between the two materials, and favourable band alignment with no spike
presented to carriers arriving at the junction. Heteroface junctions or buried homo-
junctions (Fig. 1.9d) have a p-n homojunction fronted by a highly conducting
window layer (face) semiconductor of larger bandgap (E g1 > E g2 ). This window
layer, although not photovoltaically active, is beneficial. As well as acting as current
collector from the top layer, it passivates its surface and minimises front-surface
recombination. The p+ -Alx Ga1−x As/ p-GaAs/n-GaAs cell is an important exam-
ple of this cell type. An unprotected p-n GaAs homojunction would not make an
efficient solar cell because the top layer of this direct-gap material must be very
thin to allow light to penetrate to the junction region; the dark current would then
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 32

32 M. D. Archer

be undesirably high because of the extent of recombination at the unpassivated

front surface.
Surface junctions are formed when a phase of very high carrier density (a
metal, degenerate semiconductor or a concentrated electrolyte solution) makes
contact with a phase of much lower carrier density, usually a moderately or lightly
doped semiconductor. The charge on the highly conducting side of the junction
then lies virtually ‘in’ the interface whereas that on the poorly conducting side is
spread out over a space-charge layer as usual. All the band bending and poten-
tial drop then occurs on the side of the poor conductor. Photovoltaically active
metal–semiconductor (MS) junctions can be made by choosing a metal and semi-
conductor of such work functions that the semiconductor is in depletion,14 as shown
in Fig. 1.9e. If the semiconductor surface at the MS interface is sufficiently defect-
free, a Schottky barrier is formed, in which the barrier height is purely determined
by the difference in work function of the two materials. If the density of semicon-
ductor surface states is greater than ∼1017 m−2 , a Bardeen barrier is formed in
which the surface states contain most of the charge in the semiconductor. Schottky
behaviour makes for greater band bending and is preferable.
Surface junctions are encountered in a number of guises. Depositing a thin
layer of semiconductor (say, an organic pigment) onto a metal electrode can make
an electrode assembly with photovoltaic activity. Semiconductor/electrolyte solu-
tion junctions are also surface junctions, in which the electrolyte solution plays the
role of the metal in an MS junction. These can perform very well as photoconver-
sion devices, since their dark currents are usually very small and the achievable
photovoltages are therefore high. For example, the Grätzel cell (Grätzel, 2000),
which contain a dye-sensitised TiO2 /I2 , I− junction, has an open-circuit voltage
of ∼0.7 V. However, the dark currents of true metal/semiconductor devices (which
are determined by thermionic emission from the metal) tend to be large, and the
photovoltages correspondingly poor. A better device may be made by interpolating
a very thin (<2 nm) layer of an insulator (I) between the metal film and the semi-
conductor to make an MIS junction (Fig. 1.9f). The I layer impedes thermionic
emission and improves the photovoltage without detriment to the photocurrent.
Organic cells (Fig. 1.9g) typically contain one or two thin organic layers
sandwiched between a transparent electrode and a metal. Organic materials do not
support a space charge. Consequently such cells behave electrically like capacitors,
and the band-edge energies drop linearly across the organic layer(s). Photon absorp-
tion generates excitons rather than free carriers. These generally dissociate only
at an electrode or at the junction between two dissimilar organic layers. Organic

14 Rather than in accumulation, which would create a photovoltaically inactive ohmic contact.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 33

The Past and Present 33

materials are sufficiently soft that the cell constituents may be co-blended to create
a bulk heterojunction cell with interfaces throughout the device. This increases
the chance of exciton formation within a diffusion length of the nearest junction.
However, even the best OPV cells have low open-circuit voltages for their effec-
tive bandgap, arising from their poor radiative efficiencies (i.e. high nonradiative
recombination rates), which is due to their blended structure.

1.10 Structure of this book

Martin Green, whose group at the University of New South Wales has been respon-
sible for a string of c-Si efficiency records, discusses the limits to photovoltaic
efficiency in Chapter 2 and the current status of crystalline silicon photovoltaics in
Chapter 3. Christophe Ballif, Matthieu Despeisse and Franz-Josef Haug provide a
full account of amorphous and microcrystalline silicon technology in Chapter 4.
Tim Gessert of NREL and thin-film pioneer Dieter Bonnet examine CdTe technol-
ogy in Chapter 5. Uwe Rau and Hans-Werner Schock cover CIGS technology in
Chapter 6. Masafumi Yamaguchi, who has been at the forefront of III–V solar cell
technology for over 25 years, discusses GaAs and its use in conjunction with other
III–V semiconductors in high-efficiency MJ cells in Chapter 7. Dan Credgington
describes the current status of OPV in Chapter 8. Michael Grätzel, the distinguished
founding father of DSSC, and James Durrant describe the current status of this tech-
nology in Chapter 9. Quantum well solar cells, in which III–V semiconductors are
typically used for both the quantum well and the host material, are described by
Jenny Nelson and Ned Ekins-Daukes in Chapter 10. The expert father-and-son
team of Ignacio Luque-Heredia and Antonio Luque discuss all aspects of CPV in
Chapter 11.
The final two chapters of the book are devoted to PV systems and applications:
Nicola Pearsall discusses modules, systems and applications in Chapter 12, and
Arnulf Jäger-Waldau covers manufacturers and markets in Chapter 13.

References
Adams W. G. and Day R. E. (1877), ‘The action of light on selenium’, Proc. Roy. Soc. A25,
113–117.
Ando K., Amano C., Sugiura H., Yamaguchi M. and Salates A. (1987), ‘Nonradiative e-h
recombination characteristics of mid-gap electron trap in Al x Ga1−x As (x = 0.4)
grown by molecular beam epitaxy’, Jpn. J. Appl. Phys. 26, L266–L269.
Archer M. D., Bolton J. R. and Siklos S. T. C. (1996), A review of analytic solutions for
a model p-n junction under low-injection conditions,Solar Energy Mater. Solar Cells
40, 133–176.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 34

34 M. D. Archer

Bailat J., Fesquet L., Orhan J.-B., Djeridane Y., Wolf B., Madhiger P., Steinhauser J., Benagli
S., Borrello D., Castens L., Monteduro G., Marmelo M., Dehbozorgi B., Vallat-Sauvain
E., Multone X., Romang D., Boucher J.-F., Meier J., Kroll U., Despeisse M., Bugnon
G., Ballif C., Marjanovic S., Kohnke G., Borrelli N., Koch K., Liu J., Modavis R.,
Thelen D., Vallon S., Zakharian A. and Wiedman D. (2010), ‘Recent developments
of high-efficiency micromorph tandem solar cells in KAI-M PECVD reactors’, Proc.
5th. World Conf. on Photovoltaic Energy Conversion, Valencia, Spain, 6–10 September
2010.
Barnham K. W. J. and Duggan G. (1990), ‘A new approach to high-efficiency multi-bandgap
solar cells’, J. Appl. Phys. 67, 3490–3493.
Becquerel A. E. (1839), ‘Recherches sur les effets de la radiation chimique de la lumière
solaire, au moyen des courants électriques’, Compt. Rend. Acad. Sci. 9, 145–149,
561–567.
Benzer S. (1946), The photo-diode and photo-peak characteristics in germanium’, Phys.
Rev. 70, 105.
Benzer S. (1947), ‘Excess-defect germanium contacts’, Phys. Rev. 72, 1267–1268.
Bergmann L. and Hansler J. ‘Lichtelektrische Untersuchungen an Halbleitern’, Zeits. Physik
100, 50– 79.
Bonnet D. and Rabenhorst H. (1972), ‘New results on the development of thin film p-
CdTe/n-CdS heterojunction solar cell’, Proc. 9th IEEE PVSC, Silver Springs, MD
(IEEE, 1972), pp. 129–132.
Brittlebank W. (2012), ‘Solar PV approaching grid parity’, www.climateaction.org, 10
August 2012, retrieved 12 January 2013.
Burschka J., Kessler D. A., Baranoff E., Cevey-Ha N. L., Yi C., Nazeeruddin M. K. and
Grätzel M. (2011), ‘Tris(2-(1H-pyrazol-1-yl)pyridine)cobalt(III) as p-type dopant for
organic semiconductors and its application in highly efficient solid-state dye-sensitized
solar cells’, J. Amer. Chem. Soc.133, 18042–18045.
Burschka J., Pellet N., Moon S.-J., Humphry-Baker R., Gao P., Nazeeruddin M. K. and
Grätzel M. (2013), ‘Sequential deposition as a route to high performance perovskite-
sensitized solar cells’, Nature 499, 316–319.
Carlson A. (1956), Research in Semiconductor Films, WADC Technical Report, Clevite
Corporation.
Carlson D. E. and Wronski C. R. (1976), ‘Amorphous silicon solar cells’, Appl. Phys. Lett.
28, 671–673.
Case T. W. (1920), “‘Thalofide cell” — a new photoelectric substance’, Phys. Rev. 15,
289–292.
Chapin D. M., Fuller C. S. and Pearson G. O. (1954), ‘A new silicon p–n junction
photocell for converting solar radiation into electrical power’, J. Appl. Phys. 25,
676–677.
China Briefing (2012), ‘China releases twelfth five-year plan on solar power development’,
www.china-briefing.com/news, 19 September 2012, retrieved 7 January 2013.
Crossley P. A., Noel G. T. and Wolf M. (1967), ‘Review and evaluation of past
solar cell development efforts’, Report prepared under contract for NASA.
http://ntrs.nasa.gov/archive/nasa/casi.ntrs.nasa.gov/19670022851_1967022851.pdf.
Cusano D. A. (1962), ‘Polycrystalline thin-film CdTe solar cells’, IRE Trans. Electron
Devices ED–9, 504.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 35

The Past and Present 35

Cusano D. A. (1966), ‘The performance of thin film solar cells employing photovoltaic
Cu2−x Te–CdTe heterojunctions’, Rev. Phys. Appl. 1, 195–200.
Davydov B. I. (1938), ‘Theory of rectification in the semiconductors’, Bull. Acad. Sci.
U.S.S.R. 5–6, 625–629.
Davydov B. I. (1939), ‘Contact resistance of semiconductors’, Zh. Eksp. Teor. Fiz. 9,
451–458.
DECC (2013), UK Solar PV Strategy Part 1: Roadmap to a Brighter Future, Department
of Energy and Climate Change, October 2013.
Eight19 (2012), ‘Eight19 spins out Indigo pay-as-you-go solar’, www.eight19.com, 21
August 2012, retrieved 12 January 2013.
EPIA (2013), Global Market Outlook for Photovoltaics 2013–2017.
Etgar L., Gao P., Xue Z., Peng Q., Chandiran A. K., Liu B., Nazeeruddin M. K. and Grätzel M.
(2012), ‘Mesoscopic CH3 NH3 PbI3 /TiO2 heterojunction solar cells’, J. Amer. Chem.
Soc. 134, 17396–17399.
Fink C. G. and Fogle M. E. (1934),‘A study of cuprous oxide solid photoelectric cells’,
J. Electrochem. Soc. 66, 271–322.
First Solar (2013), ‘First Solar sets new world record for CdTe solar cell efficiency’,
http://investor.firstsolar.com press release 26 February 2013, retrieved 5 April 2013.
Fraunhofer (2013), ‘World record solar cell with 44.7% efficiency’, www.ise.fraunhofer.de
23 September 2013, retrieved 30 November 2013.
Fraunhofer Institute for Solar Energy Systems ISE (2012), Photovoltaics Report, December
11, 2012.
Fritts C. E. (1883), ‘On a new format of selenium cell, and some electrical discoveries made
by its use’, Am. J. Sci. 26, 465–472.
Garrison A. D. (1923), ‘The behaviour of cuprous oxide photo-voltaic cells’, J. Phys. Chem.
27, 601–622.
Geiger P. H. (1923), ‘Spectro-photoelectrical effects in argentite: the production of an elec-
tromotive force’, Phys. Rev. 22, 461– 469.
Ghosh A. K., Morel D. L., Feng T., Shaw R. F. and Rowe C. A. (1974), ‘Photovoltaic and
rectification properties of Al/Mg phthalocyanine/Ag Schottky-barrier cells’, J. Appl.
Phys. 45, 230–236.
Grätzel M. (2000), ‘Perspectives for dye-sensitized nanocrystalline solar cells’, Progr. Pho-
tovoltaics 8, 171–185.
Green M. (1993), ‘Silicon solar cells: evolution, high-efficiency design and efficiency
enhancements’, Semiconductor Sci. Technol. 8, 1–12.
Green M. A. (2000), ‘Third generation photovoltaics: advanced structures capable of high
efficiency at low cost’, 16th. European Photovoltaic Solar Energy Conf., Glasgow, 1–5
May 2000.
Green M. A. (2009), ‘The path to 25% silicon solar cell efficiency: history of silicon cell
evolution’, Progr. Photovoltaics 17, 183–189.
Green M. A., Blakers A. W. and Osterwald C. R. (1985), ‘Characterization of high-efficiency
silicon solar cells’, J. Appl. Phys. 58, 4402–4408.
Green M. A., Emery K., Hishikawa Y., Warta W. and Dunlop E. D. (2013), ‘Solar cell
efficiency tables (version 41)’, Progr. Photovoltaics 21, 1–11.
Gremmelmaier R. (1955), ‘Gallium-arsenic photoelement’, Z. Naturforsch. A 19,
501–502.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 36

36 M. D. Archer

Grondahl L. O. (1933), ‘The copper-cuprous-oxide rectifier and photoelectric cell’, Rev.

Mod. Phys. 5, 141–168.
Grondahl L. O. and Geiger P. H. (1927), ‘A new electronic rectifier’, J. Am. Inst. Elec. Eng.
46, 215–222.
GTM Research (2012), Thin Film 2012–2016: Technology, Markets and Strategies for Sur-
vival.
GTM Research (2013), ‘Yingli gains crown as top producer in a 36 GW global PV market’,
www.greentechmedia.com 1 May 2013, retrieved 30 November 2013.
Hardin B. E., Snaith H. J. and McGehee M. D. (2012), ‘The renaissance of dye-sensitized
solar cells’, Nature Photonics 6, 162–169.
Haynos J., Allison J., Arndt R. and Meulenberg A. (1974), ‘The COMSAT non-reflective
silicon solar cell: a second generation improved cell’, Int. Conf. on Photovoltaic Power
Generation, Hamburg, p. 487.
Heliatek (2013), ‘Heliatek sets new world record of 10.7% for its organic tandem cell’,
www.heliatek.com/newscenter 24 July 2012, retrieved 5 April 2013.
Hill R. and Meakin J. D. (1985), ‘Cadmium sulphide–copper sulphide solar cells’, Ch. 5 in
Coutts T. J. and Meakin J. D., eds., Current Topics in Photovoltaics, Vol. 1, Academic
Press, London.
Hiramoto M., Fujiwarai H. and Yokoyama. M. (1991), ‘Three-layered organic solar cell with
a photoactive interlayer of co-deposited pigments’, Appl. Phys. Lett. 58, 1062–1064.
IMS Research (2012), The World Market for Concentrated PV, IMS Research, 26 September
2012.
Jackson W. D. (1955), Trans. Conf. on the Use of Solar Energy 5, Tucson: University of
Arizona Press, p.122.
Jenny D. A., Loferski J. J. and Rappaport P. O. (1956), ‘Photovoltaic effect in GaAs p-n
junctions and solar energy conversion’, Phys. Rev. 101, 1208–1209.
Kagan C. R., Mitzi D. B. and Dimitrakopoulos C. D. (1999), ‘Organic-inorganic hybrid
materials as semiconducting channels in thin-film field-effect transistors’, Science 286,
945–947.
Kazmerski L. L., White F. R. and Morgan G. K. (1976), ‘Thin-film CuInSe2 /CdS hetero-
junction solar cells’, Appl. Phys. Lett. 29, 268–270.
Kearns D. and Calvin M. (1958), ‘Photovoltaic effect and photoconductivity in laminated
organic systems’, J. Chem. Phys. 29, 950–951.
Kennard E. H. and Dieterich E. O. (1917), ‘An effect of light upon the contact potential of
selenium and cuprous oxide’, Phys. Rev. 9, 58–63.
King R. R., Bhusari D., Larrabee D., Liu X.-Q., Rehder E., Edmondson K., Cotal H., Jones
R. K., Ermer J. H., Fetzer C. M., Law D. C. and Karam N. H. (2012), ‘Solar cell
generations over 40% efficiency’, Progr. Photovoltaics 20, 801–815.
King R. R., Law D. C., Edmondson K. M., Fetzer C. M., Kinsey G. S., Yoon H., Sherif R. A.
and Karam N. H. (2007), ‘40% efficient metamorphic GaInP/GaInAs/Ge multijunction
solar cells’, Appl. Phys. Lett. 90, 183516–183519.
Kojima A., Teshima K., Shirai Y. and Miyasaka T. (2009), ‘Organometal halide per-
ovskites as visible-light sensitizers for photovoltaic cells’, J. Amer. Chem. Soc. 131,
6050–6051.
Lange B. (1930), ‘Uber eine Neue Art von Photozellen’, Phys. Zeits. 31, 964–969.
Lange B. (1938), Photoelements and Their Applications, Reinhold, New York.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 37

The Past and Present 37

Lee M. M., Teuscher J., Miyasaka T., Murakami T. N. and Snaith H. J. (2012), ‘Effi-
cient hybrid solar cells based on meso-superstructured organometal halide perovskites’,
Science 338, 643–647.
Lindmeyer J. and Alison J. F. (1973), ‘The violet cell: an improved silicon solar cell’,
COMSAT Technical Review 3, 1–22.
Liu M., Johnston M. B. And Snaith H. J. (2013), ‘Efficient planar heterojunction perovskite
solar cells by vapour deposition’, Nature 501, 395–398.
Lux Research (2012), Putting High-Concentrating Photovoltaics into Focus, June 2012.
Mahdi W. (2012), ‘Qatar to tender 200 megawatt of solar power projects in 2013’,
www.bloomberg.com, 3 December 2012, retrieved 3 January 2013.
Meier J., Dubail S., Flückiger R., Fischer D., Keppner H. and Shah A. (1994), ‘Intrinsic
microcrystalline silicon (mc-Si:H) — a promising new thin film solar cell material’,
Proc. 1st. World Conf. on Photovoltaic Energy Conversion, Hawaii, pp. 409–412.
Mishra A., Fischer M. K. R. and Bauerle P. (2009), ‘Metal-free organic dyes for dye-
sensitized solar cells: from structure–property relationships to design rules’, Angew.
Chem. Int. Ed. 48, 2474–2499.
MIT (2011), ‘Why BP Solar failed’, www.technologyreview.com 21 December 2011,
retrieved 26 January 2014.
Mitchell K., Fahrenbruch A. L. and Bube R. H. (1977), ‘Evaluation of the CdS/CdTe het-
erojunction solar cell’, J. Appl. Phys. 48, 4365–4371.
Mitzi D. B., Feild C. A., Harrison W. T. A. and Guloy A. M. (1994), ‘Conducting tin halides
with a layered organic-based perovskite structure’, Nature 369, 467–469.
Mott N. F. (1939), ‘Copper–cuprous oxide photocells’, Proc. Roy. Soc. A171, 281–285.
Ohl R. S. (1941), ‘Light-sensitive electric device’, U.S. Patent No. 2,402,662; ‘Electrical
translating device utilizing silicon’, U.S. Patent No. 2,402,839; ‘Light-sensitive device
including silicon’, U.S. Patent No. 2,443,542.
Olson J. M., Kurtz S. R., Kibbler A. E. and Faine P. (1990), ‘A 27.3% efficient
Ga0.5 In0.5 P/GaAs tandem solar cell’, Appl. Phys. Lett. 56, 623–625.
O’Regan B. and Grätzel M. (1991), ‘A low-cost, high-efficiency solar cell based on dye-
sensitized colloidal TiO2 films’, Nature 353, 737–740.
Park N.-G. (2013), ‘Organometal perovskite light absorbers toward a 20% efficiency low-
cost solid-state mesoscopic solar cell’, J. Phys. Chem. Lett. 4, 2423–2429.
Pochettino A. (1906), ‘Sul comportamento foto-elettrico dell’ antracene’, Acad. Lincei Ren-
diconti 15, 355–363.
PV Magazine (2013), ‘Solar stocks: China left for dead; FSLR, SPWR and WFR to rise’,
www.pv-magazine.com, 2 April 2013, retrieved 6 April 2013.
PV Tech (2012), ‘Around 90% of Chinese polysilicon producers stop production’, www.pv-
tech.org, 13 December 2012, retrieved 8 January 2013.
Reynolds D. C., Leies G., Antes L. L. and Marburger R. E. (1954), ‘Photovoltaic effect in
cadmium sulfide’, Phys. Rev. 96, 533–534.
Saraie J., Akiyama M. and Tanaka T. (1972), ‘Epitaxial growth of cadmium telluride by a
closed-space technique’, Jpn. J. Appl. Phys. 11, 1758–1759.
Schottky W. (1930), ‘Über den Entstehungsort der Photoelektronen in Kupfer–
Kupferoxydul–Photozellen’, Zeit. Tech. Phys. 11, 458–461.
Schottky W. (1938), ‘Halbleitertheorie der Sperrschicht’, Naturwiss. 26, 843.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 38

38 M. D. Archer

Semiconductor Today (2011), ‘First Solar raises CdTe PV cell efficiency record from 16.7%
to 17.3%’, www.semiconductor-today.com, 26 July 2011, retrieved 12 January 2013.
Shay J. L., Wagner S. and Kasper H. M. (1975), ‘Efficient CuInSe2 /CdS solar cells’, Appl.
Phys. Lett. 27, 89–90.
Sheldon H. H. and Geiger P. H. (1922), ‘The production of an E.M.F. on closed circuit, by
a light effect on argentite’, Phys. Rev. 19, 389– 390.
Siemens W. (1885), ‘On the electromotive action of illuminated selenium discovered by
Mr. Fritts, of New York’, Van Nostrand’s Engineering Magazine 32, 392n.
Smith W. (1873), ‘The action of light on selenium’, J. Soc. Telegraph Engineers 2, 31–33.
Söderström K., Bugnon G., Biron R., Pahud C., Meillaud F., Haug F.-J. and Ballif C. (2012),
‘Thin-film silicon triple-junction solar cell with 12.5% stable efficiency on innovative
flat light-scattering substrate’, J. Appl. Phys. 112, 114503–114504.
Solar Frontier (2013), ‘Solar Frontier achieves record 19.7% CIS efficiency’, www.solar-
frontier.com, 8 January 2013, retrieved 11 January 2013.
Sosnowski L., Starkiewicz J. and Simpson O. (1947), ‘Lead sulfide photoconductive cells’,
Nature 159, 818–819.
Spanggaard H. and Krebs F. C. (2004), ‘A brief history of the development of organic and
polymeric photovoltaics’, Solar Energy Mater. Solar Cells 83, 125–146.
Spear W. E. and Le Comber P. G. (1975), ‘Substitutional doping of amorphous silicon’,
Solid State Commun. 17, 1193–1196.
Stranks S. D., Peron G. E., Grancini G., Menelaou C., Alcocer M. J. P., Leijtens T.,
Herz L. M., Petrozza A. and Snaith H. J.(2013), ‘Electron-hole diffusion lengths
exceeding 1 micrometer in an organometal trihalide perovskite absorber’, Science 342,
341–344.
Takamoto T., Ikeda E., Kurita H. and Ohmori M. (1997), ‘Over 30% efficient InGaP/GaAs
tandem solar cells’, Appl. Phys. Lett. 70, 381–383.
Tang C. W. and Albrecht A. C. (1975a), ‘Photovoltaic effects of metal–chlorophyll-a–metal
sandwich cells’, J. Chem. Phys. 62, 2139–2149.
Tang C. W. and Albrecht A. C. (1975b), ‘Transient photovoltaic effects in metal–chlorophyll-
a–metal sandwich cells’, J. Chem. Phys. 63, 953–961.
Treble F. (1998), ‘Milestones in the development of crystalline silicon solar cells’, Renew-
able Energy 15, 473–478.
Wagner S., Shay J. L., Migliorato P. and Kasper H. M. (1974), ‘CuInSe2 /CdS heterojunction
photovoltaic detectors’, Appl. Phys. Lett. 25, 434–435.
Welker H. (1954), ‘Semiconducting intermetallic compounds’, Physica 20, 893–909.
Winther C. (1928), ‘Über den Becquereleffekt. I.’, Zeits. Physik. Chem. 131, 205– 213.
Wolf M. (1960), ‘Limitations and possibilities for improvement of photovoltaic energy
converters. Part I: considerations for Earth’s surface operation’, Proc. Inst. Radio Eng.
48, 1246–1263.
Woodall J. M. and Hovel H. J. (1977), ‘Isothermal etchback-regrowth method for high-
efficiency Ga1−x /Alx As–GaAs solar cells’, Appl. Phys. Lett. 30, 492–493.
Yamaguchi K., Nakayama N., Matsumoto H. and Ikegami S. (1977), ‘Cadmium sulphide–
cadmium telluride solar cell prepared by vapor phase epitaxy’, Jpn. J. Appl. Phys. 16,
1203–1211.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch01 page 39

The Past and Present 39

Yamaguchi M., Amano C., Sugiura H. and Yamamoto A. (1987), ‘High efficiency
AlGaAs/GaAs tandem solar cells’, Proc. 19th IEEE Photovoltaic Specialists Conf.,
IEEE, New York, pp. 1484–1485.
Yamaguchi M., Okuda T., Taylor S. J., Takamoto T., Ikeda E. and Kurita H. (1997), ‘Superior
radiation-resistant properties of InGaP/GaAs tandem solar cells’, Appl. Phys. Lett. 70,
1566–1568.
Yamaguchi M., Takamoto T. and Araki K. (2008), ‘Present and future of super high efficiency
multi-junction solar cells’, Proc. SPIE 6889, Physics and Simulation of Optoelectronic
Devices XVI, 688906.
Yang G. (2012), ‘Should China bail out its solar PV industry?’, www.chinadialogue.net, 12
September 2012, retrieved 4 January 2013.
Yella A., Lee H.-W., Tsao H. N., Yi C., Chandiran A. K., Nazeeruddin M. K., Diau E. W.-G.,
Yeh C.-Y., Zakeeruddin S. M. and Grätzel M., ‘Porphyrin-sensitized solar cells with
cobalt (II/III)–based redox electrolyte exceed 12% efficiency’, Science 334, 629–634.
Yu G., Pakbaz K. and Heeger A. J. (1994), ‘Semiconducting polymer diodes: large size, low
cost photodetectors with excellent visible-ultraviolet sensitivity’, Appl. Phys. Lett. 64,
3422–3424.
Zhao J., Wang A., Green M. A. and Ferrazza F. (1998), ‘Novel 19.8% efficient “honeycomb”
textured multicrystalline and 24.4% monocrystalline silicon solar cells’, Appl. Phys.
Lett. 73, 1991–1993.
ZSW (2013), ‘ZSW produces world record thin-film solar PV cell, achieves 20.8% efficiency
and overtakes multi-crystalline silicon technology’, www.solarserver.com 24 October
2013, retrieved 30 November 2013.
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 41

CHAPTER 2

LIMITS TO PHOTOVOLTAIC ENERGY

CONVERSION EFFICIENCY
MARTIN A. GREEN
Australian Centre for Advanced Photovoltaics
School of Photovoltaic and Renewable Energy Engineering
University of New South Wales
Sydney, N.S.W. Australia, 2052
[email protected]

There can be economy only where there is efficiency.

Benjamin Disraeli, 1868.

2.1 Introduction
Although the sunlight conversion efficiencies of most photovoltaic systems
installed to date have been quite modest, generally in the 10–15% range, there
is no fundamental reason why these cannot be very much higher. Historically, each
successful photovoltaic technology has evolved to ever-increasing energy conver-
sion efficiency, while simultaneously reducing cost. It seems reasonable to expect
this trend to continue as the industry grows, with conversion efficiency increasingly
becoming a key differentiator between technologies, as already apparent over the
last decade with the thin-film technologies. This makes it relevant to understand
fundamental constraints on efficiency and how these may be circumvented.
The publishing of the quantum mechanical theory of semiconductors (Wilson,
1931) coincided with a surge of interest in cuprous oxide solar cells (Grondahl,
1933), stimulating rapid progress in theoretical understanding of the photovoltaic
effect. The serendipitous discovery of the silicon p-n junction at Bell Laboratories
in the early 1940s due to its large photovoltages (Ohl, 1941; Riordan and Hoddeson,
1997) led to the development of p-n junction theory (Shockley, 1949) and its
extension to illuminated junctions (Cummerow, 1954a). The reporting of greatly
improved silicon cells in 1954 (Chapin et al., 1954) was followed by an estimation
of the limiting efficiency (Cummerow, 1954b) and investigations of the optimum
bandgap (Rittner, 1954; Trivich and Flinn, 1955; Loferski, 1956).

41
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 42

42 M. A. Green

Apart from the Trivich–Flinn analysis, subsequently reviewed, these calcula-

tions were empirical in that Shockley’s p-n junction theory was combined with
assessments of ultimately achievable values of the material parameters appear-
ing in the p-n junction equation. A significant step forward was made in 1961
(Shockley and Queisser, 1961) when a materials-independent theory of limiting
cell performance was developed. This was based on the recognition that recombi-
nation in devices of limiting performance would be entirely radiative, the inverse
of the optical absorption processes creating the cell photocurrent. Provided other
intrinsic effects do not prevent this radiative limit from being reached, all cells
of the same bandgap have the same limiting efficiency, with the perfection of the
associated cell and materials technology acting as a practical differentiator.
Even prior to this work, it was recognised that cell performance could
be improved by using multiple cells of different bandgap (Jackson, 1955).
The Shockley–Queisser approach was extended to these multiple cell systems
(Henry, 1980; Marti and Araujo, 1996), then successively to thermophotovoltaics
(Würfel and Wolfgang, 1980), hot-carrier cells (Ross and Nozik, 1982; Würfel,
1995; Würfel et al., 2005), multiple electron−hole pair (and multiple-exciton)
approaches (Werner et al., 1995), intermediate-band (Luque and Marti, 1997)
and impurity level (Brown and Green, 2002) devices, and then to up- and down-
converters (Trupke et al., 2002a; 2002b). Such ‘bottom-up’ analyses allow material-
independent ranking of these different approaches to circumventing the limits upon
conventional cells (Green, 2003) such as shown in Fig. 2.1. The most important of

100%

thermodynamic limit
74% tandem (n )
68% hot carrier
tandem (n = 6)
58% thermal, thermoPV, thermionics
54% tandem (n = 3)
49%
44% impurity PV & band, up-converters
impact ionisation
39% tandem (n = 2)
31% down-converters
single cell

Sun
0%

Figure 2.1 Limiting efficiency of various photovoltaic conversion options under global
sunlight (6000 K Sun and 300 K ambient assumed; n = number of junctions). Source: Green
and Ho (2011).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 43

Limits to Photovoltaic Energy Conversion Efficiency 43

such approaches remain the single junction and the monolithic stack of multiple-
junction tandem devices that provide the main focus of this chapter and of the
present volume.
Complementing these ‘bottom-up’ limiting efficiency calculations have been
‘top-down’ thermodynamic analyses that are both device- and system-agnostic,
based purely on the energy and entropy fluxes associated with the incident sunlight
(Landsberg and Tonge, 1980). The corresponding limit for the ‘top-down’ approach
is also shown in Fig. 2.1 and lies well above the highest efficiency from the ‘bottom-
up’ approaches. This was once thought to be because the ‘top-down’ limit was
unattainable, even in principle, since the analysis failed to include unavoidable
losses (Pauwels and De Vos, 1981; Marti and Araujo, 1996). However, from the
work of Ries (1983), it becomes clear that the difference arises from the implicit
assumption of time-symmetry in the ‘bottom-up’ approaches. Photovoltaic systems
that are capable of the thermodynamically limiting performance are possible, in
principle, although the only suggested implementation, subsequently discussed, is
very complex.
Resonating with the remark by Disraeli opening this chapter, the author has
elsewhere emphasised the likely long-term evolution of photovoltaics to high-
est possible energy conversion efficiencies, as a path to lowest possible costs
(Green, 2003). Figure 2.2a effectively demonstrates the enormous leverage of high
efficiency in reducing the cost/watt of solar modules. At the system level, when
area-related installation costs are included, the leverage increases further, as illus-
trated in Fig. 2.2b. This highlights the relevance of a clear understanding of both

US$0.10/W US$0.20/W US$0.50/W US$0.10/W US$0.20/W

100 100
Thermodynamic
80 limit 80
Efficiency,%

Efficiency,%

60 60 US$1.00/W
US$1.00/W
40 III 40 I II
Present limit

20 20 US$3.50/W
I US$3.50/W I
II II
0 100 200 300 400 500 200 100 0 100 200 300 400 500
Module Cost, US$/m2 System Cost, Module Cost,
US$/m2 US$/m2

(a) (b)
Figure 2.2 (a) Efficiency-cost trade-off for three generations of photovoltaic technology,
based on (I) silicon wafers, (II) ‘second-generation’ thin-films and (III) advanced, high-
efficiency, thin-film ‘third generation’ approaches; (b) Additional leverage provided by
high efficiency when additional area-related costs are included. Efficiency limits are shown
as bands to accommodate the increased efficiencies possible if sunlight is concentrated (or
the acceptance angle of the cell reduced). Source: Green and Ho (2011).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 44

44 M. A. Green

the efficiency limits that apply to the various photovoltaic technologies and also of
the improvements required to progress towards them.

2.2 Photovoltaic converters: essential requirements

The development of respectably efficient dye-sensitised and organic solar cells has
broadened the conception of what constitutes a photovoltaic converter. No longer
are solar cells limited to the heteroface, heterojunction and homojunction p-n and
p-i -n junction devices that previously dominated the field.
Key general features required for a solar photovoltaic converter are a set of
ground and excited states, with electrons photoexcited from the ground state able
to be replenished from one contact and those in excited states able to be extracted
by a second contact. As a practical matter, there needs to be selectivity in this
contacting process so that the contact to the ground state contact does not contact
excited states to any significant extent and vice versa (Fig. 2.3). With such contacts,
current can be generated by the device on short-circuit and voltage on open-circuit.
However, for an efficient converter, another feature is required. The voltage
able to be generated by a converter depends on its degree of excitation as measured
by the different chemical potentials of electrons in the ground and excited states.
However, electrostatics additionally dictate that the electric potential across the
device must change by a value equal to the voltage between the contacts. Moreover,
this potential change is in a direction that opposes the collection of photogenerated

excited states

(- ) (+)

ground states

Figure 2.3 Essential requirements for photovoltaics, selectively contacted ground and
excited states. Source: Green and Hansen (2002).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 45

Limits to Photovoltaic Energy Conversion Efficiency 45

carriers. What is also needed in an efficient solar cell is a way of taking up this
voltage change without affecting carrier collection.
For p-n junction devices, this mechanism is provided by the ‘built-in’ voltage
at the junction, which has a value at thermal equilibrium that depends on the product
of the doping concentration on either side of the junction. This junction potential
drop reduces under light exposure, providing the electrostatic adjustment needed
to accommodate the generation of photovoltage. If an attempt is made to exceed
the built-in voltage, the consequences depend on the conditions at the contacts. If
the doping is higher here than near the junction region, as in the mainstream silicon
wafer commercial product, additional adjustment can occur such as at the potential
step associated with the rear back surface field (BSF). The doping of a p-n junction
also provides the selective contacting requirement, with contact made selectively
to the conduction band on the n-type side of the device and to the valence band on
the p-type side of the device. The selectivity is provided by the different numbers
of conduction-band electrons and valence-band holes at the two contacts (Fig. 2.4).
For thin p-i -n structures, such as amorphous silicon devices, a similar built-
in voltage is present, corresponding to a nearly uniform field across the i -region,
with this field aiding carrier collection. This field’s reduction on illumination, as
photovoltage increases, reduces the effectiveness of carrier collection, resulting in
reduced fill factors. Doping again provides the contacting selectivity.
For dye-sensitised cells, a barrier layer at the TiO2 /front contact interface
provides a built-in potential reservoir, with an insufficient value of this potential
producing low fill factors (Kron et al., 2003; Snaith and Grätzel, 2006). Contact
to the excited states is provided by the n-type TiO2 , while contact to the ground

Transition
region
p-type n-type

Figure 2.4 Essentials of p-n junction solar cell operation. Doping provides selective contact
to the appropriate band by controlling occupancy in the contact region. Source: Green and
Hansen (2002).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 46

46 M. A. Green

states in this case is provided by a completely different process, charge transport

across an ionic liquid.
For both bilayer and bulk heterojunction organic solar cells, the work function
difference between the contacts provides the voltage reservoir, giving rise to a
built-in field at thermal equilibrium (Blom et al., 2007), similar to that in a p-i -n
junction device, with similar consequences for device operation. Photogenerated
carrier collection reduces as the field reduces under operating conditions, giving a
strong dependence of open-circuit voltage, short-circuit current and fill factor on
the original built-in field and hence the contact work function difference (Ramsdale
et al., 2002; Mihailetchi et al., 2004). Contact selectivity, particularly important
for bulk heterojunction devices since both donor and acceptor materials extend to
both contacts, is achieved by better alignment of the appropriate contact Fermi
level to either the ground or excited states, increasing the occupancy of states in
the vicinity for the desired carrier. A ‘hole-transporting’ PEDOT:PSS layer, and
often a LiF or TiO2 layer at the excited-state contacts (Lee et al., 2008) assist in
enforcing this selectivity.

2.3 Thermodynamic properties of sunlight

Black bodies feature predominantly in the theory of the limiting efficiency of solar
cells. Not only does emission from the Sun approximate that from a black-body
at 6000 K but an ideal solar cell, as a perfect absorber of sunlight at the energies
of interest, has properties that are related to those of a black body at the cell
temperature (around 300 K).
Figure 2.5a compares the actual spectrum of sunlight as measured outside the
Earth’s atmosphere (Air Mass Zero or AMO radiation) with standard terrestrial
reference spectra used in solar cell measurements (ASTM G170-03 global and
direct normal AM1.5 spectra, hereafter referred to as AM1.5G and AM1.5D).
The spectral energy content of black-body radiation was deduced by Planck
(1913) in work that changed the course of twentieth century physics. The energy
flux in black-body radiation between photon energies E 1 and E 2 is given by

E2
2
E 3d E
Ė(E 1 , E 2 ) = 3 2 (2.1)
h c E1 e kT − 1
E/

where h is Planck’s constant, c is the velocity of light and
is the etendue, a

measure of the areal and angular spread of the radiation, given by
= Aπ sin 2 θ
where A is the area of the emitter or receiver and θ is the half-angle describing the
divergence of the emitted or received beam. For direct sunlight outside the Earth’s
atmosphere,
d = Aπ/46200 (Green, 2003). For emission into a hemisphere,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 47

Limits to Photovoltaic Energy Conversion Efficiency 47

Figure 2.5 (a) Energy spectral density for the presently accepted photovoltaic reference
spectra (entropy spectral density is also shown in the lower section of the plot on an enlarged
scale as dashed lines for the direct and diluted AM0 spectrum); (b) Planckian equivalent
temperature versus wavelength for the three spectra. Source: Green and Ho (2011).

g = Aπ. In the latter case, integrating Eq. (2.1) from 0 to ∞ gives the result σ T 4
per unit area where σ , the Stefan–Boltzmann constant, equals 2π 5 k 4 /(15h 3 c3 ). A
similar expression to Eq. (2.1) applies to the photon flux Ṅ (E 1 , E 2 ) but with E 3
in the numerator replaced by E 2 .
For radiation that differs from the black-body spectrum, the same expres-
sions can still be used by defining a ‘Planckian’ temperature T (E), a function
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 48

48 M. A. Green

of energy, which gives the correct magnitude for the spectral component at each
energy. Figure 2.5b (upper curves) shows the calculated Planckian temperatures
associated with the AM0 and AM1.5 spectra. The variation in this temperature
for the AM0 spectrum is largely due to different absorption strengths in the Sun’s
photosphere for different wavelengths, resulting in the sampling of temperatures
at different depths. For the AM1.5 spectra, there is the additional impact of atmo-
spheric absorption bands, notably the very strong water vapour absorption bands
centred at about 0.95, 1.1, 1.4, 1.9 and 2.7 µm, although the latter is at too long a
wavelength to be particularly significant for solar energy conversion.
The above Planckian effective temperatures have thermodynamic significance.
If the photon occupancy number is defined as

n ph = 1/(e E/ kT − 1) (2.2)

where T is the effective Planckian temperature, the entropy flux corresponding to

the previous energy flux becomes


2
k E2 2
Ṡ(E 1 , E 2 ) = 3 2 E [(1 + n ph ) ln(1 + n ph ) − n ph ln(n ph )]d E (2.3)
h c E1
The entropy flux calculated in this way for the AM0 spectrum is also included
in Fig. 2.5a. If the Planckian temperature is constant for all wavelengths, a property
possessed solely by black-body radiation, integration gives the simple result for
the entropy flux

Ṡ(0, ∞) = (4/3) Ė(0, ∞)/T = (4/3)σ T 3

The effective temperatures shown uppermost in Fig. 2.5b are appropriate to

systems that respond only to sunlight coming directly from the Sun’s disc. The ther-
modynamic limits calculated using these temperatures apply only to Sun-tracking,
direct-light converters. Most photovoltaic devices to date respond to a much wider
range of angles and convert not only direct sunlight but light from a range of angles
incident on their surface (Fig. 2.6).
In such cases, where the converter takes no advantage of the directionality
of sunlight, the thermodynamic limit on efficiency will be the same whether the
sunlight is considered directional or spread uniformly over the entire hemisphere.
To spread light from the Sun’s disc over the whole hemisphere and maintain the
same intensity on the cell, the intensity of sunlight would have to be reduced at each
photon energy by a factor of 46,200 (equal to the square of the ratio of the Sun–Earth
distance to the Sun’s radius). Sunlight is effectively diluted by a dilution factor ξ
of 1/46,200 at each energy, giving the smaller effective Planckian temperatures
shown by the lower curves in Fig. 2.5b (note that ξ =
d /A, a relationship used
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 49

Limits to Photovoltaic Energy Conversion Efficiency 49

Ts

Tsky

dA

x
y

Figure 2.6 Illustration of radiation environment seen by an element on the cell’s surface.
As well as the direct sunlight at temperature TS , the cell is also exposed to sky radiation at a
temperature Tsky assumed to equal the ambient temperature T A . Source: Green and Hansen
(2002).

extensively in later sections). Using these modified Planckian temperatures, the

new etendue,
g , would equal Aπ.
An important question when dealing with devices responding to light from
all directions is how to treat light from directions in the sky other than those
occupied by the Sun (Fig. 2.6). The most convenient treatment is to assume a
sky temperature equal to the ambient temperature, with dilution factor (1 − ξ ).
Including the corresponding background thermal fluxes from the sky ensures the
Planckian effective temperature of the incident radiation remains above the ambient
temperature, even at long wavelengths where this would not otherwise be the case
(Fig. 2.5b).

2.4 ‘Top-down’ thermodynamic efficiency limits

If the Sun is regarded as black-body source at 6000 K and the Earth as a sink at
300 K, the Carnot efficiency on converting sunlight into electricity (or other useful
work) is 95%, as given by (1−Tsink /Tsource ). However, this limit applies only when
an infinitesimal amount of work is done by the converter, with most of the incident
energy recycled back to the Sun. In principle, this Carnot efficiency can be attained
by a photovoltaic system (an infinite tandem cell stack under maximal possible
sunlight concentration operating close to open circuit), but the infinitesimally small
power outputs involved are of no practical interest.
The problem with the Carnot calculation in this context is that, in reality, there
is no rebate for energy recycled back to the Sun. This recycled energy should
therefore be treated as a loss with attention restricted to the more confined system
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 50

50 M. A. Green

TA
.
E S .
W
.
SS
Tc
.
.
Q
. . .
EC S = Q/T A

.
S C

.S >0
G

Figure 2.7 System for calculating the Landsberg efficiency limit. Source: Green and Hansen
(2002).

of Fig. 2.7. An energy and entropy balance gives (Green, 2003)

η = Ẇ / Ė S = (1 − TA ṠS / Ė S ) − (1 − TA ṠC / Ė C ) Ė C / Ė S TA ṠG / Ė S (2.4)

where η is the energy conversion efficiency, Ẇ is the useful work (electricity) and
the Ṡ and Ė terms correspond to the associated entropy and energy fluxes. Noting
TC ≥ T A , and assuming the cell is a perfect absorber over all wavelengths emitted
by the Sun (both black bodies), the maximum value of this efficiency occurs when
ṠG = 0 and TC = T A , giving

4 TA 1 T A4
ηL = 1 − + (2.5)
3 TS 3 TS4

This limit is known as the Landsberg limit (Landsberg and Tonge, 1980).
For a 6000 K Sun and a 300 K ambient, the resulting efficiency limit is 93.3%.
While it is generally accepted that this represents an upper bound on the efficiency
of solar energy conversion, some have thought this bound cannot be reached in
principle, as previously noted, due to unavoidable entropy generation during light
absorption (making the ṠG = 0 condition fundamentally unattainable). However,
the arguments involved apply only to time-symmetric systems. It has been shown
that, in principle, the Landsberg limit can be approached arbitrarily closely by
systems that are time-asymmetric (Ries, 1983).
A subtlety is that this limit applies only to systems involving a direct exchange
with the Sun. This can be achieved by concentrating optics if sunlight is concen-
trated to the maximum extent possible (46,200 times). During such concentration,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 51

Limits to Photovoltaic Energy Conversion Efficiency 51

etendue ideally is conserved. The consequence is that, under maximal concentra-

tion, the solar intensity at the Sun’s surface is reproduced terrestrially, but over an
area 46,200 times smaller than the system aperture. Such high concentration levels
have been approached experimentally (Cooke et al., 1990). An alternative, but to
date more difficult approach giving the same result, in principle, is to restrict the
angular absorption, and hence emission, of the cell to the small range of angles
required to accept only direct sunlight (same etendue
d for incoming and emitted
radiation).
Efficiencies calculated under the above conditions are generally referred to as
efficiencies calculated under ‘maximal concentration’ although, as noted above,
light need not be concentrated, at least in principle, to attain them.
As already mentioned, most photovoltaic panels have an approximately
isotropic response, responding reasonably well to light regardless of the angle
of incidence. This gives the very practical advantages of being able to respond to
diffuse sunlight scattered by the atmosphere and also of not having to track the Sun
as it moves across the sky.
An appropriate limit in this case, where the converter makes no attempt to
exploit the directionality of sunlight, is the ‘isotropic limit’ that would apply to a
converter able to respond equally well to light from any part of the sky. As before,
the Sun could be considered as completely filling the sky but with its intensity from
each direction reduced to give the same overall flux on the device (the reduction
would be the same for all wavelengths, such as obtained by passing through a
neutral-density filter). The thermodynamics of such ‘diluted black-body radiation’
have been examined elsewhere (Landsberg and Tonge, 1979), where it is shown the
entropy-to-energy flux ratio increases by a factor χ that depends on the extent of the
dilution. Imaging the rest of the sky acting as a black-body source at temperature
TA , the general form of the Landsberg limit, presented in the form below for the
first time, becomes
  4
4 TA 4 T
ηL = 1 − χ + χ − 1 A4 (2.6)
3 TS 3 TS
where an accurate approximate expression for χ is (Green, 2011)

χ = 0.9652 − 0.2777 ln(ξ ) + 0.0348 ξ 0.9 (2.7)

For the present case of dilution by 46,200 times, χ has a value of 3.94, giving
a limiting Landsberg efficiency for isotropic converters of 73.7% for a 6000 K Sun
and a 300 K ambient, the value shown in Fig. 2.1.
Figure 2.8 shows a photovoltaic system capable in principle of reaching such
limits (Green, 2003). Each converter in the system consists of a stack of a large
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 52

52 M. A. Green

Figure 2.8 Time-asymmetric photovoltaic system using an infinite number of circulators

in combination with infinite stacks of tandem cells. Source: Green and Hansen (2002).

number of cells of different bandgap. The first converter accepts the full sunlight
intensity from the Sun via a circulator, which allows the light emitted by the cell
to be steered to the next cell. In this way, each cell can be operated infinitesimally
close to open-circuit, where it converts all wavelengths at the Carnot efficiency,
with essentially only ambient temperature black-body radiation eventually sent
back to the Sun.
Clearly, a much simplified way of implementing time-asymmetric effects will
be required if they are to be used to practical advantage. An encouraging result
is that, if only one circulator and two cell stacks are used, the maximal efficiency
increases from 86.8% to 90.7%, over half the possible benefit from time asymmetry
(Brown, 2003).

2.5 Single-cell efficiency limits

2.5.1 Trivich–Flinn limit
The fundamental trade-off associated with photovoltaic energy conversion is shown
in Fig. 2.9. Photons of energy E = hν less than the bandgap E g of the photovoltaic
converter are not absorbed while photons of energy above the bandgap create an
electron-hole pair that quickly loses any energy in excess of the bandgap (h is
Planck’s constant as before; ν the light frequency).
One of the earliest treatments of the bandgap dependence of photovoltaic
conversion efficiency analysed, very literally, the situation of Fig. 2.9 (Trivich
and Flinn, 1955). Regarding the incident sunlight as from a black-body source at
temperature TS , using the black-body photon and energy fluxes deduced by Planck
and assuming the final outcome of the absorption process was an electron-hole pair
separated by the bandgap energy, the Trivich–Flinn (TF) efficiency is given by
∞
E g Ṅ (E g , ∞) E g E g E 2 /[exp(E/kTs ) − 1] d E
ηTF = = ∞ 3 (2.8)
Ė(0, ∞) 0 E /[exp(E/kTs ) − 1] d E
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 53

Limits to Photovoltaic Energy Conversion Efficiency 53

hf >> Eg hf < E g
Eg

Figure 2.9 Photoexcitation across a semiconductor bandgap showing the fundamental

trade-off in photovoltaic conversion. Source: Green and Ho (2011).

The Bose–Einstein integrals involved can be evaluated by noting, with ε = E/kTS

∞
∞ ∞

 ∞
εn dε n −ε −ε −2ε
= ε e (1 + e +e + · · · )dε = εn e−mε dε
εg eε − 1 εg m=1 εg
(2.9)

This gives, for n = 2

∞ ∞

ε2 dε
= e−mεg (m 2 εg2 + 2εg + 2)/m 3
εg eε − 1
m=1

= εg2 Li 1 (e−εg ) + 2εg Li 2 (e−εg ) + 2Li 3 (e−εg ) (2.10)

where Lis (z) is the polylogarithm of order s, defined as (Lewin, 1981)

∞ k
 z
Li s (z) = (2.11)
ks
k=1

Importantly for efficiency calculations (Green, 2012), polylogarithms have the

general property
∂ Li s (z) Li s−1 (z) ∂ Li s (eµ )
= giving = Li s−1 (eµ ) (2.12)
∂z z ∂µ
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 54

54 M. A. Green

where µ is any variable. For integral s ≤ 1, polylogarithms can be expressed in

terms of elementary functions. Of particular relevance to the present chapter
z z
Li 1 (z) = − ln(1 − z); Li 0 (z) = ; Li −1 (z) = (2.13)
1−z (1 − z)2
For higher orders of s, the power series involved in Eq. (2.11) are quite rapidly
convergent for z < 1, the case of present interest. Note also for positive arguments
in this range
z ≤ Li 3 (z) ≤ Li 2 (z) ≤ Li 1 (z) ≤ Li 0 (z) ≤ Li −1 (z) = z/(1 − z)2 (2.14)

Hence, all orders will have roughly similar values for small z. Repeating the
previous integration for n = 3 gives

∞ 3
ε dε
ε
= εg3 Li 1 (e−εg ) + 3εg2 Li 2 (e−εg ) + 6εg Li 3 (e−εg ) + 6Li 4 (e−εg )
εg e − 1
(2.15)
Hence, the Trivich–Flinn limit expressed in polylogarithms is
εg [εg2 Li 1 (e−εg ) + 2εg Li 2 (e−εg ) + 2Li 3 (e−εg )]
ηTF = (2.16)
6Li 4 (1)
where Li4 (1) is an Euler series with value π 4 /90. The value of εg giving the max-
imum TF efficiency (εg0 ) can be found by differentiating the above numerator to
give
εg0
3
Li 0 (e−εg0 ) = εg0
2
Li 1 (e−εg0 ) + 2εg0 Li 2 (e−εg0 ) + 2Li 3 (e−εg0 ) (2.17)
This equation has the solution ε g0 = 2.1657 . . . normalised to kT S with the
corresponding maximum value of the TF efficiency of 43.88%, regardless of TS
(the assumed black-body temperature of the Sun). In terms of εg0 , it can be seen
by comparing Eqs (2.16) and (2.17) that this maximum efficiency is given by

15 εg0
4
ηTF
m
= (2.18)
π 4 eεg0 − 1
For a 6000 K black-body source, the optimum bandgap would be 1.120 eV,
about the bandgap of silicon.

2.5.2 Shockley–Queisser limit

For any actual photovoltaic converter, the voltage output will be less than the
‘bandgap voltage’ E g /q (where q is the electronic charge). An elegant approach
to finding a fundamental limit on the voltage output of a photovoltaic cell was
published by Shockley and Queisser (1961). Shockley and Queisser (SQ) realised
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 55

Limits to Photovoltaic Energy Conversion Efficiency 55

that a strongly absorbing cell in thermal equilibrium with its ambient at temper-
ature T A would be emitting black-body radiation. For photon energies above the
cell bandgap, the physical source of the emitted photons would be band-to-band
radiative recombination, the inverse of the absorption process of Fig. 2.9.
In a non-equilibrium situation with voltage across the device, SQ were aware
that, in a high-quality p-n junction device, the electron–hole product, and hence
these radiative recombination rates, would be exponentially enhanced throughout
the cell volume by the cell voltage, V , normalised to the ‘thermal voltage’, kTA /q.
Hence, a good-quality cell, for photon energies above its bandgap, would emit
exponentially enhanced black-body radiation. Assuming 100% quantum efficiency,
one electron would flow into the cell terminals for each emitted photon.
Using this insight, but a more general chemical potential model of photon
emission (Würfel, 1982), the electrical power density output from a photovoltaic
cell in the SQ limit can be expressed as (Green, 2003)
PSQ = V J



2πq V ∞ E 2d E ∞ E 2d E
= 3 2 ξ −
h c Eg exp(E/kTS ) − 1 Eg exp[(E − q V )/kTA ] − 1
(2.19)

where ξ is the black-body radiation ‘dilution’ factor. As before, ξ is unity if the

whole range of acceptance angles for the cell is filled by direct sunlight, as for a
cell operating under the maximal sunlight concentration ratio of 46,200. ξ falls to a
value of 1/46,200 if the available sunlight is considered spread over the whole ter-
restrial hemisphere, an appropriate model for a converter with an isotropic response
over this hemisphere.
The first integral can be expressed in terms of polylogarithms as in Sec-
tion 2.5.1. The second integral can be expanded as in Eqs (2.9)–(2.10) to give

∞
E 2d E
E g exp[(E − q V )/kTA ] − 1

= (kTA )E g2 Li 1 [e(q V −E g )/ kTA ] + 2(kTA )2 E g Li 2 [e(q V −E g )/ kTA ]

+ 2(kTA )3 Li 3 [e(q V −E g )/ kTA ] (2.20)

This gives the SQ efficiency limit as
15
ηSQ = [εg + t ln(γ )]{εg2 [Li 1 (e−εg ) − t Li 1 (γ )/ξ ]
π4
+ 2εg [Li 2 (e−εg ) − t 2 Li 2 (γ )/ξ ]
+ 2[Li 3 (e−εg ) − t 3 Li 3 (γ )/ξ ]} (2.21)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 56

56 M. A. Green

where γ = e(q V −E g )/ kTA and t = TA /TS . Comparison with Eq. (2.16) shows the
TF formulation is a limiting case of the SQ formulation for TA = 0. This close
relationship will be exploited in the following.
Unlike Eq. (2.16), Eq. (2.21) is a function of not one but three variables εg ,
V and t. Since these variables can be expressed in terms of standard functions
supported by common mathematical packages such as Mathematica (2008), even
in this form Eq. (2.21) could be simply solved, for example, for its global maximum
value at fixed t (built-in Mathematica programs such as FindMaximum require
only the entry of the equation, plus essentially a single line of code, to find the
solution).
The technique generally used to find the peak SQ efficiency is to fix E g , find
the optimum V for this E g , and hence steadily map out efficiency as a function of
E g (Shockley and Queisser, 1961). However, partial differentiation of Eq. (2.19)
or (2.21) with respect to E g locates a stationary point when

Li 0 [e(q V −E g )/ kT A ] = ξ Li 0 e(−E g / kTS ) (2.22)

For the case of undiluted sunlight (ξ = 1), this simplifies to the Carnot rela-
tionship noted elsewhere (Landsberg, 2000)

(E g − q Vm )/kTA = E g /kTS or Vm = E g (1 − TA /TS )/q (2.23)

where Vm is the optimum voltage for maximum power output.

This Carnot condition must prevail at the global optimum E g for the SQ
efficiency (for undiluted sunlight), but the corresponding value of V will in general
be non-optimal for other E g . Hence, imposing this condition on voltage for all
E g will result in a lower bound on available power output that will nonetheless
equal the SQ output at stationary points, including the global maximum.
Inserting Eq. (2.23) into Eq. (2.21) gives the following expression for this
lower bound (for ξ = 1)

ξ =1 15
ηSQLB = εg (1 − t)2 [εg2 Li 1 (e−εg )
π4
+ 2(1 + t)εg Li 2 (e−εg ) + 2(1 + t + t 2 )Li 3 (e−εg )] (2.24)

Differentiating Eq. (2.24) gives an analytical expression for the optimum

bandgap (alternatively, Eq. (2.21) could be partially differentiated with respect
to V with the value given by Eq. (2.23) then inserted, to give the same results). The
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 57

Limits to Photovoltaic Energy Conversion Efficiency 57

optimum bandgap is given (for ξ = 1) by

εgm
3
Li 0 (e−εgm ) = (1 − 2t)εgm
2
Li (e−εgm ) + 2(1 + t − t 2 )εgm Li 2 (e−εgm )

+ 2(1 + t + t 2 )Li 3 (e−εgm ) (2.25)

where since εgm is in normalised units, the actual bandgap E gm = εgm kTS .
Noting the TF formulation is a limiting case of the SQ formulation for TA = 0
and that small TA /TS ratios (∼0.05) prevail for solar conversion, the SQ results
can be expressed linearly in terms of TF results. The exact value of the linear
term in a power series expansion can be found by solving Eq. (2.21) by a single
Newton–Raphson iteration, using εg0 as the trial solution, giving
ξ =1
εgm (SQ) = εg0 − t[3Li 1 (e−εg0 ) − εg0 Li 0 (e−εg0 )]/
[4Li 0 (e−εg0 ) − εg0 Li −1 (e−εg0 )]
= 2.1657 − 0.422 TA /TS (2.26)

Similarly, expanding Eq. (2.24) to first order in t gives

ξ =1
ηm (SQ) = ηTF {1 − t[1 + Li 1 (e−εg0 )/Li 0 (e−εg0 )/εg0 ]}
= 0.4388 − 0.629TA /TS (2.27)

Both expressions are extremely accurate, giving the correct values of bandgap
and efficiency for t = 0.05 to within one digit in the fifth and fourth significant
figure, respectively.
For the case of diluted sunlight, of considerable practical importance since
it applies to the general photovoltaic case of non-maximal sunlight concentration
and also, as discussed below, to cases where non-radiative recombination occurs
in parallel with the modelled radiative recombination, Eq. (2.19) again prevails at
the optimum V . At this optimum, γ depends only on εg and ξ with a value γ ∗
equal to or slightly higher than ξ e−εg as defined below
ξ
γ∗ = (2.28)
eε g −1+ξ
Proceeding as before, this gives the lower bound on SQ efficiency as
15
ηSQLB = [εg + t ln(γ ∗ )]{εg2 [Li 1 (e−εg ) − t Li 1 (γ ∗ )/ξ ]
π4
+ 2εg [Li 2 (e−εg ) − t 2 Li 2 (γ ∗ )/ξ ]
+ 2[Li 3 (e−εg ) − t 3 Li 3 (γ ∗ )/ξ ]} (2.29)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 58

58 M. A. Green

Differentiating Eq. (2.29) again provides the relationship for the optimum
bandgap, εgm (SQ) normalised to kT S . Maximum efficiency is obtained when the
bandgap satisfies the following relationship

[εg + t ln(γ ∗ )][εgm

2
Li 0 (e−εgm ) + 2tεgm Li 1 (γ ∗ )/ξ + 2t 2 Li 2 (γ ∗ )/ξ ]
= εgm
2
[Li 1 (e−εgm ) − t Li 1 (γ ∗ )/ξ ] + 2εgm [Li 2 (e−εgm ) − t 2 Li 2 (γ ∗ )/ξ ]
+ 2[Li 3 (e−εgm ) − t 3 Li 3 (γ ∗ )/ξ ] (2.30)

Following the same procedure as for ξ = 1 and expressing the polylogarithms

involved in terms of elementary functions, the optimum bandgap is given by
ln(1 + A/ξ ) − (3 A/ξ ) ln(1 + ξ/A)
εgm = εg0 + t + O(t 2 ) (2.31)
4 − εg0 (1 + 1/A)
where A = (eεg0 − 1) = 7.7221 with εg0 = 2.1657, as previously evaluated and
O(t 2 ) indicates small terms of the order of t 2 . As ξ decreases from unity, the first
term in the numerator increases rapidly while the second term increases much more
slowly, giving a sign change at ξ = 0.44. For the most challenging case of small
ξ (=1/46,200), the accuracy of this expression based on only terms linear in t is
about 1% at t = 0.05. However, since ηSQ is relatively insensitive to εg near the
maximum, inserting this result for εgm into Eq. (2.29) gives the correct maximum
efficiency to close to four significant digits.
Alternatively, expanding Eq. (2.29) to terms linear in t gives
    

t A A ξ
ηSQm = ηTF 1 − ln 1 + + ln 1 + + O(t 2 ) (2.32)
εg0 ξ ξ A
The accuracy obtained using this expression is not as high as with the previous
approach, with square and cubic terms in t required for similar accuracy, for the
most challenging case of small ξ . However, Eq. (2.32) has the advantage of correctly
identifying the important dependencies. Non-ideal device properties in the form
of non-radiative recombination occurring in parallel with radiative recombination
can be modelled by further dilution of sunlight by a factor equal to the radiative
efficiency (the fraction of nett recombination in the device that is radiative).

2.5.3 Efficiencies under actual AM0 and AM1.5 spectra

The above treatments in terms of a black-body Sun allow analytical solutions
showing the important dependencies. However, limits applying to the tabulated
AM0 and AM1.5 reference spectra at the standard cell test temperature of 25◦ C
(298.15 K) are probably of more practical interest. Figure 2.10 summarises effi-
ciencies versus bandgap computed for the different spectra in the extended SQ
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 59

Limits to Photovoltaic Energy Conversion Efficiency 59

Figure 2.10 Isotropic performance limits for cells under the AM0, AM1.5G and AM1.5D
spectral as well as for the AM1.5D spectra, under various concentration levels. The Trivich–
Flinn limit for the AM1.5D spectrum is shown for comparison. Source: Green and Ho (2011).

approach. The uppermost curve is the Trivich–Flinn limit for the AM1.5D spec-
trum, which is approached for small-bandgap cells under high concentration levels
(or, more generally, severe angular restriction of the converter response). Note that
the limiting efficiency under concentration (or angular restriction) increases more
rapidly for small-bandgap cells than for large-bandgap cells (simply because the
approximately 60 mV increase in open-circuit voltage for ten times increase in
intensity, or decrease in angular acceptance, is proportionately larger for small-
bandgap cells).
For the AM0 spectrum for a cell with isotropic response, there is a single peak
(30.4% efficiency for E g = 1.245 eV). This is similar to the value of 31.0% at a
bandgap of 1.31 eV calculated for a 6000 K black-body spectrum at a cell temper-
ature of 300 K. For the terrestrial AM1.5 spectra, there are multiple peaks due to
the strong water vapour absorption bands apparent in Fig. 2.5. The most impor-
tant are the absorption bands in the 920–970 nm (1.28–1.35 eV), 1100–1160 nm
(1.07–1.13 eV) and 1300–1500 nm (0.83–0.95 eV) wavelength (energy) ranges.
Loss of radiation in the terrestrial spectra due to the first of these bands elimi-
nates the ‘natural’ peak apparent in the AM0 spectrum, creating a local maximum
on the high-energy side of the absorption band and a local minimum on the low-
energy side.
Note that from Fig. 2.10 the AM0 and AM1.5G spectra give relatively higher
efficiency for larger bandgaps than does the AM1.5D spectrum. This is due to blue
photons being selectively scattered out of the AM1.5D spectrum during passage
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 60

60 M. A. Green

Figure 2.11 Isotropic efficiency limits under the AM1.5G spectrum, and also for devices
with 1% and 0.01% radiative efficiency. Also shown are best-certified cell limits efficiencies
for various cell technologies. Source: Green and Ho (2011).

through the atmosphere. Combined with the intensity dependence already noted,
this shifts the global maximum between three different peaks, each associated with
the high-energy side of the three absorption bands already mentioned.
The solid line of Fig. 2.11 shows the limiting efficiency for a conventional cell
with isotropic response under the AM1.5G reference spectrum. Compared with an
earlier spectrum in use until 2008 (Green et al., 2009), the stronger spectral content
over the 650–900 nm range shifts the peak efficiency to higher energy, so that the
two peaks around 1.15 eV and 1.35 eV are nearly equal, with peak efficiencies of
33.6% and 33.8% respectively. For the previous reference spectrum (ASTM E892-
87, IEC60904-3:1989), the order of the peaks was reversed and both were about
1% lower in efficiency (Tiedje et al., 1984), consistent with expectations discussed
elsewhere (Green et al., 2009).
Also shown in Fig. 2.11 as dashed lines are the limiting efficiencies for 0.01%
and 1% radiative efficiencies of the cell, with radiative efficiency defined as the
fraction of nett recombination in the device that is radiative. Radiative efficiency
is a measure of the state of development of the cell material technology. Limiting
efficiencies calculated with reduced radiative efficiency are identical to those cal-
culated under a reduced intensity (further dilution) of sunlight. Hence the previous
curves under concentration (Fig. 2.10) can also be used to calculate the effects
of non-ideal radiative efficiency in this case as well (for example, the curve for
100 kW m−2 incident intensity represents the limiting performance for a cell under
1000 kW m−2 intensity, if the radiative efficiency is 10% rather than 100%).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 61

Limits to Photovoltaic Energy Conversion Efficiency 61

Given this correspondence, it is not surprising that the efficiency for low-
bandgap cells is most sensitive to low values of the radiative efficiency, since a
60 mV/decade voltage reduction with decreasing radiative efficiency has a propor-
tionately more devastating impact. Peak efficiency is pushed to higher-bandgap
cells as the radiative efficiency decreases.
Also shown in Fig. 2.11 are best-certified experimental values for a range
of different solar cell technologies (Green et al., 2011). ‘Bandgap’ in this case
corresponds to the energy on the low-energy side of the cell’s spectral response
curve, where external quantum efficiency (EQE) reaches 50% of its peak value.
Silicon (c-Si) and c-GaAs lie above the 0.01% radiative efficiency line. Actual
radiative efficiencies are close to or well above 1% for both technologies. The
difference arises since experimental devices have additional losses other than the
radiative loss assumed including, for example, reflection and resistance losses, but
also including additional radiative loss due to non-abrupt absorption thresholds
(Kirchartz et al., 2009).

2.5.4 Radiative efficiency of state-of-the-art photovoltaic cells

The limiting photovoltaic solar cell performance is obtained when cell recombi-
nation is dominated by radiative processes because there are then no non-radiative
pathways, a fact used to advantage in the SQ analysis.
The effect of parallel non-radiative processes can be described in terms of
what will be known as the external radiative efficiency (ERE). This is defined as
the fraction of the total recombination in the device that results in radiative emission
from the device. Since this fraction may vary in general with the voltage across
the device, the open-circuit voltage (Voc ) is chosen as the reference voltage. Due
to total internal reflection and photon recycling, the ERE will generally be smaller
than what will be termed the internal radiative efficiency (IRE), the fraction of
internal recombination events that are radiative, such as is determined by the ratio
of total lifetime to the radiative lifetime. For silicon devices, the IRE can be several
times the ERE (Trupke et al., 2003) but obviously the IRE:ERE ratio must decrease
towards unity as both approach 100%.
The dashed lines in Fig. 2.11 show the limiting efficiencies for ERE between
1% and 0.01%, with the best cells lying between these limits. As noted above and
to be shown below, actual ERE can be appreciably higher, since effects other than
radiative inefficiencies also decrease the experimental efficiency below the ideal
limits.
Experimental ERE can be deduced by taking advantage of a surprising yet fun-
damental reciprocal relationship recently identified (Rau, 2007). Actual solar cells
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 62

62 M. A. Green

do not emit radiation above the cell bandgap exactly as assumed in the Shockley–
Queisser analysis (exponentially enhanced by the cell voltage with the spectral
distribution of room-temperature black-body radiation). Rather, they emit radia-
tion with this distribution multiplied at each wavelength by the external quantum
efficiency (EQE) of the device, as measured for the device operating as a solar cell.
This remarkable relationship has been confirmed for a number of experimental
devices (Kirchartz and Rau, 2007; Kirchartz et al., 2008a; Kirchartz et al., 2009;
Vandewal et al., 2009), as well as being consistent with modelling prior to its
recognition (Green et al., 2001).
The cell EQE must be measured for a calibrated measurement of cell effi-
ciency, to adjust for spectral mismatch between the spectrum used for illuminating
the cell during measurement and the tabulated reference spectrum (International
Standard, 2008). This means that each calibrated cell measurement yields sufficient
information to allow calculation of ERE.
On cell open circuit, the photogeneration of carriers within a cell is bal-
anced by recombination within the device, globally if not at each point within
the device. The open-circuit voltage measured at the cell terminals is the voltage
that increases recombination within the device-active region to the level required
to balance the photocurrent able to be collected by the junction. Assuming reason-
able device quality, so parasitic resistances are not excessive, and also assuming
that the collection probability of carriers is not strongly voltage-dependent, the
photocurrent density collected by the junction on open circuit will equal the short-
circuit current density. The light emitted by the junction will be the EQE-weighted,
exponentially-enhanced black-body radiation described above, allowing the ERE
to be calculated as
 
2πq q Voc  E Q E abs E 2 dE q Voc 
h 3 c2
exp kT exp(E/ kT )−1 exp kT E Q E rel NBB (E)d E
ERE = = 
Isc /A E Q E rel N AM1.5 (E)d E
(2.33)

EQE is the value for the solar cell for near perpendicularly incident light in
either absolute or relative terms, with EQE the Lambertian weighted value over all
angles of incident light, of absolute value EQEabs and relative value EQErel . For
a high-quality cell, this will not differ greatly from the near perpendicular value.
EQE is typically measured either in absolute or relative terms with a midrange
accuracy of about 3% relative. The major contributions to the integration on the
numerator come from the long-wavelength region of the spectral response, from
the region where this response begins to fall rapidly to the low value near the
absorption threshold, while the main contribution to the denominator comes from
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 63

Limits to Photovoltaic Energy Conversion Efficiency 63

the region near the peak photon density in the AM1.5 spectrum, which is in the
600–700 nm range.
Some of the sources of inaccuracies in the evaluation of ERE have already been
suggested. An isotropic response needs to be assumed for evaluation from standard
near-perpendicular EQE data. Collection probabilities may be voltage-dependent
so that the short-circuit condition, where EQE is normally evaluated, may not
represent conditions at open circuit. The reciprocal relation itself is also only strictly
valid under conditions where the quasi-Fermi level separation at the junction is
constant (Rau, 2007), which would not be the case for resistive devices even at
open circuit, due to circulating currents. However, these effects are expected to be
minor for cells of respectable performance, particularly compared with the large
differences in ERE noted between different devices and different technologies.
ERE was calculated for a range of representative state-of-the-art cells (Green
et al., 2010; Green et al., 2011). The absolute EQEs for these cells are shown
in Fig. 2.12a, deduced from the relative EQE by normalising to the experimen-
tal short-circuit current density measured for the devices. The calculated spectral
luminescence from these devices on open-circuit is shown in Fig. 2.12b. Units in
this case are A/m2 /nm, representing the current density required on open circuit to
support the different spectral emission components shown. Due to large differences
between technologies, different scaling factors are applied to the different results,
although the same scaling factor is applied to cells of the same type, except for the
GaAs cells.
Until recently (Green et al., 2011) the most efficient non-concentrating, single-
junction cell was a 26.4% efficient GaAs cell fabricated by the Fraunhofer Institute
for Solar Energy (GaAs ISE), with parameters shown in Table 2.1 Integrating
the calculated spectral luminescence gives an ERE of 1.26%. Given that this device
was fabricated on a GaAs substrate, IRE would have been appreciably higher since
n 2 more radiation (where n is the substrate refractive index) would have been
emitted into the substrate (with most absorbed there) than into air.
In late 2010, this record was surpassed with 27.6% efficiency measured for a
thin-film GaAs device fabricated by Alta Solar (GaAs Alta). As shown in Fig. 2.12a,
this has an almost identical spectral response to cell GaAs ISE but, as in Fig. 2.12b,
the luminescent intensity on open circuit is almost 20 times higher, corresponding
to a greatly improved ERE of 22.5%. This high value suggests photons emitted
during radiative recombination events in a direction towards the rear of the device
are not wasted as in the GaAs ISE device, but that a reasonable number are reflected
back into active device regions.
With radiative recombination forming such a large fraction of total recombina-
tion in this device, further improvements in ERE will result in immediate benefits.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 64

64 M. A. Green

Figure 2.12 (a) EQE for the different cells described in the text; (b) calculated spectral
luminescence from the different cells on open circuit, with widely varying multiplication
factors. Source: Green and Ho (2011).

As ERE is increased towards 100%, an additional gain in Voc of up 40 mV is

predicted with a corresponding fill factor improvement.
The ERE for the two crystalline silicon devices investigated are not appreciably
lower than the second-rated GaAs device, with both Si devices displaying an ERE
of about 0.6%. This is similar to the electroluminescent quantum efficiency directly
measured for similar devices (Green et al., 2001), again confirming the accuracy of
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 65

Limits to Photovoltaic Energy Conversion Efficiency 65

Table 2.1 External radiative efficiency (ERE) and other relevant

performance parameters at 25◦ C for the state-of-the-art devices
(Green et al., 2010; 2011) included in the present study.

Device V oc mV isc mA/cm2 Effic. % ERE%

Si UNSW 706 42.7 25.0 0.57

Si SPWR 721 40.5 24.2 0.56
GaAs Alta 1107 29.6 27.6 22.5
GaAs ISE 1030 29.8 26.4 1.26
CIGS ZSW 740 35.4 20.3 0.19
CIGS NREL 713 34.8 19.6 0.057
CdTe* ASP 838 21.2 12.5 1.0E–4
a-Si OERL 886 16.8 10.1 5.3E–6
Dye* Sony 719 19.4 9.9 7.2E–6
OPV Konarka 816 14.5 8.3 2.7E–7
OPV Solarmer 759 15.9 8.1 3.8E–7

∗ Minimodule: results on a ‘per cell’ basis.

the present approach. Since the Sunpower device (Si SPWR) with less than 200 µm
thickness is appreciably thinner than the University of New South Wales device
(Si UNSW) with over 400 µm thickness and has possibly worse rear reflection,
the EQE is notably poorer at long wavelengths, with i sc also appreciably lower.
This is compensated by a higher Voc , due to the smaller volume of the bulk region,
probably combined with better surface passivation. From Fig. 2.12b and Table 2.1,
the two devices have almost identical open-circuit voltage luminescence.
Comparison of two recent record CIGS (copper indium gallium selenide)
devices also produces interesting results. One is a small area 20.3%-efficient device
fabricated by Zentrum für Sonnenenergie-und-Wasserstoff-Forschung (CIGS
ZSW), and the second is for a larger (1 cm2 ) device of 19.6% efficiency fabri-
cated by the US National Renewable Energy Laboratory (CIGS NREL). From
Fig. 2.12a, the two devices can be seen to have similar spectral responses, although
the CIGS ZSW device has a much stronger emission on open circuit, with ERE
of 0.19% compared to 0.06% for the NREL device. This indicates fundamen-
tally better quality material in the ZSW device, although this may be partly due
to its smaller size, given the strong dependence of CIGS cell performance on
cell area.
The next three devices discussed include recent record CdTe, a-Si and dye-
sensitised devices. The CdTe device is a small 12.5%-efficient submodule with
ERE of 1.0 × 10−4 %. Smaller cells could be expected to have higher ERE while
commercial CdTe modules, averaging about 11.4% aperture area efficiency in
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 66

66 M. A. Green

2011, would be expected to have lower ERE. The best a-Si cell to date has quite a
low ERE of 5.3 × 10−6 %. However, from Fig. 2.11, its bandgap can be seen to be
in an appropriate region for the highest possible conversion efficiency with such
low ERE.
The dye-sensitised device is again a small 9.9% efficient submodule, although
with energy conversion performance close to the best-performing small-area cell
(11.2%). Although the operational principles of a dye-sensitised cell are vastly dif-
ferent from the previous p-n junction devices, the reciprocity relationships giving
rise to Rau’s relationship are expected still to apply (Trupke et al., 1999). ERE is
7.2 × 10−6 %, similar to, but slightly higher than, the a-Si device.
The final two devices are two of the first organic photovoltaic devices to exceed
8% energy conversion efficiency, with very different energy absorption thresholds,
as apparent from Fig. 2.12a. Both devices, however, display very similar ERE in
the 3−4 × 10−7 % range. A detailed discussion of the electroluminescence from a
range of organic cells is presented elsewhere (Vandewal et al., 2009).
Even though organic light-emitting diodes (OLEDs) of quantum efficiency
>20% have been reported, the different requirements for photovoltaics (Kirchartz
et al., 2008b) greatly reduce the radiative efficiency. In particular, the blending
required to produce bulk heterojunction devices with high carrier collection greatly
reduces the radiative efficiency.
The above results show that the external radiative efficiency (ERE) can be
unambiguously deduced from standard solar cell efficiency measurements and is
a useful parameter in comparing the performance of cells of both the same and
completely different technologies. As each technology matures, ERE will evolve
towards the 100% value required for limiting performance.

2.6 Multiple-junction devices

2.6.1 The N-cell stack
Photovoltaic conversion efficiency can be increased by using cells of different
bandgaps to convert different parts of the solar spectrum. In the optimal stacked
configuration of Fig. 2.13 (Marti and Araujo, 1996), the power output from the nth
cell in an N-cell stack of radiatively-efficient devices is

+
E gn+

2πq Vn E gn E 2d E E 2d E
Pn = 3 2 ξ −
h c E gn exp(E/kTS ) − 1 E gn exp[(E − q Vn )/kTA ] − 1
(2.34)
+ is the bandgap of cell (n + 1) with E + = ∞.
where E gn gN
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 67

Limits to Photovoltaic Energy Conversion Efficiency 67

low energy pass filters

N n+1 n n-1 1

Figure 2.13 Optimal configuration for stacked tandem cells. After Marti and Araujo (1996).

Expressing this in terms of polylogarithms

2πqξ Vn
Pn = {ε2 [Li 1 (e−εgn ) − t Li 1 (γn )] + 2εgn [Li 2 (e−εgn ) − t 2 Li 2 (γn )]
h 3 c2 (kTs )3 gn
+
+ 2[Li 3 (e−εgn ) − t 3 Li 3 (γn )] − εgn
+2
[Li 1 (e−εgn ) − t Li 1 (γn+ )]
+ +
+
− 2εgn [Li 2 (e−εgn ) − t 2 Li 2 (γn+ )] − 2[Li 3 (e−εgn ) − t 3 Li 3 (γn+ )]} (2.35)

+
where γn = e(q Vn −E gn )/ kTA and γn+ = e(q Vn −E gn )/ kTA .
With the equations formulated in this way in terms of standard functions,
global maxima can be simply found with standard mathematic packages just by
entering these equations constrained by current matching if this feature is desired.
Figure 2.14 was generated for the unconstrained 3-cell case with a simple contour
plot command using Mathematica, involving essentially two steps of code.

2.6.2 Infinite stack of cells

The case of an infinite stack of tandem cells is of particular interest since the limit
in this case represents the fundamental limit on photovoltaic conversion efficiency
for a time-symmetric system (Green, 2003). For an infinite stack of cells, the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 68

68 M. A. Green

Figure 2.14 Contour plots for 47%, 48% and 49% efficiency versus normalised bandgap
εg1 , εg2 and εg3 , for a 3-cell tandem stack (unconstrained) with t = 0.05 and ξ = 1/46,200.
Peak efficiency is 49.3% for εg1 = 1.58, εg2 = 2.77 and εg3 = 4.36 (multiply by 0.517 eV
to convert to actual bandgap values for TA = 300 K, TS = 6000 K). Source: Green and Ho
(2011).

incremental current density at each wavelength is given by (Marti and Araujo,

1996; Green, 2003)

2πq E g2 E g2
di = 3 2 ξ + (1 − ξ )
h c exp(E g /kTs ) − 1 exp(E g /kTA ) − 1

E g2
− d Eg (2.36)
exp[(E g − q V )/kTA ] − 1

The extra term in (1 − ξ ) absent in Eq. (2.34) corresponds to thermal back-

ground radiation, important if low-bandgap cells (E g < kTA ln[1/ξ ]) are to be
considered in the analysis.
In terms of polylogarithms, the corresponding incremental power density out-
put becomes
2πqVE 2g
dP = [ξ Li 0 (e−E g / kTs ) + (1 − ξ )Li 0 (e−E g /TA )
h 3 c2
− Li 0 (e(q V −E g )/ kTA )]d E g (2.37)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 69

Limits to Photovoltaic Energy Conversion Efficiency 69

Partial differentiation with respect to V gives the following two equations for
the optimal value Vm
q Vm
Li 0 (e(q Vm −E g )/ kTA ) + Li −1 (e(q Vm −E g )/ kTA )
kTA
= ξ Li 0 (e−E g / kTs ) + (1 − ξ )Li 0 (e−E g / kTA )
= Li 0 (eq(Voc −E g )/ kTA ) (2.38)

At open circuit, for ξ = 1, Eq. (2.38) shows Voc = E g (1 − TA /TS )/q, a

result also noted elsewhere (Landsberg, 2000). For other values of ξ , also from
Eq. (2.38), Voc for each cell can be expressed explicitly in terms of its specific E g

q Voc = E g − kTA ln{1 + 1/[ξ Li 0 (e−E g / kTs ) + (1 − ξ )Li 0 (e−E g / kTA )]}
(2.39)

The above solution has the asymptotic values

q Voc = E g (1 − TA /Ts )/[1 + (1/ξ − 1)TA /Ts ] for E g  kTA (2.40)

and

q Voc = E g (1 − TA /Ts ) + kTA ln(ξ ) for E g  kTs (2.41)

with both solutions equivalent for ξ = 1.

As for simple ideal-diode cell models, Vm can only be expressed implicitly in
terms of either Voc or E g

 
kTA q Vm kTA 1 − γm
Vm = Voc − ln /(1 − γm ) + 1 + ln (2.42)
q kTA q 1 − γoc
where γm = e(q Vm −E g )/ kTA and γoc = e(q Voc −E g )/ kTA .
Since the terms involving Vm on the right-hand side of Eq. (2.42) are relatively
small, a good approximation allowing conversion to an explicit form is to replace
Vm on this side by Voc , which is explicitly known from Eq. (2.39). This approxi-
mation has been shown to work well for conventional cell analysis (Luque et al.,
1978; Green, 1982). It will locate a voltage Vm∗ on the output curve close to Vm to
give essentially the same, but slightly less, power output where

kTA q Voc
Vm∗ = Voc − ln /(1 − γoc) + 1 (2.43)
q kTA
Since the term in the square bracket is already an overestimate, deleting the
(1 − γm ) and the (+1) terms will generally improve accuracy, provided Voc is
constrained to values greater than kT A . Using this modified version of Eq. (2.43)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 70

70 M. A. Green

in combination with the asymptotic expansion of Eq. (2.41) gives an explicit rela-
tionship for Vm∗ which remains useful for E g ≥ kTA [1 − ln(ξ )]/(1 − TA /Ts ). For
smaller E g , only a small power contribution would be generated, so setting Vm to
zero for such E g will not cause any significant error in total output. This approxi-
mation was found to result in efficiencies which matched exact values to close to
three significant digits over the whole range of likely interest.
The resulting expression may also be integrated analytically. Inserting into
Eq. (2.37) and expanding to first order in TA , then integrating, gives the result

η∞ = 1.0 − t[ A + B ln(t) + C ln(ξ )] + O(t 2 ) (2.44)

where

B = C = −[Li 3 (1)/Li 4 (1)]/3 = −0.3702

A = 1 + [(5 − 2γ E )Li 3 (1)/6 + Li 3 (1)/3]/Li 4 (1) = 1.6508

with γ E = 0.5772 (Euler Gamma) and Li 3 (1) = ∂ Li s (1)/∂s|s=3 = −0.1981.

This expression gives reasonable accuracy (better than 3% relative) over the
range of t and ξ of likely interest. Increasing the dependence on the ln(t) term by
adjusting parameters to A = 1.1861, B = −0.4901 and C = −0.3487 gives the
best fit over this range with results accurate to close to three significant digits. This
result is to be compared to the similar ‘Landsberg’ limit on the efficiency of a system
where time-symmetry is not imposed. A simple approximate, but accurate, new
expression given earlier for this limit based on past work (Press, 1976; Landsberg
and Tonge, 1979) is

η L = 1 − t[E + D ln(ξ ) + (4/3 − E)ξ 0.9 ] + O(t 4 ) (2.45)

with E = 11/12 + Li 3 (1)/Li 4 (1)]/3 = 1.2869; D = − [Li 3 (1)/Li 4 (1)]/3 =

−0.3702.
The key difference is the ln(t) term present in the time-symmetric case, fun-
damentally due to the need to move away from open-circuit conditions in this
case.

2.7 Other high-efficiency options

Apart from multijunction cells, a range of other options has been suggested to
improve on the performance of a single-junction cell, with limiting efficiency from
some of the key approaches shown in Fig. 2.1. Each, in its own way, attempts
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 71

Limits to Photovoltaic Energy Conversion Efficiency 71

to favourably impact the fundamental trade-off of Fig. 2.9 involved in a single-

threshold device. The key options listed in Fig. 2.1 and briefly considered below
have been discussed in more detail elsewhere (Green, 2003).

2.7.1 Hot-carrier cells

One of the highest efficiency of such options is the ‘hot-carrier’ cell (Ross and
Nozik, 1982), with efficiency potential higher than a 5-tandem cell stack. The key
idea, captured in Fig. 2.15, is to avoid, or at least reduce, the loss of the energy
in excess of the bandgap imparted to electron-hole pairs generated by photons of
energy larger than the bandgap.
In a conventional cell, this extra energy is quickly lost during collisions of
the energetic particles with the atoms of the absorber (phonon emission). Within
a picosecond or less, the carriers relax back to close to the band edge, as explic-
itly assumed in the Trivich–Flinn analysis of Section 2.5.1. However, there would
appear to be no fundamental reason why this relaxation process could not be con-
siderably slowed, retaining some of this excess energy in a ‘hot’ carrier population,
as in Fig. 2.15b.
To date, the best way of doing this seems to be manipulation of the mechanical
properties of the absorber on a microscopic scale. This permits engineering of the
properties of allowed mechanical vibrations or phonons in the absorber, particularly
the energy ranges allowed for acoustic and optical phonons. Electrons give up
heat most effectively to the lattice through the relatively immobile, high-energy
optical phonons that then decay to more mobile, lower-energy acoustic phonons.
By manipulating the allowed energy ranges for these phonons, it is possible to

(a) (b)
Figure 2.15 Normally photogenerated carriers thermalise with the lattice as in (a). In hot-
carrier cells, excess energy is stored in a hot-carrier distribution, as shown in (b). Source:
Green and Hansen (2002).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 72

72 M. A. Green

slow the decay of optical into acoustic phonons, and hence the energy relaxation,
of carriers by reducing the rate of energy exchange with optical phonons.
Calculations involving the relevant experimental parameters of quantum wells
and bulk materials show that hot-carrier effects should be apparent in suitably
designed devices operating under highly concentrated sunlight (Guillemoles et al.,
2005; Aliberti et al., 2010). Development of new materials will be required to
extend this attribute to devices operating under global sunlight.

2.7.2 Multiple carrier or exciton generation

It was suggested some time ago (Deb and Saha, 1972) that photovoltaic perfor-
mance might be improved by creating multiple electron-hole pairs by high-energy
photons, as indicated in Fig. 2.16. This has the potential to substantially increase
the efficiency of the cell (Werner et al., 1995; Nozik, 2008).
Multiple electron-hole pair generation has been observed in silicon devices
(Kolodinski et al., 1994). However, the effect is too small to provide any measurable
gain in conventional silicon cell performance.
More recent interest in this effect has been stimulated by evidence for multiple
exciton generation (MEG) in Si, PbS, and PbSe quantum dots (Nozik, 2010).
Moreover there is an independent interest in replacing the dye molecules that
form the absorber in dye-sensitised solar cells by quantum dots, giving at least
a conceptual path to possible implementation of devices taking advantage of the
MEG effect.
The limiting efficiency of multiple carrier or MEG devices under maximally
concentrated direct light is quite close to the time-symmetric limit (85.9% versus

IQE

3.0

2.0

1.0

0 λ
λG
(a) (b)
Figure 2.16 (a) Schematic of the impact ionisation process whereby one energetic photon
creates multiple electron–hole pairs; (b) energetically feasible internal quantum efficiency,
where λg is the wavelength corresponding to the threshold energy for electron–hole pair
creation. Source: Green and Hansen (2002).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 73

Limits to Photovoltaic Energy Conversion Efficiency 73

86.8% for a 6000 K black-body source). However, MEG device are less suited to
global sunlight, with the limiting efficiency dropping considerably, to just above
that of a conventional 2-cell tandem stack (43.6% for a 6000 K black-body source,
compared with 42.9% for a 2-cell stack and 68.2% for the time-symmetric limit).
This is due to the larger bandgap required for efficient operation at low intensities,
reducing the gain from multiple carrier generation (Green, 2003). However, this
approach may well provide a way of improving cell performance with specific cell
technologies.

2.7.3 Impurity level and intermediate-band devices

Many years ago, Wolf (1960) suggested the use of photoexcitation through defect
levels in the normally forbidden cell bandgap to boost cell performance as shown in
Fig. 2.17. The idea was for some time controversial, since it was thought any gain
would be offset by the increased recombination introduced by the defect (Guttler
and Queisser, 1970). However, analysis of an extension of the idea, where the
defects are present in sufficient quantities to form a mid-gap band (Luque and
Marti, 1997), pushed the issue beyond doubt (Brown and Green, 2002). Quite
substantial gains in performance are possible if the defect levels/bands have quite
specific properties.
The limiting performance of a cell based on this approach is similar to that
of a 3-cell tandem, but slightly lower. This is because the effect can be modelled
by the specific connection shown in Fig. 2.18 of three different cells of energy
bandgap equal to the energy thresholds of the three different absorption processes
of Fig. 2.17 (Green, 2003).

conduction
band

impurity
3 subgap level
photons
1

valence
band

Figure 2.17 Impurity photovoltaic effect where electron-hole pairs are generated by sub-
bandgap photons. After Keevers and Green (1994).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 74

74 M. A. Green

C2
C1

C3

Figure 2.18 Equivalent circuit of the intermediate-band solar cell. The circuit elements
shown represent ideal Shockley–Queisser solar cells. Source: Green and Ho (2011).

The multiple quantum well cell (Barnham and Duggan, 1990; also Chapter
10 of this volume) and quantum-dot structures (Luque and Marti, 2010) have
become the preferred devices for investigating the implementation of this approach.
Although devices with creditable performance have been demonstrated, these have
not yet exceeded the performance of controls without the intermediate levels.

2.7.4 Up- and down-conversion

A disadvantage of introducing defects into the bulk of the cell is that they are
likely to produce additional recombination paths (Guttler and Queisser, 1970). If
the associated optical excitations are decoupled from the cell by placing an up-
converter at the rear of the cell, as in Fig. 2.19a, this potential disadvantage can be
overcome (Trupke et al., 2002a).
Similarly, Fig. 2.20a shows a schematic of how a down-converter might be
used on the front of a cell, and Fig. 2.20b shows an equivalent circuit. In this case,
the down-converter would strongly absorb light of energy more than twice the
main cell bandgap. The intermediate level of the down-converter would lie close
to halfway across its effective bandgap so that both the relaxation of electrons
to this level and from this level to lower levels produced photons just above the
bandgap energy of the main cell. With this additional constraint on energy transition
thresholds, the limiting efficiency is close to that of a 2-cell tandem (Trupke et al.,
2002b).

2.7.5 Thermal approaches

The limiting efficiency of the concentrated solar power (CSP) approach such as
used in the PS10 and PS20 solar power towers operating in Spain (Abengoa, 2011)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 75

Limits to Photovoltaic Energy Conversion Efficiency 75

Incident light

C3
C1 C2
Solar cell
Insulator
C4

Reflector

(a) (b)
Figure 2.19 (a) Solar cell and an electronically isolated up-converter. Sub-band-gap light
transmitted by the solar cell is up-converted into high-energy photons, which are subse-
quently absorbed in the solar cell; (b) equivalent circuit of the up-conversion system. The
solar cell in front of the up-converter is denoted C1. The up-converter is represented by the
three cells C2, C3 and C4. The series connected cells C3 and C4 represent the two inter-
mediate transitions while C2 represents the band-to-band transitions. After Trupke et al.
(2002a).

Incident light

Incident C3
C2 C1
Converter
Insulator C4
Solar cell
Reflector

(a) (b)
Figure 2.20 (a) Schematic diagram of the down-conversion system with the luminescence
converter located on the front surface of a solar cell. High-energy photons with h̄ν > 2E g
are absorbed by the converter and down-converted into two lower energy photons with
h̄ν > E g , which can both be absorbed by the solar cell; (b) equivalent circuit of the system.
The luminescence converter is represented by three solar cells C2, C3, and C4 representing
the band-to-band transitions and the two types of intermediate transitions, respectively.
High-energy photons are absorbed by C2, which then biases the cells C3 and C4 into the
forward direction and causes them to emit electroluminescence absorbed by the main cell
C1. After Trupke et al. (2002b).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 76

76 M. A. Green

W
TR

Sun
QT TA
R
Carnot
converter
QT
A
re-emitted

Figure 2.21 Solar thermal system based on the conversion of heat collected by an absorber
to electricity at the Carnot efficiency. Source: Green and Hansen (2002).

can be calculated from the schematic of Fig. 2.21. A black absorber receives energy
from the Sun, heating it to a high temperature, TR . Surfaces of the absorber other
than those receiving energy from the Sun have low emissivity so that only the
receiving surface loses energy back to the Sun. Heat from the absorber drives a
Carnot converter.
The limiting efficiency can be written down almost by inspection for a black-
body Sun
  
TR4 TA
ηT = 1 − 1− (2.46)
ξ TS4 TR

For ξ = 1, the optimum TR is 2544 K for TS = 6000 K and TA = 300 K,

giving a limiting efficiency of 85.4%, quite close to the time-symmetric limit of
86.8%. However, as the power able to be focussed onto the receiver decreases,
the efficiency drops rapidly, as shown by the lower curve of Fig. 2.22, to a
limiting value of only 11.7% for undiluted sunlight (including sky radiation, in
this case).
The problem is the decreasing Planckian effective temperature of the sunlight
at long wavelengths as the intensity reduces (Fig. 2.5b). The solution is to use a
receiver with selective absorption properties, so that it absorbs only those energies
higher than a threshold value, E T , determined by the crossover where the Planckian
temperature is higher than the receiver temperature.
In this case, with ε R = E T /kTR and ε S = E T /kTS , the limiting efficiency
becomes
 
∞ ε 3 dε TR4  ∞ ε 3 dε  
ε ε −1 − ε ε −1
 S e ξ TS4 R e  TA
ηTS =   ∞ ε 3 dε  1 − (2.47)
ε
TR
0 e −1
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 77

Limits to Photovoltaic Energy Conversion Efficiency 77

Figure 2.22 Limiting efficiency for the scheme of Fig. 2.21. Source: Green and Ho (2011).

As seen by the upper line in Fig. 2.22, use of such a selective absorber
significantly improves the energy conversion efficiency at low incident sunlight
intensities.
Similar energy conversion efficiency limits to those shown in Fig. 2.22 apply to
photovoltaic devices that follow a thermal route to conversion. In considering these
results, it should be borne in mind that these thermal limits are much more difficult
to attain in practice than the corresponding limits for conventional photovoltaic
devices, since the latter already take into account some of the practical difficulties
in obtaining Carnot-like efficiencies.
A solar thermophotovoltaic system uses the radiant energy from the receiver
rather than its heat energy. Figure 2.23 shows a schematic illustrating the basic
concept. An ideal solar cell with an ideal monochromatic light filter acts as an
ideal Carnot converter of radiation emitted by the receiver (Green, 2003). Hence
the scheme of Fig. 2.23 fulfils the ideal and is bounded by the same limiting
efficiencies as for Fig. 2.22b.
High expectations were held for this scheme in the late 1970s (Swanson, 1979),
but results to date have been quite modest. A good explanation of the reasons why
one might expect lower than ideal performance is given by Harder and Würfel
(2003).
A development of the idea is based on the realisation that a light-emitting diode
can convert electricity to light with greater than 100% power conversion efficiency,
if heat input to the diode is ignored. This gives rise to the ‘thermophotonic’ converter
of Fig. 2.24, which has advantages in terms of the narrow range of emission energies
of the heated device and much stronger emission at low receiver temperatures.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 78

78 M. A. Green

sunlight
NP

Figure 2.23 Thermophotovoltaic conversion with narrow passband filter. Source: Green
and Hansen (2002).

IH IC
filter

+ +
VH VC
NP PN

TH TC
Figure 2.24 Thermophotonic conversion, with narrow-pass filter and independent power
supplies. V and I are the voltages and currents at the terminals of the p-n junction devices,
while T is the device temperature and the subscripts H and C refer to the hot device heated
by sunlight and the cold receiver, respectively. Source: Green and Hansen (2002).

Another way of obtaining Carnot-like efficiencies in converting heat in the

receiver to an electric output, at least in principle, is via the thermionic effect of
Fig. 2.25 where electrons are ‘boiled off’ from the receiver (the cathode) and col-
lected by a lower work function anode. In the standard simplified analysis (Schwede
et al., 2010), electrons are collected at the anode under ‘flat-band’ conditions, where
the applied voltage is equal to the work function difference between the cathode
and anode.
Because the anode work function must be at least greater than the ratio of anode
to cathode temperature times that of the cathode work function to ensure a nett
electron flux from the cathode to anode, this results in a Carnot efficiency limit for
high cathode work functions (and low emitted flux densities). In practice, efficiency
is well below the Carnot limit, with 15% often cited as the best to date in converting
heat. Solar conversion efficiencies demonstrated to date (Ogino et al., 2004) have
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 79

Limits to Photovoltaic Energy Conversion Efficiency 79

electrons
anode
cathode

sunlight

electricity

Figure 2.25 Thermionic conversion. Sunlight heats a cathode that emits electrons which
are collected by a lower work function anode.

been very modest (well below 1% and even 0.1%). Radiative losses to the cathode
(despite attempts to reduce the emissivity of the facing surfaces of the anode and
cathode), conduction losses through wiring to the electrodes, resistive losses in the
wires and space-charge effects are key contributors to the difference (Hatsopoulos
and Gyftopoulos, 1979; Habedank, 1993).
It has been suggested that efficiency can be improved by combining both
heat and light excitation of the cathode. In one case, combined photoemission and
thermionic emission has been suggested, with the latter possibly augmented by
hot-carrier effects (Smestad, 2004). A more recent suggestion has been to use a
p-type semiconductor as the heated cathode, with above-bandgap light used to
photoexcite the cathode and sub-bandgap light and carrier relaxation losses also
used to heat the cathode (Schwede et al, 2010).
The advantage of the latter configuration is that the flatband condition is still
determined by the relatively high work function of the p-type semiconductor, estab-
lished by the majority-carrier Fermi level. However, photoexcitation moves the
minority-carrier electron quasi-Fermi level to closer to the vacuum level, enhancing
photoemission (and also reducing the minimum value of the anode work function).
The disadvantage is that to absorb above-bandgap light, the anode must also emit
it back to the Sun. For 100% radiative efficiency, this means each absorbed pho-
ton will result in a photon of energy just above the bandgap being emitted back
to the Sun, a Trivich–Flinn inefficiency. For some bandgaps, this will represent
40% of the incident energy, a huge loss before any serious attempt at conversion
begins.
Since the Carnot conversion potential of standard thermionic conversion obvi-
ously cannot be exceeded in this way, the advantage lies in offering lower possible
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 80

80 M. A. Green

operating temperatures for the converter, which may allow some of the previous
losses mentioned in traditional converters to be reduced.
The other traditional converter of heat to electricity is a thermoelectric con-
verter. The close relationship to photovoltaics is highlighted by a casual remark by
Würfel (2000) that one way of reducing heat loss to the lattice in a hot-carrier cell
is to let the lattice heat up to the hot-carrier temperature. A heated absorber with
selective energy contacts (Section 2.7.1) acts as a hot-carrier cell in the Würfel
limit (Würfel, 1995; Green, 2003). The conversion efficiency limits upon such a
cell are identical to the limits plotted in Fig. 2.22. This type of device has been
independently suggested as an improved thermoelectric converter (O’Dwyer et al.,
2005), demonstrating the close links of thermoelectrics to advanced photovoltaic
concepts.

2.8 Summary
Although most commercial single-junction solar cells operate at more than 50%
of their theoretical potential, there is scope for improving performance by edging
closer to this limit. Even more scope for performance gains arises through advanced
‘third generation’ options that overcome the basic trade-off in a standard cell: too
high a bandgap results in too little current; too low a bandgap gives too little
voltage.
Extension of the Shockley–Queisser approach offers a ‘bottom-up’ method for
analysing the ultimate potential of any new scheme, such as the recently suggested
photo-enhanced, thermionic conversion (Schwede et al., 2010). Formulation in
terms of polylogarithms, as in the present chapter, simplifies the programming
of the appropriate analyses. ‘Top-down’ thermodynamic analyses have already
proved useful in identifying the features required for the ultimate photovoltaic
converter. Calculations in terms of a black-body Sun have the virtue of timelessness
and also of evaluation of results analytically. However, results for the tabulated
reference spectra used for experimental cell measurements are also of considerable
interest, although these reference spectra are more ephemeral. For example, a
recent downscaling of the likely value of the solar constant from 1366.1 W m−2 to
1360.8 W m−2 (Kopp and Lean, 2011) is likely to have a flow-on effect to first the
standard AM0 spectra, then possibly the terrestrial spectra. Some key results for
present spectra of interest are compared in Table 2.2.
Apparent from Table 2.2 is the large difference between the limiting perfor-
mance of standard single-junction cells and the ultimate limits on solar energy
conversion. Capitalising on the potential factor of two performance gains would
see thin-film module technologies go beyond 25% energy conversion efficiency,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 81

Limits to Photovoltaic Energy Conversion Efficiency 81

Table 2.2 Summary of photovoltaic efficiency limits under black-body radiation at 300 K
and under the three ASTM G173-03 spectra at 25 ◦ C.
Blackbody
(T S = 6000 K,
T A = 300 K) AM0 AM1.5G AM1.5D

Efficiency Direct Isotropic Direct Isotropic Direct* Isotropic Direct Isotropic*

Landsberg 93.3 73.7 93.1 72.6 92.9 74.0 92.8 73.3
(Ultimate)
Time-Symmetric 86.8 68.2 86.3 66.6 86.4 68.0 86.0 67.2
(Infinite
Tandem)
Single Cell 40.9 31.0 40.7 30.4 45.2 33.8 45.1 33.4
(Bandgap eV) (1.109) (1.306) (1.006) (1.245) (1.119) (1.339) (1.117) (1.135)

∗ Included for reference.

and possibly beyond 30% in the longer term. For wafer-based technology, the
figures would be correspondingly higher.
Several options for obtaining such efficiency improvements have been
reviewed. Tandem cell stacks such as those discussed in Chapter 7 of the current vol-
ume are clearly the most promising route based on experience to date. Hot-carrier
cells have the attraction of offering higher efficiency than a 5-cell tandem stack in a
simple 2-terminal configuration, although considerable theoretical and experimen-
tal progress will be required before any performance improvement is likely to be
demonstrated with this approach. Up- and down-conversion offer the prospects for
‘supercharging’ the performance of existing cell technologies, although improve-
ments in the performance for both types of converters are required also before
any performance gain would be expected. Intermediate-band and multiple exciton
generation devices offer improvements for specific material systems.
The ‘thermal approaches’ — thermophotovoltaics, thermionics, thermoelec-
tric or hot lattice ‘hot-carrier’cells — offer tantalisingly high limiting efficiencies.
However, the gap between what is likely to be obtained in practice and the theoret-
ical potential is expected to be much larger with such approaches, because of the
much higher energy losses that are likely to occur in practice. Such approaches may
have potential in concentrating systems, but this is where tandem cell stacks have
had the most impact terrestrially and are showing unabated potential for further
development.
As is already becoming apparent as the photovoltaic industry matures, effi-
ciency will be a key driver to future cost reduction. Evidence for this is the emphasis
given by thin-film companies such as First Solar to quarterly improvements in mod-
ule efficiency (from under 8% in 2003 to over 13% in 2013) and the increasing
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 82

82 M. A. Green

number of wafer silicon companies announcing improvements in production cell

efficiency to the 19–20% range. Efficiency may be possible without economy, as
past experience has demonstrated, but to reach the economies required for photo-
voltaics to make a large impact on global energy supply, improved efficiency is
crucial.

Acknowledgement
This work has been supported by the Australian Government through the Australian
Renewable Energy Agency (ARENA). Responsibility for the views, information
or advice expressed herein is not accepted by the Australian Government.

References
Abengoa (2011), www.abengoasolar.com.
Aliberti P. Feng Y., Takeda Y., Shrestha S. K., Green M. A. and Conibeer G. (2010), ‘Inves-
tigation of theoretical efficiency limit of hot carriers solar cells with a bulk indium
nitride absorber’, J. Appl. Phys. 108, 94507–94517.
ASTM Standard Extraterrestrial Spectrum Reference E-490-00 (tabulated at
http://rredc.nrel.gov/solar/spectra/am0/).
Barnham K. and Duggan G. (1990), ‘A new approach to high-efficiency multi-band-gap
solar cells’, J. Appl. Phys. 67, 3490–3493.
Blom W. M., Mihailetchi V. D., Koster L. J. A. and Markov D. E. (2007), ‘Device physics
of polymer:fullerene bulk heterojunction solar cells’, Adv. Mater. 19, 1551–1566.
Brown A. S. (2003), ‘Ultimate efficiency limits of multiple energy threshold photovoltaic
devices’, Thesis for the Degree of Doctor of Philosophy, Centre for Third Generation
Photovoltaics and the School of Electrical Engineering, University of New South Wales.
Brown A. S. and Green M. A. (2002), ‘Impurity photovoltaic effect: fundamental energy
conversion efficiency limit’, J. Appl. Phys. 92, 1329–1336.
Brown A. S., Green M. A. and Corkish R. (2002), ‘Limiting efficiency for multi-band solar
cells containing three and four bands’, Physica E 14, 121–125.
Chapin D. M., Fuller, C. S. and Pearson G. L. (1954), ‘A new silicon p-n junction photocell
for converting solar radiation into electrical power’, J. Appl. Phys. 25, 676–677.
Cooke D., Gleckman P., Krebs H., O’Gallagher J., Sage D. And Winston R. (1990), ‘Sunlight
brighter than the Sun’, Nature 346, 802.
Cummerow R. L. (1954a), ‘Photovoltaic effect in p-n junctions’, Phys. Rev. 95, 16–21.
Cummerow R. L. (1954b), ‘Use of silicon p-n junctions for converting solar energy to
electrical energy’, Phys. Rev. 95, 561–562.
Deb S. and Saha H. (1972), ‘Secondary ionisation and its possible bearing on the perfor-
mance of a solar cell’, Solid State Electronics 15, 89–1391.
Green M. A. (1982), ‘Accuracy of analytical expressions for solar cell fill factors’, Solar
Cells 7, 337–340.
Green M. A. (2011), ‘Limiting photovoltaic efficiency under new ASTM G173-based
reference spectra’, Progr. Photovoltaics 20, 954–959.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 83

Limits to Photovoltaic Energy Conversion Efficiency 83

Green M. A. (2012), ‘Analytical treatment of Trivich–Flinn and Shockley–Queisser photo-

voltaic efficiency limits using polylogarithms’, Progr. Photovoltaics 20, 127–134.
Green M. A. and Hansen J. (1998), ‘Catalogue of photovoltaic drawings’, 1st edition, Pho-
tovoltaics Special Research Centre, University of New South Wales, Sydney.
Green M. A., Zhao J., Wang A., Reece P. J. and Gal M. (2001), ‘Efficient silicon light
emitting diodes’, Nature 412, 805–808.
Green M. A. and Hansen J. (2002), ‘Catalogue of photovoltaic drawings’, 2nd edition,
Centre for Photovoltaic Engineering, University of New South Wales, Sydney.
Green M. A. (2003), Third Generation Photovoltaics: Advanced Solar Energy Conversion,
Springer-Verlag, Berlin.
Green M. A., Emery K., Hishikawa Y. and Warta W. (2009), ‘Solar cell efficiency tables
(Version 33)’, Progr. Photovoltaics 17, 85–94.
Green M. A., Emery K., Hishikawa Y. and Warta W. (2010), ‘Solar cell efficiency tables
(Version 36)’, Progr. Photovoltaics 18, 346–352.
Green M. A., Emery K., Hishikawa Y. and Warta W. (2011), ‘Solar cell efficiency tables
(Version 37)’, Progr. Photovoltaics, 19, 84–92.
Green M. A. and Ho J. (2011), ‘Catalogue of photovoltaic drawings’, 3rd edition, School
of Photovoltaic and Renewable Energy Engineering, University of New South Wales,
Sydney.
Grondahl, L. O. (1933), ‘The copper cuprous oxide rectifier and photoelectric cell’, Rev.
Modern Phys. 5, 141.
Guillemoles J.-F., Conibeer G. and Green M. A. (2005), ‘Phononic engineering with nanos-
tructures for hot carrier solar cells’, PVSEC-15, Shanghai, pp. 375–376.
Guttler G. and Queisser H. J. (1970), ‘Impurity photovoltaic effect in silicon’, Energy Con-
version 10, 51–55.
Habedank O. D. (1993), ‘Analysis of Topaz II and space-R space nuclear power plants
using a modified thermionic model’, Thesis presented to the Faculty of the School of
Engineering of the Air Force Institute of Technology, Air University, USA.
Harder N. P. and Green M. A. (2003), ‘Thermophotonics’, Semiconductor. Sci. Tech. 18,
S270–S278.
Harder N. P. and Würfel P. (2003), ‘Theoretical limits of thermophotovoltaic generation of
electricity’, Semiconductor. Sci. Tech. 18, S151–S157.
Hatsopoulos G. N. and Gyftopoulos E. P. (1979), Thermionic Energy Conversion, Volume
2, MIT Press, Cambridge, MA.
Henry C. H. (1980), ‘Limiting efficiencies of ideal single and multiple gap terrestrial solar
cells’, J. Appl. Phys. 51, 4494–4500.
International Standard, IEC 60904–3, Edition 2 (2008), ‘Photovoltaic devices — part 3:
Measurement principles for terrestrial photovoltaic (PV) solar devices with reference
spectral irradiance data’, ISBN 2–8318–9705-X, Int. Electrotechnical Commission.
Jackson E. D. (1955), Trans. Conf. on Use of Solar Energy, Tuscon, AZ, 31 Oct–1 Nov.,
122 (also U.S. Patent No. 2,949,498, 16 Aug., 1960).
Keevers M. J. and Green M. A. (1994), ‘Efficiency improvements of silicon solar cells by
the impurity photovoltaic effect’, J. Appl. Phys. 75, 4022–4033.
Kirchartz T., Helbig A. and Rau U. (2008a), ‘Note on the interpretation of electro-
luminescence images using their spectral information’, Solar Energy Mater. Solar
Cells 92, 1621–1627.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 84

84 M. A. Green

Kirchartz T., Helbig A., Reetz W., Reuter M., Werner J. H. and Rau U. (2009), ‘Reciprocity
between electroluminescence and quantum efficiency used for the characterization of
silicon solar cells’, Progr. Photovoltaics 17, 394–402.
Kirchartz T., Mattheis J. and Rau U. (2008b), ‘Detailed balance theory of excitonic and bulk
heterojunction solar cells’, Phys. Rev. B 78, 235320.
Kirchartz T. and Rau U. (2007), ‘Electroluminescence analysis of high efficiency Cu(In,
Ga)Se2 solar cells’, J. Appl. Phys. 102, 104510.
Kolodinski S., Werner J. H., Wittfchen J. H. and Queisser H. J. (1994), ‘Quantum efficiencies
exceeding unity in silicon leading to novel selection principles for solar cell materials’,
Solar Energy Mater. Solar Cells 33, 275–286.
Kopp G. and Lean J. L. (2011), ‘A new, lower value of total solar irradiance: Evidence and
climate significance’, Geophys. Research Lett. 38, L01706.
Kron G., Rau U. and Werner J. H. (2003), ‘Influence of built-in voltage on the fill factor of
dye-sensitised solar cells’, J. Phys. Chem. B 107, 13258–13261.
Landsberg P. (2000), ‘Efficiencies of solar cells: where is Carnot hiding?’, Conf. Record,
16th European Photovoltaic Solar Energy Conf., Glasgow.
Landsberg P. T. and Tonge G. (1979), ‘Thermodynamics of the conversion of diluted radi-
ation’, J. Phys. A 12, 551–562.
Landsberg P. T. and Tong G. (1980), ‘Thermodynamic energy conversion efficiencies’, J.
Appl. Phys. 51, R1–R20.
Lee K., Kim J. Y. and Heeger A. J. (2008), ‘Titanium oxide films as multifunctional com-
ponents in bulk heterojunction ‘plastic’ solar cells’, in Brabecc, Dyakonovv, Scherfu
(eds), Organic Photovoltaics, Wiley, Weinheim.
Lewin L. (1981), ‘Polylogarithms and associated functions’, New York: North-Holland (also
see http://en.wikipedia.org/wiki/Polylogarithm).
Loferski J. J. (1956), ‘Theoretical considerations governing the choice of the optimum
semiconductor for photovoltaic energy conversion’, J. Appl. Phys. 27, 777–784.
Luque A. and Marti A. (1997), ‘Increasing the efficiency of ideal solar cells by photon
induced transitions at intermediate levels’, Phys. Rev. Lett. 78, 369.
Luque A. and Marti A. (2010), ‘The intermediate band solar cell: progress toward the
realization of an attractive concept’, Adv. Mater. 22, 160–174.
Luque A., Ruiz J. M., Cuevas A., Eguren J. and Agost J. G. (1978), Proc. 1st Commis-
sion of the European Communities Conf. on Photovoltaic Solar Energy, Luxembourg,
pp. 269–277.
Marti A. and Araujo G. (1996), ‘Limiting efficiencies for photovoltaic energy conversion
in multigap systems’, Solar Energy Mater. Solar Cells 43, 203–222.
Mathematica (2008), Version 7.0 Publisher: Wolfram Research, Inc. Champaign, Illinois.
Mihailetchi V. D., Koster L. J. A. and Blom W. M. (2004), ‘Effect of metal electrodes on the
performance of polymer:fullerene bulk heterojunction solar cells’, Appl. Phys. Lett.
85, 970–972.
Nozik A. J. (2008), ‘Fundamentals and applications of quantum-confined structures’, in
Archer M. D. and Nozik A. J. (eds), Nanostructured and Photoelectrochemical Systems
for Solar Photon Conversion, Imperial College Press, London.
Nozik A. J. (2010), ‘Nanoscience and nanostructures for photovoltaics and solar fuels’,
Nano Lett. 10, 2735–2741.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 85

Limits to Photovoltaic Energy Conversion Efficiency 85

O’Dwyer, M. F., Lewis, R. A. and Zhang C. (2005), ‘Electronic efficiency in nanostructured

thermionic and thermoelectric devices’, Phys. Rev. B 72, 205330, 1–10.
Ogino A., Muramatsu1 T. and Kando M. (2004), ‘Output characteristics of solar-power-
driven thermionic energy converter’, Jpn. J. Appl. Phys. 43, 309–314.
Ohl R. S. (1941), ‘Light sensitive electric device’, U.S. Patent No. 240252, filed 27 March
1941; ‘Light sensitive electric device including silicon’, U.S. Patent No. 2443542, filed
27 May 1941.
Parrott J. E. (1986), ‘Self-consistent detailed balance treatment of the solar cell’, IEEE Proc.
133, pp. 314–318.
Pauwels H. and de Vos A. (1981), ‘Determination of the maximum efficiency solar cell
structure’, Solid State Electronics 24, 835–843.
Planck M. (1913), Vorlesungen über die Theorie der Warmestrahlung. Leipzig: Barth [trans-
lated into English as Planck M, ‘The theory of heat radiation’, Dover, New York, 1959].
Press W. H. (1976), ‘Theoretical maximum for energy from direct and diffuse sunlight’,
Nature 264, 734–735.
Ramsdale C. M., Barker J. A., Arias A. C., MacKenzie J. D., Friend R. H. and Greenham N. C.
(2002), ‘The origin of the open-circuit voltage in polyfluorence-based photovoltaic
devices’, J. Appl. Phys. 92,. 4266–4270.
Rau U. (2007), ‘Reciprocity relation between photovoltaic quantum efficiency and electro-
luminescent emission of solar cells’, Phys. Rev. B 76, 085303.
Ries H. (1983), ‘Complete and reversible absorption of radiation’, Appl. Phys. B 32, 153.
Riordan M. and Hoddeson L. (1997), Crystal Fire: The Birth of the Information Age, Norton,
New York.
Rittner E.S. (1954), ‘Use of p-n junctions for solar energy conversion’, Phys. Rev. 96,
1708–1709.
Ross R. T. and Nozik A. J. (1982), ‘Efficiency of hot-carrier solar energy converters’ J. Appl.
Phys. 53, 3813–3818.
Schwede J. W., Bargatin, I. Riley, D. C., Hardin, B. E., Rosenthal, S. J., Sun Y., Schmitt F.,
Pianetta, P., Howe R. T. and Shen Z. X. (2010), ‘Photon-enhanced thermionic emission
for solar concentrator systems’, Nature Mater. 9, 762–767.
Shockley W. (1949), ‘The theory of p-n junctions in semiconductors and p-n junction
transistors’, Bell System Tech. J. 28, 435–89.
Shockley W. and Queisser H. J. (1961), ‘Detailed balance limit of efficiency of p-n junction
solar cells’, J. Appl. Phys. 32, 510–519.
Smestad G. P. (2004), ‘Conversion of heat and light simultaneously using a vacuum photo-
diode and the thermionic and photoelectric effects’, Solar Energy Mater. Solar Cells
82, 227–240.
Snaith H. J. and Grätzel M. (2006), ‘The role of a ‘Schottky barrier’ at an electron-collection
electrode in solid-state dye-sensitised solar cells’, Adv. Mater. 18, 1910–1914,
Swanson R. M. (1979), ‘A proposed thermophotovoltaic solar energy conversion system’,
Proc. IEEE 67, pp. 446–447.
Tiedje T., Yablonowitch E., Cody G. D. and Brooks B. G. (1984), ‘Limiting efficiency of
silicon solar cells’, IEEE Trans. Electron Devices ED-31, 711–716.
Trivich D. and Flinn P. A. (1955), ‘Maximum efficiency of solar energy conversion by
quantum processes’ in Daniels F. and Duffie J. (eds), Solar Energy Research, Thames
and Hudson, London.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch02 page 86

86 M. A. Green

Trupke T., Green M. A. and Würfel P. (2002a), ‘Improving solar cell efficiencies by the
up-conversion of sub-band-gap light’, J. Appl. Phys. 92, 4117–4122.
Trupke T., Green M. A. and Würfel P. (2002b), ‘Improving solar cell efficiencies by down-
conversion of high-energy photons’, J. Appl. Phys. 92, 1668–1674.
Trupke T., Würfel P., Uhlendorf I. and Lauermann I. (1999), ‘Electroluminescence of the
dye-sensitized solar cell’, J. Phys. Chem. B 103, 1905–1910.
Trupke T., Zhao J., Wang A., Corkish R. and Green M. A. (2003), ‘Very efficient light
emission from bulk crystalline silicon’, Appl. Phys. Lett. 82, 2996–2998.
Vandewal K., Tvingstedt K., Gadisa A., Inganäs O. and Manca J. V. (2009), ‘On the origin of
the open-circuit voltage of polymer–fullerene solar cells’, Nature Mater. 8, 904–909.
Werner J. H., Brendel R. and Queisser H. J. (1995), ‘Radiative efficiency limit of terrestrial
solar cells with internal carrier multiplication’, Appl. Phys. Lett. 67, 1028–1030.
Whitehead A. N. (1911), An Introduction to Mathematics, Henry Holt & Co., New York,
Chapter V.
Wilson A. H. (1931), ‘The theory of electronic semiconductors’, Proc. Royal Soc. London
A133, 458–491.
Wolf M. (1960), ‘Limitations and possibilities for improvement of photovoltaic solar energy
converters’, Proc. IRE 48, 1246–1263.
Würfel P. (1982), ‘The chemical potential of radiation’, J. Phys. C 15, 3967–3985.
Würfel P. (1995), ‘Is an illuminated semiconductor far from thermodynamic equilibrium’,
Solar Energy Materials and Solar Cells 38, 23–28.
Würfel P. (2000), Olympic Conf. on Third Generation Photovoltaics, Sydney (unpublished
work).
Würfel P., Brown A. S., Humphrey T. E. and Green M. A. (2005), ‘Particle conservation in
hot-carrier solar cell’, Progr. Photovoltaics 13, 277–285.
Würfel P. and Wolfgang R. (1980), ‘Upper limit of thermophotovoltaic solar-energy
conversion’, IEEE Trans. Electron Devices ED-27, 745–750.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 87

CHAPTER 3

CRYSTALLINE SILICON SOLAR CELLS

MARTIN A. GREEN
Australian Centre for Advanced Photovoltaics
School of Photovoltaic and Renewable Energy Engineering
University of New South Wales
Sydney, N.S.W. Australia, 2052
[email protected]

Vast Power of the Sun Is Tapped by Battery Using Sand Ingredient

Front page headline, New York Times, 26 April, 1954.

3.1 Overview
Front page headlines in the New York Times and the Wall Street Journal in 1954
heralded to the world the demonstration of the first reasonably efficient solar cells,
an announcement made possible by the rapid development of crystalline silicon
technology for miniaturised electronics.The majority of solar cells made since then
have been based on silicon in monocrystalline or large-grained polycrystalline
form. There are two main reasons for this. One is that silicon is an elemental
semiconductor with good stability and a well-balanced set of electronic, physical
and chemical properties, the same set of strengths that have made it the preferred
material for microelectronics. The second reason why silicon cells have been so
dominant is that the success of silicon in microelectronics created an enormous
industry where the economies of scale directly benefited the initially much smaller
photovoltaics industry.
In the early days of the industry, most silicon cells were made using thin
wafers cut from large cylindrical monocrystalline ingots prepared by the exacting
Czochralski (CZ) crystal growth process and doped to about one part per million
with boron during the ingot growth. A larger proportion now use what are referred
to as ‘multicrystalline’ wafers sliced from ingots prepared by a simpler directional
solidification technique, that produces large-grained polycrystalline ingots. To pro-
duce a cell, these boron-doped starting wafers generally have phosphorus diffused
at high temperatures a fraction of a micron into the surface to form the p-n junction
required for photovoltaic action. Metal contacts to both the n- and the p-type side

87
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 88

88 M. A. Green

of the junction are formed by screen printing metal pastes which are then densified
by firing at high temperature. Each cell is typically 12–20 cm either in diameter or
along either side if square or rectangular.
Cells generally are sold interconnected and encapsulated into a weatherproof,
glass-faced package known as a module, as shown in an exploded view in Fig. 3.1
Each module typically contains 72 cells soldered together in series. Each individual
cell gives a maximum output of about 0.63 V in sunlight. The output current depends
on cell size and the sunlight intensity (solar irradiance) but generally lies in the
4–12 A range in bright sunshine. The packaging consists of a glass/polymer lami-
nate with the positive and negative leads from the series-connected cells brought out
in a junction box attached to the module rear. Such modules have proved extremely
reliable in the field, with most manufacturers offering a 25-year warranty on the
module power output, one of the longest warranties provided for any commercial
product (saucepans have been suggested as one of the few manufactured products

rear cover
cells EVA

EVA

glass

Figure 3.1 Exploded view of a standard silicon photovoltaic module. The different layers
shown are laminated together under pressure at a temperature around 140–150 C where the
transparent EVA (ethylene vinyl acetate) softens and binds the different layers together on
cooling. Source: Green and Hansen (1998).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 89

Crystalline Silicon Solar Cells 89

with a comparable warranty period — my most recently purchased pan only having
a 10-year warranty!).
The efficiency of the cells in the module would typically lie in the 15–19%
range, appreciably lower than the fundamental ‘detailed balance’ or Shockley–
Queisser limit of 33.5% for silicon (see Chapter 2). Module efficiency is slightly
lower than that of the constituent cells due to the area lost by frames and gaps
between cells, with module efficiency generally lying in the 12–15% range (Haase
and Podewils, 2011). Over the last few years, commercial cells and modules of
significantly higher performance have been available in multi-megawatt quantities
using more advanced cell processing technology, discussed in more detail later.
Cells of 20–22% efficiency and module efficiency in the 18–20% range are now
commercially available. (Unless otherwise noted, all efficiencies quoted in this
chapter are at standard test conditions, namely with a cell temperature of 25C
under 1000 W m−2 sunlight intensity with the standard global air mass 1.5 spectral
distribution.)
This chapter discusses the historical but steadily weakening links between
silicon solar cells and the broader microelectronics industry. Also discussed are
standard and improved methods for preparing silicon cell substrates and for
processing cells to extract as much performance as possible from these at the
lowest possible overall cost. The chapter also describes progress with supported
silicon films, which have the advantage of significantly reduced silicon material
requirements.

3.2 Silicon cell development

3.2.1 The early years
The initial development of silicon photovoltaics was inextricably intertwined with
the development of the general silicon electronics field and the subsequent micro-
electronics industry. The rapid increase in interest in the properties of silicon in
the 1940s was triggered by the surprising photovoltaic properties demonstrated by
serendipitously formed p-n junctions, as described below. This increased interest
led directly to the development of point-contact and junction transistors and ulti-
mately to integrated circuits.
The earliest commercial silicon electronic devices were silicon point-contact
or ‘cat’s whisker’ diodes dating from the early 1900s (Riordan and Hoddeson,
1997). These devices rectified electrical signals at a junction formed by pressing
a thin metal wire against polycrystalline silicon (other semiconductors, such as
silicon carbide, were also used). ‘Cat’s whisker’ diodes were key components
in early radios. By the 1930s, they were largely replaced by thermionic valves.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 90

90 M. A. Green

However, the evolving field of microwaves created a renewed interest in cat’s

whisker diodes in the mid-1930s. At Bell Laboratories in the USA, Russell Ohl
guessed that impurities caused the erratic behaviour often observed whereby the
cat’s whisker only operated correctly if located on a ‘hot spot’ in the silicon.
Ohl therefore encouraged colleagues to grow purer silicon, by melting the purest
material available in a quartz capsule, and then cooling.
In one specific slow-cooled ingot, the eighteenth in the series (Riordan and
Hoddeson, 1997), Ohl and colleagues observed a surprisingly large photovoltage of
about half a volt when the ingot was illuminated by a flashlight. This ingot showed
two distinct types of properties, dubbed ‘positive’ ( p-type) and ‘negative’ (n-type),
depending on the polarity required for easy current flow between the material and
a metal wire placed on the silicon surface, and also the voltage polarity observed
under illumination. It was quickly realised that the junction between the p-type and
n-type regions, the p-n junction, caused the unusual properties. The first silicon
solar cells were formed by cutting the ingot to include sections with both a p- and
n-type region and then applying metal contacts (Ohl, 1941). These earliest silicon
solar cells, shown in Fig. 3.2a, were only a fraction of percent efficient, but were
still very much better in performance than earlier photovoltaic devices based on
selenium or cuprous oxide.
The ‘grown-in’ junctions in the earliest cells arose from the serendipitous
distribution of p-type and n-type impurities in the silicon (boron and phosphorus,
respectively) resulting from the slow solidification process (Scaff et al., 1949).
Ohl realised that more controllable ways of forming the junction would give better
performance. In the early 1950s, he was involved in experiments where surface
junctions were formed by implanting helium at high energy into the surfaces of
p-type polycrystalline silicon slices (Kingsbury and Ohl, 1952). Although this
approach produced improved cells of efficiency estimated to be up to 1% (Fig. 3.2b),
this work was soon overtaken by independent improvements in silicon technology
also made at Bell Laboratories, particularly in two areas.
The first was the development of techniques for preparing single crystals of
silicon using the Czochralski growth method. The second was the formation of
junctions by high-temperature diffusion of dopant impurities into the silicon sur-
face. Combining these two techniques, researchers at Bell Laboratories were able
to announce the first modern silicon solar cell in 1954, in what was one of the early
successes of the diffused junction approach (Chapin et al., 1954). Figure 3.2c shows
the resulting cell structure.
The impressive performance of these cells by previous standards, up to 6%
energy conversion efficiency, created enormous interest at the time, as the newspa-
per headline at the beginning of this chapter suggests, and also generated unbounded
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 91

Crystalline Silicon Solar Cells 91

top contact (+)

p-type

rear contact (-)

n-type
grown junction

(a)

He ion bombardment junction p-type

(b)

n-type
p-type

(-)
(+) (+)
(c)
Figure 3.2 (a) Silicon solar cell reported in 1941 relying on ‘grown-in’ junctions formed
by impurity segregation in recrystallised silicon melts; (b) helium-ion bombarded junction
device of 1952; (c) first modern silicon cell, reported in 1954, fabricated on single-crystalline
silicon wafers with the p-n junction formed by dopant diffusion. Source: Green (1995).

enthusiasm for future applications. This enthusiasm proved to be premature,

although the cells did find an almost immediate use in spacecraft. Space appli-
cations drove the rapid improvement in cell technology such that, by the early
1960s, cell energy conversion efficiency of about 15% under terrestrial sunlight
had been demonstrated, with the cells finding a secure market niche powering
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 92

92 M. A. Green

antireflection coating
top metal finger

n-type

p-type rear metal

contact

(a)

thin
n-type
layer

p+ layer
p-type

(b)
textured surface

n-type

p-type
p+

rear contact

(c)
Figure 3.3 (a) Space silicon cell design developed in the early 1960s which became a
standard design for over a decade; (b) shallow junction ‘violet’ cell; (c) chemically textured
non-reflecting ‘black’ cell. Source: Green (1995).

a rapidly increasing number of satellites (Wolf, 1976). The basic cell design that
evolved (Fig. 3.3a), remained unchanged from the early 1960s for almost a decade.
In the early 1970s, a reassessment of cell design at COMSAT Laboratories
showed that a shallower diffusion combined with more closely spaced metal fingers
gave a substantial improvement in the cell performance by improving the response
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 93

Crystalline Silicon Solar Cells 93

to blue wavelengths (Lindmayer and Allison, 1973). The resulting cells, shown in
Fig. 3.3b, known as ‘violet’ cells due to their characteristic colour arising from the
shorter wavelengths reflected, produced the first improvement in cell performance
for over a decade.
This improvement was augmented by the realisation that incorporating a thin
heavily doped layer under the back contact, creating a so-called ‘back surface
field’, gave unexpected benefits (Godlewski et al., 1973). This approach worked
best if the rear doped layer was formed by alloying the underlying silicon with
aluminium deposited over the cell rear. Not long afterwards, the idea of using
anisotropic chemical etches to form geometrical features on the silicon surface was
successfully demonstrated, also at COMSAT Laboratories (Haynos et al., 1974),
and resulted in a further boost in performance, taking terrestrial cell efficiency to
above 17% (Fig. 3.3c). The surface features consisted of square-based pyramids
defined by slowly etching {111} crystallographic planes. These greatly reduced
reflection from the cell surface, giving these ‘black cells’ the appearance of black
velvet after antireflection coating.
The improvements of the early 1970s came about primarily by enhancing the
ability of the cell to collect carriers generated by the incoming photons. Since
cells now appeared to be performing to close to their full potential in this regard,
it seemed that any further improvement in cell performance would come from
increased open-circuit voltage. Gains in this area became the focus of work directed
at improving cell efficiency throughout the second part of the 1970s, largely as a
result of a program directed by NASA-Lewis aimed at targeting better space cell
performance (Brandhorst and Bernatowicz, 1980).
On the commercial front, the oil embargoes of the early 1970s generated
widespread interest in alternative sources of terrestrial energy. A small terrestrial
photovoltaic industry came into existence largely as a result of the US Govern-
ment’s photovoltaic program. One component of this program (Christensen, 1985),
arguably the most successful in terms of developing the industry and its products,
involved a staged series of purchases of photovoltaic modules meeting increasingly
stringent specifications.
The first such purchase in 1975/76, known as ‘Block I’, was remarkable for the
diversity of both cell fabrication and module encapsulation approaches used in the
product supplied by four different manufacturers. One manufacturer, Spectrolab
of Sylmar, California, supplied cells where the contacts had been applied using
screen printing (Ralph, 1975), the forerunner of the billions of cells of this type
which were to follow. In the ‘Block II’ purchases under this program (1976/77),
the same company combined screen-printed cells with a laminated module design
(Fig. 3.1), a combined approach that had been adopted by almost all commercial
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 94

94 M. A. Green

Figure 3.4 Screen-printed crystalline silicon solar cell (not to scale). Source: Green (1995).

manufacturers by the early 1980s and, with relatively minor modification, remains
the present commercial standard.
The main features of a commercial screen-printed cell are shown in Fig. 3.4.
The basic cell design is similar to that of a standard space cell of the 1960s
(Fig. 3.3a), but incorporates the surface texturing of the ‘black’ cell of Fig. 3.3c as
well as the screen-printing approach to applying the front and rear contacts. Present
efficiencies are quite similar to those demonstrated by ‘black’ cells in 1974.
Since the Block II purchases of 1976/77, no major changes were made in either
the basic screen-printed approach to cell fabrication or to the cell encapsulation
approach until quite recently. Considerable attention, however, has been directed
towards reducing the cost of the silicon wafer, initially grown by the Czochralski
technique, since this accounts for a large fraction of the cost of a standard silicon
module. The most successful approach has been the simplification of the ingot
growth processes by using cruder directional solidification approaches to produce
multicrystalline ingots (Ferrazza, 1995; Rodriguez et al., 2011).
The first multicrystalline silicon cells developed specifically for the terres-
trial market were reported in 1976 (Lindmayer, 1976; Fischer and Pschunder,
1976) and commercial multicrystalline cells have been available since the late
1970s. These multicrystalline approaches involve basically a reversion to the earlier
ingot-forming approaches for crystal rectifiers, techniques pre-dating the micro-
electronics explosion. In recent years, multicrystalline silicon cells accounted for
the majority of the total market for photovoltaic product (Hering, 2011). Another
major area of emphasis has been to reduce the thickness of the silicon wafer by
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 95

Crystalline Silicon Solar Cells 95

slicing it more thinly. This resulted in the replacement of inner diameter sawing
methods traditionally favoured by the microelectronics industry by wire-cutting
approaches, described in more detail in Section 3.3.1.

3.2.2 The path to 25% cell efficiency

On the research front, it had become apparent by the late 1970s that oxide passiva-
tion of the cell surfaces was the key to improving open-circuit voltage. The early
1980s saw a series of successively improved oxide-passivated cells fabricated by
the author’s group at the University of New South Wales (UNSW) taking silicon
cell efficiency past 18%, then past 19% and finally 20%, the ‘four minute mile’ of
the photovoltaics area. The UNSW group has held the ‘world record’ for silicon
cell performance, almost without interruption, since this time, with this record until
recently standing at 25% (Green, 2009).
In 1985, the UNSW-developed microgrooved PESC cell (passivated emitter
solar cell) of Fig. 3.5a became the first silicon cell to exceed 20% energy conversion
efficiency. The same basic approach was then used by several other groups to
produce cells of similar efficiency, with commercial quantities produced for solar
car racing and for space. The approach is characterised by the use of a thin thermally
grown oxide to ‘passivate’ (reduce the electronic activity of) the top surface of
the junction diffusion layer (the emitter of the cell), combined with the use of a
shallow, high sheet resistivity phosphorus diffusion for this emitter. Another feature
is the use of photolithography to produce a relatively small contact area to this
emitter region by defining openings in the ‘passivating oxide’. Photolithography
is also used to pattern the top contact fingers and to align these fingers to the oxide
openings.
The rear of the cells borrows the ‘alloyed-aluminium back-surface-field’
approach from earlier space cells. In this approach, a layer of aluminium is
deposited on the rear of the cell and alloyed into the cell at temperatures above the
silicon–aluminium eutectic. After cooling, this produces a layer of p-type silicon
heavily doped with aluminium at the rear of the silicon substrate. This reduces rear
contact recombination. Some improvement of substrate quality also occurs during
alloying through defect and impurity ‘gettering’.
Cells of a similar quality to the first 20%-efficient PESC cell also found their
way into volume terrestrial cell manufacture through the laser-grooved buried con-
tact cell of Fig. 3.5b. This cell retains an alloyed aluminium rear and also incorpo-
rates the improvements to front surface passivation first demonstrated by the PESC
cell. To make the approach suitable for low-cost production, the photolithographic
metallisation of the PESC cell was replaced by a combination of laser grooving, to
define the areas to be metallised, followed by electroless metal plating. The oxide
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 96

96 M. A. Green

finger
microgroove

n+ thin oxide

p
p+

back contact
(a)

oxide n+
n++

p-type plated metal

p+

metal
(b)
Figure 3.5 (a) The microgrooved passivated emitter solar cell (PESC cell) of 1985, the first
silicon cell to exceed 20% efficiency; (b) buried contact solar cell. Source: Green (1995).

or other dielectric coating the top surface in this case not only serves as surface
passivation, but also as a diffusion mask to confine the heavy diffusion to the laser-
grooved areas and as a plating mask for the subsequent plating of metal into these
grooved areas. During the early 1990s, the buried contact approach produced the
highest performance terrestrial cells made in any appreciable volume, with effi-
ciency in the 17–18% range obtained using standard commercial silicon wafers
(Jordan and Nagle, 1994).
The next major laboratory improvement in silicon cell design came in the use
of oxide passivation along both the front and rear surfaces, first demonstrated in
the rear point-contact solar cell developed by Stanford University. As shown in
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 97

Crystalline Silicon Solar Cells 97

n+ busbar
p+ busbar

+
n

+
+ n
n

oxide

sunlight

Figure 3.6 Rear point-contact solar cell, which demonstrated 22% efficiency in 1988 (cell
rear shown uppermost). Source: Green (1995).

Fig. 3.6, this cell has an unusual design in that both positive and negative contacts
are on the rear surface of the cell. Although this might, at first sight, appear to
be a regression to a similar design to that used in the first modern silicon cell
of Fig. 3.2c, there is a substantial difference in the way the two types of cells
operate. The modern rear junction cells take advantage of the excellent quality of
silicon now available. Carrier diffusion lengths are several times the cell thickness,
allowing carriers photogenerated near the top surface of the cell to diffuse to the rear
contacts. In the earlier device, the junctions at top and rear surfaces are electrically
connected around the cell edge (Fig. 3.2c). Most carriers in this earlier cell are
collected by the top junction and flow around the cell edges to the rear contact. The
rear point-contact cell demonstrated 22% efficiency in 1988 and has since been
successfully commercialised, as subsequently discussed.
The next improvement in silicon cell efficiency came, again at UNSW, by
combining the earlier developments in the PESC cell sequence with the front and
rear oxide passivation first demonstrated in the rear point-contact cell. This is
possible in a number of ways as shown in Fig. 3.7. In the PERC cell (passivated
emitter and rear cell) of Fig. 3.7a, the first successfully demonstrated, rear contact
is made directly to the silicon substrate through holes in the rear passivating oxide.
This approach works reasonably well provided the substrate is sufficiently heavily
doped to ensure low contact resistance between the metal and substrate (below
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 98

98 M. A. Green

finger "inverted" pyramids

n+ n oxide
(a)
p-silicon

finger "inverted" pyramids

rear contact oxide

p+

n+ n oxide
(b) p-silicon p+
p+ p+

rear contact oxide

finger "inverted" pyramids

+
p

n+ n oxide
(c) p-silicon p
+
p+ p+ p

finger "inverted" pyramids

rear contact oxide

p+

n+ p-silicon
oxide
n p+
(d) p+ p+

rear contact oxide

Figure 3.7 A family of four related high-efficiency solar cell structures: (a) the passivated
emitter and rear cell (PERC cell); (b) the passivated emitter, rear locally diffused cell (PERL
cell) which took efficiency above 24% in the early 1990s and subsequently to 25%; (c) the
passivated emitter, rear totally diffused cell (PERT cell); and (d) the passivated emitter, rear
floating junction cell (PERF cell). Source: Green and Hansen (1998).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 99

Crystalline Silicon Solar Cells 99

about 0.5
cm resistivity for p-type substrates). The PERC cell is often suggested
as a relatively low cost way for making silicon cells above 20% efficiency, since it
is the simplest of the approaches of Fig. 3.7.
Historically, the next improvement was demonstrated by the PERL cell (pas-
sivated emitter, rear locally diffused) of Fig. 3.7b. In this case, local diffusion is
used in the area of the rear point contact to provide a minority-carrier-reflecting
region between this contact and the substrate and to reduce contact resistance. This
approach produced the first 24% efficient silicon cell in 1994 (Zhao et al., 1995) and
the world best result for silicon of 25% (Green, 2009) under the standard AM1.5
global spectrum. The PERL cell has been used in reasonably large quantities in
solar car racing and in space cells.
The third cell of Fig. 3.7c is the PERT cell (passivated emitter, rear totally
diffused). This has given performance almost equivalent to the PERL cell and also
offers some fabrication simplifications. The PERT cell has also been used in space
cell production. The final structure shown in Fig. 3.7d, the PERF cell (passivated
emitter rear floating junction) offers perhaps the best long-term potential for high
performance. This structure has produced the highest open-circuit voltage of the
cells of Fig. 3.7, with open-circuit voltage up to 720 mV demonstrated under stan-
dard test conditions (Wenham et al., 1994), together with efficiencies above 23%.
One feature of these cell designs is the very effective trapping of light within
the cell. By depositing metal over the entire rear surface of the cell but ensuring it
is displaced from the silicon substrate by an intervening layer of oxide, very high
rear reflectance is obtained for light striking this rear reflector from within the cell.
When combined with appropriate geometrical structure on the front surface of the
cell, weakly absorbed light that is reflected from this rear surface can be trapped
quite effectively within the cell, taking advantage of total internal reflection from the
front surface. This greatly extends the response of the cell to infrared wavelengths.
Cells that convert infrared wavelengths with an efficiency approaching 50% have
been demonstrated (Green et al., 1992).

3.2.3 Subsequent silicon cell developments

Although the 25% efficiency figure established by the PERL cell remains close to
the highest efficiency demonstrated for silicon under standard AM1.5 global test
conditions, there have been recent developments that demonstrate paths to future
progress.
The most efficient silicon cells to date have been made on p-type substrates.
This is largely because phosphorus diffusion is, in several ways, easier than boron
diffusion, with a wider parameter space giving good results in terms of uniformity
of diffusion and minimal degradation of the starting quality of the substrate. Oxide
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 100

100 M. A. Green

and nitride passivation of phosphorus-diffused surfaces also give good results. This
makes phosphorus the preferred dopant for the more critical top doped layer of the
cell. For the rear surface, there are fewer constraints on the required properties of
the diffusion, allowing boron diffusions to work well there.
The Stanford University rear point-contact cell of Fig. 3.6 is shown as origi-
nally conceived, without any diffusion on the surface exposed to sunlight. Adding a
non-contacted or ‘floating’ phosphorus diffusion to the top (sunlight-exposed) sur-
face improved performance, with 22.7% efficiency soon thereafter demonstrated,
long the best for any cell on an n-type silicon substrate.
In attempting to adapt their p-type wafer work to n-type substrates, the UNSW
group were able to equal this result, but again using a phosphorus diffusion along
the top surface. A ‘front surface field’ structure was used, as in the PERT cell of
Fig. 3.7c, but with the substrate n-type rather than p-type. The standard PERL
structure of Fig. 3.7b on n-type substrates with all polarities reversed gave slightly
poorer results due to the more difficult challenge involved in passivating the B-
doped top surface.
The challenge of improving cell performance on n-type substrates was sub-
sequently met by the combined efforts of the Fraunhofer Institute in Freiburg,
Germany, and ECN, Netherlands (Benick et al., 2008; 2009). Dielectrics such as
silicon oxide and nitride work well on n-type surfaces since they tend to have nett
positive electrical charge within them, like many other dielectrics, with this charge
tending to attract electrons to underlying silicon surfaces (Godfrey and Green,
1980). This also depletes holes in the surface region, reducing surface recombi-
nation rates by squeezing off the supply of these minority carriers. It was found
that Al2 O3 could be prepared with nett negative charge, repelling electrons from
the interface to the interior, similarly restricting the supply for recombination on
p-type surfaces, such as for the top surface of an n-type cell if B-doped. A sig-
nificant improvement in n-type cell performance to 23.4% was demonstrated by
using the PERL cell structure of Fig. 3.7b with all polarities reversed but with the
top surface coating (normally silicon oxide under a ZnS/MgF double layer antire-
flection coating) replaced by 30 nm Al2 O3 overlaid by 40 nm Si3 N4 (Benick et al.,
2009). Although below the best PERL cell on p-type substrates on each of the
major cell parameters (Voc , i sc , FF), this result is important in demonstrating the
ability to passivate B-doped surfaces, increasing future flexibility in cell design.
Another significant development has been the HIT (heterojunction with intrin-
sic thin layer) cell, developed by Sanyo (now Panasonic) and shown in Fig. 3.8
(Tanaka et al., 1993; Sawada et al., 1994; Maruyama et al., 2006). Here the
problems raised by diffusions and their passivation are avoided totally by using
deposited layers of hydrogenated amorphous silicon (a-Si:H) for both positive
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 101

Crystalline Silicon Solar Cells 101

metal

TCO
+
p (a-Si)
i(a-Si)

n(c-Si)

i(a-Si)
+
n (a-Si)

bottom electrode metal

Figure 3.8 HIT (heterojunction with intrinsic thin layer) cell on textured crystalline silicon
substrate (after Green and Hansen, 1998).

and negative contacts to top and rear surfaces. The hydrogen, incorporated in the
5–10% range by atomic volume, modifies the electronic structure of the a-Si:H
alloy, producing much larger bandgap (about 1.7 eV compared to 1.1 eV for crys-
talline silicon). The a-Si:H can be doped n- or p-type, although the electronic
quality of doped material tends to be poor. This different form of silicon is widely
used in its own right for thin-film solar cells (Chapter 4), making the HIT cell
structure a marriage between two successful cell technologies. Features such as
transparent conducting oxides (TCO) are also used, with these widely used in thin-
film technologies but not in standard silicon technology. The HIT also has some
similarities with III–V cell technology (Chapter 7), where heterojunctions between
semiconductors with different bandgaps are a common feature.
The advantage of using wider bandgap material is that minority-carrier con-
centrations are greatly suppressed by the wider bandgap, correspondingly reducing
recombination rates. In the HIT cell structure, all significant recombination occurs
in the wafer substrate or at its surfaces. An important additional feature of the HIT
cell is a very thin layer of intrinsic a-Si:H, only 10–20 nm wide, interposed between
the wafer and the thin doped a-Si:H regions. This thin intrinsic layer helps reduce
interfacial recombination, possibly by reducing defect states in the bandgap of the
a-Si:H in these regions that would be created by doping.
The HIT cell has a number of attractive features. One is that all processing of the
starting wafer can be effected at low temperature, allowing the initial wafer quality
to be retained during processing. The resulting cells have a bifacial structure, so
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 102

102 M. A. Green

can respond to light from either direction and, in principle, wafers of either polarity
could be used, although n-type wafers are the preferred choice.
Standard HIT cells have also demonstrated outstanding efficiencies, up to
24.7% for large area cells (Maruyama et al., 2006), until recently the best ever
result for n-type substrates. The outstanding performance parameter has been the
open-circuit voltage, with values up to 745 mV demonstrated for 23% efficient cells
(compared to 708 mV for 25% efficient PERL cells). This high voltage results from
inherently lower surface recombination due to the passivation effects of the overly-
ing a-Si:H. Since there is a very direct and general relationship between the open-
circuit voltage and the temperature coefficient of performance of the cell (Green,
2003), this gives a significant advantage at high temperature compared with stan-
dard commercial silicon cells (where Voc might be only in the 610–630 mV range).
Technically, the main disadvantages with the standard HIT cell approach are
optical. The top doped a-Si:H layer absorbs ultraviolet and blue light and, because
of its poor electronic quality, this absorption does not contribute to the cell current.
Similarly, the conducting oxide layer will absorb incoming photons, particularly at
the infrared end of the spectrum. This will additionally degrade the effectiveness
of any light-trapping scheme that attempts to boost the cell’s response to poorly
absorbed infrared light. Consequently, current output from the cell is 5–10% lower
than from other similarly high-efficiency cells. The cost of processing wafers to
cells is also reported to be moderately high (Song et al., 2010). The HIT approach
has been important in demonstrating that almost perfect surface passivation is
possible in silicon cells, showing what ultimately might be possible in the future.
The final recent improvement to be discussed relates to the Stanford rear
point-contact cell (Fig. 3.6), an important contributor to the evolution of sili-
con cell efficiency, as already noted. This cell was commercialised by SunPower
(De Ceuster et al., 2007) and now produces the highest efficiency commercial
panels. A recent re-evaluation of cell design produced impressive results (Cousins
et al., 2010), also for n-type substrates. Figure 3.9 shows the cell structure of this
‘Generation III’ product.
Improved diffusions for both contacted and non-contacted regions of the rear
of the cell have contributed to the improved performance (Cousins et al., 2010).
A combination of oxide/nitride passivation is used for non-contacted areas where
diffusions are quite light (∼1018/cm3 ). For the contact regions, a heavier doping is
used as a compromise between the optimum for those regions directly contacted by
metal and those, also in the contact region, but not so directly contacted. An impres-
sive efficiency of 24.2% was demonstrated on a large-area, commercially-sized
cell (156 cm2 ) using commercial production equipment, with an impressive Voc of
721 mV (Cousins et al., 2010).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 103

Crystalline Silicon Solar Cells 103

Figure 3.9 SunPower ‘Generation III’ cell. Source: Green and Ho (2011).

30

lambertian
Efficiency (%)

20
planar, one pass

10

0
1 10 100 1000 10000
Thickness (µ m)

Figure 3.10 Limiting efficiency of a silicon solar cell as a function of cell thickness with
and without Lambertian light trapping (global AM1.5 spectrum, 100 mW cm−2 , 25◦ C).
Source: Green (1995).

3.2.4 Opportunities for future improvements

How will silicon cell design evolve in the future? Some insight is provided by
Fig. 3.10, which shows the calculated intrinsic energy conversion efficiency bounds
on single-junction silicon solar cells, with and without ‘Lambertian’ light trapping.
In ‘Lambertian’ light-trapping schemes, the light direction within the cell is ran-
domised (Green, 1995) allowing path length enhancements to be quite readily
calculated (these reach about 50 in idealised situations). The limiting efficiency
of a silicon cell is somewhat less than the Shockley–Queisser limit of Chapter 2.
This is because, as well as intrinsic radiative recombination processes, a competing
intrinsic Auger recombination process is also present (Green, 1984; Tiedje et al.,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 104

104 M. A. Green

1984). This limits efficiency to about 29%, compared to the Shockley–Queisser

limit of 33.5%.
The best laboratory cells have demonstrated around 85% of the achievable
efficiency. In the best experimental PERL cells, performance losses of the order of
5% arise from less than ideal values of each of the short-circuit current, open-circuit
voltage and fill factor. The short-circuit current losses are most easily identified
and reduced. These come from metal finger coverage of the top surface, top surface
reflection loss, and less than perfect light trapping in the experimental cells. The
voltage loss arises from finite surface and bulk recombination in excess of the
lower limit imposed by intrinsic Auger recombination processes (Green, 1984).
The fill factor loss comes not only from ohmic series resistance loss within the cell,
but also from the same factors producing the open-circuit voltage loss. Parasitic
recombination must be sufficiently reduced so that the dominant recombination
component at the cell’s maximum power point is Auger recombination. This is
a more challenging requirement than the corresponding criterion at open-circuit
voltage (Green, 1984).
As opposed to the 25% cell efficiency demonstrated by laboratory devices,
most present manufacturers of mainstream commercial product would be very
pleased to be producing cells of 19% efficiency. Some of the difference between
laboratory and commercial cell performance is due to poorer quality of silicon
substrate material. A large component, however, is due to limits imposed by the
present screen-printing process predominantly used for commercial cell fabrica-
tion. The penalty for the processing simplicity offered by this approach is a partly
compromised cell design, since a heavily doped emitter layer is required and rear-
contact recombination and optical properties are non-ideal. Improved designs sub-
sequently discussed offer the apparently contradictory advantages of both higher
cell performance and lower overall manufacturing costs.
It seems that, eventually, efficiencies well above 20% will be the norm for
mainstream commercial cells, achieved by paying attention to the passivation of
both front and rear surfaces, by thinning the cells to reduce bulk recombination
and by modifying the crystal growth processes to produce low-cost silicon cus-
tomised for photovoltaics, particularly in its ability to withstand high-temperature
processing without loss of electronic quality.
An interesting result highlighted by Fig. 3.10 is that light trapping allows
high performance, in principle, from silicon cells that are only 1 µm thick. This
provides a justification for expecting very high performance, eventually, from the
thin, supported silicon cells discussed in Section 3.7. To approach the limiting per-
formance, the demands on bulk quality become less severe as the cell becomes thin-
ner (Green, 1995). However, the demands on light trapping and surface passivation
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 105

Crystalline Silicon Solar Cells 105

become more severe. Interestingly, Fig. 3.10 shows that the commercial imperative
to decrease wafer thickness to decrease silicon costs (SEMI, 2011) may actually
lead to higher energy conversion efficiency, with appropriate device structures.
Various approaches have been suggested that have the potential, in principle,
for exceeding even the challenging efficiency limits of Fig. 3.10. These include
the use of tandem cells, the use of high-energy photons to create more than one
electron-hole pair (Werner et al., 1994), or the use of sub-bandgap photons in
schemes such as incorporation of regions of lower bandgap (Healy and Green,
1992), multiple quantum wells (Barnham and Duggan, 1990), mid-gap impurity
levels (Wolf, 1960) or up- and/or down-conversion (Trupke et al., 2002a; 2002b).
The tandem cell approach appears the most likely to have impact in the long term,
once the problems are overcome with lattice-matching a top cell with a suitable
bandgap to silicon.

3.3 Substrate production

3.3.1 High-purity silicon
Silicon solar cell technology initially benefited directly from the economies of scale
of the silicon microelectronics industry. Silicon cells were also once able to make
use of scrap material from this industry, since the material quality requirements
in photovoltaics are less demanding than in the microelectronics field. However,
since 2006, the photovoltaic industry has used a larger volume of silicon than
microelectronics, changing forever the dynamics of pure silicon supply.
For microelectronics, the starting point for producing the required high-quality
‘semiconductor grade’ silicon is a lower grade of silicon known as ‘metallurgical
grade’, produced by the reduction in an arc furnace of quartzite by carbon, the latter
generally in the form of wood chips. This metallurgical grade silicon is about 98%
pure and is produced in large quantities for the steel and aluminium industries.
A relatively small quantity is refined for microelectronics and photovoltaics by
conversion to a volatile intermediary that can be purified by fractional distillation.
The purified intermediate compound is then decomposed to re-extract the silicon
in a similarly highly purified form.
Generally the metallurgical grade silicon is converted by hydrochloric acid
to trichlorosilane, which is then purified to 99.9999999% (nine ‘nines’) purity by
fractional distillation. Silicon is then extracted from the trichlorosilane by reducing
the latter by hydrogen at high temperature. In this process electrically heated silicon
rods are exposed to a trichlorosilane/hydrogen mixture that reacts on the surface
of the rods, depositing silicon onto them and building up their cross section. These
rods grow with a fine grain polycrystalline silicon microstructure. After the rod
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 106

106 M. A. Green

diameter has increased to the required size, the process is stopped and the rods
mechanically broken into smaller chunks, that maintain ‘nine-nines’ purity. These
chunks then become the starting point for the growth of ingots of good crystalline
quality.
In the 1970s and early 1980s, several other options for preparing silicon feed-
stock were investigated as part of a large US government photovoltaic program
supported by the Carter administration (Christensen, 1985). Approaches investi-
gated ranged from those involving radically different techniques to those exploring
only minor changes from the sequence outlined above, such as the use of different
compounds of silicon as the intermediate during the purification process. One such
process, based on the use of silane as the intermediate (Christensen, 1985), is now
used commercially. Decomposition of the intermediary gas in fluidised bed reac-
tor was also studied in this program and is now also used commercially, greatly
reducing process energy requirements. Other projects were based on the use of
upgraded metallurgical grade (UMG) silicon, a line of investigation still attracting
great attention given its appeal in circumventing the complexity of the traditional
purification approach.

3.3.2 Czochralski growth

As previously mentioned, crystalline ingots have traditionally been prepared by the
Czochralski (CZ) crystal growth process. In this process, the purified silicon chunks
are melted in a quartz crucible along with small pieces of silicon heavily doped with
boron. This produces a boron-doped melt into which a seed crystal is dipped and
slowly withdrawn (Fig. 3.11a). For high-quality crystal growth, good temperature
uniformity and slow and steady growth are required. Typically ingots are grown
to about 12.5–20 cm in diameter and 1–2 metres in length, weighing 100–200 kg.
The crystallographic orientation of the seed is transferred to the grown crystal.
Generally, for photovoltaics, the crystal is grown with a preferred orientation so
that the wafers sliced from the crystal perpendicular to the growth axis have surfaces
parallel to {100} crystallographic planes.
Prior to slicing these ingots into wafers, they were once subject to a centreless
grinding operation to remove the slight fluctuations in diameter along the length
of the ingot that occur during crystal growth. More commonly, the ingots are now
‘squared-off’ by sawing off large sections parallel to the growth axis (Fig. 3.11b),
giving ‘quasi-square’ wafers after wafering. The large pieces of silicon sawn off in
this approach are generally recycled by re-melting as feedstock for the CZ growth.
The technique traditionally used in microelectronics for sawing wafers from
ingots was based on the use of inner diameter saws. In this technique (Fig. 3.12a),
thin metal sheet blades are given dimensional solidity by being held in tension
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 107

Crystalline Silicon Solar Cells 107

seed

crystal

(a)

sawn crystal
(recycled)
wafers

(b)
Figure 3.11 (a) Czochralski (CZ) growth; (b) squared-off CZ ingot. Source: Green and
Hansen (1998).

around their outer perimeter. The cutting surface is a diamond impregnated edge
surrounding a hole within the tensioned metal sheet. This technique gives excellent
dimensional tolerance, although there are limitations arising from the thickness of
the silicon wafers that is possible to produce while still maintaining high yield.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 108

108 M. A. Green

Inner-diameter
saw blade

ingot

to spool
wafer

(a)

ingot
guide

(b)
Figure 3.12 (a) Inner diameter wafer sawing; (b) continuous wire sawing. After Dietl et al.,
1981.

Other limitations arise from the wastage of silicon as ‘kerf’ loss during cutting.
Generally, about 10–15 wafers per centimetre of ingot length were achieved by
this process.
An alternative technique now almost universally used in photovoltaics is based
on wire sawing (Fig. 3.12b). In this case, tensioned wire is used to guide an abrasive
slurry through the ingot (Schumann et al., 2009). Advantages are thinner wafers
and less surface damage for these wafers as well as lower kerf or cutting loss,
initially allowing the sawing of over 20 wafers per centimetre. With subsequent
reduction in wafer thickness to 160–180 microns in 2011 and of the diameter of
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 109

Crystalline Silicon Solar Cells 109

the wire to the 100–120 micron range, this figure has increased to over 30 wafers
per centimetre, with up to 50 wafers per centimetre of ingot length expected over
the coming decade.
There has been a recent resurgence of interest in an alternative wire sawing
approach where the wire is similarly thin, but impregnated with diamond particles
(Schmidt, 2011). This approach has the advantage of much higher cutting speed as
well as reduced kerf loss. The process also offers more potential for the recycling of
the silicon that would otherwise be wasted as kerf loss during the sawing process.
An alternative to the standard Czochralski process for producing crystalline
ingots is the floatzone (FZ) process. Although some studies have predicted superior
economics when compared to the Czochralski process for cell production due
to the elimination of consumables such as quartz crucibles, the FZ process, as
commercially implemented, is capable of accepting feedstocks only in the form of
high-quality cylindrical rods. This makes it unsuitable for using low-cost source
material.
Early interest in producing monocrystalline silicon by a directional solidifi-
cation process (Schmidt, 2011) has been recently revived and will be discussed
in the next section. Unlike the FZ process, this approach is very tolerant of low-
grade source material (Khattak et al., 1981), similar to the directional solidification
approaches to producing multicrystalline silicon, also to be discussed in the next
section.

3.3.3 Multicrystalline silicon ingots

Over recent years, most of the world’s photovoltaic production has been based
on multicrystalline silicon wafers. Several companies have developed commercial
processes for producing the precursor multicrystalline silicon ingots (Ferrazza,
1995; Rodriguez et al., 2011). Advantages over the Czochralski process are lower
capital costs, higher throughput and a higher tolerance to poor feedstock quality.
Basically, the technique involves controllably solidifying molten silicon in a
suitable container (crucible) to give silicon ingots with large columnar grains, which
generally grow from the bottom of the crucible upwards (Fig. 3.13a). Pioneers of
this approach for modern photovoltaics in the mid-1970s were Wacker Chemitronic
of Germany (Authier, 1978) and Solarex of the USA (Lindmayer, 1976). In the
1980s, several other manufacturers developed similar processes capable of produc-
ing good-quality multicrystalline material. These manufacturers differed in their
choices of crucible material, the method of loading the crucible with silicon and
the method for controlling the cooling of the melt. A good summary can be found
elsewhere (Ferrazza, 1995). The size of nominally rectilinear ingots can be very
large, presently up to 100 cm × 100 cm × 30 cm, with each ingot weighing over half
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 110

110 M. A. Green

silicon

mould
(a)

(b)
Figure 3.13 (a) Directional solidification of silicon within a mould; (b) sawing of large
ingot into smaller sub-sections. Source: Green and Hansen (1998).

a tonne (Rodriguez et al., 2011). This weight is expected to double over the coming
decade (SEMI, 2011).
The large ingots are sawn into smaller sections as shown in Fig. 3.13b, which
eventually give wafers generally 15.6–20 cm along the sides. These smaller sec-
tions can be sawn by the standard continuous wire sawing processes. The resulting
multicrystalline wafers are capable of producing cells of over 85% of the perfor-
mance of a monocrystalline cell fabricated on a CZ wafer. However, because of
the higher packing density possible due to their square or rectangular geometry,
this performance difference is largely masked at the module level, with multicrys-
talline module performance lying in the range demonstrated by modules made
from monocrystalline cells (Haase and Podewils, 2011).
Interest in producing monocrystalline rather than multicrystalline silicon
ingots by such directional solidification approaches dates back to the 1970s
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 111

Crystalline Silicon Solar Cells 111

(Schmidt, 2011; Fischer and Pschunder, 1976). However, developmental empha-

sis was diverted to the multicrystalline approach. More recently, BP Solar have
revived interest in the monocrystalline approach (Stoddard et al., 2008). Furnace
deliveries based on the monocrystalline approach began in 2011 (AMG, 2010). The
monocrystalline directional solidification approach involves seeding the bottom of
the melt of Fig. 3.13a with a crystalline template. Close control of temperature is
required to allow melt to contact this seed without consuming it.
Eventually, the whole industry may convert to this and related approaches,
given the ability to bridge the gap in cell performance between directionally solid-
ified and CZ silicon. Eventually it may be possible to produce material similar
to FZ quality, given the better control over oxygen in a directional solidification
process compared with silicon CZ growth. The process is also likely to be very tol-
erant of low-grade silicon (Khattak et al., 1981) and to provide greater purification
during directional solidification than is possible with multicrystalline ingots, pos-
sibly decreasing the purity requirements for ‘solar-grade’ feedstock. The renewed
interest in this approach has stimulated a range of other ‘controlled crystallisation’
approaches that have led to steady improvements in the quality of multicrystalline
wafers produced commercially.

3.3.4 Sheet and ribbon silicon

Although there is the potential for substantial cost reductions in both the cost
of preparing the silicon feedstock and in forming crystalline or multicrystalline
ingots from it, one unavoidable cost with the silicon wafer approach is the cost
of sawing the ingot into wafers. Several studies have suggested that the lower
bound on this cost may be something of the order of US$0.20/Watt (Christensen,
1985; Bruton et al., 1997; Song et al., 2010). This has provided the rationale for
investigating approaches that produce silicon directly in the form of self-supporting
sheets without the need for sawing (Bergin, 1980; Schulz and Sirtl, 1984; Rodriguez
et al., 2011).
The first ribbon approach to be used commercially was based on the edge-
defined film-fed growth (EFG) method of Fig. 3.14. As originally investigated
in the early 1970s, this technique involved the pulling of a thin sheet of silicon
ribbon from a strip of molten silicon formed by capillary action at the top of a
graphite dye (Fig. 3.14a). Substantially higher throughput was obtained with the
more symmetrical configuration shown in Fig. 3.14b, where the ribbon is pulled in
the form of a hollow nonagon. Individual wafers are then cut from the sides of the
nonagon, normally by laser scribing wafers from each of the sides. The material
produced is multicrystalline with elongated grains and is of a similar quality to the
standard directionally solidified multicrystalline material. Commercial cells made
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 112

112 M. A. Green

Figure 3.14 (a) Edge-defined, film-fed growth (EFG) method; (b) growth of a nonagonal
ribbon of silicon using the EFG method. Source: Green and Hansen (1998).

from this material have been available sporadically since the early 1980s. However,
after building up to a manufacturing capacity of about 200 MW/year, a decision
was made to cease production in 2009 since ‘ingot technology has now proven to
be a more economical process’ (Schott, 2009).
An even older ribbon growth process is the dendritic web approach of Fig. 3.15,
first described by Westinghouse in the 1960s. In this approach, close thermal control
is used to cause two dendrites spaced several centimetres from each other to solidify
first during the growth step. When these are drawn from the melt, a thin sheet of
molten silicon is trapped between them. This quickly solidifies to form a ribbon.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 113

Crystalline Silicon Solar Cells 113

silicon dendrites
or carbon
string

molten
silicon

Figure 3.15 Schematic illustrating the dendritic web growth process and the string ribbon
approach. Source: Green and Ho (2011).

A somewhat related approach is the string ribbon approach. In this case, the
molten silicon is trapped between two graphite strings that are drawn from the
melt. This relaxes the requirement on thermal control, compared with the previous
dendritic web approach. The string ribbon approach has been commercialised by
Evergreen Solar (Janoch et al., 1997; Wallace et al., 1997; Rodriguez et al., 2011).
Evergreen produced an estimated 157 MW of modules using this technology during
2010 while a formerly related company, Sovello, produced an additional 145 MW
(Hering, 2011).
Despite the obvious attractions of the ribbon approaches, market share steadily
declined, representing only 1.2% of production in 2010 after peaking at 5.6% in
2001 (Hering, 2011), with production now phased out. This highlights the difficulty
of establishing unique technology in an industry where mainstream manufacturers
have access to a skill and resource base far larger than that of any individual
manufacturer.

3.4 Cell processing

3.4.1 Standard process
In the previous section, standard and non-standard ways of forming the silicon
substrate were described. Most commercial substrates are formed by the wafering
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 114

114 M. A. Green

of monocrystalline and multicrystalline ingots, with a much smaller quantity of

ribbon substrates produced commercially.
Some ‘vertically integrated’ photovoltaic manufacturers prepare their own
polysilicon source material, although many bypass this and the crystal growth
step by purchasing silicon wafers. Processing begins by chemically cleaning the
starting wafers and etching their surfaces, to remove saw damage and to create
surface texture (Neuhaus and Münzer, 2007).
For monocrystalline wafers, alkaline sodium or potassium hydroxide solutions
are generally used for both steps, with a more dilute solution used for texturing.
The composition and temperature of this solution determines the texturing quality
(King and Buck, 1991), including the size of the pyramidal features resulting from
the texturing and the percentage of wafer surface area successfully covered by
such features. For multicrystalline wafers, an acidic etchant produces an effective
surface texture (Einhaus et al., 1997; Neuhaus and Münzer, 2007).
The next major stage of processing is the diffusion of the cell junction. This
is generally achieved by heating at high temperature in a tube in a furnace down
which a gas containing POCl3 is passed (Fig. 3.16). A phosphorus glass is formed
on the cell surface, allowing phosphorus dopant atoms to seep into the cell surface
by thermal diffusion. Typically, the depth of diffusion is less than 1 µm. The same
thermal diffusion approach is widely used in microelectronics. Processing for pho-
tovoltaics can involve cruder equipment and techniques, with ‘in-line’ processing
on metal belt furnaces re-emerging as a high-throughput option, with spraying or
prior spinning of a compound containing phosphorus onto the wafer surface.
After diffusion, the diffusion oxide is removed by etching. Although the dif-
fusion is required over only one surface of the wafer and processing techniques
are generally chosen to encourage such a result, phosphorus invariably seeps into
both wafer surfaces to some extent.

02 + N2
Silicon wafers

Carrier
gas

Quartz tube within furnace

Liquid POCl3

Figure 3.16 Phosphorus diffusions process. Source: Green and Hansen (1998).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 115

Crystalline Silicon Solar Cells 115

To break the connection between the phosphorus diffused into front and rear
surfaces, an ‘edge junction isolation’ step is needed to remove the thin phosphorus
layer around the edge of the wafer. This isolation was once achieved by ‘coin
stacking’ the wafers so that only their edges were exposed. The stack was then
placed in a plasma etcher to remove a small section of silicon from the wafer edge,
hence breaking the conductive link between front and rear surfaces. Single-sided
etching is now the preferred approach in which the wafer sits on the etching solution
so that only the desired surface is etched.
A silicon nitride quarter-wave antireflection coating is applied to the cell at
this stage. Not only does silicon nitride have a reasonably good refractive index
for this application, it is deposited using silane (SiH4 ) and consequently has a high
hydrogen content. This hydrogen can be released into the silicon substrate, giving
a beneficial defect passivating effect, particularly for multicrystalline cells.
Cell processing is completed by the screen printing of metal contacts onto the
front and rear surface. Silver and aluminium paste (consisting of a suspension of
fine metal particles and glass frit in an organic medium together with appropriate
binders) is squeezed through a patterned screening mesh onto the cell surface. After
application, the paste is dried at low temperature. The Ag-based paste for the top
surface is printed in a characteristic finger pattern shown in Fig. 3.17 to minimise
the resistive losses in the cell while allowing as much light as possible into it. For
the rear, two different pastes are generally applied as shown. An Ag-rich paste is
applied to a small fraction of the rear to allow soldering of the cell interconnections,
while most of the rear is covered by an Al-based paste. After these three layers

front
B
rear

Figure 3.17 Standard 156 mm × 156 mm solar cell ‘H’ metallisation pattern. Screen-printed
silver regions are shown coloured black. The pattern consists of six unit cells of size A × B.
Both front and rear views are shown. Source: Green and Ho (2011).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 116

116 M. A. Green

have been individually screened and dried, the pastes are then fired at a higher
temperature to drive off the remaining organics and to allow the metal particles
in the paste to coalesce. The glass frit is important in promoting adhesion to the
silicon substrate. Often pastes are doped with dopants such as phosphorus, in this
case to help prevent the screened contact from penetrating the thin phosphorus skin
that it is intended to contact.
This screen-printing method for applying the metal contact was borrowed
in the early 1970s from the hybrid microelectronics industry (Ralph, 1975). This
ensured the ready availability of both screen-printing equipment and the paste firing
furnaces suited to this application. Labour and equipment costs associated with this
step tend to be very low. However, the pastes themselves can be expensive (Green,
2011) and an even larger cost penalty is paid for the simplicity of this approach by
the forfeiture of the inherently available power output from the silicon wafer, as
discussed later.
After the paste firing, the cells are then ready for testing under a solar simulator.
Cells are usually graded based on their short-circuit current or current at a nominal
operating voltage, e.g. 500 mV. Generally, cells are sorted into 5% performance
bins. This sorting is required to reduce the amount of mismatch between cells
within the completed module. To a large extent, the output current of the module is
determined by that of the worst cell in the module, resulting in large power losses
within mismatched modules. Even worse, low-output cells can become reverse-
biased under some modes of module operation and destroy the module by localised
over-heating.
Silicon ribbon substrates often require modifications of the above standard
sequence. For example, the rippled surface which is a natural consequence of
the EFG ribbon growth process poses continuity hazards for screen-printed met-
allisation. To accommodate this rough surface, a novel technique was developed
whereby the metal paste was squeezed through an orifice and then dropped to the
cell surface, much the same as squeezing toothpaste from its tube onto a toothbrush.

3.4.2 Limitations of the screen-printing approach

There are four main limitations arising from the screen-printing approach to apply-
ing the front contact, causing the simplicity in processing to be at the expense of
cell performance. One limitation is that the top-surface phosphorus diffusion has to
be heavier than desirable purely from the point of view of cell performance. This is
to allow reliable low-resistance contact between the screen-printed metal and the
diffusion. Typically, sheet resistivities of this diffusion of less than 60 ohms/square
have been required (Green, 1995; De Clercq et al., 1997). Such diffusions gen-
erally reduce the quality of the silicon in the region near the cell surface, where
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 117

Crystalline Silicon Solar Cells 117

blue wavelengths in sunlight are strongly absorbed. A screen-printed cell therefore

does not respond well to blue wavelengths in sunlight, wasting up to 10% of the
possible current output through this deficiency.
The remaining three limitations relate to the geometry and conductivity of
the metal lines it is possible to produce by the standard screen-printing process.
Since the paste thickness shrinks to about one-third of its original thickness during
firing, it is very difficult to achieve metal lines with high aspect ratio (height/width),
which are the key to forming metal grids which result in low overall losses (Serreze,
1978). The nature of the screening meshes that have sufficient ruggedness for use in
commercial production means it is very difficult to achieve fine lines using screen
printing in production. Typically, 70 µm width is the minimum that can presently
be achieved cost-effectively. This limitation means that there will generally be
high shading losses in screen-printed cells owing to the large percentage (∼10%)
coverage of the front surface by the metal (Fig. 3.17). Additionally, the relatively
poor conductivity of the fired silver paste (about two times lower than that of pure
silver for large features such as busbars but up to six times lower for finer features
such as fingers (de Moor et al., 1997)) fundamentally limits the ability to optimise
metal contact design in much the same way as does the low aspect ratio previously
discussed.
Paste manufacturers are working hard to develop improved pastes that
reduce the impact of these deficiencies. Figure 3.18 shows the targeted prod-
uct introduction of one major paste manufacturer. This manufacturer anticipates
ongoing contributions to improving cell efficiency by developing pastes able to
contact increasingly lightly diffused top surface diffusions. Practical diffusion
sheet resistivity is shown as increasing from 70 ohms/square in 2010 to above
100 ohms/square by 2015. Pb-free pastes are also eventually expected to be used
across the industry.
Screen-printing equipment manufacturers have also not been idle. Modern
machines are now able to align each screen-printing step very accurately to fea-
tures on the wafer, including previously printed patterns. This makes it possible to
incorporate new design features into cells.
Figure 3.19 shows two such features of present interest. Using the alignment
capability previously mentioned, the area under the metal contact can be selectively
doped very heavily to allow good contact (Fig. 3.19a), allowing other regions of
the surface to be lightly doped, improving the cell blue response (Chunduri, 2010).
About 1% absolute gain in performance is possible (e.g. from 18% to 19%) using
this ‘selective emitter’ approach. Figure 3.19b shows another possible improvement
by ‘double printing’. By printing the metallisation twice, high aspect ratios are
possible, reducing metallisation loss. Additionally, the metal layer contacting the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 118

118 M. A. Green

120
PV17A
19 PV15A 110

Emitter, ohms/ square

PV145/7
PV142/9 100
SiNx PV156/7 Pb free
Efficiency, %

18 fire- 90
through PV173
80
TiO2
17 70
SiO2 ohms/sq.
AR 60
coat
16 50
40
15 30
1997 2000 2003 2006 2009 2012 2015

Figure 3.18 Past and planned product introduction schedule for one major paste supplier.
Shown are the different generations of paste and the emitter sheet resistivity the pastes are
designed to accommodate. The author’s estimate of likely production efficiency is based
on unscaled estimates provided by the paste manufacturer. Adapted from Laudisio et al.
(2010).

finger layer 2

layer 1

n-type n-type
n+

p-type p-type

p+ p+

rear contact
(a) (b)
Figure 3.19 (a) ‘Selective emitter’ approach; (b) ‘double printing’ to increase aspect ratio
of metal fingers. Source: Green and Ho (2011).

silicon can be optimised for this role while the second layer can be optimised for
good conductivity.
Recent rapid increases in silver prices cast doubt on the long-term role of this
metal in photovoltaics (Green, 2011). The industry seems likely to transition to
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 119

Crystalline Silicon Solar Cells 119

the use of Cu as the main conductive metal in the cell. Plating Cu to thin screen-
printed Ag lines is one possible approach, or it may be that the need for change
will stimulate the adoption of more innovative technology.

3.4.3 Laser-processed solar cells

As mentioned in Section 3.2, buried contact cells were developed as a way of
incorporating some of the efficiency improvements demonstrated in the mid-1980s
into low-cost commercial cell production sequences. This was the first use of lasers
to define fine-linewidth features in silicon cell processing.
Since that time, two other laser-based cell designs have been used either in
production or in large-scale pilot production. The first is the ‘semiconductor finger’
cell design of Fig. 3.20 which is, to some extent, a hybrid between a buried contact
and a conventional screen-printed cell. A heavily doped finger layer is formed by
laser doping in the top surface, similar to the grooves of a buried contact solar cell or
to the selective emitter regions of the laser-doped selective emitter (LDSE) cell to be
discussed below. A screen-printed Ag contact layer is then applied perpendicularly
to this semiconductor finger layer as shown. The advantages are that a light diffusion
can be used over most of the cell surface and the shading by the Ag fingers is
significantly reduced. On large area CZ cells, efficiencies in the 18–19% range
have been reported.
A second laser-based cell design is the LDSE cell, with the key processing step
shown in Fig. 3.21 (Wenham and Green, 2002). A laser is used to define the areas

bushbar metal semiconductor

finger AR coat
finger

n+
n++ p

p+

metal

Figure 3.20 Semiconductor finger solar cell. Heavily doped semiconductor regions provide
an additional finger layer that is contacted by the traditional metal finger layer. Source: Green
and Ho (2011).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 120

120 M. A. Green

Figure 3.21 Laser-doped selective emitter (LDSE) solar cell. Metal is selectively plated
to the heavily doped regions exposed during the laser doping step. Source: Green and Ho
(2011).

for top finger contact by removing the dielectric coating by melting the underlying
silicon, while simultaneously introducing dopants into the molten region. The front
contacts can then be plated as a Ni/Cu/Ag tri-layer. The cell has the advantages
of a lightly doped top surface region and of fine, highly conductive metal fingers.
Efficiency in the 19–20% range has been confirmed for cells on large area CZ
wafers fabricated using largely commercial equipment. This is expected to improve
to the 21–22% range once a similar contacting approach is applied to the rear of
the cell.

3.4.4 HIT solar cell

The HIT cell has already been discussed in Section 3.2.3 (Fig. 3.8). Sanyo intro-
duced this cell into commercial production in 1997, and both efficiency and produc-
tion volume have increased steadily since then. In 2010, an estimated 300 MW of
HIT cell modules were produced (Hering, 2011) with total area module efficiency
in the 16.1–19.0% range, well above the market average of about 14% (Haase and
Podewils, 2011). The distinctive feature of the technology is the use of amorphous
silicon (a-Si) to form heterojunctions with the n-type silicon substrate at both front
and rear contacts, as previously discussed.
With the expiry of the most important of Sanyo’s initial patents in 2010,
equipment suppliers began offering turn-key manufacturing facilities for these cells
(Strahm et al., 2010).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 121

Crystalline Silicon Solar Cells 121

3.4.5 Rear junction cells

In earlier sections, three different rear-junction devices were described: the rear-
junction cells of Figs. 3.6 and 3.9 and the early Bell Laboratories cell of 1954
(Fig. 3.2c). The latter could be described as an ‘emitter wraparound’ cell, a member
of a more general group of ‘emitter wrap through’ (EWT) cells that, like the rear-
junction cells, require no metal on the front surface. A related rear-contact device
is the ‘metal wrap through’ (MWT) device that has metal on both front and rear
surfaces. Figure 3.22 shows the key features of EWT and MWT devices.
SunPower introduced their rear-junction cell (Fig. 3.9) into production in 2003.
This was based on the high-efficiency devices earlier developed at Stanford Univer-
sity (Fig. 3.6) but with simplifications that made the device suitable for commercial

p+
n+

p+ p+

(a)

p+
n+

p+ p+

(b)
Figure 3.22 (a) Emitter wrap-through (EWT) solar cell; (b) metal wrap-through (MWT)
cell. Source: Green and Ho (2011).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 122

122 M. A. Green

production. This was the first commercial cell of efficiency above 20%. In 2008,
SunPower introduced an improved ‘Generation II’ product which took efficiency
to 22% (De Ceuster et al., 2007). This has produced commercial modules in large
volume with 18.4–19.7% total area efficiency, the highest on the market (Haase
and Podewils, 2011). More recently, a Generation III product has been reported
(Cousins, 2010) with 24% cell efficiency that is likely to take commercial module
efficiency beyond 20% for the first time. Cell processing costs are reported to be
higher than for the standard screen-printing approach (Song et al., 2010).
Commercialisation of EWT cells is not as far developed as the rear-junction
cell or even the MWT device. A key disadvantage of both compared to the rear-
junction device is the need to drill multiple holes through the wafer using a laser,
although fewer are required for MWT devices. However, both EWT and MWT
cells are more tolerant of low-quality wafers than rear-junction devices. In fact,
the first solar modules of over 17% efficiency employing multicrystalline wafers
used MWT cells (Lamers et al., 2011). In 2010, at least two manufacturers offered
MWT devices commercially (Solland, 2009; Photovoltech, 2010), although man-
ufacturing volume was not large.

3.5 Cell costs

There have been many studies of the costs of the different stages of silicon cell
production using different basic assumptions, particularly in relation to the produc-
tion volume assumed in the study and the cost of polysilicon source material. One
of the most authoritative early studies was one conducted under the auspices of
the European Union Photovoltaic Programme (Bruton et al., 1997; Bruton, 2002).
This study involved representatives of major European photovoltaic manufacturers
and research laboratories, and was valuable owing to the breadth of representation
and the diversity of approaches explored.
The key assumptions of the study were a manufacturing volume of 500 MWp
of solar cells per annum and the availability of silicon source material at US$25/kg.
The former seemed like an enormous production volume at the time when large
manufacturers were producing only 10–20 MW/year of modules. However, a man-
ufacturer producing 500 MWp/year of modules in 2010 would barely qualify for the
list of ‘Top 20 Manufacturers’ (Hering, 2011). A number of different technologies
were compared in the study. Important comparisons were between EFG ribbon,
multicrystalline and crystalline wafer technologies, between screen-printed, buried
contact and PERL cell processing sequences, in various combinations of wafers
and processing, and between two different module encapsulation approaches.
However, the results from all possible combinations were not studied (or, at
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 123

Crystalline Silicon Solar Cells 123

Table 3.1 Summary of published results of a European Commission study of manufac-

turing costs for 500 MWp/annum factory (Bruton et al., 1997).

ID Wafer Process Cell efficiency Estimated cost ( /Wp) Key variable

#1 DS SP 15% 0.91 wafer

#2 CZ SP 16% 1.25 wafer/process
#3 CZ LGBC 18% 1.15 process
#6 CZ PERL 20% 1.78 process
#7 EFG SP 14.4% 0.71 wafer

1 = US$1.2136 at time of study; 1 = US$1.44 in August 2011.

DS: Directional solidification
CZ: Czochralski growth
EFG: Edge-defined film-fed growth
SP: Screen-printed
LGBC: Laser-grooved, buried contact
PERL: Passivated emitter, rear locally diffused (acronym LBSF used in study)

least, not published), but only seven selected combinations, of which Table 3.1
shows five.
From this table, several key results can be deduced. When comparing screen-
printed cells on ribbon (EFG), multicrystalline (DS) and monocrystalline (CZ)
wafers, the ribbon produced the lowest cost of 0.71/Wp followed by the mul-
ticrystalline wafers at 0.91/Wp and the monocrystalline wafers at 1.25/Wp. The
advantage of the ribbon stemmed almost entirely from the fact that it does not need
to be sawn, as previously mentioned. As noted earlier, EFG production ceased in
2009 due to unfavourable economics. This is probably as a result of the relatively
small effort that could be devoted to the development of this technology by a single
company compared with the huge effort devoted by almost the entire industry to
the directional solidification approach.
Comparing the different processing approaches on single-crystal wafers, the
cheapest is the buried contact at 1.15/Wp, followed by the screen-printed at
1.25/Wp, followed by the PERL at 1.78/Wp. The buried contact achieved its
cost advantage over the screen-printing approach by virtue of the increased effi-
ciency giving more power per unit processing area. In terms of module costs, the
standard laminated module approach was calculated to be slightly cheaper than an
alternative resin-fill approach.
At the time these results were published, many regarded the costs deduced
as impossibly ambitious. However, actual developments have shown them to be
conservative. In 2011, average module manufacturing cost across the industry was
estimated as US$1.56/Wp (Bolman et al., 2011), or about 1.08/Wp, with the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 124

124 M. A. Green

lowest cost silicon manufacturers producing modules for around US$1.20/Wp

( 0.83/Wp). Further significant cost reductions are expected over the coming
decade (SEMI, 2011). A rough breakdown of the 2011 costs was that the cost
of the pure silicon accounted for 20–25% of the total module cost, while the cost
of converting this to a wafer accounted for a similar percentage of total costs, as did
the cost of converting to a cell. The cost of encapsulating the cell into the module
accounted for a slightly larger 25–30% of total costs (Bolman et al., 2011).

3.6 Opportunities for improvement

3.6.1 Commercial cells

The large differential between the efficiencies of a typical screen-printed com-
mercial cell (17–18%) and the best laboratory silicon cell (25%) shows the large
potential for further efficiency improvement in commercial devices.
One reason for this differential is the difference between the CZ wafers used
in commercial production and the FZ wafers until recently used for the best labora-
tory cells. CZ-grown wafers are invariably contaminated with oxygen and carbon
during growth to a much greater extent than FZ wafers, due to the use of quartz cru-
cibles and graphite heaters in the CZ process. These impurities give rise to a much
more subtle dependence on processing conditions of the important silicon mate-
rial property for producing high performance cells, the minority-carrier diffusion
length. For example, applying the high-temperature processing associated with
PERL-type sequences to CZ silicon gives a large spread in results depending upon
the supplier of the CZ material, and hence most likely on the oxygen and carbon
content (Knobloch et al., 1996). Additionally, the quality of B-doped CZ material
as compared to FZ falls off quite rapidly as the boron content is increased, which
is attributed to boron/oxygen complexes forming within the material (Schmidt and
Bothe, 2004). This reduces flexibility in cell design since it eliminates the possi-
bility of using low-resistivity CZ substrates.
Two recent developments may change this situation. One is the progress in
improving the laboratory performance of cells on n-type substrates (Section 3.2.2).
Some manufacturers have now switched to the use of n-type wafers for monocrys-
talline cells, with this expected by some to be an increasing trend over the coming
decade (SEMI, 2011). Phosphorus is the dopant in these n-type substrates and
does not form the same performance-reducing defects with oxygen as does boron.
A second change is the growing interest in producing monocrystalline silicon by
direction solidification (Section 3.3.2). Better control of oxygen content may be
possible in crystalline material prepared this way.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 125

Crystalline Silicon Solar Cells 125

The trend towards thinner cells, which arises mainly from efforts to reduce the
costs of the silicon wafer, may actually help to improve the cell efficiency, as already
noted. Thin wafers give the opportunity for back-surface fields or other rear-surface
passivation approaches to be used to improve cell performance, primarily through
increased voltage output. Given better feedstock material or cells below 150 µm
thickness, improved rear-surface passivation approaches such as demonstrated by
the PERC and PERL cells of Fig. 3.7, as well as rear-junction and HIT cells, will
become increasingly relevant.

3.6.2 Laboratory cells

For laboratory cells, as previously mentioned, an appropriate reference point for
performance is the AM1.5 detailed-balance or Shockley–Queisser efficiency limit
of 33.5% for material of the bandgap of silicon. However, it has been shown that
another intrinsic process, Auger recombination, provides a more severe funda-
mental limit for silicon than the radiative recombination processes assumed in
the detailed-balance limit (Green, 1984; Tiedje et al., 1984). Unlike the detailed-
balance limit, the Auger limit for a silicon cell is dependent on the cell thickness,
as shown in Fig. 3.10 (Green, 1995). This difference arises because the detailed-
balance calculation includes photon recycling, which makes nett recombination
rates independent of cell volume. With Lambertian light trapping, the optimum
cell performance in the Auger limit is 29% for a cell of about 80 µm thickness.
Such a cell would have an open-circuit voltage of above 760 mV, higher than the
highest value ever demonstrated for silicon of 750 mV for a HIT cell, in which
surfaces are passivated by a-Si layers.
Figure 3.23 shows the results of efficiency calculations with various amounts
of surface recombination added, characterised in terms of the open-circuit voltage
limit that this recombination would impose if it were the only recombination pro-
cess in the cell (Green, 1999). Increasing surface recombination reduces the value
of the obtainable efficiency as well as pushing the optimum cell thickness to larger
values. Figure 3.23 makes it clear that, to improve silicon cell efficiency much
beyond 26%, improved surface passivation (of both surfaces) is essential beyond
the 720 mV capability that presently limits all but HIT cells. If this improved quality
cannot be achieved, another possibility is to maintain the same quality of surface
passivation as presently achieved and reduce the effective threshold energy of the
photovoltaic process within the bulk regions of the cell. Techniques such as hav-
ing sections of the cell alloyed with germanium to reduce its bandgap (Healy and
Green, 1992) or doped with a photoactive impurity to give impurity photovoltaic
effects in the bulk region (Keevers and Green, 1994) have been suggested and
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 126

126 M. A. Green

30
no limit
680 mV
25
Efficiency (%)

720 mV

20 600 mV

640 mV

15
1 10 100 1000 10000
Thickness ( µ m)

Figure 3.23 Limiting efficiency of silicon cell with Lambertian light trapping as a function
of surface recombination velocity, characterised in terms of the voltage limit imposed by
this recombination. Source: Green (1999).

shown, in some cases, to have theoretical advantages. However, neither technique

has to date demonstrated an experimental performance advantage.
A well-proven approach for improving solar cell efficiency is the use of the
tandem cell structure. Efforts to produce tandem cells based on silicon have not yet
given good results owing to the inability to find a suitable wide-bandgap partner
that is lattice-matched to silicon (Corkish, 1991). For low-quality cells, amorphous
silicon/polycrystalline silicon tandems have given improved results over either cell
type alone (Yamamoto et al., 1997; Shah et al., 1997). However, the poorer blue
response of the a-Si cell and the inability to match the current output of a good-
quality c-Si substrate makes this an unlikely path to taking efficiency beyond 26%.
Fuller use of the available photon energy by incorporating efficient impact
ionisation process has been suggested as a way of boosting cell performance
by generating more than one electron–hole pair from one high-energy photon
(Kolodinski et al., 1993; Werner et al., 1994). However, such processes are quite
weak in silicon with increases in current density limited to less than 0.1 mA cm−2
(Green, 1987). Manipulating the details of silicon’s bandgap, for example by
alloying with germanium, may improve prospects. However, since the high-energy
photons of most interest for this process are absorbed very close to the surface of
silicon, such approaches may interfere with the ability to obtain well-passivated
surfaces. Limited experimental work with shallow germanium implants has not
given any nett performance benefit (Keevers et al., 1995).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 127

Crystalline Silicon Solar Cells 127

50.5%
47.5%
42.5% 45%
40

AM1.5G Efficiency
33%
30
29%

20

10

Free choice 0
0 1 2 3
Si bottom cell
Number of cells

Figure 3.24 Efficiency limits for cells and cell stacks with silicon as the lowermost cell
compared to the unconstrained limit for series-connected cells. While silicon is a good choice
for a single-cell material with an efficiency limit of 29% compared to the unconstrained limit
of 33%, it is an even better choice for the lowermost cell in 2- and 3-cell stacks. Source:
Green and Ho (2011).

The tandem cell approach perhaps remains the most likely option, pending
technical improvements that allow III–V or related cell technologies to be grown
on silicon. Silicon has an ideal bandgap to be the lowermost cell in a 2–4 cell
tandem (Fig. 3.24) and also can now be prepared very inexpensively in wafer
form. It would provide a clean, low-cost substrate for the subsequent growth of
high-performance cells. Low-cost deposition of the subsequent cell layers would
be essential. Progress has been reported in depositing lattice-matched III–V com-
pounds on silicon (Grassman et al., 2009), as well as fully-relaxed Ge layers on
silicon that might serve as templates for the subsequent growth of a wider range
of III–V materials (Wistey et al., 2007; Tsao, 2011).

3.7 Silicon-supported thin films

There has long been an interest in transferring the strengths demonstrated by crys-
talline silicon wafer technology to cells based on silicon thin films. Historically,
work can be divided into two phases: (i) that before the 1980s, when the benefits
of light trapping were not fully appreciated; and (ii) that after the mid-1980s, when
light trapping has been regarded as an essential feature of any silicon thin-film cell
design. The early work laboured under what is now known to be a misconception
that quite thick layers (>20 µm) of silicon would be required to give reasonable
performance due to the poor absorption characteristics of silicon arising from its
indirect bandgap (see Fig. 3.8). However, since light trapping can increase the effec-
tive optical thickness of a silicon cell by 10–50 times, this means that layers of only
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 128

128 M. A. Green

1 µm or so thickness are still inherently capable of producing similar performance

to much thicker layers (Fig. 3.10).
Approaches to producing supported silicon films can be divided into high-
temperature and low-temperature strategies depending on whether or not the sub-
strate is heated to high temperature during the silicon deposition or subsequent
processing.

3.7.1 High-temperature supported films

One of the earliest silicon-supported film approaches was the ‘silicon-on-ceramic’
approach (Christensen, 1985), whereby a ribbon of ceramic material was dipped
into a molten silicon bath or pulled across the surface of a silicon melt so that
one side was coated with silicon. This produced silicon of modest quality and the
approach suffered from difficulties in making rear contact to the cells, since the
ceramics used were insulating. This approach was discontinued in the early 1980s.
Early work by Ting and Shirley Chu involved the deposition of silicon onto a range
of foreign substrates by high-temperature chemical vapour deposition (Chu, 1977).
Operational cells were obtained using a number of substrate materials. The best
results were obtained by depositing the silicon layers on multicrystalline silicon
substrates prepared from metallurgical grade silicon. Given the previous studies
which have shown that sawing of wafers represents one of the major costs in any
wafer-type approach, the overall economics of such an approach using a wafer
substrate are questionable, regardless of the quality of this substrate. Other early
work involved the deposition of silicon onto ceramic substrates by high tempera-
ture CVD and the subsequent increase in crystal size by melting and directional
solidification (Minagawa et al., 1976).
In the post-1980 era, efforts in silicon-supported film were revitalised by the US
company AstroPower (Barnett et al., 1985). Initial work was directed at depositing
the films on ceramic-coated steel substrates. The use of the steel substrate was
soon dispensed with, followed by the use of the ceramic giving a columnar, free-
standing silicon sheet (Hall et al., 1994). A somewhat related RGS (ribbon growth
on substrate) approach (Lange and Schwirtlich, 1990; Hahn and Schönecker, 2004),
whereby the silicon layer is solidified on a re-usable substrate, is still under active
development.

3.7.2 Low-temperature approaches

One of the first papers addressing silicon photovoltaic thin films described the depo-
sition of silicon by low-temperature chemical vapour deposition onto an aluminium
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 129

Crystalline Silicon Solar Cells 129

substrate (Fang et al., 1974). A surprisingly large grain size was obtained, attributed
to eutectic reaction with the aluminium. In more recent times, laser crystallisa-
tion has been used in the active matrix liquid crystal display industry to produce
relatively small-grain polycrystalline silicon films from amorphous silicon pre-
cursors, generally deposited by low-pressure chemical vapour deposition. Grain
sizes are typically less than a micron or so, so that these films would probably
not be suitable for photovoltaics. Also, thicknesses for the active matrix display
industry tend to be only about 100 nm, which would be too thin for photovoltaic
application.
From 1989, a group at Sanyo explored the use of low-temperature solid phase
crystallisation of amorphous silicon as a technique for producing thin-film poly-
crystalline silicon cells. Good results were obtained, with 9.2% (unconfirmed) effi-
ciency reported in 1995 (Baba et al., 1995). These cells were approximately 1 cm2
in area deposited onto a textured metallic substrate and heated at approximately
600◦ C for many hours to enable the crystallisation of the originally amorphous
films. After crystallisation, the HIT structure developed by Sanyo (Fig. 3.8) was
used to complete the cell processing at low temperature.
A similar approach has been developed for depositing thin polycrystalline
films of silicon on glass (CSG), as described in more detail elsewhere (Green
et al., 2004). This produced small area modules of 10–11% efficiency (Keevers
et al., 2007). About 10 MW of 1.4 m2 modules have been produced commercially
with this approach.
Two groups pioneered the development of silicon thin films deposited directly
in microcrystalline form (µc-Si) onto glass substrates. In 1997, the University of
Neuchatel reported efficiencies of about 7% for 3 µm thick microcrystalline cells
deposited at 500◦ C (Shah et al., 1997). The cell had a p-i -n structure with the
intrinsic region comprising most of the device thickness. The cell was designed
for the intrinsic region to be depleted during normal device operation to create a
high electric field to aid carrier collection, as with a standard amorphous silicon
cell. Kaneka Corporation (Yamamoto et al., 1997) reported efficiencies of over
10% with a similar device structure (Fig. 3.25). Nearly the same efficiency has
been obtained with the total device thickness varied over the 1.5–3.5 µm range.
Both the above groups reported even higher efficiencies when amorphous sil-
icon cells were used in a tandem configuration on top of the microcrystalline
device (Fischer et al., 1997). Since this early work, the latter tandem device has
been commercialised by many groups with tandem a-Si:µc-Si cell modules of
typically 8–9% efficiency available from several sources (Haase and Podewils,
2011).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 130

130 M. A. Green

Ag

ITO

p µ c-Si

i poly-Si

n poly-Si

electrode

glass

Figure 3.25 Structure of 10.1% efficient thin-film microcrystalline solar cell developed by
Kaneka. Cell thickness is typically 1–3 µm. After Yamamoto et al. (1997).

3.8 Summary
Although crystalline silicon devices have dominated the commercial marketplace
for photovoltaics for more than three decades, there still remains scope for consider-
able improvement in both the performance and cost of these cells. Current expec-
tations are that manufacturing costs below US$0.50/Wp will soon be achieved,
without major changes in present processing sequences. The trend towards thin-
ner silicon wafers to decrease wafer cost is compatible with ongoing increases
in cell efficiency provided new cell structures are adopted and improved methods
are demonstrated in production for passivating the rear surface of the cell. The
large potential for both performance and cost reduction will make these bulk sili-
con approaches an increasingly challenging target for thin-film approaches. In this
context, good progress continues to be made with supported silicon film.
Ultimately, some way of boosting the performance of silicon cells would seem
to be required for these to continue to maintain their dominance over the coming
decades. The most promising at present would seem to be a tandem cell stack.
Such structures would take advantage of the enormous recent cost reductions in
producing silicon wafers, which could be used as clean substrates for the deposition
of thin, high-quality layers of strongly absorbing, wider-bandgap material. Such
an approach would combine the robust commercial infrastructure of silicon cell
manufacturing with the outstanding laboratory progress with tandem cell stacks.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 131

Crystalline Silicon Solar Cells 131

Acknowledgement
This work has been supported by the Australian Government through the Australian
Renewable Energy Agency (ARENA). Responsibility for the views, information
or advice expressed herein is not accepted by the Australian Government.

References
AMG (2010), ‘AMG acquires solar silicon casting technology from BP Solar International
Inc.’, Press Release, 22 December.
Authier B. (1978), ‘Poly-crystalline silicon with columnar structure’, Festkörper-probleme
XVIII, Freidr. Vieweg & Son, Wiesbaden.
Baba T., Shima M., Matsuyama T., Tsuge S., Wakisaka K. and Tsuda S. (1995), ‘9.2%
efficiency thin-film polycrystalline solar cell by a novel solid phase crystallisation
method’, Conf. Record, 13th European Photovoltaic Solar Energy Conf., Nice, October,
pp. 1708–1711.
Barnett A. M., Ford D. H., Checchi J. C., Culik S., Hall R. B., Jackson E. L., Kendall C. L.
and Rand J. A. (1997), ‘Very-large-area silicon-filmTM solar cells’, Conf. Proc., 14th
European Photovoltaic Solar Energy Conf., Barcelona, pp. 999–1002.
Barnett A. M., Hall R. B., Fardig D. A. and Culik J. S. (1985), ‘Silicon-film solar cells on
steel substrates’, Conf. Record, 18th IEEE Photovoltaic Specialists Conf., Las Vegas,
October, pp. 1094–1099.
Barnham K. W. J. and Duggan G. (1990), ‘A new approach to high-efficiency multi-band-gap
solar cells’, J. Appl. Phys. 67, 3490–3493.
Benick J., Hoex B., Dingemans G., Richter A., Hermle M. and Glunz S. (2009), ‘High
efficiency n-type silicon solar cells with front side boron emitter’, Proc. 24th European
PVSEC, September, Hamburg, 863–870.
Benick J., Hoex B., van de Sanden M. C. M., Kessels W. M. M., Schultz O. and Glunz S.
W. (2008), ‘High efficiency n-type Si solar cells on Al2 O3 -passivated boron emitters’,
Appl. Phys. Lett. 92, 253504.
Bergin D. O. (1980), ‘Shaped crystal growth — a selected bibliography’, J. Crystal Growth
50, 381–396.
Bolman C., Coffey V., Fu B., Song J., Trangucci R. and Zuboff G. (2011), ‘The true cost of
solar power: the pressure’s on’, Photon Consulting, www.photonconsulting.com.
BP Solar (1991), Data Sheet, BP Saturn Solar Cells.
Brandhorst H. W. and Bernatowicz D. T. (1980), ‘Space solar cells: high efficiency and
radiation damage’,Conf. Record, 14th IEEE Photovoltaic Specialists Conf., San Diego,
pp. 667–671.
Bruton T. M. (2002), ‘MUSIC FM five years on: fantasy or reality?’, Paper OA6.1, Proc.
PV in Europe, October, Rome.
Bruton T. M., Luthardt G., Rasch K.-D., Roy K., Dorrity I. A., Garrard B., Teale L., Alonso
J., Ugalde U., Declerq K., Nijs J., Szlufcik J., Rfauber A., Wettling W. and Vallera A.
(1997), ‘A study of the manufacture at 500 MWp p.a. of crystalline silicon photovoltaic
modules’, Conf. Record, 14th European Photovoltaic Solar Energy Conf., Barcelona,
pp. 11–16.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 132

132 M. A. Green

Chapin D. M., Fuller C. S. and Pearson G. O. (1954), ‘A new silicon p-n junction photocell
for converting solar radiation into electrical power’, J. Appl. Phys. 8, 676–677.
Christensen E. (1985, ed.), Final Report, Flat Plate Solar Array Project: 10 Years of
Progress, Jet Propulsion Laboratory, Pasadena, CA, JPL Publ. 400–279.
Chu T. L. (1977), ‘Silicon films on foreign substrates for solar cells’, J. Crystal Growth 39,
45–60.
Chunduri S. K. (2010), ‘New choices for ‘selective’ shoppers’, Photon International,
December, pp. 158–172.
Corkish R. (1991), ‘Some candidate materials for lattice-matched liquid phase epitaxial
growth on silicon’, Solar Cells 31, 537–548.
Cousins P. J., Smith D. D., Luan H. C., Manning J., Dennis T. D., Waldhauer A., Wilson
K. E., Harley G. and Mulligan G. P. (2010), ‘Gen III: improved performance at lower
cost’, 35th IEEE PVSC, June, Honolulu.
De Ceuster D., Cousins P., Rose D., Vicente D., Tipones P. and Mulligan W. (2007), ‘Low
cost, high volume production of >22% efficiency solar cells’, 22nd European Photo-
voltaic Solar Energy Conf., Milan, September, pp. 816–819.
De Clercq K., Frisson L., Szlufcik J., Nijs J. and Mertens R. (1997), ‘Design for man-
ufacturability of a silicon solar cell process based on screen printing by means of
manufacturing science techniques’, Conf. Proc., 14th European Photovoltaic Solar
Energy Conf., Barcelona, pp. 135–138.
De Moor H. H. C., Hoornstra J., Weeber A. W., Burgers A. R. and Sinke W. C. (1997), ‘Print-
ing high and fine metal lines using stencils’, Conf. Proc. 14th European Photovoltaic
Solar Energy Conf., Barcelona, pp. 404–407.
Dietl J., Helmreich D. and Sirtl E. (1981), “‘Solar” silicon’, in Crystals, Growth, Properties
and Applications, Vol. 5: Silicon, Springer-Verlag, pp. 43–107.
Einhaus R., Vazsonyi E., Szlufcik J., Nijs J. and Mertens R. (1997), ‘Isotropic texturing of
multicrystalline silicon wafers with acidic texturing solutions’, Conf. Proc., 26th IEEE
Photovoltaic Specialists Conf., pp. 167–170.
Fang P. H., Ephrath L. and Nowak W. B. (1974), ‘Polycrystalline silicon films on aluminium
sheets for solar cell application’, Appl. Phys. Lett. 25, 583.
Ferrazza F. (1995), ‘Growth and post-growth processes of multicrystalline silicon for pho-
tovoltaic use’, in Pizzini S., Strunk H. P. and Werner J. H. (eds), Polycrystalline Semi-
conductors IV — Physics, Chemistry and Technology, Solid State Phenomena” Trans.
Tech. Publ., Zug, Switzerland.
Fischer D., Keppner H., Kroll U., Torres P., Meier J., Platz R., Dubail S., Selvan J. A.
A., Vaucher N. P., Ziegler Y., Tscharner R., Hof C., Beck N., Goetz M., Pernet P.,
Goerlitzer M., Wyrsch N., Vuille J., Cuperus J. and Shah A. (1997), ‘Recent progress
of the ‘Micromorph’ tandem solar cells’, Conf. Proc., 14th European Photovoltaic
Solar Energy Conf., Barcelona, pp. 2347–2350.
Fischer H. and Pschunder W. (1976), ‘Low cost solar cells based on large area unconven-
tional silicon’, Conf. Record, 12th IEEE Photovoltaic Specialists Conf., Baton Rouge,
November, pp. 86–82.
Godfrey R. B. and Green M. A. (1980), ‘High efficiency silicon miniMIS solar cells —
design and experimental results’, IEEE Trans. Elec. Dev. ED-27, 737–745.
Godlewski M. P., Baraona C. R. and Brandhorst H. W. (1973), ‘Low-high junction theory
applied to solar cells’, 10th IEEE Photovoltaic Specialists Conf., Palo Alto, pp. 40–49.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 133

Crystalline Silicon Solar Cells 133

Grassman T. J., Brenner M. R., Rajagopalan S., Unocic R., Dehoff R., Mills M., Fraser
H. and Ringel S. A. (2009), ‘Control and elimination of nucleation-related defects in
GaP/Si(001) heteroepitaxy’, App. Phys. Lett. 94, 232106–232106-3.
Green M. A. (1984), ‘Limits on the open circuit voltage and efficiency of silicon solar cells
imposed by intrinsic Auger processes’, IEEE Trans. Electron Devices ED-31, 671–678.
Green M. A. (1987), High Efficiency Silicon Solar Cells, Trans Tech Publications,
Aedermannsdorf.
Green M. A. (1995), Silicon Solar Cells: Advanced Principles and Practice, Bridge Printery,
Sydney.
Green M. A. (1999), ‘Limiting efficiency of bulk and thin film silicon solar cells in the
presence of surface recombination’, Progr. Photovoltaics 7, 327–330.
Green M. A. (2003), ‘General temperature dependence of solar cell performance and impli-
cations for device modelling’, Progr. Photovoltaics 11, 333–340.
Green M. A. (2009), ‘The path to 25% silicon solar cell efficiency: history of silicon cell
evolution’, Progr. Photovoltaics 17, 183–189.
Green M. A. (2011), ‘Ag requirements for silicon wafer-based solar cells’, Progr. Photo-
voltaics 19, 911–916.
Green M. A., Basore P. A., Chang N., Clugston D., Egan R., Evans R., Ho J., Hogg
D., Jarnason S., Keevers M., Lasswell P., O’Sullivan J., Schubert U., Turner A.,
Wenham S. R. and Young T. (2004), ‘Crystalline silicon on glass (CSG) thin-film
solar cell modules’, Solar Energy, Special Issue on Thin Film Photovoltaics 77,
857–863.
Green M. A. and Hansen J. (1998), Catalogue of Photovoltaic Drawings, Photovoltaics
Special Research Centre, University of New South Wales, Sydney.
Green M. A. and Ho J. (2011), Catalogue of Photovoltaic Drawings, 3rd edition, School
of Photovoltaic and Renewable Energy Engineering, University of New South Wales,
Sydney.
Green M. A., Zhao J., Wang A. and Wenham S. R. (1992), ‘45% efficient silicon photovoltaic
cell under monochromatic light’, IEEE Electron Device Lett. 31, 317–318.
Haase C. and Podewils C. (2011), ‘More of everything: market survey on solar modules
2011’, Photon International, February, pp. 174–221.
Hahn G. and Schönecker A. (2004), ‘New crystalline silicon ribbon materials for photo-
voltaics’, J. Phys.: Condens. Matter 16, R1615–R1648.
Hall R. B., Barnett A. M., Brown J. E., Checchi J. C., Ford D. H., Kendall C. L., Mulligan
W. P., Rand J. A. and Ruffins T. R. (1994), ‘Columnar-grained polycrystalline solar
cell and process of manufacture’, U.S. Patent No. 5336335, August 9, 1994.
Haynos J., Allison J., Arndt R. and Meulenberg A. (1974), ‘The COMSAT non-reflective
silicon solar cell: a second generation improved cell’, Int. Conf. on Photovoltaic Power
Generation, Hamburg, p. 487.
Healy S. A. and Green M. A. (1992), ‘Efficiency enhancements in c-Si solar cells by the
incorporation of a region alloyed with germanium’, Solar Energy Mater. Solar Cells
28, 273–284.
Hering G. (2011), ‘Year of the Tiger: PV cell output roared in 2010 to over 27 GW —
beating 2006 through 2009 combined — but can the Year of the Rabbit bring more
multiples?’, Photon International, March, pp. 186–218.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 134

134 M. A. Green

Janoch R., Wallace R. and Hanoka J. I. (1997), ‘Commercialization of silicon sheet via
the string ribbon crystal growth technique’, Conf. Record, 26th IEEE Photovoltaic
Specialists Conf., Anaheim, September/October, pp. 95–98.
Jordan D. and Nagle J. P. (1994), ‘New generation of high-efficiency solar cells: develop-
ment, processing and marketing’, Progr. Photovoltaics 2, 171–176.
Keevers M. J. and Green M. A. (1994), ‘Efficiency improvements of silicon solar cells by
the impurity photovoltaic effect’, J. Appl. Phys. 75, 4022–4033.
Keevers M. J., Young T. L., Schubert U. and Green M. A. (2007), ‘10% efficient CSG min-
imodules’, 22nd European Photovoltaic Solar Energy Conf., Milan, 3–7 September,
pp. 1783–1790.
Keevers M. J., Zhang G. C., Saris F. W., Zhao J. and Green M. A. (1995), ‘Screening of
optical dopants in silicon solar cells for improved infrared response’, Conf. Proc., 13th
European Photovoltaic Solar Energy Conf., Nice, pp. 1215–1218.
Khattak C. P., Basaran M., Schmid F., D’Aiello R. V., Robinson P. H. and Firester A. H.
(1981), ‘Metallurgical-silicon substrates produced by HEM for epitaxial thin film solar
cells’, Conf. Record, 15th IEEE Photovoltaic Specialists Conf., Orlando, May 12–15,
pp. 1432–1437.
King D. L. and Buck E. M. (1991), ‘Experimental optimization of an anisotropic etch-
ing process for random texturization of silicon solar cells’, Conf. Record, 22nd IEEE
Photovoltaic Specialists Conf., Las Vegas, October, pp. 303–308.
Kingsbury E. F. and Ohl R. S. (1952), ‘Photoelectric properties of ionically bombarded
silicon’, Bell Syst. Tech. J. 31, 8092.
Knobloch J., Glunz S. W., Biro D., Warta W., Schaffer E. and Wettling W. (1996), ‘Solar
cells with efficiencies above 21% processed from Czochralski grown silicon’, Conf.
Record, 25th IEEE Photovoltaic Specialists Conf., Washington, Washington D.C., May,
pp. 405–408.
Kolodinski S., Werner J. H., Wittchen T. and Queisser H. J. (1993), ‘Quantum efficiencies
exceeding unity due to impact ionization in silicon solar cells’, Appl. Phys. Lett. 63,
2405–2407.
Lamers M. W. P. E., Tjengdrawira C., Koppes M., Bennett I. J., Bende E. E., Visser T. P.,
Kossen E., Brockholz B., Mewe A. A., Romijn I. G., Sauar E., Carnel L., Julsrud S.,
Naas T., de Jong P. C. and Weeber A. W. (2011), ‘17.9% metal-wrap-through mc-Si
cells resulting in module efficiency of 17.0%’, Progr. Photovoltaics, published online:
23 Mar 2011, DOI: 10.1002/pip.1110.
Lange H. and Schwirtlich I. A. (1990), ‘Ribbon growth on substrate (RGS) — a new approach
to high speed growth of silicon ribbons for photovoltaics’, J. Cryst. Growth 104,
p. 108.
Laudisio G., Young R., Li Z. and Getty R. (2010), ‘A view of the design challenges involved
in the development of advanced n-type contacts using lead-free chemistries’, Proc. 2nd
Workshop on Metallization for Crystalline Silicon Solar Cells — Status, Trends and
New Directions, 14–15 April, Constance, Germany.
Lindmayer J. (1976), ‘Semi-crystalline silicon solar cells’, Conf. Record, 12th IEEE Pho-
tovoltaic Specialists Conf., Baton Rouge, November, pp. 82–85.
Lindmayer J. and Allison J. (1973), ‘The violet cell: an improved silicon solar cell’, COMSAT
Tech. Rev. 3, 1–22.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 135

Crystalline Silicon Solar Cells 135

Maruyama E., Terakawa A., Taguchi M., Yoshimine Y., Ide D., Baba T., Shima M., Sakata H.
and Tanaka M. (2006), ‘Sanyo’s challenges to the development of high-efficiency HIT
solar cells and the expansion of HIT business’, 4th World Conf. on Photovoltaic Energy
Conversion (WCEP-4), May, Hawaii.
Minagawa S., Saitoh T., Warbisako T., Nakamura N., Itoh H. and Tokuyama T. (1976),
‘Fabrication and characterization of solar cells using dendritic silicon thin films grown
on alumina ceramic’, Conf. Record, 12th IEEE Photovoltaic Specialists Conf., Baton
Rouge, November, pp. 77–81.
Neuhaus D.-H. and Münzer A. (2007), ‘Industrial silicon wafer solar cells’, Adv. OptoElec-
tronics 2007, Article ID 24521, 1–15.
Ohl R. S. (1941), ‘Light sensitive electric device’, U.S. Patent No. 2,402,622, (27
March), ‘Light-sensitive electric device including silicon’, U.S. Patent No. 2,443,542
(27 May).
Photovoltech (2010), Doc. D-S&M-008, Rev. A, 22 Oct, www.photovoltech.com.
Ralph E. L. (1975), ‘Recent advancements in low cost solar cell processing’, Conf. Proc.,
11th IEEE Photovoltaic Specialists Conf., Scottsdale, AZ, pp. 315–316.
Riordan M. and Hoddeson L. (1997), Crystal Fire: The Birth of the Information Age, Norton,
New York.
Rodriguez H., Guerrero I., Koch W., Endrös A. L., Franke D., Häßler C., Kalejs J. P. and
Möller H. J. (2011), ‘Bulk crystal growth and wafering for PV’, in Luque A. and
Hegedus S. (eds) Handbook of Photovoltaic Science and Engineering, 2nd Edition,
Wiley, New York.
Sawada T., Terada N., Tsuge S., Baba T., Takahama T., Wakisaka K., Tsuda S. and Nakano S.
(1994), ‘High-efficiency a-Si/c-Si heterojunction solar cell’, Conf. Record, 1st World
Conf. Photovoltaic Energy Conversion, Hawaii, pp. 1219–1225
Scaff J. H., Theuerer H. C. and Schumachor E. E. (1949), ‘P-type and n-type silicon and
the formation of the photovoltaic barrier in silicon ingots’, Trans. Amer. Inst. Mining
Met. Engineering, June, pp. 383–388.
Schmidt F. (2011), ‘History of technologies development for silicon cost reduction’, in Palz
W., Power for the World, Pan Stanford, Singapore.
Schmidt J. and Bothe K. (2004), ‘Structure and transformation of the metastable boron- and
oxygen-related defect center in crystalline silicon’, Phys. Rev. B 69, 241071–241078.
Schott (2009), ‘Realignment of wafer business at WACKER SCHOTT Solar’, Press Release,
30 September.
Schulz M. and Sirtl E. (1984), ‘Silicon sheet’ in Landolt-Börnstein: Numerical Data and
Functional Relationships in Science and Technology, Springer-Verlag, Berlin, 17c,
Section 6.1.2.5.3, 52–54 and 442–444.
Schumann M., Orellana Perez T. and Riepe S. (2009), ‘The solar cell wafering process’,
Photovoltaics International, August, 53–59.
SEMI (2011), Crystalline silicon PV technology and manufacturing group, International
Technology Roadmap for Photovoltaics: Results 2010, 2nd Edition, SEMI PV Group
Europe, Berlin, March.
Serreze H. B. (1978), ‘Optimizing solar cell performance by simultaneous consideration of
grid pattern design and interconnect configurations’, Conf. Record, 13th IEEE Photo-
voltaic Specialists Conf., Washington, D.C., pp. 609–614.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 136

136 M. A. Green

Shah A., Meier J., Torres P., Kroll U., Fischer D., Beck N., Wyrsch N. and Keppner H.
(1997), ‘Recent progress on microcrystalline solar cells’, Conf. Record, 26th IEEE
Photovoltaic Specialists Conf., Anaheim, September/October, pp. 569–574.
Solland (2009), Preliminary Data Sheet, Sunweb, 14 May, www.sollandsolar.com.
Song J., Boas R., Bolman C., Meyers M., Rogol M. and Trangucci R. (2010), ‘The true cost
of solar power: how low can you go?’, Photon Consulting, April.
Stoddard N. G., Wu B., Witting I., Wagener, M. Park Y. Rozgonyi G. and Clark R. (2008),
‘Casting single crystal silicon: novel defect profiles from BP Solar’s Mono2TM wafers’,
Solid State Phenomena 131–133, 1–8.
Strahm B., Andrault Y., Baetzner D., Guérin C., Holmes N., Kobas M., Lachenal D., Mendes
B., Tesfai M., Wahli G., Wuensch F., Buechel A., Mai J., Schulze T. and Vogt M. (2010),
‘Progress in silicon heterojunction solar cell development and scaling for large scale
mass production use’, 25th European Photovoltaic Solar Energy Conf., September,
Valencia, Spain.
Tanaka M., Taguchi M., Takahama T., Sawada T., Kuroda S., Matsuyama T., Tsuda S.,
Takeoka A., Nakano S., Hanfusa H. and Kuwano Y. (1993), ‘Development of a new
heterojunction structure (ACJ-HIT) and its application to polycrystalline silicon solar
cells’, Progr. Photovoltaics 1, 85–92.
Tiedje T., Yablonovitch E., Cody G. D. and Brooks B. G. (1984), ‘Limiting efficiency of
silicon solar cells’, IEEE Trans. Electron Devices ED-31, 711–716.
Trupke T., Green M. A. and Würfel P. (2002a), ‘Improving solar cell efficiencies by the
up-conversion of sub-band-gap light’, J. Appl. Phys. 92, 4117–4122.
Trupke T., Green M. A. and Würfel P. (2002b), ‘Improving solar cell efficiencies by down-
conversion of high-energy photons’, J. Appl. Phys. 92, 1668–1674.
Tsao C.-Y. (2011), ‘Fabrication and characterization of Ge thin films and Ge-rich SiGe alloys
for photovoltaic applications’, PhD Thesis, University of New South Wales, submitted
on 31 March.
Wallace R. L., Hanoka J. I., Narasimha S., Kamra S. and Rohatgi A. (1997), ‘Thin silicon
string ribbon for high efficiency polycrystalline solar cells’, Conf. Record, 26th IEEE
Photovoltaic Specialists Conf., Anaheim, September/October, pp. 99–102.
Wenham S., Robinson R., Dai X., Zhao J., Wang A., Tang, Y. H., Ebong A., Honsberg
C. B. and Green M. A. (1994), ‘Rear surface effects in high efficiency silicon solar
cells’, Conf. Record, 1st World Conf. on Photovoltaic Energy Conversion, December,
pp. 1278–1282.
Wenham S. R. and Green M. A. (2002), ‘Self aligning method for forming a selective emitter
and metallization in a solar cell’, U.S. Patent No. US6,429,037 B1, 6 August.
Werner J. H., Brendel R. and Queisser H. J. (1994), ‘New upper efficiency limits for semi-
conductor solar cells’, Conf. Record, 1st World Conf. Photovoltaic Energy Conversion,
Waikoloa, December, pp. 1742–1745.
Wistey M. A., Fang Y.-Y., Tolle J., Chizmeshya A. V. G. and Kouvetakis J. (2007), ‘Chemical
routes to Ge/Si(100) structures for low temperature Si-based semiconductor applica-
tions’, Appl. Phys. Lett. 90, 82108.
Wolf M. (1960), ‘Limitations and possibilities for improvement of photovoltaic solar energy
converters’, Proc. IRE 48, 1246–1263.
Wolf M. (1976), ‘Historical development of solar cells’, in Solar Cells, Backus C. E., ed.,
New York, IEEE Press.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch03 page 137

Crystalline Silicon Solar Cells 137

Yamamoto K., Yoshimi M., Suzuki T., Okamoto Y., Tawada Y. and Nakajima A. (1997),
‘Thin film poly-Si solar cell with ‘Star Structure’ on glass substrate fabricated at
low temperature’, Conf. Record, 26th IEEE Photovoltaic Specialists Conf., Anaheim,
September/October, pp. 575–580.
Zhao J., Wang A., Altermatt P. and Green M. A. (1995), ‘24% efficient silicon solar cells
with double layer antireflection coatings and reduced resistance loss’, Appl. Phys. Lett.
66, 3636–3638.
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 139

CHAPTER 4

THIN-FILM SOLAR CELLS

BASED ON AMORPHOUS AND
MICROCRYSTALLINE SILICON
CHRISTOPHE BALLIF∗ , MATTHIEU DESPEISSE†
and FRANZ-JOSEF HAUG‡
Photovoltaics and Thin Film Electronics Laboratory
Institute of Microengineering Ecole Polytechnique
Fédérale de Lausanne Neuchâtel, Switzerland
∗[email protected]
†[email protected]
‡[email protected]

One advantage of being disorderly is that one is constantly making exciting discoveries.
A. A. Milne.

4.1 Introduction
Since the first deposition of hydrogenated amorphous silicon out of a silane dis-
charge (Chittik et al., 1969), plasma coating of thin silicon layers has played
an increasingly important role for a large number of applications in ‘macro-
electronics’, including flat panel displays (FPD), solar modules and X-ray imagers.
Indeed, thin silicon layers acting as active semiconductors were likely coated over
more than 140 million square metres in 2010, on glass plates, metal foils and plas-
tic webs. Hydrogenated amorphous silicon used as an active semiconductor n-type
channel in thin-film transistors (TFT) in active-matrix liquid crystal displays made
the bulk of this surface, whereas around 6% of the surface was dedicated to photo-
voltaic (PV) applications. The recent synergy between the FPD sector and thin-film
silicon PV allowed module producers for the first time in thin-film PV history to
purchase thin-film turnkey lines for large area (> 1 m2 ) coating. Manufacturing
costs below 0.5 /Wp (Kratzla et al., 2010; Ringbeck and Stutterlueti, 2012) with
10% module efficiency or higher are now demonstrated by several companies that
provide equipment, making thin-film silicon a low-cost technology, even at rela-
tively low production volume. Short-term perspective cost in the range of 0.35 /Wp

139
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 140

140 C. Ballif, M. Despeisse & F.-J. Haug

are even indicated. In a growing photovoltaic market, which could reach several
hundreds of GW annually, thin-film silicon should be able to remain a player
because of many inherent advantages that are the subject of this chapter.
The key feature of ‘thin-film silicon’ layers for photovoltaic application is
the possibility to manufacture a wide variety of materials with similar processes,
typically by plasma-enhanced chemical vapour deposition (PECVD):

• Hydrogenated amorphous silicon layers (a-Si:H) with a bandgap (E g ) of

1.7–1.8 eV. Alloying with germanium yields a-Six Ge1−x :H absorber layers
with decreased bandgap; addition of oxygen or carbon allows higher bandgaps.
The latter are used essentially as doped or buffer layers, as they tend to suffer
from too strong light-induced degradation for use as absorbers.
• Hydrogenated microcrystalline silicon layers (µc-Si:H): This material
typically consists of interconnected silicon nanocrystallites, forming micro-
metre-sized agglomerates. Surface defects of the nanocrystallites are passi-
vated electronically by a surrounding amorphous phase. The µc-Si:H bandgap
is similar to crystalline silicon (c-Si), i.e. 1.12 eV. As in the amorphous case,
µc-Six Ge1−x :H is used for absorbers with further reduced bandgap. Incorpo-
ration of oxygen or carbon decreases absorption and can be used to fabricate
more transparent doped layers.
• Recrystallised or polycrystalline layers, with typical grain size in the microme-
tre range and with electrical properties (in particular the mobility) closer to
those of c-Si. These layers typically exhibit high defect densities at grain
boundaries because no hydrogenated amorphous silicon phase is present to
passivate dangling bonds.
• Polycrystalline to ideally monocrystalline layers realised by epitaxial growth
on monocrystalline wafer substrates.

Table 4.1 summarises the various forms of silicon films as they are used today in
PV and FPD on industrial levels. This chapter will focus on the class of amorphous
and microcrystalline materials used as absorber and doped layers in thin-film solar
cells, as well as the related technological aspects. Notably, such layers can be
made with a variety of electronic properties and nanostructures, which depend
on the deposition process and on the substrate (i.e. its surface chemistry and its
geometry).
The chapter is organised as follows. The first section gives an overview of
device configuration, technology and history. The layers’ basic properties are
detailed in Section 4.2. Section 4.3 presents the elementary properties of amorphous
and multijunction thin-film silicon solar cells. Section 4.4 focuses on device-grade
absorber deposition of a-Si:H and µc-Si:H materials. In Section 4.5, aspects related
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 141

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 141

Table 4.1 Silicon-based materials used in thin-film Si solar cells and transistors.
Material Application Typical properties In production?

intrinsic a-Si:H channel in thin-film yes

transistor
absorber layer yes
intrinsic a- absorber layer E g down to 1.4 eV yes, in n-i- p
Si1−x Gex :H devices
intrinsic buffer layer in a-Si:H yes
a-SiC:H devices
high bandgap top cell E g > 1.7 eV, suffers no
(Yunaz et al., 2009) from strong
degradation and high
intrinsic defect density
intrinsic channel in thin-film high-mobility n and p pilot stage
µc-Si:H transistor channel
low-bandgap bottom E g at 1.12 eV, yes
cell (Meier et al., passivation of grains
1996) by a-Si:H
intrinsic µc- low-bandgap bottom lower E g than µc-Si:H
Si1−x Gex :H cell (Matsui et al.,
2010)
doped a-Si:H doped contact layers ultra-low lifetime yes
and µc-Si:H
p-SiC x :H p-type window layer higher transparency yes
(Tawada et al., 1981)
p-SiO x :H p-type window layer higher transparency not disclosed
(Sichanugrist et al.,
1983)
n- and p-µc- intermediate reflectors, in situ material*, not disclosed
SiOx :H optical layer containing
(Buehlman et al., nanocrystallites, shunt
2007; Lambertz quenching effect
et al., 2007; Cuony (Despeisse et al.,
et al., 2010) 2010)
reflector buffer layer in situ* alternative to
(Delli Veneri et al., ZnO
2010)
n-µc-SiC transparent contact E 04 at 3–3.2 eV, C
(Chen et al., 2010) incorporated in
crystalline phase
(Continued)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 142

142 C. Ballif, M. Despeisse & F.-J. Haug

Table 4.1 (Continued)

Material Application Typical properties In production?

(re)-crystallised channel in high E g and crystalline yes

layers mobility transistor structure close to
c-Si
absorber layer (Basore, like c-Si, sensitive to
2002; van Gestel grain boundaries and
et al., 2010) dislocations

∗ In situ means the material can also be deposited by PECVD.

to light trapping and rough front (transparent conducting oxide) and back electrodes
(e.g. metallic back reflectors) are presented. Section 4.6 reviews some of the newest
device architecture based on novel transparent conducting oxide and doped layers.
Finally, in Section 4.7 up-scaling, production costs and environmental impacts of
the technology are discussed.

4.1.1 Basic device configuration and definition

Figure 4.1 illustrates the variety of thin-film silicon PV devices. All cells require
a reflecting back contact and a transparent conducting oxide (TCO) at the front.
They include intrinsic layers (i -layers) with good electronic properties sandwiched
between doped p-type and n-type layers. Doped layers are required to create the
internal electric field but they exhibit poor electronic properties (carriers created
in these layers recombine almost immediately); consequently, devices make use
of p-i -n junctions rather than p-n junctions. Because the hole mobility is lower
than the electron mobility in a-Si:H, light should enter from the p-side of the
a-Si:H junction (not necessarily the case for a microcrystalline junction). Hence
when depositing on transparent substrates, the deposition sequence usually starts
with the p-layer and this cell type called p-i -n or superstrate configuration. When
depositing on an opaque substrate (e.g. steel foils or plastic sheets), the n layer is
deposited first; these cells are called n-i - p or substrate configuration. Figure 4.1
shows schematic designs.
By combining two or more cells with different bandgaps, tandem or triple
junctions can be realised. The micromorph configuration shown in panels a) and b)
of Figure 4.1 combines a high-gap amorphous cell with a low-gap microcrystalline
cell. Figure 4.2 shows a typical process sequence for the manufacture of such a
cell. Single-junction and multijunction thin-film silicon solar cell structures are
discussed in detail in Section 4.3.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 143

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 143

Superstrate (p-i-n) Substrate (n-i-p)

front ITO front ITO

Glass Glass
a-Si:H a-Si:H
front ZnO front ZnO a-SiGex:H ZnO buffer
ZnO buffer
a-Si:H a-Si:H a-SiGey:H Polymer
Metal sheet
foil
ZnO buffer
μc-Si:H μc-Si:H
Metal foil

ZnO buffer
back ZnO Metal contact
dielectric
reflector

(a) (b) (c) (d)

Figure 4.1 Schematic of various device configurations. a) p-i-n micromorph cell with thick
TCO at the front and back; b) p-i-n micromorph with thin TCO covered by reflecting metal
at the back; c) triple-junction n-i- p cell on metal foil covered with textured metal and thin
TCO; d) single-junction a-Si:H cell on textured polymer foil with a metal back reflector and
thin TCO.

PECVD PECVD Lami-

Glass P1 laser Back I-V
Front a-Si:H μc-Si:H P2 laser P3 laser Connect na on,
control/ scribe/ contact/ Flasher,
TCO p-i-n p-i-n scribe scribe tabs Junc on
washing washing reflector Sor ng
junc on junc on box

Figure 4.2 Schematic of the process steps in micromorph module production.

All schemes in Figure 4.1 show rough interfaces within the devices. These
rough interfaces are essential to ensure efficient light scattering inside the silicon
layers and thus to increase the effective light path. The light management schemes
that are used in thin-film silicon technology are discussed in Section 4.5, while
advanced concepts and substrates are presented in Section 4.6. The properties of the
front TCO and back reflector are essential for achieving high device performance;
their optimisation is as important as the PECVD deposition process of the silicon
films. Two important aspects emerge for the choice of substrate texture:
• In superstrate devices, textured TCOs control the maximum current density
via their transmission and light scattering potential.
• The texture of the growth template can lead to defects in the devices, such as
collision of growth fronts, which can result in reduced electrical performance.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 144

144 C. Ballif, M. Despeisse & F.-J. Haug

Intense research is therefore ongoing to find ideal superstrates/substrates which

allow for high current density while maintaining optimal electric performance; we
discuss this in more detail in Sections 4.3 and 4.5.

4.1.2 Uniqueness of thin-film silicon for PV: strength and weakness

The strengths of thin-film silicon PV can be summarised as follows:
• In their most simple form, modules can be made out of glass, a transparent front
contact made from zinc oxide or tin oxide, thin layers of silicon and a back
electrode. All the materials are abundant (Wadia et al., 2009) and non-toxic.
Micromorph modules can be made in compliance with norms on restriction
of hazardous substances (RoHS) and can be handled like standard window
glass at their end of life. This is presently the only technology fully adapted
for transitioning to a terawatt-scale, PV-based society.
• Compared with c-Si, the energy payback time of such modules can be reduced
significantly, especially with the use of thin µc-Si absorber layers (Fthenakis
et al., 2008a; Fthenakis et al., 2008b; Steimen, 2011). On very large scale, it
makes more sense globally to develop technologies with as low as possible
energy payback times.
• The temperature coefficient is favourable (typically around –0.2%/◦C for
a-Si:H modules and –0.25%/◦ C for micromorph modules), leading to higher
energy yield in practice, particularly in hot environments (Shah, 2010; Lechner
et al., 2010).
• Proven manufacturing technology has been developed: in recent years a huge
industrial effort has been made to industrialise thin-film silicon, tackling most
production issues. Production tools are now available to module manufacturers
(Kratzla et al., 2010; Klein et al., 2010).
• Eventually the possibility to create materials with various bandgaps should
open the route to higher efficiency multiple-junction devices.
The weakness of thin-film silicon technology lies in its relatively modest effi-
ciency, with currently confirmed best stabilised efficiencies of 13.4% and 12.5%
for p-i -n and n-i - p triple-junction cells, respectively (Kim et al. 2012; Yue et al.,
2010), and 12.3% for p-i -n tandem cells (Kroll et al., 2011). However, the latest
results from several module and equipment makers indicate that it is possible to
manufacture large area modules with 10% stabilised efficiencies in a cost-effective
way (Kratzla et al., 2010; Klein et al., 2010), and champion modules of 11% have
been reported. In contrast to other technologies, there is only a very small gap
between laboratory and real-world efficiencies. Thin-film silicon thus represents a
technology that could secure a substantial share of the PV market, in particular for
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 145

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 145

large solar parks in high irradiance areas or for integration of aesthetically pleasing
and safe PV products into buildings, but it certainly needs further increased effi-
ciency in the medium term to stay competitive with other mainstream technologies.

4.1.3 A short historical perspective

In 1976, Carlson and Wronski demonstrated the first (2.4% initial) ‘efficient’
a-Si-based PV device (Carlson and Wronski, 1976). One year later, Staebler and
Wronski reported that the photoconductivity of a-Si:H suffers on illumination
(Staebler and Wronski, 1977). This coined the term ‘Staebler–Wronski effect’
(SWE) which is often used as a synonym for any kind of light-induced degrada-
tion (LID) in thin-film silicon devices. In the 1980s, several companies started the
commercialisation of amorphous silicon modules (Solarex based on initial work
at RCA, AFG/Chronar, Arco Solar, etc.). In 1986, Arco Solar introduced the first
large-area commercial a-Si:H modules (Genesis G-400). In the 1980s and 1990s
the PV market was small (a few MWs) and the first a-Si:H modules suffered from
reliability issues, including poorly controlled LID, which prevented a successful
start to mass production. However, thanks to the simple manufacturing of a-Si:H
cells and their good low-light behaviour (the high bandgap results in a low satura-
tion current density i o ), they were successfully integrated in consumer applications
such as the pocket calculators commercialised by Sharp and Casio.
At the end of the 1980s and in the 1990s, a-Si:H material stability improved by
modifying the plasma deposition processes and, for instance, by adding additional
hydrogen to the glow discharge or by introducing very high frequency (VHF)
plasma deposition, which resulted in higher deposition rates (Curtins et al., 1987)
while maintaining a similar material quality. United Solar Ovonic was the first to
introduce triple-junction technology based on a-Si:H/a-Six Ge1−x :H/a-Si y Ge1−y :H
(Yang et al., 1994). They commercialised a flexible large-area product on steel foil.
In Japan, Fuji developed flexible modules in a-Si:H/a-Six Ge1−x :H configuration
on polyimide.
During the mid-1990s, the first efficient solar cells based on µc-Si:H were
realised by careful control of extrinsic impurities (see Section 4.2.6). Their inte-
gration with a-Si:H into high-efficiency micromorph tandem devices was demon-
strated (Meier et al., 1994); tandem devices increase stabilised efficiencies by
20% to 30% compared with single-junction amorphous silicon cells. This finding
resulted in intense research activities in various laboratories (Shah et al., 2004;
Rech et al., 2006; Schropp et al., 2007; Yan et al., 2007). An inherent advan-
tage of µc-Si:H is the fact that it is fabricated from the same hydrogen (H2 )
and silane (SiH4 ) input gases as a-Si:H. Moreover, the bandgaps of a-Si:H and
µc-Si:H coincide with the theoretical optimum of tandem devices for the solar
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 146

146 C. Ballif, M. Despeisse & F.-J. Haug

irradiation spectrum (AM1.5G) (DeVos, 1980). Microcrystalline silicon thus offers

an alternative to silicon-germanium alloys that were previously used for tandem
and triple-junction cells by several research groups and companies.
Japanese companies were the first to put massive R&D efforts into large-
scale industrialisation of thin-film silicon technology with a focus on micromorph
devices on glass. Sharp, Kaneka and MHI started production of amorphous sili-
con modules on glass, relying mostly on in-house development of the production
equipment for module manufacturing, e.g. the high-rate ladder electrode design of
MHI (Takeuchi et al., 2001). Kaneka were the first to commercialise micromorph
modules in 2001 at a pilot stage (1∼2 MW). In this time also other flexible activities
started up, e.g. Powerfilm (USA), Flexcell (Switzerland) and Xunlight (USA).
Thin-film silicon technology gained momentum in 2005 when equipment mak-
ers with experience from FPD entered the scene. The development was further
boosted by the silicon raw material shortage in 2007–2008 that prevented newcom-
ers from entering into c-Si cell manufacturing. This strong industrialisation phase
required a lot of research resources to solve up-scaling issues (cf. Section 4.7)
and start up dozens of manufacturing plants, and slowed the quest for improved
efficiencies. Recently, however, further improvements in cell performance were
achieved; most of the record cell efficiencies listed in Table 4.2 were reported by
industrial development labs, for example a new initial record of 16.3% efficiency
has been reached for a triple-junction cell (Guha et al., 2011). Close collaboration
with research institutes should offer the industry the possibility to develop and use
novel materials (e.g. new doped layers) and concepts (e.g. advanced light trapping),
which promise further gains in efficiency (Konagai et al., 2011).

4.2 Basic properties of amorphous and microcrystalline silicon alloys

This section gives a brief overview of the electronic and optical properties of thin-
film silicon. Following the discovery of a-Si:H and its first use in devices, it soon
became necessary to change some of its properties for a variety of reasons.

• Figure 4.3 shows that the bandgap of amorphous silicon is between 1.7 to
1.8 eV; this is generally considered to be too high for efficient absorption of
the solar spectrum. Attempts were therefore made to lower the bandgap by
alloying with germanium or tin, or to find new materials with lower gaps for
the construction of advanced tandem and multijunction devices.
• While the demonstration of doping was surprising because an amorphous
network should easily incorporate elements with lower or higher valence, it was
soon noted that the achievable doping efficiency is limited; films with better
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 147

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 147

6
10
Absorption coefficient (cm-1)

5
10
4
10
3
10
2 a-Si:H
10
c-Si:H
10
1 c-Si

0
10
-1
10
1.0 1.5 2.0 2.5 3.0 3.5 4.0
Photon energy (eV)
Figure 4.3 Spectral absorption of the thin-film materials a-Si:H and µc-Si:H (open squares
and circles, respectively), and c-Si (filled circles). Dashed lines illustrate the Urbach tail
between 1.25 and 1.75 eV and the indirect bandgap for energies beyond 1.75 eV.

conductivity became eventually available, but they were found to contain small
crystallites of silicon.
• Such crystallite-containing silicon is actually a material in its own right, called
microcrystalline or nanocrystalline silicon, useful not only for doped layers
but also as an absorber with a low bandgap. Figure 4.3 shows that its absorption
is similar to that of crystalline silicon, i.e. it extends into the near IR region,
but it is also relatively weak (101 –103 cm−1 between 1.5 and 1.12 eV).
• Absorption in the doped layers of solar cells does not contribute to the photo-
current. Recombination losses in doped amorphous layers are unavoidable
because defects are inherently created along with doping. In solar cell
applications, losses in the doped layers were eventually reduced by using
materials with higher bandgap, e.g. by incorporation of carbon, oxygen or
nitrogen.

In the remainder of this section these and other aspects will be addressed in more
detail.

4.2.1 Amorphous silicon

Amorphous silicon can easily be grown by evaporation or sputtering, but the
obtained material is generally too defective for use in semiconductor devices.
Higher quality amorphous silicon was obtained by deposition from SiH4 glow
discharges (Chittick et al., 1969), essentially the process called PECVD which
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 148

148 C. Ballif, M. Despeisse & F.-J. Haug

is illustrated in Figure 4.13. This material showed a thermally activated dark

conductivity indicative of transport by charge carriers that are activated over a
bandgap, and an acceptable level of photoconductivity. PECVD-grown material
differs from evaporated films by the incorporation of hydrogen. Because the hydro-
gen content can easily reach 10–15 at.%, PECVD material is normally called
hydrogenated amorphous silicon.
In contrast to crystalline silicon where the four tetrahedral valence electrons
are perfectly accommodated in the diamond lattice, there is no long-range order in
amorphous silicon. Nevertheless, a tetrahedral bonding environment reaching as far
as the first- and second-order neighbours is maintained (Laaziri et al., 1999). The
resulting defects in the amorphous silicon network are associated with dangling
bonds which contain an unpaired electron, similar to the dangling bonds at the
surface of a freshly cleaved silicon crystal. Their density can thus be measured
by the spin resonance of the unpaired electron that resides in the neutral dangling
bond (Haneman, 1968; Brodsky and Title, 1969).
Figure 4.3 illustrates that the spectral absorption of amorphous silicon shows
the behaviour of an indirect bandgap with E g = 1.7 eV, according to (E − E g )2 .
Below the bandgap, the absorption becomes exponential and is known as Urbach
absorption. The exponential part of the curve can be described by a characteristic
energy E 0 = 50 meV, often called the disorder parameter. In material of inferior
quality, a higher degree of disorder yields higher Urbach absorption, and corre-
spondingly higher values of the characteristic energy E 0 . For amorphous silicon
deposited under different conditions, the transition between the parabolic high-
energy part and the exponential Urbach region can vary significantly. Thus, the
absorption of amorphous silicon is sometimes characterised by E 04 or E 03 , the
energy where the absorption coefficient assumes a value of 104 cm−1 (or corre-
spondingly 103 cm−1 ). The absorption at energies below 1.4 eV is related to defect
levels located around mid-gap; their density can be related linearly to α1.25 , the
value of the absorption coefficient at 1.25 eV (Jackson and Amer, 1982).
Combining optical absorption measurements (Redfield, 1982) with results of
drift experiments (Tiedje et al., 1981) and photoemission studies (Jackson et al.,
1985) yields the schematic band diagram shown in Figure 4.4. As in ideal semicon-
ductors, valence and conduction bands are associated with freely moving carriers.
The band edges in amorphous silicon are not well defined but exhibit exponentially
decreasing densities of states which extend from the respective band edges into the
gap. The Urbach absorption is dominated by the characteristic energy of the valence
band tail whereas the conduction band tail is steeper with a characteristic energy
of about 30 meV (Redfield, 1982). As far as conductivity is concerned, carriers
residing in tail states are assumed to be immobile and move only after excitation
above the band edge, giving rise to the term ‘mobility gap’.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 149

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 149

23
10
22
Density of states (1/cm3eV) 10
21
10 VB-tail CB-tail
20
10
19
10
18 D+ D0 D-
10
17
10
16
10
-0.5 0.0 0.5 1.0 1.5 2.0
Energy with respect to VB edge (eV)

Figure 4.4 Schematic density of states in device-grade a-Si:H. For recombination statistics,
often two transition states D+ /D0 and D0 /D− are used rather than actual charge states.

Neutral defect states are expected at approximately mid-gap position, but

owing to the disordered character of a-Si:H, their density is normally described by
a Gaussian distribution. In its neutral state, the defect is occupied by an unpaired
electron; therefore it is usually denoted by D0 , but the defect is amphoteric and can
also acquire positive or negative charge. For the formation of the negatively-charged
defect, D− , two electrons must be accommodated in the defect state; the amount
of energy to overcome electrostatic repulsion is called the correlation energy U.
Alternatively, the defects can be described by two transition states, D+ /D0 and
D0 /D− , separated by U .

4.2.2 Light-induced degradation (LID)

In addition to defects that are inherently present directly after the growth of amor-
phous silicon, exposure to light creates additional defects that degrade the photo-
conductivity as well as the dark conductivity. Subsequent thermal annealing can
completely restore the initial state of the samples. This observation, as previously
noted, is called the Staebler–Wronski effect (SWE) (Staebler and Wronski, 1977).
The salient features are the following (Staebler and Wronski, 1981):
• The creation of light-induced defects does not depend much on the energy of
the incident light.
• Fewer defects are created in samples where a bias sweeps out the charge
carriers before recombination.
• Samples with rectifying junctions also degrade in the dark when a forward
bias is applied.
• The dependence on light intensity is sub-linear.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 150

150 C. Ballif, M. Despeisse & F.-J. Haug

Based on these observations, it was suggested that the creation of defects in

amorphous silicon is actually due to recombination of charge carriers, regardless
of their origin. In this context, three more aspects of LID are noteworthy:

• Degradation of the photoconductivity is reduced in films for which hydro-

gen is added to the discharge (Guha et al., 1981). To date, hydrogen dilution
remains an efficient way to obtain more stable absorber material for solar cells.
Nevertheless, degradation cannot be completely avoided (cf. Figure 4.9).
• The nature of the recombination process matters only little in terms of creating
defects. While non-radiative recombination via defects prevails around room
temperature, recombination is dominantly radiative at low temperature; how-
ever, the degradation of the photoconductivity is virtually unchanged (Stradins,
2003).
• The exact nature of the SWE defects and the role of hydrogen in annealing are
not clear. Defect creation by breaking of weak bonds during carrier recombi-
nation is a common model (Stutzmann et al., 1985), whereas models involving
hydrogen have also been proposed (Winer, 1990; Branz, 1999).

Recovery of an initial state by annealing, and creation of defects by recom-

bination, inspired the so-called equilibrium model; based on a large reservoir (or
pool) of weak or strained bonds (Stutzmann et al., 1985), the model assumes that at
high temperature breaking and formation of bonds takes place in thermodynamic
equilibrium with a given defect formation energy (Smith and Wagner, 1985). The
pool of weak bonds is associated with states in the valence-band tail. Their density,
in turn, varies exponentially with the disorder parameter E 0 , the Urbach energy.
Annealing at high temperature establishes equilibrium between bonds that form
during annealing and bonds that break by the recombination of thermally activated
charge carriers. At a given temperature during cooling, the kinetics of the equilib-
rium process become too slow to follow the cooling ramp. Thus, the defect density
ND (i.e. the density of broken bonds) gets frozen in at this temperature. If the
defect is associated with an energy E D above the valence-band edge, the equi-
librium value of ND can be calculated statistically by minimising the free energy
F at this freeze-in temperature. The idea of freeze-in was very well supported by
experiments where samples were cycled repeatedly to high temperature and rapidly
quenched in cold water in order to freeze their defect densities at different stages
of the cooling ramp (Street and Winer, 1989). Later additions to the model took
into account entropy effects by bond breaking (Winer, 1990) and detailed interac-
tions with hydrogen (Powell and Deane, 1996). Common to these models is the
existence of a minimum defect density which emerges as inherent consequence of
disorder. Being thus an unavoidable property of amorphous silicon, the impact of
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 151

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 151

defects must be taken into account by the device design, e.g. by charge transport
via drift rather than diffusion.

4.2.3 Doping
Doping of a-Si:H with phosphine (PH3 ) to make n-type material was already
reported with the first demonstration of a-Si:H (Chittick et al., 1969). Later, p-type
doping was also achieved by adding diborane (B2H6 ) to a glow discharge (Spear and
Le Comber, 1976), but the doping efficiency in amorphous silicon always remained
low. The assumption of amphoteric defects in the middle of the bandgap can explain
this behaviour: if n-type doping moves the Fermi level towards the conduction band,
the defect eventually lies below the Fermi level. It can be charged with a second
electron after overcoming the correlation energy U . Likewise, if p-type doping
moves the Fermi level towards the valence band, the unpaired electron residing
in the neutral defect D0 is above the Fermi level and it becomes energetically
more favourable to form a charged defect D+ . Free carriers of either type are thus
absorbed by the defect state, putting a stop to any further doping.
Figure 4.5 illustrates the energy gain by the two processes, assuming a bandgap
of 1.75 eV, a defect state at E D = 0.85 eV above the valence band, and a correlation
energy of U = 0.2 eV. The predicted defect densities according to the equilibrium
model and some experimental data are also illustrated in Figure 4.5; whenever
the Fermi level is moved too far from mid-gap, the formation energy of one type
Formation energy (eV)

1.0
0.8
D0
0.6
0.4 D+ D-
0.2
0.0

18
10
17
10
ND (cm-3)

16
10 p-type n-type
15
10
14
10
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Fermi level w.r. to VB (eV)

Figure 4.5 Formation energies of defects in their various charge states with respect to
the Fermi level (upper panel). The lower panel compares the predicted equilibrium defect
densities, squares and circles refer to reported data of p- and n-doped samples, respectively.
Data after Winer (1990).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 152

152 C. Ballif, M. Despeisse & F.-J. Haug

of charged defect starts dropping below the level of that of the neutral defect.
Consequently, the formation of this defect becomes more likely and works against
any further doping. The lowest possible defect concentrations are found in a narrow
region where the neutral defect dominates.
Note that shifts of the Fermi level are not restricted to doped materials, but also
occur throughout the intrinsic layer of a p-i -n junction. The equilibrium models
thus predict increased densities of positively and negatively charged defects close
to the p-i and i -n interfaces, respectively. Charging of the defects close to the p-i
interface gives rise to the band bending observed in Figure 4.7. The understanding
of these effects led to the introduction of buffer layers and bandgap grading close
to the interfaces where the defect densities are projected to be the highest (Guha
et al., 1989).
While the equilibrium idea is quite powerful, there is still controversy whether
such a comparatively simple model with a single defect state and three different
charge states is sufficient. Indeed, to date no model can explain consistently how
bond breaking and formation takes place on a microscopic level. Also, evidence
from cell degradation suggested early on that there is a difference between fast and
slow defects (Yang and Chen, 1993). In addition, defect densities determined by
two independent methods, such as lifetime and defect absorption measurements,
have repeatedly shown very different kinetics during degradation-annealing cycles
(Stradins, 2003), suggesting the existence of various defect states rather than iso-
lated dangling bonds.

4.2.4 The role of hydrogen in amorphous silicon

The importance of hydrogen in a-Si:H and the differences between a-Si:H and
hydrogen-free material grown by sputtering or evaporation were noticed only grad-
ually, and the exact details of the various effects of hydrogen in amorphous silicon
are still elusive despite four decades of research. The most prominent effect of
hydrogen is a reduction of the defect density; compared with about 1020 cm−3
in sputtered films, the density of dangling bond defects in PECVD material is
reduced to values between 1016 and 1017 cm−3 (Brodsky and Title, 1969; Brodsky
and Kaplan, 1979). The passivation is thought to proceed similarly to the case
of crystalline silicon where hydrogen has also been reported to provide excellent
passivation of broken bonds on the surfaces (Yablonovitch et al., 1986; Higashi
et al., 1990).
The hydrogen content and its bonding environment can be determined by
effusion (also called evolution) experiments; on heating a-Si:H above 300◦ C in
vacuum, the effusion of hydrogen from higher hydrides becomes measurable by
an increase in the hydrogen partial pressure (Brodsky et al., 1977a); effusion from
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 153

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 153

more stable monohydrides starts only beyond 450◦ C. In the previous section it was
shown that bond breaking is related to the position of the Fermi level. This explains
the observation that hydrogen effusion depends on doping (Beyer, 2003). Silicon-
hydrogen bonds also have distinct signatures in Fourier transform infrared (FTIR)
spectroscopy; stretching modes that absorb at 2010 cm−1 are generally attributed to
monohydride bonding of dense material whereas a shift of this signature towards
2090 cm−1 is attributed to di- or tri-hydrides which are often found in porous
material.
While the equilibrium models discussed in Sections 4.2.2 and 4.2.3 were
developed around the idea of a continuous random silicon network with isolated
defects, experimental evidence suggests that hydrogen can be bonded also at inter-
nal surfaces of nano-sized voids that exist even in device quality material (Mahan
et al., 1989). This is corroborated by measurements of mass density and hydrogen
content that are compatible with the mass deficit of hydrogenated di-vacancies
over a wide range of deposition conditions (Smets et al., 2007a). In this context,
the above-mentioned IR absorption signatures are attributed to shifted resonance
frequencies of hydrogen atoms that are confined to anisotropic volume defects such
as hydrogenated di- and poly-vacancies (1980–2010 cm−1, dense material), or to
quasi-free oscillations of hydrogen located on the surface of microscopic voids
(2090 cm−1, porous material). The integrated intensities of the two contributions
are used to define the microstructure factor R = I2090 /(I2090 + I2010 ) for distin-
guishing dense and porous material, but more importantly, a dominant presence of
monohydride bonding is also associated with reduced defect density (Kroll et al.,
1998).

4.2.5 Amorphous silicon alloys

Alloying amorphous silicon with carbon widens the bandgap while additions
of germanium or tin reduce it. The maximum tolerable germanium content for
absorber layers yields a gap of about 1.6 eV (Nakamura et al., 1980); strong hydro-
gen dilution can extend this range to about 1.5 eV (Matsuda et al., 1986). One of
the underlying reasons for the poor performance of a-Si1−x Gex :H is preferential
bonding of hydrogen to silicon compared with germanium, resulting in poor pas-
sivation of broken germanium bonds (Lucovsky, 1992). A second complication is
the easier dissociation of germane (GeH4 ) compared with silane, which can for
example, be compensated for by using mixtures of disilane and germane (Matsuda
and Ganguly, 1995). Because of their high defect densities, absorber layers made
from a-Si1−x Gex :H are generally graded so that material with higher bandgap but
lower defect density is close to the junctions (Guha et al., 1989; Lundszien et al.,
2002).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 154

154 C. Ballif, M. Despeisse & F.-J. Haug

Window layers of p-type a-Si1−x Cx :H were introduced by the Hamakawa

group at Osaka University (Tawada et al., 1981). Because doped layers are already
highly defective, additional defects caused by poor hydrogen passivation are some-
what less destructive; however, increasing the carbon content degrades the mobility
in p-type a-Si1−x Cx :H, thus putting an upper limit of about 25% on the carbon con-
tent. Defect creation does become an issue when a-Si1−x Cx :H is used for high-gap
absorber layers; problems due to the different reactivity of methane compared with
silane can, for example, be avoided by the use of mono-methyl-silane (CH3 -SiH3 ,
MMS) (Yunaz et al., 2009). For doped layers, amorphous nitrides and oxides have
been investigated as high-gap alternatives to carbides (Li et al., 1994). Recently,
phosphorus- and boron-doped silicon oxides received renewed attention because
they develop a microcrystalline phase when deposited with high amounts of hydro-
gen; we discuss this further in Section 4.6.

4.2.6 Microcrystalline films

PECVD-grown films with a crystalline component were first observed in studies
of doped films when high discharge powers or fluorine-containing precursor gases
were used (Usui and Kikuchi, 1979; Matsuda et al., 1980); in X-ray diffraction,
such films showed the presence of silicon crystallites with sizes between 5 and 10
nm. Because ion bombardment in high-power plasmas is often a concern, micro-
crystalline films are normally grown by an alternative method using precursor gas
mixtures of silane and large quantities of hydrogen, often referred to as hydrogen
diluted silane (Matsuda, 1983). The preferential growth of microcrystalline films
in fluorine- or hydrogen-containing plasma indicates that the deposition of µc-Si
is initiated by etching of unstable nucleation sites.
Figure 4.6 shows schematic phase diagrams of PECVD-grown silicon as the
hydrogen content changes. Without additional hydrogen, film deposition is nor-
mally completely amorphous. With the addition of hydrogen, isolated crystallites

Figure 4.6 Schematic transition from fully amorphous silicon (left) over mixed-phase mate-
rial to fully microcrystalline silicon (right) with increasing hydrogen content in the precursor
gas mix.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 155

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 155

nucleate within a matrix of amorphous material; the conductivity of the mate-

rial changes significantly when the crystallites reach the percolation threshold.
Adding yet more hydrogen yields films with a distinct transition between an amor-
phous nucleation region and conical crystalline clusters which themselves consist
of smaller grains. Finally, when using very high hydrogen content the deposition
rate is decreased drastically and the resulting films consist of a porous array of
elongated grains. Deposition of uniform films requires an interface treatment that
establishes a template for nucleation, and a hydrogen gradient during the deposi-
tion can ensure uniform crystallinity across the thickness of the film (Koh et al.,
1999a; Collins et al., 2003).
Undoped material, i.e. silicon deposited without the addition of doping gases,
often exhibits unintentional n-type conductivity, presumably due to oxygen impu-
rities. Truly intrinsic material for use as absorbers is obtained by compensation —
also called micro-doping — with [B2 H6 ]/[SiH4 ] ≈ 10−5 (Wang and Lucovsky,
1990), by suppressing oxygen impurities with specially purified source gases
(Torres et al., 1996), or by carrying out the deposition at low substrate temper-
ature where the oxygen impurities are sufficiently passivated by the incorporation
of hydrogen (Nasuno et al., 2001b).
Figure 4.3 compares the spectral absorption of microcrystalline silicon with
that of amorphous and crystalline silicon. Because µc-Si:H consists of silicon
crystallites embedded in an amorphous matrix, its spectral absorption resembles
an interpolation of these extremes. Device-grade material is usually considered
to have a crystalline fraction between 50% and 70%. The unique properties of
µc-Si:H are attributed to the excellent passivation of defects at the surface of the
nanocrystallites which is provided by the surrounding amorphous matrix. Hence,
transport of excited carriers proceeds via percolation among the crystallites; col-
lisions at the grain boundaries may reduce their mobility with respect to c-Si, but
their lifetime stays at acceptable levels because recombination is suppressed.
Doping of microcrystalline silicon with phosphine (PH3 ) is compatible with
most processing conditions, but the growth of p-type material requires special
attention; diborane (B2 H6 ) has been reported to react spontaneously and its shelf
life is limited, tri-methyl-boron (B(CH3 )3 , TMB) tends to delay the nucleation
of microcrystalline films, and boron trifluoride (BF3 ) requires very high plasma
power due to its chemical stability (Koh et al., 1999b). For a given doping gas
ratio, the conductivity in microcrystalline layers is generally about two orders of
magnitude higher than in amorphous films (Wang and Lucovsky, 1990).
Other elements (oxygen, carbon, nitrogen, germanium) can be added to modify
the properties of intrinsic or doped µc-Si layers, altering the optical and electrical
properties, as will be discussed in Section 4.6.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 156

156 C. Ballif, M. Despeisse & F.-J. Haug

4.3 Thin-film silicon solar cells

In amorphous as well as in microcrystalline silicon, the charge-carrier diffusion
lengths are much shorter than the absorption length of near-bandgap light. More-
over, in Section 4.2.1 we showed that every shift of the Fermi level out of the
mid-gap position creates additional defects in amorphous silicon. Thin-film sili-
con solar cells are therefore fabricated as drift devices using p-i -n junctions rather
than diffusion-controlled p-n junctions. The drift length, given by L drift = µEτ ,
where E is the strength of the electrical field and µ and τ are the mobility and
lifetime of the carrier, should be larger than the device thickness. At the maximum
power point and open-circuit voltage (Voc ), L drift is decreased as E is decreased.
Thin devices benefit doubly from the reduced thickness as the transit length is
reduced and the electric field is increased. Hence they are much less prone to
suffer from localised defects or SWE.

4.3.1 Single-junction solar cells

In ideal p-i -n devices, the potential variation should be linear across the intrinsic
region; however, the formation of charged defects deforms the band structure. In
most situations, the band diagram must therefore be determined numerically, for
example with software packages like ASA (Pieters et al., 2007) or AMPS (Zhu
et al., 1998). Figure 4.7 shows a band diagram for an amorphous p-i -n junction
and the front TCO, using the defect pool model in ASA. Compared with the ideal
case, the region close to the front interface shows a complicated band structure that
incorporates a Schottky barrier between the front contact and the p-layer, as well
as band distortion due to positively charged defects in the i - p region; the latter
results in a significant reduction of the electric field in the middle of the p-i -n
junction.
When additional defects are created during illumination, the field in the mid-
dle of the junction becomes further weakened. Losses in charge carrier collection
are illustrated in Figure 4.8, which shows the external quantum efficiency (EQE)
of an n-i - p cell before and after light soaking. The EQE represents the spectrally
resolved probability of creating and collecting a charge carrier from an incident
photon. Figure 4.8 shows that collection losses occur predominantly for wave-
lengths above 400 nm. Shorter wavelengths are strongly absorbed close to the
i - p interface; according to Figure 4.7, the electric field in this region is actually
enhanced by LID; thus collection here is less affected than in the bulk of the
absorber layer.
The EQE is typically measured at zero bias; the corresponding current density
represents short-circuit conditions and can be obtained by multiplying the EQE
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 157

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 157

2
TCO <p> <i> <n>

1
Potential (eV)

CB

VB
-1

<p> <i>

-2
0 50 100 150 200
Distance from TCO/p interface (nm)

Figure 4.7 Equilibrium band diagram of an amorphous silicon solar cell with a 200 nm thick
absorber layer and p- and n-layers with activation energies of 450 and 250 mV, respectively.
Potentials are shown with respect to the Fermi level. Dashed and solid lines represent ideal
and realistic band alignment, respectively. The inset illustrates the Schottky junction between
the front TCO and the p-layer.

1.0

0.8

0.6
EQE

0.4
2
14.5 mA/cm
2
0.2 13.6 mA/cm

0.0
400 500 600 700 800
Wavelength (nm)

Figure 4.8 EQE of an n-i- p a-Si:H single-junction cell without antireflection coating. Solid
and dashed lines illustrate the charge-carrier collection in the initial and stabilised states,
respectively.

by the illumination spectrum (in units of photons per wavelength interval and
per surface area), and integrating over all wavelengths. This determination of the
short-circuit current density from the EQE is normally more precise than the value
obtained from a current–voltage measurement because the EQE does not depend on
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 158

158 C. Ballif, M. Despeisse & F.-J. Haug

the cell area, whereas a reliable determination of the cell area can become difficult
for small cells, particularly if they include light-scattering interfaces.
After LID, the drift field is weakened and the collection of photogenerated
charge carriers becomes even less effective under the forward bias of operating
conditions. As a result, the solar cell efficiency is further reduced via losses in fill
factor (FF). Intuitively, Voc should be related to the potential created by the doped
layers and thus be minimally influenced by LID. However, this simplistic view
does not take into account that Voc can suffer because of increased recombination
once the layers become more defective.
Figure 4.9 illustrates the kinetics of cell efficiency degradation. The degrada-
tion scales with the absorber layer thickness: thinner layers degrade less because
the remaining field in the centre of the junction remains stronger in thinner cells,
even when weakened by additional defects. Moreover, the figure shows that cells
in which the i -layer is deposited with additional hydrogen stabilise faster and at
a higher efficiency for a given thickness (Wronski, 1996). In general, there is no
simple dependence on illumination duration, intensity or spectrum, and even at
moderate temperatures of 60 or 70◦C, the degradation is partly reversed by anneal-
ing. Therefore, standard degradation experiments are commonly carried out for
1000 hours at 50◦ C under exposure to light with the intensity and the spectrum of
sunlight. Reduced cell thickness is desirable, not only in terms of stability against
LID, but also because of throughput considerations in industrial production. A
sufficient level of light absorption must nevertheless be maintained. Absorption

1.0
85 nm
0.9
Normalised efficiency

0.8

0.7 240
nm

0.6 40
0n
m

0.5 64
0n
m

0.4
1 10 100 1000
Degradation time (h)

Figure 4.9 Degradation kinetics of solar cells with various absorber layer thicknesses; data
from Bennett et al. (1986) and Bennett and Rajan (1988). The solid line represents a cell
with a 400 nm i-layer deposited with hydrogen dilution. Data from Wronski (1996).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 159

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 159

enhancement in thin-film silicon solar cells is generally summarised by the term

‘light trapping’; we discuss this in more detail in Section 4.5.

4.3.2 Tandem and multijunction solar cells

The combination of two bandgaps in tandem solar cells makes a better use of
the solar spectrum and therefore yields higher efficiencies. Here, thin-film silicon
and its alloys have an inherent advantage because cells with different bandgaps
can be stacked monolithically, one on top of another, since no lattice matching
is required. Theoretical considerations suggest an optimum for bandgaps of 1.1
and 1.7 eV (DeVos, 1980). The upper value corresponds to a-Si:H and the lower
value is in principle accessible with a-Si1−x Gex :H, but Section 4.2.5 showed the
resulting alloy is too defective for working devices. An alternative choice is there-
fore µc-Si:H. Figure 4.10 shows cross-sections through p-i-n/p-i-n and n-i-p/n-i-p
devices.
Electrically, such cells are connected in series and therefore their current den-
sities must be matched while the output voltage of the tandem cell is obtained by
adding the voltages of the individual cells.
The design rules for multijunction cells start by looking into the current that is
generated by the bottom cell alone, and then trying to define useful distributions of
that current into two or more sub-cells. Technological throughput considerations
limit microcrystalline silicon cell thickness to between 1 and 3 µm, resulting in i sc
between 25 and 30 mA cm−2 with adequate light trapping. The usable composition
range of a-Si1−x Gex :H extends the bandgap range to about 1.4 eV, but its high
absorption can yield a similar range of current densities. Amorphous silicon cells
with typical thicknesses of 200 nm can yield about 15 mA cm−2 (cf. Figure ,4.8) and
up to 17 mA cm−2 if good light trapping schemes and antireflection coatings are
used. Without light trapping and back reflectors, a single passage of light through
the amorphous cell delivers about 10 to 11 mA cm−2. Considering these values,
the following design options are used:

• Historically, the first tandem cells combined two purely amorphous silicon
absorbers; while they do not extend the spectral response compared with a
single-junction cell, they are nevertheless advantageous because of reduced
degradation with respect to a single-junction cell of comparable thickness
(Hanak and Korsun, 1982; Bennett and Rajan, 1988; Lechner et al., 2008).
Figure 4.11 shows the repartition of 15 mA cm−2 between a thin a-Si:H top
and a relatively thick a-Si:H bottom cell.
• The combination of amorphous and microcrystalline cells in a tandem with
high total current appears unfavourable because the top cell sees only a single
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 160

160 C. Ballif, M. Despeisse & F.-J. Haug

ZnO back contact

µc-Si:H bo om cell
SiOx int. reflector
a-Si:H top cell

ZnO front contact

1 µm
glass substrate

ZnO front contact

a-Si:H top cell

ZnO int. reflector

1 µm µc-Si:H bo om cell

ZnO/Ag back contact

polymer substrate

Figure 4.10 Cross-sections through micromorph tandem cells with intermediate reflectors.
The upper panel shows a p-i-n/p-i-n structure on glass covered with a 4 µm thick ZnO front
contact. The intermediate reflector layer appears as a light band between the thin amorphous
and the thick microcrystalline films. The lower panel shows an n-i-p/n-i-p tandem cell on a
periodically textured plastic substrate. In this cell, the intermediate reflector is about 2 µm
thick. The growth sequence of both cells is from bottom to top.

pass of light and the resulting tandem would thus be heavily mismatched. This
limitation was successfully overcome by the introduction of a ZnO intermedi-
ate reflector layer between the component cells, resulting in much better top
cell current (Fischer et al., 1996; Yamamoto et al., 2002; Yamamoto et al.,
2006). An alternative material for the intermediate reflector is silicon oxide
doped with phosphorous (SiOx :P) (Buehlmann et al., 2007). An advantage
of SiOx :P is the possibility of depositing it between silicon layers without
breaking vacuum; this is discussed in more detail in Section 4.6.1.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 161

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 161

1.0
1.0

13.8 mA/cm2
0.8 0.8

13.9 mA/cm2
0.6 0.6
EQE

EQE
0.4 top: bottom:
0.4
70 nm 300 nm
2 2
7.30 mA/cm 7.35 mA/cm
0.2 0.2

0.0 0.0
400 500 600 700 800 400 500 600 700 800 900 1000 1100
Wavelength [nm] wavelength (nm)

Figure 4.11 EQEs of an a-Si:H/a-Si:H tandem cell (left) and of a micromorph cell with an
intermediate reflector layer (right). The micromorph cell has a 240 nm a-Si:H top cell, an
80 nm SiOx intermediate reflector, and a 2.4 µm thick bottom cell. The currents represent
the integrated i sc of the sub-cells. More information on these results can be found in Cubero
et al. (2008) and Boccard et al. (2011), respectively.

Figure 4.12 EQEs of triple-junction cells. The left panel represents a fully amor-
phous a-Si/a-SiGe/a-SiGe cell (Yang et al., 1997), the right panel represents an a-Si:H/
µc-Si:H/µc-Si:H cell with AR-coating (Söderström, 2013). The currents represent the inte-
grated current densities of the sub-cells.

• Thanks to the continuous variation of bandgaps in the a-Six Ge1−x :H alloy sys-
tem, a 10 mA cm−2 a-Si top cell can easily be combined with an a-Si x Ge1−x :H
bottom cell. Cell design must maximise stabilised efficiency by trading off the
germanium content vs. bottom cell thickness trade-off. Going to triple-junction
cells, somewhat higher germanium contents are admissible because of the
further reduced thicknesses of the component cells. Figure 4.12 shows that
triple-junction a-Si:H/a-Six Ge1−x /a-Si y Ge1−y cells can be matched at about
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 162

162 C. Ballif, M. Despeisse & F.-J. Haug

8.3 mA cm−2 (Guha et al., 2000). In the last two approaches no intermediate
reflector is required below the a-Si:H top cell.
• An alternative approach distributes the high current of a microcrystalline sil-
icon cell into a triple-junction cell consisting of an amorphous top cell and a
microcrystalline tandem bottom (a-Si:H/µc-Si:H/µc-Si:H). Using a slightly
thinner and therefore more stable top cell yields current matching between 9
and 10 mA cm−2 (Saito et al., 2005; Yue et al., 2010; Söderström, 2013). The
potential of going to efficiencies beyond 16% with this approach is discussed
in Section 4.6.6.

4.4 Fabrication of device-grade amorphous

and microcrystalline silicon
The deposition of thin films is based on physical vapour deposition (PVD) of
the bulk material (e.g. by evaporation or sputtering) or on chemical processes
using liquid or gaseous precursors. For the production of device-grade thin-film
silicon materials, chemical vapour deposition (CVD) is the most viable solution
today. It involves the decomposition of a mixture of gases containing silicon and
hydrogen, either in a glow-discharge for plasma-enhanced CVD (PECVD) or by
heating a filament for thermo catalytic CVD (cat-CVD), also called hot-wire CVD
(HWCVD). To date, only PECVD processes have reached industrial implementa-
tion. Consequently, their main features are reviewed in detail in this section, from
the principles of PECVD to the impact of the plasma properties on layer quality.
The transition regions from amorphous to microcrystalline silicon are then detailed,
and deposition rate and associated material quality are discussed. The principles
of HWCVD are presented in Section 4.4.4.

4.4.1 PECVD of thin-film silicon

A PECVD reactor (Figure 4.13) typically consists of a vacuum chamber equipped
with an inlet for reaction gases and an outlet to pump out unreacted and reacted
gases. The plasma excitation takes place between two parallel electrodes where
the grounded one serves as support for the substrate. This electrode also acts as a
substrate heater. The other electrode is connected to an electrical power supply and
often includes a distribution system for the reaction gases. This so-called shower
head concept yields a more uniform flux than feeding from the sides. Before deposi-
tion, the desired gas mixture flows into the chamber and a fixed pressure is obtained
by controlling the pumping aperture. Typical gases are SiH4 and H2 , while boron
and phosphorus precursors (B2 H6 , B(CH3 )3 , BF3 or PH3 ) are used for p-type and
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 163

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 163

Figure 4.13 Schematic diagram of a capacitive parallel electrode PECVD reactor.

n-type doping, respectively. A voltage is then applied to the parallel electrodes; free
electrons present in the reactor are accelerated and gas molecules are dissociated
by electron impact ionisation, generating more free electrons which trigger further
impact ionisations until the discharge reaches a stable state. Radio frequency (RF)
excitation predominantly increases the energy of the light electrons while the heavy
ions cannot follow the rapid switching of the alternating field. Thus, RF-PECVD
has the advantage of accessing high-temperature chemistry while remaining at
typical substrate temperatures of 150 to 250◦ C (cold plasma).
Electrons have much higher thermal velocities than ions and can thus reach
the electrodes faster. As a consequence, electric fields develop near the electrodes
which retard electrons and accelerate ions. These regions in front of the electrodes
are called sheaths. They extend over small distances (in the mm range at normal
deposition pressures) and drive the ion bombardment of the surfaces. The standard
RF frequency is 13.56 MHz (as allotted for industrial processes by international
authorities). Alternatively, very high frequencies (VHF, 20–150 MHz) proved use-
ful for achieving higher deposition rates (Curtins et al., 1987; Kroll et al., 1997)
and are today widely used in the thin-film silicon PV industry. Finally, microwave
excitation (2.54 GHz) is also being used in electron-cyclotron resonance PECVD
(ECR-PECVD) (Bae et al., 1998) and for inline microwave PECVD remote plasma
sources (Soppe et al., 2005).
In the plasma bulk, electron impact dissociation of SiH4 and H2 molecules
takes place, producing silicon radicals, SiHx (x = 1–3) and atomic hydrogen, H.
Radicals, molecules and ions can further interact in secondary reactions, such as
higher silane formation from the iterative insertion reaction SiH2 + Sin Hm →
Sin+1 Hm+2 , which triggers powder formation (Bano et al., 2005). Radicals diffuse
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 164

164 C. Ballif, M. Despeisse & F.-J. Haug

from the plasma zone where they are generated towards the reactor walls, the elec-
trodes and the substrate. Growth takes place under a complicated non-equilibrium
process that is governed by the temperature-dependent sticking coefficients, surface
diffusivity of silicon radicals, etching processes by atomic hydrogen, and surface
and sub-surface hydrogen recombination. Under adequate conditions, a silicon film
with a low hydrogen content of around 10 at.% is deposited.

4.4.2 Growth and layer quality: from amorphous to

microcrystalline silicon
In RF discharges, SiH3 radicals are generally considered to be the dominant species
contributing to film growth. Coverage of the growing surface with atomic hydro-
gen allows radicals to diffuse to minimum energy sites, forming a dense random
network.
Dangling bonds available for radical fixation can be created by the removal
of hydrogen from the surface by thermal excitation or by abstraction with a SiH3
radical. Radicals such as SiH2 with short lifetimes can trigger the formation of
higher-order silane-related species which yield a higher density of SiH2 bonds in
the growing a-Si:H films and nano-voids, giving rise to reduced stability against
LID. The density of such nano-voids in an amorphous silicon film can be assessed
via the microstructure factor, as defined in Section 4.2.4. Since the dissociation
energy of SiH4 into SiH3 radicals is lower than the energy required for SiH2 radical
formation, reduced electron temperature is preferred for stable amorphous silicon
growth (Nishimoto et al., 2002). High-quality amorphous silicon materials can be
realised using VHF (Benagli et al., 2009), the triode reactor configuration (Matsui
et al., 2006) and high substrate temperature for enhancing adatom mobility at the
growing surface.
While the formation of polysilanes and powder in the plasma is usually con-
sidered to degrade material quality, these conditions were also reported to yield
high-quality materials owing to the inclusion of nanoparticles in the amorphous
network (Roca i Cabarrocas, 2000). Finally, ion bombardment of the growing
surface is a crucial parameter in amorphous and microcrystalline silicon depo-
sition; ions accelerated in the plasma sheaths strongly affect the growth mecha-
nisms and film properties. Their importance depends on their energy and on the
ion-to-silicon-atom arrival ratio. For silicon, surface or bulk displacement can hap-
pen for high ion energies (∼> 50 eV) (Smets et al., 2007b; Wittmaack et al.,
2003).
The SiH4 /H2 discharge can then also yield microcrystalline silicon growth,
as described by three possible phenomenological models. In the surface diffusion
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 165

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 165

model, high H flux on the surface enhances the mobility of the deposition precur-
sors, strongly influencing the crystalline fraction of the deposited film (Matsuda
et al., 1999). In the selective etching model, amorphous and crystalline phases are
deposited simultaneously with H preferentially etching the amorphous material,
leaving behind the crystallites (Terasa et al., 2000). Finally, the chemical annealing
model postulates that H atoms induce crystallisation of amorphous films (Sriraman
et al., 2002). These three models imply that a high flux of atomic hydrogen H
with respect to the flux of silicon radicals Si favours microcrystalline silicon
growth. A well-known way to reach the transition region between amorphous and
microcrystalline silicon is therefore to use silane that is highly diluted in hydrogen,
but a high H /Si ratio can also be attained from pure silane plasma if sufficient
H is provided by high silane dissociation efficiency. The H /Si ratio required to
reach the transition zone was experimentally assessed (Dingemans et al., 2008)
and also related to the plasma silane concentration c p (Strahm et al., 2007); this
latter parameter is defined by the input silane concentration c and by the deple-
tion fraction of silane in the plasma D as c p = c(1−D). The transition zone was
shown to occur at 0.5% < c p < 1.2%; higher (lower) c p systematically results in
amorphous (microcrystalline) layers.
Figure 4.14 illustrates two optimum deposition windows for the production
of high-quality materials at the transition. In the low silane concentration regime

Figure 4.14 Left: Amorphous to microcrystalline transition zone (shaded region) for vary-
ing silane concentration and depletion fraction, with low silane concentration labelled as
Regime 1 and high silane depletion as Regime 2 (Strahm et al., 2007); right: deposition rate
of device-grade microcrystalline material versus the employed power density (data collected
from different institutes) from low pressure (<1 Torr) RF and VHF regimes (triangles and
squares) to high pressure (>1 Torr) RF and VHF regimes (circles and stars) (Smets et al.,
2008).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 166

166 C. Ballif, M. Despeisse & F.-J. Haug

(c ≤ 2.5%), transition is achieved for low depletion (D < 0.5): the film microstruc-
ture can be controlled with the silane-to-hydrogen gas flow ratio and high depo-
sition rates can be achieved only with the use of high gas flows and high power
densities. In the high silane depletion regime (D > 80%), growth in the transition
region can be achieved only for higher silane concentration (c > 5%). This regime
provides a higher efficiency of silane utilisation and lower hydrogen flows, while
the film structure is determined by the depletion which is, for example, accessi-
ble by controlling the plasma power. It must be noted that high-quality materials,
as assessed by FTPS and FTIR measurements, can be produced in the different
process windows presented in this section, and that device-grade materials typi-
cally have a Raman crystalline fraction between 50% and 70%. However, it was
recently demonstrated that similar high-quality bulk materials, as determined by
FTPS and FTIR measurements, can lead to varying device efficiencies (Bugnon
et al., 2011). To realise high-efficiency devices, the plasma process conditions must
also provide a dense material when the deposition is carried out on a textured sub-
strate in order to limit the creation of localised defects (typically 2D nano-porous
material regions) in the sharp valleys of a highly textured substrate. Very impor-
tantly, the most efficient process conditions for high-quality devices must therefore
provide not only high-quality bulk material but also a minimum density of localised
defects or cracks when deposition is carried out on a textured substrate (Bugnon
et al., 2011; Despeisse et al., 2011).

4.4.3 Deposition rate

Highest material quality and highest resilience to the substrate roughness are
generally realised for both a-Si:H and µc-Si:H at low deposition rate (1–3 Ås−1 ).
However, device-grade material must be deposited at higher rates to be eco-
nomically viable. Higher deposition rates mean increasing precursor and atomic
hydrogen fluxes to the growing surface. This can be done by simply increasing pre-
cursor gas flows and plasma power density; Figure 4.14 shows a plot of deposition
rates versus power density for device-grade µc-Si:H grown in different deposition
regimes. However, increasing the power density enhances ion bombardment, lead-
ing to amorphisation, defect creation and insufficient relaxation of growth precur-
sors. The use of VHF frequencies (40–110 MHz) was shown to yield high-quality
materials at increased deposition rates because it provides higher electron density
for lower electron temperature, enhancing the efficiency of silane dissociation in
the plasma and pushing the onset of powder formation towards higher deposition
rates. Moreover, the energy of the ions that bombard the grounded deposition sur-
face is reduced in VHF, due to decreased sheaths and potentials, giving rise to
increased surface mobility of the deposited species on the growing film.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 167

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 167

In addition to VHF, or alternatively to VHF for standard RF conditions at

13.56 MHz, the so-called high-pressure depletion (HPD) regime can be used;
high-pressure conditions result in long residence time and enhanced silane disso-
ciation, as well as a reduction of ion bombardment energy due to many ion-neutral
collisions during the acceleration of ions in the plasma sheath (Smets et al., 2008).
The HPD approach led to the production of µc-Si:H single-junction cells with
efficiencies of 9.8% and 9.13% at depositions rates of 1 nm s−1 and 2 nm s−1 ,
respectively (Mai et al., 2005; Matsui et al., 2006). Secondary gas phase reactions
and therefore powder production are enhanced at high pressure. Decreasing the
inter-electrode distance in the reactor promotes surface reactions at the electrodes
instead of volume reactions in the plasma, so that by means of this simple geomet-
ric adjustment higher processing pressure can be used without powder formation
(Strahm and Hollenstein, 2010). In addition, a multi-hollow cathode discharge
(MHC), which contains filled cavities on the surface of the electrode, has also been
proposed as a discharge method to produce higher electron density via intense
localised plasmas (Tabuchi et al., 2003). Recently, localised plasma confinement
CVD (LPC-CVD), which makes use of a special cathode with regularly arranged
nozzles and pumping holes, was demonstrated to be highly effective in the pro-
duction of high-quality material at deposition rates of 2 nm s−1, thanks to high
plasma densities (Aya et al., 2011). The use of a narrow inter-electrode gap, VHF,
MHC, LPC-CVD or HPD therefore results in higher material quality at increased
deposition rate.
As we already discussed in Section 4.4.2, high-efficiency devices rely on the
electric performance of good material and on the optics of textured substrates.
This remains a challenge for thin-film silicon: some PECVD processes can lead to
films with low bulk defect density, but their growth kinetics might yield strongly
defective areas near V-shaped regions of the substrate. Optimum plasma conditions
will therefore depend on the exact morphology on which the film is grown.

4.4.4 Hot-wire chemical vapour deposition (HWCVD)

In the HWCVD technique, source gases (e.g. SiH4 , H2 , C2 H2 ) are introduced into
a vacuum chamber with controlled flow ratio and operating pressure. Dissociation
is achieved catalytically in the vicinity of a metallic filament (typically tungsten)
heated to temperatures of 1700–1900◦C. The generated reactive species then induce
film growth on a temperature-controlled substrate facing the filament. Amorphous
silicon materials deposited using HWCVD exhibit low H content of about 3 at.%
and were reported to provide high stability against light soaking (Mahan et al.,
2002). In comparison to PECVD, HWCVD is thermal and catalytic in nature:
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 168

168 C. Ballif, M. Despeisse & F.-J. Haug

there are no ions and electric fields present and consequently there is no ion bom-
bardment of the growing surface. This is particularly important when depositing
the first part of the intrinsic layer in p-i -n microcrystalline junctions, in order to
avoid damaging the critical p-i interface. The use of intrinsic µc-Si:H buffer lay-
ers deposited by HWCVD at the p-i interface of µc-Si:H junctions resulted in
very high Voc (600 mV) and efficiency (Mai et al., 2005). While this deposition
technique is used on an experimental scale for depositing high-quality materials,
high deposition rates (>10 nm s−1) are also achievable (Mahan et al., 2002). How-
ever, the performance of HWCVD-deposited solar cells has not yet surpassed that
of PECVD-deposited solar cells. Moreover, this deposition technique has not yet
been successfully incorporated into large-scale manufacturing facilities, because
of problems with deposition uniformity and filament aging. To date, HWCVD has
thus been employed only at the laboratory scale for thin-film silicon.

4.5 Light management

A common issue for silicon-based solar cells is that the device thickness must be
kept smaller than the optical absorption length. Surprisingly, this is true for all
types of silicon, including wafer-based multicrystalline or single-crystal devices
where surface corrugations in the range of tens of micrometres are used (Redfield,
1974; Campbell and Green, 1987). Immediately after the demonstration of amor-
phous thin-film solar cells, surface textures with accordingly reduced feature size
were introduced for absorption enhancement (Deckman et al., 1983). Absorption
enhancement is equally required for nano- or microcrystalline silicon (Keppner
et al., 1999; Vetterl et al., 2000), and for polycrystalline thin-film silicon (Green
et al., 2004).
Figure 4.15 presents a naïve but intuitive illustration of the light-trapping
process; compared with a cell with flat interfaces, a textured TCO layer scatters
light into the silicon absorber layer. At the back reflector, weakly absorbed light
gets scattered a second time. Very weakly absorbed light can reach the front contact,
where a certain part undergoes total internal reflection because its incident angle

glass
TCO

silicon
back reflector

Figure 4.15 Schematic representation of light trapping by interface textures.

September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 169

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 169

with respect to the interface is beyond the critical angle of total internal reflection.
Textured interfaces thus multiply the effective absorption length of absorber layers
beyond the maximum of two passes through a flat device.
Considering a slab of weakly absorbing material with completely random
scattering interfaces, Yablonovitch proposed an upper theoretical limit of 4n r2 for
the maximum achievable absorption enhancement, n r being the refractive index
of the absorber material (Yablonovitch and Cody, 1982). It was proposed that
deterministic grating structures can exceed the 4nr2 limit in certain wavelength
ranges (Sheng et al., 1983; Gee, 2002; Yu et al., 2010). However, in solar cells,
the experimentally found path enhancement is normally between 15 and 20, much
below the value of 4n r2 of nearly 50 when a refractive index of 3.5 is assumed for
silicon. A possible explanation for this discrepancy is the neglect of the supporting
layers in most of the theoretical derivations (Haug et al., 2011).
For scattering at real surfaces, a detailed explanation of the light trapping pro-
cess is far from straightforward. The interaction of electromagnetic radiation with
matter can be treated with numerous software packages, but the random nature of
typical textures such as those shown in Figure 4.16 is difficult to describe rigor-
ously. Most approaches revert to super-cells, i.e. box-shaped regions that contain
the multi-layer stack of the solar cell above a base area of several µm2 (Rockstuhl

(a) (b)

(c) (d)

Figure 4.16 Surface morphologies of TCOs with small feature size for a-Si:H solar cells
(top) and larger features for microcrystalline cells (bottom). (a) and (c) show AP-CVD-
grown SnO2 :F of the Asahi-U and Asahi-W type, respectively (Taneda et al., 2007); (b) and
(d) show LP-CVD grown ZnO:B films. The growth processes are described in Faÿ et al.
(2005) and Faÿ et al. (2007). Scale bar: 1 µm.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 170

170 C. Ballif, M. Despeisse & F.-J. Haug

et al., 2011). Care must be taken to avoid artefacts because most calculation rou-
tines assume periodic repetition of the cells. The required computing power has so
far limited the use of rigorous models as predictive tools for optimisation; at the
time of writing the calculation of an a-Si:H cell can take hours and a full tandem
cell may take days.
Approximate methods are typically based on scalar scattering theory which
neglects the vectorial nature of electromagnetic fields. Initially developed for
remote sensing of ocean surfaces with radar waves, the scaling to visible light and
nano-textures proved very successful for solar cells (Zeman et al., 2000; Poruba
et al., 2000; Krč et al., 2003). Simulation resembles a ray-tracing approach where
scattering at each interface is described by distribution into a specular and diffuse
part for both reflection and transmission. The diffuse part is normally referred to
as haze; besides its magnitude, a description of the angular intensity distribution
is required (ARS, angle resolved scattering) which is used to describe the average
path enhancement. In scalar scattering theory, the haze depends only on a single
parameter, the root mean square roughness, which is accessible by atomic force
microscopy or optical measurements (Carniglia, 1979; Elson and Bennett, 1979).
Recently, more detailed scattering models based on Fourier theory have been pro-
posed (Dominé et al., 2010).
Adequate surface textures for light management are commonly incorporated
into one of the supporting layers, e.g. the front TCO for p-i -n devices on rigid
glass substrates. Typical examples of fluorinated tin oxide (SnO2 :F) grown by
atmospheric pressure chemical vapour deposition (APCVD) and boron-doped ZnO
(ZnO:B) deposited by low pressure CVD (LPCVD) are shown in the scanning elec-
tron microscope (SEM) images in Figure 4.16. When deposited under conditions
that favour the preferential growth of certain low-energy surfaces, these TCO mate-
rials develop facets that resemble small triangular pyramids. The feature size of
the pyramids can easily be controlled by the film thickness.
Surfaces that are dominated by pronounced V-shaped valleys between the
facets have been found to be detrimental to silicon film growth. Cross-sectional
analysis of amorphous solar cells revealed the growth of defective material above
such valleys because of reduced adatom mobility (Sakai et al., 1990). Figure 4.18
shows that this effect is pronounced in microcrystalline cells, resulting in severe
Voc and FF losses (Nasuno et al., 2001a; Li et al., 2008; Python et al., 2008). The
defective zones have also been identified to provide diffusion paths for oxygen and
other impurities (Python et al., 2010).
Figure 4.17a shows a surface where the initially pyramidal texture of the
LP-CVD ZnO shown in Figure 4.16d has been modified by a plasma treatment that
converts the V-shaped valleys between the facets into rounded valleys with smooth,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 171

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 171

(a) (b)
Figure 4.17 Surfaces with rounded features. (a) corresponds to the surface shown in Fig-
ure 4.16d after plasma treatment (Bailat et al., 2006); (b) illustrates the crater lime etch pits
of sputtered ZnO etched for 30 s in diluted HCl (Kluth et al., 1999). Scale bar: 1 µm.

Figure 4.18 Top: top-view SEM images of front electrodes (three TCOs with shapes mod-
ified from V to U by plasma treatment); bottom: TEM cross-sections of microcrystalline
silicon layers deposited on the TCOs. There is a clear correlation between the ZnO surface
morphology and the formation of defective zones within microcrystalline absorber films.
Python et al. (2008) give more details.

U-shaped bottoms. Unfortunately, U-shaped substrates yield reduced light trapping

compared with V-shaped ones; optimisation of cell efficiency must therefore trade
off high Voc and FF against reduced i sc (Bailat et al., 2006). Alternatively, valleys
with rounded bottoms can be obtained by wet chemical etching; Figure 4.17b shows
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 172

172 C. Ballif, M. Despeisse & F.-J. Haug

the crater-like texture of etch pits obtained on sputtered ZnO films after etching in
diluted hydrochloric acid (HCl) (Kluth et al., 1999).
An approach that completely separates the texture for light-confinement from
the growth process has been suggested for the n-i - p structure (Sai et al., 2011).
Starting from a flat silver reflector covered with ZnO, a relief grating is etched
into the ZnO and filled with n-doped amorphous silicon. Subsequently, the stack
is polished in order to provide a flat growth template for the microcrystalline cell.
Owing to the complete absence of valleys that induce the growth of defective zones,
the resulting device showed the same high voltage as a flat reference cell. However,
the n-doped filling-layer resulted in parasitic absorption between 650 and 750 nm.
An alternative design with an undoped filling-layer in combination with a triple-
junction device successfully overcame this limitation (Söderström et al., 2012a;
Söderström, 2013).
Traditionally, n-i - p solar cells relied on the natural textures that develop when
metals like aluminium or silver are deposited on heated substrates (Hirasaka et al.,
1990; Banerjee and Guha, 1991). Because of its relative inertness, the noble metal
silver offers an additional handle on texturing if small quantities of aluminium
are added to the silver sputtering target: by adding oxygen to the gas mixture
during sputtering from such an alloy target, the interface texture is modified by
the inclusion of alumina (Al2 O3 ) crystallites in the growing film (Franken et al.,
2007). The formation of the texture typically requires temperatures of around 300
to 400◦ C; flexible substrate materials are therefore limited to stainless steel (Baner-
jee and Guha, 1991) or polyimides (Takano and Kamoshita, 2004). However, in
order to fully exploit the promise of reduced production cost with roll-to-roll
processing, truly low-cost substrates are desirable. Substrates like polyethylene
thus require a different strategy where the substrate itself is textured; this can
be achieved by hot embossing (Fonrodona et al., 2005) or by imprinting into a
UV-curable polymer (Bailat et al., 2005; Söderström et al., 2010). The interface
texture is then transported into the device by conformal growth. An example of
a microcrystalline n-i - p cell on a periodic substrate texture obtained by roll-to-
roll embossing is shown in the transmission electron microscope (TEM) image
in Figure 4.10; Figure 4.19 shows in more detail how the moderate texture of the
grating substrate is perfectly reproduced by the sputtered silver and ZnO back con-
tact. However, throughout the microcrystalline absorber layer, the initial sinusoidal
texture develops into a structure with a much larger radius of curvature and pinched
valleys.
This change of texture is an issue for tandem cells because in the n-i - p config-
uration the amorphous sub-cell is deposited after the microcrystalline cell. While
the large features of the grating substrate can deliver acceptable isc of about 23 to
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 173

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 173

LP- CVD ZnO

p-layer

i-layer

n-layer

ZnO /Ag

polymer substrate

Figure 4.19 TEM cross-section through a microcrystalline cell on a periodically textured

back reflector obtained by embossing into a UV-curable polymer (Haug et al., 2009).

25 mA cm−2 in bottom cells with thickness varying between 1.2 and 2 µm, the flat-
tening of the interface texture is clearly inadequate for the top cell. The introduction
of a thick, textured intermediate reflector can ameliorate this situation: thanks to
the texture that ZnO develops when deposited by LP-CVD, Figure 4.10 shows
that the top cell can be grown on an adapted texture that is completely independent
from the bottom cell (Söderström et al., 2009). Tandem cells in n-i - p configuration
with stabilised efficiency up to 11.6% have been demonstrated with this approach
(Biron et al., 2013).

4.6 Advanced device architecture and record devices

In thin-film silicon PV devices, continuous developments of the constituent layers,
interfaces and of their processing conditions are essential to further improve perfor-
mance. Thanks to the intense industrialisation efforts of the past decade, conversion
efficiencies have been significantly increased with the development of efficient light
management schemes and with the introduction of low-bandgap partners to a-Si:H
to realise multiple-junction devices. At the module level, the successful incorpora-
tion of microcrystalline silicon in a-Si:H/µc-Si:H tandem modules triggered direct
efficiency improvements. In 2010, progress in the control of the multiple thin-films
that constitute such tandem devices over large deposition areas led to the production
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 174

174 C. Ballif, M. Despeisse & F.-J. Haug

of the first thin-film silicon modules on glass with stabilised efficiencies surpassing
the 10% barrier (Klein et al., 2010; Kratzla et al., 2010). Modules with up to 13.1%
initial efficiency and 10.8% stabilised efficiency were demonstrated in 2011 (Kluth
et al., 2011).
On metal foil, 12% initial efficiency and 11.2% stabilised efficiency were
reported (Banerjee et al., 2011a) for a 400 cm2 triple-junction cell (a-Si:H/a-
SiGe:H/µc-Si:H). These results demonstrate that major bricks for efficient mass
production over large areas are in place. To further increase the efficiency, and thus
the impact, of thin-film silicon technology, advanced light management schemes,
novel nanomaterials and innovative device architectures are currently being devel-
oped. Some of these recent advances are reviewed in this section, together with the
best efficiencies obtained so far at the cell level.

4.6.1 Advanced light management with intermediate reflectors

In monolithically interconnected multiple junctions, the device current is limited by
the junction with the smallest current generation. While tailoring the thicknesses
of the different absorbers adjusts the photocurrent in each junction, additional
layers can also be inserted to selectively reflect photons, thus altering the light
distribution in the cell. In micromorph solar cells, the a-Si:H top cell must be kept
thin (<300 nm) to minimise its degradation. To realise larger device currents, an
intermediate reflecting layer (IRL) can be integrated between the top and bottom
cells, to reflect photons not fully absorbed in the first pass through the top cell
(wavelengths of 550–700 nm), while ensuring transmittance of red and IR light to
the bottom cell (Yamamoto et al., 2004).
The IRL must fulfil three main roles:
• It must have sufficient transverse (out-of-plane) conductivity to avoid electrical
losses caused by its series resistance.
• In order to reflect, it must be fabricated from a low refractive index material.
Fresnel reflection occurs due to the index contrast between films; the refractive
index of silicon nr Si is approximately 3.8 at 600 nm, and n rIRL should therefore
ideally be below 2 at this wavelength.
• In addition to the refractive index, the thickness of the IRL controls the intensity
and spectral selectivity of back reflection and forward transmission into the
bottom cell via interference effects (see Figure 4.20).
The first efficient IRLs were realised using TCO layers such as ZnO (Fischer
et al., 1996). More recently, phosphorus-doped silicon oxide materials (n-SiOx :P)
with n rIRL = 1.6−2 at 600 nm and with sufficient transverse conductivity were
synthesised by PECVD, giving an in situ IRL alternative (Buehlmann et al., 2007;
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 175

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 175

Figure 4.20 Left: SEM image of a focussed ion beam cross-section of a micromorph
tandem cell with LP-CVD ZnO front and back contacts, a 250 nm thick a-Si:H top cell, a
40 nm thick silicon-rich silicon oxide IRL and a 1.1 µm thick bottom cell. The nanometre-
wide filamentary structure of the silicon phase in the SiOx material is illustrated in the
inset EFTEM image. Courtesy of Peter Cuony and Duncan Alexander, EPFL; details can
be found in Cuony et al., (2011). Right: spectral response of micromorph tandem cells with
no IRL, a 40 nm thick IRL and a 100 nm thick IRL: increasing the IRL thickness results in
enhanced (reduced) response in the 500–800 nm range in the top (bottom) cell. Courtesy of
Peter Cuony, IMT PV Lab, Neuchâtel.

Das et al., 2008). The best materials were shown to have a mixed-phase structure
with nanometre-wide silicon filaments embedded in an amorphous silicon oxide
matrix. This advanced nanostructure can be imaged by energy-filtered transmission
electron microscopy (EFTEM), as shown in Figure 4.20 (Cuony et al., 2012).
Light regions correspond to the doped silicon-rich phase that conducts, whereas
dark regions represent the silicon oxide phase that produces the material’s low
refractive index.
For optimum light harvesting in the top cell, the IRL must be combined
with efficient light scattering at the cell entrance, or it must be structured to
provide a textured interface to the incident photons, as shown in Figure 4.10
for a tandem device in the substrate configuration with an asymmetric reflector
(Söderström et al., 2010). In the superstrate configuration, i sc in 250 nm thick
top cells are typically about 11 mA cm−2 , while the integration of an IRL can
increase i sc up to 14 mA cm−2 , demonstrating the impact that such a thin layer can
have on efficiency. While n-type silicon oxide nanomaterials are straightforward
IRL solutions adaptable to mass production, consecutive approaches such as dis-
tributed Bragg reflectors and more sophisticated structures such as 3D photonic
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 176

176 C. Ballif, M. Despeisse & F.-J. Haug

crystals are also under development at the cell level to further control the spectral
selectivity and the light distribution (Bielawny et al., 2008). Nano-patterned low
refractive index dielectric materials (n r ∼ 1.3–1.5 at 600 nm) were also demon-
strated to have a high potential as efficient IRL (Boccard et al., 2013), with in
that latter case even a potential for altering the roughness of the surface on which
grows the µc-Si:H bottom cell, so as to provide in addition a smoothening of the
surface.

4.6.2 Efficient light in-coupling

To maximise current generation, reflection and absorption losses at the entrance
of the solar cell must be minimised to ensure efficient light in-coupling to the
absorber layers. In the superstrate configuration, the refractive indices of glass
(nr ∼ 1.5), TCO (n r ∼ 2) and silicon (nr ∼ 3.4−4) lead to optical reflections
and to typical transmission losses of 4%, 2% and up to 11% respectively at
the air/glass, glass/TCO and flat TCO/silicon interfaces. Textured interfaces and
transparent antireflective (AR) coatings have been developed to minimise these
reflections.
Textured TCO not only efficiently scatters light but also provides a refractive
index grading and multiple reflections of the incoming light, reducing the reflection
losses at the TCO/silicon interface. AR coatings with a refractive index interme-
diate between TCO and silicon can also be used. While titanium dioxide (TiO2 )
layers were shown to be efficient (Fujibayashi et al., 2006; Das et al., 2008), high-
gap silicon layers adjoining the TCO could also provide an AR effect by index
grading. Similarly, texturing of the glass or of a lacquer imprinted on the glass
surface (Escarre et al., 2012; Ulbrich et al., 2011) and application of AR coatings
with intermediate refractive indices can be successfully used at the air/glass and
glass/TCO interfaces.
Absorption in the non-photoactive front TCO and front p-type silicon layers
must also be minimised. Doped layers are highly defective and hardly contribute
to the photocurrent, so they are designed to be very thin (about 10 nm). Advanced
transparent doped materials are under development for use as efficient window lay-
ers. Mixed-phase p-type silicon-rich silicon oxide materials (Cuony et al., 2010)
and n-type µc-SiC:H layers (Chen et al., 2010) were recently introduced as effi-
cient window layers, leading to EQEs above 70% at 400 nm in µc-Si:H solar cells
and i sc = 29.6 mA cm−2 for a 2 µm thick device (Chen et al., 2011). Similarly,
minimum absorption must be ensured in the front TCO layer. In devices implement-
ing a µc-Si:H junction, the TCO has to be highly transparent in the 400–1100 nm
wavelength range, i.e. up to the near infrared region (NIR); this imposes restric-
tions on the carrier density as increased free carrier absorption (FCA) leads to a
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 177

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 177

reduction of red and NIR light transmission. High carrier mobility is essential to
ensure low sheet resistances (Rsheet ) for a TCO with reduced free carrier density.
Recent advances in the synthesis of doped ZnO have yielded mobilities of 40–
60 cm2 /Vs and absorption levels below 5% over the full spectral range for a layer
with Rsheet = 10 /sq, and around 2% for Rsheet = 20 /sq (Ruske et al., 2010;
Boccard et al., 2010).
Alternatively, hydrogen-doped indium oxide (In2 O3 :H) was recently reported
as a highly transparent TCO; even NIR absorption losses are low (<1%) thanks to a
high mobility of up to 120 cm2/Vs (Koida et al., 2010). These layers do not provide
a native or post-process texture, but can be used in next-generation superstrates as
thin and highly transparent TCO layers.

4.6.3 Advanced superstrates with enhanced transparency and

multi-scale texturing
Advanced superstrates with multi-scale texturing are being developed to combine
high transparency and conductivity for optimum light-in coupling, efficient light
trapping in the full spectral range and the growth of high-quality silicon. Conduction
and texture can be decoupled via a bilayer approach: a part of the TCO layer is
optimised for conduction with no restriction on its texture. Then, adequate light
scattering is achieved with a transparent film whose texture is optimised. Such
bilayer TCOs demonstrated increased cell performance thanks to better control
of transparency (Selvan et al., 2006; Ding et al., 2011). In a different approach,
textured glass can be used in combination with a thin In2 O3 :H TCO layer. As
an alternative to glass texturing, Figure 4.21 shows that the texture can equally
be applied to a transparent lacquer by nanoimprint lithography (Battaglia et al.,
2011a). The efficient light-scattering properties of the textured transparent lacquer
combined with the high transparency and conductivity of In2 O3 :H, results in a high
summed current of 26 mA cm−2 in a micromorph cell using a thin µc-Si.H bottom
cell (1.1 µm) and no AR coating.
While the textures typically implemented in thin-film silicon solar cells are dic-
tated by the underlying growth and etching kinetics of the textured TCO, nanoim-
print lithography on UV lacquer allows one to replicate an arbitrary structure made
from a transparent or opaque arbitrary master material. A large variety of light-
trapping structures can thus be designed and combined with a highly transparent
TCO. Recently, arbitrary textures from master materials were also demonstrated
to be transferable directly to ZnO front electrodes; this nanomoulding of ZnO
electrodes enables a similarly wide range of photonic structures to be directly
implemented in front electrodes in thin-film silicon solar cells (Battaglia et al.,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 178

178 C. Ballif, M. Despeisse & F.-J. Haug

Figure 4.21 Left: SEM cross-section of a-Si:H/µc-Si:H tandem solar cell, developed on
a glass superstrate implementing a nanoimprinted UV lacquer that provides large features,
a thin In2 O3 :H layer allowing for conduction and high transparency, and a low-doped LP-
CVD ZnO layer adding small pyramidal features. Courtesy of Mathieu Boccard; details
in Boccard et al. (2012). Right: tilted SEM top view image of this multi-scale textured
superstrate. Courtesy of Mathieu Boccard, IMT PV Lab, Neuchâtel.

2011b). These recent developments are important tools to determine and imple-
ment the most promising light-trapping schemes on the glass or in the electrode
itself. Some examples of periodic to random structures transferred to the electrode
or replicated on the glass substrate are shown in Figure 4.22.
Multi-scale textures are particularly interesting approaches for tandem cells
since they can combine efficient light scattering by small and large textures for
the amorphous and microcrystalline cells, respectively. Care must be taken to keep
morphologies that are adapted for the growth of the µc-Si:H cell. Two-step etch-
ing procedures on sputtered ZnO result in a surface morphology combining large
craters with smaller features (Hüpkes et al., 2010). Also, multi-scale texturing
of CVD-grown TCO (ZnO and SnO2 ) was demonstrated (Taneda et al., 2007).
A complementary approach is the combination of textured TCO with textured
glass (Bailat et al., 2010; Kroll et al., 2011) or with nanoimprinted lacquer on
glass (Boccard et al., 2012; Bessonov, 2011).
Recently, micromorph cells with record stabilised efficiencies of 11.9% (Bailat
et al., 2010) and 12.3% (Kroll et al., 2011) were demonstrated by combining
textured glass with textured ZnO. Additionally, superstrates combining both multi-
scale textures and bilayer TCOs are under development, and Figure 4.21 shows an
a-Si:H/µc-Si:H tandem cell realised on such a substrate. Superstrates with further
combinations of multi-layer TCOs and multiple textures produced in either TCO or
glass by direct texturing or nanoimprinting are still under study to further enhance
the efficiency of thin-film silicon solar cells.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 179

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 179

Figure 4.22 SEM images of (a) as-grown random pyramidal LPCVD ZnO; (b) thermally
roughened silver; and (c) alkaline-etched crystalline silicon, all realised by nanoimprinting
in a transparent lacquer. Courtesy of Jordi Escarre and Karin Söderström, IMT PV Lab,
Neuchâtel; details in Söderström et al. (2010). SEM images of (d) as-grown thick LP-CVD
ZnO; (e) a one-dimensional periodic grating; and (f) quasi-periodic hexagonal dimple pattern
of anodic oxidation of aluminium, all transferred to the surface of ZnO front electrodes via
nanomoulding. Courtesy of Corsin Battaglia; details in Battaglia et al. (2011b). (g): a-Si:H
solar cell on a nanomoulded ZnO front electrode with a surface texture corresponding to (f).

4.6.4 Cell architecture, alternative absorber layers

and advanced designs
The main driver in thin-film silicon module production is currently the a-Si:H/µc-
Si:H tandem architecture. Recently, enhanced cell and module resilience to sub-
strate roughness was achieved with the implementation of novel doped layers and
with the optimisation of µc-Si:H growth conditions (Bailat et al., 2010; Despeisse
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 180

180 C. Ballif, M. Despeisse & F.-J. Haug

et al., 2011; Bugnon et al., 2011). For instance, the implementation of silicon-rich
silicon oxide-doped layers was demonstrated to enhance the electrical performance
of thin-film silicon solar cells grown on highly textured substrates. These doped
layers were shown to mitigate the impact of localised defects on cell electrical
performance (Despeisse et al., 2011). Section 4.5 discussed how such defective
areas arise when silicon is deposited on highly textured surfaces exhibiting sharp
valleys (Python et al., 2008).
Moreover, plasma process conditions during microcrystalline silicon growth
were also shown to affect the quality and density of these localised defect regions
that occur over sharp underlying surface features (Bugnon et al., 2011), demonstrat-
ing that globally dense µc-Si:H material has to be achieved on textured substrates
to maximise efficiency. Thin devices were shown to have high potential (Schicho
et al., 2010), and the enhanced resilience of the solar cells to the substrate texture
allowed for the development of high-efficiency thin superstrate devices, with 11.3%
and 10.9% stable cells realised using 1.1 µm and 0.8 µm thick bottom cells, respec-
tively (Despeisse et al., 2011; Bailat et al., 2010). Such thin cells give stabilised
efficiencies higher than 10% in modules (Kratzla et al., 2010), and they trigger cost
reductions because of reduced deposition and reactor cleaning times, resulting in
a promising trade-off between efficiency and costs for the micromorph approach.
Alternatively, high-efficiency concepts with folded 3D cell designs (Vanecek
et al., 2010; Naughton et al., 2010) are under consideration for a-Si:H single junc-
tions and micromorph tandem devices. Novel nanostructures such as nano-columns
or nano-/micro-holes on superstrates or substrates could open possibilities; the cell
layers would be optically thick but electrically thin (cf. Figure 4.23). Accordingly,
the resulting folded cell would yield thin layers with very low LID (Vanecek et al.,
2010).

Figure 4.23 SEM image of aligned nano-pillars (left panel (Rizal et al., 2013)) and FIB
cross-section through n-i- p solar cells deposited on these pillars. The inset illustrates non-
conformal growth with preferential deposition of reflector, silicon and ITO layer at the tips
(Naughton et al., 2010).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 181

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 181

An alternative approach to higher stabilised efficiencies is the use of triple-

junction cells; however, the additional junction comes at the price of enhanced
spectral sensitivity and further increased cell thickness, i.e. higher costs or harsher
requirements on absorber deposition rates. The most efficient combinations demon-
strated so far consist of a-Si:H/a-SiGe:H/µc-Si:H or a-Si:H/µc-Si:H/µc-Si:H
devices. Both have been successfully developed in substrate n-i - p configuration
(Xu et al., 2010) by the company Unisolar. They showed stabilised cell efficiencies
beyond 12.5% (Yue et al., 2010) which translate to prototype module efficiencies
above 11% (Banerjee et al., 2011a). In the p-i -n configuration, a-Si:H/µc-Si:H/µc-
Si:H triple junction cells with stabilised efficiency of 13.4% have been presented
(Kim et al., 2013).
As outlined in Section 4.3.2, the matched cell current is lower than in tandem
junctions. This can be exploited to achieve higher stability by further thinning the
a-Si:H top cell, or by depositing it yet closer to the transition between the amorphous
and microcrystalline phases, which generally increases the bandgap and enhances
its stability against light-induced degradation at the same time (Kamei et al., 1999;
Koval et al., 1999). Research on absorber layers alloyed with carbon or oxygen for
alternative top cells is also being conducted (Yunaz et al., 2009; Inthisang et al.,
2011). Complementary research addresses materials with lower bandgap such as
µc-Si1−x Gex :H (Matsui et al., 2010) which successfully enhances the conversion
of the IR part of the spectrum. A wide range of thin-film silicon absorber materials
is therefore in development (Konagai et al., 2011), and they may lead to novel
combinations in triple- or quadruple-junction thin-film silicon solar cells to reach
beyond 16% stabilised efficiency.

4.6.5 Record devices

Table 4.2 lists the highest certified efficiencies at the time of writing (mid-2013).
Single-junction cells based on a-Si:H and µc-Si:H absorber layers reached efficien-
cies of 10.1% and 10.7% on device areas of 1 cm2, respectively. The best tandem
and triple-junction cells have reached stabilised efficiencies of 12.3% and 13.4%,
respectively. Table 4.3 gives an (incomplete) list of notable cells. These results
are quoted as reported by the respective authors, generally without certification.
Most results refer to cells in their initial states, except where the authors give their
stabilised values.

4.6.6 Roadmap for higher efficiencies

To survive the competition of crystalline silicon, thin-film silicon technology needs
to go faster towards higher efficiencies. Based on the findings of the last decade and
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 182

182 C. Ballif, M. Despeisse & F.-J. Haug

Table 4.2 Overview of certified record efficiencies in cells and modules.

Eff. FF isc Comments/
Source Structure (%) V oc (V) (%) (mA cm−2 ) Reference

Certified record cells (stabilised)

Oerlikon a-Si 10.1 0.886 67 16.75 p-i-n on LP-CVD
Solar ZnO
(Benagli et al.,
2009)
AIST a-Si 10.1 0.906 70 16.05 p-i-n on Asahi-VU
(Matsui et al.,
2013)
AIST µc-Si 11.0 0.542 73.8 27.44 (Sai et al., 2014)
IMT a-Si/µc-Si 12.6 1.382 71.3 12.83 (Boccard et al.,
2014)
United a-Si/ 12.1 2.297 69.7 7.56 n-i- p on steel,
Solar a-SiGe/a-SiGe 0.27 cm2
(Yang et al., 1997)
United a-Si/ 12.5 2.01 68.4 9.1 n-i- p on steel,
Solar µc-Si/µc-Si 0.27 cm2
(Yue et al., 2010)
LG a-Si/ 13.44 1.96 71.9 9.52 p-i-n
electronics µc-Si/µc-Si (Kim et al., 2013)
Certified record (mini-)modules (stabilised)
TEL a-Si/µc-Si 12.2 205.2 66.9 1.22 A Gen5 module,
1.43 m2 (TEL,
2014)
United a-Si/ 11.0 2.34 70.3 6.65 n-i- p on steel,
Solar a-SiGe/a-SiGe 462.3 cm2 , large
area cell
(Banerjee et al.,
2011a)
United a-Si/ 12.0 1.95 69.2 8.93 n-i- p, enc.
Solar µc-Si/µc-Si 399.8 cm2 cell
(Banerjee et al.,
2011b)

on the requirement to ‘decouple’ the electrical from the optical transport paths, a
simple calculation based on the best top and bottom cell realised so far shows that
tandem solar cells are probably limited to stabilised efficiencies between 14% and
15%. For two reasons, triple-junction solar cells such as a-Si:H/µc-Si:H/µc-Si:H
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 183

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 183

Table 4.3 Overview of noticeable efficiencies (reported by the authors, not certified).

Eff. FF isc Comments/

Source Structure (%) V oc (V) (%) (mA cm−2 ) Reference

United a-Si/ 16.3 2.242 77 9.43 n-i- pcell, 0.27 cm2

Solar a-SiGe/ (initial) (Yan et al. 2011)
µc-Si
Kaneka Triple- 15 2.28 73.5 8.93 (Yamamoto et al.,
junction (initial) 2006)
Kaneka a-Si/ 14.7 1.41 72.8 14.4 (Yamamoto et al.,
µc-Si (initial) 2004)
Oerlikon a-Si/ 14.4 1.419 70.7 14.35 (Kroll et al., 2011)
Solar µc-Si (initial)
IMT a-Si/ 14.1 1.41 71.5 14 on IO/LP-CVD,
Neuchâtel µc-Si (initial) 1.2 cm2
(Ballif et al., 2011)
IMT a-Si/ 13.0 1.94 70.3 9.5 n − i − p on
Neuchâtel µc-Si/ (stable) polished substate
µc-Si (Söderström, 2013)
Sanyo a-Si/ 13.5 1.41 75.4 12.7 (Terakawa et al.,
(ini)
µc-Si 12.2 (st) − − − 2011)

are an attractive alternative to go further; as discussed in Section 4.3.2, matching

at a lower current relieves several constraints of cell processing. The intermediate
reflector is no longer needed, and the thin top cells can be made with a very high
Voc > 1 V (Yan et al., 2003). The µc-Si:H middle cell is also thin enough to use
a recently demonstrated design for thin cells with high Voc > 600 mV (Hänni
et al., 2013). With slightly improved bottom cells and the latest advances in light
management, a stable device efficiency of 16.5% could thus be targeted with Voc
of 2.15 V, FF of 72% and i sc of 10.7 mA cm−2. Alternatively, the use of Ge alloys
for the middle cell could lead to similar results. Moving to significantly higher
efficiency will probably require the development of novel absorber materials with
higher quality, and the use of innovative passivating contacts that extract the full
voltage potential of the materials.

4.7 Industrialisation and large-area production technology

4.7.1 From small-area to large-area devices
The FPD sector showed impressive development through 2004. Indeed, at that time,
more than 10 million square metres were coated annually in the manufacturing of
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 184

184 C. Ballif, M. Despeisse & F.-J. Haug

liquid crystal displays (LCDs); the TFTs in these displays are based on amorphous
silicon as well as silicon nitride. Moreover, the FPD industry demonstrated a con-
tinuous decrease of display production costs (20% per m2 annually). It was hence
not a surprise that equipment makers, who had made a significant investment in
the development of coating tools, identified thin-film silicon solar technology as an
emerging growth market for their products, with a focus on PECVD equipment that
could be used for layer deposition. This led companies such as Oerlikon, Applied
Films (later purchased by Applied Material), Ulvac and Leybold Optics to enter
the PV market, trying to transfer several of the processes developed in research
laboratories to mass production. Most manufacturers work on Gen5 scale (1.4 m2)
but Applied Materials has also delivered tools for Gen8.5 (5.5 m2). The transition
from a mobility-controlled device (as in a TFT) to a lifetime-controlled device (as
in PV) was, however, not trivial for three related reasons:

• In a-Si:H and µc-Si:H, the material quality depends very much on the deposi-
tion conditions and rate. Reasonable rates at acceptable quality often required
equipment adaptation (e.g. narrower inter-electrode gaps for higher pressure
or measures to avoid standing-wave patterns for the case of VHF-excitation).
• By definition, µc-Si:H is a material deposited close to a phase transition; con-
sequently its homogeneous deposition on >1 m2 required significant hardware
adaptation in order to compensate for gas flow and electromagnetic inhomo-
geneities, as well as boundary effects.
• The two first steps had to be taken in conjunction with the development of
the TCO front contact (not fully developed at that time), the back contact and
laser scribing.

While solving these issues, several equipment makers were able to sign con-
tracts to sell ‘turnkey’ production lines, making thin-film silicon the first technology
offering an easy entrance into the PV market, while covering the full chain (from
glass to module). Indeed from 2006 to 2008, concerned about silicon feedstock
problems for c-Si cells manufacturers, over 50 companies announced their entrance
into thin-film silicon module manufacturing. Even though solving the three issues
noted above was more demanding than initially expected, in 2011 several com-
panies were able to manufacture modules with total-area stabilised efficiencies of
6.5–7.3% and 9–10% for a-Si:H and a-Si:H/µc-Si:H modules on glass, respec-
tively. Note that ‘total-area’ refers to a widely used module size of 1.4 m2 which
includes a non-active edge area of 6–8%. Commercial a-Si:H/a-Si1−x Gex :H/a-
Si1−y Ge y :H triple-junction modules on flexible substrates reach 8.2% stabilised
efficiencies (designated aperture area); a comparison with the record devices in
Table 4.2 shows that there is a relatively little gap between small-scale cell results
and m2 sized modules, at least for p-i -n cells.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 185

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 185

Probably no other technology has made as big an effort at transferring labora-

tory results to production-size modules, and a large part of this was facilitated by
FPD experience. The last half-decade has seen the resolution of all major produc-
tion issues and a huge amount of knowledge was gleaned on up-scaling issues. The
basic manufacturing process sequence on glass was presented in Figure 4.2. We
review some of its key aspects here and reserve a specific section for roll-to-roll
processes.

4.7.2 TCO deposition on glass

The front contact made from transparent conducting oxides (TCOs) is a key element
for modules prepared on glass. Its sheet resistance Rsheet (5–20 /sq) and the
segment width W (typically 5–15 mm) control the relative TCO resistive power
losses according to the formula P/P= Rsheet × W2 × Imax /3Vmax , where Imax and
Vmax are the current and voltage at the maximum power point of the I −V curve.
Additional losses linked to the interconnection areas (from segment to segment)
can occur as well. The TCO must scatter light efficiently to achieve high current
in devices (30–50% current increase compared with flat substrates), but if the free
carrier concentration required for lowering the sheet resistance is too high, strong
parasitic FCA will take place.
Usually, reducing the doping and thickening the layer help reduce FCA,
because it scales super-linearly to the TCO carrier concentration (Faÿ et al., 2010).
At large thicknesses, the shape of the TCO grains can lead to issues with the growth
of the subsequent devices, as detailed in Section 4.5. Advanced device process-
ing (e.g. using specially tailored doped layers, as discussed in Section 4.6) can be
used to mitigate such effects. A drawback of reducing the free-carrier concentra-
tion too much is a stronger sensitivity to moisture ingress; humidity creates trap
states at the grain boundaries, leading to a conductivity decrease and thus adequate
encapsulation schemes must be considered.
There are currently four major routes used for the fabrication of TCOs on large
area substrates (>1 m2 ). Some of the advantages and weaknesses of the various
approaches are mentioned in Table 4.4. Figure 4.16 shows images of the various
TCOs. The four routes are:
• Inline deposition of SnO2 :F by APCVD: the TCO deposition is directly inte-
grated into the cooling stage of the float glass line (McCurdy et al., 1999).
Deposition takes place between 640–590◦C (depending on the location on the
line, T > 600◦ C means the glass is still liquid). The deposition rate is around
100 nm/s. Commercially used precursors are tin-tetrachloride, dimethyl-tin-
dichloride and butyl-tin-dichloride. Usually a barrier layer against sodium
migration is included in the stack as well as a thin AR layer between the glass
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 186

186 C. Ballif, M. Despeisse & F.-J. Haug

Table 4.4 Overview of advantages and challenges linked to the use of various front TCOs
in amorphous and a-Si:H/µc-Si:H production.

TCO Advantages Challenges

SnO2 :F inline Minimal hardware costs, medium Process control, high

to high current in top cell transparency
SnO2 :F off-line Well-controlled process, medium High transparency and scattering,
to high current in top cell high temperature process
Sputter-etched Substrate suited to the growth of Uniformity, cost of hardware and
ZnO microcrystalline cells, good targets, lower current in top
light scattering for total current cell, even with IRL
LP-CVD-ZnO High current in top cell, very Control of subsequent PECVD
high with IRL, doping and growth, encapsulation
scattering easily tuneable

and the TCO. Typical TCO thickness is 700 nm. After SnO2 :F deposition, the
glass is cut and the edges are ground.
• Off-line deposition of SnO2 :F by APCVD: deposition on glass plates that are
reheated to typical temperatures of 540–590◦C. Deposition rates are 50 nm/s
and the same precursors can be used as for the inline process. Typical TCO
thickness is 700 nm.
• Off-line sputter-etched ZnO: aluminium-doped ZnO is sputtered and subse-
quently wet-etched to create crater-like nanostructures (Kluth et al., 2003).
ZnO rotary ceramic targets containing 1–2 wt.% of Al2 O3 are used for DC
sputtering (Zhu et al., 2009) at typical temperatures of 300–370◦C. After depo-
sition, the cooled substrates are etched in a 0.3–0.5% HCl solution. Typical
TCO thickness is 700 nm.
• Off-line deposition of ZnO by LPCVD (Vogler et al., 2008): deposition typi-
cally takes place at 170–200◦C, at pressures of a few mbar, using di-ethyl-zinc
(DEZ, (C2 H5 )2 Zn), water and di-borane for doping. Typical deposition rates
are 2–4 nm per second, and typical layer thickness is 1.4–2 µm.

Notably, the processes run by research laboratories might differ from those
employed by industry. At the laboratory level, all of these TCOs have demonstrated
good results. Between 2009 and 2011, certified world record efficiencies were
reported for amorphous cells deposited on LP-CVD ZnO (Benagli et al., 2009)
and for micromorph cells on LP-CVD ZnO grown on textured glass (Bailat et al.,
2010). Micromorph modules with efficiencies of over 10% have been reported
for most of the above TCOs (Klein et al., 2010; Kratzla et al., 2010; Aya et al.,
2011).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 187

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 187

4.7.3 Deposition of thin-film silicon by PECVD

This section concentrates on deposition of the silicon-based layers by PECVD
because this is the only technique implemented in production. The discussion is
moreover focussed on microcrystalline layers, which are thicker and more chal-
lenging to deposit with high quality and low cost. The challenges involved in up-
scaling PECVD processes from the laboratory to production can be summarised
as follows:

• A major issue is plasma uniformity, which is particularly critical for materials

close to the a-Si:H to µc-Si:H phase transition. On large area, the gas com-
position (e.g. ratio of Si radical H) can be non-uniform because of educts that
are not incorporated in the layers. Uniformity can be influenced by varying
the gas distribution (e.g. different gas flow at the border compared with the
centre).
• RF and VHF waves couple along the lateral dimensions of reactors and may
create standing wave effects whose suppression requires special hardware cor-
rections. As a general rule, if the size of the parallel plate electrodes are smaller
than a tenth of the excitation wavelength (i.e. 2.21 m and 0.74 m respectively
for 13.56 MHz and 40.68 MHz), electromagnetic non-uniformities start to have
a direct impact on uniformity (Sansonnens et al., 2006).
• Edge effects, including the telegraph effect, parasitic plasma at the chamber
corners and edges, and lateral duct pumping. All will influence the homo-
geneity of the coating. In some cases, they can be employed to compensate
for inhomogeneity of other origins.
• Powder formation tends to occur in µc-Si:H processes and can lead to local
non-uniformities and plasma instabilities. Local pumping might help cope
with such effects (Aya et al., 2011).
• The same chamber should be suitable for running different processes, but this
is not trivial since the effects described above tend to depend on the applied
power, pressure and gas composition. Also, contaminations from doping and
chamber cleaning have to be considered carefully (Ballutaud et al., 2004).
• A sufficient deposition rate (0.5 to, ideally, several nm/s), or very low-
cost equipment at a lower rate (0.3 to 0.5 nm/s), is required for economical
production.

Surmounting these difficulties was the main objective of much system

development during the last decade; meeting new challenges on the way was the
reason it took longer than expected to bring the technology to maturity. During this
time, several companies and equipment providers worked on deposition hardware
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 188

188 C. Ballif, M. Despeisse & F.-J. Haug

and processing for µc-Si:H with solutions based on parallel plate RF PECVD.
In most approaches, the inter-electrode gap is low (8–20 mm) and the pressure is
kept high (5–15 mbars); under these conditions, collisions slow down ions that are
accelerated in the high voltage drop at the sheath. Most deposition regimes use
high dilution (at most a few percent of silane in hydrogen). In very large coating
systems (such as Gen8.5 tools), a combination of multiple small hollow cathodes
with variable geometry may be integrated in the RF electrode and combined with a
gas diffuser consisting of holes in the showerhead plate. Additionally varying the
spacing or the gas conductance is varied, such designs can compensate standing
wave and edge effects.
A wide variety of system configurations exists (e.g. inline, vertical, batch,
cluster). TEL (formerly Oerlikon) Solar uses the design shown in Figure 4.24 where

Figure 4.24 Top: Example of a compact production system (KAI systems for PECVD,
with two process modules with stacks of 10 reactors). Bottom: Flexible amorphous silicon
modules for roof-top integration.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 189

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 189

a stack of ten reactors is placed inside a single vacuum chamber which reduces the
required components (common gas supply, pressure control, and vacuum pump
for all reactors). Only the 40.68 MHz VHF excitation is supplied from individual
power supplies. The standing wave effect is suppressed using a magnetic lens
compensation system (Sansonnens and Schmitt, 2003). Several other equipment
manufacturers, some of them not from the FPD industry, also proposed low-cost
systems for amorphous silicon, usually based on 13.56 MHz parallel plate reactors
and sometimes with lateral ingress of the processing gas (e.g. with one RF generator
powering one electrode and with two counter electrodes, allowing for coating of
four glass substrates with one generator).
Most of today’s equipment is suitable for deposition rates of 0.2–0.4 nm/s for
a-Si:H and 0.3–0.6 nm/ for µc-Si. Approaches for much higher deposition on large-
area cells use arrays of nozzle electrodes and local pumping (Aya et al., 2011) which
probably prevents excessive powder formation and plasma perturbation, or VHF at
60 MHz with ladder-shaped electrodes (Takeuchi et al. 2001; Takatsuka et al., 2005;
Mashima et al., 2006). This last configuration creates an asymmetry favouring low-
energy bombardment of the substrate. A dual frequency mode and duty cycle with
suitable spacing of the ladder allows compensating the standing wave issue. These
last two approaches have facilitated good initial module efficiencies over 10%
at high deposition rate (>2 nm/s). However, for such systems, input power and
thermal load management are not trivial.

4.7.4 Back reflectors and back contacts

Various back reflectors and back contacts are used industrially. A common back
contact for a-Si:H is a doped ZnO/aluminium stack, with typically 50–100 nm
ZnO and 200 nm aluminium. Aluminium is a medium quality reflector in the NIR
and IR, and amorphous modules can gain a few Wp/m2 when using silver instead of
aluminium. For micromorph modules, aluminium induces excessive optical losses
and must be replaced by silver. Aluminium and silver are sometimes covered
by nickel/vanadium for easier soldering. All contacts are applied by sputtering.
ZnO/silver contacts require careful optimisation, as modules that employ them
tend to be more prone to shunting. Indeed, the low sheet resistance of the metallic
back contact can allow a gain in FF. If localised shunts exist, the highly conducting
metal layer can also be detrimental since it supports larger leakage currents as a
TCO-based back reflector. This can be made, for example, with a 1.5–2 µm thick
LP-CVD ZnO with Rsheet = 10−20 /sq. In this case a white reflector must be
added as back reflector, either in the form of screen-printed white paste, or using
a white polymer in the encapsulation process.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 190

190 C. Ballif, M. Despeisse & F.-J. Haug

4.7.5 Module technology

Laser scribing following the ideas described by Nakano and co-workers (Nakano
et al., 1986; Haas et al., 2007; Bartlomé et al., 2010) is performed in the three
steps shown in Figure 4.2, referred to as P1, P2 and P3. The lasers are typically
diode-pumped solid state lasers emitting at 1064 nm or 355 nm (third harmonic)
for the P1 step, and at 532 nm (second harmonic) for the P2 and P3 processes.
Typical pulse lengths and energies are in the few ns and µJ range, with a typical
repetition rate of 30–100 kHz ensuring a scribing speed of over 1 m/s for a spot
size of 30–40 µm. The typical dead area is in the range of 200 to 350 µm. This
can be reduced with optimised mechanical and optical systems. Narrower scribing
allows for more versatility, such as thinner segment widths and the use of more
resistive TCO with less FCA. The realisation of low-voltage modules (with parallel
connection of series of segments) can also require the laser scribes to be adapted
(to go from high to low voltage instead of low to high).
The inorganic materials used in thin-film silicon modules are suitable for
long module lifetimes. In the case of a-Si:H/a-Si:H modules, systems monitored
for 14 and 16 years showed a minor efficiency decrease of 0.2% p. a. (Lechner
et al., 2010). However, care has to be taken to avoid TCO delamination (especially
SnO2 ), which can result from ion migration at high system voltage (Jansen and
Delahoy, 2003) and careful packaging is often required as some layers can be
sensitive to moisture. Most thin-film modules rely on glass/glass protection (where
none, one or both of the glasses are tempered, depending on the manufacturer)
and some on glass/back sheet protection, in which case the back sheet usually
contains a moisture-tight aluminium foil barrier. Lateral moisture ingression needs
to be controlled as well, by the use of a suitable encapsulant (modern PVB) or by
introduction of edge sealants (such as butyl, possibly in conjunction with EVA).
Moreover, good control of glass cleanliness and encapsulant adhesion is required
to avoid delamination or the ingress of water vapour.

4.7.6 Roll-to-roll technology

The advantage of roll-to-roll technology lies in flexible, lightweight unbreakable
products, which are attractive for building integration as shown in Figure 4.24. In
most cases the n-i - p configuration is used. Notably some of the best results for all
medium-size thin-film modules have been reported by United Solar working on
steel foils (12% initial on 400 cm2 , in line with the aperture area of the best module
on glass).
The dynamic roll-to-roll process requires uniformity only in one dimension,
facilitating the use of fast plasma sources, e.g. a combination of VHF and a very
narrow gap in 1D (Strobel et al., 2009; Zimmermann et al., 2010). When deposited
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 191

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 191

at high temperature, aluminium and particularly silver develop surface textures that
are suitable for light trapping (Banerjee and Guha, 1991). This ‘hot silver’ reflector
is also applicable to polyimide substrates, but low-cost plastic such as polyethylene
does not withstand the required temperatures. Alternatively, flat silver covered
by textured ZnO can act as a high-quality reflector (Guha et al., 2011), or light
trapping textures may be embossed into the plastic substrate itself (Bailat et al.,
2005; Escarré et al., 2011).
Table 4.5 summarises the status of some flexible thin-film silicon companies.
Usually, n-i - p modules incorporate EVA/ETFE protection at the front side, which
is not vapour tight but is compatible with the front ITO used by most manufac-
turers. Contrary to the case of glass substrates, there are no official turnkey sup-
pliers for roll-to-roll technology and most developments have been made in-house
by the respective companies. Helianthos uses the p-i -n structure on a sacrificial
aluminium substrate coated with a SnO2 front contact, a substrate that is also
compatible with micromorph technology (van den Donker et al., 2007).

Table 4.5 Overview of some flexible thin-film silicon suppliers and developers.

Module Capacity
Company Country Configuration efficiency (2011) Notes

Unisolar USA n-i- p triple up to 8.2% ∼180 MWp On steel foil,

(discon- a-Si/a- aperture shingle
tinued in SiGe/ area connection
2012) a-SiGe
Powerfilm USA n-i- p single 3.3% total n.a. On polyimide
area substrate
Flexcell CH n-i- p single 5% 25 MWp On PET, light
(discon- a-Si aperture trapping not yet
tinued in area introduced
2012)
Sanyo Japan n-i- p single 4.1–4.5% n.a. Amorton product
total area line
Fuji Japan n-i- p 6–7% 12 MWp On polyimide, light
tandem a- trapping by
Si/a-SiGe textured metal
Xunlight USA Triple a-Si/ 6–7% 25 MWp
Corp a-SiGe/nc- aperture
Si area
Hyet Solar NL p-i-n on (4–5) n.a. Textured AP-CVD
(ex SnO 2 SnO2 , on
Helianthos) sacrificial Al film
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 192

192 C. Ballif, M. Despeisse & F.-J. Haug

Table 4.6 Example of cost structure for a 140 MW Oerlikon Solar

THINfab II in China in 2013 for a 10–11% stabilised module efficiency
based on micromorph tandem technology (courtesy of Dr. R. Benz Oer-
likon Solar).

Contribution Amount ( /Wp ) Type of cost

Capital investment 0.08–0.12 fixed

Labour 0.01 fixed
Materials 0.12–0.15 variable
Gas 0.07–0.08 variable
Utilities 0.03 variable
Maintenance 0.03–0.04 variable
Yield loss 0.01 variable
Total module manufacturing cost 0.435–0.44

4.7.7 Costs of thin-film silicon production

Recent technological improvements, such as the use of thinner microcrystalline
absorber layers (Schicho et al., 2010; Kratzla et al., 2010; Despeisse et al., 2011),
the increase in efficiency of devices as well as the decrease in the cost of production
equipment, promise low production costs. Table 4.6 shows what a modern produc-
tion line could achieve in terms of module production costs, according to a major
equipment supplier. Production costs in the range of 0.35 /Wp should be possible
with a modest production volume of 140 MW. This calculation is given for a line
from glass to module (i.e. a line including TCO coating), and the equipment is
depreciated over seven years.
Other companies quote similar cost potential for micromorph modules. Note
that the process uses cleaning gases for the PECVD reactors. Even though such
gases are expensive, their consumption is reduced with the use of thinner micro-
crystalline absorber layers. In terms of life cycle assessment, micromorph modules
perform well. This is the case even if the reactor is cleaned with a gas such as
nitrogen trifluoride (NF3 ) (which has a high CO2 -equivalent greenhouse effect),
provided that thin layers are used and the gas is abated properly. This is also the
case for thick layers in combination with on-site fluorine (F2 ) generation (with no
greenhouse gas effect) or with mechanical cleaning. For the technology quoted
in Table 4.6, with a µc-Si:H i-layer thickness below 1 µm, an equivalent CO2
emission of 34 g/kWh is calculated, and the energy payback time at the module
level is below one year in sunny places (>1700 kWh/m2) (Steimen, 2011). These
values are in line with those reported for other thin-film technologies and, logically,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 193

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 193

lower than those given for thicker devices (Fthenakis et al., 2008a; Fthenakis et al.,
2008b; Bravi et al., 2011).

4.7.8 Conclusion and outlook

Thin-film silicon modules use abundant, non-toxic materials and have an excel-
lent temperature coefficient, promising ultra-wide-scale implementation. Thanks
to huge industrialisation efforts during the last decades, dozens of companies can
now produce hundreds of MW annually, and the best production lines are reported
to work at close to 98% production yield. Thin-film silicon is also the first thin-film
technology where efficient turnkey production lines are available for new investors
entering the market. The best micromorph modules with areas >1 m2 reach 10%
stabilised efficiencies on glass (demonstrated at the pilot line level and in some
production lines). In parallel, several companies are developing amorphous silicon
modules at ultra-low cost. There are still a lot of options to further improve efficien-
cies, including improved light trapping structures, new materials, interfaces and
doped layers as presented in this review. The large choice of materials accessible
within a single technology, as well as the possibility to easily integrate multiple
junctions, promises that significant increases in cell and module efficiency will be
achieved in coming years.

Acknowledgements
We would like to thank Z. Holman for careful reading and corrections, D. Sheel
(CVtech) and J. Hüpkes (HZB Jülich) for providing information on TCO prepa-
ration, Reinhardt Benz and Irene Steimen (Oerlikon Solar) for providing cost cal-
culation data, the personnel of PV Lab for providing figures and data, as well as
D. Alexander and M. Cantoni (CIME EPFL) and M. Leboeuf (CSEM) for creating
some of the TEM and SEM images.

References
Aya Y., Shinohara W., Matsumoto M., Murata K., Kunii T., Nakagawa M., Terakawa A. and
Tanaka M. (2011), ‘Progress of thin-film silicon photovoltaic technologies SANYO’,
Progr. Photovoltaics: Res. Applications, published online.
Bae S., Kaan Kalkan A., Cheng S. and Fonash S. (1998), ‘Characteristics of amorphous
and polycrystalline silicon films deposited at 120◦ C by electron cyclotron resonance
plasma-enhanced chemical vapor deposition’, J. Vac. Sci. Technol. A 16, 3, 1992.
Bailat J., Dominé D., Schlüchter R, Steinhauser J., Faÿ S., Freitas F., Bucher C., Feitknecht
L., Niquille X., Tscharner R., Shah A. and Ballif C. (2006), ‘High efficiency pin
microcrystalline and micromorph thin-film silicon solar cells deposited on LPCVD
ZnO coated glass substrates’, Proc. 4th World PVSEC, Honolulu, HI.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 194

194 C. Ballif, M. Despeisse & F.-J. Haug

Bailat J., Fesquet L., Orhan J.-B., Djeridane Y., Wolf B., Madliger P., Steinhauser J., Benagli,
Borrello D., Castens L., Monteduro G., Dehbozorgi B., Vallat-Sauvain E., Multone X.,
Romang D., Boucher J.-F., Meier J., Kroll U., Despeisse M., Bugnon G., Ballif C.,
Marjanovic S., Konhke G., Borrelli N., Koch K., Liu J., Modavis R., Thelen D., Vallon
S., Zakharian A. and Weidman, D. (2010), ‘Recent developments of high-efficiency
micromorph tandem solar cells in KAI-M PECVD reactors’,Proc. 5th World PVSEC,
Valencia, pp. 2720–2724, 3BO.11.5.
Bailat J., Terrazzoni-Daudrix V., Guillet J., Freitas F., Niquille X., Shah A., Ballif C., Scharf
T., Morf R., Hansen A., Fischer D., Ziegler Y. and Closset A. (2005), ‘Recent develop-
ment of solar cells on low-cost plastic substrates’, Proc. 20th EU Photovoltaic Solar
Energy Conf., Barcelona, pp. 1529–1532.
Ballif C., Barraud L., Battaglia C., Boccard M., Bugnon G., Charrière M., Cuony P.,
Despeisse M., Ding L., Escarré J., Haug F.-J., Hänni S., Löfgren L., Sculatti-Meillaud
S., Nicolay S., Parascondolo G., De Wolf S., Söderström K. and Stückelberger M.
(2011), ‘Novel Materials and Superstrates for High-Efficiency Micromorph Solar
Cells’, Proc. 26th EUPVSC, Hamburg, pp. 2384–2391, 3CO.1.1.
Ballutaud J., Bucher C., Hollenstein Ch., Howling A. A., Kroll U., Benagli S., Shah A. and
Buechel A. (2004), ‘Reduction of the boron cross-contamination for plasma deposition
of p-i-n devices in a single-chamber large area radio-frequency reactor’, Thin Solid
Films 468, 222–225.
Banerjee A., Beglau D., Su T., Pietka G., Yue G., Yan B., Yang J. and Guha S. (2011a), ‘11.0%
stable efficiency on large area, encapsulated a-Si:H and a-SiGe:H based multijunction
solar cells using VHF technology’, Proc. MRS Spring Meeting, San Francisco, CA,
A14-02.
Banerjee A. and Guha S. (1991), ‘Study of back reflectors for amorphous-silicon alloy
solar-cell application’, J. Appl. Phys. 69, 1030–1035.
Banerjee A., Liu F., Beglau D., Su T., Pietka G., Yang J. and Guha S. (2011c), ‘12.0%
efficiency on large-area, encapsulated, multijunction nc-Si:H triple cell’, Proc. 37th.
IEEE PVSC, Seattle, WA, pp. 1013–1017.
Banerjee A., Su T., Beglau D., Pietka G., Liu F., Yan B., Yang J. and Guha S. (2011b), ‘High
efficiency, large area, nanocrystalline silicon based triple junction solar cells’, Proc.
MRS Spring Meeting, San Francisco, CA, A01-03.
Bano G., Horvath P., Rozsa K. and Gallagher A. (2005), ‘The role of higher silanes in
silane-discharge particle growth’, J. Appl. Phys. 98, 013304.
Bartlomé R., Strahm B., Sinquin Y., Feltrin A. and Ballif C. (2010), ‘Laser applications in
thin-film photovoltaics’, Appl. Phys. B 100, 427–436.
Basore P.A. (2002), ‘Pilot production of thin-film crystalline silicon on glass modules’,
Proc. 29th IEEE PVSC, New Orleans, LA, pp. 49–52.
Battaglia C., Escarre J., Söderström K., Erni L., Ding L., Bugnon G., Billet A., Boccard M.,
Barraud L., De Wolf S., Haug F.-J., Despeisse M. and Ballif C. (2011a), ‘Nanoimprint
lithography for high efficiency thin-film silicon solar cells’, Nano Letters 11, 661–665.
Battaglia C., Escarre J., Söderström K., Charrière M., Despeisse M., Haug F.-J. and Ballif C.
(2011b), ‘Nanomoulding of transparent zinc oxide electrodes for efficient light trapping
in solar cells’, Nature Photonics, DOI 10.10038.
Benagli S., Borrello D., Vallat-Sauvain E., Meier J., Kroll U., Hoetzel J., Bailat J.,
Steinhauser J., Marmelo M., Monteduro G. and Castens L. (2009), ‘High efficiency
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 195

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 195

amorphous silicon devices on LPCVD-ZnO TCO prepared in industrial KAI-M R&D

reactor’, Proc. 24th. European PVSEC, Hamburg, pp. 2293–2398, 3BO.9.3.
Bennett M. S., Newton J. L. and Rajan K. (1986), ‘The influence of electric field on stability’,
Proc. 7th. European PVSEC, Sevilla, pp. 544–548.
Bennett M. S. and Rajan K. (1988), ‘The stability of multi-junction a-Si solar cells’, Proc.
20th. IEEE PVSC, Las Vegas, NV.
Bessonov A., Cho Y., Jung S.-J., Park E.-A., Hwang E.-S., Lee J.-W., Shin M. and Lee S.
(2011), ‘Nanoimprint patterning for tunable light trapping in large-area silicon solar
cells’, Solar Energy Mater. Solar Cells 95, 2886–2892.
Beyer W. (2003), ‘Diffusion and evolution of hydrogen in hydrogenated amorphous and
microcrystalline silicon’, Solar Energy Mater. Solar Cells 78, 235–267.
Bielawny A., Upping J., Miclea P. T., Wehrspohn R. B., Rockstuhl C., Lederer F., Peters
M., Steidl L., Zentel R., Lee S.-M., Knez M., Lambertz A. and Carius R. (2008), ‘3D
photonic crystal intermediate reflector for micromorph thin-film tandem cell’, Physica
Status Solidi A 205, 2796–2810.
Biron R., Hänni S., Boccard M., Pahud C., Söderström K., Duchamp M., Dunin-Borkowski
R., Bugnon G., Ding L., Nicolay S., Parascandolo G., Meillaud F., Despeisse M., Haug
F.-J. and Ballif C. (2013), ‘New progress in the fabrication of n-i- p micromorph solar
cells for opaque substrates’, Solar Energy Mater. Solar Cells 114, 147–155.
Boccard M., Battaglia C., Blondiaux N., Pugin R., Despeisse M. and Ballif C. (2013),
‘Smoothening intermediate reflecting layer for tandem thin-film solar cells’, Solar
Energy Mater. Solar Cells, http://dx.doi.org/10.1016/j.solmat.2013.03.040.
Boccard M., Cuony P., Battaglia C., Despeisse M. and C. Ballif (2010), ‘Unlinking absorp-
tion and haze in thin-film silicon solar cells’, Physica Status Solidi Rapid Research
Letters 4, 326–328.
Boccard M., Cuony P., Battaglia C., Hänni S., Nicolay S., Ding L., Benkhaira M., Bugnon G.,
Billet A., Charrière M., Söderström K., Escarré J., Sculati-Meillaud F., Despeisse M.
and Ballif C. (2012), ‘Nanometer- and micrometer-scale texturing for high-efficiency
micromorph thin-film silicon solar cells’, IEEE J. Photovoltaics 2, 83–87.
Boccard M., Despeisse M., Escarre J., Niquille X., Bugnon G., Hänni S., Bonnet-Eymard
M., Meillaud F. and Ballif C. (2014), ‘Low-refractive-index silicon-oxide interlayers
for high-stable-efficiency multi-junction thin-film silicon solar cells’, presented at the
40th IEEE PVSC, Denver.
Branz H. M. (1999), ‘The hydrogen collision model: quantitative description of metastability
in amorphous silicon’, Phys. Rev. B 59, 5498–5512.
Bravi M., Parisi M. L., Tiezzi E. and Basosi R. (2011), ‘Life cycle assessment of a micro-
morph photovoltaic system’, Energy 36, 4297–4306.
Brodsky M., Cardona M. and Cuomo J. J. (1977b), ‘Infrared and Raman spectra of silicon-
hydrogen bonds in amorphous silicon prepared by glow discharge and sputtering’,
Phys. Rev. B 16, 3556–3571.
Brodsky M., Frisch M. A. and Ziegler J. F. (1977a), ‘Quantitative analysis of hydrogen in
glow discharge amorphous silicon’, Appl. Phys. Lett. 30, 561–563.
Brodsky M. and Kaplan D. (1979), ‘Hydrogenation and the density of defect states in
amorphous silicon’, J. Non-Cryst. Solids 32, 431–435.
Brodsky M. and Title R. (1969), ‘Electron spin resonance in amorphous silicon, germanium,
and silicon carbide’, Physical Review Letters 23, 11, 581–585.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 196

196 C. Ballif, M. Despeisse & F.-J. Haug

Buehlmann P., Bailat J., Domine D., Billet A., Meillaud F., Feltrin A. and Ballif C. (2007),
‘In situ silicon oxide based intermediate reflector for thin-film silicon micromorph solar
cells’, Appl. Phys. Lett. 91, 143505.
Bugnon G., Parascandolo G., Söderström T., Bartlome R., Cuony P., Hänni S., Boccard M.,
Holovsky J., Despeisse M., Meillaud F. and Ballif C. (2011), ‘High rate deposition of
microcrystalline silicon with silicon oxide doped layers: highlighting the competing
roles of both intrinsic and extrinsic defects on the cells performances’, to be published
in Proc. 37th IEEE PVSC, Seattle, WA.
Campbell P. and Green M. A. (1987), ‘Light trapping properties of pyramidally textured
surfaces’, J. Appl. Phys. 62, 243–249.
Carlson D. E. and Wronski C. R. (1976), ‘Amorphous silicon solar cells’, Appl. Phys. Lett.
28, 671–673.
Carniglia C. K. (1979), ‘Scalar scattering theory for multilayer optical coatings’, Optical
Engineering 18, 104–115.
Chen T., Huang Y. L., Yang D., Carius R. and Finger F. (2010), ‘Microcrystalline silicon
thin-film solar cells with microcrystalline silicon carbide window layers and silicon
absorber layers both prepared by hot-wire CVD’, Physica Stat. Solidi, Rapid Research
Lett. 4, 61–63.
Chen T., Huang Y., Yang D., Carius R. and Finger F. (2011), ‘Development of micro-
crystalline silicon carbide window layers by hot-wire CVD and their applications in
microcrystalline silicon thin-film solar cells’, Thin Solid Films 519, 4523–4526.
Chittick R., Alexander J. and Sterling H. (1969), ‘The preparation and properties of amor-
phous silicon’, J. Electrochem. Soc. 116, 77–81.
Collins R., Ferlauto A., Ferreira G., Chen C., Koh J., Koval R., Lee Y., Pearce J. and Wronski
C. (2003), ‘Evolution of microstructure and phase in amorphous, protocrystalline, and
microcrystalline silicon studied by real time spectroscopic ellipsometry’, Solar Energy
Mater. Solar Cells 78, 143–180.
Cuony P., Alexander D., Perez-Wurfl Y., Despeisse M., Bugnon G., Boccard M., Söderström
T., Hessler-Wyser H., Hébert C. and Ballif C. (2012), ‘Silicon filaments in silicon oxide
for next generation photovoltaics’, Adv. Mater. 24, 1182–1186.
Cuony P., Marending M., Alexander D. T. L., Boccard M., Bugnon G., Despeisse M. and
Ballif C. (2010), ‘Mixed-phase p-type silicon oxide containing silicon nanocrystals
and its role in thin-film silicon solar cells’, Appl. Phys. Lett. 97, 213502-1–21502-3.
Curtins H., Favre M., Wyrsch N., Brechet M., Prasad K. and Shah A. (1987), ‘High-rate
deposition of hydrogenated amorphous silicon by the VHF-GD method’, Proc. 19th
IEEE PVSC, New Orleans, LA, pp. 695–698.
Das C., Lambertz A., Huepkes J., Reetz W. and Finger F. (2008), ‘A constructive combination
of antireflection and intermediate-reflector layers for a-Si/µc-Si thin-film solar cells’,
Appl. Phys. Lett. 92, 053509-1–053509-3.
Deckman H. W., Wronski C. R., Witzke H. and Yablonovitch E. (1983), ‘Optically enhanced
amorphous-silicon solar-cells’, Appl. Phys. Lett. 42, 968–970.
Delli P., Veneri L. Mercaldo V. and Usatii L. (2010), ‘Silicon oxide based n-doped layer for
improved performance of thin-film silicon solar cells’, Appl. Phys. Lett. 97, 023512–
023515.
Despeisse M., Battaglia C., Boccard M., Bugnon G., Charrière M., Cuony P., Hänni
S., Löfgren L., Meillaud F., Parascandolo G., Söderström T. and Ballif C. (2011),
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 197

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 197

‘Optimization of thin-film silicon solar cells on highly textured substrates’, Physica

Stat. Solidi A 208, 1863–1868.
Despeisse M., Bugnon G., Feltrin A., Stueckelberger M., Cuony P., Meillaud F., Billet A. and
Ballif C. (2010), ‘Resistive interlayer for improved performance of thin-film silicon
solar cells on highly textured substrate’, Appl. Phys. Lett. 96, 073507-1–073507-3.
DeVos A. (1980), ‘Detailed balance limit of the efficiency of tandem solar cells’, J. Phys.
D: Appl. Phys. 13, 839–846.
Ding L., Nicolay S., Bugnon G., Benkhaira M. and Ballif C. (2011), ‘New generation
transparent LPCVD ZnO electrodes for enhanced photocurrent in micromorph solar
cells and modules’, IEEE J. Photovoltaics 2, 88–93.
Dingemans G., van den Donker M. N., Hrunski D., Gordijn A., Kessels W. M. M. and van
de Sanden M. C. M. (2008), ‘The atomic hydrogen flux to silicon growth flux ratio
during microcrystalline silicon solar cell deposition’, Appl. Phys. Lett. 93, 111914-1–
111914-3.
Dominé D., Haug F. J., Battaglia C. and Ballif C. (2010), ‘Modeling of light scattering from
micro- and nanotextured surfaces’, J. Appl. Phys. 107, 044504–8.
Elson J. and Bennett J. (1979), ‘Relation between the angular dependence of scattering and
the statistical properties of optical surfaces’, J. Opt. Sci. Amer. 69, 31–47.
Escarré J., Söderström K., Battaglia C., Bugnon G., Parascandolo G., Nicolay S., Despeisse
M. and Ballif C. (2012), ‘Geometric light trapping scheme to improve the efficiency of
micromorph thin-film silicon solar cells’, Solar Energy Mater. Solar Cells 98, 185–190.
Faÿ S., Kroll U., Bucher C., Vallat-Sauvain E. and Shah A. (2005), ‘Low pressure chem-
ical vapour deposition of ZnO layers for thin-film solar cells: temperature-induced
morphological changes’, Solar Energy Mater. Solar Cells 86, 385–397.
Faÿ S., Steinhauser J., Nicolay S. and Ballif C. (2010), ‘Polycrystalline ZnO: B grown by
LPCVD as TCO for thin-film silicon solar cells’, Thin Solid Films 518, 2961–2966.
Faÿ S., Steinhauser J., Oliveira N., Vallat-Sauvain E. and Ballif C. (2007), ‘Opto-electronic
properties of rough LP-CVD ZnO:B for use as TCO in thin-film silicon solar cells’,
Thin Solid Films 515, 8558–8561.
Fischer D., Dubail S., Anna Selvan J. D., Pellaton Vaucher N., Platz R., Hof C., Kroll
U., Meier J., Torres P., Keppner H., Wyrsch N., Goetz M., A. Shah and Ufert K.-D.
(1996), ‘The micromorph solar cell: extending a a-Si:H technology towards thin-film
crystalline silicon’, Proc. 25th IEEE PVSC, Washington D. C., pp. 1053–1056.
Fonrodona M., Escarré J., Villar F., Soler D., Asensi J., Bertomeu J. and Andreu J. (2005),
‘PEN as substrate for new solar cell technologies’, Solar Energy Mater. Solar Cells 89,
37–47.
Franken R., Stolk R., Li H., van der Werf C., Rath J. and Schropp R. (2007), ‘Understanding
light trapping by light scattering textured back electrodes in thin-film n-i- p-type silicon
solar cells’, J. Appl. Phys. 102, 014503-1–014503-7.
Fthenakis V., Gualtero S., van der Meulen R. and Kim H. C. (2008a), ‘Comparative life-
cycle analysis of photovoltaics based on nano-materials: a proposed framework’, Proc.
MRS Spring Meeting 1041, 25–32.
Fthenakis V., Kim H. C. and Alsema E. (2008b), ‘Emissions from photovoltaic life cycles’,
Environmental Sci. Technol. 42, 2168–2174.
Fujibayashi T., Matsui T and Kondo M. (2006), ‘Improvement in quantum efficiency of
thin-film Si solar cells due to the suppression of optical reflectance at transparent
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 198

198 C. Ballif, M. Despeisse & F.-J. Haug

conductive oxide/Si interface by TiO2 /ZnO antireflection coating’, Appl. Phys. Lett.
88, 183508-1–183508-3.
Gee J. M. (2002), ‘Optically enhanced absorption in thin silicon layers using photonic
crystals’, Proc. 29th IEEE PVSC, New Orleans, LA, pp. 150–153.
Green M. A., Basore P., Chang N., Clugston D., Egan R., Evans R., Hogg D., Jarnason S.,
Keevers M. and Lasswell P. (2004), ‘Crystalline silicon on glass (CSG) thin-film solar
cell modules’, Solar Energy 77, 857–863.
Guha S., Narasimhan K. and Pietruszko S. (1981), ‘On light induced effect in amorphous
hydrogenated silicon’, J. Appl. Phys. 52, 859–860.
Guha S., Yang J. and Banerjee A. (2000), ‘Amorphous silicon alloy photovoltaic research —
present and future’, Progr. Photovoltaics: Res. Applications 8, 141–150.
Guha S., Yang J., Pawlikiewicz A., Glatfelter T., Ross R. and Ovshinsky S. (1989), ‘Band
gap profiling for improving the efficiency of amorphous silicon alloy solar cells’, Appl.
Phys. Lett. 54, 2330–2332.
Guha S., Yang J. and Yan B. (2011), ‘Amorphous and nanocrystalline silicon solar cells
and modules’, in Bhattacharya, P., Fornari, R. and Kamimura, H. (eds) Comprehensive
Semiconductor Science and Technology, Elsevier, London.
Haas S., Gordijn A. and Stiebig H. (2007), ‘High speed laser processing for monolithical
series connection of silicon thin-film modules’, Progr. Photovoltaics: Res. Applications
16, 195–203.
Hanak J. J. and Korsun V. (1982), ‘Optical stability studies of a-Si: H solar cells’, Proc.
16th IEEE PVSC, San Diego, CA, pp. 1381–1383.
Haneman D. (1968), ‘Electron paramagnetic resonance from clean single-crystal cleavage
surfaces of silicon’, Phys. Rev. 170, 705–718.
Hänni S., Battaglia C., Boccard M., Bugnon G., Cuony P., Despeisse M., Nicolay S., Meil-
laud F. and Ballif C. (2011), ‘Towards better understanding of long-term stability in
thin-film microcrystalline solar cells’, Proc. 26th PVSEC, Hamburg, pp. 2699–2703,
3AV.2.42.
Haug F.-J., Söderström K., Naqavi A. and Ballif C. (2011), ‘Resonances and absorp-
tion enhancement in thin-film silicon solar cells’, J. Appl. Phys. 109, 084516-1–
084516-8.
Haug, F.-J. Söderström, T., Python, M., Terrazzoni-Daudrix, V., Niquille, X. and Ballif, C.
(2009), ‘Development of micromorph tandem solar cells on flexible low cost plastic
substrates’, Solar Energy Mater. Solar Cells 93, 884–8877.
Higashi G., Chabal Y., Trucks G. and Raghavachari K. (1990), ‘Ideal hydrogen termination
of the Si (111) surface’, Appl. Phys. Lett. 56, 656–658.
Hirasaka M., Suzuki K., Nakatani K., Asano M., Yano M. and Okaniwa H. (1990), ‘Design
of textured Al electrode for a hydrogenated amorphous silicon solar cell’, Solar Energy
Mater. 20, 99–110.
Hüpkes J., Owen J. I., Bunte E., Zhu H., Pust S. E., Worbs J. and Jost G. (2010), ‘New
texture etching of zinc oxide: tunable light trapping for silicon thin-film solar cells’,
Proc. 5th World PVSEC, Valencia, pp. 3224–3227, 3AV.2.25.
Inthisang S., Krajangsang T., Yunaz I. A., Yamada A., Konagai M. and Wronski C. (2011),
‘Fabrication of high open circuit voltage a-Si1−x O x :H solar cells by using p-a-
Si1−x Ox :H as window layer’, Physica Stat. Solidi (c), 2990–2993.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 199

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 199

Jackson W. and Amer N. (1982), ‘Direct measurement of gap-state absorption in hydro-

genated amorphous silicon by photothermal deflection spectroscopy’, Phys. Rev. B 25,
5559–5562.
Jackson W., Kelso S., Tsai C., Allen J. and Oh S. (1985), ‘Energy dependence of the optical
matrix element in hydrogenated amorphous and crystalline silicon’, Phys. Rev. B 31,
5187–5198.
Jansen K. W. and Delahoy A. E. (2003), ‘A laboratory technique for the evaluation of
electrochemical transparent conductive oxide delamination from glass substrates’ Thin
Solid Films 423, 153–160.
Kamei T., Stradins P. and Matsuda A. (1999), ‘Effects of embedded crystallites in amorphous
silicon on light-induced defect creation’, Appl. Phys. Lett. 74, 1707–1709.
Keppner H., Meier J., Torres P., Fischer D. and Shah A. (1999), ‘Microcrystalline silicon
and micromorph tandem solar cells’, Appl. Phys. A: Mater. Sci. Processing 69,
169–177.
Kim S., Chung J.-W., Lee H., Park J., Heo Y. and Lee H.-M. (2013), ‘Remarkable progress
in thin-film silicon solar cells using high-efficiency triple-junction technology’, Solar
Energy Mater. Solar Cells 119, 26–35.
Klein S., Wieder S., Buschbaum S., Schwanitz K., Stolley T., Severin D., Obermeyer P.,
Kress M., Sommer E., Marschner T., Martini M., Noll-Baumann S., Haack J., Schmidt
U. I., Straub A., Ahmed K. and Schuegraf K. (2010), ‘Large area thin-film solar modules
with 10% efficiency for mass production’, Proc. 5th. World PVSEC, Valencia, pp. 2709–
2713, 3B.11.1.
Kluth O., Fecociu-Morariu M., Losio P., Hötzel J., Klindworth M., Eisenhammer T. and
Bailat J. (2011), ‘The way to 11% stabilized efficiency based on 1.4 m2 micromorph
tandem’, Proc. 26th PVSEC, Hamburg, pp. 2354–2358, 3BO.36.
Kluth O., Rech B., Houben L., Wieder S., Schöpe G., Beneking C., Wagner H., Löffl A.
and Schock H. W. (1999), ‘Texture-etched ZnO: Al coated glass substrates for silicon
based thin-film solar cells’, Thin Solid Films 351, 247–253.
Kluth O., Schöpe G., Hüpkes J., Agashe C., Müller J. and Rech B. (2003), ‘Modified Thorn-
ton model for magnetron sputtered zinc oxide: film structure and etching behaviour’,
Thin Solid Films, 442, 80–85.
Koh J., Ferlauto A., Rovira P., Wronski C. and Collins R. (1999a), ‘Evolutionary phase
diagrams for plasma-enhanced chemical vapor deposition of silicon thin-films from
hydrogen-diluted silane’, Appl. Phys. Lett. 75, 2286–2688.
Koh J., Fujiwara H., Koval R., Wronski C. and Collins R. (1999b), ‘Real time spectroscopic
ellipsometry studies of the nucleation and growth of p-type microcrystalline silicon
films on amorphous silicon using B2 H6 , B(CH3 )3 and BF3 dopant source gases’,
J. Appl. Phys. 85, 4141–4153.
Koida T., Sai H. and Kondo M. (2010), ‘Application of hydrogen-doped In2 O3 transparent
conductive oxide to thin-film micro-crystalline silicon solar cells’, Thin Solid Films
518, 2930–2933.
Konagai M. (2011), ‘Present status and future prospects of silicon thin-film solar cells’,
Japan. J. Appl. Phys. 50, 030001-1–030001-12.
Koval R. J., Koh J., Lu Z., Jiao L., Collins R. W. and Wronski C. R. (1999), ‘Performance
and stability of Si:H p-i-n solar cells with i-layers prepared at the thickness-dependent
amorphous-to-microcrystalline phase boundary’, Appl. Phys. Lett. 75, 1553–1555.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 200

200 C. Ballif, M. Despeisse & F.-J. Haug

Kratzla T., Zindel A. and Benz R. (2010), ‘Oerlikon Solar’s key performance drivers to grid
parity’, Proc. 5th. World PVSEC, Valencia, pp. 2807–2810, 3CO.14.1.
Krč J., Smole F. and Topič M. (2003), ‘Analysis of light scattering in amorphous Si:H
solar cells by a one-dimensional semi-coherent optical model’, Progr. Photovoltaics
11, 15–26.
Kroll U., Meier J., Torres P., Pohl J. and Shah A. (1998), ‘From amorphous to micro-
crystalline silicon films prepared by hydrogen dilution using the VHF (70 MHz) GD
technique’, J. Non-Cryst. Solids 227–230, 68–72.
Kroll U., Meier J., Steinhauser J., Fesquet L., Benagli S., Orhan J.-B., Djeridane Y., Wolf
B., Borrello D., Madliger P., Vallat-Sauvain E., Boucher J.-F., Castens L., Multone X.,
Marjanovic S., Konhke G., Borrelli N., Koch K., Liu J., Modavis R., Thelen D., Val-
lon S., Zakharian A. and Weidman D. (2011), ‘Recent developments of high-efficiency
micromorph tandem solar cells in KAI-M PECVD reactors’, Proc. 26th. PVSEC, Ham-
burg, pp. 2340–2343, 3BO.2.6.
Kroll U., Shah A., Keppner H., Meier J., Torres P. and Fischer D. (1997), ‘Potential of VHF
plasmas for low-cost production of a-Si:H solar cells’, Solar Energy Mater. Solar Cells
48, 343–350.
Laaziri K., Kycia S., Roorda S., Chicoine M., Robertson J., Wang J. and Moss S. (1999),
‘High-energy x-ray diffraction study of pure amorphous silicon’, Phys. Rev. B60,
13520–13533.
Lambertz A., Dasgupta A., Reetz W., Gordijn A., Carius R. and Finger F. (2007), ‘Micro-
crystalline silicon oxide as intermediate reflector for thin-film silicon’, Proc. 22nd
EU-PVSEC, Milan, pp. 1838–1842.
Lechner P., Frammelsberger W., Psyk W., Geyer R., Maurus H., Lundszien D., Wagner H.
and Eichhorn B. (2008), ‘Status of performance of thin-film siliocn solar cells and
modules’, Proc. 23rd EU-PVSEC, Valencia, pp. 2023–2026.
Lechner P., Geyer R., Haslauer A. and Roehrl T. (2010), ‘Long-term performance of ASI
tandem junction thin-film solar modules’, Proc. 5th. World PVSEC, Valencia, pp. 3283–
3287, 3AV.2.46.
Li H., Franken R. H., Stolk R. L., Rath J. K. and Schropp R. E. I. (2008), ‘Mechanism of
shunting of nanocrystalline silicon solar cells deposited on rough Ag/ZnO substrates’,
Solid State Phenomena 131/133, 27–32.
Li Y.-M., Jackson F., Yang L., Fieselmann B. F. and Russel L. (1994), ‘An exploratory
survey of p-layers for a-Si:H solar cells’, Proc. MRS Spring Meeting, San Francisco,
CA, pp. 663–668.
Lucovsky G. (1992), ‘Hydrogen in amorphous silicon: local bonding and vibrational prop-
erties’, J. Non-Cryst. Solids 141, 241–256.
Lundszien D., Finger F. and Wagner H. (2002), ‘Band-gap profiling in amorphous silicon-
germanium solar cells’, Appl. Phys. Lett. 80, 1655.
Mahan A. H., Williamson D. L., Nelson B. P. and Crandall R. S. (1989), ‘Small angle X-ray
scattering studies of microvoids in a-SiC:H and a-Si:H’, Solar Cells 27, 465–476.
Mahan A. H., Xu Y., Iwaniczko E., Williamson D. L., Nelson B. P. and Wang Q. (2002),
‘Amorphous silicon films and solar cells deposited by HWCVD at ultra-high deposition
rates’, J. Non-Cryst. Solids, 299–302, 2–8.
Mai Y., Klein S., Carius R., Stiebig H., Geng X. and Finger F. (2005), ‘Open circuit voltage
improvement of high-deposition-rate microcrystalline silicon solar cells by hot wire
interface layers’, Appl. Phys. Lett. 87, 073503-1–073503-12.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 201

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 201

Mashima H., Yamakoshi H., Kawamura K., Takeuchi Y., Noda M., Yonekura Y., Takatsuka
H., Uchino S. and Kawai Y. (2006), ‘Large area VHF plasma production using a ladder-
shaped electrode’, Thin Solid Films 506–507, 512–516.
Matsuda A. (1983), ‘Formation kinetics and control of microcrystallite in µc-Si: H from
glow discharge plasma’, J. of Non-Crystalline Solids 59, 767–774.
Matsuda A. (1999), ‘Growth mechanism of microcrystalline silicon obtained from reactive
plasmas’, Thin Solid Films 337, 1–6.
Matsuda A. and Ganguly G. (1995), ‘Improvement of hydrogenated amorphous silicon
germanium alloys using low power disilane–germane discharges without hydrogen
dilution’, Appl. Phys. Lett. 67, 1274–1276.
Matsuda A., Koyama M., Ikuchi N., Imanishi Y. and Tanaka K. (1986), ‘Guiding principle
in the preparation of high-photosensitive hydrogenated amorphous Si-Ge alloys from
glow-discharge plasma’, Japan. J. Appl. Phys. 25, L54–L56.
Matsuda A., Yamasaki S., Nakagawa K., Okushi H., Tanaka K., Iizima S., Matsumura M.
and Yamamoto H. (1980), ‘Electrical and structural properties of phosphorous-doped
glow-discharge Si:F and Si:H films’, Japanese J. Appl. Phys. 19, L305–L308.
Matsui T., Jia H. and Kondo M. (2010), ‘Thin-film solar cells incorporating microcrystalline
Si1−x Gex as efficient infrared absorber: an application to double junction tandem solar
cell’, Progr. Photovoltaics 18, 48–53.
Matsui T., Kondo M., Nasuno Y., Sonobe H. and Shimizu S. (2006), ‘Key issues for fab-
rication of high quality amorphous and microcrystalline solar cells’, Thin Solid Films
501, 243–246.
Matsui T., Sai H. and Kondo. (2013), ‘Development of highly stable and efficient amorphous
silicon based solar cells’, Proc. 28th European PV Conf., Paris, p. 3DO.7.2.
McCurdy R. J. (1999), ‘Successful implementation methods of atmospheric CVD on a glass
manufacturing line’, Thin Solid Films 351, 66–72.
Meier J., Dubail S., Flückinger R., Fischer D., Keppner H. and Shah A. (1994), ‘Intrinsic
microcrystalline silicon — a promising new thin-film solar cell material’, Proc. 1st
World PVSEC, Honolulu, HI, pp. 409–412.
Meier J., Spitznagel J., Kroll U., Bucher C., Faÿ S., Moriarty T. and Shah A. (2003), ‘High
efficiency amorphous and micromorph silicon solar cells’, Proc. 3rd World PVSEC,
Osaka, pp. 2801–2805.
Meier J., Torres P., Platz R., Dubail S., Kroll U., Selvan J. A., Pellaton Vaucher N., Hof
C., Fischer D., Keppner H. and Shah A. (1996), ‘On the way towards high-efficiency
thin-film silicon solar cells by the ‘micromorph’ concept’, Proc. MRS Spring Meeting,
420, 3–14.
Nakamura G., Sato K., Yukimoto Y. and Shirahata K. (1980), ‘Amorphous Si1−x Gex for
high performance solar cell’, Proc. 3rd EU-PVSEC, Cannes, pp. 835–839.
Nakano S., Matsuoka T., Kiyama S., Kawata H., Nakamura N., Nakashima Y., Tsuda
S., Nishiwak H., Ohnishi M., Nagaoka I. and Kuwano Y. (1986), ‘Laser patterning
method for integrated type a-Si solar cell submodules’, Japan. J. Appl. Phys. 25,
1936–1943.
Nasuno Y., Kondo M. and Matsuda A. (2001a), ‘Effects of substrate surface morphology
on microcrystalline silicon solar cells’, Japan. J. Appl. Phys. 40, L303–L305.
Nasuno Y., Kondo M. and Matsuda A. (2001b), ‘Passivation of oxygen-related donors
in microcrystalline silicon by low temperature deposition’, Appl. Phys. Lett. 78,
230–232.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 202

202 C. Ballif, M. Despeisse & F.-J. Haug

Naughton M. J., Kempa K., Ren Z. F., Gao Y., Rybczynski J., Argenti N., Gao W., Wang
Y., Peng Y., Naughton J. R., Burns M. J., Shepard A., Clary M., Ballif C., Haug F.-J.,
Söderström T., Cubero O. and Eminian C. (2010), ‘Efficient nano-coax based solar
cells’, Physica Stat. Solidi Rapid Research Lett. 4, 181–183.
Nishimoto T., Takai M., Miyahara H., Kondo M. and Matsuda A. (2002), ‘Amorphous silicon
solar cells deposited at high growth rate’, J. Non-cryst. Solids 299–302, 1116–1122.
Pieters B., Krč J. and Zeman M. (2007), ‘Advanced numerical simulation tool for solar
cells-ASA5’, Proc. 4th World PVSEC, Honolulu, HI, pp. 1513–1516.
Poruba A., Fejfar A., Remes Z., Springer J., Vanecek M., Kocka J., Meier J., Torres P. and
Shah A. (2000), ‘Optical absorption and light scattering in microcrystalline silicon
thin-films and solar cells’, J. Appl. Phys. 88, 148–160.
Powell M. and Deane S. (1996), ‘Defect-pool model and the hydrogen density of states in
hydrogenated amorphous silicon’, Phys. Rev. B 53, 10121–10132.
Python M., Dominé D., Söderström T., Meillaud F. and Ballif C. (2010), ‘Microcrystalline
silicon solar cells: effect of substrate temperature on cracks and their role in post
oxidation’, Progr. Photovoltaics 18, 491–499.
Python M., Vallat-Sauvain E., Bailat J., Dominé D., Fesquet L., Shah A. and Ballif C. (2008),
‘Relation between substrate surface morphology and microcrystalline silicon solar cell
performance’, J. Non-Cryst. Solids 354, 2258–2262.
Rech B., Repmann T., van den Donker M. N., Berginski M., Kilper T., Hupkes J., Calnan S.,
Stiebig H. and Wieder S. (2006), ‘Challenges in microcrystalline silicon based solar
cell technology’, Thin Solid Films 511, 548–555.
Redfield D. (1974), ‘Multiple pass thin-film silicon solar cell’, Appl. Phys. Lett. 25, 647–648.
Redfield D. (1982), ‘Energy-band tails and the optical absorption edge; the case of a-Si:H’,
Solid State Commun. 44, 1347–1349.
Ringbeck R. and Stutterlueti J. (2012), ‘BOS costs: status and optimization to reach industrial
grid parity’, Proc. 27th EU-PVSEC, Frankfurt, pp. 2961–2975, 4DP.2.2.
Rizal B., Ye F., Dhakal P., Chiles T.C., Shepard S., McMahon G., Burns M.J. and Naughton
M.J. (2013), ‘Imprint-templated nanocoax array architecture: Fabrication and utiliza-
tion’, Nano-Optics for Enhancing Light-Matter Interactions on a Molecular Scale,
NATO Science for Peace and Security Series B: Physics and Biophysics, 359–370.
Roca i Cabarrocas P. (2000), ‘Plasma enhanced chemical vapour deposition of amorphous,
polymorphous and microcrystalline silicon films’, J. Non-Cryst. Solids, 266–269,
31–37.
Rockstuhl C., Fahr S., Bittkau K., Beckers T., Carius R., Haug F.-J., Söderström T., Ballif
C. and Lederer F. (2011), ‘Comparison and optimization of randomly textured surfaces
in thin-film solar cells’, Optics Express 18, A335–A342.
Ruske F., Roczen M., Lee K., Wimmer M., Gall S., Hüpkes J., Hrunski D. and Rech B.
(2010), ‘Improved electrical transport in Al-doped zinc oxide by thermal treatment’,
J. Appl. Phys. 107, 013708–013708-8.
Sai H., Kanamori Y. and Kondo M. (2011), ‘Flattened light-scattering substrate in thin film
silicon solar cells for improved infrared response’, Appl. Phys. Lett. 98, 113502-1–
113502-3.
Sai H., Matsui T., Matsubara K., Kondo M. and Yoshida I. (2014), ‘11.0%-Efficient Thin
Film Microcrystalline Silicon Solar Cells with Honeycomb Textured Substrates’, pre-
sented at the 40th IEEE PVSC, Denver.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 203

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 203

Saito K., Sano M., Okabe S., Sugiyama S. and Ogawa K. (2005), ‘Microcrystalline silicon
solar cells fabricated by VHF plasma CVD method’, Solar Energy Mater. Solar Cells
86, 565–575.
Sakai H., Yoshida T., Hama T. and Ichikawa Y. (1990), ‘Effects of surface morphology of
transparent electrode on the open-circuit voltage in a-Si: H solar cells’, Japan. J. Appl.
Phys. 29, 630–635.
Sansonnens L., Schmidt H., Howling A. A., Hollenstein C., Ellert C. and Buechel A. (2006),
‘Application of the shaped electrode technique to a large area rectangular capacitively
coupled plasma reactor to suppress standing wave nonuniformity’, J. Vac. Sci. Technol.
A 24, 1425–1430.
Sansonnens L. and Schmitt J. (2003), ‘Shaped electrode and lens for a uniform radio-
frequency capacitive plasma’, Appl. Phys. Lett. 82, 182–184.
Schicho S., Hrunski D., van Aubel R. and Gordijn A. (2010), ‘High potential of thin (<1 µm)
a-Si:H/µc-Si:H tandem solar cells’, Progr. Photovoltaics 18, 83–89.
Schropp R. E. I., Carius R. and Beaucarne G. (2007), ‘Amorphous silicon, microcrystalline
silicon, and thin-film polycrystalline silicon solar cells’, Proc. Mater. Research Sym-
posium 32, 219–224.
Selvan J. A., Delahoy A. E., Guo S. and Li Y.-L. (2006), ‘A new light-trapping TCO for
nc-Si:H solar cells’, Solar Energy Mater. Solar Cells 90, 3371–3376.
Shah A. (2010), ‘Thin-film Silicon solar cells’, EPFL Press, ISBN 1420066749.
Shah A., Schade H., Vanecek M., Meier J., Vallat-Sauvain E., Wyrsch N., Kroll U., Droz C.
and Bailat J. (2004), ‘Thin-film silicon solar cell technology, Progr. Photovoltaics 12,
113–142.
Sheng P., Bloch A. N. and Stepleman R. S. (1983), ‘Wavelength-selective absorption
enhancement in thin-film solar cells’, Appl. Phys. Lett. 43, 579–581.
Shockley W. and Queisser H. J. (1961), ‘Detailed balance limit of efficiency of p-n junction
solar cells’, J. Appl. Phys. 32, 510–519.
Sichanugrist P., Yoshida T., Ichikawa Y. and Sakai H. (1993), ‘Amorphous silicon oxide
with microcrystalline phase’, J. Non-Cryst. Solids 164–166, 1081–1084.
Smets A. H. M., Kessels W. M. M. and van de Sanden M. C. M. (2007b), ‘The effect of ion-
surface and ion-bulk interactions during hydrogenated amorphous silicon deposition’,
J. Appl. Phys. 102, 073523-1–073523-12.
Smets A. H. M., Matsui T. and Kondo M. (2008), ‘High-rate deposition of microcrystalline
silicon p-i-n solar cells in the high pressure depletion regime’, J. Appl. Phys. 104,
034508-1–034508-11
Smets A. H. M., and van de Sanden M. C. M. (2007a), ‘Relation of the Si-H stretching
frequency to the nanostructural Si-H bulk environment’, Phys. Rev. B 76, 073202-1–
073202-4.
Smith Z. and Wagner S. (1985), ‘Intrinsic dangling-bond density in hydrogenated amorphous
silicon’, Phys. Rev. B 32, 5510–5513.
Söderström K. (2013), ’Coupling light into thin silicon layers for high-efficiency solar cells’,
PhD Thesis, EPFL No. 5714.
Söderström K., Bugnon G., Biron R., Pahud C., Meillaud F., Haug F.-J. and Ballif C. (2012b),
’Thin-film silicon triple-junction solar cell with 12.5% stable efficiency on innovative
flat light-scattering substrate’, J. Appl. Phys. 112, 114503–114503-4.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 204

204 C. Ballif, M. Despeisse & F.-J. Haug

Söderström K., Bugnon G., Haug F.-J. and Ballif C. (2012a), ‘Electrically flat/optically
rough substrates for efficiency above 10% in n-i- p thin-film silicon solar cells’, Proc.
MRS Spring Meeting, San Francisco, CA, pp. 39–44.
Söderström K., Escarré J., Cubero O., Haug F.-J., Perregaux S. and Ballif C. (2010), ‘UV
nano imprint lithography technique for the replication of back reflectors for n-i- p
thin-film silicon solar cells’, Progr. Photovoltaics 19, 202–210.
Söderström T., Haug F. J., Niquille X., Terrazoni-Daudrix V. and Ballif C. (2009), ‘Asym-
metric intermediate reflector for tandem micromorph thin-film silicon solar cells’, Appl.
Phys. Lett. 94, 063501–063501–3.
Soppe W. J., Muffler H. J., Biebericher H. C., Devilee C., Burgers A. R., Poruba A., Hodakova
L. and Vanecek M. (2005), ‘Optical and structural properties of microcrystalline silicon
grown by microwave PECVD’, Proc. 20th EU-PVSEC, Barcelona, pp. 1604–1607,
3DV.3.21.
Spear W. and Le Comber P. (1976), ‘Electronic properties of substitutionally doped amor-
phous Si and Ge’, Phil. Mag. 33, 935–949.
Sriraman S., Agarwal S., Aydil E. S. and Maroudas D. (2002), ‘Mechanism of hydrogen
induced crystallization of amorphous silicon’, Nature 418, 62–65.
Staebler D. L. and Wronski C. R. (1977), ‘Reversible conductivity changes in discharge-
produced amorphous Si’, Appl. Phys. Lett. 31, 292–294.
Staebler D. L. and Wronski C. R. (1981), ‘Stability of n-i- p amorphous silicon solar cells’,
Appl. Phys. Lett. 39, 733–735.
Steimen I. (2011), Private communication based on Oerlikon Solar process and LCA anal-
yses performed with EMPA CH.
Stradins P. (2003), ‘Light-induced degradation in a-Si: H and its relation to defect creation’,
Solar Energy Mater. Solar Cells 78, 349–367.
Strahm B. and Hollenstein C. (2010), ‘Powder formation in SiH4 -H2 discharge in large-area
capacitively coupled reactors: a study of the combined effect of interelectrode distance
and pressure’, J. Appl. Phys. 107, 023302-1–023301-7.
Strahm B., Howling A., Sansonnens L. and Hollenstein C. (2007), ‘Plasma silane concen-
tration as a determining factor for the transition from amorphous to microcrystalline
silicon in SiH4 /H2 discharges’, Plasma Sources Sci. Technol. 16, 80–89.
Street R. and Winer K. (1989), ‘Defect equilibria in undoped a-Si: H’, Phys. Rev. B 40,
6236–6249.
Strobel C., Zimmermann T., Albert M., Bartha J. W. and Kuske J. (2009), ‘Productivity
potential of an in-line deposition system for amorphous and microcrystalline silicon
solar cells’, Solar Energy Mater. Solar Cells 93, 1598–1607.
Stutzmann M., Jackson W. B. and Tsai C. C. (1985), ‘Light-induced metastable defects in
hydrogenated amorphous silicon: a systematic study’, Phys. Rev. B 32, 23–47.
Tabuchi T., Takashiri M. and Mizukami H. (2003), ‘Hollow electrode enhanced RF glow
plasma for the fast deposition of microcrystalline silicon’, Surface Coatings Technol.
173, 243–248.
Takagi T., Ueda M., Ito N., Watabe Y. and Kondo M. (2006), ‘Microcrystalline silicon
solar cells fabricated using array-antenna-type very high frequency plasma-enhanced
chemical vapor deposition system’, Japan. J. Appl. Phys., 45, 4003–4005.
Takano A. and Kamoshita T. (2004), ‘Light-weight and large-area solar cell production
technology’, Japan. J. Appl. Phys. 34, 7976–7983.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 205

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 205

Takatsuka H., Takeuchi Y., Yamauchi Y., Shioya T. and Kawai Y. (2005), ‘Production of
large area VHF plasma using ladder-shaped electrode’, Surface Coatings Technol.
2005, 972–975.
Takeuchi Y., Nawata Y., Ogawa K., Serizawa A., Yamauchi Y. and Murata M. (2001),
‘Preparation of large uniform amorphous silicon films by VHF-PECVD using a ladder-
shaped antenna’, Thin Solid Films 386, 133–136.
Taneda N., Oyama T. and Sato K. (2007), ‘Light scattering effects of highly textured trans-
parent conductive oxide films’, Techical Digest 17th PVSEC, Fukuoka, pp. 309–312,
5O-B6-01.
Tawada Y., Okamoto H. and Hamakawa Y. (1981), ‘a-SiC:H/a-Si:H heterojunction
solar cell having more than 7.1% conversion efficiency’, Appl. Phys. Lett. 39,
237–239.
TEL (2014), TEL Solar press release, 9 July 2014.
Terakawa A., Hishida M., Aya Y., Shinohara W., Kitahara A., Yoneda H., Iseki M. and
Tanaka M. (2011), ‘R&D on thin-film silicon solar cells in Sanyo’, Proc. 26th PVSEC,
Hamburg, pp. 2362–2365, 3BO.4.2.
Terasa R., Albert M., Grüger H., Haiduk A. and Kottwitz A. (2000), ‘Investigation of growth
mechanisms of microcrystalline silicon in the very high frequency range’, J. Non-cryst.
Solids 266–269, 95–99.
Tiedje T., Cebulka J., Morel D. and Abeles B. (1981), ‘Evidence for exponential band tails
in amorphous silicon hydride’, Phys. Rev. Lett. 46, 1425–1428.
Torres P., Meier J., Flückiger R., Kroll U., Selvan J., Keppner H., Shah A., Littelwood S.,
Kelly I. and Giannoules P. (1996), ‘Device grade microcrystalline silicon owing to
reduced oxygen contamination’, Appl. Phys. Lett. 69, 1373–1375.
Ulbrich C., Zahren C., Noll J., Lambertz A., Gerber A. and Rau U. (2011), ‘Analysis of the
short-circuit current gains by anti-reflective texture coating on silicon thin-film solar
cells’, Proc. 26th PVSEC, Hamburg, pp. 302–305, 1CV.3.29.
Usui S. and Kikuchi M. (1979), ‘Properties of heavily doped GD-Si with low resistivity’,
J. Non-Cryst. Solids 34, 1–11.
van den Donker M. N., Gordijn A., Stiebig H., Finger F., Rech B., Stannowski B., Bartlb R.,
Halmers E. A. G., Schlatmann R. and Jongerden G. J. (2007), ‘Flexible amorphous and
microcrystalline silicon tandem solar modules in the temporary superstrate concept’,
Solar Energy Mater. Solar Cells 91, 572–580.
van Gestel D., Chahal M., van der Wilt P. C., Qiu Y., Gordon I., Im J. S. and Portmans J.
(2010), ‘Thin-film polycrystalline silicon solar cells with low intragrain defect density
made via laser crystallization and epitaxial growth’, Proc. 35th IEEE PVSC, Honolulu,
HI, pp. 279–282.
Vanecek M., Neykova N., Babchenko O., Purkrt A., Poruba A., Remes Z., Holovsky J.,
Hruska K., Meier J. and Kroll U. (2010), ‘New 3-dimensional nanostructured thin-film
silicon solar cells’, Proc. 5th World PVSEC, Valencia, pp. 2763–2766, 3CO.12.2.
Vepřek S. and Marečeka V. (1968), ‘The preparation of thin layers of Ge and Si by chemical
hydrogen plasma transport’, Solid-State Electronics 11, 683–684.
Vetterl O., Finger F., Carius R., Hapke P., Houben L., Kluth O., Lambertz A., Mück A.,
Rech B. and Wagner H. (2000), ‘Intrinsic microcrystalline silicon: a new material for
photovoltaics’, Solar Energy Mater. Solar Cells 62, 97–108.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 206

206 C. Ballif, M. Despeisse & F.-J. Haug

Vogler B., Kerschbaumer J., Kuhn H., Mark A., Poppeller M., Schneider S., Zimin D. and
Zindel A. (2008), ‘TCO 1200◦ C Oerlikon production tool for transparent conductive
oxide thin-films’, Proc. 23rd EU-PVSEC, Valencia, pp. 2492–2493.
Wadia C., Alivisatos A. P. and Kammen D. M. (2009), ‘Materials availability expands the
opportunity for large-dcale photovoltaics deployment’, Environmental Sci. Technol.
43, 2072–2077.
Wang C. and Lucovsky G. (1990), ‘Intrinsic microcrystalline silicon deposited by remote
PE-CVD: a new thin-film photovoltaic material’, Proc. 21st IEEE PVSC, Orlando, FL,
pp. 1614–1618.
Winer K. (1990), ‘Defect formation in a-Si:H’, Phys. Rev. B 41, 12150–12161.
Wittmaack K. (2003), ‘Analytical description of the sputtering yields of silicon bombarded
with normally incident ions’, Phys. Rev. B 68, 235211–235221.
Wronski C. (1996), ‘Amorphous silicon technology: coming of age’, Solar Energy Mater.
Solar Cells 41, 427–439.
Xu X., Zhang J., Beglau D., Su T., Ehlert S., Li Y., Pietka G., Worrel C., Lord K., Yue G.,
Yan B., Banerjee A., Yang J. and Guha S. (2010), ‘High efficiency large-area a-SiGe:H
and nc-Si:H based multi-junction solar cells: a comparative study’, Proc. 5th World
PVSEC, Valencia, pp. 2783–2787, 3CO.13.1.
Yablonovitch E., Allara D., Chang C., Gmitter T. and Bright T. (1986), ‘Unusually low
surface-recombination velocity on silicon and germanium surfaces’, Phys. Rev. Lett.
57, 249–252.
Yablonovitch E. and Cody G. D. (1982), ‘Intensity enhancement in textured optical sheets
for solar cells’, IEEE Trans. Electron Devices, 29, 300–305.
Yamamoto K., Nakajima A., Yoshimi M., Sawada T., Fukuda S., Suezaki T., Ichikawa M.,
Goto M., Meguro T., Matsuda T., Sasaki T. and Tawada Y. (2006), ‘High efficiency
thin-film silicon hybrid cell and module with newly developed innovative interlayer’,
Proc. 4th World PVSEC, Honolulu, HI, pp. 1489–1492.
Yamamoto K., Nakajima A., Yoshimi M., Sawada T., Fukuda S., Suezaki T., Ichikawa M.,
Koi Y., Goto M., Meguro T., Matsuda T., Kondo M., Sasaki T. and Tawada Y. (2004),
‘A high efficiency thin-film silicon solar cell and module’, Solar Energy 77, 939–949.
Yamamoto K., Toshimi M., Suzuki T., Tawada Y., Okamoto T. and Nakajima A. (1998),
‘Thin-film poly-Si solar cell on glass substrate fabricated at low temperature’, Proc.
MRS Spring Meeting, San Francisco, CA, pp. 131–138.
Yamamoto K., Yoshimi M., Tawada Y., Fukuda S., Sawada T., Meguro T., Takata H., Suezaki
T., Koi Y. and Hayashi K. (2002), ‘Large area thin-film Si module’, Solar Energy Mater.
Solar Cells 74, 449–455.
Yan B., Yang J. and Guha S. (2003), ‘Effect of hydrogen dilution on the open-circuit voltage
of hydrogenated amorphous silicon solar cells’, Appl. Phys. Lett. 83, 782–784.
Yan B., Yue G. and Guha S. (2007), ‘Status of nc-Si:H solar cells at United Solar and
roadmap for manufacturing a-Si:H and nc-Si:H based solar panels’, Proc. MRS Spring
Meeting, San Francisco, CA, A15–01.
Yan B., Yue G., Sivec L., Yang J., Guha S. and Jiang C.-S. (2011), ‘Innovative dual function
nc-SiOx :H layer leading to a >16% efficient multi-junction thin-film silicon solar cell’,
Appl. Phys. Lett. 99, 113512-1–113512-3.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch04 page 207

Thin-Film Solar Cells based on Amorphous and Microcrystalline Silicon 207

Yang J., Banerjee A., Glatfelter T., Hoffman K., Xu X. and Guha S. (1994), ‘Progress in
triple-junction amorphous silicon based alloy solar cells and modules using hydrogen
dilution’, Proc. 1st World PVSEC, Honolulu, HI, pp. 380–385.
Yang J., Banerjee A. and Guha S. (1997), ‘Triple-junction amorphous silicon alloy solar
cell with 14.6% initial and 13.0% stable conversion efficiencies’, Appl. Phys. Lett. 70,
2975–2977.
Yang L. and Chen L. (1993), “‘Fast” and “slow” metastable defects in hydrogenated amor-
phous silicon’, Appl. Phys. Lett. 63, 400–402.
Yu Z., Raman A. and Fan S. (2010), ‘Fundamental limit of light trapping in grating struc-
tures’, Optics Express 18, A366–A380.
Yue G., Sivec L., Yan B., Yang J. and Guha S. (2010), ‘High efficiency hydrogenated
nanocrystalline silicon based solar cells deposited by optimized Ag/ZnO back reflec-
tors’, Proc. 25th EU-PVSEC, Valencia, pp. 3196–3200, 3AV.2.17.
Yunaz I. A., Hashizume K., Miyajima S., Yamada A. and Konagai M. (2009), ‘Fabrication
of amorphous silicon carbide films using VHF-PECVD for triple-junction thin-film
solar cell applications’, Solar Energy Mater. Solar Cells 93, 1056–1061.
Zeman M., van Swaaij R. A. C. M. M., Metselaar J. W. and Schropp R. E. I. (2000), ‘Optical
modeling of a-Si:H solar cells with rough interfaces: effect of back contact and interface
roughness’, J. Appl. Phys. 88, 6436–6443.
Zhu H., Bunte E., Hüpkes J., Siekmann H. and Huang S. M. (2009), ‘Aluminium doped zinc
oxide sputtered from rotatable dual magnetrons for thin-film silicon solar cells’, Thin
Solid Films 517, 3161–3166.
Zhu H., Kalkan A., Hou J. and Fonash S. (1998), ‘Applications of AMPS-1D for solar cell
simulation’, Proc. AIP Conf. 462, pp. 309–314.
Zimmermann T., Strobel C., Albert M., Beyer W., Gordijn A., Flikweert A. J., Kuske J. and
Bartha J. W. (2010), ‘Inline deposition of microcrystalline silicon solar cells using a
linear plasma source’, Physica Stat. Solidi C 7, 1097–1100.
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 209

CHAPTER 5

POLYCRYSTALLINE CADMIUM
TELLURIDE PHOTOVOLTAIC DEVICES
TIMOTHY A. GESSERT
National Renewable Energy Laboratory
Golden, Colorado, USA 8040
[email protected]

and

DIETER BONNET
Private Consultant, formerly Antec Solar

5.1 Introduction
Thin-film photovoltaic (PV) devices based on CdTe absorbers represent one of
the fastest growing segments of all PV technologies (Mehata, 2010; Coggeshall
and Margolis, 2011). It is even more remarkable that most of this impact has
occurred within the past few years. Much of the reason for this rapid development
can be traced to two factors: (i) Thin-film PV has been designed specifically to
embody production advantages over historic PV products (i.e. wafer-based tech-
nologies); (ii) CdTe PV presently enjoys production advantages over other thin-film
technologies.
In this chapter, we review the development and present state of CdTe PV
technology to describe the advantages that have made this technology one of two
dominant forces in the present PV market. Although much is understood regarding
the physics and material science of the CdTe thin-film device, we will see that
producing PV modules that are cost-effective in the present market(s) remains a
complex technological and financial undertaking. Furthermore, although current
CdTe commercial technology demonstrates the lowest specific production cost per
watt of any PV technology, we will also suggest that additional improvements and
further market penetration remain likely.
Finally, we will argue that as the economic viability of CdTe PV products
improves further, the impacts of less-obvious considerations will emerge. These

209
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 210

210 T. A. Gessert & D. Bonnet

include developing reliability extending beyond the warranty period of typically

25 years, the likelihood of large-scale mineral resource availability, principally of
Te, and how CdTe PV sustainability compares with other base-load energy sources
in terms of energy payback time and toxicity.

5.2 Brief history of CdTe PV devices

One of the first uses of CdTe was for gamma and X-ray detectors, and it was through
research related to these applications that much of the early material understanding
was established (Zanio, 1978). Additionally, because the photoelectric absorption
coefficient of CdTe is similar to that of skin tissue, several uses related to nuclear
medicine have been investigated (Meyer et al., 1972; Garcia et al., 1974).
The first notable investigations of CdTe as a solar absorber were published by
Vodakov et al. (1960) and Naumov and Nikolaeva (1961). Although few details
are provided in these papers, they reported devices with efficiencies up to ∼6%
which were fabricated by diffusing semitransparent surface coatings into crystalline
n-CdTe.
More detailed work was reported by Cusano (1963), who was seeking to
understand the formation and operation of (what would eventually be called) the
Cu x S/CdS cell. In this work, Cusano substituted n-CdTe for the more typically used
n-CdS to form Cu2−x Te/CdTe devices. Both thin-film and crystalline n-CdTe layers
were used, and the process involved dipping the CdTe into a solution containing
Cu(I) ions.
These studies provided considerable insight into how the p-type Cu2−x Te
layer is formed during cell processing, and indicated that the Cu2−x Te/CdTe device
performs more like a heterojunction than a CdTe homojunction. This study also
revealed several problems related to CdTe PV devices that remain today. Notewor-
thy among these were a close relationship between effects occurring at the CdTe
contact electrodes and junction characteristics, and the observation that the best
II–VI devices almost always involved the use of Cu.
In the same year, Nicoll (1963) became one of the first to report on the use of
close-space vapour transport (CSVT) for the deposition of CdTe. During the next
few years, measurements of minority-carrier lifetimes indicated that the maximum
lifetime of holes in n-CdTe is only about 10 ns, with even shorter electron lifetimes
observed in p-CdTe (Cusano and Lorenz, 1964; Cusano, 1967; Artobolevskaya
et al., 1968). Subsequent modelling suggested that efficiencies > 10% were unlikely
for CdTe homojunction devices without drift assistance unless the minority-carrier
lifetimes were at least 100 ns (Bell et al., 1975).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 211

Polycrystalline Cadmium Telluride Photovoltaic Devices 211

This benchmark lifetime was an order of magnitude longer than that observed
for the lowest-doped, monocrystalline CdTe. Because the homojunction emitter
would have to be heavily doped to reduce series resistance effects, most research
efforts during the last 40 years have been directed toward development of CdTe
heterojunction devices. Recently, however, reports have emerged suggesting that
monocrystalline homojunction CdTe-based devices may demonstrate a significant
opportunity for multijunction concentrator PV devices (Carmody et al., 2010).
In the CdTe heterojunction configuration, the choice for an appropriate het-
eroface material is limited by several considerations. For example, because of the
near-optimum bandgap of CdTe, the heterojunction should be designed so that
most of the absorption occurs within the CdTe bulk. Thus, the heteroface partner
must act as a highly transparent, low-resistance window layer and not be respon-
sible for carrier generation. Because all known wide-gap, low-resistivity window
materials are n-type, the CdTe layer must be p-type. The choice of window layer
is further constrained in that it is better for it to have a small lattice mismatch with
CdTe to avoid excessive interface recombination. Finally, to promote long-term
stability, the window material should be composed of elements that are slow to
diffuse into CdTe, preferably with the cation from Group II (e.g. Zn, Cd, Hg) to
optimise conduction-band alignment.
Although these considerations provide a guide for the choice of the wide-
bandgap heteroface window, a wide range of materials have been investigated,
including CdS (Bonnet and Rabenhorst, 1972; Yamaguchi et al., 1977; Bube et al.,
1977), ZnO (Aranovich et al., 1977) and ZnSe (Bube et al., 1975). It is notewor-
thy that the 5–6%-efficient CdS/CdTe devices fabricated by Bonnet in 1972 were
not only the first CdS/CdTe devices reported but were also the first all-thin-film
CdS/CdTe devices to be made, in that gas-phase transport was used to deposit thin
films of both layers.
Some of the other window materials may eventually demonstrate advantages
over CdS. However, the most efficient cells produced to date have the CdS/CdTe
configuration (Bube, 1988; Chu, 1988; Wu, 2004; Green, 2013). Additionally,
because the CdS/CdTe configuration represents a 3-element system (rather than
4-element, or more), it may produce fewer potential native defects, and ulti-
mately present fewer unforeseen industrial (production) problems arising from
interdiffusion.

5.3 Initial attempts towards commercial modules

With the success of the CdS/CdTe all-thin-film device reported by Bonnet and
Rabenhorst in 1972, many entrepreneurs who had previously worked to develop
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 212

212 T. A. Gessert & D. Bonnet

Table 5.1 Historical and present commercial companies involved in CdTe PV.
Approximate CdTe absorber
time period Name Location deposition
1975–1984 Photon Power Inc. El Paso, USA Chemical spraying
1984–1992 Photon Energy Inc. Golden, USA
1992–1997 Golden Photon Inc.
1977–c.1997 Matsushita Corp. Japan Screen-printed sublimation
1980–1984 Monosolar Inc. Santa Monica, USA Electrodeposition
1984–1999 BP Solar London, UK
1999–2002 BP/Solarex Baltimore, USA
1980–1989 AMETEK Harleysville, USA Electrodeposition
1994–2002 ANTEC GmbH Arnstadt, Germany Vacuum sublimation
2002–present Antec Solar Energy
1990–1999 Solar Cells Inc. Toledo, USA Vacuum sublimation
1999–present First Solar
2003–2007 Solar Fields Calyxo Toledo, USA Non-vacuum sublimation
2007–present USA
2006 Ziax Corp. Arvada, USA Vacuum sublimation
2006–2008 Primestar Solar
2008–present Primestar GE Solar
2006–present Solexant Corp. San José, USA Non-vacuum printing
2007–2009 AVA Solar Fort Collins, USA Vacuum sublimation
2009–2012 Abound Solar
2008–present Xunlight 26 Solar Toledo, USA Sputtering
2008–present SunPrint/Alion Inc Richmond, USA Non-vacuum printing
2008–present Advanced Solar Hangzhou, China Vacuum sublimation
Power
2007–present WK Solar Group Toledo, USA Vacuum sublimation
2008–c.2011 Clean Cell Int. Sunnyvale, USA Vacuum sublimation
2009–present REEL Solar Menlo Park, USA Electrodeposition
2010–present Encore Solar Fremont, USA Electrodeposition
In some cases, several entries are listed, indicating name transitions and/or involvements by different
corporate entities. This table has been assembled from various sources, including reports, websites and
personal communications, and is intended to be representative rather than exhaustive.

Cu x S/CdS PV products began to consider if CdTe might demonstrate advantages

compared with Cux S-based devices. Table 5.1 highlights early attempts at CdTe
commercialisation.
As Table 5.1 shows, several significant commercial endeavours were initi-
ated in the ten years between 1975 and 1985 to develop thin-film CdTe PV prod-
ucts. Many of the early attempts were based on a belief that chemical and/or
atmospheric pressure processes were a more attractive commercialisation avenue
compared with the vacuum-sublimation processes used to produce the laboratory
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 213

Polycrystalline Cadmium Telluride Photovoltaic Devices 213

devices of the early 1970s. Although the only companies presently in commercial
(or pre-commercial) production are using vacuum sublimation, several compa-
nies are now re-exploring the potential advantages of non-vacuum CdTe deposi-
tion processes. Braun and Skinner (2007) provide a more detailed discussion of
the successes and difficulties that can be encountered during CdTe PV module
commercialisation.

5.4 Review of present commercial industry/device designs

Nearly all of the companies listed in Table 5.1 have been involved in developing
or producing thin-film CdTe devices configured in the superstrate design which is
shown in Fig. 5.1. In this design, light enters through a transparent glass superstrate

Figure 5.1 Schematic diagram showing main functional components of a typical thin-film
CdTe PV solar cell device configured as a superstrate structure. The arrow on the right
indicates the direction and approximate location of the junction electric field.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 214

214 T. A. Gessert & D. Bonnet

and passes through one or more transparent conducting oxide (TCO) layers and a
CdS window layer. It is then absorbed in the CdTe absorber layer.
Because the glass superstrate provides the mechanical surface onto which all
subsequent layers are deposited, the sequence of layer deposition is TCO, then CdS,
then CdTe, and then back contact. Although the superstrate design is currently the
most widely used, and the present world record thin-film CdTe device is a super-
strate structure (Rose et al., 1999; Wu, 2004; Green et al., 2013; Gloeckler et al.,
2013), the substrate design continues to attract both research and commercial inter-
est. Here the device is typically grown on a non-transparent substrate that is often
either a Mo or Mo-coated stainless steel foil. This results in a deposition sequence
opposite to the superstrate device (i.e. back contact, then CdTe, then CdS, then
TCO, then encapsulation). Although the currently reported maximum performance
of substrate devices is significantly lower than that of high-performance superstrate
devices (∼13% efficiency vs. ∼20% efficiency: Romeo et al., 1992; Singh et al.,
1999; Dhereet al., 2012; Duenow et al., 2012; Gretener et al., 2012; Kranz et al.,
2013; Green et al., 2013), the substrate configuration has advantages, in particular in
avoiding the use of glass, which could be exploited in certain commercial products.
These include applications requiring high specific power (i.e. watts/kg), the ability
to produce flexible products for aerospace and/or building-integrated PV markets,
and/or envisioned use of roll-to-roll processing during module manufacture.
Figure 5.2 shows a scanning electron microscopy (SEM) cross-section of a
superstrate CdTe device produced at NREL. This particular device was produced
on barium silicate glass (Corning 7059), and had an SnO2 :F TCO layer and SnO2
buffer layer produced by MOCVD, a CdS layer produced by chemical bath deposi-
tion (CBD), a CdTe layer produced by close-space sublimation (CSS), a dry CdCl2
activation process, a chemical pre-contact etch step and a graphite-paste contact.

5.5 General CdTe material properties

One reason many technologists believe CdTe PV devices embody industrial advan-
tages is that the phase diagram of this material is very simple. As Fig. 5.3 shows,
single-phase CdTe can exist only very near the 50/50% Cd/Te ratio. If the com-
position deviates from this stoichiometric ratio during crystal growth, the CdTe
material will phase-separate into CdTe and Cd or Te — whichever is present in
the greater abundance. Because the vapour pressure of both Cd and Te are much
higher than that of CdTe, any elemental regions will tend to re-evaporate before
they can be incorporated into the growing crystal. However, if a sufficient Cd or
Te overpressure exists, and/or if the rate of crystal growth exceeds the elemental
evaporation rates, elemental inclusions can be incorporated. If the temperature is
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 215

Polycrystalline Cadmium Telluride Photovoltaic Devices 215

Figure 5.2 SEM micrograph showing cross-section of CdS/CdTe superstrate device pro-
duced at NREL. The layers have been colour-enhanced for clarity.

Figure 5.3 Phase diagram of CdTe. Source: Zanio (1978).

sufficiently high (>324◦C for Cd and >449◦C for Te), liquid elemental regions
can reside on the surface of the crystal. The CdTe itself will melt, as shown by the
outside of the overall curve, and the melting point is a function of off-stoichiometry.
The highest melting point at ∼1092◦ C is for stoichiometric CdTe.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 216

216 T. A. Gessert & D. Bonnet

Figure 5.4 Temperature vs. composition diagram for CdTe showing extent of electrically
active centres for both Cd- and Te- rich concentration. Source: Zanio (1978).

Although single-phase CdTe exists only very near the 50/50% composition, the
material can sustain a Cd or Te deficiency to a small extent. As shown in Fig. 5.4,
this off-stoichiometry can extend to a maximum value of about 5 × 1017 cm−3
before the Te or Cd deficiency will spontaneously produce separate Cd or Te
phases. Recalling that most materials have a density of atoms on the order of
1×1022 cm−3 , this means that the maximum off-stoichiometry that can be sustained
in crystalline CdTe at thermodynamic equilibrium is only about 0.01%. Analysis
of off-stoichiometry also suggests that, for equilibrium conditions, temperatures
greater than 600◦C can yield higher excess Te or Cd than temperatures below 500◦C
(Greenberg, 1996). Depending on the concentration of other defects, the excess Te
or Cd can manifest in polycrystalline films as a material with net acceptors or net
donors. Therefore, if a high degree of stoichiometry control can be maintained
at high growth temperatures, it may be possible to have these defects assist with
various types of junction functionality. However, we shall see that the influence of
the CdCl2 treatment and Cu incorporation during contacting also has a significant
impact on electrical properties.
In addition to a relatively simple diagram space, another important reason
CdTe possesses production advantage is that it demonstrates a property known as
congruent sublimation and condensation. When a source of bulk CdTe is heated to a
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 217

Polycrystalline Cadmium Telluride Photovoltaic Devices 217

temperature at which the CdTe surface sublimes into Cd and Te2 gas, the Cd and Te
atoms evaporate at roughly the same rate (i.e. congruent sublimation). Although
the vapour pressure of Cd is higher than that of Te2 at any given temperature,
Te2 has two atoms, so the arrival rates of Te and Cd atoms at a surface are very
similar. Furthermore, when Cd and Te2 vapours condense onto a (cooler) surface,
they again do so at about the same rate (i.e. congruent condensation). These two
congruent processes enable a bulk source of stoichiometric CdTe to evaporate and
condense into a film of nearly stoichiometric CdTe. If for some reason Cd and Te
arrive at a growing film surface at different rates, then because the vapour pressure
of Cd (or Te2 ) above a pure Cd (or Te) surface is much higher than that of either
above a CdTe surface, any depositing Cd or Te surface atoms that do not quickly
coordinate into the CdTe material will re-evaporate from the surface, limiting the
formation of Cd or Te secondary phases.
The previous discussion pertains to high-purity, crystalline CdTe formed
under equilibrium conditions. However, for thin-film PV devices, the CdTe will be
formed using less pure materials and ambients, and relatively non-equilibrium con-
ditions. These differences can result in polycrystalline thin films that can contain
numerous structural and impurity defects that can affect the material properties of
the film. Luckily, because the kinetics of film growth for CdTe are fairly rapid, even
high-rate film growth processes tend to retain much of the structural regularity that
is observed for bulk crystalline materials.
Although CdTe is ambipolar (i.e. it can be doped both n-type and p-type),
the material tends to self-compensate. This means if one attempts to incorporate
acceptors by substituting Group I dopants on the cation [Cd] site, or Group V
dopants on the anion [Te] site, a dopant concentration will be reached where the
lattice will spontaneously create compensating donors, with the result that the
further change in net acceptor concentration is minimal (Zunger, 2003).
Production of epitaxial CdTe films using molecular beam epitaxy (MBE) tech-
niques can achieve n- and p-type doping into the mid-1018 cm−3 range using
extrinsic dopants (Taterenko et al., 1993; Dhese et al., 1994). For polycrystalline
PV devices, the CdTe is typically deposited without extrinsic dopants added. The
as-deposited CdTe is often believed to be p-type as a result of cadmium vacancies
(VCd ); however, the electrical resistance of as-deposited CdTe is very high, so the
electrical properties of as-deposited polycrystalline thin-films are not well estab-
lished. Most device technologists know only that a working junction results if the
CdTe is deposited onto an n-type heteroface layer, such as CdS, and a back contact
is fabricated onto the CdTe. The conclusion is that the CdTe must have been, or
have become, p-type. As we will see in the following discussion, for some CdTe
deposition processes, it is possible that the as-deposited CdTe may be p-type,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 218

218 T. A. Gessert & D. Bonnet

intrinsic, or even n-type, and it is the post-deposition processes (CdCl2 processing

and/or contacting with a Cu(I)-containing layer) that establish the necessary p-type
properties.

5.6 Layer-specific process description for superstrate CdTe devices

5.6.1 Superstrate materials
CdTe thin-film PV devices are typically grown on a glass superstrate. At this
time, nearly all published research on CdTe PV has been performed using barium
silicate glass (e.g. Corning 7059), borosilicate glass or soda-lime float glass. There
are presently no definitive studies that compare material and/or performance differ-
ences resulting from these different glass choices. This is partly because different
glass materials have different strain points, and thus can allow higher-temperature
processes, and partly because some transparent conducting oxides (TCOs) are
only available on certain types of glass. The following discussion notes some of
the parameters that should be considered when choosing a glass superstrate for a
CdTe PV device.

Temperature stability
CdTe technologists suggest that higher substrate temperatures minimise the defect
density in the CdTe film. Although firm evidence of this has not been reported,
it is true that the devices with higher performance tend to have had the CdTe
deposited at higher temperature. In the case of commonly used soda-lime float glass
substrates, it is essential to keep the processing temperatures below 550–575◦ C to
avoid mechanical distortion due to the low strain and softening points, and to
avoid altering the degree of temper of the as-fabricated glass. For some industrial
processes, the temperature may be much lower to avoid even small distortions in
the glass and/or to limit loss of glass temper. For higher processing temperatures,
borosilicate (and historically, barium silicate/Corning 7059) glass substrates have
been used for laboratory devices to allow studies of material functionality at higher
temperature. Because of the size of the potential market for thin-film CdTe PV,
many commercial glass companies are considering the use of other types of glass
compositions that may demonstrate higher strain points than soda-lime glass, or
may embody other commercial advantages.

Coefficient of thermal expansion

A difference in the coefficient of thermal expansion (CTE) between the substrate
and the film can induce stress in the deposited layers following deposition and
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 219

Polycrystalline Cadmium Telluride Photovoltaic Devices 219

cooling. This gives rise to structural defects and can also affect the adhesion of the
film(s). Although this is another area where studies have not been conducted and/or
reported, it remains prudent to choose a glass with a CTE that closely matches that
of the material(s) being deposited.

Glass composition and impurity diffusion

Because glass constituents can diffuse out from the glass at elevated processing
temperatures and/or during long-term field deployment, the effect on the device
layers of mobile impurities from the glass must be considered. For alkali glasses,
impurities (e.g. sodium, potassium, calcium etc.) can diffuse into the films dur-
ing processing. If impurity diffusion is found to impact underlying device layers
adversely, additional diffusion-barrier layers and/or glass compositions that miti-
gate these issues may be required. It may also be possible to produce barrier layers
that incorporate designs which reduce optical reflection, especially at non-normal
incidence. Alternatively, and as in CIGS PV device technology, improved under-
standing of the CdTe material may indicate that certain types of alkali diffusion
could be beneficial.

5.6.2 TCO layer

Following deposition of any alkali diffusion barrier layer(s), the next layer
deposited onto the glass is the TCO layer. Because this provides a lateral con-
duction pathway for electrons from the n-type side of the device, this must be
highly conductive. However, because incident light must pass through the TCO
layer for subsequent absorption in the junction, it must also be highly transparent.
Presently, most CdTe modules incorporate a TCO layer that has a sheet resistiv-
ity of 10–15 ohms/square, and optical transparency of ≥90% within the spectral
region ∼600–700 nm.
For many production CdTe technologies, fluorine-doped SnO2 (SnO2 :F) has
been the historical choice for the TCO layer. SnO2 :F-coated glass has been widely
available for many years because of its use in low-emissivity (low-e) glazing for
residential and appliance applications (such as commercial/residential windows,
and heat-reflecting glass in oven windows). SnO2 :F is chemically inert, is not
greatly affected by environmental moisture or diffusion from other materials in
the PV device, and may provide a degree of protection from Na diffusion from
soda-lime glass. Finally, SnO2 :F demonstrates very high stability in its electrical
and optical properties at CdTe processing temperatures (>600◦ C).
Detriments of SnO2 :F include that it has lower visible transmission than either
ITO (indium tin oxide) or ZnO:Al, and this limits current generation. Although a
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 220

220 T. A. Gessert & D. Bonnet

complete explanation for this difference has not been reported, it partly arises
because the maximum electron mobility of present commercial-grade SnO2 :F is
lower than in other TCOs (in the range of 20–25 cm2 V−1 sec−1 ). This means that
to achieve both low sheet resistance and high optical transmission, higher carrier
concentration must be used in the film (∼7 × 1020 cm−3 ). This higher carrier
concentration will produce more free-carrier absorption, limiting the transmitted
near infrared (NIR) light entering the PV device, and thus the short-circuit current.
Many research groups are presently investigating avenues to improve the per-
formance of SnO2 :F, and/or develop alternative TCOs for CdTe devices (Dhere
et al., 2010). For example, a recent high-performance thin-film CdTe device uses a
Cd2 SnO4 TCO layer (Wu et al., 1996; Coutts et al., 2000; Wu, 2004). An alternative
avenue to improve the transmission of TCOs for PV applications involves forming
TCO alloys with higher dielectric permittivity (Gessert et al., 2011). Because the
dielectric permittivity affects the plasma wavelength, this technique can yield a
TCO with much less NIR absorption.

5.6.3 Buffer layer

In most high-performance CdTe (and CIGS) PV devices, a high-resistance oxide
layer is placed between the TCO and the CdS layers. This layer is often called a
buffer layer (BL) but also has been called a high-resistance transparent (HRT) layer.
Many attributes have been ascribed to this layer, and it is not presently clear if all
yield benefits in all CdTe device processes, or if certain BL attributes assist only
certain types of device designs and/or fabrication processes. Suggested benefits
include:

• The BL allows thinner CdS layers to be used. One proposed reason for this is
that the space charge on the n-type side of the junction can expand into the
BL when the CdS becomes too thin to balance the charge needed to maintain
the space-charge width in the p-type side of the junction. This enables the
space charge in the p-CdTe region to remain sufficiently wide as the CdS
layer is thinned, thereby avoiding device functionality dominated by voltage-
dependent collection (i.e. low fill factor arising from loss of minority-carrier
collection at high forward bias). A different explanation is that the BL reduces
surface recombination at the CdS/TCO interface. The suggested benefit is that
reduced recombination at this interface may improve device performance if
thinner CdS layers position this interface closer to the CdS/CdTe junction
region.
• The BL provides tolerance to shunt (or short) paths from the back contact to
the front contact. These pathways will be present if the CdTe layer contains
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 221

Polycrystalline Cadmium Telluride Photovoltaic Devices 221

cracks or pinholes that may form during CdTe deposition, or during pre-contact
treatments such as chemical etching. These shunt paths are particularly prob-
lematic if vacuum-deposited metals are used in the contact (as opposed to
mechanically applied paste contacts).
• Certain types of BL can allow higher temperatures or longer times to be used
during the CdCl2 treatment. The issue here is that during the CdCl2 process
(see Section 5.6.7), delamination can occur between the CdS and TCO layers
if the CdCl2 process temperature is excessively high. Indeed, the optimum
CdCl2 processing conditions are typically defined as the maximum time and
temperature that can be tolerated before delamination. It has been suggested
this benefit may be more likely for a chemically active BL, such as Zn2 SnO4
or related alloys that can form mixed alloys with the CdS.
• A chemically active buffer layer, such as Zn2 SnO4 or related alloys, may
consume some of the CdS layer, reducing the amount of (optically absorbing)
CdS, and producing in its stead a wider bandgap CdZnS material (Wu, 2004).

The particular choices of glass superstrate, TCO, buffer layer and CdS will
significantly affect the amount of light that can enter the junction region of a CdTe
thin-film PV device, and thus will impact the device short-circuit current. This
is illustrated in Fig. 5.5, which compares the quantum efficiency consistent with
the 16.7% CdTe device produced at NREL (i sc =∼ 26 mA cm−2, dotted quantum
efficiency curve) with a historic commercial device (i sc =∼ 19 mA cm−2 , solid
curve) (Wu et al., 2002; Green et al., 2012).
The comparison reveals significant loss differences in both the NIR region
(600–850 nm) and the UV region (300–600 nm). About one half of the ∼3 mA cm−2
NIR loss is believed to be due to absorption in the (high-Fe) commercial glass and
the other half is due to the free-carrier absorption of commercial TCO. For the
NREL device shown, barium silicate glass (Corning 7059) and a Cd2 SnO4 TCO
layer were used to minimise both these NIR losses. It should also be noted that the
commercial TCO modelled for this figure (using Drude Theory approximations)
might actually suggest less NIR absorption than typical commercial SnO2 :F. The
difference in UV loss between the commercial and NREL devices shown in this
figure (∼4 mA cm−2) is primarily due to the use of relatively thick layers (typ-
ically ∼150–300 nm) of CdS in the commercial device. In the NREL device, a
Zn2 SnO4 -alloy BL was used to allow a CdS layer that is typically thinned to ∼70 nm
thick. From this discussion, it is clear that additional performance benefits remain
possible in commercial modules by optimisation of the CdS thickness. However,
this will require improved understanding of the function(s) and optimisation of
specific BLs.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 222

222 T. A. Gessert & D. Bonnet

Figure 5.5 Comparison of quantum efficiency curves from the 16.7% efficiency CdS/CdTe
device at NREL (i sc =∼ 26 mA cm−2 ) and a representative commercial CdS/CdTe device
(i sc =∼ 19 mA cm−2 ). The red triangles indicate the approximate amount of loss in each
region. The figure also shows several different types of glass superstrate and a modelled
absorption curve for a typical commercial TCO.

5.6.4 CdS layer

Although the CdS layer is formally the n-type heteroface partner of the CdS/CdTe
device, for most device processes the resulting electrical junction is cited between
the CdTe absorber layer and an intermixed CdSTe layer that forms during high-
temperature CdTe deposition. Because of this intermixing, the device is not
generally considered a heterojunction — but instead a quasi-homojunction (Dhere
et al., 2008). Although materials other than CdS have been tested for the heteroface
partner layer, CdS has been found to yield the highest device performance. More-
over, and unlike the situation in CuInSe2 -alloy devices, because a much thicker
Cd-containing layer still exists in the device (the CdTe layer!), there has been little
motivation to eliminate the Cd from the heteroface layer based solely on perceived
toxicity concerns.
Many different CdS deposition processes can be used to make the CdTe device.
The previous NREL world-record cell uses chemical-bath deposition (CBD) (Rose
et al., 1999). However, sputter deposition, evaporation and various forms of gas-
phase transport (close-space sublimation, vapour-transport deposition etc.) are
also used by research and industry. When processes other than CBD are used,
CdTe technologists have found that adding oxygen to the CdS deposition process
appears to produce beneficial diffusion during the subsequent CdTe deposition
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 223

Polycrystalline Cadmium Telluride Photovoltaic Devices 223

process. In the case of sputter deposition of CdS, oxygen addition can also lead
to the formation of a nanocrystalline CdS phase (Wu et al., 2002). It is believed
that quantum-confinement effects yield a wider bandgap for nanocrystalline CdS,
allowing increased device short-circuit current from the CdS:O/CdTe device.

5.6.5 CdTe layer

A primary production advantage of CdTe, relative to other known thin-film absorber
layers, is that it can be deposited very quickly (at rates >20 µm/min). This means
that the length of the CdTe deposition zone (and related deposition hardware costs)
can be much smaller compared with other thin-film PV technologies that require
slower deposition rates.
A variety of deposition techniques have been used to deposit CdTe including:
various forms of vacuum sublimation (e.g. evaporation, close-spaced sublimation,
vapour transport deposition) (Chu et al., 1992; Sasala et al., 1996; Rose et al., 1999;
McCandless et al., 1999), sputtering (Li et al., 1993; Wendt et al., 1998), electrode-
position (Fulop et al., 1982; Basol, 1984), screen printing (Suyama et al., 1990),
spray techniques (Jordan, 1993; Kester et al., 1996) and printing of nanoparticles
(Gur et al., 2005). Following proper post-deposition processing, nearly all these
techniques have yielded devices with efficiency > 10%. The first method used to
deposit CdTe films for high-efficiency solar cell applications used vacuum subli-
mation in which a crystal of CdTe was heated in a partial vacuum to its sublimation
point, and the vapour was transported by a gas and re-condensed on a substrate
(Bonnet and Rabenhorst, 1972). Figure 5.6 illustrates two laboratory-scale meth-
ods that use higher-rate vacuum sublimation for deposition of CdTe (and possibly
CdS). Powell et al. (1999) provided a description of one high-rate commercial vac-
uum sublimation process. Because of the dominant importance of various methods
of high-rate vacuum sublimation in today’s commercial CdTe industry, the follow
discussion will focus on parameters relevant to this deposition process only.
One of the primary parameters that affects CdTe material properties is
the deposition temperature. Material properties can be categorised broadly
into films deposited at low temperature (<∼425◦ C), intermediate temperature
(∼450–575◦C) and high temperature (∼600–625◦C). Because of void formation
and substrate limitations at temperatures greater than ∼625◦ C, this substrate tem-
perature represents the approximate upper limit for present high-rate vacuum sub-
limation processes. Although devices have been produced on as-deposited CdTe
and heat-treated films (without CdCl2 treatment), it is worth noting that the only
device to exceed 10% efficiency without CdCl2 was produced from CdTe layers
deposited in the high-temperature range (Tyan and Perez-Albuerne, 1982).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 224

224 T. A. Gessert & D. Bonnet

Figure 5.6 (a) Cross-section schematic of research-scale CSS system; (b) cross-section of
research-scale gas-phase transport system.

Low-temperature growth yields films with maximum grain diameters

<∼0.5 µm, and which are highly oriented and exhibit significant inbuilt stress
(Moutinho et al., 1999; McCandless et al., 1999). Generally, these films must
undergo higher-temperature heat treatment (or more typically, the CdCl2 treat-
ment as described in Section 5.6.7) to yield high device performance. In contrast,
intermediate- and high-temperature growth yields films with maximum grain diam-
eters >∼0.5 µm that demonstrate progressively less orientation and lower inbuilt
stress as the substrate temperature increases (Moutinho et al., 1999; McCand-
less et al., 1999). Presently, because all commercial modules are produced using
soda-lime glass that has an effective upper temperature limit in the range of ∼550–
575◦ C (owing to production issues related to glass softening), most commercial
CdTe films are produced in the intermediate temperature range.
Closely linked to substrate temperature, the deposition ambient exerts a sig-
nificant effect on material properties. Although sublimation occurs primarily in a
relatively non-reactive environment (usually He, Ar, or N2 ), deposition in a small
partial pressure of O2 (either intentional or unintentional) will lead to higher device
performance (Tyan et al., 1984) and nucleation of smaller grains (Rose, 1997). As
growth proceeds, these smaller grains yield a denser film that provides resistance
to shunting between the front and back contacts, and may reduce recombination
at deep defect states (e.g. OTe is thermodynamically favoured over VTe ) (Corwine
et al., 2005). Following nucleation, grain size increases with film thickness such
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 225

Polycrystalline Cadmium Telluride Photovoltaic Devices 225

that the maximum grain size tends to be equal to the film thickness. However, the
initially small grains have significant implications for superstrate devices because
the grain size nearest to the junction region (where light absorption occurs) is
generally small, and therefore issues arising from interface and/or grain-boundary
recombination will be more pronounced.
Reported electrical properties of as-deposited polycrystalline CdTe films are
based primarily on resistivity and hot-probe studies (Zanio, 1978; McCandless,
2003; Dhere and Li, 2012). These suggest that low deposition temperatures produce
n-type films with resistivity at or above ∼108 ohm cm. Intermediate-temperature
deposition yields as-deposited resistivity in the range of 107 –108 ohm cm, and
the films tend to be intrinsic but can demonstrate either n- or p-type character-
istics. Research devices that have been produced at high temperatures suggest
as-deposited films are slightly p-type (Dhere et al., 2008). The transition of con-
ductivity type from n to p with increasing temperature may be linked to the forma-
tion of more cadmium vacancy defects. The CdCl2 treatment reduces the resistivity
by 3–4 orders of magnitude and can assist conversion of intrinsic or n-type films
to lightly p-type (this is discussed more fully in Section 5.6.7). Although more
detailed information on the carrier concentration and type of as-deposited films
would be valuable, results of CV and Hall analyses of as-deposited and CdCl2 -
treated CdTe films are generally inconclusive because the low carrier concentration
produces surface and interface depletion regions that are equal to or wider than
typical film thicknesses. Therefore, nearly all of the electrical data reported for
polycrystalline CdTe are for films that have been treated after deposition, typically
by CdCl2 and Cu contacting.
As we note below, the assumed interplay between different process steps is one
of the things that make CdTe a difficult material to process reproducibly. However,
we will also note that improved understanding of any of these process steps may
not only lead to their elimination, but also elimination of related process issues.

5.6.6 CdSTe layer

CdTe substrate temperatures higher than ∼425◦C yield progressively more inter-
diffusion between the CdTe and the underlying CdS layer, forming what is often
called the CdSTe-alloy region (or CdSTe layer) (Moutinho et al., 1999; McCand-
less et al., 1999). For devices formed at high temperature, the CdSTe-alloy region
forms primarily during CdTe growth, and is altered further during the CdCl2 pro-
cess (discussed below). If a low-temperature CdTe deposition process is used, the
CdSTe-alloy region is typically established entirely during the CdCl2 treatment
(Moutinho et al., 1999). Although most high-performance CdTe devices contain
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 226

226 T. A. Gessert & D. Bonnet

Figure 5.7 SEM cross section of CdS/CdTe device illustrating the effect of oxygen in
the CdS on CdS consumption during CdTe deposition: (a) CSS-deposited CdS containing
no oxygen. Note significant consumption of CdS; (b) CSS-deposited CdS that has been
post-deposition treated to incorporate oxygen. Note much less consumption of CdS; (c)
CBD-deposited CdS. Note very little consumption of CdS. Source: Albin et al. (2002).

a CdSTe-alloy region, its function and preferred attributes remain areas of active
debate.
In addition to the CdTe deposition temperature and CdCl2 treatment (McCan-
dless et al., 1999), another important parameter affecting the CdSTe-alloy region
is the amount of oxygen available within this region (Albin et al., 2002). Figure 5.7
shows that the amount of oxygen in the CdS layer alters the interdiffusion between
the CdS and CdTe layers during high-temperature CdTe deposition. In general,
adding oxygen to the CdS layer reduces the amount of CdS consumed during
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 227

Polycrystalline Cadmium Telluride Photovoltaic Devices 227

high-temperature CdTe deposition. The reduced CdS consumption not only limits
the extent of the CdSTe alloy that forms on the CdTe side of the device but also
appears to limit the amount of Te that diffuses into the CdS layer.
It has been suggested that the resulting moderation of the extent of the CdSTe
thickness by oxygen limits the depth of the CdSTe/CdTe interface, which limits the
depth of the quasi-homojunction (Dhere et al., 1996; Dhere et al., 2008). One pos-
sible explanation for these effects of oxygen is that oxygen acts as an isoelectronic
substitutional defect for Te and/or S vacancies. The formation of these defects could
limit diffusion, and possibly reduce interface recombination. However, at this time,
the effects of this diffusion on junction formation and ultimate cell performance
are not well understood.

5.6.7 CdCl2 activation

Since about 1985, nearly all research and commercial thin-film CdTe PV devices
have incorporated a CdCl2 treatment or activation step. The incorporation of this
step is one of the main reasons why so many different groups, using different
CdTe deposition processes and temperatures, have been able to achieve conver-
sion efficiencies greater than 10%. Many call the CdCl2 process an equaliser step
because it results in CdTe polycrystalline thin-film material that demonstrates
minority-carrier lifetimes that are similar and sufficiently high to enable good
device performance.
For PVD CdTe (e.g. GPT, evaporated, sputtered, etc.), the CdCl2 process
usually involves spraying a CdCl2 -saturated methanol or water solution onto the
CdTe surface after deposition, or dipping the CdTe into a CdCl2 solution, followed
by drying and annealing in an oxygen-containing ambient at ∼400◦ C for ∼10 min.
The CdCl2 treatment can also be performed by combining the application and
thermal steps into a CdCl2 vapour process at ∼400◦ C. For solution-based CdTe
deposition processes, compounds that contain chlorine can be included in the CdTe
precursors with results similar to a post-deposition CdCl2 treatment.
For small-grained CdTe (∼<0.5 µm diameter) produced from solution or
by low-temperature CdTe growth, the CdCl2 process typically leads to complete
recrystallisation that produces larger polycrystalline grains (Mountiho et al., 1999;
McCandless et al., 1999). Although for larger-grained (higher-temperature CdTe
growth) material, the CdCl2 process does not appear to produce significant recrys-
tallisation, both small- and large-grained material demonstrate higher minority-
carrier lifetime after CdCl2 treatment. For large-grained material, the improvement
is due primarily to passivation at grain boundaries (Gessert et al., 2001).
The particular defect modifications associated with the CdCl2 treatment con-
tinue to be debated (Corwine et al., 2005). In general, adding either excess Cd
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 228

228 T. A. Gessert & D. Bonnet

or Cl to CdTe results in the formation of donor defects (Zanio, 1978). However,

many technologists associate improved p-type CdTe material quality following the
CdCl2 treatment with the formation of what has become known as the Cl A-centre,
which is believed to be an acceptor level located ∼150 meV above the valence
band, and ascribed to a Cl-on-Te vacancy defect (ClTe ) paired with a Cd vacancy
defect (VCd -ClTe ) (Zanio, 1978; Halliday et al., 2001; Valdna et al., 2001; Wei and
Zhang, 2002).
One artefact of the CdCl2 process that is often not discussed relates to residuals
left behind on the CdTe surface. It has been shown that, if the CdCl2 process is
wet (application of a saturated methanol or aqueous solution), residuals are either
cadmium oxychlorides or oxytellurides (Gessert et al., 2001), as shown in Fig. 5.8.
These residuals are often removed during the pre-contact chemical etching step(s),
and are therefore not a major concern for wet-contact processing. If, however, a dry
contacting process is to be used, sufficient removal of CdCl2 residuals requires the
use of a dry removal process (e.g. ion-beam or ion-etch processes) (Gessert et al.,
2001). If a vapour process is used for the CdCl2 treatment, the resulting residual
is CdCl2 , and this can be removed using a thermal pre-contact processing (Waters
et al., 1988).
In addition to the effect of CdCl2 treatment on the electrical properties of
the CdTe layer, it also alters CdSTe interdiffusion. Figures 5.9a and 5.9b com-
pare secondary ion mass spectrometry (SIMS) of two CdS/CdTe devices that were
deposited at two different CSS substrate temperatures. Both figures also show the
sulphur (S) profile before and after a CdCl2 process. By comparing these pro-
files, one notes that the extent of S diffusion from the CdS layer during the CdCl2
treatment is much greater for the device shown in Fig. 5.9a where the CdTe was

(a) (b)
Figure 5.8 SEM micrographs of CdCl2 oxychloride residuals from wet CdCl2 processing
treatment: (a) low magnification; (b) higher magnification. Source: Gessert et al. (2001).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 229

Polycrystalline Cadmium Telluride Photovoltaic Devices 229

(a)

(b)
Figure 5.9 SIMS analysis of the effect of CdCl2 treatment on S diffusion as a function of
CdTe deposition temperature. (a) 550◦ C; (b) 600◦ C. Source: Dhere et al. (1996).

deposited at lower temperature (550◦ C). Recalling that the amount of oxygen avail-
able at the CdS/CdTe interface also affects the extent of the CdSTe layer, the result
suggests controlling the formation of the CdSTe layer requires (at least) controlling
the CdTe temperature, the CdCl2 treatment temperature and time, and the amount
of available oxygen. The interrelationship between these parameters begins to
explain why developing a reproducible and cost-effective CdTe PV module pro-
cess requires a deep understanding of the mechanisms at work and the ability to
control these mechanisms.
The above description of how the CdCl2 treatment affects CdSTe and CdTe
material properties provides clues to how the treatment also affects device perfor-
mance. Figures 5.10a and 5.10b show the effect of the CdCl2 treatment process
on device performance as a function of CdTe CSS substrate temperature (Dhere
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 230

230 T. A. Gessert & D. Bonnet

Figure 5.10 Effect of a wet CdCl2 treatment on (a) device Voc and (b) i sc as a function of
CSS CdTe substrate temperature. Source: Dhere et al. (1996).

et al., 1996). The figure shows that while higher CdTe deposition temperatures
lead to improved device performance, the CdCl2 process improves performance at
all substrate temperatures. In this data set, a device process parameter (substrate
temperature) was altered to yield a device that responds better to a given (wet)
CdCl2 treatment. However, depending on process constraints, the CdCl2 treatment
could also be varied to produce a higher performance for a given device process.
Because these process alterations can be subtle, and are often made without full
knowledge of what aspect of cell functionality is being altered, it may be difficult to
know precisely how a specific CdCl2 process parameter may be affecting junction
performance.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 231

Polycrystalline Cadmium Telluride Photovoltaic Devices 231

5.6.8 Back contact

For superstrate CdTe PV devices, the last step in device formation is fabrication of
the back contact. Until about 2000, most CdTe technologists believed the primary
function of the back contact was to establish a low-resistance pathway for elec-
trons to enter the CdTe layer during device operation. It was believed that the
semiconductor junction properties (n-type and p-type regions, respective space-
charge regions, and minority-carrier lifetimes) were established prior to contacting.
However, more recent studies have shown that the back contact process not only
provides a low-resistance current pathway, but significantly alters the electrical
properties of the underlying CdTe layer, thereby significantly affecting junction
functionality.
The first consideration in developing a back-contacting process is to determine
if/how any residuals from the CdCl2 process should be removed. Depending on how
the CdCl2 process was performed, this removal can be easy or difficult: removing
CdCl2 residual following a vapour process is easy, while removing Cd oxychloride
residuals after a wet process is more difficult (Gessert et al., 2001). Removal process
options include thermal, chemical and physical (e.g. ion-beam treatment) methods.
Historically, a chemical option has been preferred because it often produces a Te-
rich layer that assists certain types of contact formation (Li et al., 1999; Levi et al.,
2000). It has also been suggested that gettering of excess Cu by etch-formed Te
inclusions can enhance device stability (Albin et al., 2006). In addition, one also
needs to consider that the process used to remove CdCl2 residuals must also produce
the CdTe surface stoichiometry required for specific contact functionality — a p-
ZnTe contact interface requires a stoichiometric CdTe surface, whereas many other
contact designs benefit from a Te-rich surface layer. Finally, it has been suggested
that certain chemical surface treatments may be effective at reducing the effect
of weak diodes and/or microshunts that may exist in some types of CdTe layers
(Karpov et al., 2003).
Once the type of surface treatment is determined, one must next realise that
the electrical quality of the as-deposited and/or CdCl2 -treated bulk CdTe layer is
insufficient for effective junction operation. Depending on the CdTe source material
used, the specifics of the CdTe deposition, and CdCl2 treatment processes, the net
acceptor density will generally be too low (<10−12 cm−3 ) for optimal device oper-
ation, and/or the material may be n-type. Furthermore, even if the acceptor density
is sufficiently high, the minority-carrier lifetime is generally too short (<0.5 ns).
Most contacting processes utilise a contact interface layer (CIFL) that can
diffuse one or more dopant species into the CdTe layer. Although Cu has historically
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 232

232 T. A. Gessert & D. Bonnet

been used for the active diffusing dopant species, other Group I species (Au or Ag)
or Group V species (P, As, Sb or Bi) have also demonstrated potential (Zanio, 1978).
While the precise defect formation that occurs during dopant diffusion remains
debated, it is generally accepted that a successful diffusion alters the electrical
properties of the underlying CdTe layer so that it becomes sufficiently p-type to
establish a strong field (i.e. narrow space-charge region) in the ∼0.5–1 µm of the
device nearest the CdS layer (Gessert et al., 2006). Furthermore, Cu diffusion
at appropriate temperatures (∼250–350◦C) has been found to increase electron
lifetime near the CdS/CdTe interface (Gessert et al., 2009; Gessert et al., 2010).
As mentioned in Section 5.6.5, it is possible that the CdCl2 -treated CdTe
layers may be intrinsic or n-type. This suggestion is supported by contact studies
where dopant diffusion can be limited to very low concentrations. These show,
for CdTe deposited at intermediate temperatures (450–575◦C), devices formed
without Cu dopant diffusion from the contact typically demonstrate very little (if
any) rectification (Gessert et al., 2002). However, when even a small amount of
Cu is allowed to diffuse from the contact, strong junction rectification is observed,
suggesting that contact diffusion may be converting the layer from n-type to p-type.
Additional differences in the response of the CdTe layer to contact diffusion
are observed between layers grown at intermediate (450–575◦C) and high temper-
ature (600–625◦C). Capacitance–voltage analysis of devices grown at intermediate
temperatures, shown in Fig. 5.11, indicates that initial dopant diffusion from the
contact causes the junction space charge to contract systematically. This suggests
that even the initial dopant diffusion from the contact leads to an increase in net
acceptor concentration (Gessert et al., 2005; Gessert et al., 2006]. In contrast, con-
tact diffusion into material grown at high temperatures (600–625◦C) causes the
space charge to first expand, and then contract (Fig. 5.11). This suggests that the
net acceptor density decreases with initial dopant diffusion, and thereafter increases
(Chin et al., 2010).
It has recently been reported that this seemingly confusing functionality can
be explained by noting that higher deposition temperature will possibly yield CdTe
layers in which the effective ionisation energy changes with the extent of Cu dif-
fusion (Ma et al., 2011). Specifically, high-temperature deposition yields CdTe
with more cadmium vacancy defects (VCd ) than would be present at intermedi-
ate temperatures. Further, the VCd exists as a double acceptor, with both a lower
(∼130 meV) and a higher ionisation energy (∼210 meV). As Cu diffuses into the
CdTe, it preferentially substitutes onto available VCd sites and becomes an acceptor
CuCd . However, the ionisation energy of CuCd (∼220 meV) is greater than either
of the VCd activation energies, and so the observed net acceptor density (NA )
decreases. This reduction in NA continues until the available cadmium vacancies
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 233

Polycrystalline Cadmium Telluride Photovoltaic Devices 233

Figure 5.11 Capacitance–voltage analysis of two sets of devices produced at intermedi-

ate temperatures (450–575◦ C, labelled Hot Pocket (HP) and VTD) and one set produced at
higher temperature labelled CSS. For the two sets produced at intermediate temperature, the
space charge decreases systematically with increasing contacting temperature (i.e. increas-
ing Cu incorporation). However, for the device produced at high temperature (labelled CSS),
the space charge initially increases and then decreases, suggesting that the net acceptor den-
sity initially decreases and then increases.

are consumed. At this point, remaining Cu can either displace the Cd in the CdTe
lattice (the formation energy of CuCd in CdTe is ∼1.31 eV) to form additional CuCd
acceptors, or form Cu interstitial donors (Cui ) (Wei and Zhang, 2002). Recent stud-
ies have shown, depending on chemical potential (i.e. whether CdTe is formed with
excess Cd or Te concentration), the CuCd defect formation is preferred over Cui
up to a certain defect concentration, at which point formation of the compensating
Cui will commence and pin the Fermi level (Ma et al., 2011). Further, depending
on the particular thermal process conditions following Cu diffusion, the Cui can
begin to dominate, type-converting the material back to n-type.
In contrast to doping the CdTe layer from the contact and producing the n − p
junction functionality described above, an alternative contact design is intended to
yield a low-resistance CdTe/CIFL interface while sustaining a CdTe layer that is
electrically intrinsic (i.e. not extrinsically doped p-type) while still demonstrating
a long minority-carrier lifetime (i.e. n −i − p junction functionality) (Sites and Pan,
2007). For this junction design, if the intrinsic CdTe layer demonstrates sufficiently
long lifetime, it may also be advantageous for the CIFL to function as an electron
reflector. Modelling suggests wider-bandgap tellurides or their alloys (e.g. ZnTe,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 234

234 T. A. Gessert & D. Bonnet

CdZnTe or CdMgTe) could be appropriate CIFL choices in this application (Hsiao

and Sites, 2009).
Regardless of the type of junction functionality with which the contacting
process must align, achieving high device performance requires that the resistance
at the CdTe/CIFL interface be minimised. The strategies to accomplish this fall
in one of two broad categories: Category 1 — A CIFL is chosen that will form a
relatively low barrier height between the CIFL and CdTe, while the barrier width
is reduced as much as possible by increasing p-type doping in the CdTe through
diffusion from the CIFL (Sites and Pan, 2007). CIFLs that fall into this category
include those where a Te or CuTe layer is formed or deposited, and Cu is diffused
from or through it into the CdTe; Category 2 — choosing a CIFL that facilitates
valence-band alignment between the CdTe and CIFL, and can also be degenerately
doped p-type. For this contact, holes in the valence band will not experience a
barrier as they move across the CdTe/CIFL interface (i.e. the interface will be a
low-resistance pathway), while transport from the (degenerately doped) CIFL into
an outer metallisation will be via low-resistance quantum-mechanical tunnelling
(Gessert et al., 2003; Gessert et al., 2007). One CIFL that falls into this category is
p-ZnTe:Cu; however CIFLs that include SbTe or alloys of HgTe may also benefit
from this functionality.
Once an appropriate CIFL is deposited onto the CdTe, metal layers are gener-
ally deposited on top of the CIFL. These layers provide several functions including
reducing lateral resistivity, limiting moisture ingress, and/or providing functional-
ity to enhance stability. For commercial modules, these outer layers must also be
developed to provide attributes such as allowance for high deposition temperature
without diffusion, high adhesion and low cost.

5.7 Where is the junction?

Having established a basic appreciation of what each layer of a CdTe PV device may
do in various device designs, a final question remains as to where the semiconductor
junction is actually located. We now know that the specific parameters of the CdS
and CdTe deposition combine with the parameters of the CdCl2 treatment to affect
the formation and extent of the CdSTe layer. Furthermore, we know that diffusion
from the CIFL additionally alters the electrical properties of the CdTe, thereby
altering the junction. However, we still have not discussed where the junction is.
To understand this better, we must also consider that when sulphur is diffused
into the CdTe layer (forming a CdSTe interdiffused layer), the bandgap of the
CdSTe alloy decreases relative to CdTe or CdS. This is called band bowing, and
is shown schematically in Fig. 5.12a. At a temperature of ∼600◦ C, depending on
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 235

Polycrystalline Cadmium Telluride Photovoltaic Devices 235

(a)

(b)
Figure 5.12 (a) Diagram illustrating the effect of band bowing and phase separation in
the CdS x Te1−x layer. The circle at x =∼ 0.1 and 620◦ C indicates the most likely sulphur
concentration, based on the effective bandgap and the available single-phase region; (b)
diagram illustrating the location of the junction in a CdS/CdTe PV device from a historic
(top) and more recent (bottom) perspective.

the stoichiometry of the CdSTe alloy, this can exist as a single-phase, hexagonal
material (wurtzite, ∼85–100% sulphur), a single-phase cubic material (zincblende,
0–∼15% sulphur), or a mixed-phase CdSTe material (between ∼15 and ∼85%
sulphur) (Nunoue et al., 1990). During intermixing of the CdS and CdTe layers
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 236

236 T. A. Gessert & D. Bonnet

at the CdTe deposition temperature, it is believed that all three of these phases
will be present in what has so far been described simply as the CdSTe layer.
Many technologists believe that it is reasonable to assume the CdSTe layer will be
primarily in the wurtzite (i.e. S-rich) phase nearest the CdS layer, primarily in the
zincblende (Te-rich) phase nearest the CdTe layer, and in the mixed phase between
these two (primarily) single-phase endpoints. However, only a few studies of this
region have been reported (Dhere et al., 2008).
For the type of devices produced at NREL, we believe the electrical junction
forms between the non-alloyed CdTe, and a diffusion-formed zincblende CdSTe
(Dhere et al., 2008). Analysis of the IR absorption characteristics of quantum
efficiency data indicates the effective sulphur concentration in the junction region is
∼10%. This is estimated by calculating the effective bandgap from device quantum
efficiency (∼1.46 eV), while accounting for the known band bowing (Compaan
et al., 1996). This estimation process is shown graphically in Fig. 5.12a. Two
schematic diagrams illustrating both the historic and more recent understanding of
the layer structure, as well as the associated electric fields, are shown in Fig. 5.12b.

5.8 Considerations for large-scale deployment

The previous discussion suggests that CdTe thin-film PV modules are a viable
option for large-scale, long-term, renewable electricity generation. This is because
the specific cost of these thin-film modules is (or is quickly becoming) less than
the price of delivering electricity in many parts of the world. Although economic
viability is a necessary element for both continued production expansion and soci-
etal acceptance, the promise of thin-film PV as a source of sustainable base energy
requires additional issues to be considered. Even though a thorough treatment of
any of the following topics is beyond the scope of this chapter, the following discus-
sion suggests why many believe these topics are rapidly becoming just as important
as specific cost in the widespread establishment of any PV energy infrastructure.

5.8.1 Reliability
As any PV technology matures from the laboratory to commercial production, a
point is reached when product reliability must be understood well enough to assign a
product warranty. Presently, most CdTe thin-film PV module manufacturers like to
align module warranties with established PV alternatives (particularly wafer-based
crystalline Si). This level of reliability is typically consistent with the panel deliv-
ering at least 80% of its initial power after 25 years of field deployment. Assuming
a constant exponential degradation, this equates to about 0.9% power loss per
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 237

Polycrystalline Cadmium Telluride Photovoltaic Devices 237

year. Because thin-film manufacturers need to ensure their products demonstrate

reliability consistent with their warranty, developing techniques that can rapidly,
accurately and cost-effectively probe reliability expectations remains one of the
more critical challenges for both public- and private-sector PV researchers.
Another evolving value consideration relates to expectations of post-warranty
deployment. Like many older cars that are still on the road after their warranty
has expired, many PV modules have exceeded their warranty period, but remain
usefully deployed. Although degraded modules may not be appropriate for many
applications, modules with de-rated performance will continue to have signifi-
cant value in many other off-grid and/or remote-power applications. It therefore
seems reasonable to assume that designing CdTe module reliability for time periods
exceeding the warranty could impart (marketing) advantages to new PV products.

5.8.2 Minerals availability

As production of all types of PV technologies expand, there is increasing debate
as to whether minerals availability will become a significant factor in module cost.
This is especially true for present thin-film alternatives because each uses a mineral
component that is perceived to be relatively scarce. The minerals most typically
mentioned in this context are Ge (for some amorphous Si devices), In (for CIGS
devices) and Te (for CdTe devices). A direct answer in this debate is complicated
because, like other (secondary) minerals that are considered scarce, most of these
minerals are presently co-processed along with other (primary) minerals.
Because of this interdependence, parameters that will influence Te availabil-
ity include the following: (i) the amount of production of the primary ore (i.e.,
presently Cu is the primary ore for Te); (ii) concentration of the secondary mineral
in the primary ore; (iii) industry incentive to co-process the secondary mineral (i.e.
emerging and/or multiple markets for the secondary ore); and (iv) environmen-
tal, governmental or practical constraints related to extracting the mineral. The
main message is that the business plan of anyone considering the establishment of
a large-scale production facility for any PV technology should include thorough
advisement from a reputable mineral geologist.

5.8.3 Environmental considerations

It is well known that although PV modules do not produce toxic waste or greenhouse
gas when operating, toxic products and/or by-products can be produced during the
initial production and recycling phases of their life cycle. This waste is the primary
component of what is often referred to as the ‘total life-cycle eco-toxicity’ of any
PV technology. Although this eco-toxicity derives from many sources, much of
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 238

238 T. A. Gessert & D. Bonnet

it depends on the particular chemicals used to process subcomponents, as well as

the burning of fossil fuels to mine, transport or refine raw materials, or provide the
electricity used to produce the modules.
Realising that different module-production methods may use different
amounts and types of chemicals, and that different countries use different amounts,
types and mixtures of fossil fuels to produce and deliver their electricity, it is easy
to see how associating a given level of eco-toxicity with any particular PV technol-
ogy involves significant assumptions about both production process and location.
Furthermore, the energy mix will continue to evolve as more PV-derived elec-
tricity is used to fuel the production of PV modules (often called the breeder
PV energy-balance scenario). Nevertheless, reports are beginning to suggest some
useful guidance for technology-dependent reduction in PV eco-toxicity (Fthenakis,
2004; Raguei et al., 2007; Held and Ilg, 2008). For example, a primary source of
eco-toxicity in glass-encapsulated CdTe thin-film modules arises from the produc-
tion of the glass. This suggests that significant opportunities may exist to reduce
both eco-toxicity (as well as to reduce the energy used to make the module) by
developing module designs that incorporate either less glass or glass produced in
such a way to generate less eco-toxicity.

5.8.4 Energy payback time

The energy payback time (EPT) of a PV module is the amount of time a module
must produce power to recover the energy it took to produce the module initially.
Although assumptions vary, the calculation of the energy to produce the module
should be as inclusive as possible, accounting for everything from the energy
needed to mine, transport, refine, produce and deliver all module subcomponents
to that required to deposit/assemble/package the module, deploy it, and eventually
recycle it at the end of its life.
Within this definition, if a particular module is expected to last for 25 years,
and it demonstrates an EPT of five years, one can roughly expect to get about five
times more energy out of the module than was used to produce it. Although this
ratio of energy returned-to-invested sounds great, a five-year EPT may still be too
long when one accounts for current production-growth scenarios.
The growth scenario discussed previously might be consistent with PV pro-
duction and efficiency levels of several years ago. In this case, as long as the PV
module produced more energy in its expected lifetime (∼15–25 years) than it took
to produce the module, it was considered a net-energy producer. In contrast, we
presently have PV production technologies and module efficiencies that yield EPTs
of two years — or less. An EPT of two years would allow a module with a 25-year
warranty to produce more than ten times the original energy investment during its
lifetime.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 239

Polycrystalline Cadmium Telluride Photovoltaic Devices 239

Although this sounds both environmentally and financially attractive, a serious

complication arises when the growth in PV production becomes so rapid that the
expected doubling time of production rate becomes shorter than the EPT. In this
rapid-growth scenario, an energy balance at any given time during production
expansion can only be achieved if the EPT is equal to or less than the expected
production doubling time. Otherwise, additional amounts of traditional energy
resources (e.g. fossil fuels) may have to be supplied to provide the energy to
sustain the rapid growth in PV production. Hence, if PV is to be a net energy
producer during this period of rapidly expanding production, the EPT may need
to be significantly less than the production doubling time. Specifically, the 46%
exponential growth in PV production observed during the past ten years indicates
that EPTs need to be less than two years. Therefore, the primary question is this:
what is the EPT for different PV technologies, and specifically for thin-film CdTe
PV technologies?
Recent reports have indicated that most thin-film technologies may have
advantages over wafer Si in this respect, with CdTe demonstrating EPTs of between
six months and one year (Raugei et al., 2007; Held and Ilg, 2008).

5.9 Conclusions
This chapter has presented a technological description of components and process
alternatives consistent with the present generation of CdTe thin-film PV devices. We
have also suggested why this technology may embody considerable opportunity for
becoming a significant part of future large-scale electricity production. However,
even when economic requirements are met, the establishment of a sustainable
energy-production infrastructure from CdTe technologies will require additional
consideration of longer-term issues. These include reliability, materials availability,
environmental considerations and energy payback time.

Acknowledgements
The authors wish to thank R. G. Dhere, D. S. Albin, T. M. Barnes, J. N. Duenow,
S.-H. Wei and T. J. Coutts of NREL for assistance and thoughtful suggestions
regarding the preparation of this chapter. Portions of this work were supported
under DOE Contract No. DE-AC36-08-GO28308 to NREL.

References
Albin D. S., Demtsu S. H. and McMahon T. J. (2006), ‘Film thickness and chemical pro-
cessing effects on the stability of cadmium telluride solar cells’, Thin Solid Films 515,
2659–2668.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 240

240 T. A. Gessert & D. Bonnet

Albin D. S., Yan Y. and Al-Jassim M. M. (2002), ‘The effect of oxygen on interface
microstructure evolution in CdS/CdTe solar cells’, Progr. Photovoltaics 10, 309–322.
Aranovich J., Golmayo D., Fahrenbruch A. L. and Bube R. H. (1980), ‘Photovoltaic prop-
erties of ZnO–CdTe heterojunctions prepared by spray pyrolysis’, J. Appl. Phys. 51,
4260–4268.
Artobolevskaya E. S., Afanaseva E. A., Vodopyanov L. K. and Sushkov V. P. (1967), ‘Pho-
toelectromagnetic effect in cadmium telluride’, Sov. Phys. Semicond. 1, 1531.
Basol B. M. (1984), ‘High-efficiency electroplated heterojunction solar cell’, J. Appl. Phys.
55, 601–603.
Bell R. O., Serreze H. B. and Wald F.V. (1975), ‘A new look at CdTe solar cells’, Proc. 11th
IEEE Photovoltaic Specialists Conf., Scotsdale, AZ, pp. 497–502.
Bonnet D. and Rabenhorst H. (1972), ‘New results on the development of thin film p-
CdTe/n-CdS heterojunction solar cell’, Proc. 9th IEEE Photovoltaic Specialists Conf.,
Silver Springs, MD, pp. 129–132.
Braun G. W. and Skinner D. E. (2007), ‘Experience scaling-up manufacturing of emerging
photovoltaic technologies’, Natl. Renewable Energy Laboratory Subcontract Report,
NREL/SR-640-39165, January 2007.
Bube R. H. (1988), ‘CdTe junction phenomena’, Solar Cells 23, 1–17.
Bube R. H., Buch F., Fahrenbruch A. L., Ma Y. Y. and Mitchell K. W. (1977), ‘Photovoltaic
energy conversion with n-CdS- p-CdTe heterojunctions and other II–VI junctions’,
IEEE Trans. Electron. Dev. ED-24, 487–492.
Bube R. H., Fahrenbruch A., Aranovich J., Buch F., Chu M. and Mitchell K. (1975), ‘Applied
research in II–VI compound materials for heterojunction solar cells’, NSF Report No.
NSF/RANN/SE/AER-75-1679/75/4.
Carmody M., Mallick S., Margetis J., Kodama R., Biegala T., Xu D., Bechmann P., Garland
J. W. and Sivananthan S. (2010), ‘Single-crystal II–VI on Si single-junction and tandem
solar cells’, Appl. Phys. Lett. 96, 153502–153502-3.
Chin K. K., Gessert T. A. and Wei S.-H. (2010), ‘The roles of Cu impurity states in CdTe
thin film solar cells’, Proc. 35th IEEE Photovoltaic Specialists Conf., Piscataway, NJ,
pp. 1915–1918.
Chu T. L. (1988), ‘Cadmium telluride solar cells’, in Coutts T. J. and Meakin J. D., (eds),
Vol. 3 of Current Topics in Photovoltaics, Academic Press, New York.
Chu T. L., Chu S. S., Britt J., Ferekides C., Wang C., Wu C. Q. and Ullal H. S. (1992),
‘14.6% Efficient thin-film cadmium telluride heterojunction solar cell’, IEEE Electron
Dev. Lett. 13, 303–304.
Coggeshall C. and Margolis R. (2011), Solar Vision Study, U.S. Dept. of Energy Report.
Compaan D., Feng Z., Contreas-Puente G., Narayanswamy C. and Fisher A. (1996),
‘Properties of pulsed laser deposited CdSx Te1−x films on glass’, MRS Proc. 426,
367–372.
Corwine C., Sites J. R., Gessert T. A., Metzger W. K., Dippo P., Li J., Duda A. and Teeter
G. (2005), ‘CdTe photoluminescence: comparison of solar-cell material with surface-
modified single crystals’, Appl. Phys Lett. 86, 221909–221909-3.
Coutts T. J., Young D. L. and Li X. (2000), ‘Characterization of transparent conducting
oxides’, MRS Bulletin 25, 58–65.
Cusano D. A. (1963), ‘CdTe solar cells and photovoltaic heterojunctions in II–IV Com-
pounds’, Solid-State Electron. 6, 217–218.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 241

Polycrystalline Cadmium Telluride Photovoltaic Devices 241

Cusano D. A. (1967), ‘Thin film studies and electro-optical effects’, in Aven M. and Prener
J. S. (eds.), Physics and Chemistry of II–VI Compounds, North-Holland Publishing
Co., Amsterdam, Ch. 14, pp. 709–766.
Cusano D. A. and Lorenz M. R. (1964), ‘CdTe hole lifetime from the photovoltaic effect’,
Solid State Commun. 2, 125–128.
Dhere R. G., Albin D. S., Rose D. H., Asher S. E., Jones K. M., Al-Jassim M. M., Moutinho
H. R. and Sheldon P. (1996), ‘Intermixing at the CdS/CdTe interface and its effect on
device performance’, MRS Proc. 426, 361–366.
Dhere R. G., Bonnet-Eymard M., Charlet E., Peter E., Duenow J. N., Li J. V., Kuciauskas D.
and Gessert T. A. (2011), ‘CdTe solar cell with industrial Al:ZnO on soda-lime glass’,
Thin Solid Films 519, 7142–7145.
Dhere R. G., Duenow J. N., DeHart C. M., Li J. V., Kuciauskas D. and Gessert T. A.,
(2012), ‘Development of substrate structure CdTe photovoltaic devices with perfor-
mance exceeding 10%’, Proc. 38th IEEE Photovoltaic Specialists Conf., IEEE, Piscat-
away, NJ, pp. 3208–3211.
Dhere R. G. and Li X. (2012), NREL, private communication.
Dhere R. G., Zhang Y., Romero M. J., Asher S. E., Young M., To B., Noufi R. and Gessert
T. A. (2008), ‘Investigation of junction properties of CdS/CdTe solar cells and their
correlation to device properties’, Proc. 33rd IEEE Photovoltaic Specialists Conf., San
Diego, CA, Manuscript No. 279.
Dhese K. A., Devine P., Ashenford D. E., Nicholls J. E., Scott C. G., Sands D. and Lunn B.
(1994), ‘Photoluminescence and p-type conductivity in CdTe:N grown by molecular
beam epitaxy’, J. Appl. Phys. 76, 5423–5428.
Duenow J. N., Dhere R. G., Kuciauskas D., Li J. V., Pankow J. W., Dippo P. C., DeHart C.
M. and Gessert T. A., ‘Oxygen incorporation during fabrication of substrate CdTe pho-
tovoltaic devices’, Proc. 38th IEEE Photovoltaic Specialists Conf., IEEE, Piscataway,
NJ, pp. 3225–3229.
Fthenakis V. M. (2004), ‘Life cycle impact analysis of cadmium in CdTe PV production’,
Renewable and Sustainable Energy Reviews 8, 303–334.
Fulop G., Doty M., Meyers P., Betz J. and Liu C. H. (1982), ‘High-efficiency electrodeposited
cadmium telluride solar cells’, Appl. Phys. Lett. 40, 327–328.
Garcia D. A., Entine G. and Tow D. E. (1974), ‘Detection of small bone abscesses with a
high-resolution cadmium telluride probe’, J. Nucl. Med. 15, 892–895.
Gessert T. A., Asher S., Johnston S., Duda A., Young M. R. and Moriarty T. (2006), ‘For-
mation of ZnTe:Cu/Ti contacts at high temperature for CdS/CdTe devices’, Proc. 4th
World Conf. Photovoltaic Energy Conversion, pp. 432–435.
Gessert T. A., Asher S., Johnston S., Young M., Dippo P. and Corwine C. (2007), ‘Analysis
of CdS/CdTe devices incorporating a ZnTe:Cu/Ti contact’, Thin Solid Films 515, 6103–
6106.
Gessert T. A., Burst J., Li X., Scott M. and Coutts T. J. (2011), ‘Advantages of transparent
conducting oxide thin films with controlled permittivity for thin film photovoltaic solar
cells’, Thin Solid Films 519, 7146–7148.
Gessert T. A., Dhere R. G., Duenow J. N., Kuciauskas D., Danevce A. and Bergeson J. D.
(2010), ‘Comparison of minority carrier lifetime measurements in superstrate and sub-
strate CdTe PV devices’, Proc. 37th IEEE Photovoltaic Specialists Conf., IEEE, Pis-
cataway, NJ, pp. 335–339.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 242

242 T. A. Gessert & D. Bonnet

Gessert T. A., Metzger W. K., Dippo P., Asher S. E., Dhere R. G. and Young M. R. (2009),
‘Dependence of carrier lifetime on Cu-contacting temperature and ZnTe:Cu thickness
in CdS/CdTe thin film solar cells’, Thin Solid Films 517, 2370–2371.
Gessert T. A., Perkins C. L., Asher S. E., Duda A. and Young M. R. (2003), ‘Study of
ZnTe:Cu/metal interfaces in CdS/CdTe photovoltaic solar cells’, Mater. Res. Soc.
Symp., Vol. 796, pp. 79–84.
Gessert T. A., Romero M. J., Johnston S., Keys B. and Dippo P. (2002), ‘Spectroscopic
cathodoluminescence studies of the ZnTe:Cu contact process for CdS/CdTe solar cells’,
Proc. 29th IEEE Photovoltaic Specialists Conf., IEEE, Piscataway, NJ, pp. 535–538.
Gessert T. A., Romero M. J., Perkins C. L., Asher S. E., Matson R., Moutinho H. and
Rose D. (2001), ‘Microscopic residuals on polycrystalline CdTe following wet CdCl2
treatment’, MRS Proc. 668, H1.10.1–H1.10.6.
Gessert T. A., Smith S., Moriarty T., Young M., Asher S., Johnston S., Duda A. and DeHart C.
(2005), ‘Evolution of CdS/CdTe device performance during Cu diffusion’, Proc. 31st
IEEE Photovoltaic Specialists Conf., IEEE, Piscataway, NJ, pp. 291–294.
Gloeckler M., Sankin I. and Zhao Z. (2013), ‘CdTe solar cells at the threshold to 20%
efficiency’, IEEE J. Photovoltaics 3, 1389–1398.
Green M. A., Emery K., Hishikawa Y., Warta W. and Dunlop E. D. (2012), ‘Solar cell
efficiency tables (Version 39)’, Progr. Photovoltaics 20, 12–20.
Green M. A., Emery K., Hishikawa Y., Warta W. and Dunlop E. D. (2013), ‘Solar cell
efficiency tables (Version 42)’, Progr. Photovoltaics 21, 827–837.
Greenberg J. H. (1996), ‘P–T–X phase equilibrium and vapor pressure scanning of non-
stoichiometry in CdTe’, J. Cryst. Growth 161, 1–11.
Gretener C., Perrenoud J., Kranz L., Kneer L., Schmitt R., Buecheler S. and Tiwari N. (2012),
‘CdTe/CdS thin film solar cells grown in substrate configuration’, Progr. Photovoltaics
21, 1580–1586.
Gur I., Fromer N. A., Geier M. L. and Alivisatos A. P. (2005), ‘Air-stable all-inorganic
nanocrystal solar cells processed from solution’, Science 310, 462–465.
Halliday D. P., Potter M. D. G., Boyle D. S. and Durose K. (2001), ‘Photoluminescence
characterization of ion implanted CdTe’, MRS Proc.668, H1.8.1.
Held M. and Ilg R. (2008), ‘Life cycle assessment (LCA) of CdTe thin film PV modules and
material flow analysis (MFA) of cadmium within EU27’, Proc. 23rd Eur. Photovoltaic
Solar Energy Conf., Sept. 1–5, Valencia.
Hsiao K. J. and Sites J. R. (2009), ‘Electron reflector strategy for CdTe solar cells’, Proc.
34th IEEE Photovoltaic Specialists Conf., Fort Collins, CO, pp. 1846–1850.
Jordan J. F. (1993), ‘Photovoltaic cell and method’, U.S. Patent No. 5261968.
Karpov V. G., Shvydka D., Roussillon Y. and Compaan A. D. (2003), ‘The mesoscale physics
of large-area photovoltaics’, Proc. 3rd World Conf. Photovoltaic Energy Conversion,
WCPEC–3, Osaka, pp. 495–498.
Kester J. L., Albright S., Kaydanov V., Ribelin R., Woods L. M. and Phillips J. A. (1996),
‘CdTe solar cells: electronic and morphological properties’, AIP Conf. Proc. Ser. 394,
162–169.
Kranz L., Schmitt R., Gretener C., Perrenoud J., Pianezzi F., Uhl A. R., Keller D., Buecheler
S. and Tiwari A.N. (2013), ‘Progress towards 14% efficient CdTe solar cells in substrate
configuration’, Proc. 39th IEEE Photovoltaic Specialists Conf., Tampa, FL, 1644–
1648.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 243

Polycrystalline Cadmium Telluride Photovoltaic Devices 243

Levi D., Albin D. and King D. (2000), ‘Influence of surface composition on back-contact
performance in CdTe/CdS PV devices’, Progr. Photovoltaics 8, 591–602.
Li X., Gessert T. A., Matson R. J., Hall J. F. and Coutts T. J. (1993), ‘Microstructural study
of sputter-deposited CdTe thin films’, J. Vac. Sci. Technol. A 12, 1608–1613.
Li X., Niles D. W., Hasoon F. S., Matson R. J. and Sheldon P. (1999), ‘Effect of nitric–
phosphoric acid etches on materials properties and back-contact formation in CdTe-
based solar cells’, J. Vac. Sci. Technol. A 17, 805–809.
Ma J., Wei S.-H., Gessert T. A. and Chin K. K. (2011), ‘Carrier density and compensation
in semiconductors with multiple dopants and multiple transition energy levels: case of
Cu impurities in CdTe’, Phys. Rev. B 83, 245207–245214.
McCandless B. (2003), ‘Effects of treatments on CdTe film conductivity’, NREL Subcon-
tract Report #ADJ-1-30630-12, 23 December.
McCandless B. E., Youm I. and Birkmire R. W. (1999), ‘Optimization of vapor post-
deposition processing for evaporated CdS/CdTe solar cells’, Progr. Photovoltaics 7,
21–30.
Mehata S. (2010, ‘PV news annual data collection result: 2010 cell, module production
explodes past 20 GW’, PV News 29, May 2010.
Meyer E., Martini M. and Sternberg J. (1972), ‘Measurement of the disappearance rate of
Se-75 sodium selenite in the eye of the rat by a CdTe medical probe’, IEEE Trans.
Nucl. Sci. 19, 237–243.
Moutinho H. R., Dhere R. G., Al-Jassim M. M., Levi D. H. and Kazmerski L. L. (1999),
‘Investigation of induced recrystallization and stress in close-spaced sublimated and
radio-frequency magnetron sputtered CdTe thin films’, J. Vac. Sci. Technol. A 17,
1793–1798.
Naumov G. P. and Nikolaeva O. V. (1961), ‘The efficiency of transformation of direct solar
radiation energy into electric energy using a CdTe photocell’, Soviet Phys. Solid State
3, 2718.
Nicoll F. H. (1963), ‘The use of close spacing in chemical-transport systems for growing
epitaxial layers of semiconductors’, J. Electrochem. Soc. 110, 1165–1167.
Nunoue S.-Y., Hemmi T. and Kato E. (1990), ‘Mass spectrometric study of the phase bound-
aries of the CdS–CdTe system’, J. Electrochem. Soc., 137, 1248–1251.
Powell R. C., Dorer G. L., Reiter N. A., McMaster H. A., Cox S. M. and Kahle T. D.
(1999), ‘Apparatus and method for depositing a material on a substrate’, U.S. Patent
No. 5945163.
Raugei M., Bargigli S. and Ulgiati S. (2007), ‘Life cycle assessment and energy pay-back
time of advanced photovoltaic modules: CdTe and CIS compared to poly-Si’, Energy
32, 1310–1318.
Romeo N., Bosio A. and Canevari V. (1992), ‘Large crystalline grain CdTe thin films for
photovoltaic application’, Int. J. Solar Energy 12, 183–186.
Rose D. H (1997), ‘The effects of oxygen on CdTe–absorber solar cells deposited by close-
spaced sublimation’, PhD Thesis, University of Colorado, Boulder, CO.
Rose D. H., Hasoon F. S., Dhere R. G., Albin D. S., Ribelin R. M., Li X. S., Mahathongdy Y.,
Gessert T. A. and Sheldon P. (1999), ‘Fabrication procedures and process sensitivities
for CdS/CdTe solar cells’, Progr. Photovoltaics 7, 331–340.
Sasala R. A., Powell R. C., Dorer G. L. and Reiter N. (1996), ‘Recent progress in CdTe
solar cell research at SCI’, AIP Conf. Proc Ser. 394, 171–186.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch05 page 244

244 T. A. Gessert & D. Bonnet

Singh V. P., McClure J. C., Lush G. B., Wang W., Wang X., Thompson G. W. and Clark
E. (1999), ‘Thin film CdTe-CdS heterojunction solar cells on lightweight metal sub-
strates’, Solar Energy Mater. Solar Cells 59, 145–161.
Sites J. and Pan J. (2007), ‘Strategies to increase CdTe solar-cell voltage’, Thin Solid Films
515, 6099–6102.
Suyama N., Arita T., Nishiyama Y., Ueno N., Kitamura S. and Murosono M. (1990),
‘CdS/CdTe solar cells by the screen-printing–sintering technique’, Proc. 21st IEEE
Photovoltaic Specialists Conf., Kissimmee, FL, pp. 498–503.
Tatarenko S., Bassani F., Saminadayar K., Cox R. T., Jouneau P. H. and Megnea N. M.
(1993), ‘Indium doping of (001), (111) and (211) CdTe layers grown by molecular
beam epitaxy’, J. Cryst. Growth 127, 318–322.
Tyan Y.-S. and Perez-Albuerne E. A (1982), ‘Efficient thin-film CdS/CdTe solar cells’, Proc.
16th IEEE Photovoltaic Specialists Conf., San Diego, CA, pp. 794–800.
Tyan Y.-S., Vazan F. and Barge T. S. (1984), ‘Effect of oxygen on thin-film CdS/CdTe solar
cells’, Proc. 17th IEEE Photovoltaic Specialists Conf., Kissimmee, FL, pp. 840–844.
Valdna V., Hiie J. and Gavrilov A. (2001), ‘Defects in Cl-doped CdTe thin films’, Solid State
Phenom. 80, 155–162.
Vodakov Y. A., Lomakina G. A., Naumov G. P. and Maslakovets Y. P. (1960), ‘Properties
of p − n junctions in cadmium telluride photocells’, Soviet Phys. Solid State 2, 11.
Waters D. M., Niles D., Gessert T., Albin D., Rose D. and Sheldon P. (1988), ‘Surface analysis
of CdTe after various pre-contact treatments’, Proc. 2nd World Conf. Photovoltaic Solar
Energy Conversion, Vienna, European Commission, Luxembourg, p. 1031.
Wei S.-H. and Zhang S. B. (2002), ‘Chemical trends of defect formation and doping limit
in II–VI semiconductors’, Phys. Rev. B 66, 155211–155221.
Wendt R., Fischer A., Grecu D. and Compaan A. D. (1998), ‘Improvement of CdTe solar cell
performance with discharge control during film deposition by magnetron sputtering’,
J. Appl. Phys. 84, 2920–2925.
Wu X. (2004), ‘High efficiency polycrystalline CdTe thin-film solar cells’, Solar Energy 77,
803–814.
Wu X., Dhere R. G., Yan Y., Romero M. J., Shang Y., Zhou J., DeHart C., Duda A., Perkins
C. and To B. (2002), ‘High efficiency polycrystalline CdTe thin-film solar cells with an
oxygenated amorphous CdS (a-CdS:O) window layer’, Proc. 29th IEEE Photovoltaic
Specialists Conf., New Orleans, LA, pp. 531–535.
Wu X., Keane J. C., Dhere R. G., DeHart C., Albin D. A., Duda A., Gessert T. A., Asher
A. S., Levi D. H. and Sheldon P. (2002) ‘16.5% Efficient CdS/CdTe polycrystalline
thin-film solar cell,’ Proc. 17th European Photovoltaic Solar Energy Conf., October
22–26, 2001,WIP-Renewable Energies, Munich, pp. 995–1000.
Wu X., Mulligan W. P. and Coutts T. J. (1996), ‘Recent developments in RF sputtered
cadmium stannate films’, Thin Solid Films 286, 274–276.
Yamaguchi K., Nakayama N., Matsumoto H. and Ikegami S. (1977), ‘CdS–CdTe solar cell
prepared by vapor phase epitaxy’, Jpn. J. Appl. Physics 16, 1203–1211.
Zanio K. (1978), Cadmium Telluride, Vol. 13 in Semiconductors and Semimetals, Academic
Press, New York.
Zunger A. (2003), ‘Practical doping principles’, Appl. Phys. Lett. 83, 57–59.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 245

CHAPTER 6

Cu(In,Ga)Se2 AND RELATED

SOLAR CELLS
UWE RAU∗ and HANS W. SCHOCK†
∗IEK5-Photovoltaik, Forschungszentrum Jülich GmbH, Germany
†Helmholtz-Zentrum Berlin für Materialien und Energie, Germany
∗[email protected]
†[email protected]

Mephistopheles: Wer will Lebendigs erkennen und beschreiben,

Sucht erst den Geist heraus zu treiben,
Dann hat er die Teile in der Hand,
Fehlt leider ! nur das geistige Band.
Encheiresin naturae nennt’s die Chemie,
Spottet ihrer selbst und weiß nicht wie.
Johann Wolfgang von Goethe, Faust, 1808.

6.1 Introduction
From the early days of photovoltaics until today, thin-film solar cells have always
competed with technologies based on single-crystal materials such as Si and GaAs.
Owing to their amorphous or polycrystalline nature, thin-film solar cells always
suffered from power conversion efficiencies lower than those of the bulk technolo-
gies. This drawback was and still is counterbalanced by several inherent advantages
of thin-film technologies. As in the early years of photovoltaics, space applications
were the driving force for the development of solar cells, the argument in favour
of thin films was their potential lighter weight as compared to bulk materials.
An extended interest in solar cells as a source of renewable energy emerged
in the mid-1970s as the limitations of fossil energy resources were widely recog-
nised. For terrestrial power applications the cost arguments and the superior energy
balance strongly favoured thin films. However, from the various materials under
consideration since the 1950s and 1960s, only three thin-film technologies, namely
amorphous (a-)Si and the polycrystalline heterojunction systems CdS/CdTe, and
CdS/CuInSe2 , have achieved industrial production in any volume.
CuInSe2 was synthesised for the first time by Hahn in 1953 (Hahn et al., 1953).
In 1974, this material was proposed as a photovoltaic material (Wagner et al., 1974)

245
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 246

246 U. Rau & H. W. Schock

with a power conversion efficiency of 12% demonstrated for a single-crystal cell.

Thin-film development achieved a historical milestone in the years 1980 to 1982,
when Boeing Corp. boosted the efficiencies of thin-film solar cells obtained from
a three-source co-evaporation process from 5.7% (Mickelsen and Chen, 1980)
to over 10% (Mickelsen and Chen, 1982). (We will discuss the co-evaporation
process in Section 6.3.1.) The Boeing result was surpassed in 1987 by ARCO
Solar with a long-standing record efficiency for a thin-film cell of 14.1% (Mitchell
et al., 1988). This company used a different approach for absorber preparation,
namely the selenisation of stacked metal layers by H2 Se (see Section 6.3.1). The
lack of reproducibility and resulting low production yield considerably delayed the
pilot production envisaged at that time. It took a further ten years before Siemens
Solar Industries in the US, after inheriting the ARCO Solar approach, entered
the stage of production. In 1998 they produced the first commercially available
Cu(In,Ga)Se2 solar modules. During the early 2000s, pilot production lines were
set up at various places in the US, Europe and Japan. During the first boom years
of photovoltaics, 2003–2007, most of these attempts resulted in industrial-scale
production capacities. However, from 2008 up to the time of writing (2013) the
overproduction crisis in the photovoltaic industry has led to a considerable shake
out of cell, module and equipment manufacturers, which has also affected the
producers of Cu(In,Ga)Se2 solar modules.
On the laboratory scale, the successful series of record efficiencies established
by the National Renewable Energy Laboratory (Contreras et al., 1999; Contreras
et al., 2005; Repins et al., 2008) was eventually topped by the Centre for Solar
Energy and Hydrogen Research in 2011 with a confirmed cell efficiency of 20.3%
on a 0.5 cm2 laboratory cell (Jackson et al., 2011). A remarkable improvement in
flexible cell efficiency was demonstrated by the Swiss Federal Laboratories for
Materials Science and Technology (EMPA) (Čhirila, 2013) with a 20.4%-efficient
small cell (0.5 cm2 ) on a polymide substrate. The Centre for Solar Energy and
Hydrogen improved the laboratory cell efficiency to 20.8%. A record (4 × 4 cm2 )
mini-module of 17.4% efficiency was developed by Solibro Research (Wallin et al.,
2012) and champion modules from various industrial production lines exceed
the efficiency level of 14%. With the perspective of production costs lower than
0.5$/Wp combined with high module efficiencies, Cu(In,Ga)Se2 still plays the role
of one of the leading thin-film photovoltaic technologies.
This chapter aims to summarise our present knowledge of Cu(In,Ga)Se2-based
heterojunction solar cells. We focus on four main areas: (i) The description of basic
material properties such as crystal properties, phase diagram and defect physics
in Section 6.2; (ii) cell technology starting from the growth of the polycrystalline
Cu(In,Ga)Se2 absorber up to device finishing by heterojunction formation and
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 247

Cu(In,Ga)Se2 and Related Solar Cells 247

window layer deposition (Section 6.3). This section also discusses options which
can be used to design the electronic properties of the absorber material as well
as basic technologies for module production; (iii) the electronic properties of
the finished heterostructure and some methods of analysing them (Section 6.4);
(iv) finally, Section 6.5 discusses the photovoltaic potential of wide-gap chalcopy-
rites, namely CuGaSe2 and CuInS2 , as well as that of the pentenary alloy system
Cu(In,Ga)(S,Se)2 and the possibility of building graded-gap structures with these
alloys. Regarding In and Ga as rare materials has stimulated the development of
the kesterite compounds, in which indium and gallium are replaced by a pair of
Group II and Group IV elements such as Zn–Sn.
This chapter can only very briefly cover those scientific issues that are rele-
vant for photovoltaic applications. For other important points and for more detailed
information, we refer the reader to the literature. More about the structural prop-
erties of Cu(In,Ga)Se2 can be found, for example, in Shay and Wernick (1975),
Kazmerski and Wagner (1985), Coutts et al. (1986), Rockett and Birkmire (1991),
Schock (1996), Bube (1998) and Rau and Schock (1999). Interface properties
of Cu(In,Ga)Se2 and related compounds have been reviewed by Scheer (1997).
For up-scaling and module technologies see, for example, Dimmler and Schock
(1996), and for economic aspects, see Zweibel (1995). A comprehensive overview
of Cu(In,Ga)Se2 (and CdTe)-based photovoltaics is further given in the book by
Scheer and Schock (2011).

6.2 Material properties

6.2.1 Basics
CuInSe2 and CuGaSe2 , the materials that form the alloy Cu(In,Ga)Se2 , belong
to the semiconducting I–III–VI2 materials family that crystallise in the tetragonal
chalcopyrite structure. The chalcopyrite structure of, for example, CuInSe2 , is
obtained from the cubic zinc-blende structure of II–VI materials like ZnSe by
occupying the Zn sites alternately with Cu and In atoms. Figure 6.1 compares the
two unit cells of the zinc-blende structure with the chalcopyrite unit cell. Each I
(Cu) or III (In) atom has four bonds to the VI atom (Se). In turn each Se atom has
two bonds to Cu and two to In. Because the strengths of the I–VI and III–VI bonds
are in general different, the ratio of the lattice constants c/a is not exactly 2:1. The
quantity 2 − c/a (which is −0.01 in CuInSe2 , +0.04 in CuGaSe2 ) is a measure of
the tetragonal distortion in chalcopyrites.
The bandgap energies of I–III–VI2 chalcopyrites are considerably smaller
than those of their binary analogues (this is the binary material where the I/III ele-
ments are replaced by their average II element; thus ZnSe is the binary analogue of
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 248

248 U. Rau & H. W. Schock

Se
Se In
Zn Cu

(a) (b)
Figure 6.1 Unit cells of chalcogenide compounds. (a) Sphalerite or zinc-blende structure
of ZnSe (two unit cells); (b) chalcopyrite structure of CuInSe2 . The metal sites in the two
unit cells of the sphalerite structure of ZnSe are alternately occupied by Cu and In in the
chalcopyrite structure.

2.4 CuGaS2

2.2

2.0
1.8
Eg /eV CuGaSe2
1.6
CuInS2
1.4
1.2 CuInSe 2

1.0
5.4 5.5 5.6 5.7 5.8
a/Angstroms

Figure 6.2 Bandgap energies E g vs. lattice constants a of Cu(In,Ga)(S,Se)2 system.

CuGaSe2 and Zn0.5 Cd0.5 Se that of CuInSe2 ). This difference is because the Cu 3d
band, together with the Se 4 p band, forms the uppermost valence band in the Cu-
chalcopyrites, which is not so in II–VI compounds. However, the system of copper
chalcopyrites covers a wide range of bandgap energies E g from 1.04 eV in CuInSe2
up to 2.4 eV in CuGaS2 , covering most of the visible spectrum. The kesterite com-
pounds Cu2 ZnSnS4 and Cu2 ZnSnSe4 have similar bandgaps to the corresponding
chalcopyrite compounds CuInS2 and CuInSe2 . Figure 6.2 summarises lattice con-
stants a and bandgap energies E g of this system. Any desired alloys between these
compounds can be produced, as there is no miscibility gap in the entire system. We
will discuss the status and prospects of this system in more detail in Section 6.5.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 249

Cu(In,Ga)Se2 and Related Solar Cells 249

6.2.2 Phase diagram

Compared with all other materials used for thin-film photovoltaics, Cu(In,Ga)Se2
has by far the most complicated phase diagram. Figure 6.3 displays the ternary
phase diagram, which comprises all ternary Cu–In–Ga compounds. This complex
ternary phase diagram can be reduced to a simpler pseudo-binary phase diagram
along the tie line between Cu2Se and In2Se3 (solid line in Fig. 6.3). Figure 6.4 shows
the phase diagram of CuInSe2 given by Haalboom et al. (1997). This investigation
had a special focus on temperatures and compositions relevant for the preparation
of thin films. The phase diagram in Fig. 6.4 shows the four different phases which
have been found to be relevant in this range: the α-phase (CuInSe2 ), the β-phase
(CuIn3 Se5 ), the δ-phase (the high-temperature sphalerite phase) and Cu2−y Se.
An interesting point is that all neighbouring phases to the α-phase have a similar
structure. The β-phase is actually a defect chalcopyrite phase built by ordered
arrays of defect pairs (Cu vacancies VCu and In–Cu anti-sites InCu ). Similarly,
Cu2−y Se can be viewed as constructed from the chalcopyrite by using Cu–In anti-
sites CuIn and Cu interstitials Cui . The transition to the sphalerite phase arises
from disordering the cation (Cu, In) sub-lattice, and leads back to the zinc-blende
structure (cf. Fig. 6.1a).
The existence range of the α-phase in pure CuInSe2 on the quasi-binary tie line
Cu2 Se–In2 Se3 extends from a Cu content of 24% to 24.5%. Thus, the existence
range of single-phase CuInSe2 is astonishingly small and does even not include the

CuIn5Se8
Se CuIn3Se5

10
Cu2In4Se 7
90
Cu3In5Se9
80 20
CuInSe2
70 30
CuSe2 In2Se3
60 40
In6Se 7
CuSe 50 50 InSe
Cu3Se2 40 60

Cu2Se 70

20 80

10 90

Cu In
90 80 70 60 50 40 30 20 10
Cu11In9

Figure 6.3 Ternary phase diagram of the Cu–In–Se system.

September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 250

250 U. Rau & H. W. Schock

β+ δ
δ α+δ
600
α+

Temperature T/°C
Cu 2 Se(HT)
β
400

α+ β α
α+
200 Cu2 Se(RT)

15 20 25 30
Cu content (at%)
single-phase region DTA cooling
two-phase region DTA heating

Figure 6.4 Quasi-binary phase diagram of CuInSe2 established by Differential Thermal

Analysis (DTA) and microscopic phase analysis. Note that at 25% Cu no single phase exists.
After Haalboom et al. (1997).

stoichiometric composition of 25% Cu. The Cu content of absorbers for thin-film

solar cells typically varies between 22 and 24 at.% Cu. At the growth temperature
this region lies within the single-phase region of the α-phase. However, at room
temperature it lies in the two-phase α + β region of the equilibrium phase diagram
in Haalboom et al. (1997). Hence one would expect a tendency for phase separation
in photovoltaic-grade CuInSe2 after deposition. Fortunately, it turns out that partial
replacement of In with Ga, as well as the use of Na-containing substrates, consid-
erably widens the single-phase region in terms of (In + Ga)/(In + Ga + Cu) ratios
(Herberholz et al., 1999). Thus, the phase diagram hints at the substantial improve-
ments actually achieved in recent years by the use of Na-containing substrates, as
well as by the use of Cu(In,Ga)Se2 alloys.

6.2.3 Defect physics of Cu(In,Ga)Se2

Basics
The role of defects in the ternary compound CuInSe2 , and even more in
Cu(In,Ga)Se2 , is of special importance because of the large number of possible
intrinsic defects and the role of deep recombination centres in the performance of
the solar cells. For insight into the defect physics of Cu(In,Ga)Se2, see Cahen (1987)
and Burgelman et al. (1997). The challenge of defect physics in Cu(In,Ga)Se2 ,
according to Zhang et al. (1998), is to explain three unusual effects in this
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 251

Cu(In,Ga)Se2 and Related Solar Cells 251

material: (i) the ability to dope Cu(In,Ga)Se2 with native defects; (ii) the structural
tolerance to large off-stoichiometries; and (iii) the electrically neutral nature of the
structural defects. It is obvious that the explanation of these effects significantly
contributes to the explanation of the photovoltaic performance of this material. It
is known that the doping of CuInSe2 is controlled by intrinsic defects. Samples
with p-type conductivity are grown if the material is Cu-poor and annealed under
high Se vapour pressure, whereas Cu-rich material with Se deficiency tends to be
n-type (Migliorato et al., 1975; Noufi et al., 1984). Thus, the Se vacancy VSe is
considered to be the dominant donor in n-type material (and also the compensat-
ing donor in p-type material), and the Cu vacancy VCu the dominant acceptor in
Cu-poor p-type material.

Theoretical considerations
By calculating the metal-related defects in CuInSe2 and CuGaSe2 , Zhang et al.
(1998) found that the defect formation energies for some intrinsic defects are so
low that they can be heavily influenced by the chemical potential of the components
(i.e. by the composition of the material) as well as by the electrochemical potential
of the electrons. For VCu in Cu-poor and stoichiometric material, the calculated
formation energy actually becomes negative. This would imply the spontaneous
formation of large numbers of these defects under equilibrium conditions. Low
(but positive) formation energies are also found for the Cu-on-In anti-site CuIn in
Cu-rich material (this defect is a shallow acceptor which could be responsible for
the p-type conductivity of Cu-rich, non-Se-deficient CuInSe2 ). The dependence
of the defect formation energies on the electron Fermi level could explain the
strong tendency of CuInSe2 to self-compensation and the difficulties of achieving
extrinsic doping. The work of Zhang et al. (1998) provides a good theoretical basis
for the calculation of defect formation energies and defect transition energies,
which exhibit good agreement with experimentally obtained data.
Further important results in Zhang et al. (1997) are the formation energies
of defect complexes such as (2Vcu ,InCu ), (CuIn ,InCu ) and (2Cui ,CuIn ), where Cui
is an interstitial Cu atom. These formation energies are even lower than those of
the corresponding isolated defects. Interestingly, (2Vcu ,InCu ) does not exhibit an
electronic transition within the forbidden gap, in contrast to the isolated InCu -anti-
site, which is a deep recombination centre. As the (2Vcu ,InCu ) complex is most
likely to occur in In-rich material, it can accommodate a large amount of excess In
(or likewise deficient Cu) and, at same time, maintain the electrical performance
of the material. Furthermore, ordered arrays of this complex can be thought as
the building blocks of a series of Cu–In–Se compounds such as CuIn3 Se5 and
CuIn5 Se8 (Zhang et al., 1997).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 252

252 U. Rau & H. W. Schock

Additionally, the microscopic reasons for metastable changes have been inten-
sively investigated in recent years. The (Vcu ,VSe ) divacancy complex (Lany and
Zunger, 2006) and the InCu anti-site defect (Lany and Zunger, 2008) have been iden-
tified as sources for the metastabilities. Since both defects exist in multiple charge
states their influence on the electronic behaviour of CIGS solar cells is rather com-
plex. For example, even in equilibrium the divacancy complex has three different
charge configurations, namely (Vcu ,VSe )+ , (Vcu ,VSe )− , and (Vcu ,VSe )3− , depen-
dent on the position in the band diagram of the ZnO/CdS/CIGS heterostructure
(Urbaniak and Igalson, 2009; Siebentritt et al., 2010).

Device-relevant defects
Let us now concentrate on the defects experimentally detected in photovoltaic
grade (and thus In-rich) polycrystalline films. In-rich material is in general highly
compensated, with a net acceptor concentration of the order of 1016 cm−3 . The
shallow acceptor level VCu (which lies about 30 meV above the valence band)
is assumed to be the main dopant in this material. As compensating donors,
the Se-vacancy VSe as well as the double donor InCu are considered. The most
prominent defect is an acceptor level about 270–300 meV above the valence band,
which has been reported by several groups from deep-level transient spectroscopy
(Igalson and Schock, 1996) and admittance spectroscopy (Schmitt et al., 1995;
Walter et al., 1996b). This defect is also present in single crystals (Igalson et al.,
1995).
As an example, Fig. 6.5 displays a defect density spectrum obtained from
admittance spectroscopy by the method of Walter et al. (1996a). The transition
at ∼300 meV exhibits a broadened energy distribution with a tail in the defect
density towards larger energies. This tail-like distribution is best described by a
characteristic energy E*, as shown in Fig. 6.5. This defect is detected, not only in
In-rich, but in equal amounts also in Cu-rich polycrystalline materials (Herberholz,
1998). An assignment of this defect to the CuIn anti-site is in agreement with the
theoretical calculations of Zhang et al. (1998) as well as with the proposition of
several experimentalists. The importance of this transition derives from the fact that
its concentration is related to the open-circuit voltage of the device (Herberholz
et al., 1997a) and that the defect seems to be involved in the defect metastability
(Igalson and Schock, 1996) (cf. Section 6.4.5).
The lower-energy transition in Fig. 6.5 is attributed to interfacial defects rather
than to a bulk defect (Herberholz et al., 1998) because its activation energy can vary
between 50 meV and 250 meV depending on air-annealing prior to the measurement
(Rau et al., 1999a). Thus, the activation energy of this transition measures the depth
∆E Fn from the vacuum level of the (electron) Fermi level and the conduction-band
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 253

Cu(In,Ga)Se2 and Related Solar Cells 253

18
10
interface defect N 1

E
Defect density Dt /cm–3 eV–1

E
17 bulk defect N 2 E
10 E

16
10

high-energy tail
D t ~ exp (–E/E*)
15
10
0 .1 0.2 0.3 0.4 0.5 0.6
Activation energy E / eV

Figure 6.5 Defect density spectrum obtained from admittance spectroscopy of a ZnO–
CdS–CuInGaSe2 heterojunction. The peaks N1 and N2 can be related to interface and bulk
defects (see inset).

energy at the Cu(In,Ga)Se2 surface (Herberholz et al., 1998), as shown in the top
right-hand of Fig. 6.5.

6.3 Cell and module technology

6.3.1 Structure of the heterojunction solar cell
The complete layer sequence of a ZnO/CdS/Cu(In,Ga)Se2 heterojunction device
is shown in Fig. 6.6. It consists of a typically 1 µm thick Mo layer deposited on
a soda-lime glass substrate and serving as the back contact for the solar cell. The
Cu(In,Ga)Se2 is deposited on top of the Mo back electrode as the photovoltaic
absorber material. This layer has a thickness of 1–2 µm. The heterojunction is then
completed by chemical bath deposition (CBD) of CdS (typically about 50 nm)
and sputter deposition of a nominally undoped (intrinsic) i -ZnO layer (usually of
thickness 50–70 nm) and then a heavily doped ZnO layer. As ZnO has a bandgap
energy of 3.2 eV it is transparent for the main part of the solar spectrum and therefore
serves as the window layer of the solar cell.
We will first mention four important technological innovations which, during
the last decade, have led to a considerable improvement of the efficiencies. These
steps are the key elements of the present Cu(In,Ga)Se2 technology.

• The film quality has been substantially improved by the crystallisation mech-
anism induced by the presence of Cu y Se (y < 2). This process is further
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 254

254 U. Rau & H. W. Schock

ZnO:Al

i-ZnO
CdS

Cu(In,Ga)Se2
Mo

glass

Figure 6.6 Schematic layer sequence of a standard ZnO-CdS-Cu(In,Ga)Se2 thin-film

solar cell.

supported by a substrate temperature close to the softening point of the glass

substrate (Stolt et al., 1993).
• The glass substrate has been changed from Na-free glass to Na-containing
soda-lime glass (Hedström et al., 1993; Stolt et al., 1993). This has led to an
enormous improvement of the efficiency and reliability of the solar cells, as
well as to a larger process tolerance. It was first assumed that this improvement
was due to better match of thermal expansion coefficients, but the beneficial
impact of Na — diffusing from the substrate through the Mo back contact —
on the growth of the absorber layer and its structural and electrical properties
was soon recognised.
• Initially, the absorbers consisted of pure CuInSe2 . The partial replacement of
In with Ga (Devaney et al., 1990) made a further noticeable improvement,
which increased the bandgap of the absorber from 1.04 eV to 1.1–1.2 eV for
high-efficiency devices. The benefit of 20–30% Ga incorporation stems not
only from the better bandgap match to the solar spectrum but also from the
improved electronic quality of Cu(In,Ga)Se2 as compared with pure CuInSe2
(Hanna et al., 2000; Herberholz et al., 1999).
• The counter-electrode for the CuInSe2 absorber of the earlier cells was a 2 µm
thick CdS layer laid down by Physical Vapour Deposition (PVD). This has
been replaced by a combination of a 50 nm thin CdS buffer layer laid down
by chemical bath deposition (Potter et al., 1985; Birkmire et al., 1989; Mauch
et al., 1991) and a highly conductive ZnO window layer.

The effect of these four items on the electronic properties and performance
of Cu(In,Ga)Se2 solar cells will be considered in detail below, as we discuss the
preparation of a Cu(In,Ga)Se2 solar cell step by step.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 255

Cu(In,Ga)Se2 and Related Solar Cells 255

6.3.2 Absorber preparation techniques

Basics
The preparation of Cu(In,Ga)Se2 -based solar cells starts with the deposition of the
absorber material on an Mo-coated glass substrate (preferably soda-lime glass).
The properties of the Mo film and the choice of the glass substrate are of primary
importance for the final device quality, because of the importance of Na, which
diffuses from the glass through the Mo film into the growing absorber material.
Some processes use blocking layers such as SiN x , SiO2 or Cr between the glass
substrate and the Mo film to prevent the out-diffusion of Na. Instead, Na-containing
precursors like NaF, Na2 Se or NaS are now deposited prior to absorber growth to
provide a controlled, more homogeneous, incorporation of Na into the film. The
control of Na incorporation in the film from precursor layers allows the use of other
substrates like metal or polymer foils. There seems to be no fundamental efficiency
limitation due to the substrate as long as a proper supply of sodium is provided.
During absorber deposition, a MoSe2 film forms at the Mo surface (Wada
et al., 1996; Takei et al., 1996). MoSe2 is a layered semiconductor with p-type
conduction, a bandgap of 1.3 eV and weak van der Waals bonding along the c-axis.
If the layer were oriented parallel to the plane of contact, the MoSe2 would inhibit
adhesion of the absorber as well as leading to unfavourable electronic transport.
Fortunately, the c-axis is found to be in parallel with, and the van der Waals planes
thus perpendicular to, the interface (Wada et al., 1996). Because of the larger
bandgap of the MoSe2 compared with that of standard Cu(In,Ga)Se2 films, the
MoSe2 layer provides an electronic mirror for the photogenerated electrons and at
the same time provides a low-resistance contact for the holes (see Section 6.4.1).
Photovoltaic-grade Cu(In,Ga)Se2 films have a slightly In-rich overall composi-
tion. The allowed stoichiometry deviations are astonishingly large, yielding a wide
process window with respect to composition. Devices with efficiencies above 14%
are obtained from absorbers with (In + Ga)/(In + Ga + Cu) ratios between 52% and
64% if the sample contains Na (Ruckh et al., 1994a). Cu-rich Cu(In,Ga)Se2 shows
the segregation of a secondary Cu2−y Se phase preferentially at the surface of the
absorber film. The metallic nature of this phase does not allow the formation of effi-
cient heterojunctions. Even after removal of the secondary phase from the surface
by etching the absorber in KCN, the utility of Cu-rich material for photovoltaic
applications is limited, probably due to the high doping density of 1018 cm−3 in
the bulk and the surface defects. However, the importance of the Cu-rich com-
position is given by its role during film growth. Cu-rich films have grain sizes in
excess of 1 µm whereas In-rich films have much smaller grains. A model for film
growth under Cu-rich compositions comprises the role of Cu2−y Se as a flux agent
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 256

256 U. Rau & H. W. Schock

Re-evaporation Vapour

Condensation
CuSe Liquid
Growth
CuInSe2 Solid

CuSe

CuInSe2

Mo
Substrate

Figure 6.7 Schematic illustration of the growth of a Cu(In,Ga)Se2 film under Cu-rich con-
ditions. A quasi-liquid Cu–Se phase acts as a flux in a vapour-liquid solid growth mechanism.

during the growth process of co-evaporated films (Klenk et al., 1993). This model
for the growth of Cu(In,Ga)Se2 in the presence of a quasi-liquid surface film of
Cu y Se is highlighted in Fig. 6.7. For Cu(In,Ga)Se2 prepared by selenisation, the
role of Cu2−y Se is similar (Probst et al., 1996); therefore growth processes for
high-quality absorber materials have to go through a copper-rich stage and end
with an indium-rich composition.

Co-evaporation processes
The absorber material yielding the highest efficiencies is Cu(In,Ga)Se2 with
a Ga/(Ga + In) content of ∼20%, prepared by co-evaporation from elemental
sources. Figure 6.8 sketches a co-evaporation set-up as used for the preparation of
laboratory-scale solar cells and mini-modules. The process requires a maximum
substrate temperature of ∼550◦ C for a certain time during film growth, preferably
towards the end of growth. One advantage of the evaporation route is that material
deposition and film formation are performed during the same processing step. A
feed-back loop based on a quadrupole mass spectrometer or an atomic absorption
spectrometer controls the rate of each source. The composition of the deposited
material with regard to the metals corresponds to their evaporation rates, whereas
Se is always evaporated in excess. This precise control over the deposition rates
allows for a wide range of variations and optimisations with different sub-steps or
stages for film deposition and growth. These sequences are defined by the evapo-
ration rates of the different sources and the substrate temperature during the course
of deposition. Figure 6.9 illustrates some of the possibilities, starting with a simple
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 257

Cu(In,Ga)Se2 and Related Solar Cells 257

Figure 6.8 Arrangement for the deposition of Cu(In,Ga)Se2 films by co-evaporation on a

heated substrate. The rates of the sources are controlled by mass spectrometry.

(a) single layer (c) three stage

(b) bilayer (d) graded bandgap

Figure 6.9 Schematic rate and substrate temperature profiles for co-evaporation processes.
All processes lead to single-phase films. (a) Single-layer process without Cu-rich growth
step; (b) bilayer process (‘Boeing recipe’) with Cu-rich growth at the start; (c) three-stage
inverted process with intermediate Cu-rich growth; (d) growth of graded-gap films under
Cu-poor conditions.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 258

258 U. Rau & H. W. Schock

single-step process where all rates as well as the substrate temperature are kept
constant during the whole process (Fig. 6.9a).
Advanced preparation sequences always include a Cu-rich stage during the
growth process and end up with an In-rich overall composition in order to combine
the large grains of the Cu-rich stage with the otherwise more favourable electronic
properties of the In-rich composition. The first example of this kind of procedure
is the so-called Boeing or bilayer process (Mickelsen and Chen, 1980), which
starts with the deposition of Cu-rich Cu(In,Ga)Se2 and ends with an excess In
rate, as illustrated in Fig. 6.9b. Another possibility is the inverted process, where
first (In,Ga)2 Se3 (likewise In, Ga, and Se from elemental sources to form that
compound) is deposited at a lower temperatures (typically around 300◦ C). Then
Cu and Se are evaporated at an elevated temperature until an overall composition
close to stoichiometry is reached (Kessler et al., 1992). This process leads to
smoother film morphology than the bilayer process. The most successful version
of the inverted process is the so-called three-stage process (Gabor et al., 1994)
shown in Fig. 6.9c This process puts the deposition of In, Ga, and Se at the end of
an inverted process to ensure the overall In-rich composition of the film even if the
material is Cu-rich during the second stage. The three-stage process currently leads
to the best solar cells. Variations of the Ga/In-ratio during deposition, as shown in
Fig. 6.9d, allow the design of graded-bandgap structures (Gabor et al., 1996).

Selenisation processes
The second class of absorber preparation routes is based on the separation of
deposition and compound formation into two different processing steps. High effi-
ciencies are obtained from absorber prepared by the selenisation of metal precursors
in H2 Se (Binsma and Van der Linden, 1982; Chu et al., 1984; Kapur et al., 1987)
and by rapid thermal processing of stacked elemental layers in a Se atmosphere
(Probst et al., 1996). These sequential processes have the advantage that large-
area deposition techniques such as sputtering can be used for the deposition of
the materials. The Cu(In,Ga)Se2 film formation then requires a second step, the
selenisation.
The very first large-area modules were prepared by the selenisation of metal
precursors in the presence of H2 Se some 25 years ago (Mitchell et al., 1988). A
modification of this process provided the first commercially available Cu(In,Ga)Se2
solar cells, manufactured by Siemens Solar Industries. This process is schemati-
cally drawn in Fig. 6.10. First, a stacked layer of Cu, In and Ga is sputter-deposited
on the Mo-coated glass substrate. Then selenisation takes place under H2 Se. To
improve device performance, a second thermal process under H2 S is added, result-
ing in an absorber that is Cu(In,Ga)(S,Se)2 rather than Cu(In,Ga)Se2 .
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 259

Cu(In,Ga)Se2 and Related Solar Cells 259

In

Cu
DEPOSITION
Mo
glass

heating

H2 Se

SELENISATION

H2S inlet

Figure 6.10 Illustration of the sequential process. Stacked metal layers are selenised and
converted into CuInSe2 in an H2 Se atmosphere.

A variation of this method that avoids the use of the toxic H2Se during selenisa-
tion is the rapid thermal processing of stacked elemental layers (Probst et al., 1996).
Here the precursor includes a layer of evaporated elemental Se. The stack is then
selenised by a rapid thermal process (RTP) in either an inert or a Se atmosphere.
The highest efficiencies are obtained if the RTP is performed in an S-containing
atmosphere (either pure S or H2 S).
On the laboratory scale, the efficiencies of cells made by these preparation
routes are smaller by about 3% (absolute) as compared with the record values.
However, on the module level, co-evaporated and sequentially prepared absorbers
have about the same efficiency. Sequential processes need two or even three stages
for absorber completion. These additional processing steps may counterbalance
the advantage of easier element deposition by sputtering. Also the detailed and
sophisticated control over composition and growth achieved during co-evaporation
is not possible for the selenisation process. Fortunately, the distribution of the
elements within the film grown during the selenisation process turns out to be
close to what one could think to be an optimum, especially if the process includes
the sulphurisation stage. Since the formation of CuInSe2 is much faster than that of
CuGaSe2 , and because film growth starts from the top, Ga is concentrated towards
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 260

260 U. Rau & H. W. Schock

the back surface of the film. An increasing Ga content implies an increase in

bandgap energy. This introduces a back-surface field, improving carrier collection
at the same time as minimising back-surface recombination. In turn, S from the
sulphurisation step is found preferentially towards the front surface of the film,
where it reduces recombination losses and also increases the absorber bandgap in
the space-charge region (SCR) of the heterojunction.

Other absorber deposition processes

Besides selenisation and co-evaporation, other deposition methods have been stud-
ied, either to obtain films with very high quality or to reduce cost of film deposition
on large areas. Methods that are used to form epitaxial III–V compound films, such
as molecular beam epitaxy (MBE) (Niki et al., 1994) or metal organic chemical
vapour deposition (MOCVD) (Gallon et al., 1998) have revealed interesting fea-
tures for fundamental studies, such as phase segregation and defect formation, but
cannot be used to form the base material for high-efficiency solar cells.
Attempts to develop so-called low-cost processes include electrodeposition,
(Abken et al., 1998; Lincot et al., 1998), screen printing and particle deposition
(Eberspacher et al., 1998). Electrodeposition can be carried out in either one or
two steps. The crucial step is final film formation in a high-temperature annealing
process. The recrystallisation process competes with the decomposition of the
material, so process optimisation is quite difficult. Cells with good efficiencies
were obtained by electrodeposition of a Cu-rich CuInSe2 film and subsequent
conditioning by a vacuum evaporation step of In(Se) (Ramanathan et al., 1998a).
Films prepared by spray pyrolysis did not lead to high-performance devices.

Influence of alkali ions sodium and potassium

The outstanding role played by Na in the growth of Cu(In,Ga)Se2 films was recog-
nised some years ago (Hedström et al., 1993; Stolt et al., 1993; Ruckh et al.,
1994a). In most cases, the Na comes from the glass substrate and diffuses into the
absorber. But there are also approaches where Na is incorporated by the use of
Na-containing precursors such as NaSe (Holz et al., 1994; Nakada et al., 1997),
Na2 O2 (Ruckh et al., 1994a), NaF (Contreras et al., 1997a) or Na2 S (Nakada et al.,
1998). Other alkali precursors have been investigated by Contreras et al. (1997a),
who found that Na-containing precursors yield the best cell efficiencies. The most
obvious effects of Na incorporation are better film morphology and higher film
conductivity (Ruckh et al., 1994a). Furthermore, the incorporation of Na induces
beneficial changes in the defect distribution of the absorber films (Keyes et al.,
1997; Rau et al., 1998a).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 261

Cu(In,Ga)Se2 and Related Solar Cells 261

There are many explanations for the beneficial impact of Na, and it is likely
that the incorporation of Na results in a variety of consequences. During film
growth, the incorporation of Na leads to the formation of NaSex compounds. This
slows down the growth of CuInSe2 and could at the same time facilitate the incor-
poration of Se into the film (Braunger et al., 1998b). Also the widening of the
existence range of the α-(CuInSe2 ) phase in the phase diagram, discussed above,
as well as the reported larger tolerance to the Cu/(In + Ga) ratio of Na-containing
thin films, could be explained in this picture. Furthermore, the higher conductivity
of Na-containing films could result from the diminished number of compensat-
ing Vse donors. Wolf et al. (1998) investigated the influence of Na incorporation
on CuInSe2 film formation from stacked elemental layers by means of thin-film
calorimetry. The addition of Na inhibits the growth of CuInSe2 at temperatures
below 380◦C. The retarded phase formation is responsible for the better morphol-
ogy in the case of Na-containing samples.
Another explanation put forward by Kronik et al. (1998) is that Na pro-
motes oxygenation and passivation of grain boundaries. This could account for
the observed enhancement of the net film doping by Na incorporation, through the
diminished positive charge at the grain boundaries. It has in fact been observed
that the surfaces of Na-containing films are more prone to oxygenation than are
Na-free films (Braunger et al., 1998a).
The above explanations deal with the role of Na during growth. However, the
amount of Na in device-quality Cu(In,Ga)Se2 films is of the order 0.1 at.%, which
is a concentration of 1020 cm−3 (Niles et al., 1997), and one may ask the question:
where are these tremendous quantities of Na in the finished absorber? The elec-
tronic effect, i.e. the change of effective doping resulting from Na incorporation,
is achieved at concentrations of ∼1016 cm−3 , four orders of magnitude below the
absolute Na content. It has long been believed that the main part of the Na is situ-
ated at the film surface and the grain boundaries. Final evidence for this hypothesis
was obtained by Niles et al. (1999) with the help of high spatial resolution Auger
electron spectroscopy.
Heske et al. (1996, 1997) investigated the behaviour of Na on the surface of
polycrystalline Cu(In,Ga)Se2 films by X-ray photoelectron spectroscopy (XPS).
They found two different species of Na: (i) The first, denoted ‘reacted’, was
observed on the air-exposed sample or after storing the sputter-cleaned sample
for three days in an ultra-high vacuum (UHV). The second, denoted ‘metallic’,
was found on clean samples either after annealing at 410 K in UHV or after delib-
erate Na deposition from a metallic source. The latter species is considered to be
the active one during crystal growth. In addition, Heske et al. found an increase of
band bending of ∼150 meV induced by the deposition of Na. This finding, as well
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 262

262 U. Rau & H. W. Schock

as the occurrence of two different Na species, is consistent with results obtained

from vacuum-cleaved single crystals (Klein and Jaegermann, 1996).
Another interpretation of the beneficial effect of Na is based on the incorpora-
tion of Na into the Cu(In,Ga)Se2 lattice (Niles et al., 1997). Niles and co-workers
identified Na–Se bonds by means of XPS and concluded that the Na is built into
the lattice, replacing In or Ga. The extrinsic defect NaIn/Ga should then act as an
acceptor and improve the p-type conductivity. The incorporation of Na into the
Cu(In,Ga)Se2 lattice is supported by X-ray diffraction measurements that indicate
an increased volume of the unit cell (Contreras et al., 1997a). Here, the authors
assume that Na in a Cu site prevents the formation of the deep double donor InCu .
Schroeder and Rockett (1997) found that Na driven into epitaxial Cu(In,Ga)Se2
films at a temperature of 550◦ C decreases the degree of compensation by up to a
factor 104 . Schroeder and Rockett attributed their findings to an Na-enhanced reor-
ganisation of the defects, which allows them to build electrically passive clusters.
We see from these numerous approaches that, despite the significance of Na
incorporation, the benefit is far from being explained in terms of simple models.
However, we feel that in view of the amounts of Na (∼0.1 at.%) necessary for
optimum film preparation, arguments based on its effect on film growth are slightly
favoured over those based on the incorporation of Na into the completed film.
A substantial new finding is the role of potassium as key to the improvement
of the quality of CIGS films (Würz et al., 2012). A post-deposition treatment of Na
containing films with K (Čhirila, 2013) led to the first device on a flexible polymer
substrate with efficiency >20%. This treatment appears to passivate the surface of
the grains by replacing Cu by K in the Cu-depleted surface (see below).

Influence of oxygen
Air annealing has been an important process step, crucial for good efficiency,
especially of the early solar cells based on CuInSe2 . Also, though often not men-
tioned explicitly, an oxygenation step is still used for most of the present-day
high-efficiency devices. The beneficial effect of oxygen was explained within the
defect chemical model of Cahen and Noufi (1989). In this model, the surface
defects at grain boundaries are positively charged Se vacancies VSe (Fig. 6.11a)
During air annealing, these sites are passivated by O atoms (Fig. 6.11b). Because
of the decreased charge at the grain boundary, the band bending and the recombina-
tion probability for photogenerated electrons are reduced. The surface donors and
their neutralisation by oxygen are important for the free Cu(In,Ga)Se2 surface as
well as for the formation of the CdS/Cu(In,Ga)Se2 interface (Kronik et al., 2000).
Electrical analysis of oxidised and unoxidised samples revealed the validity of the
Cahen–Noufi model for the earlier CdS/CuInSe2 devices (Sasala and Sites, 1993),
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 263

Cu(In,Ga)Se2 and Related Solar Cells 263

Ec

(a) Eg

Ev

(b)
Eg

Figure 6.11 Band diagram of the conduction- and valence-band energies across a sin-
gle grain of Cu(In,Ga)Se2 . (a) The electronic states at the grain boundaries are positively
charged. This surface charge is compensated by the negative charges in the depleted grain.
This induces the band-bending electronic states at the grain boundaries shown at the left-
hand grain boundary, and the defect chemical equivalent, dangling bonds, shown at the
right-hand boundary; (b) oxygen passivates these dangling bonds and reduces the band
bending.

as well as for the later ZnO/CdS/Cu(In,Ga)Se2 heterostructures (Rau et al., 1999b;

Kronik et al., 1998).
The intriguing interplay between surface oxygenation and the deposition of the
CdS buffer layer is visualised in Fig. 6.12. In the initial state (Fig. 6.12a), the film
surface as well as the grain boundaries are electrically active due to the positively
charged dangling bonds. These charges create a large space-charge region within
the grain. Air annealing passivates the dangling bonds at both interfaces. The bands
become essentially flat, and space charge essentially vanishes (Fig. 6.12b). Even-
tually, the chemical bath removes the passivating oxygen and thus re-establishes
the beneficial type inversion of the film surface (Fig. 6.12c).

6.3.3 The free Cu(In,Ga)Se2 surface

The surface properties of CIGS thin films are especially important, as this surface
becomes the active interface of the completed solar cell. However, the band dia-
gram of the ZnO/CdS/Cu(In,Ga)Se2 heterojunction, especially the detailed struc-
ture close to the CdS/Cu(In,Ga)Se2 interface, is still under debate. Figure 6.13
depicts three different possibilities: (a) the ordered defect compound model of
Schmid et al. (1993); (b) the surface-state model of Rau et al. (1999a); and (c) the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 264

264 U. Rau & H. W. Schock

(a) (b) (c)

Figure 6.12 Illustration of grain boundary and interface charges during the fabrication of a
Cu(In,Ga)Se2 /CdS heterojunction. (a) On as-prepared films, both grain boundaries and the
film surface are positively charged because of the presence of the dangling bonds; (b) air
exposure or air annealing neutralises these charges; (c) the oxide passivation of the surface
is removed by the chemical bath deposition process of CdS. The re-established positive
charges give rise to a type inversion of the CuInSe2 surface.

Ec

EF Ev EF EF
∆Ev
∆Ev

Figure 6.13 Models for the CuInSe2 surface: (a) Segregation of a Cu-poor CuIn3 Se5 ODC
on the surface; (b) band bending due to surface charges; (c) band bending induces Cu-
depletion at the surface, creating a surface defect layer. The valence-band energy is lowered
due to Cu depletion. The defect layer provides an internal barrier Φi to the electron transport
(Niemegeers et al., 1998).

defect layer model of Niemegeers et al. (1998) and Herberholz et al. (1999). The
free surfaces of as-grown Cu(In,Ga)Se2 films exhibit two prominent features:
• The valence band-edge energy E v lies above the surface-Fermi level E F by
about 1.1 eV for CuInSe2 films (Schmid et al., 1993). This energy is larger than
the bandgap energy E ga of the bulk of the absorber material. This was taken
as an indication of a widening of bandgap at the surface of the film. For the
surfaces of Cu(In1−x Gax )Se2 thin films it was found that E F − E v = 0.8 eV
(almost independent of the Ga content if x > 0) (Schmid et al., 1996b).
• The surface composition of Cu-poor CuInSe2 , as well Cu(In,Ga)Se2 films,
corresponds to a surface composition of (Ga + In)/(Ga + In + Cu) of about 0.75
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 265

Cu(In,Ga)Se2 and Related Solar Cells 265

for a range of bulk compositions of 0.5 < (Ga + In)/(Ga + In + Cu) < 0.75.
Both observations have led to the assumption that a phase segregation of
Cu(In,Ga)3Se5 , the so-called Ordered Defect Compound (ODC), occurs at the
surface of the films. The segregation of this β-phase would be compatible
with the phase diagram (see Fig. 6.3) and, as this material displays n-type
conductivity, could yield the explanation for the surface type inversion.

Unfortunately, the existence of a separate phase on top of standard

Cu(In,Ga)Se2 thin films has, to our knowledge, not yet been confirmed by
structural methods such as X-ray diffraction, high resolution transmission elec-
tron microscopy or electron diffraction. Furthermore, if the surface phase exhib-
ited the weak n-type conductivity of bulk Cu(In,Ga)3 Se5 , simple charge neutrality
estimates (Herberholz et al., 1999) show this would not be sufficient to achieve
type inversion. The space-charge width in a CIGS absorber of doping density
3 × 1016 cm−3 is approximately 300 nm. This would require a charge density of
2 × 1018 cm−3 in a 15 nm thick ODC layer to warrant charge neutrality. This
required n-type doping density is considerably more than that usually found in
Cu(In,Ga)3Se5 compounds.
Based on these arguments, another picture of the surface of Cu(In,Ga)Se2 thin
films and of junction formation has emerged. As sketched in Fig. 6.13b, the type
inversion can be viewed as resulting from the presence of shallow surface donors.
This is the classical Bardeen picture (Bardeen, 1947) of Fermi level pinning by
electronic states at semiconductor surfaces. Here, a surface-charge density of a few
times 1012 cm−2 eV−1 is sufficient to pin the Fermi level at the neutrality level of
free semiconductor surfaces. The positively charged surface donors in Fig. 6.13b
are expected to be present in the metal-terminated (112) surface of CuInSe2 because
of the dangling bond to the missing Se (Cahen and Noufi, 1989). The type inversion
vanishes on air exposure because the surface donors are passivated by the reaction
of oxygen with the metal-terminated surface, as discussed in Section 6.3.2 in the
context of the Cahen–Noufi model.
Surface states also play an important role in the completed heterostructure,
where they become interface states at the absorber/buffer interface. As we noted
above, one prominent feature of the defect spectrum of the ZnO/CdS/CIGS het-
erostructure (transition N1 in Fig. 6.5) arises from the charging and discharging of
these states.
The defect layer model shown in Fig. 6.13c represents in some sense a compro-
mise between the ODC model and the surface defect model, as it takes into account
a modification of the band structure due to the Cu deficiency of the surface as well
as the presence of positively charged surface states due to the missing surface Se.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 266

266 U. Rau & H. W. Schock

However, the defect layer model considers the surface layer not as n-type bulk
material (as does the ODC model) but as a p+ -layer (cf. Fig. 6.13c). Furthermore,
the defect layer is viewed not as the origin, but rather as the consequence of the
natural surface type inversion. In contrast to the ODC model, surface states are
responsible for the band bending. In turn, this band bending leads to the liberation
of Cu from its lattice sites and to Cu migration towards the neutral part of the
film (Herberholz et al., 1999). The remaining copper vacancies V− Cu close to the
surface result in a high density of acceptor states, i.e. the p+ -defect layer at the film
surface.

6.3.4 Grain boundaries

As CIGS is a polycrystalline semiconductor the question of which way grain bound-
aries (GBs) influence the device performance and the question of how it is pos-
sible that a material with a grain size <1 µm can have a performance similar to
multicrystalline Si (grain size 1 mm) led to considerable research effort dur-
ing the last decade. Many aspects have been reviewed by Rau et al. (2009). The
electronic activity of GBs in CIGS appears controlled almost equally by crystallog-
raphy and defect chemistry. Most GBs in properly grown films are of low energy,
most are even twin boundaries (Abou-Ras et al., 2007). Even GBs with a less
favourable crystallography leading to a high concentration of structural defects
may possibly not have as many electronically active defects (Yan et al., 2007).
Thus, GBs in CIGS do not seem to be very recombination-active from the start,
which already provides a good base for a reasonable photovoltaic device. On top
of that, the beneficial effect of O and Na appears to result at least partly from
GB passivation, as discussed above. Additionally, it has been suggested that an
important ingredient for the high performance of polycrystalline CIGS thin-film
solar cells stems from an internal valence band offset ∆E v at the GB, much like
that at the CIGS surface (Persson and Zunger, 2003). This fact, sometimes referred
to as a neutral barrier, provides a barrier to majority carrier transport and there-
fore prevents recombination with minority carriers. In fact such neutral barriers
have been found experimentally (Siebentritt et al., 2006; Hafemeister et al., 2010).
Though being certainly an additional beneficial feature, this valence-band offset
cannot be the decisive reason for the low recombination activity of GBs in CIGS
as has been shown by extensive numerical electronic simulations (Gloeckler et al.,
2005; Taretto and Rau, 2008).

6.3.5 Heterojunction formation

The ZnO/CdS/CIGS solar cell structure is usually completed by chemical bath
deposition (CBD) of CdS (typically 50 nm) and by the sputter deposition of a
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 267

Cu(In,Ga)Se2 and Related Solar Cells 267

nominally undoped (intrinsic) i -ZnO layer (usually of thickness 50–70 nm) and
then by heavily doped ZnO. This three-step process appears at the moment to be
mandatory for high-efficiency devices, but a convincing explanation of the need
for such a relatively complicated three-layer structure, especially the role of the
i -ZnO, is not available at the moment.

Buffer layer deposition

Surface passivation and junction formation is most easily achieved by the CBD
deposition of a thin CdS film from a chemical solution containing Cd ions and
thiourea (Kessler et al., 1992). The CdS layer has several advantages:

• CBD deposition of CdS provides complete coverage of the rough polycrys-

talline absorber surface at a film thickness of only 10 nm (Friedlmeier et al.,
1996).
• The layer provides protection against damage and chemical reactions resulting
from the subsequent ZnO deposition process.
• The chemical bath removes the natural oxide from the film surface (Kessler
et al., 1992) and thus re-establishes positively charged surface states and, as a
consequence, the natural type inversion at the CdS/Cu(In,Ga)Se2 interface.
• The Cd ions, reacting first with the absorber surface, remove elemental Se,
possibly by the formation of CdSe.
• The Cd ions also diffuse to a certain extent into the Cu-poor surface layer of the
absorber material (Ramanathan et al., 1998b; Wada et al., 1998), where they
possibly form CdCu donors, thus providing positive charges which support the
type inversion of the buffer/absorber interface.
• As we discuss below, the open-circuit voltage limitations imposed by interface
recombination can be overcome by a low surface recombination velocity in
addition to the type inversion of the absorber surface. Thus one might conclude
that interface states (except those shallow surface donors responsible for the
type inversion) are also passivated by the chemical bath.

Owing to the favourable properties of CdS as a heterojunction partner and the

chemistry of the CBD process, it is difficult to find a replacement, although avoiding
CdS and the chemical bath step would be advantageous from the point of view of
production. First, a toxic material such as CdS requires additional safety regulations
and, second, the chemical bath deposition does not fit with the vacuum deposition
steps of an in-line module fabrication. Research and development in this area relates
to two issues: (i) the search for alternative materials for chemical deposition, and
(ii) the development of ways to deposit the front electrode without an intermediate
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 268

268 U. Rau & H. W. Schock

step in a chemical bath. Promising materials to replace CdS are In(OH,S), Zn(OH,S)
and ZnSe. However, all these materials require additional precautions to be taken for
the preparation of the absorber surface or front electrode deposition. Dry methods
like evaporation or sputtering of (In,S) or ZnS are developed as a prerequisite for
inline fabrication of cells without breaking the vacuum.

Window layer deposition

The most commonly used material for the preparation of the front electrode is
ZnO doped with B or Al. In some cases, doping with Ga or In is claimed to
be advantageous. The first large-area modules produced by ARCO Solar (later
Siemens Solar Industries) had a ZnO:B window layer deposited by chemical vapour
deposition (CVD). Later production facilities at Boeing and EuroCIS use sputtering
processes. Present pilot production lines also favour sputtering. As mentioned
above, an undoped i -ZnO layer with a thickness of about 100 nm is needed at the
heterojunction in order to achieve optimum performance. Stability of the ZnO layer
in humid environments is a major concern. Indium tin oxide (ITO) is superior in
this respect.

6.3.6 Module production and commercialisation

Monolithic interconnects
One inherent advantage of thin-film technology for photovoltaics is the possibility
of using monolithic integration for series connection of individual cells within a
module. In contrast, bulk Si solar cells must be provided with a front metal grid,
and each of these front contacts has to be connected to the back contact of the next
cell to make a series connection. The interconnect scheme, shown in Fig. 6.14, has

ZnO
(1 µm)

CdS
(0.05 µm) CIGS (2 µm)

Mo (0.5 µm)

glass (2–3 mm)

1st patterning

2nd patterning
3rd patterning

Figure 6.14 Interconnect scheme and patterning of a Cu(In,Ga)Se2 (CIGS) based module.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 269

Cu(In,Ga)Se2 and Related Solar Cells 269

to ensure that the front ZnO of one cell is connected to the back Mo contact of the
next one. Three different patterning steps are necessary to obtain this connection.
The first interrupts the Mo back contact by a series of periodical scribes and
thus defines the width of the cells, which is about 0.5–1 cm. A laser is normally
used to pattern the Mo in this way. The second patterning step is performed after
absorber and buffer deposition, and the final one after window deposition. Scribing
of the semiconductor layer is performed by mechanical scribing or laser scribing.
The total width of the interconnect depends not only on the scribing tools, but also
on the reproducibility of the scribing lines along the entire module. The typical
interconnect width is of the order of 300 µm. Thus, about 3–5% of the cell area
must be sacrificed to the interconnects.

Module fabrication
Figure 6.15 shows the typical sequence for the production of a Cu(In,Ga)Se2 mod-
ule. The technologies for absorber, buffer and window deposition used for module
production are the same as those discussed above for the production of small labo-
ratory cells. However, the challenge of modules is to transform the laboratory-scale
technologies to much larger areas. Basically, the scheme in Fig. 6.15 applies to both
of the two concepts currently used for large-area absorber preparation, selenisation

Mo sputtering

laser patterning

absorber
deposition

chemical
deposition of
CdS

patterning

ZnO deposition

patterning

Figure 6.15 Process sequence for the fabrication process of a Cu(In,Ga)Se2 module.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 270

270 U. Rau & H. W. Schock

multiple sources
coated
raw plates
plates

atmosphere vacuum deposition vacuum atmosphere

conditioning

Figure 6.16 In-line evaporation system for the deposition of CuInSe2 thin film. The line
source evaporates the material from top to bottom (Dimmler and Schock, 1996).

and co-evaporation. The selenisation process uses as much off-the-shelf equipment

and processing as possible (e.g. sputtering of the metal precursors) for fabricating
Cu(In,Ga)Se2 films. For co-evaporation on large areas, the Centre for Solar Energy
and Hydrogen Research in Stuttgart (ZSW) has designed its own equipment, shown
in Fig. 6.16, for an in-line process. Line-shaped evaporation sources allow contin-
uous deposition of large-area, high-quality Cu(In,Ga)Se2 films. The relatively high
substrate temperatures that are necessary for high-quality material impose prob-
lems in handling very large-area glass sheets. Future process optimisation therefore
implies reduction of the substrate temperature.
A bottleneck for the production is the deposition of the buffer layer in a chem-
ical bath. First, it is not straightforward to integrate this process in a line consisting
mainly of dry PVD processes. Second, it would be favourable to replace the CdS
layer used at present. Replacement of CdS by a Zn or In compound from solution
solves part of the problem (Kushiya, 1999). Complete, Cd-free dry processes have
been demonstrated on large areas, but a large enough process window has still to
be demonstrated for high-throughput production of large-area modules.
The transparent ZnO front electrode is put down by either CVD or sputtering.
Each method has specific advantages as regards process tolerance, throughput, cost
and film properties. The cell widths within the module — and therefore the relative
losses from the patterning — mainly depend on the sheet resistance of the ZnO.
Module encapsulation is an important issue because module stability depends
on proper protection against humidity. Low-iron cover glasses provide good pro-
tection, and hermetic sealing of the edges is mandatory to obtain stable modules.
Today’s production lines use either co-evaporation of the elements or seleni-
sation of metal layers. Modules with sizes of about 1 m2 reach efficiencies up to
16% with either absorber deposition method. Typical efficiencies in production
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 271

Cu(In,Ga)Se2 and Related Solar Cells 271

are about 13%. For reaching competitive cost the production volume has to
exceed 100 MW/year; Solar Frontier has established a capacity of about 1GW/year
(Kushiya, 2013).

6.4 Device physics

The device physics underlying electronic transport in thin-film solar cells is the
same as for (say) bulk silicon solar cells. However, in view of the fact that thin-
film Cu(In,Ga)Se2 (and also CdTe cells) are heterojunction cells, and because the
absorber layer is only a few times thicker than the space-charge region of the
heterojunction, we shall develop a description which concentrates more on these
specific features.

6.4.1 Band diagram

The equilibrium band diagram of the ZnO/CdS/Cu(In,Ga)Se2/Mo heterostruc-
ture in Fig. 6.17 shows the conduction- and valence-band energies E c,v of the
Cu(In,Ga)Se2 absorber, the CdS buffer layer and the ZnO window. The latter
consists of the intrinsic and the highly Al-doped layer. Here, we completely

i -ZnO CdS Cu(In,Ga)Se 2

Ec

E E
E
E Fn
EF
EV
E

Figure 6.17 Band diagram of the CIGS heterojunction showing the conduction and valence
p
band-edge energies E c and E v . The quantities Φb and Φbn are the barriers for holes and
electrons, ∆E Fn is the energy distance between the Fermi level and the conduction-band
energy at the CdS/CuIn(Ga)Se2 heterointerface, ∆E vab and ∆E cab are the valence-band
and conduction-band discontinuities at the buffer/absorber interface, and w p and wn are the
widths of the space-charge regions in the p-type absorber and the n-type window/buffer
respectively.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 272

272 U. Rau & H. W. Schock

neglect the polycrystalline nature of the semiconductor materials, which in princi-

ple requires a two- or three-dimensional band diagram. We will restrict ourselves
in the following to the implication of the one-dimensional diagram of Fig. 6.17.
Even in the one-dimensional model, some details of the band diagram are still
not perfectly clear. The diagram in Fig. 6.17 concentrates on the heterojunction and
does not show the contact between the Mo and Cu(In,Ga)Se2 at the back side of
the absorber. Another feature under debate but neglected here is a 10–30 nm thick
defect layer on top of the Cu(In,Ga)Se2 absorber, already discussed in Section 6.3.3.
The energetic quantities describing the band diagram in Fig. 6.17 are the
bandgap energies E gx , where x = a, b, w for the absorber, buffer and window,
respectively. The conduction/valence band offsets between the semiconductors are
xy p
denoted ∆E c/v . The built-in or diffusion voltage of the p-type absorber is V D
n
whereas that of the n-type window/buffer is the sum of the contributions V Db from
n and V n from the window layer. Note that the quantities V p/n as
the buffer V Db Dw D
drawn in Fig. 6.17 are zero-bias quantities and change when an external voltage is
p p p
applied. The important barriers Φb and Φbn can be calculated from Φb = q Vb +ς p
and Φbn = ∆E cab + ∆E Fn = ∆E cwb + ∆E b1 n
+ ∆Eb2
n
, where ξ p = E F − E v refers
to the neutral bulk of the absorber and ∆E Fn = E v − E F to the Cu(In,Ga)Se2 /CdS
p
interface. We discuss the impact of Φb on interface recombination and that of Φbn
on the fill factor in Section 6.4.3. Rau et al. (1999a) give a simplified approach to
p
describing the heterojunction and computing Φb and Φbn .
The most important quantities to be considered in the band diagram are the
band discontinuities between the different heterojunction partners. Band disconti-
nuities in terms of valence-band offsets ∆E vab between semiconductors a and b are
usually determined by photoelectron spectroscopy (for a discussion with respect to
Cu-chalcopyrite surfaces and interfaces, see Scheer, 1997). The valence-band offset
between a (011)-oriented Cu(In,Ga)Se2 single crystal and CdS deposited by PVD
at room temperature is determined as ∆E vab = −0.8 (±0.2) eV (Nelson et al., 1993;
Löher et al., 1995), and therefore ∆E cab = E gCdS − E gCIS +∆E vab ≈ 0.55 eV, with the
bandgaps E gCdS ≈ 2.4 eV and E gCIS ≈ 1.05 eV of CdS and CuInSe2 , respectively.
Several authors have investigated the valence-band discontinuity between poly-
crystalline Cu(In,Ga)Se2 films and CdS, and found values between 0.6 and 1.3 eV
with a clear centre of mass around 0.9 eV, corresponding to a conduction-band
offset of 0.45 eV (Scheer, 1997). Wei and Zunger (1993) calculated a theoretical
value of ∆E vab = 1.03 eV, which would lead to ∆E cab ≈ 0.3 eV.
The band alignment of polycrystalline CuInSe2 and Cu(In,Ga)Se2 alloys was
examined by Schmid et al. (1993, 1996a), who found that the valence-band offset is
almost independent of the Ga content. In turn, the increase of the absorber bandgap
leads to a change of ∆E cab from positive to negative values. The conduction-band
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 273

Cu(In,Ga)Se2 and Related Solar Cells 273

offset between the CdS buffer and the ZnO window layer was determined by Ruckh
et al. (1994b) to be 0.4 eV.

6.4.2 Short-circuit current

Optical losses
The short-current density i sc that can be obtained from the standard 100 mW cm−2
solar spectrum (AM1.5) is determined, on the one hand, by optical losses, that
is, by the fact that photons from a part of the spectrum are either not absorbed
in the solar cell or are absorbed without generation of electron-hole pairs. On the
other hand, not all photogenerated electron-hole pairs contribute to i sc because they
recombine before they are collected. We denote these as recombination losses.
Figure 6.18 shows the maximum short-circuit current that can be obtained from
the standard AM 1.5G spectrum (Green, 1995, Appendix C) for photon energies
hν > E g , i.e. the photons that can contribute to the short-circuit current of a
semiconductor of bandgap energy E g . For pure polycrystalline CuInSe2 with E g =
1.04 eV, this value is 46.8 mA cm−2. For E g = 1.11 eV, the bandgap energy of the
best Cu(In,Ga)Se2 solar cell, i sc = 43.6 mA cm−2 . Now we estimate how much
absorber material is needed to achieve this photocurrent. The light absorption in
a semiconductor is described by the Lambert–Beer law. The irradiance I decays
exponentially with depth x into the semiconductor according to

I (x) = Io exp(−αx) (6.1)

50
Cu(In,Ga)Se2 48 α–1 >1.25 µm
α–1 > 0.5µm
40
46
∆isc=
isc/mA cm–2

–2
30 CuInSe2 3 mA cm
44

20 CuInS 2 42
∆isc=0.9 mAcm –2

1.1 1.2
CuGaSe 2
10
ZnO

CdS
0
1.0 1.5 2.0 2.5 3.0 3.5
Bandgap energy Eg/eV
Figure 6.18 Short-circuit current density from the AM 1.5 solar spectrum and correspon-
dence to the bandgap energies of various chalcopyrite compounds, of a typical Cu(In,Ga)Se2
alloy (E g = 1.12 eV), and of the heterojunction partners. The inset shows the losses that
occur if less than 1.25/0.5 µm material is available for light absorption.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 274

274 U. Rau & H. W. Schock

where Io is the incident irradiance and α the absorption coefficient. For direct
semiconductors, α depends on the photon energy hν according to


hν − E g
α(hν) = a (6.2)
hν
The absorption coefficient of Cu(ln,Ga)Se2 with a low Ga content is reason-
ably described by Eq. 6.2 and a ≈ 8 × 104 eV1/2 cm−1 . By reorganising Eq. 6.2
we can calculate the excess energy
 
hv 2
∆hv = hv − E g = α (6.3)
a
of photons that have an absorption coefficient larger than a given α. For
instance, with absorption lengths L α = α −1 = 1.255 µm and 0.5 µm, we have
∆hν = 10 meV and 62.5 meV. The inset in Fig. 6.18 shows that the losses ∆i sc cor-
responding to the photons that are not absorbed within 1.25 or 0.5 µm of CuInSe2
are 0.9 mA cm−2 and 3.0 mA cm−2, respectively. The lower-energy photons are
either absorbed at the back-metal/absorber interface or reflected out of the cell.
Thus a typical absorber of thickness 1.5 µm absorbs all the light from the solar
spectrum except for a negligible remnant corresponding to less than 1 mA cm−2 .
Next, we have to recognise that light absorbed in the ZnO window layer does
not contribute to the photocurrent. This loss affects absorption and photogeneration
for photons of energy >3.2 eV (the bandgap energy of ZnO). As shown in Fig. 6.18,
this loss of high-energy photons costs about 1.3 mA cm−2. In addition, photons in
the energy range hν < 1.4 eV are absorbed by free carriers within the conduction
band of the window material. High-efficiency heterojunction solar cells avoid free-
carrier absorption by optimising the conductivity of the front electrode, not by a
high concentration, but by the high mobility of the free carriers. In any case, for
wide-gap absorbers with E g > 1.4 eV this loss can be neglected.
Another portion of the solar light is absorbed in the buffer layer. If, for instance,
a CdS buffer layer caused a sharp cut-off of the spectral response at the bandgap
of 2.4 eV, only a total of 38 mA cm−2 or 35.5 mA cm−2 would be available for
the short-circuit current of a CuInSe2 or Cu(In,Ga)Se2 (E g = 1.11 eV) solar cell,
respectively. However, measurements of the external quantum efficiency (EQE)
of a typical ZnO/CdS/Cu(In,Ga)Se2 heterostructure reveal that the EQE typically
drops by a factor of only ∼0.8 in the wavelength range between the bandgap of
CdS and that of the ZnO window layer. About 70–80% of the photons in the
wavelength range 440–510 nm contribute to i sc because the thin buffer layer does
not absorb all photons and about 50% of the electron-hole pairs created in the
buffer layer still contribute to the photocurrent (the hole recombination probability
at the buffer/absorber interface is relatively low (Engelhardt et al., 1999)).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 275

Cu(In,Ga)Se2 and Related Solar Cells 275

Collection loss analysis

The most common characterisation method of solar cells other than current–voltage
analysis is the measurement of the quantum efficiency. The EQE at a given wave-
length λ is defined as the number of electron-hole pairs contributing to the pho-
tocurrent divided by the number of photons incident on the cell. A quantitative
evaluation of the EQE can be used to determine the diffusion length L e if the data
are corrected for reflection losses and absorption losses in the window material
and if the absorption data of the absorber material are known for the wavelength
regime where the absorption length is in the order of L e (Arora et al., 1980). This
analysis has been performed in the past by several authors for different types of
devices (Klenk and Schock, 1994; Parisi et al., 1998).
An alternative way to determine the diffusion length in solar cells is provided
by Electron Beam Induced Current (EBIC) measurements. Two approaches are
possible: planar EBIC, where the electron beam is scanned over the device surface,
and junction EBIC, where the device is cleaved and the beam is scanned along the
cross section (Jäger-Waldau et al., 1991). For CIGS, the values for L e extracted
from EQE and EBIC measurements are ∼0.5–1.5 µm.

6.4.3 Open-circuit voltage

Diode characteristics
At open circuit, no current flows across the device and all photogenerated charge
carriers have to recombine within the solar cell. The possible recombination paths
for the photogenerated charge carriers in the Cu(In,Ga)Se2 absorber are indicated
in the band diagram of Fig. 6.19. Here we have considered recombination in the
neutral bulk (A) and at the back surface of the absorber (A ), recombination in
the space-charge region (B), and recombination at the buffer/absorber interface
(C). The dotted lines indicate that the latter two mechanisms may be enhanced by
tunnelling in the presence of a high built-in electrical field.
At the back contact we have drawn the thin MoSe2 layer which forms during
the first minutes of absorber deposition. As drawn here, the MoSe2 has a small
conduction-band offset with respect to the Cu(In,Ga)Se2 bulk material and a small
Schottky barrier at the Mo back contact. Both features are beneficial for device per-
formance, because the conduction-band offset between the Cu(In,Ga)Se2 absorber
and the MoSe2 acts as an electronic mirror (the so-called back-surface field) for
the photogenerated electrons and diminishes back-surface recombination, and the
narrow Schottky barrier provides no substantial resistance for holes between the
absorber and the metallic back contact. We emphasise, however, that the details of
this band diagram are still under debate.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 276

276 U. Rau & H. W. Schock

MoSe2
Cu(In,Ga)Se2

CSd

i–ZnO
Mo

A′
C B A qV

φpb

Figure 6.19 Recombination paths in a CdS–Cu(In,Ga)Se2 junction. The paths A and A

represent bulk and back-contact recombination. B and C result from space-charge and inter-
face recombination. The dotted arrows indicate tunnelling.

The basic equations for the recombination processes (A–C) can be found in
Bube (1992). All recombination current densities i rec for processes A–C can be
written in the form of a diode law
   
qV
i rec = i o exp −1 (6.4)
βkT
where V is the applied voltage, β the diode quality factor, and kT/q the thermal
voltage. The saturation current density i o is in general a thermally activated quantity
and may be written in the form
 
Ẽ a
i o = i oo exp (6.5)
kT

where Ẽ a is the activation energy and the prefactor i oo is only weakly temperature-
dependent. The quantities i o and β depend on the details of each recombination
mechanism. Since mechanisms A–C are connected in parallel, the strongest one
will dominate the recombination loss.
At open circuit, the total recombination current density i rec exactly compen-
sates the short-circuit current density i sc . Hence we can write the open-circuit
voltage as
   
β Ẽ a βkT i oo Ea βkT i oo
Voc = − ln = − ln (6.6)
q q i sc q q i sc
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 277

Cu(In,Ga)Se2 and Related Solar Cells 277

where we have assumed that Voc > 3βkT /q, which allows us to consider only the
exponential term in Eq. (6.4). We have also replaced the activation energy Ẽ a by
E a = β Ẽ a , which will prove in the following to be the ‘true’ activation energy of
the carrier recombination processes.
We now discuss the recombination processes A–C in more detail.

Recombination in the absorber

In the following we shall assume an n + − p junction, i.e. that the doping density on
the n-side is much higher than on the p-side. Shockley’s diode equation for such a
single-sided junction yields the saturation current density for recombination in the
neutral region of the ( p-type) absorber. Knowing the square of the intrinsic carrier
density n 2i = Nc Nv exp(−E g /kT ) we calculate the open-circuit voltage as
 
Eg kT q De Nc Nv
Voc = − ln (6.7)
q q i sc NA L e
where De is the diffusion constant for electrons, and Nc/v the effective density
of states in the conductance/valence band. For our calculations we have used the
values Nc = 6.7 × 1017 cm−3 and Nv = 1.5 × 1019 cm−3 resulting from the
density-of-states effective masses m e = 0.09 m o and m h = 0.71 m o for electrons
and holes, respectively, where m o is the free electron mass (Neumann, 1986). The
quantity NA is the acceptor density, and L e is the diffusion length of the electrons.
If this becomes comparable with the thickness d of the quasi-neutral region (QNR)
of the absorber, the recombination velocity Sb at the back contact has to be taken
into account (recombination path A in Fig. 6.19), and L e in Eq. (6.7) has to be
replaced by
cosh(l −1 ) + sb sinh(l −1 )
L eff = L e (6.8)
sb cosh(l −1 ) + sinh(l −1 )
where sb = Sb L e De and l = L e /d.
Since the width of the SCR in thin-film solar cells is comparable with the
film thickness, recombination in the SCR is important. The Voc -limitation due to
recombination in the SCR of the absorber may be written in a form comparable to
Eq. (6.7), namely
 √ 
Eg 2kT kT De π/2 Nc Nv
Voc = − ln (6.9)
q q i sc E m L 2e
where Em = (2q NA Vbm /o s )1/2 is the electrical field at the position of maxi-
mum recombination. The quantity Em depends on the doping density NA , the band
bending Vbm , and the dielectric constant s of the absorber. The dependence of
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 278

278 U. Rau & H. W. Schock

Eqs (6.7) and (6.9) on the doping density NA is equal in that an increase of NA by
one order of magnitude yields an increase of Voc of ∆Voc = (kT ln 10)/q ≈ 60 mV.
However, improving the open-circuit voltage by increasing the doping density is
limited by the increased Auger recombination in the QNR and the enhancement
of tunnelling in the SCR (see Green, 1996a and below).
Note that in Eq. (6.9) the activation energy E a is given by E a = A Ẽ a = 2 Ẽ a =
2(E g /2) = E g whereas in the diode equation for space-charge recombination, the
saturation current density is i o ∝ exp(E g /2kT ) and the activation energy is only
E g /2. This demonstrates that we have to correct the activation energies obtained
from, for example, Arrhenius plots of the saturation current density for the effect of
non-ideal diode behaviour in order to obtain the activation energy relevant to Voc .
In Fig. 6.20 we display the open-circuit voltage limitations given by Eqs (6.7)
and (6.9) for a Cu(In,Ga)Se2 solar cell with an absorber layer thickness of
1.5 µm, a bandgap energy E g of 1.11 eV and a short-circuit current density i sc
of 35.4 mA cm−2. The top and the bottom axes, showing the electron diffusion
length L e and the lifetime τe , are connected by L e = (De τe )1/2 and a diffusion
constant which is here assumed to be De = 2.59 cm2 s−1 . As the open-circuit
voltages in Eqs (6.7) and (6.9) can be shifted by the bandgap energy, we have used
the right-hand axis of Fig. 6.20 to display the difference (E g /q) − Voc .

Diffusion length Le / µm
0.5 1 5

Sb= 10 2 cm s –1 0.40
0.70
Open-circuit voltage Voc

0.45
E g /q – Voc (V)

0.65 Sb = 105 cm s−1

0.50
0.60

0.55
0.55

0.60
0.1 1 10 100
Lifetime τe /ns

Figure 6.20 Correlation of the open-circuit voltage with lifetimes and diffusion lengths for
a device with a bandgap of 1.12 eV. Solid lines are the results of Eq. (6.9) for L e < 0.7 µm.
For L e > 0.7 µm, Eq. (6.7) holds if effective diffusion lengths that take back-surface recom-
bination into account are introduced. The lines with symbols are the results of a complete
device simulation.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 279

Cu(In,Ga)Se2 and Related Solar Cells 279

For the high-efficiency Cu(In,Ga)Se2 solar cell of Contreras et al. (1999),

this difference is only (1.12–0.68) V = 0.44 V. The open-circuit voltage of this
device requires a lifetime of 30 ns or more, corresponding to a diffusion length
of over 2 µm, thus exceeding the absorber thickness. Hence recombination at the
back contact also has some influence on Voc — a recombination velocity Sb >
105 cm s−1 would hardly allow a Voc of 680 mV.
Since the open-circuit voltage of reasonable Cu(In,Ga)Se2 devices (E g /q −
Voc ≈ 0.5 V) is just at the threshold between SCR and QNR recombination, we have
also conducted some numerical simulations using the software package SCAPS-1D
(Niemegeers and Burgelman, 1996). Results for assumed back-surface recombi-
nation velocities Sb = 102 cm s−1 and 105 cm s−1 are also displayed in Fig. 6.20.
Here we see that recombination can be well described only outside a transition
regime of 1 ns < τe < 30 ns (0.5 µm < L e < 3 µm) by the analytical approaches
for SCR or QNR recombination. Within this parameter range, the recombination
paths A, A and B contribute to recombination. Note that we have suppressed the
interface recombination (path C) by setting the recombination velocity for holes at
the front contact to S p = 102 cm s−1 and assuming a hole barrier Φb = 1 eV.
p

Effective lifetimes for polycrystalline semiconductors

Cu(In,Ga)Se2 solar cells are based on polycrystalline absorbers. Electronic
transport in such devices is not completely covered by one-dimensional models.
However, quasi-one-dimensional approaches are possible as long as the influence
of grain boundaries on the recombination and charge distribution is not too strong.
A first-order approximation is the replacement of the minority-carrier lifetime τe
poly
by an effective lifetime τeff , which includes the interface recombination velocity
Sg at the grain boundaries. This is given by

1 1 2Sg
= + (6.10)
poly
τeff τeb g

where τeb is the minority-carrier lifetime within the grain volume and g denotes the
grain size. With the help of Eq. (6.10), we can still use Eqs (6.8) and (6.10) if we
poly
also use the effective diffusion length L eff for polycrystalline materials, given by
−2
+ 2Sg /(Dn g)]−1/2
poly
L eff = [(L mono
eff ) (6.11)

instead of L eff = L mono

eff . For more details and especially for the limitations of
Eqs (6.10) and (6.11), see Green (1996b), Brendel and Rau (1999) and Jensen
et al. (2000).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 280

280 U. Rau & H. W. Schock

Distribution of recombination centres

An approach to describing the temperature dependence of current–voltage curves
which is useful for Cu(In,Ga)Se2 devices was introduced in Walter et al. (1996b).
This approach does not use recombination centres of a single energy within the
forbidden gap, but rather a distribution of the form DT (E) = DT 0 exp(−E/kT ∗ ),
where the centres are exponentially distributed in energy. The defect density DT (E)
has units of cm−3eV−1 and kT ∗ denotes the characteristic energy of the exponential
distribution. The characteristic energy E ∗ = kT ∗ is also seen in the defect density
spectra obtained from admittance spectroscopy (see Fig. 6.5). Rau et al. (2000) give
a rigorous mathematical treatment for the recombination current density under this
assumption. The recombination current density can be written in the form
   
qV Eg
i rec = i oo exp exp − (6.12)
βkT βkT
where the pre-exponential term i oo is weakly temperature-dependent, and the diode
quality factor is given by
 
1 1 T
= 1+ (6.13)
β 2 T∗
The importance of this approach is on the one hand that a defect distribution
with a characteristic energy kT* of the order of 100–150 mV is often observed in
Cu(In,Ga)Se2 as well as in CuGaSe2 . On the other hand, it has been shown by
Walter et al. (1996b) and Engelhardt et al. (1998) that the temperature dependence
of the current–voltage characteristics of high-efficiency Cu(In,Ga)Se2 solar cells
in the temperature range 200 K < T < 350 K is reasonably well described by
Eq. (6.16). For lower temperatures and for CuGaSe2 absorbers, the contribution of
tunnelling becomes more important (Nadenau et al., 2000).

Interface recombination
Here we consider the simple case of an inverted interface, as shown in Fig. 6.13.
If the recombination centres are not too close to the conduction band, the recom-
bination rate is dominated by the concentration p|i f of free holes at the interface.
From the band diagram and taking the interface recombination velocity for holes
p
as S p and the diffusion potential at the p-side of the heterojunction as VD , the
recombination current density is given by
 p
  p

qV D + ξ Φb − q V
i rec = qS p p|i f = qS p Nv exp − = qS p Nv exp −
kT kT
(6.14)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 281

Cu(In,Ga)Se2 and Related Solar Cells 281

with ξ = E F − E v in the QNR of the Cu(In,Ga)Se2 absorber. In general, a voltage

applied over the heterojunction does not drop only across the p-type part of the
junction. Rather, a change ∆V in the externally applied bias is shared between the
p p
p-type and n-type part according to ∆V = (1−α)∆V D +α∆V Dn where ∆V D /∆V Dn
is the ratio of the diffusion potential in the p-type and n-type components. Note
that the calculation of the voltage share between the two heterojunction partners
in a complicated heterojunction like that shown in Fig. 6.17 is not straightforward.
Here we simply use a linear approach with the coefficient α (0 ≤ α ≤ 1). The
diode law can be written equivalently by the use of a voltage-dependent barrier
p p p
Φb (V ) = Φb0 + αV , where Φb0 is the barrier at zero bias. With this voltage-
dependent barrier, we can rewrite the recombination current density as
 p
  
Φb0 + αq V qV
i rec = qS p Nv exp − exp
kT kT
 p
  
Φb0 q V (1 − α)
= qS p Nv exp − exp (6.15)
kT kT
By comparison of the coefficients we find that the coefficient α is linked to the
diode quality factor by β = (1 − α)−1 . Finally, we write the open-circuit voltage
for interface recombination as
p  
Φb f βkT qS p Nv
Voc = − ln (6.16)
q q i sc
p p p
where Φb f = βΦb0 ≥ Φb0 is the so-called flat-band barrier.
Early CuInSe2 solar cells consisted of a thick CdS window layer laid down by
physical vapour deposition onto the absorber. The doping density in the CdS and
the CuInSe2 absorber were approximately equal, so the band bending in the two
heterojunction partners was also equal. In terms of Eq. (6.16) and the band diagram,
p
the open-circuit voltage was determined by a relatively low value of Φb and a high
value of S p . The use of a highly doped ZnO layer and a CdS buffer layer of only
several tens of nm in thickness in actual high-efficiency ZnO/CdS/Cu(In,Ga)Se2
devices increases the band bending in the absorber and decreases the recombination
velocity for holes at the hetero-interface. For these recent state-of-the-art devices,
it is believed that recombination in the bulk of the absorber is the dominant loss
mechanism.

Tunnelling
In the presence of high electrical fields, interface recombination as well as space-
charge recombination may be enhanced by tunnelling. As shown in Fig. 6.19, the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 282

282 U. Rau & H. W. Schock

charge carriers do not have to overcome the entire energetic barrier, but only a part
of it, in order to recombine. We denote this transport path as tunnelling-enhanced
recombination because it can be described in very similar terms to the classical
recombination mechanism merely by using modified recombination rates. A useful
quantity within the theory of thermally assisted tunnelling is the tunnelling energy
q h̄
E 00 = (NA,o s /m∗)1/2 (6.17)
2
The tunnelling energy E 00 (Padovani and Stratton, 1966) depends only on the
material parameters NA , the permittivity o s and the effective tunnelling mass m ∗ .
First we consider the case of tunnelling-enhanced bulk recombination in the
SCR. Here the modified recombination rate for tunnelling-assisted recombination
can be integrated over the width of the space-charge region. For an exponential
distribution of trap states, a convenient form of the diode law gives the diode quality
factor as (Rau, 1999)
 
2
E 00
1 1 T
= 1+ ∗ − (6.18)
β 2 T 3(kT )2

The recombination current density is again expressed by Eq. (6.12), but with the
quality factor given by Eq. (6.18).
The case of tunnelling-enhanced interface recombination can be treated by
analogy with the thermionic field emission theory of Schottky contacts (Padovani
and Stratton, 1966). The recombination current density is
√    p
   
πq Vb (x)E 00 qV −Φb ξ 1
i rec = qS p Nv exp exp exp −1
kT cosh(E 00 /kT ) βkT βkT kT β
(6.19)

and the quality factor

 
E 00 E 00
β= coth (6.20)
kT kT
is given in Padovani and Stratton (1966). Tunnelling-enhanced interface recombi-
nation is often invoked in interpretations of current–voltage curves of wide-bandgap
materials such as CuInS2 (Hengel et al., 2000).

Recombination loss analysis

The analysis of the temperature dependence of current–voltage curves is one way
to get closer access to the recombination mechanism relevant to the open-circuit
voltage. Usually, current–voltage curves are recorded with different illumination
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 283

Cu(In,Ga)Se2 and Related Solar Cells 283

1.2
Eg = 1.24 eV

Open-circuit voltage Voc /V 1.0

0.8

0.6

0.4

0 100 200 300

Temperature T/K

Figure 6.21 The temperature dependence of the open-circuit voltage under various illumi-
nation intensities extrapolates at 0 K to the bandgap energy.

intensities and at different temperatures. One evaluation method makes use of

the temperature dependence of the open-circuit voltage. Following Eq. (6.7), the
open-circuit voltages obtained at different temperatures T but identical illumi-
nation intensities (corresponding to roughly the same i sc ) display linear temper-
ature dependence. Figure 6.21 shows that the Voc(T ) curves in the temperature
range 200–350 K extrapolate to an activation energy of 1.24 eV, indicating that the
bandgap energy of the Cu(In,Ga)Se2 absorber material is the relevant activation
energy for recombination.
Such an evaluation works well as long as the diode quality factor in Eq. (6.6)
is not too temperature-dependent. An evaluation scheme that is also able to invoke
temperature-dependent quality factors makes use of the fact that the saturation
current density of all recombination processes discussed above can be written in
the form
   
Ẽ a Ea
i o = i oo exp = i oo exp (6.21)
kT βkT

Reorganisation of Eq. (6.21) yields

 
io Ea
β ln = (6.22)
i oo kT
p
where E a = Φb for interface and E a = E g for bulk recombination. Thus a
modified Arrhenius plot of β ln i o vs. 1/T should yield a straight line with the
relevant activation energy E a as the slope.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 284

284 U. Rau & H. W. Schock

Temperature T/K
300 200 100

– 20 CuInSe 2
Ea =1.04 eV
ln (io /mA cm –2 )

–40
Cu(In,Ga)Se2
– 60 Ea =1.21 eV

– 80
CuGaSe2
–100 Ea =1.67 eV

0.00 4 0 .0 06 0 .0 08 0.01
Inverse temperature 1/T (1/K)
(a)

0.7 CuInSe 2
E00 = 3.9 meV
Inverse quality factor 1/

0.6

0.5
Cu(In,Ga)Se 2
E 00 = 6.5 meV
0.4

CuGaSe2
0.3 E00 = 16 meV
10 0 150 2 00 25 0 300
Temperature T/K
(b)
Figure 6.22 (a) Arrhenius plots of the reverse saturation current multiplied by the diode
quality factor. Straight lines indicating an activation energy equal to the respective bandgap
energy are evidence for bulk recombination; (b) the experimental data of the inverse quality
factor vs. temperature can be fitted to the model that includes tunnelling.

Figure 6.22a compares such modified Arrhenius plots of the saturation current
densities i o derived from three different Cu(In,Ga)Se2 based heterojunctions. The
three curves stem from absorbers with different Ga contents x = 0, 0.28 and 1.
The activation energies E a = 1.04 eV, 1.21 eV and 1.67 eV extracted from the plots
correspond to the bandgap energies of the respective absorber materials, indicating
that recombination in all these devices occurs in the volume of the absorber.
The corrected Arrhenius plots yield the activation energy of the recombination
process whether or not the transport processes are enhanced by tunnelling. More
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 285

Cu(In,Ga)Se2 and Related Solar Cells 285

specific information on the transport mechanism is obtained from the temperature

dependence of the quality factor β. We evaluate the quality factors with the help of
Eq. (6.18), as shown in Fig. 6.22b. From the fits we find for the CuInSe2 (x = 0)
cell E 00 = 3.9 meV and kT ∗ = 95 meV, for the Cu(In,Ga)Se2 cell E 00 = 6.5 meV
and kT ∗ = 74 meV, and for CuGaSe2 E 00 = 16 meV and kT ∗ = 281 meV. Thus
an increasing Ga content in Cu(In,Ga)Se2 , in general, increases the contribution of
tunnelling as expressed by the increase of the characteristic tunnelling energy E 00
in Fig. 6.22b. However, for CuInSe2 and Cu(In,Ga)Se2 with moderate Ga content
at room temperature the tunnelling contribution to recombination is insignificant.
In contrast, for CuGaSe2 one extracts significant values of E 00 (Nadenau et al.,
2000).

6.4.4 Fill factor

The fill factor ηfill of a solar cell can be expressed in a simple way as long as
the solar cell is well described by a diode law. Green (1986) gives the following
phenomenological expression for the fill factor
1 − ln(v oc − 0.7)
ηfill
0
= (6.23)
1 + v oc
where v oc is the dimensionless quantity v oc = qV oc /βkT and the fill factor ηfill
0

results from the diode law alone. Thus ηfill depends on temperature as well as on the
0

quality factor of the diode. Effects from series resistance Rs and shunt resistance
Rsh also contribute to the fill factor losses. A good approximation is given by
 x 
x v oc + 0.7 ηfill
ηfill = ηfill 1 − (6.24)
voc rsh
where ηfill
x
= ηfill
0 (1 − r ), r = R i /V and r = R i /V . The description of
s s s sc oc sh sh sc oc
Cu(In,Ga)Se2 solar cells in terms of Eqs (6.23) and (6.24) works reasonably well.
For example, the typical high-efficiency cell of Contreras et al. (1999) has a fill
factor of 79% and the values calculated from Eqs (6.23) and (6.24) are 78.0% (Voc =
678 mV, i sc = 35.2 mA cm−2, Rs = 0.2 Ω cm2 , Rsh = 104 Ω cm2 , β = 1.5).
The dependence of the fill factor on the quality factor (bearing in mind that
this also determines the open-circuit voltage) highlights the importance of this
parameter for the output power of the solar cell. The diode quality factor βL obtained
from the illuminated current–voltage curve is often different from βD of the dark
curve, with βL > βD . One explanation for this important fact could be the finite
barrier Φbn for the electron transport across the CdS/Cu(In,Ga)Se2 heterointerface.
Looking at Fig. 6.17, a band offset ∆E cab ≈ 0.3 eV between the absorber and
the buffer layer seems to represent a substantial barrier hindering the collection
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 286

286 U. Rau & H. W. Schock

of photogenerated electrons. The buffer/absorber interface shown in Fig. 6.6 can

be roughly looked at as a Schottky barrier with a height Φbn = ∆E cab + ∆E Fn
and the photocurrent as the reverse current over this barrier. A simple calculation
shows that the barrier does not provide a large series resistance to the cell as
long as Φbn < 0.5 eV. The conductance of a Schottky contact with an effective
Richardson constant A∗ = 100 A cm−2 K−2 is G = (qA∗ T /k) exp[−Φbn /kT ] =
1.4 S cm−2 for Φbn = 0.5 eV and at a temperature T = 300 K. A series resistance of
0.7 Ω cm2 is acceptable for a device with a reasonable open-circuit voltage. It has
been shown by numerical calculations that a conduction-band offset ∆E cab below
0.4 eV does not affect the device performance (Niemegeers et al., 1995). However,
at lower temperatures or for devices where the type inversion at the Cu(In,Ga)Se2
is less pronounced, i.e. the quantity ∆E Fn is larger than 200 meV, the barrier Φbn
becomes important as shown by experiments (Schmidt et al., 2000) and numerical
simulations (Topic et al., 1997).

6.4.5 Electronic metastabilities

The slow change (over hours and days) of the open-circuit voltage of Cu(In,Ga)Se2-
based solar cells during illumination is a commonly observed phenomenon
(Ruberto and Rothwarf, 1987; Sasala and Sites, 1993). Fortunately, it turns out
that in most cases the open-circuit voltage increases with illumination time, a sit-
uation which is more favourable than that encountered in a-Si:H (Staebler and
Wronski, 1977).
A first model for this slow change in open-circuit voltage was proposed in
Ruberto and Rothwarf (1987). This relies on the reduction of interface recombi-
nation at the CdS/Cu(In,Ga)Se2 interface by additional charges introduced into
the CdS buffer layer either by illumination under open-circuit conditions or by
application of forward bias in the dark. The model is based on the assumption that
interface recombination is the dominant recombination mechanism in the solar
cells. The increase of positive charges in the buffer layer is assumed to increase the
p
barrier Φb and thus reduce interface recombination. However, as we noted above,
the open-circuit voltages of high-efficiency devices are limited by recombination in
the bulk (i.e. in the SCR) rather than at the interface. Since these devices also show
light-soaking effects, another mechanism, possibly additional to that proposed in
Ruberto and Rothwarf (1987), must be at work.
An important observation is that of persistent photoconductivity in
Cu(In,Ga)Se2 thin films (Rau et al., 1998b) and single crystals (Seifert et al., 1997).
Meyer et al. (1999) relate the trapping of electrons as the origin of persistent photo-
conductivity (Fig. 6.23a) to the persistent increase of the charge density in the SCR
of the heterojunction, as shown in Fig. 6.23b. This leads to another model for the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 287

Cu(In,Ga)Se2 and Related Solar Cells 287

wp

Ec

EF
recombination
Ev zone
(a) (b)
Figure 6.23 Illustration of persistent changes of (a) the density of free charge carriers in
the bulk; (b) the charge density in the space-charge region.

open-circuit voltage relaxation in Cu(In,Ga)Se2 solar cells: the gradual decrease of

the electrical field in the SCR leads to a decrease of space-charge recombination,
and finally to the increase of the open-circuit voltage during illumination.
The (Vcu ,VSe ) divacancy complex (Lany and Zunger, 2006) and the InCu anti-
site (Lany and Zunger, 2008) have been identified as sources for the metastable
behaviour of Cu(In,Ga)Se2 . Both defects exist in multiple charge states and, con-
sequently, their influence on the electronic behaviour of CIGS solar cells is rather
complex. Figure 6.24a shows that even in the equilibrium band diagram of the
heterostructure the divacancy complex has three different charge configurations,
namely (Vcu ,VSe )+ , (Vcu ,VSe )− and (Vcu ,VSe )3− (Urbaniak and Igalson, 2009;
Siebentritt et al., 2010). Injection of excess minority-charge carriers via illumi-
nation or electrical bias, i.e. electrons into the neutral bulk, as well as holes into
the inverted CdS/CIGS interface region, will have manifold consequences on the
charge distribution and on the electrical properties of the device (Fig. 6.24b).
The reaction (VCu ,VSe )+ + e− → (VCu , VSe )− + h+ will preferably take
place in the neutral bulk and at the edge of the SCR, leading to an increase of net
doping density and thereby reducing the series resistance, as well as the width of
the SCR, as discussed above.
In contrast, the reaction (VCu ,VSe )3− + 2h+ → (VCu , VSe )− taking place
closer to the interface region will decrease the net doping and widen the SCR.
Furthermore, the capture of holes into the (VCu , VSe )3− state will further diminish
the negative charge density in the region close to the interface. As a consequence,
the collection/injection barrier (Kniese et al., 2008) for electrons at the CdS/CIGS
interface will be decreased. Eventually, injection of minority carriers as sketched
in Fig. 6.24b will result in at least two quite different effects: (i) the reduction of the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 288

288 U. Rau & H. W. Schock

CIGS
+
Ec e
-

VV
CdS
energy
VV
3- EF EFn
VV
-
EF
EFp
+
h

x-coordinate x-coordinate
(a) (b)

Figure 6.24 (a) Sketch of the equilibrium band diagram of the CdS/CIGS heterojunction
involving the three charge states (VCu ,VSe )+ , (VCu ,VSe )− , and (VCu ,VSe )3− of the diva-
cancy; (b) under light or current bias, excess holes diminish the negative charge close to the
CdS/CIGS interface and excess electrons increase the effective doping in the bulk of the
material.

recombination current, and (ii) the reduction of the series resistance. Notably, these
two effects correspond to the commonly observed features during light soaking
of CIGS modules or solar cells: the increase of the open-circuit voltage and the
increase of the fill factor.

6.5 Wide-gap chalcopyrites

The system Cu(In,Ga)(S,Se)2 provides the possibility of fabricating alloys with
a wide range of bandgap energies between the 1.04 eV of pure CuInSe2 up to
2.4 eV for CuGaS2 . The achievement of higher voltages in the individual cells by
increasing the bandgap of the absorber material is important for thin-film modules.
An ideal range for the open-circuit voltage would be between 1.4 and 1.6 eV.
This increased voltage would reduce the number of scribes needed for monolithic
integration of the cells into a module. The thickness of the front and back electrodes
could also be reduced because of the reduced current density. Higher bandgap
materials also lower the temperature coefficient dPmax /dT of the maximum output
power Pmax , i.e. the loss of conversion efficiency with increasing cell temperature.
Higher bandgap materials thus perform better at elevated temperatures. However,
the highest efficiency in the chalcopyrite system is achieved with the relatively
low-bandgap energy E g of 1.12 eV, and attempts to maintain the high efficiency
level at bandgaps of E g > 1.3 eV have so far failed.
Table 6.1 compares the output parameters of the best chalcopyrite-based solar
cells. This compilation clearly shows the superiority of Cu(In,Ga)Se2 with a rela-
tively low Ga content. The fact that the best CuInSe2 devices have an efficiency of
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 289

Cu(In,Ga)Se2 and Related Solar Cells 289

Table 6.1 Operating parameters of the best Cu(In,Ga)Se2 , CuGaSe2 and Cu(In,Ga)S2
cells.

Material Eg /eV η /% V oc /mV isc /mA cm−2 η fill /% A/cm2 Ref.

Cu(In,Ga)Se2 1.12 20.3a 730 35.7 77.7 0.5 1

Cu(In,Ga)Se2 1.18b 20.4 736 35.1 78.9 0.5 4
Cu(In,Ga)Se2 20.8 5
CuGaSe2 1.68 9.53a 905 14.88 70.8 0.471 2
Cu(In,Ga)S2 1.53 12.6a 879 20.4 70.0 0.5 3

a Confirmed total area values; b estimate. References: 1. Jackson et al. (2011); 2. Young et al. (2003);
3. Merdes et al. (2011); 4. Čhirila et al. (2013); 5. ZSW press release (2013).

Cu(In,Ga)Se 2
20
18
Efficiency η / %

16
14 Cu(In,Ga)S 2

12
10
8
1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7
Band gap energy Eg /eV

Figure 6.25 Efficiency η vs. bandgap energy E g for several series of Cu(In1−x Gax )Se2
devices with varying Ga content x (0 ≤ x ≤ 1). The plot is redrawn after Contreras et al.
(2012), and the data correspond to original data from this reference (full circles) and earlier
data from Contreras et al. (2005) (open circles), Eisenbarth et al. (2009) (open triangles),
as well as the record data from Jackson et al. (2011) (open diamonds). Additionally shown
are Cu(In,Ga)S2 results (stars) from Merdes et al. (2011).

3% below that of the best Cu(In,Ga)Se2 device is due not only to the less favourable
bandgap energy but also to the lack of the beneficial effect of small amounts of Ga
on film growth, discussed above.
The difficulty of obtaining wide-gap devices with high efficiencies is illustrated
by plotting the absorber bandgap of a series of chalcopyrite alloys vs. the attained
efficiency. Figure 6.25 shows that efficiencies in the 18–20% range peak sharply
at bandgap energies E g ∼ 1.1−1.2 eV.
One reason for the low performance of wide-gap devices is the less favourable
band offset constellation at the absorber/CdS-buffer interface. Figure 6.26 shows
the band diagram of a CuGaSe2-based heterojunction. As the increase of bandgap in
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 290

290 U. Rau & H. W. Schock

Ec

Ec Ec

EF
Ev

Ev

Figure 6.26 Energy band diagram for the ZnO/CdS/CuGaSe2 heterojunction.

going from CuInSe2 to CuGaSe2 takes place almost exclusively by increase of the
energy of the conduction band, the positive band offset ∆E cab between the absorber
and the buffer in Fig. 6.17 turns into a negative one in Fig. 6.26. This implies that
p
the barrier Φb that hinders the holes from the absorber from recombining with
the electrons from the buffer does not increase proportionally with increase in the
bandgap energy. Thus the importance of interface recombination (dominated by the
p
barrier Φb ) grows considerably relative to that of bulk recombination (dominated
by E g of the absorber; see Herberholz et al., 1997b).
Using the same arguments with respect to the MoSe2 /absorber interface, the
back-surface field produced by this type of back contact turns into its contrary
because the built-in field acts in the opposite direction when the conduction-band
energy of the absorber is increased. At the moment, this drawback of Cu(In,Ga)Se2
with high Ga contents seems to be of minor importance and does not explain the
poorer performance of CuInSe2 devices either. More substantial are the changes
of the electronic quality of the bulk material with increasing x. Figure 6.27 (Hanna
et al., 2000) compares the defect density spectra of different CuIn1−x Gax Se2 alloys
with x = 0, 0.26, and 0.57.
Two features are obvious:

• The bulk defect at activation energy E a = 300 meV displays a minimum defect
density D for the composition x = 0.26. Higher bandgap material (x = 0.56),
as well as lower bandgap material (x = 0), has higher defect densities. It might
not be incidental that the superior electronic quality of the material is found at
that composition which is used to produce the highest-efficiency Cu(In,Ga)Se2
devices.
• The interface-related peak at lower activation energies for the compositions
x = 0 and 0.26 is no longer visible at x = 0.56. This is just what we would
expect from the band diagram shown in Fig. 6.28b: the Fermi level at the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 291

Cu(In,Ga)Se2 and Related Solar Cells

defect density D T / eV–1 cm–1 291

16
16 9x10
10 17 5x10 16 –1 –3
–1
eV cm
–3
1x10 eV cm
–1 –3
eV cm
10 16

10 15
N2
N1 x=0 x = 0.26 x = 0.58
0.2 0.4 0.2 0.4 0.2 0.4
activation energyEa / eV
(a) (b) (c)
Figure 6.27 Admittance defect density spectra of solar cells with different CuIn1−x Gax Se2
alloy compositions. The Ga content is varied from (a) x = 0 to (c) x = 0.58. The lowest
density for the acceptor defect with an activation energy of ∼300 meV is found at (b)
x = 0.26. Source: Hanna et al. (2000).

interface moves towards mid-gap and the type inversion is no longer seen in
admittance spectroscopy.
Thus we see that bulk as well as interface properties change when going from
low Ga content Cu(In,Ga)Se2 towards higher bandgap compositions over a limit
which appears to be at x = 0.3, corresponding to a bandgap energy of 1.3 eV. A
similar limit for S/Se alloying is not yet well established. It appears, however, that
making use of the full alloy system Cu(In,Ga)(S,Se)2 enables us to go beyond the
limit of 1.3 eV while preserving a high efficiency level.

6.5.1 CuGaSe2
CuGaSe2 has a bandgap of 1.68 eV and therefore would represent an ideal partner
for CuInSe2 in an all-chalcopyrite tandem structure. However, a reasonable effi-
ciency for the top cell of any tandem structure is about 15%, far higher than has
been reached by the present polycrystalline CuGaSe2 technology. Despite intense
research during the last two decades the record efficiency is only 9.5% (Young
et al., 2003). In principle, the electronic properties of the material are not so far
from those of CuInSe2 . However, in detail, all the differences quantitatively point
in a less favourable direction. In general, the net doping density NA in CuGaSe2
appears too high (Jasenek et al., 2000). Together with the charge of deeper defects,
the high doping density leads to a strong electrical field in the space-charge region,
which enhances recombination by tunnelling.
Furthermore, a structural difference exists between the bulk CuGaSe2 (chal-
copyrite) phase and the surface (defect chalcopyrite) phase. Regardless of whether
or not this 135-phase is perfectly formed at the film surface or is a defect layer,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 292

292 U. Rau & H. W. Schock

Ga/In
or S/Se
ratio

(a)

(b)
Figure 6.28 Band diagram of an optimised graded-gap device with an increasing Ga/In
ratio towards the back surface (region I) and an increasing S/Se-ratio towards the front
(region III). Region II has no grading. The dotted lines correspond to the conduction- and
valence-band energies of the non-graded device.

a lattice mismatch between the surface layer and the bulk material could account
for the increased defect density, which seems operative for CuGaSe2 as well as for
Cu(In,Ga)Se2 alloys with high Ga contents (Contreras et al., 1997b).

6.5.2 Cu(In,Ga)S2
The major difference between CuInS2 and Cu(In,Ga)Se2 is that the former cannot
be prepared with an overall Cu-poor composition. Cu-poor CuInS2 displays an
extremely low conductivity, making it almost unusable as a photovoltaic absorber
material (Walter et al., 1992). Even at small deviations from stoichiometry on the
In-rich side, segregation of the spinel phase is observed (Walter and Schock, 1993).
Instead, the material of choice is Cu-rich CuInS2 . As in the case of CuInSe2 , a Cu-
rich preparation route implies the removal of the unavoidable secondary phase
(here, Cu2−y S) by etching the absorber in KCN (Scheer et al., 1993). Such an
etch may involve some damage to the absorber surface as well as the introduc-
tion of shunt paths between the front and back electrodes. However, as shown
in Table 6.1, the best Cu(In,Ga)S2 device has an efficiency above 12% (Merdes
et al., 2011). Remarkably, this is achieved by a sulphurisation process rather than
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 293

Cu(In,Ga)Se2 and Related Solar Cells 293

by co-evaporation. The higher quality of the material from sulphurisation might

be due to the higher activity of sulphur, and the consequently lower concentration
of sulphur vacancies. Although results from co-evaporated samples come close to
those of sulphurised cells, CuInS2 is at the moment the only chalcopyrite for which
two-step preparation proves itself to be superior to co-evaporation. The main limi-
tation of efficiency results from an open-circuit voltage that is too low as compared
with the bandgap. By adding Zn (Walter et al., 1995) or Ga (Neisser et al., 1999)
to the absorber layer, voltages above 0.8 V can be achieved. In fact, mastering the
difficult task of incorporating small amounts of Ga (∼5%) in the active part of
the absorber layer was the key for the recent improvement of efficiencies (Merdes
et al., 2011).

6.5.3 Graded-gap devices

An interesting property of the CuIn1−x Gax (S1−y Se y )2 alloy system is the possibil-
ity of designing graded-gap structures that optimise the electronic properties of the
final device (Gray and Lee, 1994; Dhingra and Rothwarf, 1996; Gabor et al., 1996;
Dullweber et al., 2000). Such bandgap gradings are achieved during co-evaporation
by the control of the elemental sources, but selenisation/sulphurisation processes
also lead to beneficial compositional gradings.
The art of designing optimum bandgap gradings is to push back charge carriers
from critical regions, i.e. regions with high recombination probability within the
device. Such critical regions are (i) the interfaces between the back contact and
the absorber layer; (ii) the heterojunction, including the absorber/buffer interface.
Figure 6.28 shows a band diagram of a grading structure that fulfils the requirements
for minimising recombination losses.
To keep the back contact region clear from electrons, one can use a Ga/In
grading. The increase of the Ga/(Ga + In) mole fraction x causes a movement of the
conduction-band minimum upward with respect to its position in pure CuInSe2 . An
increase of x towards the back surface leads to a gradual increase of the conduction-
band energy, as illustrated in Fig. 6.28a and b. The resulting back-surface field, as
in the Cu(In,Ga)Se2 /MoSe2 hetero-contact, drives photogenerated electrons away
from the metallic back contact towards the buffer/absorber junction.
The minimisation of junction recombination, both at the point of equal capture
rates of holes and electrons and at the metallurgical interface between absorber and
buffer, requires a larger bandgap towards the front contact of the absorber. If one
had the choice, one would clearly favour a decrease of the valence-band energy, as
shown in Fig. 6.28, over an increase of the conduction-band energy. This favours a
grading with the help of S/Se alloying, as at least a part of the increasing bandgap
energy is supported by a decrease of valence-bandedge energy. The decreased
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 294

294 U. Rau & H. W. Schock

p
valence-band-edge energy in Fig. 6.28 leads to an increase of the barrier Φb and
thus minimises interface recombination. The favourable combination of In/Ga-
grading towards the back contact and S/Se grading towards the front contact appears
to be an inherent feature of the combined selenisation/sulphurisation processes,
which has led to the highest efficiency devices obtained to date from sequential
processing of stacked metallic or stacked elemental precursors.

6.6 Kesterite (CZTS) solar cells

The ongoing discussions about the availability of indium and gallium in a ter-
awatt future have driven research on indium-free materials with properties similar
to those of the very efficient chalcopyrite compounds. The kesterites have a sim-
ilar crystal structure to the chalcopyrites, with the difference that the Group III
elements In or Ga are replaced by a pair of Group II and Group VI elements,
preferably the abundant elements Zn and Sn. These compounds are the basis of
copper–zinc–tin–sulphide/selenide (CZTS) solar cells. A review (Polizzotti et al.,
2013) and the analysis of actual devices present a discerning picture of the material.
In the quaternary compound Cu2ZnSn(S,Se)4 , disorder caused by secondary phases
makes the presence of anti-site defects very likely and the control of composition is
an additional challenge. The resulting band tails and potential fluctuations severely
affect the open-circuit voltage (Gokmen et al., 2013). In growing thin CZTS films,
conditions that reflect the region of the phase diagram where the occurrence of
defects and secondary phases is minimised should be chosen. In open systems,
considerable loss of Sn is observed (Weber, 2011). Therefore, methods that allow
higher chalcogenide pressures in combination with suitable precursors are prefer-
able for making high-quality films. Nevertheless, the maximum efficiency of 12.6%
(Wang, 2013) for CZTS devices is as yet much lower than that of chalcopyrite CIGS
devices, where an amazing value of 20.8% has been reached.

6.7 Conclusions
The objective of this chapter was not only to describe the achievements of
Cu(In,Ga)Se2 -based solar cells, but also to give an account of our present under-
standing of the physical properties of the materials involved and the electronic
behaviour of the devices. The fortunate situation of Cu(In,Ga)Se2 , which is in a
leading position among polycrystalline thin-film solar cell materials, arises from
the benign, forgiving nature of the bulk materials and interfaces. Nevertheless, we
want to draw the attention of the reader also to the work that has still to be done.
Know-how must be transferred into know-why.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 295

Cu(In,Ga)Se2 and Related Solar Cells 295

Three of the four cornerstones for the recent achievements mentioned in Sec-
tion 6.3 concern the growth of the films: the optimised deposition conditions, and
the incorporation of Na and Ga. However, there is still no detailed model available
definitively to describe the growth of Cu(In,Ga)Se2 , and especially the impact of
Na, which in our opinion is the most important of the different ingredients avail-
able to tune the electronic properties of the absorber. A clearer understanding of
Cu(In,Ga)Se2 growth would allow us to find optimised conditions in the wide
parameter space available, and thus to reduce the number of recombination centres
and compensating donors and optimise the number of shallow acceptors.
The deposition of the buffer layer, or more generally speaking, the forma-
tion of the heterojunction, is another critical issue. The surface chemistry during
heterojunction formation and post-deposition treatments is decisive for the final
device performance. Both processes greatly affect not only the surface defects (i.e.,
recombination and charge), and therefore the charge distribution in the device, but
also the defects in the bulk of the absorber. Concentrated effort and major progress
in these tasks would not only allow us to push the best efficiencies further above
21%, but would also provide a sound knowledge base for the various attempts at
commercialisation of Cu(In,Ga)Se2 solar cells.

References
Abken A., Heinemann F., Kampmann A., Leinkühler G., Rechid J., Sittinger V., Wietler T.
and Reineke-Koch R. (1998), ‘Large area electrodeposition of Cu(In,Ga)Se2 precursors
for the fabrication of thin film solar cells’, Proc. 2nd World Conf. on Photovoltaic Solar
Energy Conversion, Vienna, European Commission, pp. 1133–1136.
Abou-Ras D., Schorr S. and Schock H. W. (2007), ‘Grain-size distributions and grain bound-
aries of chalcopyrite-type thin films,’ J. Appl. Crystallogr. 40, 841–848.
Arora N. D., Chamberlain S. G. and Roulston D. J. (1980), ‘Diffusion length determination
in p-n junction diodes and solar cells’, Appl. Phys. Lett. 37, 325–327.
Bardeen J. (1947), ‘Surface states and rectification at a metal/semiconductor contact’, Phys.
Rev. 71, 717–727.
Binsma J. J. M. and van der Linden H. A. (1982), ‘Preparation of thin CuInS2 films via a
two-stage process’, Thin Solid Films 97, 237–243.
Birkmire R. W., McCandless B. E., Shafarman W. N. and Varrin R. D. (1989), ‘Approaches
for high efficiency CuInSe2 solar cells’, Proc. 9th Eur. Photovoltaic Solar Energy
Conf., Freiburg, Kluwer Academic Publishers, Dordrecht, pp. 134–137.
Braunger D., Hariskos D. and Schock H. W. (1998a), ‘Na-related stability issues in highly
efficient polycrystalline Cu(In,Ga)Se2 solar cells’, Proc. 2nd World Conf. on Photo-
voltaic Solar Energy Conversion, Vienna, European Commission, pp. 511–514.
Braunger D., Zweigart S. and Schock H. W. (1998b), ‘The influence of Na and Ga on
the incorporation of the chalcogen in polycrystalline Cu(In,Ga)(S,Se)2 thin-films for
photovoltaic applications’, Proc. 2nd World Conf. on Photovoltaic Solar Energy Con-
version, Vienna, European Commission, pp. 1113–1116.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 296

296 U. Rau & H. W. Schock

Brendel R. and Rau U. (1999), ‘Injection and collection diffusion lengths of polycrystalline
thin-film solar cells’, Solid State Phenomena 67–68, 81–86.
Bube R. H. (1992), Photoelectronic Properties of Semiconductors, Cambridge University
Press, Cambridge.
Bube R. H. (1998), Photovoltaic Materials, Imperial College Press, London.
Burgelman M., Engelhardt F., Guillemoles J.-F., Herberholz R., Igalson M., Klenk R.,
Lampert M., Meyer T., Nadenau V., Niemegeers A., Parisi J., Rau U., Schock H.-
W., Schmitt M., Seifert O., Walter T. and Zott S. (1997), ‘Defects in Cu(In,Ga)Se2
semiconductors and their role in the device performance of thin-film solar cells’, Progr.
Photovoltaics 5, 121–130.
Cahen D. (1987), ‘Some thoughts about defect chemistry in ternaries’, Proc. 7th Int.
Conf. on Ternary and Multinary Compounds, Mat. Res. Soc., Pittsburgh, PA,
pp. 433–442.
Cahen D. and Noufi R. (1989), ‘Defect chemical explanation for the effect of air anneal on
CdS/CuInSe2 solar cell performance’, Appl. Phys. Lett. 54, 558–560.
Čhirila A., Reinhard P., Pianezzi F., Bloesch P., Uhl A. R., Fella C., Kranz L., Keller
D., Gretener C., Hagendorfer H., Jaeger D., Erni R., Nishiwaki S., Buecheler S. and
Tiwari A. N. (2013), ‘Potassium-induced surface modification of Cu(In,Ga)Se2 thin
films for high-efficiency solar cells’, Nature Mater. 12, 1107–1111.
Chu T. L., Chu S. C., Lin S. C. and Yue J. (1984), ‘Large grain copper indium diselenide
films’, J. Electrochem. Soc. 131, 2182–2184.
Contreras M. A., Egaas B., Dippo P., Webb J., Granata J., Ramanathan K., Asher S., Swart-
zlander A. and Noufi R. (1997a), ‘On the role of Na and modifications to Cu(In,Ga)Se
absorber materials using thin MF (M = Na, K, Cs) precursor layers’, Conf. Record
26th IEEE Photovoltaic Specialists Conf., Anaheim, CA, IEEE Press, Piscataway,
pp. 359–362.
Contreras M. A., Egaas B., Ramanathan K., Hiltner J., Swartzlander A., Hasoon F. and
Noufi R. (1999), ‘Towards 20% efficiency in Cu(In,Ga)Se2 polycrystalline solar cells’,
Progr. Photovoltaics 7, 311–316.
Contreras M. A., Mansfield L. M., Egaas B., Li J., Romero M., Noufi R., Rudiger-Voigt E.
and Mannstadt W. (2012), ‘Wide bandgap Cu(In,Ga)Se2 solar cells with improved
energy conversion efficiency,’ Progr. Photovoltaics 20, 843–850.
Contreras M. A., Ramanathan K., AbuShama J., Hasoon F., Young D. L., Egaas B. and
Noufi R. (2005), ‘Diode characteristics in ZnO/CdS/Cu(In(1−x)Gax )Se2 solar cells,’
Progr. Photovoltaics 13, 209–216.
Contreras M. A., Wiesner H., Tuttle J., Ramanathan K. and Noufi R. (1997b), ‘Issues of
the chalcopyrite/defect-chalcopyrite junction model for high-efficiency Cu(In,Ga)Se2
solar cells’, Solar Energy Mater. Solar Cells 49, 239–247.
Coutts T. J., Kazmerski L. L. and Wagner S. (1986, ed.), Ternary Chalcopyrite Semicon-
ductors: Growth, Electronic Properties, and Applications, Elsevier, Amsterdam.
Devaney W. E., Chen W. S., Steward J. M. and Mickelson R. A. (1990), ‘Structure and
properties of high efficiency ZnO/CdZnS/CuInGaSe2 solar cells’, IEEE Trans. Electron
Devices ED–37, 428–433.
Dhingra A. and Rothwarf A. (1996), ‘Computer simulation and modeling of graded bandgap
CuInSe2 /CdS based solar cells’, IEEE Trans. Electron Devices 43, 613–621.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 297

Cu(In,Ga)Se2 and Related Solar Cells 297

Dimmler B. and Schock H. W. (1996), ‘Scaling-up of CIS technology for thin-film solar
modules’, Progr. Photovoltaics 4, 425–433.
Dullweber T., Hanna G., Shams-Kolahi W., Schwartzlander A., Contreras M. A., Noufi R.
and Schock H. W. (2000), ‘Study of the effect of gallium grading in Cu(In,Ga)Se2 ’,
Thin Solid Films 361–362, 478–481.
Eberspacher C., Pauls K. L. and Fredric C. V. (1998), ‘Improved processes for forming
CuInSe2 films’, Proc. 2nd World Conf. on Photovoltaic Solar Energy Conversion,
Vienna, European Commission, pp. 303–306.
Eisenbarth T., Unold T., Caballero R., Kaufmann C. A., Abou-Ras D. and Schock H. W.
(2009), ‘Origin of defects in CuIn1−x Gax Se2 solar cells with varied Ga content,’ Thin
Solid Films 517, 2244–2247.
Engelhardt F., Bornemann L., Köntges M., Meyer Th., Parisi J., Pschorr-Schoberer E., Hahn
B., Gebhardt W., Riedl W. and Rau U. (1999), ‘Cu(In,Ga)Se2 solar cells with a ZnSe
buffer layer — interface characterization by quantum efficiency measurements’, Progr.
Photovoltaics 7, 423–436.
Engelhardt F., Schmidt M., Meyer Th., Seifert O., Parisi J. and Rau U. (1998), ‘Metastable
electrical transport in Cu(In,Ga)Se2 thin films and ZnO/CdS/ Cu(In,Ga)Se2 het-
erostructures’, Phys. Lett. A 245, 489–493.
Friedlmeier T. M., Braunger D., Hariskos D., Kaiser M., Wanka H. N. and Schock H. W.
(1996), ‘Nucleation and growth of the CdS buffer layer on Cu(In,Ga)Se2 thin films’,
Conf. Record 25th IEEE Photovoltaic Specialists Conf., Washington D.C., IEEE Press,
Piscataway, pp. 845–848.
Gabor A. M., Tuttle J. R., Albin D. S., Contreras M. A., Noufi R. and Hermann A. M. (1994),
‘High-efficiency CuInx Ga1−x Se2 solar cells from (Inx Ga1−x )2 Se3 precursors’, Appl.
Phys. Lett. 65, 198–200.
Gabor A. M., Tuttle J. R., Bode M. H., Franz A., Tennant A. L., Contreras M. A., Noufi R.,
Jensen D. G. and Hermann A. M (1996), ‘Bandgap engineering in Cu(In,Ga)Se2 thin
films grown from (In,Ga)2 Se3 precursors’, Solar Energy Mater. Solar Cells 41, 247–
260.
Gallon P. N. R., Orsal G., Artaud M. C. and Duchemin S. (1998), ‘Studies of CuInSe2 and
CuGaSe2 thin films grown by MOCVD from three organometallic sources’, Proc. 2nd
World Conf. on Photovoltaic Solar Energy Conversion, Vienna, European Commission,
pp. 515–518.
Gilson M., Bacewicz R. and Schock H. W. (1995), “‘Dangling bonds” in CuInSe2 and related
compounds’, Proc. 13th Eur. Photovoltaic Solar Energy Conf., Nice, H. S. Stephens
& Associates, Bedford, pp. 2076–2079.
Gloeckler M., Sites J. R. and Metzger W. K. (2005), ‘Grain-boundary recombination in
Cu(In,Ga)Se2 solar cells,’ J. Appl. Phys., 98, 113704–113704-10.
Gokmen T., Gunawan O., Todorov T. K. and Mitzi D. B. (2013), ‘Band tailing and efficiency
limitation in kesterite solar cells’, Appl. Phys. Lett. 103, 103506.
Gray J. L. and Lee Y. J. (1994), ‘Numerical modeling of graded bandgap CIGS solar cells’,
Proc. 1st World Conf. on Photovoltaic Solar Energy Conversion, Waikoloa, IEEE Press,
Piscataway, NJ, pp. 123–126.
Green M. A. (1986), Solar Cells: Operating Principles, Technology and System Applica-
tions, University of New South Wales, Sydney.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 298

298 U. Rau & H. W. Schock

Green M. A. (1995), Silicon Solar Cells: Advanced Principles and Practice, University of
New South Wales, Sydney.
Green M. A. (1996a), ‘Depletion region recombination in silicon thin-film multilayer solar
cells’, Progr. Photovoltaics 4, 375–380.
Green M. A. (1996b), ‘Bounds upon grain boundary effects in minority carrier semiconduc-
tor devices: a rigorous ‘perturbation’ approach with application to silicon solar cells’,
J. Appl. Phys. 80, 1515–1521.
Haalboom T., Gödecke T., Ernst F., Rühle M., Herberholz R., Schock H.-W., Beilharz C.
and Benz K. W. (1997), ‘Phase relations and microstructure in bulk materials and thin
films of the ternary system Cu–InSe’, Inst. Phys. Conf. Ser. 152E, 249–252.
Hafemeister M., Siebentritt S., Albert J., Lux-Steiner M. C. and Sadewasser S. (2010),
‘Large neutral barrier at grain boundaries in chalcopyrite thin films,’ Phys. Rev. Lett.
104, 196602–196606.
Hahn H., Frank G., Klingler W., Meyer A. and Störger G. (1953), ‘Über einige ternäre
Chalkogenide mit Chalkopyritstruktur’, Z. anorg. u. allg. Chemie 271, 153–170.
Hanna G., Jasenek A., Rau U. and Schock H. W (2000), ‘Open circuit voltage limitations in
CuIn1−x Gax Se2 thin film solar cells — dependence on alloy compositions’, Physica
Status Solidi A. 179, R7–R8.
Hedström J., Ohlsen, H., Bodegard M., Kylner A., Stolt L., Hariskos D., Ruckh M. and
Schock H.-W. (1993), ‘ZnO/CdS/Cu(In,Ga)Se2 thin film solar cells with improved per-
formance’, Conf. Record 23rd IEEE Photovoltaic Specialists Conf., Louisville, IEEE
Press, Piscataway, NJ, pp. 364–371.
Hengel I., Neisser A., Klenk R. and Lux-Steiner C.-M. (2000), ‘Current transport in
CuInS2 :Ga/CdS/ZnO solar cells’, Thin Solid Films 361–362, 458–462.
Herberholz R. (1998), ‘Defect characterisation in chalcopyrite-based heterostructures’, Inst.
Phys. Conf. Ser. 152E, 733–740.
Herberholz R., Braunger D., Schock H. W., Haalboom T. and Ernst F. (1997a), ‘Performance
and defects in Cu(In,Ga)Se2 heterojunctions: combining electrical and structural mea-
surements’, Proc. 14th Eur. Photovoltaic Solar Energy Conf., Barcelona, H. S. Stephens
& Associates, Bedford, pp. 1246–1249.
Herberholz R., Nadenau V., Rühle U., Köble C., Schock H. W. and Dimmler B. (1997b),
‘Prospects of wide-gap chalcopyrites for thin film photovoltaic modules’, Solar Energy
Mater. Solar Cells 49, 227–237.
Herberholz R., Igalson M. and Schock H. W. (1998), ‘Distinction between bulk and inter-
face states in CuInSe2 /CdS/ZnO by space charge spectroscopy’, J. Appl. Phys. 83,
318–325.
Herberholz R., Rau U., Schock H. W., Hallboom T., Gödecke T., Ernst F., Beilharz C., Benz
K. W. and Cahen D. (1999), ‘Phase segregation, Cu migration and junction formation
in Cu(In,Ga)Se2 ’, Eur. Phys. J.: Appl. Phys. 6, 131–139.
Heske C., Fink R., Umbach E., Riedl W. and Karg F. (1996), ‘Na-induced effects on the
electronic structure and composition of Cu(In,Ga)Se2 thin-film surfaces’, Appl. Phys.
Lett. 68, 3431–3432.
Heske C., Richter G., Chen Z., Fink R., Umbach E., Riedl W. and Karg F. (1997), ‘Influence
of Na and H2 O on the surface properties of Cu(In,Ga)Se2 thin films’, J. Appl. Phys.
82, 2411–2420.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 299

Cu(In,Ga)Se2 and Related Solar Cells 299

Holz J., Karg F. and von Phillipsborn H. (1994), ‘The effect of substrate impurities on the
electronic conductivity in CIGS thin films, Proc. 12th Eur. Photovoltaic Solar Energy
Conf., Amsterdam, H. S. Stephens & Associates, Bedford, pp. 1592–1595.
Igalson M. and Schock H. W. (1996), ‘The metastable changes of the trap spectra of CuInSe2 -
based photovoltaic devices’, J. Appl. Phys. 80, 5765–5769.
Jackson P., Hariskos D., Lotter E., Paetel S., Wuerz R., Menner R., Wischmann W. and
Powalla M. (2011), ‘New world record efficiency for Cu(In,Ga)Se2 thin-film solar
cells beyond 20%,’ Progr. Photovoltaics 19, 894–897.
Jäger-Waldau G., Schmid D. and Jäger-Waldau A. (1991), ‘Diffusion length measurement
of heterojunction thin films by junction-EBIC’, J. Phys. IV 1, C6, 131–132.
Jasenek A., Rau U., Nadenau V. and Schock H. W. (2000), ‘Electronic properties of
CuGaSe2 -based heterojunction solar cells. Part II — Defect spectroscopy’, J. Appl.
Phys. 87, 594–602.
Jensen N., Rau U., Hausner R. M., Uppal S., Oberbeck L., Bergmann R. B. and Werner J. H.
(2000), ‘Recombination mechanisms in amorphous silicon/crystalline silicon hetero-
junction solar cells’, J. Appl. Phys. 87, 2639–2645.
Kapur V. K., Basol B. M. and Tseng E. S. (1987), ‘Low-cost methods for the production of
semiconductor films for CuInSe2 /CdS solar cells’, Solar Cells 21, 65–72.
Kazmerski L. L. and Wagner S. (1985), ‘Cu-ternary chalcopyrite solar cells’, in Coutts T. J.
and Meakin J. D., Current Topics in Photovoltaics, Academic Press, Orlando.
Kessler J., Velthaus K. O., Ruckh M., Laichinger R., Schock H. W., Lincot D., Ortega R.
and Vedel J. (1992), ‘Chemical bath deposition of CdS on CuInSe2 , etching effects and
growth kinetics’, Proc. 6th Int. Photovoltaic Solar Energy Conf., New Delhi, Oxford
IBH Publishing, New Delhi, pp. 1005–1010.
Keyes B. M., Hasoon F., Dippo P., Balcioglu A. and Aboulfotuh F. (1997), ‘Influence of Na on
the electro-optical properties of Cu(In,Ga)Se2 ’, Conf. Record 26th IEEE Photovoltaic
Specialists Conf., Anaheim, IEEE Press, Piscataway, NJ, pp. 479–482.
Klein A. and Jaegermann W. (1996), ‘Chemical interaction of Na-cleaved (001) surfaces of
CuInSe2 ’, J. Appl. Phys. 80, 5039–5043.
Klenk R. and Schock H. W. (1994), ‘Photocurrent collection in thin film solar cells —
calculation and characterization for CuGaSe2 /(Zn,Cd)S’, Proc. 12th Eur. Photovoltaic
Solar Energy Conf., Amsterdam, H. S. Stephens & Associates, Bedford, pp.1588–1591.
Klenk R., Walter T., Schock H. W. and Cahen D. (1993), ‘A model for the successful growth
of polycrystalline films of CuInSe2 by multisource physical vacuum evaporation’, Adv.
Mat. 5, 114–119.
Kronik L., Cahen D. and Schock H. W. (1998), ‘Effects of sodium on polycrystalline
Cu(In,Ga)Se2 and its solar cell performance’, Adv. Mater. 10, 31–36.
Kronik L., Rau U., Guillemoles J.-F., Braunger D., Schock H. W. and Cahen D. (2000),
‘Interface redox engineering of Cu(In,Ga)Se2 -based solar cells: oxygen, sodium, and
chemical bath effects’, Thin Solid Films 361–362, 353–359.
Kushiya K. (2013), ‘CIS-based thin-film PV technology in solar frontier K.K.’, Solar Energy
Mater. Solar Cells 122, 309–313.
Kushiya K., Tachiyuli M., Nagoya Y., Fujimaki A., Sang B., Okumara D., Satoh M. and
Yamase O. (1999), ‘Progress in large-area Cu(In,Ga)Se2 -based thin-film modules with
a Zn(O,S,OH) x buffer layer’, Tech. Digest 11th Int. Sci. Eng. Conf., Tokyo University
of Agriculture and Technology, Tokyo, Japan, pp. 637–640.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 300

300 U. Rau & H. W. Schock

Lany S. and Zunger A. (2006), ‘Light- and bias-induced metastabilities in Cu(In,Ga)Se2

based solar cells caused by the (VSe -VCu ) vacancy complex,’ J. Appl. Phys. 100,
113725–113740.
Lany S. and Zunger A. (2008), ‘Intrinsic DX centers in ternary chalcopyrite semiconductors’,
Phys. Rev. Lett. 100, 016401–016405.
Lincot D., Guillemoles J.-F., Cowache P., Marlot A., Lepiller C., Canava B., Yousfi F. B.
and Vedel J. (1998), ‘Solution deposition technologies for thin film solar cells: status
and perspectives’, Proc. 2nd World Conf. on Photovoltaic Solar Energy Conversion,
Vienna, European Commission, pp. 440–445.
Löher T., Jaegermann W. and Pettenkofer C. (1995), ‘Formation and electronic properties
of the CdS/CuInSe2 (001) heterointerface studied by synchrotron-induced photoemis-
sion’, J. Appl. Phys. 77, 731–737.
Mauch R. H., Ruckh. M., Hedström J., Lincot D., Kessler J., Klinger R., Stolt L. Vedel J. and
Schock H. W. (1991), ‘High efficiency Zno/CdS/CuInSe2 solar cells — recent results
of EUROCIS collaboration’, Proc. 9th Eur. Photovoltaic Solar Energy Conf., Lisbon,
Portugal; Kluwer Academic Publishers, Dordrecht, pp. 1415–1419.
Merdes S., Mainz R., Klaer J., Meeder, A., Rodriguez-Alvarez H., Schock H. W., Lux-
Steiner M. C. and Klenk R. (2011), ‘12.6% efficient CdS/Cu(In,Ga)S2 -based solar cell
with an open circuit voltage of 879 mV prepared by a rapid thermal process’, Solar
Energy Mater. Solar Cells 95, 864–869.
Meyer T., Schmidt M., Engelhardt F., Parisi J. and Rau U. (1999), ‘A model for the open-
circuit voltage relaxation in Cu(In,Ga)Se2 heterojunction solar cells’, Eur. Phys. J.:
Appl. Phys. 8, 43–52.
Mickelsen R. A. and Chen W. S. (1980), ‘High photocurrent polycrystalline thin-film
CdS/CuInSe2 solar cell’, Appl. Phys. Lett. 36, 371–373.
Mickelsen R. A. and Chen W. S. (1982), ‘Polycrystalline thin-film CuInSe2 solar cells’,
Conf. Record 16th IEEE Photovoltaic Specialists Conf., San Diego, CA, IEEE Press,
Piscataway, NJ, pp. 781–785.
Migliorato P., Shay J. L., Kasper H. M. and Wagner S. (1975), ‘Analysis of the electrical
and luminescent properties of CuInSe2 ’, J. Appl. Phys. 46, 1777–1782.
Mitchell K. C., Ermer J. and Pier D. (1988), ‘Single and tandem junction CuInSe2 cell
and module technology’, Conf. Record 20th IEEE Photovoltaic Specialists Conf., Las
Vegas, IEEE Press, Piscataway, NJ, pp.1384–1389.
Nadenau V., Rau U., Jasenek A. and Schock H. W. (2000), ‘Electronic properties of
CuGaSe2 -based heterojunction solar cells. Part I — Transport analysis’, J. Appl. Phys.
87, 584–593.
Nakada T., Iga D., Ohbo H. and Kunioka A. (1997), ‘Effects of sodium on Cu(In,Ga)Se2 -
based thin films and solar cells’, Jpn. J. Appl. Phys. 36, 732–737.
Nakada T., Mise T., Kume T. and Kunioka A. (1998), ‘Superstrate type Cu(In,Ga)Se2 thin
film solar cells with ZnO buffer layers — a novel approach to 10% efficiency’, Proc. 2nd
World Conf. on Photovoltaic Solar Energy Conversion, Vienna, European Commission,
pp. 413–417.
Neisser A., Hengel I., Klenk R., Matthes Th. W., Alvarez-Garcia J., Perez-Rodriguez A.,
Romano-Rodriguez A. and Lux-Steiner M. C. (1999), ‘Effect of Ga incorporation in
sequentially prepared CuInS2 thin film absorbers’, Tech. Digest 11th Int. Sci. Eng.
Conf., Tokyo University of Agriculture and Technology, Tokyo, Japan, pp. 955–956.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 301

Cu(In,Ga)Se2 and Related Solar Cells 301

Nelson A. J., Schwerdtfeger C. R., Wei S.-H., Zunger A., Rioux D., Patel R. and Höchst H.
(1993), ‘Theoretical and experimental studies of the ZnSe/CuInSe2 heterojunction
band offset’, Appl. Phys. Lett. 62, 2557–2559.
Neumann H. (1986), ‘Optical properties and electronic band structure of CuInSe2 ’, Solar
Cells 16, 317–333.
Niemegeers A. and Burgelman M. (1996), ‘Numerical modelling of ac-characteristics of
CdTe and CIS solar cells’, Proc. 25nd IEEE Photovoltaic Specialists Conf., Washington
D.C., April 1996, IEEE, New York, NY, pp. 901–904.
Niemegeers A., Burgelman M. and de Vos A. (1995), ‘On the CdS/CuInSe2 conduction
band discontinuity’, Appl. Phys. Lett. 67, 843–845.
Niemegeers A., Burgelman M., Herberholz R., Rau U., Hariskos D. and Schock H. W.
(1998), ‘Model for electronic transport in Cu(In,Ga)Se2 solar cells’, Progr. Photo-
voltaics 6, 407–421.
Niki S., Fons P. J., Yamada A., Suzuki R., Ohdaira T., Ishibashi S. and Oyanagai H. (1994),
‘High quality CuInSe2 epitaxial films — molecular beam epitaxial growth and intrinsic
properties’, Inst. Phys Conf. Ser. 152E, 221–227.
Niles D. W., Al-Jassim M. and Ramanathan K. (1999), ‘Direct observation of Na and
O impurities at grain surfaces of CuInSe2 thin films’, J. Vac. Sci. Technol. A 17,
291–296.
Niles D. W., Ramanathan K., Haason F., Noufi R., Tielsh B. J. and Fulghum J. E. (1997),
‘Na impurity chemistry in photovoltaic CIGS thin films: investigation with X-ray pho-
toelectron spectroscopy’, J. Vac. Sci. Technol. A 15, 3044–3049.
Noufi R., Axton R., Herrington C. and Deb S. K. (1984), ‘Electronic properties versus
composition of thin films of CuInSe2 ’, Appl. Phys. Lett. 45, 668–670.
Padovani F. A. and Stratton R. (1966), ‘Field and thermionic field emission in Schottky
barriers’, Solid State Electronics 9, 695–707.
Parisi J., Hilburger D., Schmitt M. and Rau U. (1998), ‘Quantum efficiency and admittance
spectroscopy on Cu(In,Ga)Se2 solar cells’, Solar Energy Mater. Solar Cells 50, 79–85.
Persson C. and Zunger A. (2003), ‘Anomalous grain boundary physics in polycrystalline
CuInSe2 : the existence of a hole barrier,’ Phys. Rev. Lett., 91, 266401–266405.
Polizzotti A., Repins I. L., Noufi R., Wei S.-H. and Mitzi D. B. (2013), ‘The state and future
prospects of kesterite photovoltaics’, Energy Environ. Sci. 6, 3171–3182.
Potter R. R., Eberspacher C. and Fabick L. B. (1985), ‘Device analysis of CuInSe2 / (Cd,Zn)S
solar cells’, Conf. Record 18th IEEE Photovoltaic Specialists Conf., Las Vegas, NV,
IEEE Press, Piscataway, pp. 1659–1664.
Probst V., Karg F., Rimmasch J., Riedl W., Stetter W., Harms H. and Eibl O. (1996),
‘Advanced stacked elemental layer progress for Cu(InGa)Se2 thin film photovoltaic
devices’, Mat. Res. Soc. Symp. Proc. 426, 165–176.
Ramanathan K., Bhattacharya R. N., Granata J., Webb J., Niles D., Contreras M. A.,
Wiesner H., Haason F. S. and Noufi R. (1998a), ‘Advances in CIGS solar cell research
at NREL’, Conf. Record 26th IEEE Photovoltaic Specialists Conf., Anaheim, CA, IEEE
Press, Piscataway, NJ, pp. 319–325.
Ramanathan K., Wiesner H., Asher S., Niles D., Bhattacharya R. N., Keane J., Contreras
M. A. and Noufi R. (1998b), ‘High efficiency Cu(In,Ga)Se2 thin film solar cells with-
out intermediate buffer layers’, Proc. 2nd World Conf. on Photovoltaic Solar Energy
Conversion, Vienna, European Commission, pp. 477–481.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 302

302 U. Rau & H. W. Schock

Rau U. (1999), ‘Tunneling-enhanced recombination in Cu(In,Ga)Se2 heterojunction solar

cells’, Appl. Phys. Lett. 74, 111–113.
Rau U., Braunger D., Herberholz R., Schock H. W., Guillemoles J.-F., Kronik L. and Cahen
D. (1999a), ‘Oxygenation and air-annealing effects on the electronic properties of
Cu(In,Ga)Se2 films and devices’, J. Appl. Phys. 86, 497–505.
Rau U., Braunger D. and Schock H. W. (1999b), ‘Air-annealing effects on polycrystalline
Cu(In,Ga)Se2 heterojunctions’, Solid State Phenomena 67–68, 409–414.
Rau U., Jasenek A., Schock H. W., Engelhardt F. and Meyer T. (2000), ‘Electronic loss mech-
anisms in chalcopyrite based heterojunction solar cells’, Thin Solid Films 361–362,
298–302.
Rau U., Schmitt M., Engelhardt F., Seifert O., Parisi J., Riedl W., Rimmasch J. and Karg F.
(1998a), ‘Impact of Na and S incorporation on the electronic transport mechanisms of
Cu(In,Ga)Se2 solar cells’, Solid State Commun. 107, 59–63.
Rau U., Schmitt M., Parisi J., Riedl W. and Karg F. (1998b), ‘Persistent photoconductivity in
Cu(In,Ga)Se2 heterojunctions and thin films prepared by sequential deposition’, Appl.
Phys. Lett. 73, 223–226.
Rau U. and Schock H. W. (1999), ‘Electronic properties of Cu(In,Ga)Se2 heterojunction
solar cells — recent achievements, current understanding and future challenges’, Appl.
Phys. A 69, 131–147.
Rau U., Taretto K. and Siebentritt S. (2009), ‘Grain boundaries in Cu(In,Ga)(Se,S)2 thin-film
solar cells’, Appl. Phys. A — Mater. Sci. Process. 96, 221–234.
Repins I., Contreras M. A., Egaas B., DeHart C., Scharf J., Perkins C. L., To B. and Noufi R.
(2008), ‘19.9%-efficient ZnO/CdS/CuInGaSe2 solar cell with 81.2% fill factor,’ Progr.
Photovoltaics 16, 235–239.
Rincon C., Bellabarba C., Gonzalez J. and Sanchez Perez G. (1986), ‘Optical properties and
characterization of CuInSe2 ’, Solar Cells 16, 335–349.
Rincon C. and Wasim S. M. (1987), ‘Defect chemistry of AI BIII CVI 2 chalcopyrite semicon-
ducting compounds’, Proc. 7th Int. Conf. Ternary and Multinary Compounds, Mater.
Res. Soc., Pittsburgh, PA, pp. 443–452.
Rockett A. and Birkmire R. W. (1991), ‘CuInSe2 for photovoltaic applications’, J. Appl.
Phys. 70, 81–97.
Ruberto M. N. and Rothwarf A. (1987), ‘Time-dependent open-circuit voltage in
CuInSe2 /CdS solar cells: theory and experiment’, J. Appl. Phys. 61, 4662–4669.
Ruckh M., Schmid D., Kaiser M., Schäffler R., Walter T. and Schock H. W. (1994a), ‘Influ-
ence of substrates on the electrical properties of Cu(In,Ga)Se2 thin films’, Proc. 1st
World Conf. on Photovoltaic Solar Energy Conversion, Waikoloa, IEEE Press, Piscat-
away, NJ, pp. 156–159.
Ruckh M., Schmid D. and Schock H. W. (1994b), ‘Photoemission studies of the ZnO/CdS
interface’, J. Appl. Phys. 76, 5945–5948.
Sasala R. A. and Sites J. R. (1993), ‘Time-dependent voltage in CuInSe2 and CdTe solar
cells’, Conf. Record 23rd IEEE Photovoltaic Specialists Conf., Louisville, KT, IEEE
Press, Piscataway, NJ, pp. 543–548.
Scheer R. (1997), ‘Surface and interface properties of Cu-chalcopyrite semiconductors and
devices’, Research Trends in Vacuum Sci. Technol. 2, 77–112.
Scheer R. and Schock H. W. (2011), Chalcogenide Photovoltaics, Wiley-VCh Verlag,
Weinheim, Germany.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 303

Cu(In,Ga)Se2 and Related Solar Cells 303

Scheer R., Walter T., Schock H. W., Fearheiley M. L. and Lewerenz H. J. (1993), ‘CuInS2
based thin film solar cells with 10.2% efficiency’, Appl. Phys. Lett. 63, 3294–3296.
Schmid D., Ruckh M., Grunwald F. and Schock H. W. (1993), ‘Chalcopyrite/defect chal-
copyrite heterojunctions on basis of CuInSe2 ’, J. Appl. Phys. 73, 2902–2909.
Schmid D., Ruckh M., Grunwald F. and Schock H. W. (1996a), ‘Photoemission stud-
ies on Cu(In,Ga)Se2 thin films and related binary selenides’, Appl. Surf. Sci. 103,
409–429.
Schmid D., Ruckh M. and Schock H. W. (1996b), ‘A comprehensive characterization of the
interfaces in Mo/CIS/CdS/ZnO solar cell structures’, Solar Energy Mater. Solar Cells
41, 281–294.
Schmidt M., Braunger D., Schäffler R., Schock H. W. and Rau U. (2000), ‘Influence of
damp heat on the electrical properties of Cu(In,Ga)Se2 solar cells’, Thin Solid Films
361–362, 283–287.
Schmitt M., Rau U. and Parisi J. (1995), ‘Investigation of deep trap levels in CuInSe2 solar
cells by temperature dependent admittance measurements’, Proc. 13th Eur. Photo-
voltaic Solar Energy Conf., Nice, H. S. Stephens & Associates, Bedford, pp. 1969–
1972.
Schock H. W. (1996), ‘Thin film photovoltaics’, Appl. Surf. Sci. 92, 606–616.
Schroeder D. J. and Rockett A. A. (1997), ‘Electronic effects of sodium in epitaxial
CuIn1−x Gax Se2 ’, J. Appl. Phys. 82, 5982–5985.
Seifert O., Engelhardt F., Meyer Th., Hirsch M. T., Parisi J., Beilharz C., Schmitt M. and
Rau U. (1997), ‘Observation of a metastability in the DC and AC electrical transport
properties of Cu(In,Ga)Se2 single crystals, thin films and solar cells’, Inst. Phys. Conf.
Ser. 152E, 253–256.
Shay J. L. and Wernick J. H. (1975), Ternary Chalcopyrite Semiconductors: Growth, Elec-
tronic Properties, and Applications, Pergamon Press, Oxford.
Siebentritt S., Igalson M., Persson C. and Lany S. (2010), ‘The electronic structure of
chalcopyrites-bands, point defects and grain boundaries,’ Progr. Photovoltaics 18, 390–
410.
Siebentritt S., Sadewasser S., Wimmer M., Leendertz C., Eisenbarth T. and Lux-Steiner
M. C. (2006), ‘Evidence for a neutral grain-boundary barrier in chalcopyrites’, Phys.
Rev. Lett. 97, 146601–146605.
Staebler D. L. and Wronski C. R. (1977), ‘Reversible conductivity changes in discharge-
produced amorphous Si’, Appl. Phys. Lett. 31, 292–294.
Stolt L., Hedström J., Kessler J., Ruckh M., Velthaus K. O. and Schock H. W. (1993),
‘ZnO/CdS/CuInSe2 thin-film solar cells with improved performance’, Appl. Phys. Lett.
62, 597–599.
Suryawanshi M. P., Agawane G. L., Bhosale S. M., Shin S. W., Patil P. S., Kim J. H. and
Moholkar A. V. (2013), ‘CZTS based thin film solar cells: a status review’, Mater.
Technol. 28, 98–109.
Takei R., Tanino H., Chichibu S. and Nakanishi H. (1996), ‘Depth profiles of spatially-
resolved Raman spectra of a CuInSe2 -based thin-film solar cell’, J. Appl. Phys. 79,
2793–2795.
Taretto K. and Rau U. (2008), ‘Numerical simulation of carrier collection and recombi-
nation at grain boundaries in Cu(In,Ga)Se2 solar cells,’ J. Appl. Phys. 103, 094523–
094523-11.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 304

304 U. Rau & H. W. Schock

Topic M., Smole F. and Furlan J. (1997), ‘Examination of blocking current–voltage behavior
through defect chalcopyrite layer in ZnO/CdS/Cu(In,Ga)Se2 solar cell’, Solar Energy
Mater. Solar Cells, 311–317.
Urbaniak A. and Igalson M. (2009), ‘Creation and relaxation of light- and bias-induced
metastabilities in Cu(In,Ga)Se2 ’, J. Appl. Phys. 106, 063720.
Wada T., Hayashi S., Hashimoto Y., Nishiwaki S., Sato T., Negami T. and Nishitani M.
(1998), ‘High efficiency Cu(In,Ga)Se2 (CIGS) solar cells with improved CIGS sur-
face’, Proc. 2nd World Conf. on Photovoltaic Solar Energy Conversion, Vienna, Euro-
pean Commission, pp. 403–408.
Wada T., Kohara N., Negami T. and Nishitani M. (1996), ‘Chemical and structural char-
acterization of Cu(In,Ga)Se2 /Mo Interface in Cu(In,Ga)Se2 solar cells’, Jpn. J. Appl.
Phys. 35, 1253–1256.
Wagner S., Shay J. L., Migliorato P. and Kasper H. M. (1974), ‘CuInSe2 /CdS heterojunction
photovoltaic detectors’, Appl. Phys. Lett. 25, 434–435.
Wallin E., Malm U., Jarmar T., Lundberg O., Edoff M. and Stolt L. (2012), ‘World-record
Cu(In,Ga)Se2 -based thin-film sub-module with 17.4% efficiency,’ Progr. Photovoltaics
20, 851–854.
Walter T., Braunger D., Hariskos D., Köble C. and Schock H. W. (1995), ‘CuInS2 : film
growth, devices, submodules and perspectives’, Proc. 13th Eur. Photovoltaic Solar
Energy Conf., Nice, H. S. Stephens & Associates, Bedford, pp. 597–600.
Walter T., Content A., Velthaus K. O. and Schock H. W. (1992), ‘Solar cells based on
CuIn(Se,S)2 ’, Solar Energy Mater. Solar Cells 26, 357–368.
Walter T., Herberholz R., Müller C. and Schock H. W. (1996a), ‘Determination of defect
distributions from admittance measurements and application to Cu(In,Ga)Se2 based
heterojunctions’, J. Appl. Phys. 80, 4411–4420.
Walter T., Herberholz R. and Schock H. W. (1996b), ‘Distribution of defects in polycrys-
talline thin films’, Solid State Phenomena 51, 309–316.
Walter T. and Schock H. W. (1993), ‘Structural and electrical investigations of the anion
exchange in polycrystalline CuIn(S,Se)2 thin films’, Jpn. J. Appl. Phys. 32–3, 116–119.
Wang W., Winkler M. T., Gunawan O., Gokmen T., Todorov T. K., Zhu Y. and
Mitzi D. B., ‘Device characteristics of CZTSSe thin-film solar cells with 12.6%
efficiency’, Advanced Energy Mater., published online: 27 Nov. 2013, DOI:
10.1002/aenm.201301465.
Weber A., Mainz R. and Schock H. W. (2010), ‘On the Sn loss from thin films of the material
system Cu–Zn–Sn–S in high vacuum’, J. Appl. Phys. 107, 013516–013516.
Wei S. H. and Zunger A. (1993), ‘Band offsets at the CdS/CuInSe2 heterojunction’, Appl.
Phys. Lett. 63, 2549–2551.
Wolf D., Müller G., Stetter W. and Karg F. (1998), ‘In-situ investigation of Cu–In–Se
reactions: impact of Na on CIS formation’, Proc. 2nd World Conf. on Photovoltaic
Solar Energy Conversion, Vienna, European Commission, pp. 2426–2429.
Würz R., Eicke A., Kessler F., Paetel S., Efimenko S. and Schlegel C. (2012), ‘CIGS thin-
film solar cells and modules on enamelled steel substrates’, Solar Energy Mater. Solar
Cells 100, 132–137.
Yan Y., Jones K. M., Jiang C. S., Wu X. Z., Noufi R. and Al-Jassim M. M. (2007), ‘Under-
standing the defect physics in polycrystalline photovoltaic materials,’ Physica B 401,
25–32.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch06 page 305

Cu(In,Ga)Se2 and Related Solar Cells 305

Young D. L., Keane J., Duda A., AbuShama J. A. M., Perkins C. L., Romero M. and
Noufi R. (2003), ‘Improved performance in ZnO/CdS/CuGaSe2 thin-film solar cells’,
Progr. Photovoltaics 11, 535–541.
Zhang S. B., Wei S. H. and Zunger A. (1997), ‘Stabilization of ternary compounds via
ordered arrays of defect pairs’, Phys. Rev. Lett. 78, 4059–4062.
Zhang S. B., Wei S. H., Zunger A. and Katayama-Yoshiba H. (1998), ‘Defect physics of the
CuInSe2 chalcopyrite semiconductor’, Phys. Rev B. 57, 9642–9656.
Zweibel K. (1995), ‘Thin films: past, present, future’, Progr. Photovoltaics 3, 279–293.
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 307

CHAPTER 7

SUPER-HIGH-EFFICIENCY III–V
TANDEM AND MULTIJUNCTION CELLS
MASAFUMI YAMAGUCHI
Toyota Technological Institute, Nagoya 468–8511, Japan
[email protected]

The reasonable man adapts himself to the world: the unreasonable one persists in trying
to adapt the world to himself. Therefore all progress depends on the unreasonable man.
George Bernard Shaw, Man and Superman, 1903.

7.1 Introduction
Although solar electricity, including solar photovoltaics, is expected to make a
great contribution as a major energy source, providing a share of about 20% of
global electric power in 2050 and about 70% in 2100 (WBGU, 2003), nuclear
power is still a major energy source because of its huge power-generation capacity
and relatively low electricity cost. In order to realise the vision of a solar electricity
future, high-performance solar cells are very attractive.
The development of high-performance solar cells offers a promising path-
way toward achieving high power per unit cost for many applications. Substantial
increases in conversion efficiency can be realised by using multijunction solar cells
rather than single-junction cells. The principles of multijunction cells were sug-
gested as long ago as 1955 (Jackson, 1955), and they have been investigated since
1960 (Wolf, 1960), as shown in the historical timeline of Table 7.1. However, no
significant progress was made in multijunction cell conversion efficiency during
the period 1960–75 because of poor thin-film fabrication technologies. It is thanks
to progress in the technology of liquid-phase epitaxy (LPE) and vapour-phase
epitaxy and to optical devices such as semiconductor lasers that more efficient
AlGaAs/GaAs tandem cells were developed during the 1980s. These included tun-
nel junctions (Hutchby et al., 1985) and metal interconnections (Ludowise et al.,
1982; Flores, 1983; Chung et al., 1989).
At that time, the predicted efficiency of close to 30% (Fan et al., 1982) was
not achieved because of difficulties in making high-performance, stable tunnel

307
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 308

308 M. Yamaguchi

Table 7.1 Progress of III–V tandem solar cell technologies.

Location
Date Milestone or individual

1955 Principle of multijunction solar cells Jackson

1960 Proposal of multijunction solar cells Wolf
1982 Efficiency calculation for tandem cells MIT
1982 15.1% AlGaAs/GaAs two-junction (2-J) cell (1 Sun) RTI
1987 Proposal of double-hetero structure tunnel junction NTT
for multijunction interconnection
1987 20.2% AlGaAs/GaAs 2-J cell (1 Sun) NTT
1988 29.6% GaAs/Si 2-J cell (mechanically stacked, Sandia
350 Suns concentration)
1989 27.6% GaAlAs/GaAs 2-J cell (1 Sun) Varian
1989 32.6% GaAs/GaSb concentrator 2-J cell Boeing
(mechanically stacked, 100 Suns concentration)
1990 Proposal of InGaP as top cell material NREL
1993 29.5% InGaP/GaAs 2-J cell (1 Sun) NREL
1994 30.2% InGaP/GaAs 2-J cell (180 Suns) NREL
1996 30.3% InGaP/GaAs 2-J cell (1 Sun) Japan Energy
1997 Discovery of radiation resistance of InGaP top cell Toyota T.I.
1997 33.3% InGaP/GaAs//InGaAs 3-J cell (1 Sun) Japan Energy,
(mechanically stacked) Sumitomo, Toyota T.I.
1997 Commercial satellite with 2-J cells Hughes
2000 31.7% InGaP/InGaAs/Ge 3-J cell (1 Sun) Japan Energy
2000 32.4% InGaP/GaAs/Ge 3-J cell (42 Suns) Spectrolab
2001 33.5% InGaP/GaAs/InGaAs 3-J cell (308 Suns) Fraunhofer ISE
(mechanically stacked)
2003 32.0% InGaP/GaAs/Ge 3-J cell (1 Sun) Spectrolab
2004 38.9% InGaP/InGaAs/Ge 3-J cell (489 Suns) Sharp, Toyota T.I.
2005 38.9% InGaP/InGaAs/Ge 3-J cell (236 Suns) Spectrolab
2006 31.5% large-area (5445 cm2 ) InGaP/InGaAs/Ge 3-J Daido Steel, Daido
cell module (outdoor) Metal, Sharp,
Toyota T.I.
2006 40.7% InGaP/GaAs/Ge 3-J cell (236 Suns) Spectrolab
2008 33.8% InGaP/GaAs/InGaAs 3-J cell (1 Sun) NREL
2009 41.1% InGaP/InGaAs/Ge 3-J cell (454 Suns) Fraunhofer ISE
2009 41.6% InGaP/InGaAs/Ge 3-J cell (364 Suns) Spectrolab
2009 35.8% InGaP/GaAs/InGaAs 3-J cell (1 Sun) Sharp
2010 42.3% InGaP/GaAs/InGaAs 3-J cell (406 Suns) Spire
2011 36.9% InGaP/GaAs/InGaAs 3-J cell (1 Sun) Sharp
2012 43.5% InGaP/GaAs/GaInNAs 3-J cell (418 Suns) Solar Junction
2013 37.9% InGaP/GaAs/InGaAs 3-J cell (1 Sun) Sharp
2013 44.4% InGaP/GaAs/InGaAs 3-J cell (240–300 Suns) Sharp
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 309

Super-High-Efficiency III–V Tandem and Multijunction Cells 309

junctions (Yamaguchi et al., 1987), and also because of the presence of oxygen-
related defects in the AlGaAs (Ando et al., 1987). High-performance, stable tunnel
junctions with a double-hetero (DH) structure (in which the GaAs tunnel junction
is sandwiched between AlGaAs layers) were developed by Sugiura et al. (1988)
of NTT Electrical Communications Laboratories (NTT). The use of InGaP for
the top cell was introduced by Olson et al. (1990) of National Renewable Energy
Laboratory (NREL), and as a result, a GaInP/GaAs tandem cell with an efficiency
of 29.5%, though a small area of only 0.25 cm2 , was made by Bertness et al.
(1994). Monolithically grown InGaP/GaAs two-junction solar cells achieved a
then-record efficiency of 30.3% in 1997 (Takamoto et al., 1997a) at 1 Sun AM
1.5. As regards concentrator systems, over 30% efficiency was attained by the
mechanically stacked GaAs/GaSb cells of Fraas et al. (1990).
InGaP/GaAs-based multijunction (MJ) solar cells have drawn increased atten-
tion for space applications because of the superior radiation resistance of InGaP
top cells and materials, which was discovered by the author and co-workers (Yam-
aguchi et al., 1997). In addition we also achieved the possibility of high conversion
efficiency of over 30%. The commercial satellite HS 601HP with two-junction
GaInP/GaAs on Ge solar arrays was launched in 1997 (Brown et al., 1997).
InGaP/GaAs-based MJ solar cells have also drawn increased attention for
terrestrial applications because operating MJ cells under concentrated sun-
light has great potential for providing high-performance, low-cost solar mod-
ules. For concentrator applications, the cell contact grid structure should be
designed so as to reduce the energy loss due to series resistance; in this way
38.9% (AM 1.5G, 489 Suns) efficiency was demonstrated by Sharp (Takamoto
et al., 2005). The achievement of 41.1% efficiency under 454 Suns with
In0.65 Ga0.35 P/In0.17 GaAs0.83 /Ge three-junction concentrator cells by Fraunhofer
ISE (Bett et al., 2009) and 41.6% efficiency under 364 Suns with lattice-matched
InGaP/InGaAs/Ge three-junction concentrator cells by Spectrolab (King et al.,
2009) has also been reported. Later, 42.3% efficiency under 406 Suns with bifa-
cial epitaxially grown InGaP/GaAs/InGaAs three-junction concentrator cells was
reported by Spire (Wojtczuk et al., 2010); and Solar Junction (Solar Junction, 2011;
Sabnis, 2012) achieved 43.5% efficiency under 418 Suns with lattice-matched
InGaP/GaAs/GaInNAs three-junction concentrator cells. Most recently, 37.9%
efficiency under 1 Sun and 44.4% efficiency under 240–300 Suns have been demon-
strated with InGaP/GaAs/InGaAs three-junction cells by Sharp (Sharp, 2013;
Sasaki, 2013; Yamaguchi and Luque, 2013; Green et al., 2013). In addition, high-
efficiency large-area (5445 cm2) concentrator InGaP/InGaAs/Ge three-junction
modules for 500 Suns used with an outdoor efficiency of 31.5% (Araki et al., 2005)
have been developed as a result of incorporating high-efficiency InGaP/InGaAs/Ge
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 310

310 M. Yamaguchi

three-junction cells, low-optical-loss Fresnel lenses and homogenisers, and design-

ing low-thermal-conductivity modules.
In this chapter, we describe the principles of MJ solar cells, candidate mate-
rials, epitaxial technologies, MJ cell configurations and cell interconnection tech-
nologies. We also review recent progress with laboratory cells and make some
predictions about the future of MJ cells.

7.2 Principles of super-high-efficiency multijunction solar cells

7.2.1 Conversion efficiency analysis and candidate materials
for high-efficiency multijunction solar cells
While single-junction cells may be capable of attaining AM 1.5 efficiencies of
up to 27%, MJ structures were recognised early on as being capable of realising
efficiencies of up to 35% (Hutchby et al., 1985). Figure 7.1 shows the principle of
wide photoresponse using MJ solar cells for the case of a triple-junction cell. Solar
cells with different bandgaps are stacked one on top of the other so that the cell
facing the Sun has the largest bandgap (in this example, this is the InGaP top cell).
This top cell absorbs all the photons at and above its bandgap energy and transmits

Figure 7.1 Principle of wide photoresponse by using a multijunction solar cell, for the case
of an InGaP/InGaAs/Ge triple-junction solar cell.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 311

Super-High-Efficiency III–V Tandem and Multijunction Cells 311

Figure 7.2 Current–voltage curve and spectral response for an InGaP/InGaAs/Ge mono-
lithic, two-terminal three-junction cell (Takamoto et al., 2000).

the less energetic photons to the cells below. The next cell in the stack (here the
GaAs middle cell) absorbs all the transmitted photons with energies equal to or
greater than its bandgap energy, and transmits the rest downward in the stack (in this
example, to the Ge bottom cell). As shown in Fig. 7.2, the current–voltage curve and
spectral response for an InGaP/GaAs/Ge monolithic, two-terminal triple-junction
cell (Takamoto et al., 2000) shows the wideband photoresponse of MJ cells. In
principle, any number of cells can be used in tandem.
Computer analysis of the performance of MJ solar cells has been carried out by
several researchers (Loferski, 1976; Lamorte and Abbott, 1980; Mitchell, 1981; Fan
et al., 1982; Nell and Barnett, 1987; Amano et al., 1989; Letay and Bett, 2001).
The following explanations are based on the findings of Letay and Bett (2001).
Figure 7.3 shows their AM 1.5 iso-efficiency plots for a three-cell, two-terminal
tandem structure with Ge as the bottom cell at 25◦ C and 1 Sun. The maximum
theoretical efficiency for this system is 38% at AM 1.5. For optimal efficiencies in
the two-terminal structure, the allowable range of bandgaps for the top and bottom
cells is very narrow. The top cell must have a bandgap of about 1.8 eV, and the
middle cell about 1.1 eV. In this case, one of the candidate material combinations
is InGaP/InGaAs/Ge. However, because the optimal bandgap combination, which
is 1.8 eV/1.1 eV/0.66 eV, is a lattice-mismatched system, 1.85 eV/1.4 eV/0.66 eV
and 2 eV/1.4 eV/0.66 eV lattice-matched systems have been realistically devel-
oped. The optimal efficiencies of such systems are 31–34.5%. In these cases,
InGaP/GaAs/Ge, AlInGaP/GaAs/Ge and AlGaAs/GaAs/Ge are candidate material
combinations.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 312

312 M. Yamaguchi

Figure 7.3 Calculated AM 1.5 iso-efficiency plots for a three-cell, two-terminal tandem
structure with Ge as the bottom cell at 25◦ C and 1 Sun. From Letay and Bett (2001).

2.00
Three cells
Bottom-cell band gap = 1.00 eV
Series-connected
AM1
Middle-cell band gap (eV)

27 C
1.75

1.50
41.1%
39.0%
37.0%
35.0%

1.25
1.75 2.00 2.25 2.50
Top-cell band gap (eV)

Figure 7.4 Calculated AM 1 iso-efficiency plots at 27◦ C and 1 Sun for the three-cell tandem
structure with the cells connected in series. The bottom cell has a fixed energy gap of 1.0 eV
(Fan et al., 1982).

Figure 7.4 shows the AM 1 iso-efficiency plots at 27◦C for the three-cell
tandem structure with the cells connected in series (Fan et al., 1982). The curves are
plotted for a bottom-cell bandgap of 1.0 eV, because the maximum calculated effi-
ciencies are obtained for a range of 0.95–1.0 eV. The optimal top/middle/bottom
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 313

Super-High-Efficiency III–V Tandem and Multijunction Cells 313

60

50
conversion efficiency / %

40

theory (Conc.)
30 theory (1 Sun)
realised (Conc.)
realised (1 Sun)

20
1 2 3 4 5 6

number of junctions
Figure 7.5 Theoretical conversion efficiencies of single-junction and multijunction solar
cells in comparison with experimentally realised efficiencies for 1 Sun intensity and under
concentration.

cell bandgap combination is 1.95/1.45/1.0 eV for the series-connected structure,

and the maximum theoretical efficiency at AM 1 is 41.1%. In this case, (Al)In
GaP/GaAs/InGaAs, (Al)InGaP/GaAs/InGaAsP, (Al)InGaP/GaAs/InGaAsN, (Al)
InGaP/GaAs/Si and AlGaAs/GaAs/Si are candidate material combinations.
Figure 7.5 compares theoretical and realised energy efficiences for single-
junction and MJ solar cells, under unconcentrated and concentrated sunlight. Mul-
tijunction solar cells have the potential for achieving ultra-high conversion effi-
ciencies of over 50%, and are promising for both space and terrestrial applications
due to their wide photoresponse.

7.2.2 Cost analysis

The allowable cost per unit area of solar cell modules depends strongly on module
efficiency (Bowler and Wolf, 1980; Yamaguchi et al., 1994). For example, a 30%-
efficient cell costing 3.5 times as much as a 15%-efficient cell of the same area will
yield equivalent overall photovoltaic (PV) system costs (Bowler and Wolf, 1980).
Therefore high-efficiency cells will have a substantial economic advantage over
low-efficiency cells as long as the cost of fabricating them is low enough. For space
applications, high-efficiency cells also have significant payload advantages.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 314

314 M. Yamaguchi

40
x 0.1 Factory
Equipment
30 Labour
Cost(US$/Wp)

Others
Dopant
20
V source
V' source
x 0 .1
10 Substrate

0
Present Mass MOCVD MBE CBE CBE +
production (future) (future) (future) concentrator

Figure 7.6 Cost estimates for monolithically integrated two-junction solar cells consisting
of a III–V compound solar cell combined with a Si cell, fabricated by metal-organic chemical
vapour deposition (MOCVD), molecular beam epitaxy (MBE) or chemical beam epitaxy
(CBE) (Yamaguchi et al., 1994).

The use of concentrating systems can further enhance the cost advantage of
high-efficiency cells. Figure 7.6 shows cost estimates for monolithically integrated
two-junction solar cells with a III–V compound solar cell and a Si cell, fabricated
by metal–organic chemical vapour deposition (MOCVD), molecular beam epitaxy
(MBE) or chemical beam epitaxy (CBE) (Yamaguchi et al., 1994). Through source
material cost reduction combined with the use of the tandem structure and a con-
centrator system, high-efficiency cells costing less than $1/Wp should be possible.
Concentrator operation is very effective for cost reduction of solar cell modules
and thus also of PV systems. Figure 7.7 shows a summary of the estimated cost for
the concentrator PV systems vs. concentration ratio (Yamaguchi, 2003). As shown
in Fig. 7.7, concentrator PV systems using MJ solar cells have great potential for
cost reduction to under 50 ¢/Wp if one could fabricate 35%-efficient modules for
5000 Suns use.
The effectiveness of high-performance, low-cost concentrator PV systems is
also discussed in Chapter 11 of this book by Luque-Heredia and Luque, and sep-
arately by Yamaguchi and Luque (1999).

7.3 Epitaxial technologies for growing III–V compound cells

Figure 7.8 shows the chronological improvements in the efficiencies of GaAs solar
cells fabricated by LPE, MOCVD and MBE. Liquid-phase epitaxy was used to
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 315

Super-High-Efficiency III–V Tandem and Multijunction Cells 315

Figure 7.7 Summary of estimated cost for the concentrator PV systems vs. concentration
ratio (Yamaguchi, 2003).

30
Laboratory
conversion efficiency (%)

25 MOCVD
MBE
20
LPE LPE
MOCV
15

10 Manufactured
(AM0)
5

0
1970 1980 1990
year

Figure 7.8 Chronological improvements in the efficiencies of GaAs solar cells fabricated
by the LPE, MOCVD and MBE methods.

fabricate GaAs solar cells in 1972 because it produces high-quality epitaxial film
and has a simple growth system. However, it is not as useful for devices that
involve multilayers because of the difficulty of controlling layer thickness, doping,
composition and speed of throughput. Since 1977, MOCVD has been used to
fabricate large-area GaAs solar cells because it is capable of large-scale, large-area
production and has good reproducibility and controllability.
Table 7.2 compares the advantages and disadvantages of the various epitax-
ial technologies. While LPE can produce high-quality epitaxial films, MOCVD is
effective for the large-scale, large-area production of solar cells. Molecular beam
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 316

316 M. Yamaguchi

Table 7.2 Advantages and disadvantages of epitaxial technologies.

Characteristics LPE MOCVD MBE CBE

Quality **** *** ** ***

Multiple quantum well (MQW) * *** **** ****
Abrupt interface * *** **** ****
Heavy doping ** ** *** ****
Large area * **** ** ***
Throughput *** **** ** ***
Efficient use of source materials *** ** ** ****
Equipment cost *** ** ** *

****Excellent; ***very good; **fairly good; *bad.

+ + + +

p+ p+ p+ p+
top cell

metal
n n interconnect n – n
interconnect
tunnel

++ +
n n + –
++ + n
p p + +
p p n p
bottom cell

n n p n

– – + –

2-terminal cell 3-terminal cell 4-terminal cell

Figure 7.9 Schematic diagrams of various configurations of two-junction cells.

epitaxy and CBE are advantageous for realising novel multilayer structures such as
MJ solar cells because they provide excellent controllability of monolayer abrupt-
ness and thickness due to the nature of the beam (Yamaguchi et al., 1994). However,
so far there have been few reports of CBE-grown solar cells.

7.4 Monolithic vs. multi-terminal connection modes

Figure 7.9 shows the connection options for two-junction cells: the two cells can
be connected to form two-terminal, three-terminal or four-terminal devices. In a
monolithic, two-terminal device, the cells are connected in series with an opti-
cally transparent tunnel junction intercell electrical connection. In a two-terminal
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 317

Super-High-Efficiency III–V Tandem and Multijunction Cells 317

structure, only one external circuit load is needed, but the photocurrents in the
two cells must be equal for optimal operation. Key issues for maximum-efficiency
monolithic cascade cells (two-terminal MJ cells series connected with tunnel junc-
tion) are the formation of tunnel junctions of high performance and stability for cell
interconnection, and the growth of optimum bandgap top- and bottom-cell struc-
tures on lattice-mismatched substrates, without permitting propagation of delete-
rious misfit and thermal stress-induced dislocations.
In contrast, the photocurrents in three-and four-terminal cells do not have to
be equal. However, it is very difficult to connect three-terminal devices in series, so
three-terminal tandem cells do not appear to be viable. In the four-terminal case,
two separate external circuit loads are used. Since the two individual cells are not
coupled, the photocurrents do not have to be the same. Consequently, a much larger
range of bandgap energy combinations is possible, and the changes in photocurrents
with changing solar spectral distributions do not pose serious limits. This approach
avoids the problems of lattice-mismatched epitaxial growth, current matching and
the internal electrical connection of the two-terminal device. Important issues for
obtaining high-efficiency mechanically stacked cells are the development of MJ
cell fabrication techniques such as thinning the top cell, bonding the bottom cell
to the top cell, and cell connections.

7.5 Key issues for realising high-efficiency multijunction solar cells

7.5.1 Selection of cell materials and improvement of quality
Multijunction cells with different bandgaps are stacked in tandem so that the cells
cover a wide wavelength region from 300 nm to 1800 nm. Cell materials are selected
by choosing bandgap energies close to the optimal bandgap energy combination
found from theoretical efficiency calculations, and by considering lattice match-
ing to substrates and reducing impurity problems. Since, for example, oxygen
in AlGaAs acts as recombination centre (Ando et al., 1987), reduction in resid-
ual oxygen is necessary for realising higher efficiencies. Figure 7.10 shows the
minority-carrier diffusion length dependence of GaAs single-junction solar cell
efficiency. It is clear that a high minority-carrier diffusion length L (minority-
carrier lifetime τ = L 2/D, where D is the minority-carrier diffusion coefficient)
is substantially necessary to realise high-efficiency solar cells.
Figure 7.11 shows the carrier concentration dependence of minority-carrier
lifetime in p-type and n-type GaAs (Ahrenkiel et al., 1993). The minority-carrier
lifetime τ depends on the carrier concentration N of solar cell layers as expressed by

τ = 1/B N (7.1)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 318

318 M. Yamaguchi

30

25

20
efficiency / %

junction depth 0.3 μm

15

10

0
0.1 1 10 100
minority-carrier diffusion length / μm
Figure 7.10 Minority-carrier diffusion length dependence of GaAs single-junction solar
cell efficiency.

Figure 7.11 Carrier concentration dependence of minority-carrier lifetime in p-type and

n-type GaAs (Ahrenkiel et al., 1993).

where B is the radiative recombination coefficient. B is 2 × 10−10 cm3 s−1 for

GaAs and 1.7 × 10−10 cm3 s−1 for InGaP. The carrier concentration of cell layers
is optimised by considering minority-carrier lifetimes, built-in potential and series
resistance of p-n junction diodes.
Selection of cell materials, especially the top-cell materials, is also important
for achieving high-efficiency tandem cells. It has been found by the authors (Ando
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 319

Super-High-Efficiency III–V Tandem and Multijunction Cells 319

1000
S=0 (cm/s)
Relative PL Intensity (Arb. Unit)

5800
Growth Temp. (°C)
100 104
700
650 105
750
10 106
Without
Buffer Layer 3x106
1

Without
0.1
AlInP Window

0.01
0.01 0.1 1 10 100
Minority Carrier Lifetime τ (ns)
Figure 7.12 Changes in photoluminescence (PL) intensity of the solar cell active layer as a
function of the minority-carrier lifetime and surface recombination velocity in InGaP, grown
by MOCVD.

et al., 1987) that an oxygen-related defect in the AlGaAs top-cell materials acts as
a recombination centre. As a top-cell material lattice-matched to GaAs or Ge sub-
strates, InGaP has some advantages (Olson et al., 1990), such as a lower interface
recombination velocity, a lower oxygen-related defect problem and a good window
layer material compared with AlGaAs. The top-cell characteristics depend on the
minority-carrier lifetime in the top-cell layers.
Figure 7.12 shows changes in photoluminescence (PL) intensity of the solar
cell active layer as a function of the minority-carrier lifetime τ of the p-InGaP base
layer grown by MOCVD and surface recombination velocity S. The lowest level of
S was obtained by introducing an AlInP window layer, and the highest minority-
carrier lifetime τ was obtained by introducing a buffer layer and optimising the
growth temperature. The best conversion efficiency of the InGaP single-junction
cell was 18.5% (Yang et al., 1997).

7.5.2 Lattice matching between cell materials and substrates

Lattice mismatching of cell materials to substrates should be minimised because
misfit dislocations are generated in the upper cell layers by lattice mismatch, which
decreases cell efficiency. Figure 7.13 shows the calculated and experimental dislo-
cation density dependence of the minority-carrier lifetime in GaAs (Yamaguchi and
Amano, 1985). The dependence of the dislocation density Nd on minority-carrier
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 320

320 M. Yamaguchi

Figure 7.13 Calculated and experimental dislocation density dependence of minority-

carrier lifetimes in GaAs, InP and InGaP. NTT model (Eq. (7.2)), NTT data and Ringel’s
data relate to Yamaguchi and Amano (1985), Yamaguchi et al. (2001) and Carlin et al.
(2000).

lifetime τ was calculated from the following equation (Yamaguchi and Amano,
1985):

1/τ = 1/τr + 1/τ0 + π 3 D Nd /4 (7.2)

where D is the minority-carrier diffusion coefficient, τr is the radiative recombina-

tion lifetime and τ0 is the minority-carrier lifetime associated with recombination
at other unknown defects.
Application of an InGaAs middle cell (Takamoto et al., 2003) lattice-matched
to Ge substrates has been demonstrated to increase the open-circuit voltage (Voc)
due to improved lattice matching, and also the short-circuit current density (i sc )
due to the decrease in bandgap energy of the middle cell.

7.5.3 Effectiveness of wide-bandgap windows and back-surface

field layers
Figure 7.14 shows the effect of surface recombination velocity on the short-circuit
current density of In0.14Ga0.86 As homojunction solar cells as a function of junction
depth; because GaAs-related materials have high surface recombination velocity
of about 107 cm/s−1 , formation of a shallow junction with a junction depth of less
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 321

Super-High-Efficiency III–V Tandem and Multijunction Cells 321

35
S=1E+4 cm/s
–2

S=1E+5 cm/s
30
short-circuit current density / A cm

S=1E+6 cm/s
S=1E+7 cm/s
25

20

15

10

0
0 0.5 1 1.5 2 2.5
junction depth / μm

Figure 7.14 Surface recombination effect velocity on the short-circuit current density of a
In0.14 Ga0.86 As homojunction solar cell as a function of junction depth.

than 50 nm is necessary. In order to decrease the efficiency drop due to front and
rear surface recombination as shown in Fig. 7.15, formation of a heteroface or
double-hetero structure is necessary.
Figure 7.15 shows the changes in Voc and i sc of InGaP single-junction cells as
a function of the potential barrier E. A wide-bandgap back-surface field (BSF)
layer (Takamoto et al., 2003) is found to be the most effective for confinement of
minority carriers.

7.5.4 Low-loss tunnel junctions for intercell connection and preventing

impurity diffusion from the tunnel junction
Another important issue for realising high-efficiency monolithic-cascade type tan-
dem cells is the achievement of optically and electrically low-loss interconnection
of two or more cells. A degenerately doped tunnel junction is attractive because
creating it only involves one extra step in the growth process. To minimise optical
absorption and achieve higher short-circuit current density for the bottom cell, as
shown in Fig. 7.16, the formation of physically thin wide-bandgap tunnel junc-
tions is necessary. In order to form a wide-bandgap tunnel junction, doping of
higher concentration impurities into the tunnel junction is necessary because the
tunnelling current decreases exponentially with increase in bandgap energy.
In addition, impurity diffusion from a highly doped tunnel junction during
overgrowth of the top cell increases the resistivity of the tunnel junction. As shown
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 322

322 M. Yamaguchi

Figure 7.15 Change in Voc and i sc of InGaP single-junction cells as a function of the
potential barrier E.

in Fig. 7.17, a double-hetero (DH) structure was found by the authors to be useful
for preventing diffusion (Sugiura et al., 1988). An InGaP tunnel junction was tried
for the first time for an InGaP/GaAs tandem cell in our work (Takamotoet al.,
1997b). Zn and Si were used as p-type and n-type dopants, respectively. The peak
tunnelling current of the InGaP tunnel junction increased from 5 mA cm−2 up to
2 A cm−2 on making a DH structure with AlInP barriers. Effective suppression of
the Zn diffusion from tunnel junction by the InGaP tunnel junction with the AlInP-
DH structure is thought to be due to the lower diffusion coefficient (Takamoto
et al., 1999) for Zn in the wider bandgap energy materials such as the AlInP barrier
layer and InGaP tunnel junction layer. In conclusion, the InGaP tunnel junction is
very effective for obtaining high tunnelling currents, and the DH structure is useful
for preventing diffusion.
Table 7.3 summarises factors of importance in achieving MJ cells of very high
efficiency.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 323

Super-High-Efficiency III–V Tandem and Multijunction Cells

–2 323

104
tunnel peak current density / A cm

3
10

–2
isc of bottom cell / mA cm
2 15
10
10

1 10

10–1
10–2 5
10–3
–4
10 0
0.5 1.0 1.5 2.0
bandgap energy / eV
Figure 7.16 Calculated tunnel peak current density and short-circuit current density of a
GaAs bottom cell as a function of the bandgap energy of the tunnel junction.

10
–2
tunnel peak current density / A cm

10–1

–2
10

–3
10

10
500 600 700 800
o
annealing temperature / C
Figure 7.17 Annealing temperature dependence of tunnel peak current densities for double-
hetero structure tunnel diodes. X is the Al mole fraction in the Al x Ga1−x As barrier layers.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 324

324 M. Yamaguchi

Table 7.3 Key issues for realising super-high-efficiency multijunction solar cells.

Key issue Past Present Future

Top cell AlGaAs InGaP AlInGaP

materials
Third cell None Ge InGaAsN etc.
materials
Bottom cell GaAs Ge Si, Ge, InGaAs
materials
Substrate GaAs Ge Si, Ge, GaAs, metal
Tunnel junction DH-structure DH-structure InGaP TJ DH-structure InGaP or
(TJ) GaAs TJ GaAs TJ
Lattice GaAs middle InGaAs middle cell (In)GaAs middle cell
matching cell
Carrier InGaP-BSF AlInP-BSF Wide-gap-BSF
confinement Quantum dots (QDs)
Photon None None Bragg reflector, QDs,
confinement etc.
Others None Inverted epitaxial Inverted epitaxial
growth (i.e. epitaxial growth, epitaxial lift
growth on substrate off (i.e. thin film
from top cell to peeled off)
bottom cell)

7.6 High-efficiency InGaP/GaAs/Ge three-junction solar cells

and their space applications
7.6.1 Development of high-efficiency InGaP/GaAs/Ge three-junction
solar cells
As one of the projects in the Sunshine Programme in Japan, a research and devel-
opment (R&D) project for super-high-efficiency MJ solar cells was started in 1990.
The conversion efficiency of InGaP/GaAs-based multijunction solar cells was
improved under this programme by the application of the following technologies:

• Selection and high-quality growth of InGaP as a top-cell material.

• Use of double-hetero structures and wide-bandgap tunnel junctions for cell
interconnection.
• Precise lattice matching of the InGaP top cell and the InGaAs middle cell with
the Ge substrate.
• Use of AlInP as a back-surface-field layer for the InGaP top cell.
• Use of InGaP-Ge heteroface structure bottom cells.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 325

Super-High-Efficiency III–V Tandem and Multijunction Cells 325

Approaches for high-efficiency

triple-junction cells

(1) Wide-gap tunnel junction

with double heterostructure
High transmittance
High potential barrier
(2) Combination of Ge cell with
InGaP 1st hetero-layer
Shallow junction
(3) Precise lattice matching by
adding 1% indium
No misfit dislocations
(4) Widening top-cell bandgap
(developing 1.96 eV
AlInGaP)
Increase of Voc

Figure 7.18 Schematic illustration of a triple-junction cell and approaches for improving
its efficiency.

As a result of the above innovations and performance improvements, the

1 Sun conversion efficiency of InGaP/(In)GaAs/Ge three-junction solar cells was
improved to 31.7% (AM 1.5G) by the turn of the century (Takamoto et al.,
2000). Figure 7.18 shows a schematic illustration of the InGaP/(In)GaAs/Ge triple-
junction solar cell and the key technologies for improving its conversion efficiency.
Although most recently, a 1 Sun efficiency of 37.9% has been achieved with
InGaP/GaAs/InGaAs three-junction solar cells by Sharp (Sharp, 2013; Sasaki,
2013; Yamaguchi and Luque, 2013; Green et al., 2013), the key technologies for
improving efficiency are based on the following key issues.

Wide-bandgap tunnel junction

Use of a wide-bandgap tunnel junction which consists of the double-hetero struc-
ture p-Al(Ga)InP/ p-AlGaAs/n-(Al)InGaP/n-Al(Ga)InP increases the amount of
light incident on the (In)GaAs middle cell and produces effective potential bar-
riers for minority carriers generated in the top and middle cells. Both the open-
circuit voltage and the short-circuit current of the cells are improved by such
wide-gap tunnel junctions without suffering absorption and recombination losses
(Takamoto et al., 1997b). It is difficult to obtain a high tunnelling peak current
with wide-gap tunnel junctions, so thinning the depletion layer width by formation
of a highly doped junction is necessary. Since impurity diffusion occurs during
growth of the top cell (Sugiura et al., 1988), carbon and silicon, which both have
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 326

326 M. Yamaguchi

low diffusion coefficients, are used as dopants for p-type AlGaAs and n-type
(Al)InGaP, respectively. Furthermore, the double-hetero structure appears to sup-
press impurity diffusion from the highly doped tunnel junction (Takamoto et al.,
1999). The second tunnel junction between the middle and bottom cells consists
of p-InGaP/ p-(In)GaAs/n-(In)GaAs/n-InGaP, which has a wider bandgap than the
middle cell materials.

Heteroface-structure Ge bottom cells

InGaP/GaAs cell layers are normally grown on a p-type Ge substrate. A n- p junc-
tion is formed automatically during MOCVD growth by diffusion of the V-group
atom from the first layer grown on the Ge substrate, so the material of the first het-
erolayer is important for the performance of the Ge bottom cell. An InGaP layer is
thought to be suitable as material for the first heterolayer, because phosphorus has
a lower diffusion coefficient in Ge than arsenic, and indium has a lower solubility
in Ge than gallium.
Figure 7.19 shows the change in spectral response of the triple-junction cell
on changing the first hetero-growth layer on Ge from GaAs to InGaP. The quantum
efficiency of the Ge bottom cell was improved by using the InGaP hetero-growth
layer. In the case of the GaAs hetero-growth layer, the junction depth was measured
to be around 1 µm. On the other hand, the thickness of the n-type layer produced
by phosphorus from the InGaP layer was 0.1 µm. Takamoto et al. (2003, 2005)
confirmed that the increase in Ge quantum efficiency was due to this reduction in
junction depth.

Figure 7.19 Change in the spectral response due to modification of the first heterolayer
from GaAs to InGaP (without antireflection coating).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 327

Super-High-Efficiency III–V Tandem and Multijunction Cells 327

0% In 1% In 2% In

100 µm

Figure 7.20 Surface morphology of InGaAs with various indium compositions grown
on Ge.

Precise lattice-matching to the Ge substrate

Although the 0.08% lattice mismatch between GaAs and Ge is thought to be negligi-
bly small, misfit dislocations are found to be generated in thick GaAs layers, which
diminishes cell performance. By adding about 1% indium into the InGaP/GaAs
cell layers, all cell layers are lattice-matched precisely to the Ge substrate. As a
result, the crosshatch pattern caused by misfit dislocations due to lattice mismatch
disappears in the surface morphology of the cells, as shown in Fig. 7.20. Elimi-
nating the misfit dislocations in this way was found to influence the open-circuit
voltage of the cell but not the short-circuit current. In addition, the wavelength
of the absorption edge increased and the short-circuit current of the both top and
middle cells was increased.

7.6.2 Radiation resistance of InGaP-based multijunction solar cells

Radiation in space is severe, particularly in the Van Allen radiation belt, and lat-
tice defects are induced in semiconductors by the high-energy electron and proton
irradiation of devices sited in this belt. These defects cause a decrease in the out-
put power of solar cells deployed in space. Figure 7.21 shows the effectiveness
of radiation resistance and high conversion efficiency of space cells in order to
increase power density (W m−2 ) for space missions. Before 1995, single-crystal
Si solar cells were used in space but their BOL (beginning-of-life) and EOL (end-
of-life) efficiencies (about 15% and 10%, respectively) were low and insufficient.
Developing higher conversion efficiency and highly radiation-resistant space cells
was necessary for widespread applications of PV in space missions. InGaP/GaAs-
based MJ solar cells have attracted increased attention in this regard because of
the possibility of high conversion efficiencies of over 40% and good radiation
resistance.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 328

328 M. Yamaguchi

degradation rate 0% degradation rate 20%

degradation rate 40%

600
power density / W m–2

500
400
300
200
100
0
10 15 20 25 30 35 40
BOL efficiency / %
Figure 7.21 Effectiveness of radiation resistance and high conversion efficiency of space
cells in increasing the power density of space missions. Beginning-of-life efficiency, BOL,
is the efficiency of the space solar cells before space satellite launching.

0.8
15
1 MeV - 10 e / cm –2
0.78 photo-injection: 100 mW / cm

75°C
power ratio PI / P0

0.76
50°C
0.74 25°C

0.72

0.7
0 10 20 30 40
photo-injection time / min

Figure 7.22 Maximum power recovery of the InGaP/GaAs tandem cell due to light illu-
mination at various temperatures.

Figure 7.22 shows the recovery of maximum power produced by light of

100 mW cm−2 intensity at various temperatures for InGaP/GaAs tandem cells that
have been irradiated with a 1-MeV 1 × 1015 cm−2 electron flux (Yamaguchi et al.
1997). The ratios of maximum power after injection, PI , to maximum power before
irradiation, P0 , are shown as a function of injection time. Even at room temperature,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 329

Super-High-Efficiency III–V Tandem and Multijunction Cells 329

Figure 7.23 Deep-level transient spectroscopy (DLTS) spectrum of trap H2 (Ev + 0.55 eV)
for various injection times at 25◦ C with an injection density of 100 mA cm−2 .

photoinjection-enhanced annealing of radiation damage to InGaP/GaAs tandem

cells was observed. The recovery ratio increases with increase in ambient tem-
perature within the operating range for space use. Such a recovery is attributed
to damage recovery in the InGaP top cell layer (Yamaguchi et al., 1997). These
results indicate InGaP/GaAs tandem cells have superior radiation-resistance under
device operation conditions.
Figure 7.23 shows DLTS (deep-level transient spectroscopy) spectra of the
hole trap H2 (Ev+0.55 eV) for various injection times at 25◦ C with a forward
bias injection density of 100 mA cm−2 . Khan et al. (2000) found from DLTS
measurements that this major defect level recovers under forward bias or light
illumination, i.e. the defect signal decreases with prolonged light exposure. More-
over, the H2 centre was confirmed to act as a recombination centre by using
the double carrier pulse DLTS method. The enhancement of defect annealing in
the InGaP top-cell layer under minority-carrier injection conditions is thought to
occur as a result of the non-radiative electron-hole recombination process (Lang
et al., 1976), whose energy E R enhances the defect motion. The thermal activa-
tion energy E A (1.1 eV) of the defect is reduced to E I (0.48 ∼ 0.54 eV) by an
amount E R (0.56 ∼ 0.62 eV). Thus electronic energy from a recombination event
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 330

330 M. Yamaguchi

InGaP/GaAs-3J
efficiency / %

silicon

year

Figure 7.24 Sharp space solar cell conversion efficiency heritage. CONV., BSF, BSFR,
NRS/BSF (IBF) show conventional p-n junction, back-surface-field, back-surface-field-
reflector, non-reflective surface/back-surface field (improved back-surface field) structures
for Si space solar cells. AHES shows advanced high-efficiency solar cell structures for space
cells. BOL and EOL show beginning-of-life and end-of-life, respectively. All efficiencies
for 1 Sun AM 0 under standard conditions.

can be channelled into the lattice vibration mode which drives the defect motion:
EI = E A − ER .

7.6.3 Space applications of InGaP/GaAs/Ge three-junction solar cells

Advanced technologies for high-efficiency cells and the discovery of the superior
radiation resistance of InGaP-based materials are contributing to the industrialisa-
tion of InGaP-based MJ space solar cells in Japan; InGaP/GaAs/Ge three-junction
cells have been commercialised for space use in Japan since 2002 (Takamoto et al.,
2005; Imaizumi et al., 2005). Figure 7.24 shows the Sharp space solar cell conver-
sion efficiency heritage up to 2003.

7.7 Multijunction solar cells: recent results

InGaP/GaAs-based MJ solar cells have drawn increased attention for terrestrial
applications because concentrator operation of MJ cells has great potential for
providing high-performance, low-cost solar cell modules. For concentrator appli-
cations, the cell contact grid structure should be designed to reduce the energy loss
due to series resistance. Efficiencies have steadily increased: Sharp demonstrated
38.9% efficiency at 489 Suns (Takamoto et al., 2005). Fraunhofer ISE achieved
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 331

Super-High-Efficiency III–V Tandem and Multijunction Cells 331

41.1% efficiency under 454 Suns with In0.65 Ga0.35 P/In0.17 GaAs0.83 /Ge 3-J cells
(Bett et al., 2009), and Spectrolab reached 41.6% efficiency under 364 Suns with
lattice-matched InGaP/InGaAs/Ge 3-J cells (King et al., 2009). More recently,
Spire achieved 42.3% efficiency under 406 Suns with bifacial epitaxially grown
InGaP/GaAs/InGaAs three-junction cells (Wojtczuk et al., 2010) and Solar Junc-
tion (Solar Junction, 2011; Sabnis, 2012) reached 43.5% efficiency under 418 Suns
with lattice-matched InGaP/GaAs/GaInNAs 3-J cells. At the time of writing, the
world record is held by Sharp with their InGaP/GaAs/InGaAs 3-J solar cells, which
are 44.4% efficient at 302 Suns (Sharp, 2013; Sasaki, 2013; Yamaguchi and Luque,
2013; Green et al., 2013).
The 1 Sun efficiency of triple-junction solar cells is also improving. A world-
record efficiency (37.9%) at 1 Sun (AM 1.5 G) has recently been realised with
inverted epitaxially grown InGaP/GaAs/InGaAs three-junction cells by Sharp
(Sharp, 2013; Sasaki, 2013; Yamaguchi and Luque, 2013; Green et al., 2013).
Figure 7.25 shows the fabrication process of this cell (Takamoto et al., 2010;
Yoshida et al., 2011) and Fig. 7.26 shows its current–voltage curve. Figure 7.27
shows the chronological improvement in the conversion efficiencies of III–V com-
pound MJ solar cells under 1 Sun and concentrator conditions.

Figure 7.25 Fabrication process of inverted epitaxially grown InGaP/GaAs/InGaAs 3-J

solar cells (Takamoto et al., 2010; Yoshida et al., 2011).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 332

332 M. Yamaguchi

Figure 7.26 Current–voltage curve of world-record efficiency InGaP/GaAs/InGaAs cell as

measured by the AIST (Sharp, 2013; Sasaki, 2013; Yamaguchi and Luque, 2013).

target > 48%

45
number of junctions III–V, concentrator
1-junction 44.4％
■
2-junction this time
40
3-junction

33.3%
efficiency / %

35 JE, Sumitomo 37.9％

TTI this time

35.8％
30 previous
30.3% III–V, 1 Sun AM1.5G
JE

25
single Si, 1 Sun AM1.5G

20
1985 1990 1995 2000 2005 2010 2015
year
Figure 7.27 Chronological improvements in conversion efficiencies of III–V compound
multijunction solar cells under 1 Sun and concentrator conditions (Yamaguchi and Luque,
2013).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 333

Super-High-Efficiency III–V Tandem and Multijunction Cells 333

7.8 Future directions

Multijunction solar cells will be widely used in space because of their high
conversion efficiency and good radiation resistance. However, in order to apply
super-high-efficiency cells widely on Earth, it will be necessary to improve
their conversion efficiency and reduce their cost. As described in Chapter 11,
concentrator PV (CPV) systems with several times more annual power-generation
capability than conventional crystalline silicon flat-plate systems will open a new
market for apartment or building rooftop applications. Other interesting applica-
tions are in agriculture and large-scale PV power plants.
Currently, efficiencies of over 40% are achieved by concentrator MJ solar cells,
as shown in Fig. 7.28. Concentrator four-junction or five-junction solar cells have
great potential for realising super-high efficiencies of over 50% (Goetzberger et al.,
2001; Yamaguchi, 2004; updated by using Solar Efficiency Tables [Green et al.,
2011]). Lattice-mismatched and III–V–N compounds are thought to be promising
materials for realising more than 50% efficiency. As a three-junction combination,
InGaP/InGaAs/Ge cells on Ge substrates will be widely used because this system
has already been developed. The four-junction combination of a top cell with a
bandgap energy E g = 2.0 eV, a GaAs second-layer cell, a third-layer cell with a
material of bandgap 1.05 eV and a Ge bottom cell would be lattice-matched to Ge

50

III−V multijunction
40
efficiency / %

30

thin-film Si
CIS
20 dye-sensitised
cryst. Si
organic
10
a-Si

0
1940 1960 1980 2000 2020 2040 2060
year
Figure 7.28 Predictions of future solar cell efficiencies (Goetzberger et al., 2001; Yam-
aguchi, 2004; updated by using Solar Cell Efficiency Tables [Green et al., 2013]).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 334

334 M. Yamaguchi

1:Si crystal
2:thin-film
3:concentrator
100
2
Electricity Cost (Yen/kWh)

3
10

1
1995 2000 2005 2010 2015 2020 2025 2030
Year

Figure 7.29 Scenario of electricity cost reduction by developing concentrator solar cells
(Yamaguchi, 2003).

substrates and would have a theoretical efficiency of about 42% under 1 Sun AM
0, and over 47% under 500 Suns AM 1.5 G (Kurtz et al., 1997).
In conclusion, as shown in Fig. 7.29 (Yamaguchi, 2004), we would like to
contribute to commercialisation of CPV technologies as the third-generation PV
technologies in succession to the first-generation crystalline Si PV and the second-
generation thin-film PV technologies.

Acknowledgements
This work was partially supported by the Japanese New Energy and Industrial Tech-
nology Development Organization (NEDO) under METI (Ministry of Economy,
Trade and Industry). The author thanks members of the Toyota Institute, Sharp,
Daido Steel, JAXA, JAEA, the University of Tokyo, Meijo University, Kyushu
University and Miyazaki University for their collaboration and cooperation.

References
Ahrenkiel R. K., Keyes B. M., Durbin S. M. and Gray J. L. (1993), ‘Recombination lifetime
and performance of III–V compound photovoltaic devices’, Proc. 23rd IEEE Photo-
voltaic Specialists Conf., IEEE, New York, NY, pp. 42–51.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 335

Super-High-Efficiency III–V Tandem and Multijunction Cells 335

Amano C., Sugiura H., Yamaguchi M. and Hane K. (1989), ‘Fabrication and numerical anal-
ysis of AlGaAs/GaAs tandem solar cells with tunnel interconnections’, IEEE Trans.
Electron Devices ED–36, 1026–1035.
Ando K., Amano C., Sugiura H., Yamaguchi M. and Saletes A. (1987), ‘Non-radiative
e-h recombination characteristics of mid-gap electron trap in Alx Ga1−x As (x = 0.4)
grown by molecular beam epitaxy’, Jpn. J. Appl. Phys. 26, L266–L269.
Araki K., Uozumi H., Egami T., Hiramatsu M., Miyazaki Y., Kemmoku Y., Akisawa A.,
Ekins-Daukes N. J., Lee H.-S. and Yamaguchi M. (2005), ‘Development of concentrator
modules with dome-shaped Fresnel lenses and triple-junction concentrator cells’,
Progr. Photovoltaics 13, 513–527.
Bertness K. A., Kurtz S. R., Friedman D. J., Kibbler A. E., Kramer C. and Olson J.
M. (1994), ‘29.5%-Efficiency GaInP/GaAs tandem solar cells’, Appl. Phys. Lett. 65,
989–991.
Bett A. W., Dimroth F., Guter W., Hoheisel R., Oliva E., Philips S. P., Schone J., Siefer G.,
Steiner M., Wekkeli A., Welser E., Meusel M., Kostler W. and Strobl G. (2009), ‘Highest
efficiency multijunction solar cell for terrestrial and space applications’, Proc. 24th
European Photovoltaic Solar Energy Conf., WIP, Munich, pp. 1–6.
Bowler D. L. and Wolf M. (1980), ‘Interactions of efficiency and material requirements for
terrestrial silicon solar cells’, IEEE T Compon. Hybr., CHMT–3, 464–472.
Brown M. R., Goldhammer L. J., Goodelle G. S., Lortz C. U., Perron J. N., Powe J. S.
and Schwartz J. A. (1997), ‘Characterization testing of dual junction GaInP2 /GaAs/Ge
solar cell assemblies’, Proc. 26th IEEE Photovoltaic Specialists Conf., IEEE, New
York, NY, pp. 805–810.
Carlin J. A., Hudait M. K., Ringel S. S., Wilt D. M., Clark E. B., Leitz C. W., Currie M.,
Langdo T. and Fitzgerald E. A. (2000), ‘High-efficiency GaAs-on-Si solar cells with
high Voc using graded GeSi bufers’, Proc. 28th IEEE Photovoltaic Specialists Conf.,
IEEE, New York, NY, pp. 1006–1011.
Chung B.-C., Virshup G. F., Hikido S. and Kaminar N. R. (1989), ‘27.6% efficiency (1 Sun,
air mass 1.5) monolithic Al0.37 Ga0.63 As/GaAs two-junction cascade solar cell with
prismatic cover glass’, Appl. Phys. Lett. 55, 1741–1743.
Fan J. C. C., Tsaur B. Y. and Palm B. J. (1982), ‘Optical design of high-efficiency tan-
dem cells’, Proc. 16th IEEE Photovoltaic Specialists Conf., IEEE, New York, NY,
pp. 692–701.
Flores C. (1983), ‘A three-terminal double junction GaAs/GaAlAs cascade solar cells’,
IEEE Electr. Device L. EDL–4, 96–99.
Fraas L. M., Avery J. E., Martin J., Sundaram V. S., Girard G., Dinh V. T., Davenport T. M.,
Yerkes J. W. and O’Neill M. J. (1990), ‘Over 35% efficient GaAs/GaSb tandem solar
cells’, IEEE Trans. Electr. Devices 37, 443–449.
Goetzberger G., Luther J. and Willeke G. (2002), ‘Solar cells: past, present, future’, Solar
Energy Mater. Solar Cells, 74, 1–11.
Green M., Emery K., Hishikawa Y., Warta W. and Dunlop E. D. (2013), ‘Solar efficiency
tables (Version 42)’, Progr. Photovoltaics 21, 827–837.
Hutchby J. A., Markunas R. J., Timmons M. L., Chiang P. K. and Bedair S. M. (1985), ‘A
review of multijunction concentrator solar cells’, Proc.18th IEEE Photovoltaic Spe-
cialists Conf., IEEE, New York, NY, pp. 20–27.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 336

336 M. Yamaguchi

Imaizumi M., Matsuda S., Kawakita S., Sumita T., Takamoto T., Ohshima T. and
Yamaguchi M. (2005), ‘Activity and current status of R&D on space solar cells in
Japan’, Progr. Photovoltaics 13, 529–543.
Jackson E. D. (1955), ‘Areas for improving of the semiconductor solar energy converter’,
Trans. Conf. on the Use of Solar Energy 5, University of Arizona Press, Tucson (1958),
pp. 122–126.
Khan A., Yamaguchi M., Bourgoin J. C. and Takamoto T. (2000), ‘Room-temperature
minority-carrier injection-enhanced recovery of radiation-induced defects in p-InGaP
and solar cells’, Appl. Phys. Lett. 76, 2559–2561.
King R. R., Boca A., Hong W., Liu X.-Q., Bhusari D., Larrabee D., Edmondson K. M.,
Law D. C., Fetzer C. M., Mesropian S. and Karam N. H. (2009), ‘Bandgap-engineered
architectures for high-efficiency multijunction concentrator solar cells’, Proc. 24th
European Photovoltaic Solar Energy Conf., WIP, Munich, pp. 55–61.
Kurtz S. R., Myers D. and Olson J. M. (1997), ‘Projected performance of three-and four-
junction device using GaAs and GaInP’, Proc. 26th IEEE Photovoltaic Specialists
Conf., IEEE, New York, NY, pp. 875–878.
Lamorte M. F. and Abbott D. H. (1980), ‘Computer modeling of a two-junction monolithic
cascade solar cell’, IEEE Trans. Electr. Devices ED–25, 831–840.
Lang D. V., Kimerling L. C. and Leung S. Y. (1976), ‘Recombination-enhanced annealing
of the E1 and E2 defect levels in 1-MeV-electron-irradiated n-GaAs’, J. Appl. Phys.
47, 3587–3591.
Letay G. and Bett A. W. (2001), ‘Theoretical investigation III–V multijunction solar cells’,
Proc. 17th European Photovoltaic Solar Energy Conf., WIP, Munich, pp. 178–181.
Loferski J. J. (1976), ‘Tandem photovoltaic solar cells and increased energy conversion
efficiency’, Conf. Record 12th IEEE Photovoltaic Specialists Conf., Baton Rouge.
IEEE Press, Piscataway, NJ, pp. 957–961.
Ludowise M. J., LaRue R. A., Borden P. G., Gregory P. E. and Dietz W. T. (1982), ‘High-
efficiency organometallic vapor phase epitaxy AlGaAs/GaAs monolithic cascade solar
cell using metal interconnects’, Appl. Phys. Lett. 41, 550–552.
Mitchell K. W. (1981), ‘High efficiency concentrator cells’, Conf. Record 15th IEEE
Photovoltaic Specialists Conf., Kissimmee. IEEE Press, Piscataway, NJ, pp. 142–146.
Nell M. E. and Barnett A. M. (1987), ‘The spectral p-n junction model for tandem solar-cell
design, IEEE Trans. Electr. Devices ED–34, 257–266.
Olson J. M., Kurtz S. R. and Kibbler A. E. (1990), ‘A 27.3% efficient Ga0.5 In0.5 P/GaAs
tandem solar cell’, Appl. Phys. Lett. 56, 623–625.
Sabnis V., Yuen H. and Wiemer M. (2012), ‘High-efficiency multijunction solar cells
employing dilute nitrides’, Proc. 9th Int. Conf. on Concentrating Photovoltaics Sys-
tems, Toledo, Spain.
Sasaki K., Agui T., Nakaido K., Takahashi N., Onitsuka R. and Takamoto T. (2013), ‘Devel-
opment of InGaP/GaAs/InGaAs inverted triple junction concentrator solar cells’, Proc.
9th Int. Conf. on Concentrating Photovoltaics Systems, Miyazaki, Japan.
Sharp (2013), http://www.sharp.co.jp.
Solar Junction (2011), http://www.sj-solar.com.
Sugiura H., Amano C., Yamamoto A. and Yamaguchi M. (1988), ‘Double heterostructure
GaAs tunnel junction for AlGaAs/GaAs tandem solar cells’, Jpn. J. Appl. Phys. 27,
269–272.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 337

Super-High-Efficiency III–V Tandem and Multijunction Cells 337

Takamoto T., Agui T., Ikeda M. and Kurita E. (2000), ‘High-efficiency InGaP/GaAs tandem
solar cells on Ge substrates’, Proc. 28th IEEE Photovoltaic Specialists Conf., IEEE,
New York, NY, pp. 976–981.
Takamoto T., Agui T., Kamimura K., Kaneiwa M., Imaizumi M., Matsuda S. and
Yamaguchi M. (2003), ‘Multijunction solar cell technologies — high efficiency,
radiation resistance and concentrator applications’, Proc. 3rd World Conf. Photovoltaic
Energy Conversion, WCPEC-3, Osaka, pp. 581–586.
Takamoto T., Agui T., Yoshida A., Nakaido K., Juso H., Sasaki K., Nakamora K.,
Yamaguchi H., Kodama T., Washio H., Imaizumi M. and Takahashi M. (2010),
‘World’s highest efficiency triple-junction solar cells fabricated by inverted layers trans-
fer process’, Proc. 35th IEEE Photovoltaic Specialists Conf., IEEE, New York, NY,
pp. 412–417.
Takamoto T., Ikeda E., Kurita H. and Ohmori M. (1997a), ‘Over 30% efficient InGaP/GaAs
tandem solar cells’, Appl. Phys. Lett. 70, 381–383.
Takamoto T., Ikeda E., Kurita H., Ohmori M. and Yamaguchi M. (1997b), ‘Two-terminal
monolithic InGaP/GaAs tandem solar cells with a high conversion efficiency of over
30%’, Jpn. J. Appl. Phys. 36, 6215–6220.
Takamoto T., Kaneiwa M., Imaizumi M. and Yamaguchi M. (2005), ‘InGaP/GaAs-based
multijunction solar cells’, Progr. Photovoltaics 13, 495–511.
Takamoto T., Yamaguchi M., Ikeda E., Agui T., Kurita H. and Al-Jassim M. (1999), ‘Mecha-
nism of Zn and Si diffusion from highly doped tunnel junction for InGaP/GaAs tandem
solar cells’, J. Appl. Phys. 85, 1481–1485.
WBGU (German Advisory Council on Global Change) (2003), World in Transition —
Towards Sustainable Energy Systems, Earthscan, London, http://wbgu.de.
Wojtczuk S., Chiu P., Zhang X., Derkacs D., Harris C., Pulver D. and Timmons
M. (2010), ‘InGaP/GaAs/InGaAs 41% concentrator cells under bi-facial epi-
growth’, Proc. 35th IEEE Photovoltaic Specialists Conf., IEEE, New York, NY,
pp. 1259–1264.
Wolf M. (1960), ‘Limitations and possibilities for improvement of photovoltaic solar energy
converters’, Proc. Inst. Radio Engineers 48, 1246–1263.
Yamaguchi M. (2003), ‘III–V compound multijunction solar cells: present and future’, Solar
Energy Mater. Solar Cells, 75, 261–269.
Yamaguchi M. (2004), ‘Toward 50% efficiency III–V compound multijunction and concen-
trator solar cells’, Proc. 19th European Photovoltaic Solar Energy Conf., WIP, Munich,
xl–xlii.
Yamaguchi M. and Amano C. (1985), ‘Efficiency calculations of thin film GaAs solar cells
on Si substrates’, J. Appl. Phys. 58, 3601–3606.
Yamaguchi M., Amano C., Sugiura H. and Yamamoto A. (1987), ‘High-efficiency
AlGaAs/GaAs tandem solar cells’, Proc. 19th IEEE Photovoltaic Specialists Conf.,
IEEE, New York, NY, pp. 1484–1485.
Yamaguchi M. and Luque, A. (1999), ‘High efficiency and high concentration in photo-
voltaics’, IEEE Trans. Electron Devices 46, 2139–2144.
Yamaguchi M. and Luque A. (2013), ‘Outline of Europe–Japan collaborative research on
concentrator photovoltaics’, Proc. 39th IEEE Photovoltaic Specialists Conf., IEEE,
New York, NY.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch07 page 338

338 M. Yamaguchi

Yamaguchi M., Ohmachi Y., Oh’hara T., Kadota Y., Imaizumi M. and Matsuda S., (2001),
‘GaAs solar cells grown on Si substrates for space use’, Progr. Photovoltaics 9,
191–201.
Yamaguchi M., Okuda T., Taylor S. J., Takamoto T., Ikeda E. and Kurita H. (1997), ‘Superior
radiation-resistant properties of InGaP/GaAs tandem solar cells’, Appl. Phys. Lett.
70, 1566–1568.
Yamaguchi M., Warabisako T. and Sugiura H. (1994), ‘CBE as a breakthrough technology
for PV solar energy applications’, J. Crystal Growth 136, 29–36.
Yang M.-J., Yamaguchi M., Takamoto T., Ikeda E., Kurita H. and Ohmori M. (1997), ‘Photo-
luminescence analysis of InGaP top cells for high-efficiency multijunction solar cells’,
Solar Energy Mater. Solar Cells 45, 331–339.
Yoshida A., Agui T., Nakaido K., Murasawa K., Juso H., Sasaki K. and Takamoto T. (2011),
‘Development of InGaP/GaAs/InGaAs inverted triple junction solar cells for concen-
trator application’, Extended Abstracts of 21st International Photovoltaic Science and
Engineering Conf., Fukuoka, Japan, 28 November–2 December, 4B–4O–01.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 339

CHAPTER 8

ORGANIC PHOTOVOLTAICS
DAN CREDGINGTON
Cavendish Laboratory, JJ Thomson Avenue
Cambridge CB3 0HE, UK
[email protected]

One day carpenters will fashion plastic trees,

And artists will paint their nylon leaves.
Anonymous.

8.1 Introduction
In this chapter, we move our focus away from the numerous elemental and crys-
talline semiconductors which dominate the existing technological landscape of
photovoltaics. For these materials, it is the specific arrangement of elements in a
(usually) crystalline lattice which gives rise to delocalised electronic states and
resulting band structure. The quality of the band structure therefore depends rather
sensitively on how such lattices are formed, and the necessary delocalisation of
electrons leads to materials that are highly sensitive to defects — including both
impurities and crystallite boundaries. The creation of high-performance materials
by this approach is an intrinsically intolerant, and therefore expensive, process. As
is detailed elsewhere in this volume, the primary brake on uptake of photovoltaics
for the last several decades has been the relatively high cost of silicon crystal, which
must be processed from a purified melt at extremely high temperature. While the
price of silicon crystal has reduced considerably as the demand for solar energy
has grown, routes that circumvent high-tolerance manufacturing may lead to even
cheaper photovoltaic devices.
An alternative approach is to utilise materials in which semiconducting
behaviour is intrinsic, rather than an emergent property of lattice formation.
Molecular semiconductors are such a class of materials, whereby pre-synthesised
molecules with desirable electronic structure are used to form the light-absorbing
layer of the photovoltaic device. Interactions between molecules lead to broaden-
ing of the molecular orbitals, but this is a relatively small perturbation, which does
not typically lead to three-dimensional extended electronic states. The advantage

339
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 340

340 D. Credgington

of such an approach is that the electronic properties of molecular semiconduc-

tors are largely ‘hard-wired’ into the molecule, and do not rely on a particular
crystal structure being formed. This renders them significantly more tolerant to
impurities, packing defects or grain boundary effects and so allows far greater
scope for low-cost processing. Decoupling electronic and physical structure also
casts bandgap engineering as a problem of molecular design, rather than crystal
composition, and thus renders it accessible to the vast field of synthetic chemistry. In
particular, the tremendous flexibility of organic (carbon-based) chemistry has led to
the dominance of organic molecules within molecular semiconductor technology.
In this chapter, we examine the nature of semiconducting behaviour in organic
molecules, the challenges and advantages associated with their use in photovoltaic
devices, and the progress made in recent years to bring this technology to fruition.
This chapter is split into two main themes. Sections 8.2 and 8.3 provide an overview
of organic semiconductors, their operation in photovoltaic devices and the current
state of the art. These sections are intended to provide a broad introduction to
the technology. Sections 8.4 and 8.5 examine in more detail the processes gov-
erning organic photovoltaic (OPV) operation, in terms of current understanding
and empirical observation. These sections are intended to provide a more in-depth
description of current OPV research, and highlight some of the questions that
remain within the field.

8.2 Basic concepts in organic semiconductors

8.2.1 Chemistry and conjugation
Semiconducting and metallic organic materials are based on chains and sheets of
sp2 -hybridised carbon. Neutral carbon contains six electrons, two tightly bound in
1s orbitals and four valence electrons in a 2s 2 , 2 p2 configuration. This configu-
ration is maintained only in atomic carbon; in a carbon-based molecule the total
bond energy may be reduced by the formation of ‘sp’ hybrid states between the
filled 2s- and partially filled 2 p-shells. Three primary hybrid states are possible:
(i) tetrahedrally aligned sp3 orbitals as found in diamond and saturated polymers;
(ii) planar hexagonal sp2 orbitals as found in graphite and all organic semiconduc-
tors; and (iii) linear sp orbitals as found in linear acetylenic carbon (carbyne) —
a highly reactive allotrope which is usually synthesised in short lengths capped
with other organic groups. Figure 8.1 shows schematics of these configurations.
Since only three of the four valence electrons hybridise in the sp2 configuration,
the final electron occupies a perpendicular 2 pz orbital. The sharing of sp orbitals
between adjacent carbons leads to strong, highly directional ‘σ -bonding’, which
allows the formation of the stable carbon chains, sheets and branched structures
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 341

Organic Photovoltaics 341

Figure 8.1 Schematics of the electronic orbitals of a) sp3 hybridisation; b) sp2 hybridis-
ation; and c) sp hybridisation, with examples showing the a) diamond, b) graphene and c)
heptayne, octayne and dodecayne linear acetylenic carbon structures, with various capping
groups.

which form the basis of the rich field of synthetic organic chemistry, examples of
which are shown in Fig. 8.2.
Hybridisation between the remaining 2 pz orbitals gives rise to weaker π
(bonding) or π ∗ (anti-bonding) orbitals, the former arising from a symmetric
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 342

342 D. Credgington

Figure 8.2 Chemical structures for a range of commonly used small-molecule and poly-
mer semiconductors. CuPc = Copper phthalocyanine; SubPc = boron subphthalocyanine
chloride.

superposition of 2pz orbitals, the latter from an anti-symmetric superposition.

Such hybridisation typically encourages a preferred rotational orientation between
the main π-bonded units, which induces planarisation of the molecule and
consequently allows delocalisation of π/π ∗ -states over many carbon atoms — a
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 343

Organic Photovoltaics 343

process known as conjugation. In the symmetric (π) configuration, significant

electron density is present between neighbouring carbon atoms, stabilising this
configuration compared with the anti-symmetric (π ∗ ) one. In a simple two-carbon
system, the two unpaired 2 pz electrons will therefore tend to occupy the two spin-
allowed π states, leaving the unoccupied π ∗ states above a forbidden energy gap
E g . Increasing the conjugation length has the effect of broadening the ensemble of
hybrid states into a quasi-continuous ‘band’ and narrowing this gap: for example,
controlling the degree of polymerisation of ethylene can vary E g from 6.7 eV for
the monomer to around 1.5 eV for stretch-oriented polyacetylene (with effective
conjugation over more than 20 carbon atoms) as illustrated in Fig. 8.3 (Townsend
et al., 1985). In the solid state, a similar effect is observed when the π orbitals
of neighbouring molecules overlap to allow delocalisation over several molecular

Figure 8.3 Schematic of energy-level splitting and absorption in alkenes with increas-
ing conjugation length, highlighting the lowest-energy optical transitions from HOMO to
LUMO. The bandgap of polyacetylene depends on both the number of monomers n and the
effective conjugation length in the polymer. Arrows represent spin-paired electrons.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 344

344 D. Credgington

layers. Such ‘π −π stacking’ is important for transport (see Section 8.4.3) and also
leads to a lower optical bandgap. A combination of intra- and inter-molecular delo-
calisation can therefore lead to molecules with absorption in the visible spectral
range.
Since thermal excitation from π to π* is negligible for visible-light transitions,
such materials behave as intrinsic semiconductors with the ‘bandgap’ defined by
E g , the difference between the highest occupied molecular orbital (HOMO) and
lowest unoccupied molecular orbital (LUMO). Doping of the molecule, or injec-
tion of external charge, is therefore necessary to initiate and exploit conduction.
In addition, each conjugated region will not necessarily extend over the whole
molecule (particularly in the case of larger polymers), limiting the practical extent
of delocalisation. However, as we shall see in Section 8.4.4, numerous alternative
routes exist for engineering the HOMO–LUMO offset, and a key advantage of these
materials is that E g is a molecular property, so may be defined before the material
is incorporated into a photovoltaic device. In addition, most organic molecules
are direct-bandgap materials with low symmetry, and thus exhibit extremely high
absorption coefficients — a film of organic material 100–200 nm thick is usually
sufficient to achieve complete absorption at the primary π − π* transition. How-
ever, the molecular (rather than band-like) nature of absorption usually means that
the optical density is peaked rather than step-like at this transition — ideal for dyes,
but not for photovoltaic absorbers.

8.2.2 Excited states

Optical excitation in conjugated organic molecules occurs by absorption of light
with photon energy greater than E g , promoting an electron from the ground state
to the vibronic manifold of the first electronic excited state, and leaving behind a
hole. In general, this electron-hole pair will remain coulombically bound, forming
a neutral molecular excited state referred to as an exciton. The behaviour of excited
states in organic semiconductors is distinct from that of inorganic materials, for
the reasons outlined below.

• Organic semiconductors are low dielectric constant () materials

For example, pentacene ∼ 4, PPV ∼ 2 (Martens et al., 1999) whereas Si ∼ 12,
GaAs ∼ 13. In materials with high , dielectric screening reduces the effective
Coulomb attraction of the excited electron-hole pair, leading to very low binding
energies and characteristic radii greater than the material’s lattice spacing — i.e.
electrons and holes are delocalised over many unit cells. This subset of excitons,
described by and named after Gregory Wannier (Wannier, 1937), are characterised
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 345

Organic Photovoltaics 345

by binding energies of a few kT at room temperature, and thus may undergo spon-
taneous dissociation into free charges via interactions with lattice phonons.
In materials with low , the coulomb attraction between the electron and hole is
not strongly screened, leading to tightly bound Frenkel-type excitons (named after
Jakov Frenkel [Frenkel, 1931]). These have binding energies of order 0.1–1 eV
and characteristic radii of the same order as the unit cell (for crystals) or individual
monomers/molecular units (for molecular materials). Dissociation of these excitons
into free charge carriers is not typically spontaneous at room temperature, and
must be achieved by other means, which we discuss in Section 8.3.1. If they do not
dissociate, excitons in organic materials typically exhibit luminescence lifetimes
of a few nanoseconds. Since one of the most obvious effects of encouraging exciton
dissociation is to quench this luminescence, the term ‘exciton quenching’ is often
used to describe processes leading to exciton dissociation, whether or not the
material in question is intrinsically luminescent.
• Organic materials are electronically ‘soft’

Charge carriers moving through an organic material are usually localised to a few
carbon atoms, and so act as large perturbations to the local electronic structure.
Polarisation of the local medium (enhanced by low ) and the breaking of local
conjugation by the occupancy of the π* anti-bonding state leads to geometric relax-
ation of the molecule. The fundamental charge carrier in organic materials is thus
a quasi-particle, comprising the charge self-localised by its associated molecular
distortion: this entity is called a polaron. The ‘binding’ energy of the polaron (as
compared with the energy of a free charge without distortion) is estimated to be of
order 100 meV, and is determined by the competition between the energy gained
by relaxation of the surrounding medium around the charged excitation and the
energy cost of decreasing delocalisation. A consequence of this self-localisation
is that the migration of polarons between polymer chains or conjugated regions
involves significant reorganisation energy, and is limited by the electronic coupling
between conjugated regions. This relaxation also produces a significant Stokes
shift between absorption and emission spectra, which is particularly beneficial for
reducing self-absorption in organic light-emitting diodes (LEDs), but represents
a potential loss of free energy in a photovoltaic. The extent of relaxation is gov-
erned by the physical properties of the particular molecule; thus physically ‘stiff’
materials tend to exhibit lower polaron binding energies and faster charge transport.

• Transport of carriers is activation-limited

The polaronic nature of charge carriers leads to self-localisation, and to move from
one conjugated molecule (or conjugated segment) to another therefore requires
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 346

346 D. Credgington

additional thermal energy. This energy is associated with the geometric relaxation
of the ‘acceptor’ molecule combined with the geometric rebound of the ‘donor’
molecule. In addition, the probability for a carrier to transfer is also dependent on the
wave-function overlap between donor and acceptor states. Since organic molecules
(and particularly polymers) typically exhibit low symmetry, transport can be highly
anisotropic. For a stiff conjugated polymer, intra-chain transport will be faster than
transport in the π − π stacking direction, while lateral transport through saturated
side groups can be very poor. For example, intra-chain hole mobilities as high as
600 cm2 /Vs have been reported for ladder-type poly( p-phenylene) (Prins et al.,
2006), whereas measured mobilities are of order 1–10 cm2/Vs (Li et al., 2012)
in solid films, and depend strongly on polymer orientation (Lee et al., 2011).
Such transport also depends on temperature and electronic disorder, which we will
discuss in more detail in Section 8.4.2.
• Transport of excitons is disorder-limited
The migration of excitons through the semiconductor occurs via transfer of the
bound polaron pair from one conjugated molecule/chain segment to another
through either Förster or Dexter energy transfer. Förster transfer (Förster, 1948)
involves short-range dipole–dipole interactions equivalent to the exchange of a
virtual photon; Dexter transfer (Dexter, 1953) involves direct exchange of elec-
trons. Requiring only coupling between dipoles, Förster transfer of singlet excitons
is the dominant process in photovoltaics and is typically restricted to a range of
5–20 nm over the lifetime of the exciton (Halls et al., 1996; Stubinger and Brutting,
2001), depending on the material in question. This range is limited by the degree
of overlap between the material’s absorption and emission spectra (reduced by
polaronic relaxation), and by disorder in the energetic landscape accessible to the
exciton. Since Dexter transfer requires wave-function overlap between the donat-
ing and accepting species, it is usually dominant only in situations where optical
transitions are forbidden, for example in the migration of spin-triplet excitons.

8.2.3 Thin-film processing

Small organic molecules may be sublimed under vacuum at relatively low temper-
ature, enabling the creation of well-defined films and layer structures without the
need for high-temperature, high-purity processing. However, the scalability and
cost of such an approach must be balanced against its benefits, as we outline in
Section 8.5.6. An alternative approach is possible since organic semiconductors are
usually synthesised using wet chemistry from soluble precursors. Solubility can
therefore be specifically designed into the molecule without significantly affect-
ing its electronic structure (as illustrated in Fig. 8.4). Since conjugated molecules
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 347

Organic Photovoltaics 347

Figure 8.4 Typical routes to solubilisation. Insoluble poly( p-phenylenevinylene) (PPV) is

formed in situ from a soluble precursor via a leaving-group reaction. Insoluble polythio-
phene and pentacene are rendered soluble by the inclusion of flexible side chains.

are usually quite stiff due to the planarising effect of their π-bonding, solubilisa-
tion is usually achieved by attaching flexible saturated side groups along the main
backbone, whose chemistry determines the range of compatible solvents. For sim-
plicity, we will here refer to both the conjugated segments of a polymer chain, and
the conjugated carbon scaffold of organic small molecules as the ‘backbone’ of
the structure.
Many organic molecules may therefore be formed into inks and processed from
solution. This approach is of particular importance for depositing large semicon-
ducting molecules such as conjugated polymers, for which vacuum sublimation is
impossible. As solid films, these polymeric semiconductors are flexible and highly
resistant to fracture, lending the collection of technologies which exploit them the
umbrella term ‘plastic electronics’.
In solution, the freedom of organic molecules to explore a range of config-
urations is enhanced by the presence of the solvent. Deposition from solution is
therefore thermodynamically analogous to a rapid quench, as the evaporation of
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 348

348 D. Credgington

solvent molecules quickly restricts this freedom. Depending on the details of the
deposition (and particularly on the evaporation rate of the solvent), the resulting
material is typically left in a metastable configuration, with a mix of glassy and
more ordered regions. This effect is even more prevalent in systems comprising
mixtures of multiple materials. Since electronic structure is influenced by both
local conjugation length and intermolecular interactions, solution deposition leads
unavoidably to significant spatial and energetic disorder. This intrinsic disorder has
implications for many photovoltaic properties, as we discuss in Section 8.4.2.
Deposition of organic semiconductors from solution can take advantage of
the wide range of thin-film coating technologies already employed by existing
industries. These include inkjet and screen-printing for individual substrates and
extend to continuous roll-to-roll manufacture, for example using slot-die, gravure
printing or doctor-blading. The existence of mature, low-cost, large-area manu-
facturing routes means that scaling-up of organic semiconductor technology is
widely expected to be a relatively cheap process, if the materials set can also reach
maturity. We discuss the progress made to date in large-area OPV manufacture in
Section 8.5.6.

8.2.4 Device structure

The short optical absorption lengths (∼100–200 nm) associated with organic semi-
conductors, combined with the ease with which they may be deposited over large
areas, renders them highly suited for use in thin-film devices. When employed
as photovoltaics, the conventional arrangement is a ‘sandwich’ structure with the
photoactive layer deposited on top of one electrode, and capped with a second.
The low carrier mobilities typical of organic molecules generally preclude the use
of laterally separated electrodes in efficient devices. To allow light to enter the
absorbing layer, at least one electrode must be optically transparent. Indium tin
oxide (ITO) is commonly used in this case, and is usually deposited as the ‘bot-
tom’ (though sun-facing in the completed device) electrode of the sandwich onto
a glass or plastic substrate. It is also common to choose a highly reflective metal
such as Al or Ag as the ‘top’ electrode to allow transmitted photons a second pass
through the device. It has become conventional to refer to as ‘normal’ or ‘non-
inverted’ a device where the transparent electrode is primarily used to collect hole
current and the reflective electrode used to collect electron current. Where this
assignment is reversed, the device is referred to as ‘inverted’. While the inverted
structure is a relatively recent development, it is generally thought that equivalent
efficiencies are achievable in both arrangements, but that inverted structures may
allow for more stable devices (see Section 8.5.1). Figure 8.5 shows conventional
and inverted device structures.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 349

Organic Photovoltaics 349

Figure 8.5 Conventional (a) and inverted (b) device structures, indicating stack order and
interlayer positions.

Figure 8.6 Poly(3,4-ethylenedioxythiophene) and poly(styrenesulphonate) — synthesised

as a water-based colloidal suspension to form PEDOT:PSS.

8.2.5 Interface control

Whichever the orientation, the device electrodes have to be made selective for
the appropriate carrier, or significant current may be lost as carriers diffuse to
the ‘wrong’ electrode. For silicon solar cells, selectivity is imposed by heavy
n- or p-doping close to the contact. Since direct doping of organic semicon-
ductors remains difficult, particularly from solution, introducing such selectiv-
ity is usually encouraged by introducing thin interfacial layers (‘interlayers’)
between the electrode and the photoactive layer. Interlayers may comprise both
organic and inorganic materials, and may fulfil additional functions such as pre-
venting exciton quenching at the electrodes and modifying the electrode work
function (see Section 8.3.2). The most common interlayer material is the water-
soluble composite poly(3,4-ethylenedioxythiophene): poly(styrenesulphonate)
(PEDOT:PSS), a mixture of hole-conducting PEDOT doped by PSS, illustrated
in Fig. 8.6.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 350

350 D. Credgington

8.3 Overview of organic photovoltaic devices

Organic photovoltaics aim to take advantage of the properties of organic semicon-
ductors described above, while mitigating their pitfalls. In particular, the aim is to
use the flexibility of organic chemistry to design direct-bandgap semiconductors
with absorption tuned to closely match the solar spectrum. By choosing solu-
ble organic molecules, large-area, low-temperature production techniques become
available, and by utilising materials with high absorption coefficients, materials
usage and module weight may be significantly reduced. Combined, these factors
have the potential to considerably reduce the cost of photovoltaic power generation.
However, a number of challenges must be overcome to achieve these goals.
Photogeneration of dissociated charge carriers from excitons is not spontaneous,
transport of dissociated carriers is characterised by low mobilities as a result of
polaronic self-localisation and electronic disorder, individual materials typically
exhibit narrow absorption features, and the molecular origin of their semiconduct-
ing properties means that performance is sensitive to chemical degradation.
In the following sections, we briefly illustrate how these issues have been
addressed and give an overview of the state of the art, starting with the process of
exciton dissociation, which represented the most significant obstacle to the field.

8.3.1 Exciton dissociation

Homogeneous devices
The first classes of organic molecules to be considered seriously as photovoltaic
materials were the porphyrins and phthalocyanines (shown in Fig. 8.2). Chloro-
phyll, the familiar green pigment responsible for the collection of solar energy in
plants, is a dihydroporphyrin (a chlorin), so the prospect of emulating the highly
evolved natural process of photosynthesis in an artificial solar cell inspired consid-
erable investigation. The use of thin films of chlorophyll (Meilanov et al., 1970a;
Meilanov et al., 1970b; Tang and Albrecht, 1975) and related ‘natural’ molecules
(Delacote et al., 1964; Kearns and Calvin, 1958) in photovoltaic cells was ini-
tially disappointing, with external quantum efficiencies (EQEs) less than 0.1%,
and power conversion efficiencies (PCEs) throughout the 1960s and 1970s gen-
erally less than 0.01%. Parallel investigation on ‘artificial’ semiconductors such
as anthracene (Kallmann and Pope, 1959) and pentacene (Silinsh et al., 1974)
yielded similar results. Several useful concepts were introduced, the most impor-
tant being the use of electrodes with dissimilar work functions to break the sym-
metry of the device. This asymmetry generates a ‘built-in’ electric field which
drives oppositely charged carriers in opposite directions, and allows photovoltages
in the 200–500 mV range to be developed. A sufficiently high internal field could
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 351

Organic Photovoltaics 351

also be used directly to split excitons, which was found to be consistent with the
electric-field-induced ionisation mechanism proposed by Onsager, and later refined
by Braun (Braun, 1984; Onsager, 1934). However, the primary limit on efficiency
remained the competition between the short exciton lifetime (a few ns) and the low
rate of spontaneous exciton dissociation under normal operating conditions.

Planar heterojunctions
A number of pieces of evidence pointed to a possible solution. In single-material
devices, dissociation was found to be more likely at surfaces or impurities, and in
the presence of water or oxygen. It was also discovered that electron transfer from
organic pigments to other semiconductors — such as from chlorophyll to ZnO —
could be efficient (Tributsch and Calvin, 1971). The implication of these obser-
vations was that exciton dissociation could be encouraged by allowing charges
to transfer to a nearby heterogeneous molecule. This was confirmed by Tang and
co-workers (Tang, 1986), who showed that the photocurrent of an organic photo-
voltaic could be increased by orders of magnitude if two different semiconducting
molecules were combined to form a bilayer, or ‘planar’ heterojunction (shown in
Fig. 8.7a). Rather than relying on thermal energy or an electric field to ionise the
excitons, this structure drives dissociation by using the gain in free energy associ-
ated with the transfer of an electron from the LUMO of the ‘donor’ semiconductor
to the LUMO of the ‘acceptor’ semiconductor (or the transfer of a hole from accep-
tor HOMO to donor HOMO). In the language of crystalline semiconductors, a type
II heterojunction is required. Figure 8.8 illustrates this process.
To a first approximation, the band-edge offset must be greater than or equal to
the exciton binding energy, a point we shall discuss in more detail in Section 8.4.4.

Figure 8.7 Heterojunction structures, showing schematics of: a) planar heterojunction; and
b) bulk heterojunction photoactive layers in a conventional structure.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 352

352 D. Credgington

Figure 8.8 Photogeneration in a planar heterojunction device, illustrating: a) exciton for-

mation and diffusion in the donor; b) formation and dissociation of a ‘charge-transfer’ state
across the heterojunction; c) collection of dissociated charges at the electrodes. Solid lines
represent single-carrier molecular energy levels, dotted lines represent (induced) excitonic
energy levels, and dashed lines indicate a coulombically bound state. Interlayer energy levels
have been chosen so as to be selective for electrons or holes and the LUMO:LUMO offset
is highlighted.

By careful choice of materials — CuPc and the perylene derivative ‘PV’ in this case
(shown in Fig. 8.2) — devices with PCEs of around 1% were realised. In addition,
the use of two semiconductors opened the possibility of combining materials with
complementary absorption spectra in a single device.

Bulk heterojunctions
While a significant advance, the EQE of planar heterojunction devices peaked at
around 15%, since only those excitons able to diffuse to the heterojunction could
be dissociated. With the diffusion length of singlet excitons limited to approxi-
mately 10 nm, and absorption lengths of order 100 nm, we must often accept either
incomplete light harvesting or incomplete exciton harvesting when using such a
bilayer configuration.
This problem was solved in the mid-1990s with the introduction of the ‘bulk’
heterojunction (BHJ) structure (Halls et al., 1995; Yu et al., 1995), the philosophy
of which is to ‘bring the heterojunction to the exciton’. By blending the donor and
acceptor materials together in a single layer, it is possible to produce a distributed
heterojunction extending throughout the device, such that every newly generated
exciton is within one diffusion length of an interface (Fig. 8.7b).
The discovery of this approach coincided with the development of highly
effective soluble fullerene derivatives (small molecules based on buckyballs) as
acceptors (Sariciftci et al., 1992). These favoured the use of organic polymers
as donors, since the entangled polymer chains can prevent regions of the film
becoming electrically isolated, and can also encourage de-mixing (discussed in
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 353

Organic Photovoltaics 353

Section 8.4.1). By combining the polymer poly[2-methoxy-5-(2-ethyl-hexyloxy)-

1,4-phenylene vinylene] (MEH-PPV) and C60 , PCEs of nearly 3% were realised,
with peak EQE of 30% (Yu et al., 1995). The recent dominance of the organic
photovoltaics field by polymer:fullerene devices largely stems from this result. It
was subsequently discovered that blends of the fullerene derivative [6,6]-phenyl-
C61 -butyric acid methyl ester (PC60BM) with the soluble semicrystalline thiophene
donor poly(3-hexylthiophene) (P3HT) could achieve PCEs approaching 5% after
a low temperature (140◦C) anneal (Ma et al., 2005; Padinger et al., 2003). Such
high efficiencies, particularly compared with those achieved in previous decades,
suggested that commercial exploitation of OPV was a genuine possibility, and
confirmed fullerenes as the acceptor of choice. Since fullerenes are usually weakly
absorbing owing to their spherical symmetry, it has also become common to focus
on the donor as the primary absorber, and the LUMO–LUMO offset as the main
driving force for charge separation.
The BHJ structure has disadvantages, however. While continuous single-
component pathways from the interface to the electrodes are always present in a
bilayer heterojunction, in a BHJ this is not guaranteed, meaning isolated ‘islands’
of donor or acceptor may form. In addition, both electrons and holes remain in
close proximity to the heterojunction as they travel to the electrodes, allowing
dissociated carriers to recombine across an interface. The ideal bulk heterojunc-
tion is therefore likely to be an interpenetrating network of de-mixed donor and
acceptor phases on a number of length scales, but since microstructure forms spon-
taneously during deposition this is very difficult to control. In addition, since a
solution-cast microstructure is usually metastable, its long-term stability is always
of concern. Direct nanoscale patterning of the semiconducting materials could
solve this issue, but to date no low-cost scalable technology has achieved com-
parable results to simple thermodynamically mediated self-assembly. Because of
its importance, influencing BHJ microstructure by controlling the thermodynam-
ics of film formation still remains at the forefront of device engineering after
nearly 20 years (Campoy-Quiles et al., 2008; Chirvase et al., 2004; Lee et al.,
2008; Li et al., 2005; Moulé and Meerholz, 2008; Nguyen et al., 2007; Peet et al.,
2007; Shaheen et al., 2001; Xin et al., 2010). We discuss this topic further in
Section 8.4.1.
Despite the inherent complexity, the advantages of the BHJ continue to out-
weigh the disadvantages. By utilising absorption from low-bandgap polymers and
asymmetric (and so more strongly absorbing) fullerenes, device currents in poly-
mer:fullerene cells can exceed 15 mA cm−2 under solar illumination, e.g. utilising
PDPP-TTT and PTB7 (Bronstein et al., 2011; He et al., 2012), and internal quan-
tum efficiencies can reach nearly 100%, e.g. utilising PCDTBT (Park et al., 2009),
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 354

354 D. Credgington

the structures of which are shown in Fig. 8.2. This means that the technology
has advanced from quantum yields of 10−3 to a point where nearly every exciton
can reach the heterointerface and dissociate, and under short-circuit conditions the
resulting carriers escape the device without loss.

8.3.2 Photovoltage
For any photovoltaic material, the absorption of photons with energy greater than
the semiconductor bandgap leads to thermalisation losses as electrons/holes relax to
the conduction/valence band edges (or LUMO/HOMO in this case). The maximum
free energy difference per carrier that remains after transport to the electrodes
defines the open-circuit voltage (Voc) of the cell. This is generally limited by either
the electrodes or the materials, as we now discuss.

Electrode-limited regime
After the development of the bulk heterojunction solved the problem of exciton
dissociation, several authors observed that a correlation existed between Voc and
the work function difference of the cell’s electrodes,  (Brabec et al., 2002;
Mihailetchi et al., 2003; Mihailetchi et al., 2004). The implication was that the
energy difference between the donor HOMO and acceptor LUMO (often referred
to as the heterojunction bandgap, E HJ ) is reduced to  when the carrier trans-
fers to the electrodes. This picture became known as the metal–insulator–metal
(MIM) model, and drove the development of materials to achieve increasingly
large . It is now common to use interlayers comprising soluble conducting
polymers (Aernouts et al., 2004), p-type oxides (Irwin et al., 2008; Zilberberg
et al., 2012) and surface treatments to increase the work function of the hole-
collecting electrode, and to use salts (Brabec et al., 2002), polymers (Zhou et al.,
2012), n-type oxides (Kim et al., 2006) and low work-function metals as inter-
layers to decrease the work function of the electron-collecting electrode. These
layers can provide an additional benefit if they act to increase the ‘selectivity’ of
the appropriate contact — i.e. to block the transport of electrons diffusing to the
hole-collecting contact, and vice versa.

Materials-limited regime
The drive to increase , and the development of lower-bandgap absorbers, has
revealed that where  is greater than E HJ , it is the latter that limits the available
free energy (see Fig. 8.9a). Outside the MIM regime, applicable where  < EHJ ,
numerous studies have shown correlations between the achievable Voc and E HJ , the
most well-known being the study by Scharber and co-workers (Scharber et al.,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 355

Organic Photovoltaics 355

Figure 8.9 a) Dependence of the work function of polymer-coated substrate (ELP/SUB)

on the work function of bare substrate (SUB ), for four polymers: P3HT, TFB, P10AF,
and PFO. The transition from electrode-limited (1:1 linear correlation) to materials-limited
(no dependence on SUB ) performance is clear. Reproduced from Tengstedt et al. (2006).
b) Experimentally observed correlation between Voc of different bulk-heterojunction solar
cells plotted versus the oxidation potential/HOMO position for the donor polymer used in
each device. The solid line is a linear fit with slope of one. Reproduced from Scharber et al.
(2006).

2006) reproduced in Fig. 8.9b. The correlation is not exact — in all cases the
measured difference between donor HOMO and acceptor LUMO overestimates
the achievable Voc — but it provides an excellent design rule. In Section 8.4.5 we
discuss in more detail the origin of this discrepancy, and how the switch between
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 356

356 D. Credgington

electrode-limited and materials-limited performance may be interpreted. For now,

we note that maximising E HJ has allowed Voc in excess of 1 V to be achieved in
single-layer devices (Najari et al., 2012), while cells exhibiting efficient photogen-
eration can achieve 0.9 V (Park et al., 2009).

8.3.3 Maximum efficiency

Since absorber bandgap, LUMO–LUMO (or HOMO–HOMO) offset and
heterojunction bandgap are interdependent, it is not possible to simultaneously
optimise energetic structure for high Voc, broad absorption and efficient exciton dis-
sociation. Numerous studies exist which aim derive an equivalent of the Shockley–
Queisser limit for the maximum PCE of a practical organic photovoltaic, balancing
carrier generation with free energy loss and helping to guide the synthesis of new
molecules with more favourable energetics. Estimates for maximum achievable
PCE range from 15–20% (Dennler et al., 2009; Giebink et al., 2011), though tend
to be revised upward year on year as limiting processes are circumvented or better
understood. Such calculations are typically based on the physical efficiency lim-
its described in Chapter 2 of this volume, combined with additional constraints
such as the free energy lost during exciton dissociation and the finite bandwidth of
molecular absorbers (Deibel and Dyakonov, 2010).

8.3.4 State of the art

The rapid progress of materials design and solar cell performance over the last
decade means that any examples of ‘top’ efficiencies described herein will rapidly
become out of date. However, it is instructive to compare the relative performance of
different cell types and provide a snapshot of the field. Until recently, most organic
photovoltaics were produced over small areas in university research laboratories,
with relative flexibility regarding testing conditions. Robust certification of large-
area (>1 cm2 ) devices under standard temperature, humidity and spectrum (Reese
et al., 2011) is therefore still rare for all but the most eye-catching ‘headline’
cells. Because of this, throughout this chapter we make no distinction between
self-reported and certified performance, nor do we exclude reports on small-area
(<1 cm2 ) devices. The reader should therefore not give undue weight to incremental
differences between performance metrics, which may arise simply from slightly
different testing conditions. Figure 8.10 shows the progress in certified solar cell
efficiency since 2000, collated by the US National Renewable Energy Laboratory
(NREL). At the time of writing (July 2013), the world record was held by Heliatek’s
tandem evaporated small molecule OPV (Heliatek, 2013).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 357

Organic Photovoltaics 357

UCLA-Sumitomo
12 Single Junction Mitsubishi Chemical
Tandem
Solarmer
Heliatek
10
Certified PCE (%)

Konarka Heliatek

8 Plextronics
University Groningen UCLA
Siemens Konarka
6
Sumitomo Chemical

4 Heliatek
NREL/Konarka

2 University Dresden

University Linz

0
2000 2004 2008 2012
Year

Figure 8.10 Progress of certified organic photovoltaic efficiencies over time since 2000,
including both single-junction and tandem devices. Labels indicate institution/company of
origin. Data from Green et al. (2013).

Bilayer heterojunctions
Bilayers are mainly of interest for studying interface physics in a well-defined
system, meaning that thermally evaporated small molecules are particularly well
suited to this architecture. Solution-processed bilayers require either strict solvent
orthogonality (i.e. two materials with mutually exclusive solubility) or a means
of rendering one layer insoluble before deposition of the next — for instance by
cross-linking (Png et al., 2010) or removal of solubilising groups (Bradley, 1987).
Failure to do so risks rapid mixing of the bilayer components. Even where this
is initially prevented, inter-diffusion of materials over time is often unavoidable
(Treat et al., 2011).
The best small-molecule absorbers remain the phthalocyanine and sub-
phthalocyanine derivatives (see Fig. 8.2), which achieve good light harvesting
in exceptionally thin (∼10–15 nm) films. When combined with the acceptor C60 ,
PCEs of order 3–4% are achievable (Lin et al., 2012; Peumans and Forrest, 2001).

Small-molecule bulk heterojunctions

Small organic molecules may be processed from solution, or co-evaporated under
vacuum, to produce a mixed film. However, the as-deposited film is typically highly
amorphous owing to the ease of intermixing, and requires further treatment to
encourage the coarsening of donor and acceptor domains through crystallisation.
The highest efficiency for small-molecule cells with published structures has been
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 358

358 D. Credgington

achieved using a solution-processed oligothiophene donor with a fullerene accep-

tor, reaching 6.7% PCE (Sun et al., 2012). However, it is important to point out
the exceptional performance of small-molecule devices reported by developers in
the commercial sector, detailed below.

All-polymer bulk heterojunctions

Blends of conjugated polymers form the archetypical bulk heterojunction: two
de-mixed interpenetrating networks of relatively pure material, with domains eas-
ily distinguishable using high-resolution microscopy. We consider the mechanism
for this de-mixing in more detail in Section 8.4.1. The development of poly-
mer:polymer blends has historically been hampered by the difficulty of produc-
ing acceptor polymers with high electron mobility, leading to excitons remaining
trapped at the heterointerface. Recent development of new high-mobility poly-
mer acceptors containing naphthalene diimide electron-accepting units have led
to significant improvements in the current generation of devices, with efficiencies
increasing from 1–2% up to 6.4% (Polyera, 2013).

Polymer:small-molecule bulk heterojunctions

Since the discovery of the efficient P3HT:PCBM system, the development of new
conjugated polymer donors in combination with soluble fullerene acceptors has
been the main focus of the organic photovoltaics field, while the optoelectronic and
thermodynamic complexity of such systems has been responsible for the majority
of organic semiconductor research output. The most efficient systems to date are
also the most efficient lab-scale organic devices yet reported, with PCE now exceed-
ing 9% for single-junction devices comprising alternating ‘push-pull’ copolymers
with PC70 BM (He et al., 2012), and approaching 9.5% for multijunction devices
(Li et al., 2013).

Commercial developers
For the last two decades, research into new materials and device optimisation has
been undertaken mainly within university research laboratories. It is worth noting
that within the last few years, this trend has changed: the most efficient organic
photovoltaics now reported are produced almost exclusively by commercial devel-
opers. As noted in connection with Fig. 8.10, the record for any OPV device cur-
rently lies with Heliatek (2013) and their 12%-efficient evaporated small-molecule
tandem solar cell.
Mitsubishi Chemical have also achieved certified PCEs of 11% with a
solution-processed small-molecule heterojunction, which is reported to be a
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 359

Organic Photovoltaics 359

single-junction device (Green et al., 2013). However, in all cases the materials
and device structure are not known. It is therefore not clear in which direction the
future of the OPV field lies, but it is certain that the materials on which it is based
are still far from maturity.

8.4 Device physics of organic photovoltaics

In the following sections, we will revisit the key operating principles of organic
photovoltaics and give an overview of the current state of understanding, begin-
ning with the process of thin-film formation, then focussing on the links between
emergent device properties and molecular/morphological design.

8.4.1 Morphology of the photoactive layer

Thermodynamics and self-assembly
Many of the organic photovoltaic devices described throughout this chapter rely on
a two-phase blend of semiconducting polymers or small molecules to form their
photoactive layer. Homogeneous mixtures are also common, and may likewise
display useful properties absent in the component parts. PEDOT:PSS is one such
example, the two component polymers exchanging charge to create a soluble,
conductive composite.
The mixing process for dissimilar materials can be understood within a frame-
work set out in two papers published simultaneously by Flory (1941) and Huggins
(1941). The Flory–Huggins approach gives a basic understanding of the possible
behaviours and phases in a two-component blend by modifying the expression for
the Gibbs free energy of a mixture to include molecules of different molecular
weight.
The Gibbs free energy of a particular system is given by the difference between
its enthalpy, H , and TS, a measure of the system’s disorder, where T is temperature
and S the entropy of the system:

G = H − TS (8.1)

This quantity estimates the energy currently available to a system to do work,

or to allow transitions to other states. The change in free energy, G, associated
with moving from one state to another is therefore an indication of the nature of the
transition: only transitions for which G < 0 are thermodynamically favourable.
For a mixture, an initial state of a pure A phase and a pure B phase and a final state
of a mixed AB phase are considered, with the change in free energy on mixing
given by G mix , which in the Flory–Huggins approximation is considered as a
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 360

360 D. Credgington

sum of two terms:

G mix = G AB − (G A + G B ) = G local − T Strans (8.2)

where G X is the free energy associated with phase X, G local captures the changes
to the local chemistry and motions of the mixed units, including any changes
to the local entropy due to volume changes, and Strans captures the change in
translational entropy of the system. By assuming that G local arises primarily from
enthalpic interactions between a given molecule and the mean field originating from
the rest of the mixture, it may be shown that


φA φB
G mix = RT V χφA φB + ln φA + ln φB (8.3)
NA NB
where V is the volume of the system, R the molar gas constant, φA and φB the
volume fractions of component A and B, and NA and NB their degrees of poly-
merisation. χ is the dimensionless ‘Flory–Huggins’ parameter, which represents
the change in local free energy per reference unit (monomer, in this case). The
first term represents local enthalpy change, while the second and third terms repre-
sent entropy change and will be negative. Homogeneous mixing will occur only if
G mix < 0, so compatible molecules (χ < 0) will always mix, but incompatible
molecules (χ > 0) will only do so for sufficiently small N. This implies that a
homogeneous mix of dissimilar polymers can usually lower its free energy via
spinodal decomposition into a phase-separated morphology, with little entropic
penalty. Figure 8.11a shows an example of such a morphology.
Real blend systems are always more complicated, as illustrated in Fig. 8.11c,
in particular because it has been observed empirically that at least one semicrys-
talline component is required for a device to be efficient. If phase separation can
also be driven by crystallisation, then even small molecules may spontaneously
de-mix, given sufficient thermodynamic freedom. For example, the small size of
fullerene acceptors means they are usually miscible in amorphous conjugated poly-
mers (or within the amorphous fraction of semicrystalline polymers) up to around
50% by volume, as might be expected from a simple Flory–Huggins description.
Beyond this miscibility threshold, purer fullerene domains develop which may
subsequently crystallise, adding significant complexity to the system.
For a two-component semicrystalline blend at least five phases are therefore
expected: two pure phases per component, one crystalline, one amorphous, plus a
mixed amorphous phase. The widely studied P3HT:PCBM system exhibits such
behaviour, with the final phase structure and domain size depending crucially on
the thermodynamic trajectory. The somewhat unusual system of poly(2,5-bis(3-
hexadecylthiophen-2-yl)thieno[3,2-b]thiophene) (PBTTT):PCBM adds a sixth
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 361

Organic Photovoltaics 361

Figure 8.11 a) Topography (vertical range 81 nm) of a (1:1) PFO/F8BT thin film; b) F8BT
fluorescence (arbitrary scale) from the same region of the blend, illustrating the classic
phase-separated structure of a de-mixed polymer:polymer blend. Adapted from Chappell
et al. (2003). c) Sketch illustrating the much finer microstructure formed by the blend of
semicrystalline P3HT (lines) and PCBM (dots) with increasing fullerene loading: (i) with
no PCBM; (ii) with PCBM dissolved in the amorphous fraction of the P3HT film without
disrupting the typical crystallite spacing (L P , of order 10 nm); (iii) with sufficient PCBM
inclusion to swell the amorphous phase and increase L P ; (iv) saturation of amorphous-
phase swelling and the onset of PCBM aggregation between P3HT crystallites. Adapted
from Kohn et al. (2013).

phase, a co-crystal of PBTTT and PCBM, which is strongly implicated in exci-

ton dissociation (see Fig. 8.12). The result in both cases is a hierarchical bulk
heterojunction, with highly intermixed regions in quasi-equilibrium with purer
phases and crystalline regions, the effects of which are discussed below.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 362

362 D. Credgington

Figure 8.12 a) Phase diagram of the pBTTT/PCBM/co-crystal system, showing two eutec-
tics: at ∼1 wt% PCBM (pBTTT/co-crystal binary) and 43 wt% PCBM (co-crystal/PCBM
binary; b Photoluminescence quenching (PLQ) is correlated with the onset of co-crystal for-
mation beyond the first eutectic, as indicated by Ht(co−crystal), whereas long-lived charge
generation (OD) and collection (JSC) are correlated with the formation of crystalline
PCBM domains beyond the second eutectic. Reprinted from Jamieson et al. (2012).

Ensuring a bulk heterojunction with beneficial morphology develops

spontaneously following deposition requires a well-controlled thermodynamic tra-
jectory, which in turn requires control of solution composition, temperature and
evaporation rate as well as the interactions between component molecules. Tech-
niques to improve device morphology therefore include:

• Varying the relative solubility of the blended materials — either by altering

the molecules themselves (Park et al., 2006), changing the type or temperature
of the solvent (Liu et al., 2001), or adding co-solvents (Lee et al., 2008).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 363

Organic Photovoltaics 363

• Increasing the thermodynamic freedom of the blend to allow coarsening —

either by annealing above its glass transition temperature (Ma et al., 2005),
by swelling the blend using a solvent vapour (Li et al., 2007), or simply by
restricting the solvent evaporation rate during deposition.
• Altering the interaction between molecules, for instance by encouraging/
discouraging planarisation, enhancing/suppressing stacking or by affecting
nucleation and growth of crystallites (Ballantyne et al., 2007; Kim et al.,
2006).

Top-down patterning
The difficulty of generating a beneficial heterojunction structure using thermo-
dynamics alone has led to interest in enforcing a particular nanostructure on the
blend. To achieve this, a huge variety of techniques have been explored for pat-
terning organic semiconductors (Brédas et al., 2009). These include:

• Techniques which utilise pre-patterned substrates or stamps to guide or define

the final structures, typically over large areas, such as site-selective chemical
vapour deposition on patterned precursors (Cacialli and Bruschi, 1996), micro-
moulding in capillaries (Rogers et al., 1998), micro-contact printing (Rogers
et al., 1999) and nanoimprint lithography (Chou et al., 1996; Dong et al.,
2006).
• Techniques which rely on self-assembly to pre-form or form in situ a beneficial
morphology, for instance using electrodeposition (Yun et al., 2004), electro-
spinning (Kameoka et al., 2004) and all forms of pre-aggregation in solu-
tion, including colloid, nanowhisker or vesicle formation. This may extend
to templating using self-assembled diblock copolymer structures (Chappell
et al., 2003) and inorganic scaffolds (Martens et al., 1999).
• Techniques which pattern directly to an arbitrary design, typically over rela-
tively small areas, including electron-beam lithography (Stabile et al., 2007),
as well as scanning-probe technologies such as dip-pen nanolithography (via
electrostatic or electrochemical deposition), scanning tunnelling microscopy
(Granström, 1998), nanothermal lithography (Fenwick et al., 2009) and scan-
ning near-field optical lithography (Credgington et al., 2010).

Each technique has particular strengths, which may be in terms of flexibility,

resolution, or the amount of damage inflicted on the active polymer, though
most are not easily scalable and few are able to achieve three-dimensional
structures.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 364

364 D. Credgington

The ideal bulk heterojunction

A more fundamental difficulty when employing top-down approaches is that the
ideal bulk heterojunction structure is not currently known. However, it must have
some general properties:

1. A region of pure donor material adjacent to the hole-extracting contact to provide

a barrier to the extraction of electrons.
2. A region of pure acceptor material adjacent to the electron-extracting contact
to provide a barrier to the extraction of holes.
3. Continuous percolation pathways through high-mobility donor/acceptor phases
from every interface to the electrodes.
4. Transport pathways aligned with the internal field, to enhance the contribution
of carrier drift.
5. Interfaces distributed throughout the photoactive layer, such that there is a high
probability of every photogenerated exciton reaching the heterojunction.
6. Thickness greater than the absorption depth of above-bandgap photons

Criteria 1 and 2 above improve the selectivity of the contacts, and reduce the
need for interlayer materials to be carrier-blocking. Criteria 3 and 5 are particularly
difficult to achieve simultaneously, since even in semicrystalline materials excitons
rarely diffuse much further than 10 nm, while criterion 6 requires a film thickness
of ∼200 nm. In principle, these requirements dictate an interdigitated structure of
pure donor and acceptor material with a very high aspect ratio, which has been
attempted by a variety of means using the techniques outlined above.
However, recent improvements in polymer:fullerene solar cell performance
have shown that neither pure donor and acceptor phases, nor well-structured het-
erojunction networks, are necessary for efficient operation. Regions of intermixed
donor and acceptor relax the need for single-phase percolation pathways, and sep-
arate excitons very efficiently. There is also evidence, discussed in Sections 8.4.4
and 8.4.5, that a hierarchical structure may provide an energetic driving force to
move carriers away from the interface. The assumption that a pure two-component
bulk heterojunction is ideal is therefore likely to be too simplistic, and the inclusion
of mixed or amorphous phases may be necessary for efficient operation.

8.4.2 Electronic disorder

One of the benefits of using molecular semiconductors is that removal of defects
is a relatively simple chemical purification process. While separation of very
similar molecules is difficult — for instance C60 and C70 , or polymers with
similar molecular weights — the electronically active backbone of the organic
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 365

Organic Photovoltaics 365

semiconductor is usually defect-free. This feature has several implications:

• Impurity tolerance. Since there is no lattice to disrupt, electronically inactive
impurities have little effect on electronic structure.
• Passive interfaces. Junctions between molecular layers are not associated with
broken bonds or lattice mismatch, and so are not detrimental to performance.
This underpins the ability of organic heterojunctions to function, and allows
relatively facile interface engineering by the insertion of interlayers.
• Both semicrystalline and amorphous materials may be employed, since the
formation of a well-ordered lattice is not necessary.
Organic semiconductors are nevertheless usually highly disordered systems,
despite their low intrinsic molecular disorder. This is because spatial and energetic
disorder can be significant, as we outline below.

Configurational disorder
Although all small organic molecules, or monomers, of a given material are struc-
turally identical, electronic interactions between two neighbouring molecules are
strongly dependent on their relative position (Linares et al., 2010). Since as-
deposited films are usually amorphous, or at most semicrystalline, the relative
position and orientation of neighbouring molecules varies significantly through-
out the film. This spatial disorder leads to highly anisotropic electronic coupling
between molecules, which impacts on charge transport. Similarly, since coupling
strengths between molecules affect the energetics of intermolecular and delocalised
states, this spatial disorder may contribute to energetic disorder, as each molecule
experiences a slightly different local environment.

Conformational disorder
For flexible molecules, variation in conformation also leads to spatial and energetic
disorder. This is particularly applicable to semiconducting polymers, where charge
delocalisation is dependent on conjugation length. Kinks or twists in the polymer
chain will break conjugation, limiting the extent of delocalisation and narrowing
the molecular ‘bands’. Chain-extended polymers are therefore optically distinct
from coiled polymers within the same sample (Clark et al., 2009), meaning that
variations in processing history and polydispersity, as well as random variation in
conjugation length, result in additional energetic disorder.

Dynamic disorder
The mechanisms above describe ‘static’ disorder, which will be valid at low
temperatures where molecular positions are fixed. At elevated temperature,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 366

366 D. Credgington

dynamic spatial disorder is induced by the presence of phonons. Since organic

semiconductors are typically flexible, and adjacent molecules not strongly bound
together, this thermally induced disorder can be large enough to perturb elec-
tronic coupling between molecules significantly — affecting both electronic tran-
sitions and charge transport. Since couplings strengths vary in time, modelling such
behaviour is extremely complex, meaning dynamic effects are often approximated
as an additional temperature-dependent contribution to static disorder (Bässler and
Kohler, 2012).

8.4.3 Charge transport in a disordered material

Charge transport in organic semiconductors is quite different from that of their
inorganic counterparts, where a three-dimensional periodic lattice structure often
exists. The charges in the latter are strongly delocalised and their movement is
interrupted only by scattering with phonons or impurities; therefore their mobil-
ity decreases with increasing temperature. In polymer semiconductors, spatial and
energetic disorder prevents long-range band-like transport. Charge transport in
organic semiconductors is instead described by quantum-mechanical tunnelling
between localised states, and so depends on the ability of the charge carriers to
access states in their immediate vicinity. As we will discuss below, the availability
of accessible states is dependent on temperature, local electric field and carrier
density. We note that most models for carrier transport assume a rigid lattice struc-
ture, and so implicitly ignore the effects of dynamic disorder and polaronic self-
localisation, which will act to reduce carrier mobility. In general this simplification
is only important where quantitative predictions are sought. In addition, we will
not consider here the effects of anisotropic coupling between molecules (which
will lead to anisotropic mobility), or include the contribution of intra-molecular
transport, which we assume will only serve to reduce the per-state localisation.

Miller–Abrahams and Mott variable-range hopping

The Miller–Abrahams model was originally developed to describe conduction
through impurities in inorganic semiconductors (Miller and Abrahams, 1960).
Transport is assumed to occur in two steps — an initial phonon-assisted hop ‘up’, to
conserve energy, followed by lateral tunnelling to the new site. The charge transfer
rate Wi j for hopping from site i (with energy E i ) to site j (with energy E j ) is
given by

exp(−(E j − E i )/k B T ), E j − E i ≥ 0
Wi j = ν0 exp(−Ri j / ) (8.4)
1, E j − Ei < 0
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 367

Organic Photovoltaics 367

Figure 8.13 Schematic illustration of charge (circles) hopping between localised states
(lines). Hopping from site A to site B represents nearest-neighbour hopping, while hopping
to sites C or D represents variable-range hopping, with lower energetic activation competing
with lower tunnelling probability.

where ν0 is the hopping attempt frequency, is the localisation length, and Ri j

is the distance between sites i and j . The first exponential term describes the
tunnelling probability, while the second exponential term accounts for thermal
activation, with the assumption that there is no activation barrier for ‘downhill’
hops. Such a system is illustrated schematically in Fig. 8.13.
The Miller–Abrahams model was used by Mott to describe conduction in
a d-dimensional system with uncorrelated spatial and energetic disorder and a
constant density of states (DoS). By integrating Wij over available states in d + 1
dimensions (d spatial and one energetic), the conductivity σ of such a system was
found to follow the expression
1
σ ∝ exp[−(T0 /T ) d+1 ] (8.5)

where T0 is a parameter proportional to the reciprocal of the DoS at the Fermi

energy. This expression is therefore valid for variable-range hopping close to the
Fermi energy, where the assumption of constant DoS holds. The key prediction
of this model, and those below which derive from it, is that transport in organic
semiconductor devices is faster at elevated temperatures.

Gaussian disorder model

A more realistic treatment of the effect of energetic disorder suggests several mod-
ifications to the constant DoS model. Uncorrelated disorder around a mean value is
likely to give rise to a Gaussian ‘band’ of localised states, as illustrated in Fig. 8.14a.
If instead disorder arises from a single species, the localised states may exist in
a separate ‘impurity band’ within the forbidden energy gap — the original target
for Miller–Abrahams theory. The scheme shown in Fig. 8.14b represents a band
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 368

368 D. Credgington

Figure 8.14 Models for transport in a disordered (electron) density of states. Localised
states are indicated in red, delocalised states in grey. a) Fully localised Gaussian DoS;
b) Band-like DoS with localised donor-like states within the bandgap; c) Transition from
localised to delocalised states within a continuous DoS, demarcated by an abrupt mobility
edge.

of lower-lying ‘donor’ states, separated from delocalised states of the conduction

band. Figure 8.14c presents the situation where the presence of disorder induces a
tail of localised states at the bottom of the conduction band. The threshold between
localised and delocalised states is termed the mobility edge (discussed below).
By assuming Miller–Abrahams hopping rates between localised states, and
employing Monte Carlo simulations, Bässler derived an expression for the mobility
µ of a material described by a distribution of localised states characterised by two
Gaussian distributions: one spatial, with width
and one energetic, with width σ

 √
4 2 exp[C(σ̂ 2 −
2 ) √
E],
≥ 1.5
µ(E, T ) = µ0 exp − σ̂ (8.6)
9 exp[C(σ̂ 2 − 2.25) E],
< 1.5

where µ0 is the zero-field mobility at infinite temperature, σ̂ ≡ σ/kB T , C is a

constant related to the typical lattice spacing and E is electric field (Bässler, 1993).
This Gaussian disorder model (GDM) predicts that the mobility is dependent
on the field by a factor of exp(E 1/2 ), which arises from the increased number of
‘downhill’ hops made available by the applied field. It was later found that the
agreement with experiment could be improved by taking the correlation between
spatial and energetic disorder into consideration — the correlated disorder model
(CDM) (Dunlap et al., 1996). Further detail may be added by considering the
effect of state filling as the carrier density increases. Low-lying states with low
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 369

Organic Photovoltaics 369

transition probabilities will be filled first, implying that the mobility of subsequent
carriers progressively increases. Such a modification was developed by Pasveer
and co-workers (Pasveer et al., 2005).

Multiple trapping and release

Since the number of states at a given energy is likely to correlate with their local-
isation, we may also consider a model whereby sparse, localised ‘tail states’ give
way to increasingly delocalised ‘band-like’ states as energies approach the unper-
turbed HOMO/LUMO level. This tail is typically treated either as a Gaussian
or, more simply, as a decaying exponential. Extracting analytical expressions for
temperature-, field- and density-dependent mobility is difficult in this case, and sta-
tistical methods such as Monte Carlo simulations are required to understand charge
transport. The behaviour of carriers in such a regime is usually simplified by defin-
ing a mobility edge, as illustrated in Fig. 8.14c. States below this edge are treated
as localised, while states above the mobility edge are treated as delocalised and
mobile. Carriers in localised states are therefore trapped until thermally released to
the extended states, while free carriers in the extended states can be trapped to the
localised states (Nelson, 2003). Transport of charge carriers in this model depends
on the spatial and energetic distribution of localised sites, which may be included
as a parameter, or extracted if sufficiently detailed data are available (MacKenzie
et al., 2012).

Numerical device simulations

Based on such models of carrier mobility, simulations of the current–voltage curves
of complete devices are possible. The literature on such simulations is extensive, but
we note here three important concepts. At low internal field, transport is primarily
diffusive, with the diffusion coefficient D assumed to be given by either the Einstein
relation (Eq. (8.7)) or the generalised Einstein relation (Eq. (8.8)) (Reese et al.,
2010):
D/µ = kT /q (8.7)
D/µ = g(n, T )kT /q (8.8)
where µ is the zero-field mobility and g is a function of the carrier density (n) and
temperature which attempts to account for the disordered nature of the DoS. The
need for such an ‘enhancement factor’ in practical devices is contested, however
(Wetzelaer et al., 2011a).
Since organic photovoltaics are thin-film (∼100 nm) devices with low dielec-
tric constants, they are subject to significant internal fields under operation, of the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 370

370 D. Credgington

order 107 V m−1 . Carrier drift is therefore a significant process and usually domi-
nates even at modest applied voltage, such that forward-biased organic diodes are
usually found to be space-charge limited (Mihailetchi et al., 2005).
Therefore, complete drift–diffusion simulations are required to capture the
main features of device operation in the power-generating quadrant, solving (in
the 1D case)
dn(x)
i n = i diffusion + i drift = q D + qµn E(x)n(x) (8.9)
dx
under the continuity condition
di n (x)
+ G(x) − R(x) = 0 (8.10)
dx
where i n is the current density due to electrons, µn is the electron mobility, G
and R are local generation and recombination rates for electrons, and we assume
equivalent expressions for holes with a requirement for charge neutrality. It is usual
to base such a model on an ‘effective medium’ representing the combined contri-
bution of each blend component, which can be made more realistic by including
localised variations in the DoS or separate contributions from different thermody-
namic phases.
In practice, obtaining sufficiently detailed data over a range of time and temper-
ature scales to isolate all of the internal processes contributing to charge transport
is extremely difficult (Bässler and Kohler, 2012).

8.4.4 Carrier generation

Optical bandgap
In order to generate large current densities, it is first necessary for a solar cell
to absorb a large fraction of the solar spectrum, which requires a semiconductor
with a low optical bandgap (of order 1–1.5 eV). However, as outlined in Sec-
tion 8.2.1, most small organic molecules and polymers are found to have relatively
large HOMO–LUMO offsets. Figure 8.15 illustrates this with the prototypical
homopolymer P3HT, which has an absorption edge at around 1.9 eV (650 nm).
Even assuming complete absorption of all photons with energies greater than this,
such a large bandgap covers only 20% of the AM1.5G solar photon flux. This is
compounded by the well-defined (rather than band-like) electronic transitions in
molecular materials, which lead to relatively low absorption bandwidths.
The simplest method to collect additional photons is to combine donor and
acceptor species with complementary absorption. For polymer:polymer cells, this is
possible, but for the more efficient polymer:fullerene (or small-molecule:fullerene)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 371

Organic Photovoltaics 371

Figure 8.15 AM1.5G solar spectrum and associated normalised cumulative photocurrent,
assuming that all photons below the given wavelength are converted to carriers and collected.
The absorption spectra of P3HT and PCPDTBT thin films are included for comparison.

systems, changing the acceptor is more difficult. Since C60 and its soluble deriva-
tives have spherical symmetry, their extinction coefficients are very low in the
visible range. Substituting C60 by the prolate C70 fullerene significantly increases
absorption in the visible and ultraviolet (UV), but at higher materials cost. To
date, it has not been possible to replace fullerenes successfully with more highly
absorbing acceptor materials.
The primary responsibility for absorption thus lies with the donor material.
Increasing conjugation lengths to reduce the HOMO–LUMO gap can only be
taken so far before molecules become too large or too stiff to deposit. Equally,
including heteroatoms within the (primarily hydrocarbon) organic backbone shifts
both HOMO and LUMO energies, but not necessarily their offset (Son et al.,
2011). This problem has been solved by coupling together electron-rich and
electron-withdrawing units in a ‘push-pull’ (or, confusingly ‘donor-acceptor’) con-
figuration, which can lead to the molecular HOMO states resting primarily on
the electron-rich unit, while the molecular LUMO states reside primarily on the
electron-withdrawing unit.
Figure 8.16 shows an example of this behaviour, illustrating the shift in electron
density for a 3CDTBT oligomer (analogous to the backbone of PCDTBT) from the
cyclopentadithiophene ‘push’ units to the benzothiadiazole ‘pull’ units. The result
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 372

372 D. Credgington

Figure 8.16 Shift in electron density on a 3CDTBT oligomer on excitation from the
ground (HOMO) to the first excited state (LUMO), calculated using time-dependent den-
sity functional theory. Electron density moves from red to blue regions. Image courtesy of
Sheridan Few.

has been a class of push-pull small molecules and polymers with absorption now
extending to beyond 900 nm (Bronstein et al., 2011; Mühlbacher et al., 2006; Zhong
et al., 2013). An additional benefit is that the higher-lying electronic transitions are
still accessible, which can broaden the absorption profile of such materials over
much of the visible range. The best performing polymer:fullerene cells utilise the
push-pull polymer PTB7 in combination with PC70BM to achieve >70% EQE from
400 to 750 nm, resulting in short-circuit current densities in excess of 17 mA cm−2
(He et al., 2012).

Exciton dissociation
While absorption of photons is a necessary precursor to current generation, the early
work on single-material devices showed that it was not sufficient. The necessity
for a type 2 (staggered) heterojunction indicates that a driving force is required to
separate charge, which represents a loss of available free energy. For example, in
the P3HT:PCBM system, around 800 meV of free energy is lost per carrier during
dissociation. Considerable effort has been expended to understand how this energy
loss may be minimised, while maintaining good separation efficiency (Clarke and
Durrant, 2010). We illustrate the main processes in Fig. 8.17, and refer to the rates
for those processes below.
When excitons reach the heterointerface, charge transfer from donor to accep-
tor is exceptionally fast, with the process complete within 50 fs — i.e. k CT  kPL .
Studies of blend photoluminescence show that only 5–10 wt% of a good acceptor
mixed homogeneously with a donor is necessary to quench over 90% of exci-
tons. However, the result of such quenching is not necessarily dissociated charges.
The low dielectric constant of organic semiconductors means that the exciton may
remain bound across the heterointerface in a charge-transfer (CT) state. The spatial
separation necessitated by the interface means the binding energy of the CT state
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 373

Organic Photovoltaics 373

Figure 8.17 Energy-level diagram illustrating the competing processes (with rates k i )
involved in photogeneration following absorption of a photon with energy hν. Exciton
diffusion and intersystem crossing of the donor singlet are ignored. Solid arrows repre-
sent processes increasing charge separation; dashed arrows represent relaxation towards the
ground state S0 . kCT : Dissociation of relaxed (or hot, kCT∗ ) singlet excitons (S) to form
hot interfacial charge-transfer (CT) states. kCS : Dissociation of relaxed (or hot, kCS∗ and
kCS∗∗ ) CT states to form fully dissociated charge-separated (CS) states. kISC : Intersystem
crossing between singlet and triplet CT states. k triplet : Generation of molecular triplet exci-
tons from triplet CT states, including possible re-dissociation at the heterojunction. k PL/Ph :
Relaxation of singlet/triplet excitons to the ground state. k GR : Geminate recombination
of CT states. kNGR : Non-geminate recombination of dissociated carriers, either directly
to the ground state or, more likely, via re-formation of CT states at the heterointerface.
k therm : Thermal relaxation to the corresponding lowest energy excited state. Since the CT
state involves partially separated carriers, there is unlikely to be a significant difference in
exchange interaction between 1 CT and 3 CT. Adapted from Clarke and Durrant (2010).

will be less than that of the molecular exciton, but still greater than the available
thermal energy; estimates for CT binding energies vary from around 0.1 to 0.5 eV
depending on the assumed separation and local polarisation. Excitons occupying
singlet CT states are observed to decay to the ground state within a few nanosec-
onds (a process usually termed geminate recombination, kGR in Fig. 8.17). While
the majority of CT states decay non-radiatively, some couple radiatively to the
ground state and produce CT luminescence, which provides a useful tool for their
study (Tvingstedt et al., 2009; Vandewal et al., 2009). The mechanism by which
CT excitons avoid geminate recombination and dissociate remains unclear, but
several routes have been proposed, as we now discuss.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 374

374 D. Credgington

Field-assisted ionisation
In many organic systems, macroscopic electric fields can distort the local energy
landscape and enhance separation efficiency. This process, observed in single-
component systems (Dick et al., 1994; Silinsh et al., 1974), polymer:polymer
systems (Gonzalez-Rabade et al., 2009), and some systems with fullerene acceptors
(Credgington et al., 2012; Veldman et al., 2008), is usually understood in terms
of Onsager–Braun theory. This describes the separation of electrons and holes in a
dielectric medium by calculating the probability that a Brownian random walk will
separate an electron-hole pair by more than their Coulomb capture radius — defined
by Onsager as the distance at which the Coulomb binding energy equals the thermal
energy kT. Excess energy provided by the incident photon (or heterojunction) would
generate an initially ‘hot’ carrier, and thus increase the chance of exceeding this
radius. Any macroscopic electric field will also lower the Coulomb potential in
the direction of the field, and so aid dissociation. Within this theory, the rate of
dissociation for a geminate pair kCS as a function of electric field E was derived
by Braun (1984) as:
 
3 µ
q b2 b3
kCS (E) = exp(−E/kT ) 1 + b + + + · · · (8.11)
4π s
a 3 3 18

where µ
is the spatially averaged sum of the electron and hole mobilities, s

is the spatially averaged dielectric constant, E is the Coulomb attraction of the

electron-hole pair after thermalisation at a distance a and b = q 3 E/8π s
(kT )2 ,
where the final terms arise from a power-series approximation to a first-order Bessel
function.
The efficiency of dissociation thus depends on the cumulative dissociation
probability over the CT state lifetime. While effective at describing spontaneous
dissociation in single-component systems, applying this theory to predict quantita-
tive yields of dissociated carriers in heterojunction systems remains problematic.
In these systems, carrier generation is often observed to be independent of temper-
ature, and the CT lifetimes (or local mobilities) which must be assumed are usu-
ally unphysical. Dissociation is empirically observed to be far more efficient (and
thus far less dependent on macroscopic electric field) than might be expected —
i.e. k CS > kGR . Part of this difficulty arises from the assumed Coulomb capture
radius, which for a material with s ∼3–4 is 10–20 nm. This ignores the signif-
icant influence of static and dynamic energetic disorder, which can change the
energy landscape experienced by a carrier by ∼100 meV over this length scale. In
addition, the local electric field at the heterojunction will be influenced by the ori-
entation of the junction with respect to the macroscopic field, by space charge in the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 375

Organic Photovoltaics 375

photoactive layer and by the presence of interfacial dipoles/polarisation (Rogers

et al., 1998).
The physical mechanisms allowing efficient exciton dissociation at organic
heterojunctions (and particularly in fullerene-based systems) are thus not well
understood. Several empirical observations help to shape our picture of the process,
however.

Generation from CT states

As we have described, CT states present a paradox. When long-lived, they do
not ionise efficiently and so are considered to be dead ends in the generation of
dissociated carriers. However, since the CT state is often radiatively coupled (albeit
weakly) to the ground state, it is possibly to optically excite these states directly.
A number of studies have examined the quantum efficiency of polaron generation
from such sub-bandgap absorption, and concluded that kCS = kCS∗ : i.e. that there
is no difference in quantum efficiency for polaron generation from CT excitons as
compared with those generated from molecular excitons (Lee et al., 2010).
This apparent contradiction suggests that the behaviour of CT states when they
are first generated plays a critical role in their subsequent dissociation. The initially
excited CT state is not geometrically relaxed, and thus corresponds to a relatively
delocalised excitation. Subsequent geometric relaxation of those CT states which
do not dissociate leads to a significant increase in their binding energy. The dif-
ference between CT absorption and emission spectra indicates this increase is of
order 0.3 eV, due to both reorganisation of the molecular structure and increased
coulomb attraction (Gélinas et al., 2011). Equilibrated CT states are therefore
tightly bound, and unlikely to dissociate, consistent with observation. However,
non-equilibrium CT species — which are generated either by direct photon absorp-
tion or by charge-transfer from a molecular exciton — can achieve high dissociation
yields. For instance, Bakulin and co-workers (Bakulin et al., 2012) showed that
exciting long-lived relaxed CT states with a low energy light pulse recreates these
more delocalised species, which subsequently separate into dissociated carriers.
Separation of carriers therefore requires that more energy is available than
is needed to form a relaxed CT exciton. Since this excess represents a loss of
free energy available to do electrical work (as discussed in Section 8.4.5), there is
considerable interest in establishing the minimum excess energy needed to drive
polaron dissociation. Since we are mainly concerned with systems with fullerene
acceptors, this is usually described in terms of the minimum LUMO–LUMO offset.
Empirical observations suggest that for many blend systems, a correlation
exists between LUMO–LUMO offset and the yield of long-lived dissociated
polarons (Shoaee et al., 2010; Shoaee et al., 2013). However, different classes of
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 376

376 D. Credgington

material exhibit different ‘minimum’ driving energies, suggesting that molecular

structure and blend morphology are significant complicating factors in this pro-
cess, and may obscure the exact role of energetics. It is also speculated that the
dipole moment associated with transitions to the excited state may play a part,
and has been cited as a possible reason why devices based on modern push-pull
donors achieve excellent dissociation efficiency with offsets of only 100–200 meV
(Carsten et al., 2011). While the underlying mechanism remains uncertain, a rel-
atively safe empirical design rule is that around 300 meV of driving energy is
usually sufficient to achieve electron (or hole) transfer and polaron dissociation in
a well-optimised organic blend.

Hot exciton effects

In some crystalline organic systems, including the fullerenes, absorption of a pho-
ton with energy well above the bandgap can lead to spontaneous generation of
polarons — in accordance with the Onsager–Braun model of exciton dissociation.
In a heterojunction system, and where absorption occurs away from the heteroint-
erface, this effect is expected to be small: since excitons diffusing to the interface on
picosecond to nanosecond timescales are likely to be thermally relaxed, they will
have no memory of the initial photon energy. However, if absorption occurs directly
at the interface, dissociation of the ‘hot’ exciton may compete with thermalisation
(kStherm ≈ kCT∗ ), since both can be femtosecond processes. In bulk heterojunction
systems with highly intermixed donor and acceptor phases, absorption directly at
the heterointerface is common, and the additional thermal energy has been shown
to assist the dissociation process. In particular, it has been observed that ‘cold’
excitons lead to the population of lower-lying CT states, which may dissociate less
efficiently than higher-lying CT states populated by ‘hot’ excitons (kCS∗∗ > kCS∗ )
(Dimitrov et al., 2012; Grancini et al., 2013).

Suppression of geminate recombination

Geminate recombination is a severely limiting process for organic solar cells, and
while current evidence suggests that it can be suppressed in empirically optimised
materials by a sufficiently large band offset at the heterojunction, this has a sig-
nificant cost in free energy. However, the generation of CT states (if not their
dissociation) appears to be a highly overdriven reaction — once an exciton reaches
the heterojunction, the quantum yield of charge transfer is unity. This suggests
that much of the free energy of charge transfer could be better spent moving the
geminate pair further apart. Inspiration for this approach comes from biological
photosynthesis, which employs an energetic cascade to drive electrons away from
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 377

Organic Photovoltaics 377

the chlorophyll absorbers. Each step costs energy, but decreases the associated
back-recombination rate. Considerable effort has therefore been directed towards
including spacer molecules or groups at the heterointerface to increase the electron-
hole separation of the CT state and thus increase its lifetime. This necessarily
reduces the rate of CT state formation, but since there are no competing processes
on the timescales involved, this need not be restrictive.
A related approach may already have been implemented empirically in hier-
archical polymer:fullerene bulk heterojunctions, where amorphous and crystalline
regions exist together. Since crystalline phases typically exhibit narrower bandgaps
than more amorphous phases, the existence of a crystalline region adjacent to the
more amorphous interface region may also drive carriers away from the interface
and thus reduce the geminate recombination rate (Jamieson et al., 2012).

8.4.5 Free energy per carrier

When a photon with energy E ph is absorbed by the photoactive layer and generates
carriers, only the free energy of those carriers reaching the contacts can be used
to do work. The remainder is lost, generally to phonons or emitted photons. The
maximum free energy available is therefore qVoc, since at open circuit we require
negligible carrier flow. Maximising Voc therefore involves minimising the free
energy lost by carriers as they transit the device. We have already discussed the
losses of free energy associated with carrier formation, which are:
1. Relaxation of the initially hot exciton back to the band edges. This is unavoidable
for absorbers with a single bandgap, and sets a limit on how small that bandgap
can be made.
2. The excess energy required to drive charge separation at the heterojunction.
We now discuss a third, which is common to all photovoltaic devices. The
free energy available at open circuit is the difference in the chemical potentials (or
quasi-Fermi energies) of electrons and holes at their respective extracting contacts,
which incorporates both the enthalpy associated with the occupancy of excited
states and the entropic cost of maintaining thermochemical equilibrium between
the solar cell and light source. In the Shockley–Queisser treatment of an idealised
solar cell, this equilibrium is maintained only by the radiative recombination of
carriers. As discussed in Chapter 2 of this volume, for a device illuminated by the
Sun, this gives a voltage loss of around 0.3 V arising primarily from the mismatch
between the solid angle subtended by the Sun at the Earth and the solid angle of
re-radiation from the device. Explicit in this analysis is the scope for excitons and
photons to exchange energy multiple times before the photon finally escapes or
electrons and holes are collected at their electrodes. However, any non-radiative
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 378

378 D. Credgington

recombination processes will necessarily increase the recombination rate, and so

reduce the available voltage.
Within this framework, we may understand the empirically observed correla-
tions between device energetics and Voc . We begin by assuming that the hetero-
junction can be approximated as an effective medium with a single, well-defined
pair of HOMO and LUMO states and no disorder, and that the reduction in free
energy due to recombination may be treated as an empirical offset parameterised
as V . We will also assume for now that illumination intensity is fixed.
Since most of these recombination processes are a function of carrier density
or cell voltage (or both), they increase with the filling fraction of the DoS. The point
at which the total recombination current i rec balances the rate of carrier generation
i gen sets the equilibrium filling fraction, and thus the maximum quasi-Fermi energy
splitting. Organic solar cells therefore experience a voltage loss V beyond the
0.3 V predicted by the Shockley–Queisser analysis, owing to both non-radiative
recombination and disorder.

Metal–insulator–metal regime
In the metal–insulator–metal (MIM) regime, the electrode work functions lie within
the heterojunction bandgap. Since organic semiconductors are usually intrinsic,
equilibration between the electrodes is likely to deplete the entire semiconduc-
tor thickness, preserving the mismatch between the heterojunction energy levels
and the electrode work functions. Carriers reaching an electrode must therefore
thermalise to the work function, reducing their enthalpy. As such, the quasi-Fermi
energies in the bulk are limited by , giving
1
Voc = (h − e ) − V (8.12)
q
where e/ h are the work functions of the electron/hole extracting contacts,
respectively.

Materials-limited regime
If the electrode work functions lie outside the heterojunction bandgap, charge may
be transferred from the electrodes to empty states in the semiconductor, caus-
ing sharp band bending. This allows equilibration between the electrodes and the
HOMO/LUMO levels of the heterojunction without the build-up of space charge
in the majority of the semiconductor (provided the offset is not too great). In this
regime, the quasi-Fermi energies at the contacts are ‘pinned’ to states in the semi-
conductor (Crispin et al., 2006). For well-defined (donor) HOMO and (acceptor)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 379

Organic Photovoltaics 379

LUMO levels, Voc will be primarily governed by the relative difference between
these states, which we have termed E HJ . These energies are usually estimated
from pure donor or acceptor films using electrochemical or spectroscopic tech-
niques such as cyclic voltammetry or ultraviolet photoelectron spectroscopy. As
such, it is common for the measured quantities Ionisation Potential (IP) and Elec-
tron Affinity (EA) to be quoted instead of E HOMO and E LUMO . The latter are not
generally known, since blending donor and acceptor together may induce a change
in energetics, either through changes in molecular packing or the introduction of
interfacial dipoles. Where these changes may be ignored, however, we expect
1
Voc = (IPdonor − EAacceptor ) − V (8.13)
q
which is a widely observed correlation. A more accurate measure of the HOMO–
LUMO gap may be obtained by using the weak luminescence of the CT state as
an in situ probe of E HJ , thereby implicitly including the effects of blending. The
CT energy (E CT ) may be estimated by measuring either the electroluminescence
of the blend (which will be dominated by CT emission), or by isolating sub-
bandgap absorption features arising from CT states. The most accurate correlation
between energetics and open-circuit voltage is therefore given by (Faist et al., 2012;
Vandewal et al., 2008)
1
Voc = E CT −V (8.14)
q
where V in equation (8.14) is reduced by any contribution from the coulomb
binding energy of the CT state.

Disorder and recombination

All of the approaches above give good empirical correlation with the measured
open-circuit voltage within the limits of their assumptions. However, all require
an empirical offset V of order 300–500 mV, which varies between systems (and
depends on measurement technique). This is because we have not yet accounted for
the influence of recombination and disorder on the quasi-Fermi energy positions.
If the molecular HOMO and LUMO bands are broadened by disorder, then
the experimentally determined CT state energy, or IP and EA, will correspond
to some average energy located mid-band, as will any reasonable definition of
E HJ . Photogenerated carriers will therefore be able to thermalise into the tail of
empty states below this energy, with an associated loss of enthalpy. Where the
characteristic disorder width is greater than the width of the Fermi distribution,
such that the quasi-Fermi energies lie within the disordered DoS, this enthalpic
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 380

380 D. Credgington

Figure 8.18 Schematic illustrating the effect of DoS width on achievable Voc . For a fixed
number of photoexcited carriers, the wider LUMO DoS (a) exhibits a lower electron quasi-
Fermi energy E F n1 than the narrower DoS (b) with quasi-Fermi energy E F n2 . Both LUMO
distributions share a common centre.

loss will dominate V (Blakesley and Neher, 2011). At constant carrier density,
V will therefore be determined by the extent of disorder, as illustrated in Fig. 8.18.
The carrier density in a solar cell under operation is not a constant, however. At
open circuit, where no net current flows out of the device, the internal carrier density
will be set by the equilibrium between carrier generation and carrier recombina-
tion. Within a real solar cell, numerous additional recombination pathways exist
beyond the radiative recombination described in the Shockley–Queisser approach.
These include geminate recombination of CT excitons, collection of carriers at
the ‘wrong’ electrode, non-radiative recombination between dissociated polarons
(non-geminate recombination), recombination through trap sites and direct leak-
age currents, all of which can be important in organic solar cells and all of which
can vary in severity between devices (Credgington and Durrant, 2012).
A general expression for the open-circuit voltage must therefore include not
only the average heterojunction bandgap, but also the characteristic disorder width
and the severity of recombination losses. For a device limited by both geminate
and Langevin-type recombination (see below) and with fixed DoS N (assumed to
be the same for electrons and holes), Koster et al. (2005) derived the expression


E HJ kT (1 − P)γ N 2
Voc = − ln (8.15)
q q PG
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 381

Organic Photovoltaics 381

where G is the generation rate of excitons per unit volume, P their dissociation
probability and γ a measure of the likelihood of Langevin recombination. Includ-
ing the effects of a general, disordered DoS and more complex recombination
mechanisms is not straightforward (Kirchartz et al., 2011), but in the case where
the dominant recombination process gives rise to a rate which varies as a power law
in carrier density, and the DoS varies exponentially with energy, the open-circuit
voltage may be expressed as


E HJ ηkT JBI
Voc = − ln (8.16)
q q Jgen
where η is an experimentally determined function of the DoS shape and recombina-
tion order, and JBI is related to the recombination rate at V = E HJ /q (Credgington
and Durrant, 2012).

Optimising open-circuit voltage

Routes to optimise Voc rely on minimising the free-energy losses described above,
and illustrated in Fig. 8.19. Minimising thermalisation losses requires a multijunc-
tion solar cell, with each junction absorbing progressively lower energy photons.
The free-energy loss per photon is thus considerably reduced. Table 8.1 gives the
thermodynamic limits for various junction arrangements, assuming only radiative
losses, spontaneous exciton dissociation and that the solar spectrum can be approx-
imated to a black body at 6000 K (Vos, 1980). The continuous multijunction is the
limit of an infinite series of thin solar cells with bandgaps varying continuously
from infinity to zero.
Minimising the loss at the LUMO–LUMO offset requires minimising the
required driving force for exciton dissociation. This is limited by the need to over-
come the CT-state binding energy, but there are indications that a driving force of
100–200 meV may be sufficient for some materials systems.
Minimising disorder widths is more difficult without employing more expen-
sive fabrication methods. The two most practical routes are reducing impuri-
ties/unwanted isomorphs during synthesis (including reducing the polydispersity
of polymers) and increasing the tendency of materials to crystallise. This must be
weighed against the requirement for thermal stability — a high driving force for
crystallisation may lead to Ostwald ripening over time (see Section 8.5.1).
Since the Shockley–Queisser approach considers only radiative recombina-
tion, it is often commented that a good solar cell is also a good LED. This must be
understood to mean that if carriers can be made to recombine radiatively (which
is a relatively slow process) then all other non-radiative recombination pathways
have implicitly been eliminated and the cell can operate closer to its maximum
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 382

382 D. Credgington

Figure 8.19 Free energy losses at a donor-acceptor heterojunction after absorption of a

photon with energy hν. a) Thermal relaxation to the donor bandgap E g ; b) exciton separation
at the heterojunction; c) relaxation into the disordered DoS, with electron/hole quasi-Fermi
energies E F n, p . For simplicity, energetic disorder in the donor LUMO and acceptor HOMO
is not considered, nor is relaxation of the exciton into the disordered DoS. Equilibration
between the electrode work function and semiconductor Fermi energies is assumed (i.e. the
system is not within the MIM regime).

Table 8.1 Theoretical efficiency limits for single and mul-

tijunction solar cells under 1 Sun and maximally concen-
trated sunlight (in brackets).

Maximum PCE under 1:1

Solar cell type (45,900:1) concentration

Single junction 30% (40%)

Tandem 42% (55%)
Triple junction 49% (63%)
Continuous multijunction 68% (86%)

thermodynamic efficiency. Non-radiative recombination losses can most simply

be reduced by improving the integrity of the device stack, to reduce leakage
through pinholes and spikes, and by engineering the electrode interfaces to reduce
surface recombination. However, methods to reduce the more important Langevin
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 383

Organic Photovoltaics 383

(Eq. (8.17)) or Shockley–Read–Hall (SRH) (Eq. (8.18)) recombination currents

are in their relative infancy, since the rate-limiting steps for these processes are not
yet well understood. These recombination currents are given by


q(np − n 2i ) µn + µ p
iL = (8.17)
A 2
q Nt Cn C p (np − n 2i )
iSRH = (8.18)
A[Cn (n + n 1 ) + C p ( p + p1 )]
where A is cell area, n, p are the density of electrons and holes in LUMO/HOMO
bands and µn , µ p their mobilities, n i is the intrinsic electron density, Nt is the
trap density, Cn , C p are the electron and hole capture cross-sections and n 1 , p1
are auxillary parameters such that n 1 p1 = n 2i . Note that Eq. (8.17) tends towards
Eq. (8.18) as the mobility of either carrier drops towards zero, and that Eq. (8.18)
simplifies considerably if one carrier is in significant excess.
It is known that certain materials systems (particularly those comprising
semicrystalline polythiophenes and fullerenes) exhibit recombination rates which
are significantly suppressed with respect to those expected from a simple mobility-
limited Langevin recombination model (Credgington and Durrant, 2012). Whether
this is due to morphological or energetic effects remains contentious, but it suggests
that there is scope to reduce recombination rates in other materials systems.
Finally, note that we have not yet considered the influence of light intensity
on recombination rate. The dominant recombination mechanism sets the ideality
factor η for a device, defined as (Wetzelaer et al., 2011b)


kT dlni ph −1
η= (8.19)
q dV oc
where i ph is the photogenerated current density (often replaced simply with illumi-
nation intensity). Recombination between mobile carrier populations (Langevin-
type) leads to an ideality of 1, whereas recombination through traps (SRH-type)
and at interfaces, or where one carrier is in significant excess, leads to a charac-
teristic ideality >1. Suppressing non-Langevin recombination processes therefore
has the additional benefit that efficiency under low-light conditions can more eas-
ily be maintained. Considerable emphasis has therefore been placed on balancing
carrier mobilities, balancing extraction efficiencies at the electrodes and removing
energetic traps.

8.4.6 Carrier collection efficiency

While in Section 8.4.4 we considered the maximum yield of dissociated carriers,
and in Section 8.4.5 the maximum free energy available per carrier pair, neither
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 384

384 D. Credgington

uniquely sets the efficiency of an OPV at its maximum power point. This is deter-
mined by how EQE varies with internal field — i.e. the fill factor of the cell.
As we have outlined, the losses within any heterojunction solar cell arise from
either: (i) Relaxation of excitons which fail to diffuse to, and separate at, a het-
erojunction; (ii) Recombination of geminate pairs formed at the heterojunction
which fail to fully dissociate; or (iii) Recombination of fully dissociated carri-
ers via surface recombination, Langevin-type bulk recombination and SRH-type
recombination.

Field-dependent exciton lifetime

In most efficient bulk heterojunction solar cells, the quenching efficiency of exci-
tons on the donor molecule is close to unity, implying that the relaxation of excitons
during diffusion is not a strongly limiting process (Jamieson et al., 2012; Morana
et al., 2010). Due to their significant binding energy, the intrinsic lifetime of the
exciton is only affected by internal fields well in excess of those experienced during
normal operation.

Field-dependent geminate recombination

The extent to which the dissociation of geminate pairs is affected by macroscopic
electric fields within the photoactive layer is currently the subject of significant
uncertainty. The mechanism for this process is assumed to be a variation on the
Onsager–Braun model of geminate pair dissociation, such that the Coulomb barrier
to dissociation is lowered in the direction of the internal field. Device models that
include field-dependent geminate recombination losses are widely employed to
analyse the output curves of OPV devices, and have provided support for the
importance of such losses (Andersson et al., 2011; Mandoc et al., 2007; Ooi et al.,
2008; Peumans and Forrest, 2004; Veldman et al., 2008).
In polymer:polymer bulk heterojunctions, there is good experimental evidence
that geminate pair separation efficiency can be enhanced by the presence of a
macroscopic electric field (Gonzalez-Rabade et al., 2009).
For heterojunctions employing a fullerene acceptor, the evidence that gemi-
nate pair separation efficiency is influenced by macroscopic electric fields is much
less clear-cut, and appears to be dependent on both materials selection and pro-
cessing regime. Direct experimental measurements of field-dependent geminate
losses have often observed these to vary under strong reverse bias — important for
devices used as photodetectors, but of less significance for the operation of photo-
voltaic cells (Tvingstedt et al., 2010). Measurements in the power-generating quad-
rant give mixed results — some authors conclude that the variation of geminate
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 385

Organic Photovoltaics 385

recombination rate with internal electric field is sufficient to significantly impact

the fill factor (Albrecht et al., 2012; Credgington et al., 2012), others that the effect
is negligible (Etzold et al., 2011; Kniepert et al., 2011; Shuttle et al., 2010; Street
et al., 2010). The fundamental mechanism that determines whether a given device
suffers from this loss process therefore remains unclear. Nevertheless, the highest
efficiency bulk heterojunction OPVs can achieve fill factors of greater than 70%
(He et al., 2012), implying that the dependence of geminate recombination rate on
internal field can be suppressed, albeit empirically.

Field-dependent non-geminate recombination

Recombination of dissociated carriers will lead to further losses in extraction
efficiency. In a device with highly carrier-selective electrodes, such that surface
recombination is minimised, recombination of carriers back across the hetero-
junction will be the dominant loss pathway. While the mechanisms of geminate
and non-geminate recombination are physically related, since both proceed via
an intermediate CT-like state, they are usually treated as separable due to their
differing timescales. Geminate recombination is usually a nanosecond process,
and non-geminate recombination lifetimes can extend to microseconds or even
milliseconds, meaning that the non-geminate recombination rate is more strongly
affected by local carrier density.
The non-geminate recombination rate in an OPV device is determined by sev-
eral field-dependent processes. Under constant illumination, increasing internal
field leads to shorter carrier dwell times and therefore lower equilibrium carrier
densities, reducing recombination rate. In addition, since non-geminate recombi-
nation is likely to proceed via CT-state formation, high internal fields may act to
dissociate some of these CT states and regenerate free carriers — particularly since
dissociated carriers may form longer-lived triplet CT excitons (with reference to
Fig. 8.17, if k triplet is low, kGR may be suppressed for triplet CT states). However,
observing such an effect is difficult since under high field, carrier densities, and so
recombination rates, are already low. Increasing internal field also alters the equi-
librium carrier distribution, affecting the spatial overlap between electron and hole
populations. Depending on the zero-field carrier distribution within a given device,
this may either increase or decrease the recombination rate as internal field varies.
The low mobilities and high internal fields common in OPV materials mean
that carrier distributions are also dependent on illumination intensity. Therefore the
equilibrium carrier distribution at a given applied bias in the dark differs from that
present at the same bias under illumination (Shuttle et al., 2010), and so the principle
of superposition does not normally hold for OPVs. That is, the current–voltage
curve of an illuminated cell cannot usually be separated into dark current and
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 386

386 D. Credgington

photocurrent components, making the study of non-geminate recombination rates

difficult. As such, analyses of recombination rates typically have to rely on time-
resolved, rather than steady state, measurements (Albrecht et al., 2012; Credging-
ton and Durrant, 2012; Marsh et al., 2010).
While control of non-geminate recombination rates remains largely empirical,
there is now a growing library of experimental techniques capable of quantify-
ing both geminate and non-geminate recombination processes, and their variation
across the current–voltage curve. Reductions in recombination rate are correlated
with improved carrier collection efficiency and higher fill factors, as might be
expected, though it is not yet clear which methods for achieving this will emerge
as the most fruitful. As noted above, the fill factor can exceed 70% in the highest
efficiency devices, meaning that suppression of recombination losses is clearly
possible.

8.5 Beyond the state of the art

In this section, we detail directions in which the OPV field is moving beyond the
current state of the art, both in terms of focus, new technology and new physics.

8.5.1 Stability
Device lifetime is a critical concern for any commercial photovoltaic technology.
With the development of lab-scale OPVs with PCEs exceeding 10%, significant
attention has been focussed on the stability of organic devices. To compete directly
with existing technology, lifetimes approaching 100,000 hours under illumination
are sought. As with device performance, the rate of progress in this area has been
rapid, meaning that the state-of-the art described here will quickly become out-
of-date. We therefore outline the recent progress in developing stable devices,
and refer the reader to the series of reviews by Krebs and co-workers for a more
comprehensive outlook (Jørgensen et al., 2012). In particular, we do not consider
the interplay between stability and manufacturing process and the evolution of
encapsulation methods, which are not unique to organic photovoltaics.
In the following, we divide our discussion by degradation target, beginning
with the photoactive layer and extending to the electrodes, including the impact of
interfacial layers.

Photoactive layer: Photostability

The function of organic semiconductors is determined by their molecular structure.
Chemical changes to that structure are usually detrimental to device performance,
so the main concern for the long-term stability in OPVs is the ability of the
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 387

Organic Photovoltaics 387

organic layers to endure prolonged exposure to illumination and elevated temper-

ature without chemical degradation. This is of particular concern in luminescent
organic materials, where excited states can persist for relatively long times and
photon energies (particularly in the UV) may be comparable to the strengths of
intra-molecular bonds. In this respect, experience with organic LEDs (OLEDs) is
extremely encouraging. The lifetimes of solution-processed OLED materials can
now exceed 1,000,000 hours, suggesting there is no practical limit to the intrinsic
stability of organic semiconductors (Dupont, 2009).
Nevertheless, current OPV materials systems fall short of the stability of
OLEDs. One of the most important reactions limiting lifetime are the interac-
tions between excited organic molecules and ambient oxygen. The prototypical
poly(phenylene vinylene)-related polymer semiconductors (PPVs) were observed
to degrade within minutes on exposure to light, arising from attack by oxygen on
both the solubilising side chains and the main vinylene backbone units (Chambon
et al., 2007).
With the growing library of semiconducting polymers and experience with
OLEDs and OPV devices, certain design rules have emerged for the design of sta-
ble materials — which has until recently been of secondary concern to efficiency.
For example, by estimating the degradation rate of individual monomer units under
a given photon flux, Manceau and co-workers developed an empirical ranking of
donor and acceptor units in terms of their intrinsic photostability, reproduced in
Fig. 8.20 (Manceau et al., 2011). The least stable, fluorene, is widely observed to
undergo oxidation to fluorenone regardless of the side-chain configuration (Griso-
rio et al., 2011). This was particularly detrimental to the blue OLEDs in which it was
first used, since fluorenone is a green emitter and acts as a trap for charge transport.
The continuous addition of new monomer units to the available library of
molecular design means that optimised stability remains something of a moving
target. However, studies such as these suggest that certain units, such as fluo-
rene, should be avoided in the design of stable materials, and that stability may be
enhanced by the substitution of silicon in place of carbon where possible. In addi-
tion, even for the most stable units, the addition of solubilising side chains provides
a general route for photodegradation. For example, significantly longer lifetimes
are observed for the PPV-replacing poly(alkyl-thiophenes) (Manceau et al., 2010a),
which was one reason for the P3HT:PCBM system being developed as a possible
target for scale-up and large-area processing. Nevertheless, photodegradation is
still observed, and is linked to attack by oxygen on the solubilising hexyl groups
(Manceau et al., 2009).
Since side chains are almost always required to render an organic semi-
conductor soluble, removing them entirely to improve stability is not possible.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 388

388 D. Credgington

Figure 8.20 ‘Rule-of-thumb’ photostability ranking of donor and acceptor monomers for
OPV polymers. Adapted from Manceau et al. (2011).

One option to circumvent this problem is to include thermocleavable solubilising

groups, which may be removed post-deposition by heating. Such approaches are
not yet widespread, but have been shown to improve photostability (Manceau et al.,
2010b).
Since the degradation of organic semiconductors is linked to the presence
of both an excited molecule and oxygen (degradation rates for material kept in
the dark are very slow), the removal of either improves stability significantly.
Removing excited states is relatively straightforward — the addition of an elec-
tron or hole acceptor rapidly quenches excitons. The photostability of blended
heterojunction films is therefore significantly higher than that of pristine materials
(Neugebauer et al., 2000), although PCBM itself may also photodegrade (Reese
et al., 2010). Removal of oxygen relies on good encapsulation combined with
oxygen-scavenging ‘getter’ materials.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 389

Organic Photovoltaics 389

Photoactive layer: Morphological stability

As we outlined in Section 8.4.1, the multiphase structure formed during the
controlled deposition of a bulk heterojunction is thermodynamically unstable.
Domains comprising equilibrium phases will Ostwald-ripen at the expense of non-
equilibrium structures, and both donor and acceptor materials may form crystallites.
These can reduce the heterointerface area, reducing carrier generation, and cut per-
colation pathways for carrier transport. The timescale over which these processes
happen will be determined by operating temperature and the size of the molecules
in question. In particular, small fullerene acceptors are extremely mobile within
amorphous polymer domains at elevated temperature (Treat et al., 2011), meaning
that formation of fullerene crystallites is a significant degradation process within
polymer:fullerene photovoltaics.
Stabilising morphology has been attempted by a number of routes. The first is
to tether donor and acceptor molecules together as block-copolymers, which may
be tailored to form a thermodynamically stable interpenetrating network (Segalman
et al., 2009). However, these devices have yet to achieve PCEs beyond 2%, meaning
that morphological stability in such systems currently comes at too high a cost.
An alternative is to constrain the diffusion of blend components post-
deposition by means of cross-linking between polymer chains using the solu-
bilising side groups. This may be achieved using a photo- or thermally curable
cross-linking agent as a tertiary component in the blend (Liu et al., 2012; Png
et al., 2010), although this must be carefully targeted to avoid damaging the semi-
conducting backbone. It has also been possible to chemically engineer polymers
with side groups that are inherently cross-linkable (Charas et al., 2006). Neither
technique has yet been explored fully in the highest efficiency polymer:fullerene
systems, but these approaches can in principle be applied universally. However,
cross-linking the polymer may not entirely prevent the much smaller fullerene from
diffusing through the blend and aggregating. Employing cross-linkable fullerenes
is one route to address this problem (Drees et al., 2005). Another is to remove the
side groups altogether, since these are responsible for a significant proportion of
the blend’s free volume, and thus facilitate both polymer reptation and fullerene
diffusion.
Thermo-cleavable side groups, as outlined above, can therefore also aid
morphological stability, as well as photostability. P3HT-derivatives incorporat-
ing thermo-cleavable carboxylic esters, blended with PCBM, exhibit significantly
enhanced stability after cleaving above 200◦C (Krebs and Spanggaard, 2005). How-
ever, while P3HT:PCBM typically requires an annealing step at around 150◦ C to
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 390

390 D. Credgington

achieve the good efficiency, annealing has been shown to be detrimental for many
better performing systems. It therefore remains to be seen whether thermo-cleaving
can be achieved at temperatures low enough to avoid this process.
An advantage of all of the approaches outlined above is that they also render
the blend insoluble, allowing for multiple thin-film stacks to be deposited without
the need for orthogonal solvents. This ability is crucial for developing multijunction
cells, as described in Section 8.5.2.

Electron-collecting electrode
The photoactive layer is not the only device component susceptible to degrada-
tion. Historically, the electron-collecting electrode has been formed of an evap-
orated metal. The requirement for this metal to have a low work function (see
Section 8.4.5) means that highly reactive materials such as calcium and aluminium
are needed. Such metals react in the presence of water or oxygen, creating an insu-
lating metal oxide layer which grows from pinholes and the edges of the electrode,
eventually spreading over the entire device. As the oxide increases in thickness,
the transport barrier created eventually impedes the extraction of electrons from
the entire active layer (Glatthaar et al., 2007).
One method to circumvent this problem is to use a less reactive, high-work-
function metal such as Ag or Au. These must be combined with a second mate-
rial, such as a conducting metal oxide, to achieve a low work function at the
organic:electrode interface. An alternative is to use an inverted structure, such that
Ag or Au, sometimes combined with a high-work-function interlayer, acts as the
hole-collecting electrode while a stable conductive oxide such as TiO2 is used to
extract electrons. The use of oxides as low-work-function electrodes has driven
interest in inverted device structures, since they are intrinsically stable to oxygen
degradation. In addition, while somewhat less conductive than TiO2 , solution-
processed titanium sub-oxide (TiOx ) has been shown to act as an oxygen ‘getter’
which also improves the photostability of the active layer (Li et al., 2011).

Hole-collecting electrode
While interfaces between ITO and organic materials have been shown to impact
stability, its relatively low work function means that it is unusual to use ITO without
a high-work-function hole-transporting layer — PEDOT:PSS (Fig. 8.6) being by
far the most widely used. However, PEDOT:PSS has been implicated in device
degradation through several mechanisms.
Since the conductivity of PEDOT:PSS relies on p-doping of the PEDOT
chain by the charged PSS component, phase separation, structural degradation
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 391

Organic Photovoltaics 391

and electrochemical reactions that interfere with this charge exchange reduce the
conductivity of the PEDOT:PSS layer over time (Vitoratos et al., 2009). Commer-
cial PEDOT:PSS formulations can be highly acidic, such that ambient humidity can
significantly increase the rate of corrosion of other material components — partic-
ularly low-work-function metals, but also ITO (de Jong et al., 2000). PEDOT:PSS
is also highly hygroscopic, enhancing this effect even when minimal atmospheric
water is present.
As a result, there have been significant efforts to try and replace PEDOT:PSS
with a stable, transparent, high-work-function hole conductor. Examples of possi-
ble replacements include vacuum-deposited NiO (Irwin et al., 2008), and solution-
or vacuum-deposited MoO3 (Zilberberg et al., 2012). Both have been shown to
significantly enhance the stability of devices, particularly under humid conditions.
As in the previous section, an alternative approach is to switch to an inverted
structure such that a transparent electron-collecting electrode is used, and holes are
collected through a stable, high-work-function metal (Ag or Au, potentially with
an additional interlayer).
A huge number of other electron and hole-transporting interlayers have been
examined in the OPV and OLED fields, though many have not been assessed for
stability. For details, we refer the reader to the existing literature (Steim et al.,
2010).

8.5.2 Multijunction organic photovoltaics

As discussed in Section 8.4.5, thermalisation losses and narrow spectral response
both limit the performance of OPVs. For any solar cell technology, a common
route to reducing thermalisation losses and broadening spectral response is to
absorb high-energy photons in a wide-bandgap semiconductor before absorbing
low-energy photons in a low-bandgap semiconductor — i.e. to construct a tandem
cell. A detailed analysis of tandem cells is beyond the scope of this chapter; instead
we outline current progress towards realising such cells using organic materials.
The basic stacked tandem structure is shown in Fig. 8.21, and requires pre-
cise control over each layer of the stack. Evaporated small-molecule devices are
therefore particularly well suited for combining into the tandem architecture, and
the highest-efficiency organic solar cells to date are produced in this way. In par-
ticular, controlled doping of individual layers using molecular dopants is possible
(Lüssem et al., 2013), and the optical field structure can be precisely controlled by
introducing transparent spacer layers to ensure efficient absorption by the photoac-
tive components (Riede et al., 2011). This flexibility also means that there is no
technological barrier to producing evaporated multijunction organic cells (Sullivan
et al., 2013).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 392

392 D. Credgington

Figure 8.21 a) Physical structure; and b) energetic structure of a tandem heterojunction

solar cell. High-energy (blue) photons are absorbed by the donor (D) of the front, wide-
bandgap sub-cell while low-energy (red) photons pass through to the narrow-bandgap donor
of the second sub-cell. Optimum performance requires lossless recombination (i.e. matched
electron/hole Fermi energies (E F n, p )) and matched electron/hole currents at the common
recombination contact. In this example, identical acceptors (A) are assumed.

Achieving solution-deposited tandem cells is complicated by the need to pre-

vent subsequent layers dissolving those beneath them. This can be achieved by
rendering each layer insoluble before deposition of the next (through cross-linking
or similar), by choosing orthogonal solvents for the deposition of each layer, or by
depositing one or more layers separately, then combining the resulting films using
soft lithography.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 393

Organic Photovoltaics 393

In either approach, of critical importance is the choice of material for the

recombination contact, which must allow holes from the top sub-cell to recom-
bine with electrons from the bottom sub-cell, or vice versa. In early attempts at
tandem cells, a metallic interlayer or metal nanoparticle layer was used (Dennler
et al., 2006); metals provide an optimum recombination contact since electron and
hole quasi-Fermi energies are identical, meaning that no free energy is lost during
recombination.
However, their broad parasitic absorption is a significant disadvantage, as is
the need to interrupt solution processing to deposit thermally evaporated metals.
More recently, evaporated devices have employed a heavily doped organic p-n
junction to fulfil this role (Timmreck et al., 2010), whereas solution-processed
devices have utilised transparent metal oxides such as TiOx and ZnO in combination
with PEDOT:PSS as the recombination layer (Li et al., 2013); both routes exhibit
negligible free-energy losses, and also provide interconnects which are transparent
in the visible range. In a solution-processed inverted structure, the oxide layer can
also act to ‘reset’ the solubility of the stack, allowing repeat use of solvents for
each sub-cell.
The complexity of tandem fabrication means that research-scale structures
remain relatively rare, and developing tandems with efficiencies approaching the
sum of their sub-cells is a significant challenge. However, there are now sufficient
examples of success to confirm that such an approach can bear fruit. As a result,
the discovery of new low-bandgap tandem-ready materials has proceeded at pace
(Bronstein et al., 2011; Zhong et al., 2013), with the understanding that meth-
ods of tandem cell fabrication can be developed in parallel, and await a mature
materials set.

8.5.3 Hybrid photovoltaics

One of the most significant functional problems for organic solar cells is their low
dielectric constant. This enhances geminate recombination in almost all materials
systems except those incorporating fullerene-based electron acceptors. However,
as we have already noted, fullerenes are poorly absorbing and can be expen-
sive to produce. One promising approach is therefore to replace fullerenes with
a high-dielectric constant inorganic acceptor, creating a ‘hybrid’ solar cell, which
combines the advantages of a highly absorbing organic donor with a stable and
highly conductive acceptor. This approach is closely related to that taken by dye-
sensitised solar cells (Hagfeldt et al., 2010).
The commonly used acceptors are TiO2 and ZnO, which may be deposited
from a precursor or grown in situ. Since an organic donor is still required, a BHJ
architecture must be used to ensure efficient exciton harvesting. It is not usually
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 394

394 D. Credgington

possible to form a hybrid heterojunction by in situ self-assembly, as it is for organic

systems, though routes do exist (Dowland et al., 2011). The inorganic material is
therefore either pre-formed into particles which are deposited within an organic
matrix (Beek et al., 2004), or deposited first to form a mesoporous or nanostructured
scaffold around which a soluble organic is deposited (Peiro et al., 2006). The
latter process renders hybrid devices very well suited for top-down heterojunction
patterning.
While stable and relatively easy to produce, oxide semiconductors absorb pri-
marily in the UV and so do not improve the spectral response of the solar cell. An
alternative is to use semiconducting nanocrystals, the absorption of which can be
tuned by controlling both the nanocrystal composition and size. In combination
with polymers, CdSe is the most commonly studied nanocrystal-forming material
owing to its controllable and relatively straightforward synthesis. The first poly-
mer:nanocrystal hybrid solar cell, investigated in 1996, used CdSe as the acceptor,
although the maximum EQE achieved was only 12% (Greenham et al., 1996).
There are two main difficulties when blending isotropic nanocrystals (also referred
to as ‘quantum dots’) and polymers. The first is that as-synthesised nanocrystals
are ‘capped’ with long, solubilising ligands, which insulate the particle electri-
cally. These need to be replaced with shorter, more conducting groups (or removed
entirely) to allow charge to tunnel between particles. The second is that a balance
must be struck between nanoparticle aggregation, to create pathways for charge
transport, and dispersion, to maintain a high heterojunction area. Thus the ability to
control particle shape offers an additional route to tune device performance. Using
anisotropic rod- and tetrapod-shaped nanocrystals can therefore allow simultane-
ous formation of a charge-transporting network while achieving efficient exciton
quenching (Huynh et al., 2002; Sun et al., 2003).
In addition to CdSe, silicon nanocrystals have attracted considerable attention
in hybrid devices (Liu et al., 2008), and efforts to absorb more infrared light have
spurred interest in low-bandgap materials combinations such as CdTe and PbS
(Kumar and Nann, 2004; Watt et al., 2005).

8.5.4 Singlet fission

One option for reducing thermalisation losses when absorbing high-energy photons
is to combine wide and narrow bandgap absorbers in a tandem (or multijunction)
solar cell, as described in Section 8.5.2. However, this necessarily increases fabrica-
tion complexity. An emerging strategy is to employ an organic donor material that
undergoes spontaneous singlet fission — whereby a singlet exciton donates some
of its energy to a neighbouring chromophore, converting both to triplets, and where
it does so more rapidly than the singlet quenches at a heterojunction or emits. This
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 395

Organic Photovoltaics 395

strategy is particularly appropriate for organic semiconductors, since their tightly

confined excitons necessarily lead to an increase in exchange energies between
singlets and triplets. If both triplets can be ionised, quantum efficiencies of up to
200% are possible, and in combination with a low-bandgap absorber, theoretical
efficiencies can reach ∼50%, exceeding the Shockley–Queisser single-junction
limit (Smith and Michl, 2010).
Singlet fission was first discovered in the 1960s in polyacenes, primarily via
indirect observations, and considered a scientific curiosity (Singh et al., 1965).
More recently, singlet fission has been thought of as a means to enhance solar cell
performance, and modern ultrafast spectroscopy has allowed direct observation
of singlet fission and its kinetics. In the five-ringed polyacene pentacene, singlet
fission occurs with near unity probability, since it proceeds on timescales of around
80 fs (Wilson et al., 2011), out-competing all other decay channels including exci-
ton thermalisation and charge transfer.
Harvesting these triplet excitons requires a heterojunction, since they are
still tightly bound, with the associated heterojunction bandgap chosen such that
Etriplet > E CT . The large optical bandgap of pentacene (around 1.8 eV) requires
that it be paired with a low-bandgap absorber, typically acting as acceptor. Tune-
able low-bandgap quantum-dot acceptors such as PbS and PbSe are particularly
suited in this role, allowing absorption into the infrared and leading to efficient
hybrid devices (Ehrler et al., 2012b; Ehrler et al., 2012c). The implementation of
such a hybrid singlet-fission device is illustrated in Fig. 8.22, which bears similar-
ities with a tandem device. By analogy with the requirement for current matching
between sub-cells in a tandem device, here we require voltage matching between
the triplet energy level of the singlet fission material with the CT energy of the
heterojunction.
While photons absorbed by pentacene do contribute to the photocurrent
of such devices, internal quantum efficiencies currently remain below those of
pentacene:fullerene devices and well below the maximum possible efficiencies,
emphasising that this approach is in its relative infancy.
Singlet fission also suggests a way to increase the performance of existing
low-bandgap solar cells, such as those made from silicon, as it allows for better
utilisation of high-energy photons. The first efforts towards dissociating triplets
generated via singlet fission in this way have recently been achieved using amor-
phous silicon (Ehrler et al., 2012a).

8.5.5 Multi-component systems

The complexity of binary blends has been sufficient to drive OPV research for the
last two decades, and may yet result in the optimum materials system. However,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 396

396 D. Credgington

Figure 8.22 Singlet-fission processes and triplet harvesting. a) Schematic of indirect (blue)
and direct (green) pathways for the fission of one singlet into two triplets, which are arranged
on neighbouring chromophores in a pure singlet configuration. Adapted from Smith and
Michl (2010). b) Photogeneration in a singlet fission device: i) singlet exciton formation
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 397

Organic Photovoltaics 397

Figure 8.22 (continued from caption on previous page) in the donor following absorption
of a Singlet-fission processes and triplet harvesting. a) Schematic of indirect (blue) and
direct (green) pathways for the fission of one singlet into two triplets, which are arranged on
neighbouring chromophores in a pure singlet configuration. Adapted from Smith and Michl
(2010). b) Photogeneration in a singlet fission device: i) singlet exciton formation in the
donor following absorption of a photon with energy greater than the donor bandgap E D ; ii)
singlet fission resulting in two coupled triplet excitons; iii) diffusion of triplet excitons to the
heterojunction forming two triplet CT states, which subsequently dissociate; iv) collection of
dissociated charges arising from the dissociated triplets and absorption of low-energy pho-
tons in the acceptor. Free-carrier generation is typically spontaneous in inorganic acceptors.
Solid lines represent single-carrier molecular energy levels. Dotted lines represent excitonic
‘virtual’ energy levels where the exciton binding energy has been schematically included.
Dashed lines indicate coulombically bound carriers. Interlayers and competing processes
have been ignored for clarity.

there is in principle no additional processing complexity involved in preparing

an ink with three or more functional components. Early work on small-molecule
additives was initially directed towards incorporating these materials into the het-
erointerface, though they are now usually seen primarily as co-solvents. Neverthe-
less, such a process is still viable and can produce ternary systems.
Incorporation of block copolymers into a two-component blend has also been
attempted as a means of interface control. Spontaneous formation of polymer-rich
‘skin’ layers is commonly observed during film deposition, suggesting that modi-
fication of the electrode interfaces may be possible by incorporating materials that
preferentially wet the electrode surface. Even more simply, combining components
of different regioregularity or molecular weight offers a route to controlling the
relative crystallinity within a blend (Campoy-Quiles et al., 2009). More so than any
other sub-field of OPV research, the parameter space for multi-component systems
remains largely unexplored.

8.5.6 Scale-up
The increasing activity of industrial research groups in the OPV field is paral-
leled by increasing focus on applying scalable thin-film coating techniques to their
manufacture, and developing design rules for taking lab-scale devices to large
scale production. While much of this activity occurs within the commercial sector,
meaning that limited information is publicly available, scaling up OPV technology
requires several key transitions. We briefly outline these here, and refer the reader
to the growing literature on OPV scale-up for further details (Krebs, 2009; Krebs
et al., 2009).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 398

398 D. Credgington

Scalable deposition
The first such transition is to move away from small-area serial production tech-
niques such as spin-casting. While suitable for lab-scale devices, this requires rela-
tively high solution wastage. In addition, variations in device properties result from
changes in evaporation rate, wet-film thickness, shear rates and nucleation surfaces
for crystallisation, which in turn depend on the spin rate, wetting behaviour and
solution viscosity. Spin-casting can be very reproducible within a given laboratory
environment for a given material batch, but it is very difficult to know the exact
parameters of a deposition such that it can be reproduced in a different location.
Large-area, roll-to-roll (R2R) compatible techniques are therefore preferred
for scalable and reproducible manufacture, as illustrated in Fig. 8.23. Numerous
processes have been explored, including screen-printing, doctor blading, wire-bar
coating, inkjet printing, knife-over-edge coating, slot-die coating, gravure printing
and spray coating (Krebs, 2009). Roll-to-roll is particularly appealing for solution-
based techniques since in-line assays can be used to correct for variations in film
formation caused by changing ambient conditions and solution composition.
In terms of reproducibility, it may be that the advantage lies with all-vacuum
R2R processing of thermally evaporated small molecules, as has been developed
by Heliatek (2013). While thermal evaporation under vacuum is a more technically
intensive route than solution processing, provided that all processing steps can be

Figure 8.23 A roll of printed OPV modules produced in a continuous R2R process. The
inset shows the layer stack, comprising PET-ITO-ZnO-P3HT:PCBM-PEDOT:PSS-printed
silver paint and encapsulated using R2R lamination post-production. Reproduced from
Krebs et al. (2010).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 399

Organic Photovoltaics 399

achieved without breaking vacuum there is little fundamental difference between

the two methods. Solution processing allows for more complex thermodynamic
control of active layer deposition, and a wider range of available materials, whereas
thermal evaporation allows for tighter control of multilayers and higher material
purity. Furthermore, lessons learned from lab-scale evaporated devices may be
directly translated to R2R, whereas the need to limit the use of environmentally
damaging heavy metal catalysts and halogenated solvents may lead to a more
difficult translation from lab-scale solution processing.
In addition to re-optimising device production for new deposition processes,
a transition to flexible substrates with flexible electrodes and encapsulation is
required. Current lab-scale devices tend to employ glass-on-glass encapsulation,
which offers exceptionally low permeability to oxygen and water. Flexible bar-
rier layers of similar quality and low enough cost remain elusive, and are seen as a
significant challenge for achieving long module (rather than just material) lifetimes.

Large-area devices
The primary limit on efficient scale-up of OPV module size is series resistance.
Since organic semiconductors are low-mobility materials, it is not currently fea-
sible to employ laterally spaced high-conductivity electrodes, which are used in
crystalline solar cells to collect carriers while allowing the passage of light. It is
thus necessary to employ at least one transparent electrode with complete coverage
of the photoactive layer. The only material yet found to combine high optical trans-
mission (>80%) with acceptable sheet resistance (<15ohm/square) and a scalable
deposition route is ITO.
The relatively high sheet resistance of ITO requires that modules are formed
of multiple strips of photoactive material connected in series. Roll-to-roll print-
ing techniques can reliably ensure electrode spacing of 1–2 mm between strips,
and strip widths of ∼15 mm are empirically found to optimise between the fill
factor and light-harvesting efficiency. For this reason, lab-scale devices with area
>1 cm × 1 cm are considered to give a reasonable indication of the achievable per-
formance of a larger-area module. This architecture has disadvantages, however,
since the intercell gap represents dead module area and shading of a single strip can
lead to a significant drop in module performance unless additional bypass diodes
are used.
Because solution-processed ITO of sufficient quality is not currently available,
ITO is typically pre-sputtered onto PET film and the composite used as the web
for R2R printing. Such an approach has significant drawbacks since indium is
expensive and sputtering is a relatively costly deposition process; ITO/PET is
currently estimated to comprise half of the cost of large-area OPV manufacture.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 400

400 D. Credgington

By comparison, the polymers and fullerene photoactive components account for

only around 10% of module cost, despite the lack of a sufficiently mature materials
set to drive a reduction in their manufacturing costs (Nielsen et al., 2010).
As a result, there is a significant drive to find cheap ITO replacements for
transparent electrodes. Promising candidates include conducting polymers simi-
lar to PEDOT:PSS, graphene, carbon nanotubes, composites of highly conductive
metallic nanowire networks dispersed in a lower-conductivity matrix and alterna-
tive transparent conducting oxides.

8.6 Conclusions
In this chapter we have traced the progress of organic photovoltaics from their early
development, through the various routes found to increase their performance, to
the state-of-the-art. We have set out the current scientific understanding and engi-
neering progress in the use of organic semiconductors in thin-film solar cells,
and highlighted both where there is broad consensus on the electronic processes
involved in light absorption, charge generation, transport and collection, and where
uncertainty still remains. This growing understanding and continued technologi-
cal achievement indicate that excellent performance is achievable (towards that
of current crystalline silicon technology). This will require continued engineering
development to achieve long-lasting, highly efficient, mass-produced devices, but
recent rapid advances give cause for great optimism. The most promising outlook
comes from the sheer size of the space still to explore. Crystalline silicon has been
a stable materials platform for four decades, yet the main driver of increased OPV
performance has been the continual discovery of new organic semiconductors.
There is no indication yet that the ideal OPV materials system has been found, and
until it has, the process of true engineering optimisation remains ahead. Twenty
years from their first discovery, there is every reason to believe that organic pho-
tovoltaics have a bright future.

Acknowledgement
Compiling this chapter would not have been possible without many fruitful dis-
cussions between myself and Professor Sir Richard Friend, who has provided
invaluable insights from his long experience working in the organic electronics
field, and whom I acknowledge unreservedly.

References
Aernouts T., Vanlaeke P., Geens W., Poortmans J., Heremans P., Borghs S., Mertens R.,
Andriessen R. and Leenders L. (2004), ‘Printable anodes for flexible organic solar cell
modules’, Thin Solid Films 451, 22–25.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 401

Organic Photovoltaics 401

Albrecht S., Schindler W., Kurpiers J., Kniepert J., Blakesley J. C., Dumsch I., Allard S.,
Fostiropoulos K., Scherf U. and Neher D. (2012), ‘On the field dependence of free
charge carrier generation and recombination in blends of pcpdtbt/PC70BM: influence
of solvent additives’, J. Phys. Chem. Lett. 3, 640–645.
Andersson L. M., Müller C., Badada B. H., Zhang F., Würfel U. and Inganäs O. (2011),
‘Mobility and fill factor correlation in geminate recombination limited solar cells’,
J. Appl. Phys. 110, 024509.
Bakulin A. A., Rao A., Pavelyev V. G., van Loosdrecht P. H., Pshenichnikov M. S.,
Niedzialek D., Cornil J., Beljonne D. and Friend R. H. (2012), ‘The role of driv-
ing energy and delocalized states for charge separation in organic semiconductors’,
Science 335, 1340–1344.
Ballantyne A. M., Chen L. C., Nelson J., Bradley D. D. C., Astuti Y., Maurano A., Shuttle
C. G., Durrant J. R., Heeney M., Duffy W. and McCulloch I. (2007), ‘Studies of highly
regioregular poly(3-hexylselenophene) for photovoltaic applications’, Adv. Mater. 19,
4544–4547.
Bässler H. (1993), ‘Charge transport in disordered organic photoconductors a Monte Carlo
simulation study’, Phys. Status Solidi B 175, 15–56.
Bässler H. and Kohler A. (2012), ‘Charge transport in organic semiconductors’, Top. Curr.
Chem. 312, 1–65.
Beek W. J. E., Wienk M. M. and Janssen R. A. J. (2004), ‘Efficient hybrid solar cells from
zinc oxide nanoparticles and a conjugated polymer’, Adv. Mater. 16, 1009–1013.
Blakesley J. C. and Neher D. (2011), ‘Relationship between energetic disorder and open-
circuit voltage in bulk heterojunction organic solar cells’, Phys. Rev. B 84, 075210.
Brabec C. J., Shaheen S. E., Winder C., Sariciftci N. S. and Denk P. (2002), ‘Effect of
LiF/metal electrodes on the performance of plastic solar cells’, App. Phys. Lett. 80,
1288–1290.
Bradley D. D. C. (1987), ‘Precursor-route poly( p-phenylenevinylene): Polymer character-
isation and control of electronic properties’, J. Phys. D: Appl. Phys. 20, 1389–1410.
Braun C. L. (1984), ‘Electric field assisted dissociation of charge transfer states as a mech-
anism of photocarrier production’, J. Chem. Phys. 80, 4157–4161.
Brédas J.-L., Norton J. E., Cornil J. and Coropceanu V. (2009), ‘Molecular understanding
of organic solar cells: the challenges’, Acc. Chem. Res. 42, 1691–1699.
Bronstein H., Chen Z., Ashraf R. S., Zhang W., Du J., Durrant J. R., Shakya Tuladhar
P., Song K., Watkins S. E., Geerts Y., Wienk M. M., Janssen R. A. J., Anthopoulos
T., Sirringhaus H., Heeney M. and McCulloch I. (2011), ‘Thieno[3,2-b]thiophene–
diketopyrrolopyrrole-containing polymers for high-performance organic field-effect
transistors and organic photovoltaic devices’, J. Amer. Chem. Soc. 133, 3272–3275.
Cacialli F. and Bruschi P. (1996), ‘Site-selective chemical-vapor-deposition of submicron-
wide conducting polypyrrole films: morphological investigations with the scanning
electron and the atomic force microscope’, J. Appl. Phys. 80, 70–75.
Campoy-Quiles M., Ferenczi T., Agostinelli T., Etchegoin P. G., Kim Y., Anthopoulos T.
D., Stavrinou P. N., Bradley D. D. C. and Nelson J. (2008), ‘Morphology evolution
via self-organization and lateral and vertical diffusion in polymer:fullerene solar cell
blends’, Nat. Mater. 7, 158–164.
Campoy-Quiles M., Kanai Y., El-Basaty A., Sakai H. and Murata H. (2009), ‘Ternary mixing:
a simple method to tailor the morphology of organic solar cells’, Org. Electronics 10,
1120–1132.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 402

402 D. Credgington

Carsten B., Szarko J. M., Son H. J., Wang W., Lu L., He F., Rolczynski B. S., Lou S. J.,
Chen L. X. and Yu L. (2011), ‘Examining the effect of the dipole moment on charge
separation in donor–acceptor polymers for organic photovoltaic applications’, J. Amer.
Chem. Soc. 133, 20468–20475.
Chambon S., Rivaton A., Gardette J.-L., Firon M. and Lutsen L. (2007), ‘Aging of a donor
conjugated polymer: photochemical studies of the degradation of poly[2-methoxy-5-
(3 , 7 -dimethyloctyloxy)-1,4-phenylenevinylene]’, J. Polymer Sci. Part A: Polymer
Chem. 45, 317–331.
Chappell J., Lidzey D. G., Jukes P. C., Higgins A. M., Thompson R. L., O’Connor S., Grizzi
I., Fletcher R., O’Brien J., Geoghegan M. and Jones R. A. L. (2003), ‘Correlating
structure with fluorescence emission in phase-separated conjugated-polymer blends’,
Nat. Mater. 2, 616–621.
Charas A., Alves H., Alcacer L. and Morgado J. (2006), ‘Use of cross-linkable polyfluorene
in the fabrication of multilayer polyfluorene-based light-emitting diodes with improved
efficiency’, Appl. Phys. Lett. 89, 143519.
Chirvase D., Parisi J., Hummelen J. C. and Dyakonov V. (2004), ‘Influence of nanomor-
phology on the photovoltaic action of polymer-fullerene composites’, Nanotechnology
15, 1317.
Chou S. Y., Krauss P. R. and Renstrom P. J. (1996), ‘Imprint lithography with 25-nanometer
resolution’, Science 272, 85–87.
Clark J., Chang J. F., Spano F. C., Friend R. H. and Silva C. (2009), ‘Determining exci-
ton bandwidth and film microstructure in polythiophene films using linear absorption
spectroscopy’, Appl. Phys. Lett. 94, 163306.
Clarke T. M. and Durrant J. R. (2010), ‘Charge photogeneration in organic solar cells’,
Chem. Rev. 110, 6736–6767.
Credgington D. and Durrant J. R. (2012), ‘Insights from transient optoelectronic analyses
on the open-circuit voltage of organic solar cells’, J. Phys. Chem. Lett. 3, 1465–1478.
Credgington D., Fenwick O., Charas A., Morgado J., Suhling K. and Cacialli F. (2010),
‘High-resolution scanning near-field optical lithography of conjugated polymers’, Adv.
Funct. Mater. 20, 2842–2847.
Credgington D., Jamieson F. C., Walker B., Nguyen T.-Q. and Durrant J. R. (2012), ‘Quantifi-
cation of geminate and non-geminate recombination losses within a solution-processed
small-molecule bulk heterojunction solar cell’, Adv. Mater. 24, 2135–2141.
Crispin A., Crispin X., Fahlman M., Berggren M. and Salaneck W. R. (2006), ‘Transition
between energy level alignment regimes at a low band gap polymer-electrode inter-
faces’, Appl. Phys. Lett. 89, 213503.
de Jong M. P., van Ijzendoorn L. J. and de Voigt M. J. A. (2000), ‘Stability of the inter-
face between indium-tin-oxide and poly(3,4-ethylenedioxythiophene)/poly (styrene-
sulfonate) in polymer light-emitting diodes’, Appl. Phys. Lett. 77, 2255–2257.
Deibel C. and Dyakonov V. (2010), ‘Polymer:fullerene bulk heterojunction solar cells’, Rep.
Progr. Phys. 73, 096401.
Delacote G. M., Fillard J. P. and Marco F. J. (1964), ‘Electron injection in thin films of
copper phtalocyanine’, Solid State Commun. 2, 373–376.
Dennler G., Prall H.-J., Koeppe R., Egginger M., Autengruber R. and Sariciftci N. S.
(2006), ‘Enhanced spectral coverage in tandem organic solar cells’, Appl. Phys. Lett.
89, 073502.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 403

Organic Photovoltaics 403

Dennler G., Scharber M. C. and Brabec C. J. (2009), ‘Polymer-fullerene bulk-heterojunction

solar cells’, Adv. Mater. 21, 1323–1338.
Dexter D. L. (1953), ‘A theory of sensitized luminescence in solids’, J. Chem. Phys. 21,
836–850.
Dick D., Wei X., Jeglinski S., Benner R. E., Vardeny Z. V., Moses D., Srdanov V. I. and
Wudl F. (1994), ‘Transient spectroscopy of excitons and polarons in C60 films from
femtoseconds to milliseconds’, Phys. Rev. Lett. 73, 2760–2763.
Dimitrov S. D., Nielsen C. B., Shoaee S., Shakya Tuladhar P., Du J., McCulloch I. and
Durrant J. R. (2012), ‘Efficient charge photogeneration by the dissociation of PC70BM
excitons in polymer/fullerene solar cells’, J. Phys. Chem. Lett. 3, 140–144.
Dong B., Lu N., Zelsmann M., Kehagias N., Fuchs H., Sotomayor Torres C. M. and Chi L. F.
(2006), ‘Fabrication of high-density, large-area conducting-polymer nanostructures’,
Adv. Funct. Mater. 16, 1937–1942.
Dowland S., Lutz T., Ward A., King S. P., Sudlow A., Hill M. S., Molloy K. C. and Haque S.
A. (2011), ‘Direct growth of metal sulfide nanoparticle networks in solid-state polymer
films for hybrid inorganic–organic solar cells’, Adv. Mater. 23, 2739–2744.
Drees M., Hoppe H., Winder C., Neugebauer H., Sariciftci N. S., Schwinger W., Schaf-
fler F., Topf C., Scharber M. C., Zhu Z. and Gaudiana R. (2005), ‘Stabilization of
the nanomorphology of polymer-fullerene “bulk heterojunction” blends using a novel
polymerizable fullerene derivative’, J. Mater. Chem. 15, 5158–5163.
Dunlap D. H., Parris P. E. and Kenkre V. M. (1996), ‘Charge-dipole model for the universal
field dependence of mobilities in molecularly doped polymers’, Phys. Rev. Lett. 77,
542–545.
Dupont (2009), ‘DuPont displays surpasses million hour milestone for lifetime of new
OLED material’, 27 May 2009, www2.dupont.com.
Ehrler B., Musselman K. P., Bohm M. L., Friend R. H. and Greenham N. C. (2012a), ‘Hybrid
pentacene/a-silicon solar cells utilizing multiple carrier generation via singlet exciton
fission’, Appl. Phys. Lett. 101, 153507.
Ehrler B., Walker B. J., Böhm M. L., Wilson M. W. B., Vaynzof Y., Friend R. H. and
Greenham N. C. (2012b), ‘In situ measurement of exciton energy in hybrid singlet-
fission solar cells’, Nature Commun. 3, 1019.
Ehrler B., Wilson M. W., Rao A., Friend R. H. and Greenham N. C. (2012c), ‘Singlet exciton
fission-sensitized infrared quantum dot solar cells’, Nano Letters 12, 1053–1057.
Etzold F., Howard I. A., Mauer R., Meister M., Kim T.-D., Lee K.-S., Baek N. S. and Laquai F.
(2011), ‘Ultrafast exciton dissociation followed by nongeminate charge recombination
in PCDTBT:PCBM photovoltaic blends’, J. Amer. Chem. Soc. 133, 9469–9479.
Faist M. A., Kirchartz T., Gong W., Ashraf R. S., McCulloch I., de Mello J. C., Ekins-
Daukes N. J., Bradley D. D. C. and Nelson J. (2012), ‘Competition between the charge
transfer state and the singlet states of donor or acceptor limiting the efficiency in
polymer:fullerene solar cells’, J. Amer. Chem. Soc. 134, 685–692.
Fenwick O., Bozec L., Credgington D., Hammiche A., Lazzerini G. M., Silberberg Y. R.
and Cacialli F. (2009), ‘Thermochemical nanopatterning of organic semiconductors’,
Nature Nanotechnology 4, 664–668.
Flory P. J. (1941), ‘Thermodynamics of high polymer solutions’, J. Chem. Phys. 9, 660–661.
Förster T. (1948), ‘Zwischenmolekulare energiewanderung und fluoreszenz’, Annalen der
Physik 437, 55–75.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 404

404 D. Credgington

Frenkel J. (1931), ‘On the transformation of light into heat in solids. I’, Phys. Rev. 37, 17–44.
Gélinas S., Paré-Labrosse O., Brosseau C.-N., Albert-Seifried S., McNeill C. R., Kirov
K. R., Howard I. A., Leonelli R., Friend R. H. and Silva C. (2011), ‘The binding
energy of charge-transfer excitons localized at polymeric semiconductor heterojunc-
tions’, J. Phys. Chem. C 115, 7114–7119.
Giebink N. C., Wiederrecht G. P., Wasielewski M. R. and Forrest S. R. (2011), ‘Thermody-
namic efficiency limit of excitonic solar cells’, Phys. Rev. B 83, 195326.
Glatthaar M., Riede M., Keegan N., Sylvester-Hvid K., Zimmermann B., Niggemann M.,
Hinsch A. and Gombert A. (2007), ‘Efficiency limiting factors of organic bulk het-
erojunction solar cells identified by electrical impedance spectroscopy’, Solar Energy
Mater. Solar Cells 91, 390–393.
Gonzalez-Rabade A., Morteani A. C. and Friend R. H. (2009), ‘Correlation of heterojunction
luminescence quenching and photocurrent in polymer-blend photovoltaic diodes’, Adv.
Mater. 21, 3924–3927.
Grancini G., Maiuri M., Fazzi D., Petrozza A., Egelhaaf H. J., Brida D., Cerullo G. and
Lanzani G. (2013), ‘Hot exciton dissociation in polymer solar cells’, Nat. Mater. 12,
29–33.
Granström M. (1998), ‘Micropatterned luminescent polymer films’, Acta Polymerica 49,
514–517.
Green M. A., Emery K., Hishikawa Y., Warta W. and Dunlop E. D. (2013), ‘Solar cell
efficiency tables (version 41)’, Progr. Photovoltaics 21, 1–11.
Greenham N. C., Peng X. and Alivisatos A. P. (1996), ‘Charge separation and transport in
conjugated-polymer/semiconductor-nanocrystal composites studied by photolumines-
cence quenching and photoconductivity’, Phys. Rev. B 54, 17628–17637.
Grisorio R., Allegretta G., Mastrorilli P. and Suranna G. P. (2011), ‘On the degradation pro-
cess involving polyfluorenes and the factors governing their spectral stability’, Macro-
molecules 44, 7977–7986.
Hagfeldt A., Boschloo G., Sun L., Kloo L. and Pettersson H. (2010), ‘Dye-sensitized solar
cells’, Chem. Rev. 110, 6595–6663.
Halls J. J. M., Pichler K., Friend R. H., Moratti S. C. and Holmes A. B. (1996), ‘Exciton
diffusion and dissociation in a poly( p-phenylenevinylene)/C60 heterojunction photo-
voltaic cell’, Appl. Phys. Lett. 68, 3120–3122.
Halls J. J. M., Walsh C. A., Greenham N. C., Marseglia E. A., Friend R. H., Moratti S. C. and
Holmes A. B. (1995), ‘Efficient photodiodes from interpenetrating polymer networks’,
Nature 376, 498–500.
He Z., Zhong C., Su S., Xu M., Wu H. and Cao Y. (2012), ‘Enhanced power-conversion
efficiency in polymer solar cells using an inverted device structure’, Nat. Photonics 6,
591–595.
Heliatek (2013), ‘Heliatek consolidates its technology leadership by establishing a new
world record for organic solar technology with a cell efficiency of 12%’, 16 January
2013, www.heliatek.com.
Huggins M. L. (1941), ‘Solutions of long chain compounds’, J. Chem. Phys. 9, 440.
Huynh W. U., Dittmer J. J. and Alivisatos A. P. (2002), ‘Hybrid nanorod-polymer solar
cells’, Science 295, 2425–2427.
Irwin M. D., Buchholz B., Hains A. W., Chang R. P. H. and Marks T. J. (2008),
‘P-type semiconducting nickel oxide as an efficiency-enhancing anode interfacial
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 405

Organic Photovoltaics 405

layer in polymer bulk-heterojunction solar cells’, Proc. Natl. Acad. Sci. U S A 105,
2783–2787.
Jamieson F. C., Domingo E. B., McCarthy-Ward T., Heeney M., Stingelin N. and Durrant
J. R. (2012), ‘Fullerene crystallisation as a key driver of charge separation in poly-
mer/fullerene bulk heterojunction solar cells’, Chem. Sci. 3, 485–492.
Jørgensen M., Norrman K., Gevorgyan S. A., Tromholt T., Andreasen B. and Krebs F. C.
(2012), ‘Stability of polymer solar cells’, Adv. Mater. 24, 580–612.
Kallmann H. and Pope M. (1959), ‘Photovoltaic effect in organic crystals’, J. Chem. Phys.
30, 585–586.
Kameoka J., Czaplewski D., Liu H. and Craighead H. G. (2004), ‘Polymeric nanowire
architecture’, J. Mater. Chem. 14, 1503–1505.
Kearns D. and Calvin M. (1958), ‘Photovoltaic effect and photoconductivity in laminated
organic systems’, J. Chem. Phys. 29, 950–951.
Kim J. Y., Kim S. H., Lee H. H., Lee K., Ma W., Gong X. and Heeger A. J. (2006),
‘New architecture for high-efficiency polymer photovoltaic cells using solution-based
titanium oxide as an optical spacer’, Adv. Mater. 18, 572–576.
Kim Y., Cook S., Tuladhar S. M., Choulis S. A., Nelson J., Durrant J. R., Bradley D. D. C.,
Giles M., McCulloch I., Ha C.-S. and Ree M. (2006), ‘A strong regioregularity effect in
self-organizing conjugated polymer films and high-efficiency polythiophene:fullerene
solar cells’, Nat. Mater. 5, 197–203.
Kirchartz T., Pieters B. E., Kirkpatrick J., Rau U. and Nelson J. (2011), ‘Recombination via
tail states in polythiophene:fullerene solar cells’, Phys. Rev. B 83, 115209.
Kniepert J., Schubert M., Blakesley J. C. and Neher D. (2011), ‘Photogeneration and recom-
bination in P3HT/PCBM solar cells probed by time-delayed collection field experi-
ments’, J. Phys. Chem. Lett. 2, 700–705.
Kohn P., Rong Z., Scherer K. H., Sepe A., Sommer M., Müller-Buschbaum P., Friend R. H.,
Steiner U. and Hüttner S. (2013), ‘Crystallization-induced 10-nm structure formation
in P3HT/PCBM blends’, Macromolecules 46, 4002–4013.
Koster L. J. A., Mihailetchi V. D., Ramaker R. and Blom P. W. M. (2005), ‘Light intensity
dependence of open-circuit voltage of polymer:fullerene solar cells’, Appl. Phys. Lett.
86, 123509.
Krebs F. C. (2009), ‘Fabrication and processing of polymer solar cells: a review of printing
and coating techniques’, Solar Energy Mater. Solar Cells 93, 394–412.
Krebs F. C., Fyenbo J. and Jorgensen M. (2010), ‘Product integration of compact roll-to-roll
processed polymer solar cell modules: methods and manufacture using flexographic
printing, slot-die coating and rotary screen printing’, J. Mater. Chem. 20, 8994–9001.
Krebs F. C., Gevorgyan S. A. and Alstrup J. (2009), ‘A roll-to-roll process to flexible polymer
solar cells: model studies, manufacture and operational stability studies’, J. Mater.
Chem. 19, 5442–5451.
Krebs F. C. and Spanggaard H. (2005), ‘Significant improvement of polymer solar cell
stability’, Chemistry of Materials 17, 5235–5237.
Kumar S. and Nann T. (2004), ‘First solar cells based on CdTe nanoparticle/MEH-PPV
composites’, J. Mater. Res. 19, 1990–1994.
Lee J., Vandewal K., Yost S. R., Bahlke M. E., Goris L., Baldo M. A., Manca J. V. and
Voorhis T. V. (2010), ‘Charge transfer state versus hot exciton dissociation in polymer-
fullerene blended solar cells’, J. Amer. Chem. Soc. 132, 11878–11880.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 406

406 D. Credgington

Lee J. K., Ma W. L., Brabec C. J., Yuen J., Moon J. S., Kim J. Y., Lee K., Bazan G. C.
and Heeger A. J. (2008), ‘Processing additives for improved efficiency from bulk
heterojunction solar cells’, J. Amer. Chem. Soc. 130, 3619–3623.
Lee M. J., Gupta D., Zhao N., Heeney M., McCulloch I. and Sirringhaus H. (2011),
‘Anisotropy of charge transport in a uniaxially aligned and chain-extended, high-
mobility, conjugated polymer semiconductor’, Adv. Funct. Mater. 21, 932–940.
Li G., Shrotriya V., Yao Y. and Yang Y. (2005), ‘Investigation of annealing effects and film
thickness dependence of polymer solar cells based on poly(3-hexylthiophene)’, J. Appl.
Phys. 98, 043704.
Li G., Yao Y., Yang H., Shrotriya V., Yang G. and Yang Y. (2007), “‘Solvent annealing”
effect in polymer solar cells based on poly(3-hexylthiophene) and methanofullerenes’,
Adv. Funct. Mater. 17, 1636–1644.
Li J., Kim S., Edington S., Nedy J., Cho S., Lee K., Heeger A. J., Gupta M. C. and Yates Jr J. T.
(2011), ‘A study of stabilization of P3HT/PCBM organic solar cells by photochemical
active TiOx layer’, Solar Energy Mater. Solar Cells 95, 1123–1130.
Li J., Zhao Y., Tan H. S., Guo Y., Di C.-A., Yu G., Liu Y., Lin M., Lim S. H., Zhou Y., Su
H. and Ong B. S. (2012), ‘A stable solution-processed polymer semiconductor with
record high-mobility for printed transistors’, Sci. Rep. 2, 754.
Li W., Furlan A., Hendriks K. H., Wienk M. M. and Janssen R. A. J. (2013), ‘Efficient
tandem and triple-junction polymer solar cells’, J. Amer. Chem. Soc. 135, 5529–5532.
Lin C.-F., Liu S.-W., Lee C.-C., Hunag J.-C., Su W.-C., Chiu T.-L., Chen C.-T. and Lee J.-H.
(2012), ‘Open-circuit voltage and efficiency improvement of subphthalocyanine-based
organic photovoltaic device through deposition rate control’, Solar Energy Mater. Solar
Cells 103, 69–75.
Linares M., Beljonne D., Cornil J. r. m., Lancaster K., Bré das J.-L., Verlaak S., Mityashin
A., Heremans P., Fuchs A., Lennartz C., Idé J., Méreau R. L., Aurel P., Ducasse L. and
Castet F. (2010), ‘On the interface dipole at the pentacene–fullerene heterojunction: a
theoretical study’, J. Phys. Chem. C 114, 3215–3224.
Liu B., Png R.-Q., Zhao L.-H., Chua L.-L., Friend R. H. and Ho P. K. H. (2012), ‘High
internal quantum efficiency in fullerene solar cells based on crosslinked polymer donor
networks’, Nature Commun. 3, 1321.
Liu C.-Y., Holman Z. C. and Kortshagen U. R. (2008), ‘Hybrid solar cells from P3HT and
silicon nanocrystals’, Nano Letters 9, 449–452.
Liu J., Shi Y. J. and Yang Y. (2001), ‘Solvation-induced morphology effects on the perfor-
mance of polymer-based photovoltaic devices’, Adv. Funct. Mater. 11, 420–424.
Lüssem B., Riede M. and Leo K. (2013), ‘Doping of organic semiconductors’, Phys. Stat.
Solidi A 210, 9–43.
Ma W., Yang C., Gong X., Lee K. and Heeger A. J. (2005), ‘Thermally stable, efficient poly-
mer solar cells with nanoscale control of the interpenetrating network morphology’,
Adv. Funct. Mater. 15, 1617–1622.
MacKenzie R. C. I., Shuttle C. G., Chabinyc M. L. and Nelson J. (2012), ‘Extracting micro-
scopic device parameters from transient photocurrent measurements of P3HT:PCBM
solar cells’, Adv. Energy Mater. 2, 662–669.
Manceau M., Bundgaard E., Carle J. E., Hagemann O., Helgesen M., Sondergaard R.,
Jorgensen M. and Krebs F. C. (2011), ‘Photochemical stability of small π-conjugated
polymers for polymer solar cells: a rule of thumb’, J. Mater. Chem. 21, 4132–4141.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 407

Organic Photovoltaics 407

Manceau M., Chambon S., Rivaton A., Gardette J.-L., Guillerez S. and Lemaître N. (2010a),
‘Effects of long-term uv-visible light irradiation in the absence of oxygen on P3HT
and P3HT:PCBM blend’, Solar Energy Mater. Solar Cells 94, 1572–1577.
Manceau M., Helgesen M. and Krebs F. C. (2010b), ‘Thermo-cleavable polymers: materials
with enhanced photochemical stability’, Polym. Degrad. Stabil. 95, 2666–2669.
Manceau M., Rivaton A., Gardette J.-L., Guillerez S. and Lemaître N. (2009), ‘The mecha-
nism of photo- and therMoOxidation of poly(3-hexylthiophene) (P3HT) reconsidered’,
Polym. Degrad. Stabil. 94, 898–907.
Mandoc M. M., Veurman W., Koster L. J. A., de Boer B. and Blom P. W. M. (2007), ‘Origin
of the reduced fill factor and photocurrent in MDMO-PPV:PCNEPV all-polymer solar
cells’, Adv. Funct. Mater. 17, 2167–2173.
Marsh R. A., Hodgkiss J. M. and Friend R. H. (2010), ‘Direct measurement of electric
field-assisted charge separation in polymer:fullerene photovoltaic diodes’, Adv. Mater.
22, 3672–3676.
Martens H. C. F., Brom H. B. and Blom P. W. M. (1999), ‘Frequency-dependent electrical
response of holes in poly( p-phenylene vinylene)’, Phys. Rev. B 60, R8489–R8492.
Meilanov I. S., Benderskii V. A. and Bliumenfel’d L. A. (1970a), ‘Photoelectric properties
of chlorophyll a and b layers. I. Photocurrents during constant illumination’, Biofizika
15, 822–827.
Meilanov I. S., Benderskii V. A. and Bliumenfel’d L. A. (1970b), ‘Photoelectric properties
of chlorophyll a and b layers. II. Photocurrents during impulse illumination’, Biofizika
15, 958–964.
Mihailetchi V. D., Blom P. W. M., Hummelen J. C. and Rispens M. T. (2003), ‘Cathode
dependence of the open-circuit voltage of polymer:fullerene bulk heterojunction solar
cells’, J. Appl. Phys. 94, 6849–6854.
Mihailetchi V. D., Koster L. J. A. and Blom P. W. M. (2004), ‘Effect of metal electrodes
on the performance of polymer:fullerene bulk heterojunction solar cells’, Appl. Phys.
Lett. 85, 970–972.
Mihailetchi V. D., Wildeman J. and Blom P. W. M. (2005), ‘Space-charge limited photocur-
rent’, Phys. Rev. Lett. 94, 126602.
Miller A. and Abrahams E. (1960), ‘Impurity conduction at low concentrations’, Phys. Rev.
120, 745–755.
Morana M., Azimi H., Dennler G., Egelhaaf H.-J., Scharber M., Forberich K., Hauch J.,
Gaudiana R., Waller D., Zhu Z., Hingerl K., van Bavel S. S., Loos J. and Brabec C. J.
(2010), ‘Nanomorphology and charge generation in bulk heterojunctions based on low-
bandgap dithiophene polymers with different bridging atoms’, Adv. Funct. Mater. 20,
1180–1188.
Moulé A. J. and Meerholz K. (2008), ‘Controlling morphology in polymer–fullerene mix-
tures’, Adv. Mater. 20, 240–245.
Mühlbacher D., Scharber M., Morana M., Zhu Z., Waller D., Gaudiana R. and Brabec C.
(2006), ‘High photovoltaic performance of a low-bandgap polymer’, Adv. Mater. 18,
2884–2889.
Najari A., Berrouard P., Ottone C., Boivin M., Zou Y., Gendron D., Caron W.-O., Legros
P., Allen C. N., Sadki S. and Leclerc M. (2012), ‘High open-circuit voltage solar cells
based on new thieno[3,4-c]pyrrole-4,6-dione and 2,7-carbazole copolymers’, Macro-
molecules 45, 1833–1838.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 408

408 D. Credgington

Nelson J. (2003), ‘Diffusion-limited recombination in polymer-fullerene blends and its

influence on photocurrent collection’, Phys. Rev. B 67, 155209.
Neugebauer H., Brabec C., Hummelen J. C. and Sariciftci N. S. (2000), ‘Stability and
photodegradation mechanisms of conjugated polymer/fullerene plastic solar cells’,
Solar Energy Mater. Solar Cells 61, 35–42.
Nguyen L. H., Hoppe H., Erb T., Günes S., Gobsch G. and Sariciftci N. S. (2007), ‘Effects of
annealing on the nanomorphology and performance of poly(alkylthiophene):fullerene
bulk-heterojunction solar cells’, Adv. Funct. Mater. 17, 1071–1078.
Nielsen T. D., Cruickshank C., Foged S., Thorsen J. and Krebs F. C. (2010), ‘Business,
market and intellectual property analysis of polymer solar cells’, Solar Energy Mater.
Solar Cells 94, 1553–1571.
Onsager L. (1934), ‘Deviations from Ohm’s law in weak electrolytes’, J. Chem. Phys. 2,
599–615.
Ooi Z. E., Tam T. L., Sellinger A. and deMello J. C. (2008), ‘Field-dependent carrier
generation in bulk heterojunction solar cells’, Energy Environ. Sci. 1, 300–309.
Padinger F., Rittberger R. S. and Sariciftci N. S. (2003), ‘Effects of postproduction treatment
on plastic solar cells’, Adv. Funct. Mater. 13, 85–88.
Park S. H., Roy A., Beaupre S., Cho S., Coates N., Moon J. S., Moses D., Leclerc M., Lee
K. and Heeger A. J. (2009), ‘Bulk heterojunction solar cells with internal quantum
efficiency approaching 100%’, Nat. Photonics 3, 297–302.
Park Y. D., Kim D. H., Jang Y., Cho J. H., Hwang M., Lee H. S., Lim J. A. and Cho K.
(2006), ‘Effect of side chain length on molecular ordering and field-effect mobility in
poly(3-alkylthiophene) transistors’, Org. Electronics 7, 514–520.
Pasveer W. F., Cottaar J., Tanase C., Coehoorn R., Bobbert P. A., Blom P. W. M., de Leeuw
D. M. and Michels M. A. J. (2005), ‘Unified description of charge-carrier mobilities
in disordered semiconducting polymers’, Phys. Rev. Lett. 94, 206601.
Peet J., Kim J. Y., Coates N. E., Ma W. L., Moses D., Heeger A. J. and Bazan G. C. (2007),
‘Efficiency enhancement in low-bandgap polymer solar cells by processing with alkane
dithiols’, Nat. Mater. 6, 497–500.
Peiro A. M., Ravirajan P., Govender K., Boyle D. S., O’Brien P., Bradley D. D. C., Nelson
J. and Durrant J. R. (2006), ‘Hybrid polymer/metal oxide solar cells based on ZnO
columnar structures’, J. Mater. Chem. 16, 2088–2096.
Peumans P. and Forrest S. R. (2001), ‘Very-high-efficiency double-heterostructure copper
phthalocyanine/C60 photovoltaic cells’, Appl. Phys. Lett. 79, 126–128.
Peumans P. and Forrest S. R. (2004), ‘Separation of geminate charge-pairs at donor–acceptor
interfaces in disordered solids’, Chem. Phys. Lett. 398, 27–31.
Png R.-Q., Chia P.-J., Tang J.-C., Liu B., Sivaramakrishnan S., Zhou M., Khong S.-H., Chan
H. S. O., Burroughes J. H., Chua L.-L., Friend R. H. and Ho P. K. H. (2010), ‘High-
performance polymer semiconducting heterostructure devices by nitrene-mediated
photocrosslinking of alkyl side chains’, Nat. Mater. 9, 152–158.
Polyera (2013), ‘Polyera achieves 6.4% all-polymer organic solar cells’, 12 April 2013,
www.polyera.com.
Prins P., Grozema F. C., Schins J. M., Patil S., Scherf U. and Siebbeles L. D. A. (2006),
‘High intrachain hole mobility on molecular wires of ladder-type poly( p-phenylenes)’,
Phys. Rev. Lett. 96, 146601.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 409

Organic Photovoltaics 409

Reese M. O., Gevorgyan S. A., Jørgensen M., Bundgaard E., Kurtz S. R., Ginley D. S.,
Olson D. C., Lloyd M. T., Morvillo P., Katz E. A., Elschner A., Haillant O., Currier T.
R., Shrotriya V., Hermenau M., Riede M., R. Kirov K., Trimmel G., Rath T., Inganäs
O., Zhang F., Andersson M., Tvingstedt K., Lira-Cantu M., Laird D., McGuiness C.,
Gowrisanker S., Pannone M., Xiao M., Hauch J., Steim R., DeLongchamp D. M.,
Rösch R., Hoppe H., Espinosa N., Urbina A., Yaman-Uzunoglu G., Bonekamp J.-B.,
van Breemen A. J. J. M., Girotto C., Voroshazi E. and Krebs F. C. (2011), ‘Consensus
stability testing protocols for organic photovoltaic materials and devices’, Solar Energy
Mater. Solar Cells 95, 1253–1267.
Reese M. O., Nardes A. M., Rupert B. L., Larsen R. E., Olson D. C., Lloyd M. T., Shaheen
S. E., Ginley D. S., Rumbles G. and Kopidakis N. (2010), ‘Photoinduced degradation
of polymer and polymer–fullerene active layers: experiment and theory’, Adv. Funct.
Mater. 20, 3476–3483.
Riede M., Uhrich C., Widmer J., Timmreck R., Wynands D., Schwartz G., Gnehr W.-M.,
Hildebrandt D., Weiss A., Hwang J., Sundarraj S., Erk P., Pfeiffer M. and Leo K.
(2011), ‘Efficient organic tandem solar cells based on small molecules’, Adv. Funct.
Mater. 21, 3019–3028.
Rogers J. A., Bao Z., Makhija A. and Braun P. (1999), ‘Printing process suitable for reel-to-
reel production of high-performance organic transistors and circuits’, Adv. Mater. 11,
741–745.
Rogers J. A., Bao Z. and Raju V. R. (1998), ‘Nonphotolithographic fabrication of organic
transistors with micron feature sizes’, Appl. Phys. Lett. 72, 2716–2718.
Sariciftci N. S., Smilowitz L., Heeger A. J. and Wudl F. (1992), ‘Photoinduced electron
transfer from a conducting polymer to buckminsterfullerene’, Science 258, 1474–1476.
Scharber M. C., Mühlbacher D., Koppe M., Denk P., Waldauf C., Heeger A. J. and Brabec
C. J. (2006), ‘Design rules for donors in bulk-heterojunction solar cells — towards
10% energy-conversion efficiency’, Adv. Mater. 18, 789–794.
Segalman R. A., McCulloch B., Kirmayer S. and Urban J. J. (2009), ‘Block copolymers for
organic optoelectronics’, Macromolecules 42, 9205–9216.
Shaheen S. E., Brabec C. J., Sariciftci N. S., Padinger F., Fromherz T. and Hummelen J. C.
(2001), ‘2.5% efficient organic plastic solar cells’, Appl. Phys. Lett. 78, 841–843.
Shoaee S., Clarke T. M., Huang C., Barlow S., Marder S. R., Heeney M., McCulloch I.
and Durrant J. R. (2010), ‘Acceptor energy level control of charge photogeneration in
organic donor/acceptor blends’, J. Amer. Chem. Soc. 132, 12919–12926.
Shoaee S., Subramaniyan S., Xin H., Keiderling C., Tuladhar P. S., Jamieson F., Jenekhe S.
A. and Durrant J. R. (2013), ‘Charge photogeneration for a series of thiazolo-thiazole
donor polymers blended with the fullerene electron acceptors PCBM and ICBA’, Adv.
Funct. Mater. 23, 3286–3298.
Shuttle C. G., Hamilton R., O’Regan B. C., Nelson J. and Durrant J. R. (2010), ‘Charge-
density-based analysis of the current–voltage response of polythiophene/fullerene pho-
tovoltaic devices’, Proc. Natl. Acad. Sci. USA 107, 16448–16452.
Silinsh E. A., Belkind A. I., Balode D. R., Biseniece A. J., Grechov V. V., Taure L. F.,
Kurik M. V., Vertzymacha J. I. and Bok I. (1974), ‘Photoelectrical properties, energy
level spectra, and photogeneration mechanisms of pentacene’, Phys. Stat. Solidi A 25,
339–347.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 410

410 D. Credgington

Singh S., Jones W. J., Siebrand W., Stoicheff B. P. and Schneider W. G. (1965), ‘Laser
generation of excitons and fluorescence in anthracene crystals’, J. Chem. Phys. 42,
330–342.
Smith M. B. and Michl J. (2010), ‘Singlet fission’, Chem. Rev. 110, 6891–6936.
Son H. J., Wang W., Xu T., Liang Y., Wu Y., Li G. and Yu L. (2011), ‘Synthesis of flu-
orinated polythienothiophene-co-benzodithiophenes and effect of fluorination on the
photovoltaic properties’, J. Amer. Chem. Soc. 133, 1885–1894.
Stabile R., Camposeo A., Persano L., Tavazzi S., Cingolani R. and Pisignano D. (2007),
‘Organic-based distributed feedback lasers by direct electron-beam lithography on
conjugated polymers’, Appl. Phys. Lett. 91, 101110.
Steim R., Kogler F. R. and Brabec C. J. (2010), ‘Interface materials for organic solar cells’,
J. Mater. Chem. 20, 2499–2512.
Street R. A., Cowan S. and Heeger A. J. (2010), ‘Experimental test for geminate recombi-
nation applied to organic solar cells’, Phys. Rev. B 82, 121301.
Stubinger T. and Brutting W. (2001), ‘Exciton diffusion and optical interference in organic
donor–acceptor photovoltaic cells’, J. Appl. Phys. 90, 3632–3641.
Sullivan P., Schumann S., Da Campo R., Howells T., Duraud A., Shipman M., Hatton R.
A. and Jones T. S. (2013), ‘Ultra-high voltage multijunction organic solar cells for
low-power electronic applications’, Adv. Energy Mater. 3, 239–244.
Sun B., Marx E. and Greenham N. C. (2003), ‘Photovoltaic devices using blends of branched
CdSe nanoparticles and conjugated polymers’, Nano Letters 3, 961–963.
Sun Y., Welch G. C., Leong W. L., Takacs C. J., Bazan G. C. and Heeger A. J. (2012),
‘Solution-processed small-molecule solar cells with 6.7% efficiency’, Nat. Mater. 11,
44–48.
Tang C. W. (1986), ‘Two-layer organic photovoltaic cell’, Appl. Phys. Lett. 48, 183–185.
Tang C. W. and Albrecht A. C. (1975), ‘Photovoltaic effects of metal–chlorophyll-a–metal
sandwich cells’, J. Chem. Phys. 62, 2139–2149.
Tengstedt C., Osikowicz W., Salaneck W. R., Parker I. D., Hsu C.-H. and Fahlman M. (2006),
‘Fermi-level pinning at conjugated polymer interfaces’, Appl. Phys. Lett. 88, 053502.
Timmreck R., Olthof S., Leo K. and Riede M. K. (2010), ‘Highly doped layers as efficient
electron-hole recombination contacts for tandem organic solar cells’, J. Appl. Phys.
108, 033108.
Townsend P. D., Pereira C. M., Bradley D. D. C., Horton M. E. and Friend R. H. (1985),
‘Increase in chain conjugation length in highly oriented Durham-route polyacetylene’,
J. Phys. C: Solid State Physics 18, L283–L289.
Treat N. D., Brady M. A., Smith G., Toney M. F., Kramer E. J., Hawker C. J. and Chabinyc M.
L. (2011), ‘Interdiffusion of PCBM and P3HT reveals miscibility in a photovoltaically
active blend’, Adv. Energy Mater. 1, 82–89.
Tributsch H. and Calvin M. (1971), ‘Electrochemistry of excited molecules: photo-
electrochemical reactions of chlorophylls*’, Photochem. Photobiol. 14, 95–112.
Tvingstedt K., Vandewal K., Gadisa A., Zhang F., Manca J. and Inganäs O. (2009), ‘Elec-
troluminescence from charge transfer states in polymer solar cells’, J. Amer. Chem.
Soc. 131, 11819–11824.
Tvingstedt K., Vandewal K., Zhang F. and Inganás O. (2010), ‘On the dissociation efficiency
of charge transfer excitons and Frenkel excitons in organic solar cells: A luminescence
quenching study’, J. Phys. Chem C 114, 21824–21832.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 411

Organic Photovoltaics 411

Vandewal K., Gadisa A., Oosterbaan W. D., Bertho S., Banishoeib F., Van Severen I.,
Lutsen L., Cleij T. J., Vanderzande D. and Manca J. V. (2008), ‘The relation between
open-circuit voltage and the onset of photocurrent generation by charge-transfer absorp-
tion in polymer:fullerene bulk heterojunction solar cells’, Adv. Funct. Mater. 18, 2064–
2070.
Vandewal K., Tvingstedt K., Gadisa A., Inganäs O. and Manca J. V. (2009), ‘On the
origin of the open-circuit voltage of polymer-fullerene solar cells’, Nat. Mater. 8,
904–909.
Veldman D., İpek Ö., Meskers S. C. J., Sweelssen J., Koetse M. M., Veenstra S. C., Kroon
J. M., van Bavel S. S., Loos J. and Janssen R. A. J. (2008), ‘Compositional and electric
field dependence of the dissociation of charge transfer excitons in alternating polyflu-
orene copolymer/fullerene blends’, J. Amer. Chem. Soc. 130, 7721–7735.
Vitoratos E., Sakkopoulos S., Dalas E., Paliatsas N., Karageorgopoulos D., Petraki F., Ken-
nou S. and Choulis S. A. (2009), ‘Thermal degradation mechanisms of PEDOT:PSS’,
Org. Electronics 10, 61–66.
Vos A. D. (1980), ‘Detailed balance limit of the efficiency of tandem solar cells’, J. Phys.
D: Appl. Phys. 13, 839–846.
Wannier G. H. (1937), ‘The structure of electronic excitation levels in insulating crystals’,
Phys. Rev. 52, 191–197.
Watt A., Blake D., Warner J. H., Thomsen E. A., Tavenner E. L., Rubinsztein-Dunlop H.
and Meredith P. (2005), ‘Lead sulfide nanocrystal: conducting polymer solar cells’,
J. Phys. D: Appl. Phys. 38, 2006–2012.
Wetzelaer G. A. H., Koster L. J. A. and Blom P. W. M. (2011a), ‘Validity of the Einstein
relation in disordered organic semiconductors’, Phys. Rev. Lett. 107, 066605.
Wetzelaer G. A. H., Kuik M., Lenes M. and Blom P. W. M. (2011b), ‘Origin of the dark-
current ideality factor in polymer:fullerene bulk heterojunction solar cells’, Appl. Phys.
Lett. 99, 153506.
Wilson M. W. B., Rao A., Clark J., Kumar R. S. S., Brida D., Cerullo G. and Friend R. H.
(2011), ‘Ultrafast dynamics of exciton fission in polycrystalline pentacene’, J. Amer.
Chem. Soc. 133, 11830–11833.
Xin H., Reid O. G., Ren G., Kim F. S., Ginger D. S. and Jenekhe S. A. (2010), ‘Poly-
mer nanowire/fullerene bulk heterojunction solar cells: how nanostructure determines
photovoltaic properties’, ACS Nano 4, 1861–1872.
Yu G., Gao J., Hummelen J. C., Wudl F. and Heeger A. J. (1995), ‘Polymer photovoltaic
cells: enhanced efficiencies via a network of internal donor-acceptor heterojunctions’,
Science 270, 1789–1791.
Yun M., Myung N. V., Vasquez R. P., Lee C., Menke E. and Penner R. M. (2004), Electro-
chemically grown wires for individually addressable sensor arrays’, Nano Letters 4,
419–422.
Zhong H., Li Z., Deledalle F., Fregoso E. C., Shahid M., Fei Z., Nielsen C. B., Yaacobi-Gross
N., Rossbauer S., Anthopoulos T. D., Durrant J. R. and Heeney M. (2013), ‘Fused
dithienogermolodithiophene low band gap polymers for high-performance organic
solar cells without processing additives’, J. Amer. Chem. Soc. 135, 2040–2043.
Zhou Y., Fuentes-Hernandez C., Shim J., Meyer J., Giordano A. J., Li H., Winget P.,
Papadopoulos T., Cheun H., Kim J., Fenoll M., Dindar A., Haske W., Najafabadi
E., Khan T. M., Sojoudi H., Barlow S., Graham S., Brédas J.-L., Marder S. R., Kahn A.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch08 page 412

412 D. Credgington

and Kippelen B. (2012), ‘A universal method to produce low-work function electrodes

for organic electronics’, Science 336, 327–332.
Zilberberg K., Gharbi H., Behrendt A., Trost S. and Riedl T. (2012), ‘Low-temperature,
solution-processed MoOx for efficient and stable organic solar cells’, ACS Appl. Mater.
Interfaces 4, 1164–1168.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 413

CHAPTER 9

DYE- AND PEROVSKITE-SENSITISED

MESOSCOPIC SOLAR CELLS
MICHAEL GRÄTZEL
Laboratory for Photonics and Interfaces
Swiss Federal Institute of Technology
CH–1015 Lausanne, Switzerland
[email protected]

and

JAMES R. DURRANT
Department of Chemistry, Imperial College London SW7 2AZ, UK
Materials Science Centre,
College of Engineering, University of Swansea,
Swansea SA12 7Ax, UK
[email protected]

On the arid lands there will spring up industrial colonies without smoke and without
smokestacks; forests of glass tubes will extend over the plains and glass buildings will rise
everywhere; inside of these will take place the photochemical processes that hitherto have
been the guarded secret of the plants, but that will have been mastered by human industry
which will know how to make them bear even more abundant fruit than nature, for nature
is not in a hurry and mankind is. And if in a distant future the supply of coal becomes
completely exhausted, civilisation will not be checked by that, for life and civilisation will
continue as long as the Sun shines!
Giacomo Ciamician, Eighth International Congress of Applied Chemistry,
Washington and New York, September 1912.

9.1 Introduction
Photovoltaic (PV) devices are based on charge separation at an interface of two
materials of different conduction type or mechanism. To date, this field has been
dominated by solid-state junction devices, usually made of silicon, and profiting
from the experience and material availability resulting from the semiconductor
industry. The dominance of the photovoltaic field by inorganic solid-state junction

413
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 414

414 M. Grätzel & J. R. Durrant

devices is now being challenged by the emergence of a range of new device con-
cepts, including devices based on nanocrystalline and conducting polymer films.
These devices offer the prospect of very low-cost fabrication and present a range
of attractive features that will facilitate market entry. It is now possible to depart
completely from the classical solid-state junction device by replacing the phase
contacting the semiconductor by an electrolyte, liquid, gel or solid, thereby form-
ing a photoelectrochemical cell. The phenomenal progress realised recently in the
fabrication and characterisation of nanocrystalline materials has opened up vast
new opportunities for these systems. Contrary to expectation, devices based on
interpenetrating networks of mesoscopic semiconductors have shown strikingly
high conversion efficiencies, competing with those of conventional devices. The
prototype of this family of devices is the dye-sensitised solar cell (DSSC), which
realises the optical absorption and charge-separation processes by the association
of a sensitiser as light-absorbing material with a wide-bandgap semiconductor of
nanocrystalline morphology (O’Regan and Grätzel, 1991).
Recently, solid-state embodiments of mesoscopic solar have emerged where
the sensitiser is replaced by a semiconductor quantum dot (Grätzel et al., 2012)
or a perovskite pigment. Solid-state cells based on perovskites as light harvesters
have stunned the PV community by their amazing performance, certified power
conversion efficiencies having reached 14.14% only within one year after their
inception (Burschka et al., 2013).

9.2 Historical background

The history of the sensitisation of semiconductors to light of wavelength longer than
that corresponding to the semiconductor bandgap has been presented elsewhere
(McEvoy and Grätzel, 1994; Hagfeldt and Grätzel, 2000). It represents an inter-
esting convergence of photography and photoelectrochemistry, both of which rely
on photoinduced charge separation at a liquid–solid interface. The silver halides
used in photography have bandgaps of the order of 2.7–3.2 eV, and are therefore
insensitive to much of the visible spectrum, as are the metal oxide films now used
in DSSCs.
The first panchromatic film able to render the image of a scene realistically into
black and white followed on the work of Vogel in Berlin after 1873 (West, 1874),
in which he associated dyes with the halide semiconductor grains. The first sen-
sitisation of a photoelectrode followed shortly thereafter, using similar chemistry
(Moser, 1887). However, the clear recognition of the parallels between the two pro-
cedures — realisation that the same dyes can in principle function in both systems
(Namba and Hishiki, 1965) and verification that their operating mechanism is by
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 415

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 415

injection of electrons from photoexcited dye molecules into the conduction band
of the n−type semiconductor substrates (Gerischer and Tributsch, 1968) — dates
only to the 1960s. In the years that followed it became recognised that the dye
could function most efficiently if chemisorbed on the surface of the semiconduc-
tor (Tsubomura et al., 1976; Anderson et al., 1979; Dare-Edwards et al., 1980).
The concept of using dispersed particles to provide a sufficient interface area then
emerged (Dung et al., 1984), and was subsequently employed for photoelectrodes
(Desilvestro et al., 1985).
Titanium dioxide (TiO2 ) quickly became the semiconductor of choice for
the photoelectrode on account of its many advantages for sensitised photochem-
istry and photoelectron chemistry: it is a low-cost, widely available, non-toxic and
biocompatible material, and as such is even used in healthcare products as well as
industrial applications such as paint pigmentation. Initial studies of the TiO2 -based
DSSC employed tris-bipyridyl ruthenium(II) dyes, which are paradigm sensitisers
for photochemical studies, functionalised by the addition of carboxylate groups to
attach the chromophore to the oxide substrate by chemisorption. Progress there-
after was incremental, a synergy of structure, substrate roughness and morphol-
ogy, dye photophysics and electrolyte redox chemistry, until the announcement
in 1991 (O’Regan and Grätzel, 1991) of a sensitised electrochemical photovoltaic
device with an energy conversion efficiency of 7.1% under solar illumination. That
evolution has continued progressively since then, with certified efficiencies now
over 12%. These advances in efficiency have been complemented by significant
advances in processability, cost and stability which have greatly enhanced the
commercial viability of this technology.

9.3 Mode of function of dye-sensitised solar cells

9.3.1 Device configuration
Figure 9.1 is a schematic of the components of a DSSC. At the heart of the sys-
tem is a mesoporous oxide layer composed of nanometre-sized particles that have
been sintered together to allow electronic conduction to take place. The material
of choice has been TiO2 (anatase) although alternative wide-gap oxides such as
ZnO, SnO2 and Nb2 O5 have also been investigated (Hoyer and Weller, 1995; Fer-
rere et al., 1997; Rensmo et al., 1997; Sayama et al., 1998; Green et al., 2005).
Attached to the surface of the nanocrystalline film is a monolayer of a sensitiser
dye. Photoexcitation of the latter results in the injection of an electron into the
conduction band of the oxide, generating the dye cation. The original state of the
dye is subsequently restored by electron donation from the electrolyte; this step
is often referred to as the regeneration reaction. The electrolyte usually comprises
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 416

416 M. Grätzel & J. R. Durrant

nanocrystalline sensitiser I–/I3– based platinised TCO-

TiO2 film dye electrolyte coated glass
TCO-coated
glass

light -
I
I3

external circuit

electrons

Figure 9.1 Schematic of a liquid electrolyte dye-sensitised solar cell. Photoexcitation of the
sensitiser dye is followed by electron injection into the conduction band of the mesoporous
oxide semiconductor, and electron transport through the metal oxide film to the TCO-coated,
glass, working electrode. The dye molecule is regenerated by the redox system, which is
itself regenerated at the platinised counter electrode by electrons passed through the external
circuit.

an iodide/triiodide redox couple dissolved in a liquid organic solvent, although

attention is increasingly focusing on alternatives for the solvent, including ionic
liquids, gelled electrolytes and polymer electrolytes (Stathatos et al., 2003; Wang
et al., 2003b; Nogueira et al., 2004; Durr et al., 2005) as well as alternatives to the
iodide/iodine redox couple which are now yielding very exciting device efficien-
cies (Yella et al., 2011). The regeneration of the sensitiser by iodide intercepts the
recapture of the injected electron by the oxidised dye. The iodide is in turn regener-
ated by the reduction of triiodide at the counter electrode, with the electrical circuit
being completed via electron migration through the external load.
The high surface area of the mesoporous metal oxide film is critical to efficient
device performance as it allows strong absorption of solar irradiation to be achieved
by only a monolayer of adsorbed sensitiser dye. Whereas a dye monolayer absorbed
on a flat interface exhibits only negligible light absorption (the optical absorption
cross-sectional areas for molecular dyes being typically two to three orders of
magnitude smaller than their physical cross-sections), the use of a mesoporous film
dramatically enhances the interfacial surface area over the geometric surface area,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 417

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 417

by up to a 1000-fold for a 10 µm thick film, leading to high visible-light absorbance

from the many successive monolayers of adsorbed dye in the optical path. Another
advantage of the use of a dye monolayer is that there is no requirement for exciton
diffusion to the dye/metal oxide interface, and also the non-radiative quenching of
excited states often associated with thicker molecular films is avoided. The high
surface area of such mesoporous films does, however, have a significant downside,
as it also enhances interfacial charge-recombination losses, a topic we return to in
more detail below.
A recent alternative embodiment of the DSSC concept is the replacement of
the redox electrolyte with a solid-state hole conductor, which may be either inor-
ganic (Tennakone et al., 1995; O’Regan and Schwartz, 1998) or organic (Bach
et al., 1998), thereby avoiding the use of a redox electrolyte. Such solid-state sen-
sitised heterojunctions can be regarded as functionally intermediate between redox
electrolyte-based photoelectrochemical DSSCs and the organic bulk heterojunc-
tions described in Chapter 6. For more than a decade, power conversion efficiencies
(PCEs) measured under AM 1.5 standard reporting conditions for such solid-state
DSSCs were below those of liquid electrolyte-based devices, the highest reported
PCE value for the former being 7.2% (Burschka et al., 2011) as compared with
12.3% for the latter (Yella et al., 2011). However, the situation has changed dramat-
ically with the advent of perovskite light harvesters, which now reach efficiencies
of 15% in a solid hole-conductor-based device.

9.3.2 Device fabrication

Dye-sensitised solar cells are typically fabricated on transparent conducting oxide
(TCO) glass substrates, enabling light irradiance through this substrate under pho-
tovoltaic operation. The conductive coating typically used is fluorine-doped SnO2
(FTO), preferred over its indium-doped analogue (ITO) for reasons of lower cost
and enhanced stability. Prior to deposition of the mesoporous TiO2 film, a dense
TiO2 film may be deposited to act as a hole-blocking layer, preventing recombi-
nation (shunt resistance) losses between electrons in the FTO and oxidised redox
couple.
The TiO2 nanoparticles are typically fabricated by the aqueous hydrolysis
of titanium alkoxide precursors, followed by autoclaving at temperatures up to
240◦ C to achieved the desired nanoparticle dimensions and crystallinity (anatase)
(Barbe et al., 1997). The nanoparticles are deposited as a colloidal suspension
by screen-printing or by spreading with a doctor blade, followed by sintering at
∼450◦C to ensure good interparticle connectivity. The film porosity is controlled
by the addition of an organic filler such as carbowax to the suspension prior to
deposition; this filler is subsequently burnt off during the sintering step. Figure 9.2
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 418

418 M. Grätzel & J. R. Durrant

Figure 9.2 Scanning electron micrograph of a typical mesoscopic TiO2 film employed in
DSSCs. Note the bipyramidal shape of the particles, with (101) oriented facets exposed.
The average particle size is 20 nm.

shows a scanning electron micrograph of a typical mesoporous TiO2 film. Typical

film thicknesses are 5–20 µm, with TiO2 mass 1–4 mg cm−2, film porosity 50–65%,
average pore size 15 nm and particle diameter 15–20 nm.
The classic sensitiser dye employed in DSSCs is a ruthenium(II) bipyridyl
dye, cis-bis(isothiocyanato)-bis(2,2-bipyridyl-4,4-dicarboxylato)-ruthenium(II),
which is usually referred to as ‘N3’, or in its partially deprotonated form (a di-tetra
butylammonium salt) as ‘N719’ (Nazeeruddin et al., 1993). Figure 9.3 shows the
structure of this dye. The incorporation of carboxylate groups allows ligation to the
film surface via the formation of bidendate and ester linkages, whilst the (–NCS)
groups enhance the absorption of visible light. Adsorption of the dye to the meso-
porous film is achieved by simple immersion of the TiO2 film in a solution of dye,
which results in the conformal adsorption of a self-assembled dye monolayer to
the film surface. The counter electrode is fabricated from FTO-coated glass, with
the addition of a Pt catalyst to catalyse the reduction of the redox electrolyte at this
electrode. Electrical contact between working and counter electrodes is achieved by
the redox electrolyte, with capillary forces being sufficient to ensure the electrolyte
efficiently penetrates the film pores.

9.3.3 Energetics of operation

Figure 9.4 illustrates the energetics of DSSC operation. In contrast to silicon and
organic devices, the high concentration (∼0.5 M) of mobile ions in the electrolyte
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 419

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 419

COOH

COO-
N
S
C
N N

Ru

N N
C
S N
COO-

COOH

Figure 9.3 Chemical structure of the N719 ruthenium complex used as a charge-transfer
sensitiser in dye-sensitised solar cells. The N3 dye has the same structure, except that all
four carboxylates are protonated.

effectively screens out any macroscopic electric fields. Charge separation is there-
fore primarily driven by the inherent energetics (oxidation/reduction potentials) of
the different species at the TiO2 /dye/electrolyte interface rather than by the pres-
ence of macroscopic electrostatic potential energy gradients. Similarly, charge-
transport processes are primarily diffusive, driven by concentration gradients in
the device generated by the photoinduced charge separation. Photoinduced charge
separation occurs at the dye-sensitised TiO2 /electrolyte interface. Electron injec-
tion requires the dye excited state to be more reducing than the TiO2 conduc-
tion band edge, enabling transfer of an electron from photoexcited dye into the
metal oxide. This energetic requirement is equivalent to the dye excited-state low-
est unoccupied molecular orbital (LUMO) having a lower electron affinity than
the electrode conduction-band edge. Similarly, regeneration of the dye ground
state by the redox couple requires the dye cation to be more oxidising than the
iodide/triiodide redox couple. Charge separation in DSSCs can be regarded as a
two-step redox cascade, resulting in the injection of electrons into the TiO2 elec-
trode and the concomitant oxidation of the redox electrolyte. Figure 9.4 shows
typical values for the interfacial energetics in DSSCs. Both charge-separation
reactions are thermodynamically downhill and can be achieved with near unity
quantum efficiency in efficient devices.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 420

420 M. Grätzel & J. R. Durrant

conducting sensitiser
glass TiO2 dye electrolyte cathode
U/V
vs. NHE CB injection +
S*/ S
–0.5
EF
maximum
0 voltage
hν
Red /Ox
0.5 diffusion
regeneration
1.0 +
S/S

e
–
e–

Figure 9.4 Energetics of operation of DSSCs. The primary free-energy losses are associ-
ated with electron injection from the excited sensitiser into the TiO2 conduction band and
regeneration of the dye by the redox couple. The voltage output of the device is approx-
imately given by the splitting between the TiO2 Fermi level (dashed line) and the redox
potential of the redox electrolyte.

Following charge separation, charge collection from the device requires trans-
port of the photogenerated charges to their respective electrodes. For an efficient
DSSC under AM 1.5 solar irradiance, these charge fluxes are of the order of
20 mA cm−2 . The high ionic concentrations in the device effectively screen out
any macroscopic electric fields, thereby removing any significant drift compo-
nent of these transport processes. The transport of both electrons and redox ions
is therefore primarily driven by diffusion processes resulting from concentration
gradients. Under optimum conditions (e.g. good TiO2 nanoparticle interconnec-
tions and a low-viscosity electrolyte), these charge-transport processes (electrons
towards the FTO working electrode and the oxidised component of redox couple
towards the counter electrode) can be efficiently driven with only modest con-
centration gradients, and therefore only small free-energy losses (<50 meV). At
the counter electrode, the oxidised redox couple is re-reduced, a platinum catalyst
typically being used to enable this reaction to proceed with minimal overpotential
(again <50 meV). A similarly ohmic contact is achieved at the TiO2 /FTO inter-
face. It is apparent that the energetics at the TiO2 /dye/electrolyte interface are of
primary importance in determining the overall device energetics.
Power output from the DSSC requires not only efficient charge collection by
the electrodes but the generation of a photovoltage, corresponding to a free-energy
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 421

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 421

difference between the working and counter electrodes. In the dark at equilibrium,
the Fermi energy of the TiO2 electrode (corresponding to the free energy of elec-
trons in this film after thermalisation) equilibrates with the mid-point potential
of the redox couple, resulting in zero output voltage. Under these conditions, the
TiO2 Fermi level lies deep within the bandgap of the semiconductor, and the film
is effectively insulating, with a negligible electron density in the TiO2 conduction
band. Photoexcitation results in electron injection into the TiO2 conduction band
and concomitant hole injection into (that is, oxidation of) the redox electrolyte. The
high concentrations of oxidised and reduced redox couple present in the electrolyte
in the dark mean that this photooxidation process does not result in a significant
change in redox potential of the electrolyte, which remains effectively fixed at
its dark, resting value. In contrast, electron injection into the TiO2 conduction
band results a dramatic increase in electron density (from approx. ≤1013 cm−3 to
≥1018 cm−3 ), raising the TiO2 Fermi level towards the conduction-band edge, as
illustrated in Fig. 9.4, and allowing the film to become conducting. This shift of the
TiO2 Fermi level under irradiation increases the free energy of injected electrons
and is responsible for the generation of the photovoltage in the external circuit.
The mid-point potential of the redox couple is given by the Nernst equation,
and is therefore dependent on the relative concentrations of iodide and iodine. The
concentrations of these species required for efficient device function are in turn
constrained by kinetic requirements of dye regeneration at the working electrode,
and iodide regeneration at the counter electrode, as discussed below. Typical con-
centrations of these species for the iodine/iodide couple are in the range 0.1–0.7 M
iodide and 10–200 mM iodine, constraining the mid-point potential of this elec-
trolyte to ∼0.4 V vs. NHE. It should furthermore be noted that in the presence
of excess iodide, the iodine is primarily present in the form I− 3 , resulting in this
electrolyte often being referred to as the iodide/triiodide redox couple.
Determining the energetics of the TiO2 conduction band is more complex. As
with most oxides, the surface of TiO2 may be more or less protonated depending
on the pH of the surrounding medium. The resultant changes in surface charge
cause the surface potential to exhibit a Nernstian dependence on effective pH,
shifting by 60 mV per pH unit (Rothenberger et al., 1992). In bulk metal oxides,
surface charge can result in significant bending of the conduction and valence
bands adjacent to the surface. However, in the mesoporous TiO2 films employed in
DSSCs, the nanoparticles are too small to support significant band bending. As a
consequence, the whole conduction band of such mesoporous films shifts with the
surface potential. At pH 1 in aqueous solutions, the conduction band edge for TiO2
has been reported at −0.06 V vs. NHE, shifting negatively as the pH is increased
(Hengerer et al., 2000). Dye-sensitised solar cells typically employ organic rather
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 422

422 M. Grätzel & J. R. Durrant

than aqueous electrolytes, complicating quantification of the effective pH. Never-

theless, studies in organic solvents have demonstrated shifts of the conduction band
of mesoporous TiO2 films of up to 1 V depending on the concentration of potential-
determining ions (primarily small cations such as protons or lithium cations) in the
electrolyte (Redmond and Fitzmaurice, 1993). For this reason, the concentration
of such potential-determining ions in the electrolyte plays a key role in determin-
ing the energetics of the dye-sensitised interface, and thereby device performance.
Additives added to the electrolyte to determine such energetics include Li+ , gua-
nadinium ions, N-methylbenzimidazol and t-butyl pyridine (which functions as a
base). Further influence on the interfacial energetics can be achieved by variation
of the extent of protonation of the sensitiser dye. A further consideration is the
presence of a tail of sub-bandgap states in the TiO2 which function as localised
electron trap states, as discussed elsewhere (O’Regan and Durrant, 2009).
The choice of suitable sensitiser dye energetics is essential to achieve suitable
matching to the metal oxide and redox couple. The excited-state oxidation potential
(Um (S+ /S*)) must be sufficiently negative to achieve efficient electron injection
into the TiO2 conduction band, whilst the ground-state oxidation potential must be
sufficiently positive to oxidise the redox couple. The redox properties of adsorbed
sensitiser dyes may differ significantly from those measured in solution, mainly
due to the high surface-charge densities and dipoles present at this interface (Zaban
et al., 1998). Notwithstanding this, typical dye redox potentials empirically found
to be compatible with efficient device function are Um (S+ /S) >0.6 V vs. NHE.

9.3.4 Kinetics of operation

Figure 9.5 is a photochemical view of the function of a DSSC, illustrating the
sequence of electron transfer and charge-transport processes which result in pho-
tovoltaic device function. In addition to the forward electron transfer and transport
processes, this figure also illustrates several competing loss pathways, shown as
black arrows. These loss pathways include decay of the dye excited state to ground,
and charge recombination of injected electrons with dye cations and with the redox
couple. Each charge-transfer step results in an increased spatial separation of elec-
trons and holes, increasing the lifetime of the charge-separated state, but at the
expense of reducing the free energy stored in this state. This functionality exhibits
a close parallel to function of photosynthetic reaction centres. As in natural photo-
synthesis, kinetic competitions between the various forward and loss pathways are
critical to determining the quantum efficiencies of charge separation and collection,
and are therefore key factors determining energy conversion efficiency.
The theory of interfacial electron transfer, and the dependence of electron-
transfer rate on interfacial energetics, are discussed in Volume 3 of this book series,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 423

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 423

G/eV

Dye*
electron injection ps
+ –
Dye / e TiO2

regeneration reaction
1 ~1 µs
decay to ground – –
hν ~ns I3 / e TiO2
recombination to dye*,
µs – ms charge transport to
electrodes, ms – s
recombination to
electrolyte, ms – s
Dye
0

Figure 9.5 State diagram representation of the kinetics of DSSC function. Forward pro-
cesses of light absorption, electron injection, dye regeneration and charge transport are
indicated by red arrows. The competing loss pathways of excited-state decay to ground and
electron recombination with dye cations and oxidised redox couple are shown in black.
The vertical scale corresponds to the free energy G stored in the charge-separated states.
Note the free energy of injected electrons is determined by the Fermi level of the TiO2 ;
the figure is drawn assuming a TiO2 Fermi level 0.8 V above the chemical potential of the
redox electrolyte, corresponding to the Fermi level of a typical iodide/iodine-based DSSC
at open circuit. It should be noted that at maximum power point, the Fermi level is lower
due to charge transport to the electrodes (also shown in diagram).

and the dynamics of charge transport in mesoscopic electrodes are addressed in

Chapter 11 of that volume. In this chapter we will therefore confine ourselves only
to consideration of the impact of these dynamics on the performance of DSSCs.
More detailed reviews of these dynamics have been reported elsewhere (Hagfeldt
et al., 2010; Listorti et al., 2011).
Efficient electron injection requires the rate of electron injection to exceed
excited-state decay to ground. Typical rates of dye excited-state decay to ground are
in the range 107 –1010 s−1 . The rate of electron injection depends on the electronic
coupling between the dye excited-state LUMO orbital and accepting states in the
TiO2 , and on the relative energetics of these states. In model system studies of
dye-sensitised metal oxide films, electron injection rates of >1012 s−1 have been
reported for a range of sensitiser dyes, consistent with efficient electron injection
(Anderson and Lian, 2005; Durrant et al., 2006). However it should be noted
that fast electron-injection dynamics require both strong electronic coupling of
the dye LUMO orbital to the metal oxide conduction-band states, and a sufficient
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 424

424 M. Grätzel & J. R. Durrant

free-energy difference to drive the reaction (Schmickler, 1996). As such, electron-

injection dynamics are dependent on the energetics of the TiO2 conduction band,
and therefore on the concentration of potential-determining ions (e.g. Li+ ) in the
electrolyte. Omission of such ions from the electrolyte can result in an insufficient
energetic driving force, reducing the quantum yield of charge injection, and thereby
reducing device photocurrent.
It should be noted that the heavy metal ion in the ruthenium bipyridyl dyes
widely employed in DSSCs results in an ultrafast (up to 1013 s−1 ) intersystem
crossing from the initially formed singlet excited state to a triplet state. This relax-
ation process reduces the excited-state free energy by ∼400 meV. While several
studies have reported sub-picosecond electron injection from the singlet excited
state of these dyes, electron injection from the triplet state is significantly slower
(1010 –1011 s−1 ), consistent with the lower free energy of this excited state, but still
fast compared with decay of the triplet state to ground (107 –108 s−1 ) (Anderson
and Lian, 2005; Durrant et al., 2006). Studies of complete devices have suggested
that this relatively slow electron injection from the dye triplet state may be the
dominating injection pathway in efficient DSSCs (Haque et al., 2005; Koops et al.,
2009).
Efficient dye regeneration requires the rate of re-reduction of the dye cation
by the redox couple to exceed that of the charge recombination of injected elec-
trons with these dye cations. This recombination reaction has been shown to be
strongly dependent on the electron density in the TiO2 electrode (and therefore light
intensity and cell voltage), accelerating by at least an order of magnitude between
short-circuit and open-circuit conditions (Durrant et al., 2006). It is furthermore
dependent on the spatial separation of the dye cation (highest occupied molecu-
lar orbital [HOMO]) orbital from the metal oxide surface, with the rate constant
decaying exponentially with distance, consistent with electron tunneling theory.
The regeneration reaction is dependent on the iodide concentration, electrolyte
viscosity and dye structure. For the N719 sensitiser dye, and employing a low-
viscosity electrolyte such as acetonitrile, the regeneration reaction has a half-time
of ∼1 µs, sufficiently fast to compete effectively with the competing recombina-
tion reaction and ensuring that the regeneration reaction can be achieved with unit
quantum efficiency (Green et al., 2005).
Efficient charge collection by the external circuit requires the time constant
for electron transport within the TiO2 matrix to be faster than charge recombina-
tion of injected electrons with the redox couple. Electron transport is a diffusive
process, strongly influenced by electron trapping in localised sub-bandgap states,
resulting in the dynamics being strongly dependent on the position of the TiO2
electron Fermi level: raising the Fermi level towards the conduction-band edge
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 425

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 425

results in increased trap filling. Typical electron-transport times under solar irra-
diation are of the order of milliseconds (Peter and Wijayantha, 2000; Frank et al.,
2004). If a low-viscosity solvent such as acetonitrile is used, transport of the oxi-
dised redox couple to the counter electrode is not rate-limiting. However, the use
of higher viscosity solvents more suitable for practical device applications (for rea-
sons of device stability) can result in significantly lower ionic diffusion constants,
with the resultant iodide/triiodide concentration gradients causing significant free-
energy (series resistance) losses, and also potentially accelerating interfacial charge
recombination.
Given the relatively slow timescale for charge transport in DSSCs compared
with most other photovoltaic devices, and the extensive interfacial area available
for charge recombination in the device (due to its mesoscopic structure), it is
remarkable that the quantum efficiency of charge collection can approach unity. The
key factor enabling this high efficiency is the slow rate constant for the interfacial
charge recombination of injected electrons with the oxidised redox couple. This
reaction is a multi-electron reaction, most simply being described by the equation

I3− + 2e− → 3I− (9.1)

which must therefore proceed via one or more intermediate states. The mechanism
of this reaction has been extensively studied, and while the details remain somewhat
controversial, it is apparent that without a suitable catalyst such as platinum, one or
more of the intermediate steps exhibits a significant activation barrier, resulting in
a slow overall rate constant for this reaction. The low rate constant for this recom-
bination reaction on TiO2 , contrasting to the facile electrochemistry of this redox
couple on the platinised counter electrode, is a key factor behind the remarkable
efficiencies achieved to date for DSSCs. Similarly, interface engineering through
sensitiser design to minimise this recombination rate constant has been key to
recent successes with alternative one electron couples (Feldt et al., 2010). Nev-
ertheless, Eq. (9.1) is the primary recombination pathway in DSSCs. The flux of
this recombination pathway increases with increasing electron density in the TiO2
electrode (and therefore with the TiO2 Fermi level or cell voltage). In the dark it is
responsible for the diode-like leakage current observed in current–voltage scans,
while under illumination it is the primary factor limiting the voltage output of the
device.
The kinetic competition between charge transport and recombination in
DSSCs has been analysed in terms of an effective carrier diffusion length L n ,
given by


L n = Deff τ (9.2)
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 426

426 M. Grätzel & J. R. Durrant

where Deff is the effective electron diffusion coefficient, and τ the electron lifetime
due to the charge-recombination reaction given by Eq. (9.1) (Peter and Wijayantha,
2000). Deff increases with light intensity (due to the increased electron density in
the TiO2 film) whilst τ shows a proportional decrease, resulting in L n being largely
independent of light intensity. Typical values for L n are 5–20 µm, and even longer
near the optimum power point for cells with >10% conversion efficiency, consistent
with the high carrier-collection efficiencies observed in efficient DSSCs.
It is important to emphasise that the energetics and kinetics of DSSC function
are not independent considerations. The kinetics of the interfacial electron-transfer
dynamics depend strongly on the energetics of the TiO2 /dye/electrolyte interface
and on the density of electrons in the TiO2 (i.e. the TiO2 Fermi level). Raising the
energy of the TiO2 conduction band reduces recombination losses (as for a given
TiO2 Fermi level, the electron density in the TiO2 film will be lower), and therefore
may give a high cell output voltage, but at the expense of a lower free energy driving
force for charge separation, which may result in a lower quantum efficiency for
charge generation and therefore a lower output current. In practice, modulation of
these energetics and kinetics to achieve optimum device performance remains one
of the key challenges in DSSC research and development.

9.4 DSSC research and development

9.4.1 Panchromatic sensitisers
The ideal sensitiser for a single-junction photovoltaic cell converting standard
global AM 1.5 sunlight to electricity should absorb all light below a threshold
wavelength of about 920 nm. In addition, it must also carry attachment groups
such as carboxylate or phosphonate to firmly graft it to the semiconductor oxide
surface. On excitation, it should inject electrons into the solid with a quantum
yield of unity. The energy level of the excited state should be well matched to
the lower bound of the conduction band of the oxide to minimise energetic losses
during the electron-transfer reaction. Its redox potential should be sufficiently high
that it can be regenerated via electron donation from the redox electrolyte or the
hole conductor. Finally, it should be stable enough to sustain about 100 million
turnover cycles, corresponding to about twenty years of exposure to natural light.
A single-junction device with such a sensitiser could reach a maximum conversion
efficiency of 32% in global AM 1.5 sunlight.
Much of the research in DSSC dye chemistry is devoted to the identification
and synthesis of sensitisers matching these requirements, while retaining stability
in the photoelectrochemical and photovoltaic environment. A number of recent
reviews reflects the impressive scientific thrust of this research (Imahori et al.,
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 427

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 427

2009; Hagfeldt et al., 2010; Mishra et al., 2010; Ning et al., 2010; Walter et al.,
2010; Qin and Peng, 2012; McEvoy and Grätzel, 2013). The attachment group of
the dye ensures that it spontaneously assembles as a molecular layer on exposing
the oxide film to the dye solution. This molecular dispersion ensures a high prob-
ability that, once a photon is absorbed, the excited state of the dye molecule will
relax by electron injection into the semiconductor conduction band. However, the
optical absorption of a single monolayer of dye is weak, a fact which originally was
cited as ruling out the possibility of high-efficiency sensitised devices, as it was
assumed that smooth substrate surfaces would be imperative in order to avoid the
recombination loss mechanism associated with rough or polycrystalline structures
in solid-state photovoltaics. This objection was invalidated by recognising that the
injection process places electrons in the semiconductor lattice, spatially separated
from the positive charge carriers in the electrolyte or a hole carrier. Hence, mas-
tering the interface is key to slow down the rate of charge recombination. The dye
molecules themselves, if judiciously designed, can block the access of the redox
shuttle to the interface providing a barrier for charge recombination. By now, the
use of nanocrystalline thin-film structures with a roughness factor of >1000 has
become standard practice.
Polypyridyl complexes of ruthenium and osmium have over many years main-
tained a lead in photovoltaic performance in terms of both conversion yield and
long-term stability. Initially, interest focused on sensitisers having the general struc-
ture ML2 (X)2 where L stands for 2,2 -bipyridyl-4,4-dicarboxylic acid, M is Ru or
Os, and X represents a halide, cyanide, thiocyanate, acetyl acetonate, thiocarba-
mate or water substituent. In particular, the ruthenium complex cis-RuL2 (NCS)2 ,
known as N3 dye, became the paradigm heterogeneous charge-transfer sensitiser
for mesoporous solar cells (Nazeeruddin et al., 1993). The absorption spectrum of
fully protonated N3 has maxima at 518 and 380 nm, with extinction coefficients of
1.3 × 104 M−1 cm−1 and 1.33 × 104 M−1 cm−1 , respectively. The complex emits
at 750 nm, the excited-state lifetime being 60 ns. The optical transition has MLCT
(metal-to-ligand charge-transfer) character: excitation of the dye involves transfer
of an electron from the metal to the π ∗ orbital of the surface-anchoring carboxy-
lated bipyridyl ligand, from where it is released in a timescale of femtoseconds
to picoseconds into the conduction band of TiO2 , generating electric charges with
near unit quantum yield.
Discovered in 1993, the photovoltaic performance of N3 was unmatched for
eight years by virtually hundreds of other complexes that have been synthesised and
tested. However, in 2001 the ‘black dye’ tri(cyanato)-2,22 -terpyridyl-4,44 -9-
tricarboxylate)Ru(II) achieved 10.4% AM1.5 solar-to-power conversion efficiency
in full sunlight (Nazeeruddin et al., 2001). Conversion efficiencies have meanwhile
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 428

428 M. Grätzel & J. R. Durrant

isc 17 mA cm–2 isc 22 mA cm–2 isc 26.7 mA cm–2

Figure 9.6 Spectral response of the photocurrent for the N3, black dye and DX1 sensitisers.
The incident photon-to-current conversion efficiency (IPCE) is plotted as a function of
excitation wavelength. The structure of the sensitisers and their short-circuit photocurrent
is shown below the graph.

been improved further, the current record validated by an accredited laboratory

being 11.9% (Han et al., 2012). In the latter case an organic co-sensitiser was
employed to enhance the light harvesting in the blue spectral region.
Figure 9.6 compares the spectral response of the photocurrent observed with
the N3 dye, the black dye and the DX1 sensitisers. The incident photon-to-current
conversion efficiency (IPCE), also referred to as the external quantum efficiency
(EQE) of the DSSC, is plotted as a function of excitation wavelength. Both
chromophores show very high IPCE values in the visible range. However, the
response of the black dye extends 100 nm further into the infrared (IR) than that
of N3. The photocurrent onset is close to 920 nm, near the optimal threshold for
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 429

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 429

single-junction converters. From there the IPCE rises gradually until at 700 nm it
reaches a plateau of ∼80%. If one accounts for reflection and absorption losses in
the conducting glass, the conversion of incident photons to electric current is prac-
tically quantitative over the whole visible domain. From the overlap integral of the
curves in Fig. 9.6 with the AM 1.5 solar spectrum, the short-circuit photocurrents
of the N3, black dye and DX1 sensitisers cells are predicted to be 17 mA cm−2,
22 mA cm−2 and 27.6 mA cm−2 under standard AM 1.5 G reporting conditions,
respectively, in agreement with experimental observations.
Thus, over the last decade a whole new family of judiciously engineered ruthe-
nium complexes has emerged that show excellent photovoltaic performance (Wang
et al., 2005a; Reddy et al., 2007; Grätzel, 2009; Kalyanasundaram and Grätzel,
2010). The spectral response of the photocurrent for of the DX1 dye, shown by the
blue curve in Fig. 9.6, illustrates the amazing advance in this area (Kinoshita et al.,
2013). By replacing the three thiocyanate groups in the black dye with two chloride
ions and one phosphine ligand, a new panchromatic sensitiser, coded DX1, was
fashioned that harvests sunlight over the whole visible and near-IR range up to
1000 nm, closely matching the spectral response of silicon photovoltaic cells and
generating a short circuit photocurrent of 27.6 mA cm−2 .
Recently, new donor acceptor porphyrins have been developed, in particular
the YD2-o-C8 zinc complex, whose PCE exceeds that of the black dye when used
in conjunction with Co((II/III)(bipy)3 complexes as a redox couple (Yella et al.,
2011). The Nernst potential of this redox mediator is more positive than that of
the iodide/triiodide couple, yielding substantial gains in open-circuit potential of
the cells. The performance of YD2-o-C8 is enhanced by judicious substitution
of two meso-positions at the tetrapyrrole ring with a donor and acceptor moiety
and the two remaining meso-positions bearing phenyl substituents with two bulky
alkoxy groups in ortho-positions. This increases the light-harvesting capacity of the
porphyrin and blocks the unwanted interfacial electron back transfer. Further gain
in efficiency was obtained via the use of a co-sensitiser, coded Y123, absorbing
strongly in the green spectral region. The structure of the two sensitisers is shown
in Fig. 9.7, which presents a plot of the photocurrent density versus voltage for a
mixture of the two dyes, adsorbed at the surface of the nanocrystalline titania film
and the IPCE spectra for this. The i sc , Voc and fill factor values under standard
air mass (AM 1.5) reporting conditions were 17.8 mA cm−2 , 0.935 V and 0.74,
respectively, yielding a power conversion efficiency of 12.3%.

9.4.2 Present status of DSSC performance

The solar-to-electric power conversion efficiency (PCE) of PV cells is calculated
from the short-circuit photocurrent density (i sc ), the open-circuit photovoltage
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 430

430 M. Grätzel & J. R. Durrant

C6H13O

C6H13 C8H17O OC8H17

C6H13O

NC
N N OH
O

S S
N Zn N OH

N N O
C6H13 C6H13
C6H13O

C6H13 C8H17O OC8H17

C6H13O

Figure 9.7 Top left: current–voltage characteristics of a YD2-o-C8/Y123 co-sensitised

DSSC, measured under standard AM 1.5 G sunlight under various light intensities. The
molar YD2-o-C8/Y123 ratio in the dye solution was 7. Top right: spectral response of the
incident photon-to-electric current conversion efficiency (IPCE) for YD2-o-C8 (red dots),
Y123 (blue triangles) and YD2-o-C8/Y123 co-sensitised nanocrystalline TiO2 films (black
squares). Bottom left: chemical structure of the sensitiser YD2-o-C8. Bottom right: chemical
structure of the sensitiser Y123.

(Voc ), the fill factor (ηfill ) and the standard incident solar irradiance (E so =
1000 W m−2 ) as

PCE = i sc × Voc × ηfill /E so (9.3)

The PCE value of 12.3% achieved with the YD2-o-C8/Y123 co-sensitised tita-
nia films present a new record for laboratory cells based on redox electrolytes. How-
ever, solid-state DSSC embodiments based on perovskite light harvesters achieve
currently a PCE of 15% (Burschka et al., 2013). Due to significant industrial upscal-
ing efforts, the conversion efficiency of DSSC modules has also been rapidly rising
over the last few years. Using the black ruthenium dye along with an organic co-
sensitiser, Sony Corporation has reached a certified PCE efficiency under AM
1.5 standard conditions of 9.9% (Green et al., 2011). This implies that scaling up
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 431

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 431

dye-sensitised laboratory cells by a factor of at least 100 does not entail a significant
decrease in photovoltaic performance, the loss being mainly due to interconnects
or collector grids reducing the active device area.
Further substantial gains in the PCE of the DSSC can be expected in the near
term by implementing the substantial increase in the open-circuit photovoltage
that can be achieved with cobalt complexes rather than iodide/triiodide as redox
electrolyte. For example, a very high Voc of 1.1 V was recently reached with the
cobalt bisbipyridyl pyrazol complex (Yum et al., 2012) whose redox potential
matches well that of the sensitiser ground state, reducing the free-energy loss in
the regeneration step from 0.6 eV for I− /I−3 to 0.2 eV, as shown by the energy level
diagram in Fig. 9.8.

9.4.3 Tandem cells

A feature that makes the DSSC particularly attractive for tandem cell application is
that its optical transmission and short-circuit photocurrent can be readily adjusted
by changing the film thickness, pore size, nature of the dye, and dye loading. This,
along with the ease of forming layered structures (for example, by producing the
mesoscopic oxide films using screen-printing or doctor blading methods), renders
the DSSC particularly well suited for the fabrication of tandem solar cell structures
that capture the solar spectrum in an optimal fashion. Note that for a two-level
tandem cell a conversion efficiency of up to 45% can be reached, the optimal
bandgaps for the top and bottom cells being around 1.65 eV and 1 eV, respectively.

Figure 9.8 Energy-level diagram of a DSSC. Shown are the ground and excited-state redox
potentials for a typical sensitiser as well as for the iodide/triiodide redox couple and a
series of cobalt complexes; bpy-pz and py-pz represent bipyridylpyrazol and pyridylpyrazol,
respectively.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 432

432 M. Grätzel & J. R. Durrant

Several publications have dealt with the use of stacked DSSC configurations in
which two dyes absorbing different parts of the solar spectrum serve as sensitisers
(Durr et al., 2004; Kubo et al., 2004). We have demonstrated that a tandem device
comprising a DSSC as a top cell for high-energy photons and a copper indium
gallium selenide (CIGS) thin-film bottom cell, capturing the red and near-IR solar
emission respectively, produces AM 1.5 solar-to-electric conversion efficiencies
greater than 15% (Liska et al., 2006).
Figure 9.9 shows the I –V curve and other characteristics obtained under
AM 1.5 insolation for one of the first embodiments of this two-terminal tandem
device. The performance of the tandem was clearly superior to that of the individ-
ual cells, despite the fact that the short-circuit currents of the two cells were not
perfectly matched, as evidenced by the shoulder in the I –V curve. Likewise, no
effort was made to minimise the optical losses. This leaves no doubt that further
efficiency gains reaching well beyond the 20% mark can be expected from the
fructuous marriage of these two thin-film PV technologies (Liska et al., 2006).
Combining a relatively low-cost thin-film CIGS substrate cell with a DSSC super-
strate cell may be a cheaper method of achieving efficiencies above 15% than use
of a high-efficiency CIGS cell alone.
However, in order to minimise optical and ohmic losses it is mandatory to
integrate the two cells in a monolithic device in which they are series-connected

–14

–12
current density / mA cm–2

–10

–8

–6

–4

–2

2
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
potential / V

Figure 9.9 Photocurrent density–voltage characteristics under AM 1.5 full sunlight

(1000 mW cm−2 ) for a two-terminal tandem DSSC/CIGS cell. The PCE is 15.8%. Adapted
from Liska et al. (2006).
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 433

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 433

through a transparent recombination layer. Attempts in this direction have been

made (Wenger et al., 2009) but the ZnO front contact of the CIGS cell was unstable
in the presence of the iodide/triidode-based electrolyte. The advent of perovskite-
sensitised solid-state cells has rekindled interest in such two-cell tandem structures,
and several groups are actively working on devices of this type. Indeed, perovskite
photovoltaics are very attractive candidates for the top cell of a two-cell tandem
with CIGS or silicon as a bottom cell because their bandgap can be adjusted to a
value between 2.1 and 1.5 eV, for example by using mixtures of iodide and bromide
as anions (Noh et al., 2013). Moreover, the open-circuit voltages of perovskite cells
are amazingly high, reaching 1.1 V for the 1.5 eV bandgap, close to the theoretical
maximum of 1.2 V. Realising tandem cells with PCEs >20% has therefore come
within close reach.

9.4.4 Stability
Unlike amorphous silicon, which suffers from degradation due to the well-known
Staebler–Wronski effect, the intrinsic stability of the DSSC has been confirmed
by extensive accelerated light-soaking tests carried out over the last fifteen years.
One major issue that has been settled during this period is that the sensitisers
employed in the current DSSC embodiments can sustain twenty years of outdoor
service without significant degradation. However, as new and more advanced dye
structures emerge, and in order to avoid repeating these lengthy tests every time
the sensitiser is modified, kinetic criteria have been developed to allow prediction
of long-term sensitiser performance.
Figure 9.10 illustrates the catalytic cycle that the sensitiser undergoes during
cell operation. Critical for stability are side reactions that occur from the excited
state (S*) or the oxidised state of the dye (S+ ), which would compete with electron
injection from the excited dye into the conduction band of the mesoscopic oxide

s h•

regeneration excitation
k reg

s+ s*
k1
2
kinj

product product
decomposition injection decomposition

Figure 9.10 The catalytic cycle of the sensitiser during DSSC operation.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 434

434 M. Grätzel & J. R. Durrant

and with the regeneration of the sensitiser. These destructive channels are assumed
to follow first or pseudo-first-order kinetics and are assigned the rate constants
k1 and k2 . Introducing the two branching ratios P1 = kinj /(k1 + k inj) and P2 =
k reg /(k2 + kreg ), where kinj and kreg are the first-order or pseudo-first-order rate
constants for the injection and regeneration process, respectively, the fraction of
the sensitiser molecules that survives n cycles is given by the product (P1 × P2 )n .
A simple calculation (Grätzel, 2006) shows that the sum of the branching ratios for
the two bleeding channels should not exceed 1 × 10−8 in order for the sensitiser
to turn over 100 million times, corresponding to a lifetime of at least twenty years.
The turnover frequency of the dye in the working DSSC, averaged diurnally and
seasonally, is about 0.16 s−1 .
For most of the common sensitisers, the rate constant for electron injection
from the excited state of the dye to the conduction band of the TiO2 particles
is in the picosecond or femtosecond range. Assuming k inj = 1010 –1013 s−1 , any
destructive side reaction should have k 1 < 102 s−1 . Ruthenium sensitisers of the
N719 or C106 type readily satisfy this condition as the decomposition from the
excited-state level occurs at a much lower rate than the 102 s−1 limit. Precise kinetic
information has also been gathered for the second destructive channel involving
the oxidised state of the sensitiser, the key parameter being the ratio k2 /kreg of the
rate constants for the degradation of the oxidised form of the sensitiser and its
regeneration. The S+ state of the sensitiser can readily be produced by chemical
or electrochemical oxidation and its lifetime can be independently determined
by absorption spectroscopy. A typical value of k 2 is around 10−4 s−1 while the
regeneration rate constant is at least in the 105 s−1 range. Hence the branching ratio
is well below the limit of 10−8 , which can be tolerated to achieve the 100 million
turnovers and a twenty-year lifetime for the sensitiser. Apart from the photoinduced
degradation of the sensitiser during the catalytic cycle, its chemical reaction with
constituents of the electrolyte or its desorption or its mere reorganisation on the
surface to a non-productive conformation could possibly harm cell performance
in the long term. An appropriate way to rule out the existence of any problems of
this type is to carry out accelerated tests under heat stress.
Many long-term tests that have been performed with the N3-type ruthenium
complexes have confirmed the extraordinary stability of these charge-transfer sen-
sitisers. For example, a European consortium supported by the Joule program
(Hinsch et al., 2001) confirmed cell photocurrent stability during 8,500 hours of
light soaking at 2.5 Suns, corresponding to about 56 million turnovers of the dye
without any significant degradation. More recently, DSSC lifetimes of over 20
years have been projected from continuous light-soaking tests performed over
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 435

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 435

20,000 hours (Harakirisun and Desilvestro, 2011). These results corroborate the
projections from the kinetic considerations made above.
A more difficult task has been to achieve stability under prolonged stress at
higher temperatures, i.e. 80–85◦ C. The introduction of hydrophobic sensitisers
has been particularly rewarding in allowing DSSCs to meet, for the first time,
the specifications laid out for outdoor applications of silicon photovoltaic cells
(Wang et al., 2003a). In addition these dyes show enhanced extinction coeffi-
cients owing to the extension of the π-conjugation onto one of the bipy ligands
by styrene moieties. Taking advantage of these properties and using a novel robust
electrolyte formulation, a ≥8% efficient DSSC that shows strikingly stable perfor-
mance under both prolonged thermal stress and light soaking was realised (Wang
et al., 2005b; Sauvage et al., 2011). Dyesol recently announced completion of a
high-temperature test carried out with a new electrolyte formulation based on sul-
fone, a polar non-toxic high boiling point solvent and using the above mentioned
Y123 dye as a sensitiser. Impressively, stable performance was maintained over
3000 hours, establishing beyond any doubt that the DSSC can survive long-term
heat exposure without any significant degradation.
While impressive progress has been made in the development of stable, non-
volatile electrolyte formulations, the efficiencies of these systems are presently
in the 7–10% range (Sauvage et al., 2011), below the 11.9% reached with Ru-
based sensitisers and volatile solvents. Future research efforts will be dedicated
to reducing the performance gap between these systems. The focus will be on
hole conductors and solvent-free electrolytes, in particular ionic liquids. The latter
are a particularly attractive choice for the first commercial modules, owing to
their high stability, negligible vapour pressure and excellent compatibility with the
environment. Fujikura (Arakawa et al., 2009) has announced excellent stability data
with cells based on solid (gel) ‘nanocomposite’ electrolytes based on a mixture
of ionic liquids and silica nanoparticles. Modules showing 8% PCE passed all
accelerated stability tests required by the IEC 61646 protocol for outdoor PV
panels including exposure to 85% humidity at 85◦ C for 1000 hours, long-term light
soaking and cycling the temperature of the modules between −40◦ C and +90◦ C.
These results confirm the extraordinary robustness of dye-sensitised solar cells
when employed in conjunction with appropriately designed electrolytes, showing
their aptness for outdoor deployment.
In keeping with the encouraging results obtained by Fujikura, the Israeli com-
pany 3GSolar announced that its DSSC modules have operated on the company’s
rooftop continuously for two years and they continue to perform to the same stan-
dard as on the day they were placed outdoors (3GSolar, 2011). Importantly, 3GSolar
and ECN (DeWild-Scholten and Veltkamp, 2007) have also performed life-cycle
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 436

436 M. Grätzel & J. R. Durrant

analysis, which shows the energy payback time for the DSSC to be less than one
year in southern European climates as compared with about three years for silicon
solar cells at that time; a more recent analysis shows the latter has improved to
1.3–1.5 years (Mann et al., 2013).

9.4.5 Organic dyes

Over the last two decades, ruthenium complexes endowed with appropriate ligands
and anchoring groups have by far been the preferred choice of charge-transfer sensi-
tisers for mesoscopic solar cells. In recent years, however, there has been a surge of
interest in organic donor–acceptor dyes (Qin et al., 2008). The groups of Arakawa
and Uchida have made remarkable advances in the development of organic dyes for
use in DSSCs (Horiuchi et al., 2004; Hara et al., 2005). The majority of these sen-
sitisers are of the ‘push-pull’ type comprising a donor and acceptor group bridged
by a π-conducting moiety. Coumarin or polyene-type sensitisers have attained
PCE values of up to 9.2% in full sunlight (Ito et al., 2006), increasing to 9.5% for
the indoline dye D205 (Ito et al., 2008). Figure 9.11 shows examples of typical
structural elements of such D–π–A dyes. Numerous representatives of this class of
compounds have been synthesised (Yum et al., 2009), and PCE values up to 9.8%
have been reported with iodide-based electrolytes (Zhang et al., 2009). This rapid
development has led to the discovery of a new class of D–π–A dyes where the

SnO2 TiO 2 Dye OMeTAD

Gold
Glass
OCH3 OCH3

OCH3 OCH3

h+ HH3COCO
3 NN N OCH3 N OCH3
Light H3CO N N OCH3

e-
H3CO OCH3
N OCH3
N OCH3

OCH3 OCH3

Figure 9.11 Cross-sectional view of a solid-state DSSC containing the hole conductor
spiro-OMeTAD, the structure of which is indicated on the right. Left-hand drawing courtesy
of B. O’Regan.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 437

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 437

π-bridge is constituted by a porphyrin moiety (Hsieh et al., 2010; Bessho et al.,

2010).
The number of suitable options for D–π–A dye structures being very large,
state-of-the-art theoretical chemical calculations are being employed as a guide for
the selection of the most promising candidates for synthesis. The recently devel-
oped scaled opposite-spin configuration interaction singles-doubles technique,
abbreviated as SOS-CIS(D) (Rhee and Head-Gordon, 2007), appears to offer great
accuracy in calculating the UV–vis spectra of novel D–π–A structures (Casanova
et al., 2010) This new method has advantages over time-dependent density func-
tional theory in computing the energy of charge transfer type optical excitations and
will assist the experimentalists in the judicious selection of molecular components
to engineer the best-performing push charge-transfer sensitisers.
When considering organic dyes for use in DSSCs, porphyrins and phthalo-
cyanines have attracted particular attention, the former because of the analogy
with natural photosynthetic processes, the latter because of their photochemical
and phototherapeutic applications. Through judicious molecular engineering, por-
phyrins have become the leading photosensitiser with PCE values reaching 12.3%,
as shown in Fig. 9.7 (Yella et al., 2011). Phthalocyanines show intense absorp-
tion bands in the 600–800 nm spectral region. Previous problems with aggregation
and the unsuitable energetic position of the LUMO level, which was too low for
electron transfer to the TiO2 conduction band to occur efficiently, have now been
largely overcome (Ince et al., 2012), rendering this class of sensitisers also very
attractive for DSSC applications in buildings where the attractive blue-green tinge
of the glass panels is very much in demand by architects.

9.4.6 Quantum dots as sensitisers

Semiconductor quantum dots (QDs) are another attractive option for panchromatic
sensitisers. These are II–VI and III–V type semiconductor particles whose size is
small enough to produce quantum confinement effects. The absorption spectrum of
such quantum dots can be adjusted by changing the particle size. Thus, the bandgap
of materials such as InAs and PbS can be adapted to attain the value of 1.35 eV,
which is ideal for a single-junction solar quantum converter. During the last two
decades a wealth of information has been gathered on the physical properties of
QD materials and research is being pursued very actively. With liquid electrolytes,
photocorrosion of the quantum dots is likely to happen. However, quantum dots
are expected to display higher stability in solid-state heterojunction devices (Plass
et al., 2002). Hence, recent work has focused on solid-state cells where a layer of
quantum dots deposited on a mesoscopic oxide film such as ZnO or TiO2 forms a
depleted heterojunction (Pattantyus-Abraham et al., 2010; Barkhouse et al., 2011;
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 438

438 M. Grätzel & J. R. Durrant

Sargent, 2012). These simple devices function surprisingly well, exhibiting high
short-circuit photocurrents that exceed 20 mA cm−2 under full sunlight, the highest
certified PCE being 7.5%.
Multiple exciton generation (MEG) can be obtained from the absorption of a
single photon by a quantum dot if the photon energy is at least two times higher
than its bandgap (Nozik, 2003; Schaller and Klimov, 2004). The challenge is to
find ways to collect the excitons before they recombine. As recombination occurs
on a picosecond time scale, the use of mesoporous oxides as electron collectors
presents a promising strategy, because the electron transfer from the quantum dot
to the conduction band of the oxide electrode occurs within femtoseconds (Plass
et al., 2002). This opens up research avenues that ultimately may lead to photon
converters reaching external quantum efficiencies over 100%. Recent results on
depleted heterojunctions of PbSe(ZnO) confirm this expectation (Semonin et al.,
2011). A calculation based on Henry’s model (Henry, 1980) shows that the maxi-
mum conversion efficiency of a single-junction cell could be increased from 34%
to 44% by exploiting MEG effects.

9.4.7 Mesoporous oxide film development

When the dye-sensitised nanocrystalline solar cell was first presented, perhaps the
most puzzling phenomenon was the highly efficient charge transport through the
nanocrystalline TiO2 layer. Mesoporous electrodes differ greatly from their com-
pact analogues because: 1) the inherent conductivity of the film is very low; 2) the
small size of the nanocrystalline particles does not support a built-in electrical field;
and 3) the electrolyte penetrates the porous film all the way to the back contact,
making the semiconductor/electrolyte interface essentially three-dimensional. The
mechanism of charge transport in mesoporous systems is still under debate and sev-
eral interpretations based on the Montrol–Scher model for random displacement
of charge carriers in disordered solids have been advanced (Nelson and Chandler,
2004). However, the ‘effective’ electron diffusion coefficient is expected to depend
on a number of factors such as trap filling and space-charge compensation by ionic
motion in the electrolyte. The factors controlling the rate of charge-carrier percola-
tion across the nanocrystalline film are presently under intense scrutiny. Intensity-
modulated impedance spectroscopy has proved to be an elegant and powerful tool
(Bisquert, 2002; Kubo et al., 2002; Cass et al., 2003; Frank et al., 2004; Peter,
2009; Fabregat-Santiago et al., 2005) to address these and other important ques-
tions related to the characteristic time constants for charge-carrier transport and
reaction dynamics in dye-sensitised nanocrystalline solar cells.
On the materials science side, future research has been directed towards
synthesising structures with a higher degree of order than a random assembly
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 439

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 439

of nanoparticles. A desirable morphology of the films would have the mesoporous

channels or nanorods aligned in parallel to each other and vertically with respect
to the TCO glass current collector. This would facilitate pore diffusion, give easier
access to the film surface, avoid grain boundaries and allow the junction to be
formed under better control. One approach to fabricate such oxide structures is
based on surfactant template-assisted preparation of TiO2 nanotubes, as described
by Jiu et al. (2005), another the use of mesoporous TiO2 single crystals (Crossland
et al., 2013). These and the hybrid nanorod–polymer composite cells developed by
Alivisatos and co-workers (Huynh et al., 2002) have confirmed the superior pho-
tovoltaic performance of such films as compared with random-particle networks.

9.4.8 Molecular engineering of the interface

The large contact area of the junction in nanocrystalline solar cells renders manda-
tory the grasp and control of interfacial effects for future improvement of cell
performance. The nature of the exposed surface planes of the oxide and the mode
of interaction with the dye is the first important information to gather. As far as
the adsorption of the N3 dye on TiO2 is concerned, this is now well understood.
The prevalent orientation of the anatase surface planes is (101) and the sensi-
tiser is adsorbed through two of the four carboxylate groups, at least one of them
being anchored via a bidentate configuration bridging two adjacent titanium sites
(Nazeeruddin et al., 2000). Molecular dynamic calculations employing a classical
force field have been carried out to predict the equilibrium geometry of the adsorbed
sensitiser state (Burnside et al., 1998; Shklover et al., 1998). More sophisticated
first-principle density functional calculations have also been undertaken (Vittadini
et al., 1998) to model the surface interactions of TiO2 with simple adsorbates as
well as the surface reconstruction effects resulting from the adsorption. The latter
approach is particularly promising and will provide an important tool for future
theoretical investigations.
Synthetic efforts focus on the molecular engineering of sensitisers that enhance
charge separation at the oxide solution interface. The structural features of the dye
should match the requirements for current rectification: by analogy to the photofield
effect in transistors, the gate for unidirectional electron flow from the electrolyte
through the junction and into the oxide is opened by the photoexcitation of the
sensitiser. The reverse charge flow, i.e. recapture of the electron by the electrolyte,
could be impaired by judicious design of the sensitiser, which is an insulator in the
ground state. The latter should form a tightly packed monolayer blocking the dark
current. The gain in open-circuit voltage can be calculated from the diode equation

Voc = (nRT /F) ln[(i sc /i o ) − 1] (9.4)

September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 440

440 M. Grätzel & J. R. Durrant

where n is the ideality factor, whose value is between 1 and 2 for DSSCs, and
i o is the reverse saturation current. Thus for each order of magnitude decrease
in the dark current, the gain in V oc would be between 59 and 118 mV at room
temperature. Work in this direction is indispensable to raise the efficiency of
the DSSC significantly over the 15% limit with the currently employed redox
electrolytes.

9.5 Solid-state mesoscopic cells based on molecular dyes or

perovskite pigments as sensitisers
9.5.1 Mesoscopic photovoltaics based on molecular sensitisers
Research on solid-state DSSCs has gained considerable momentum in recent years
as this embodiment is attractive for realising flexible PV cells in a roll-to-roll
production. The most successful p-type organic conductor employed to date is
spiro-OMeTAD, which has a work function of ∼5.1 eV and hole mobility of 2 ×
10−4 cm2 s−1 (see Fig. 9.11 for the structure of this material). First reported in
1998, conversion efficiencies of solid-state cells incorporating this hole conductor
have increased dramatically over the last few years, from a fraction of a percent
(Bach et al., 1998) to over 4% (Schmidt-Mende et al., 2005a), reaching presently
7.2% (Burschka et al., 2011). The main challenge posed by these cells has been fast
interfacial electron-hole recombination, reducing the electron diffusion length to
a few microns (Kruger et al., 2003) as compared with 20–100 µm for electrolyte-
based DSSCs. This restricts the film thickness employed in these cells to only
2 µm, which is insufficient for the adsorbed sensitiser to harvest all incident visible
sunlight, thus reducing the photocurrent. The dye monolayer can block this back
reaction to some extent because it is electrically insulating (Snaith et al., 2005).
Hence, current efforts are directed towards molecular engineering of the interface
to improve the compactness and order of the monolayer and prevent charge carriers
from recombining.
Another difficulty that has been encountered has been the filling of the porous
network with the hole conductor. This impediment may be overcome by devel-
oping oxide films having regular mesoporous channels aligned perpendicular to
the current collector. On the other hand, the Voc values obtained with solid-state
DSSCs are high, reaching nearly 1 V, owing to a better match of the hole conductor
work function with the redox potential of the sensitiser than that of the electrolyte.
The future of these solid hole-conductor systems thus looks very bright if the
recombination and pore-filling problems can be solved.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 441

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 441

9.5.2 Mesoscopic photovoltaics based on metal halide perovskites

The recent introduction of lead (Kojima et al., 2009) and tin halide (Chung et al.,
2012) perovskites as light harvesters and solid hole conductors in mesoscopic solar
cells has triggered one of the most exciting developments in the history of pho-
tovoltaics. While these compounds have been known for 120 years (Wells, 1893)
and have been extensively studied for optoelectronic applications (Hirasawa et al.,
1994) in, for example, light-emitting diodes (Chondroudis and Mitzi, 1999) and
thin-film transistors (Kagan et al., 1999), surprisingly their photovoltaic properties
have not been examined until very recently, even though it was remarked in 1958
that Cs3 PbI3 is a semiconductor exhibiting photoconductivity (Moller, 1958).
Photovoltaic investigations on metal halide perovskites started only in 2009
when methylammonium lead halides, of the composition CH3 NH3 PbX3 (X =
I, Br), were introduced as replacement for the dye sensitiser in liquid electrolyte-
based DSSCs. The PCE was initially 3.8% and device stability was poor. By chang-
ing the electrolyte Im et al. (2011) improved the performance and stability, attaining
a PCE of 6.5%. Importantly it became apparent that the CH3 NH3 PbI3 perovskite
is a better light harvester than the molecular N719 ruthenium sensitisers, enabling
the thickness of the mesoscopic titania film to be reduced significantly.
A key advance was subsequently made by replacing the liquid electrolyte with
the solid-state hole conductor spiro-OMeTAD (Kim et al., 2012; Lee et al., 2012).
Not only did the conversion efficiencies practically double but the cell stability
was vastly improved by avoiding the liquid solvent which dissolves the perovskite.
The studies by Lee et al. (2012) and Etgar et al. (2012) revealed that the perovskite
itself acts as an ambipolar semiconductor, obviating the need to employ a TiO2
scaffold or organic hole conductor. This prompted further intensive investigations,
which yielded a wealth of new information on these exciting systems. Today the
best-performing cells reach a PCE of 15% (Burschka et al., 2013), which is more
than four times higher than the PCE value attained in 2009 — an unprecedented
rapid rise in efficiency for a photovoltaic. Figure 9.12 shows a scanning electron
microscope (SEM) view of the cross-section of the cell and a photocurrent–voltage
curve for the device. The field is presently in a very dynamic state and reports on
further improvements in performance are expected in the near future.

9.6 Pilot production of modules, field tests and commercial

DSSC development
Prototypes of monolithic Z-type interconnected DSSC modules, fabricated by
Aisin Seiki in Japan, use carbon as a back contact to cut costs. Comparative field
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 442

442 M. Grätzel & J. R. Durrant

Figure 9.12 Top: Cross-sectional SEM photograph of a mesoscopic solar cell using
CH3 NH3 PI3 as light harvester. The perovskite is infiltrated into the mesopores of the TiO2
scaffold. FTO and HTM signify fluorine-doped tin dioxide and hole-transport material.
Bottom: i–V curve of the best-performing device.

tests of these modules and polycrystalline silicon (pc-Si) have been running for
several years (Figure 9.13 shows a photograph of the test station). The test results
revealed the advantages of the DSSC as compared with silicon modules under
realistic outdoor conditions: for equal rating under standard test conditions (STC),
the DSSC modules produced 20–30% more energy than pc-Si modules (Toyoda,
2006). Recent data presented by Sony (Noda, 2012) show that today’s DSSC mod-
ule, having a standard AM 1.5 conversion efficiency of 10%, is expected to produce
on average over one year the same electricity as a 15% standard-rated pc Si module.
The superior performance of the DSSC can be ascribed to the following factors:

• The DSSC efficiency is practically temperature-independent in the range

25–65◦ C while that of monocrystalline and pc-Si declines by ∼20% over
the same range.
• Outdoor measurements indicate that light capture by the DSSC is less sensitive
to the angle of incidence, although this needs to be further assessed. The DSSC
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 443

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 443

Figure 9.13 Outdoor field tests of DSSC modules by Fujikura Corporation near Tokyo,
Japan.

is more efficient than pc-Si in diffuse light or cloudy conditions. This advantage
is further enhanced for glass modules, which collect light in a bifacial way
from all angles.
While it is up to the commercial supplier to set the final price for such modules,
it is clear that the DSSC shares the cost advantages of all thin-film devices. In addi-
tion it uses no high-vacuum, cost-intensive steps in manufacture and only cheap
and readily available materials. Although it might be thought that the ruthenium-
based sensitiser adds high material cost, its contribution to cost is < US$ 0.01 W−1 p
because of the small amount used. Also, the present organic sensitisers match or
even surpass the performance of Ru complexes. Given these advantages at compa-
rable conversion efficiency, module costs below 0.5 W−1 p are realistic targets even
for production plants having well below GW capacity. The DSSC has thus become
a viable contender for large-scale solar energy conversion systems on the basis of
cost, efficiency, stability and availability, as well as environmental compatibility.
The British company G24i has started commercial sales of a flexible DSSC model.
Today the production rate is in the MW per year range.

9.7 Outlook
Using a principle derived from natural photosynthesis, mesoscopic injection solar
cells and in particular the DSSC, have become a credible alternative to solid-state
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 444

444 M. Grätzel & J. R. Durrant

p-n junction devices. Conversion efficiencies of 12.3% and 15% have already been
obtained with single-junction liquid and solid-state mesoscopic solar cells, respec-
tively, on the laboratory scale, but there is ample room for further improvement.
Future research will focus on improving short-circuit current by extending the light
response of the sensitisers in the near-IR spectral region. Substantial gains in the
open-circuit voltage are expected from introducing ordered oxide mesostructures
and controlling the interfacial charge recombination by judicious engineering on
the molecular level. Hybrid cells based on solid-state inorganic and organic hole
conductors are an attractive option, in particular for the flexible DSSC embodiment.
Solar cells based on perovskite light absorbers are showing very promising effi-
ciencies and are rapidly developing into a particularly exciting new research area.
Mesoscopic dye-sensitised cells are well suited for a whole realm of appli-
cations, ranging from the low-power market to large-scale applications. Their
excellent performance in diffuse light gives them a competitive edge over silicon
in providing electric power for stand-alone electronic equipment both indoors and
outdoors. Dye-sensitized solar cells are already being applied in building-integrated
PV and this will become a fertile field of future commercial development.
As the epigraph at the chapter opening shows, at the 1912 IUPAC conference
in Washington a hundred years ago, the famous Italian photochemist Professor
Giacomo Ciamician predicted that mankind would unravel the secrets of photo-
synthesis and apply the principles used by plants to harvest solar energy in glass
buildings. His visionary thoughts now appear close to becoming a reality.

Acknowledgements
We gratefully acknowledge the insights and support from the many co-workers
with whom we have had the pleasure of working on these devices, and particularly
thank Dr Brian O’Regan for his help in preparing this manuscript.

References
3GSolar (2011), ‘3GSolar announces record outdoor reliability of third generation photo-
voltaic technology’, http:// www.3Gsolar.com.
Anderson N. A. and Lian T. Q. (2005), ‘Ultrafast electron transfer at the molecule-
semiconductor nanoparticle interface’, Annu. Rev. Phys. Chem. 56, 491–519.
Anderson S., Constable E. C., Dare-Edwards M. P., Goodenough J. B., Hamnett A. and
Seddon K. R. (1979), ‘Chemical modification of a titanium(IV) oxide electrode to give
stable dye sensitization without a supersensitiser’, Nature 280, 571–573.
Arakawa H., Yamaguchi T., Okada K., Matsui K., Kitamura T. and Tanabe N. (2009), ‘Highly
durable dye-sensitised solar cells’, Fujikura Tech. Rev. 2009, 55–59.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 445

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 445

Bach U., Lupo D., Comte P., Moser J. E., Weissortel F. and Salbeck J. (1998), ‘Solid-state
dye-sensitised mesoporous TiO2 solar cells with high photon-to-electron conversion
efficiencies’, Nature 395, 583–585.
Barbe C. J., Arendse F., Comte P., Jirousek M., Lenzmann F. and Shklover V. (1997),
‘Nanocrystalline titanium oxide electrodes for photovoltaic applications’, J. Amer.
Ceramic Soc. 80, 3157–3171.
Barkhouse D. A. R., Debnath R., Kramer, I. J., Zhitomirsky D., Pattantyus-Abraham A. G.,
Levina L., Etgar L., Grätzel M. and Sargent E. H. (2011), ‘Depleted bulk heterojunction
colloidal quantum dot photovoltaics’, Adv. Mater. 23, 3134–3140.
Bessho T., Zakeeruddin S. M., Yeh C.-Y., Diau E. W. G. and Grätzel M. (2010), ‘Highly
efficient mesoscopic dye-sensitised solar cells based on donor-acceptor-substituted
porphyrins’, Angew. Chem. Int. Ed. 49, 6646–6649.
Bisquert J. (2002), ‘Theory of the impedance of electron diffusion and recombination in a
thin layer’, J. Phys. Chem. B 106, 325–333.
Burnside S. D., Shklover V., Barbe C., Comte P., Arendse F. and Brooks K. (1998),
‘Self-organisation of TiO2 nanoparticles in thin films’, Chem. Mater. 10, 2419–2425.
Burschka A., Dualeh F., Kessler F., Baranoff E., Cevey-Ha N.-L., Yi Md. C., Nazeeruddin
M. K. and Grätzel M. (2011), ‘Tris(2-(1H-pyrazol-1-yl) pyridine) cobalt(III) as p-type
dopant for organic semiconductors and its application in highly efficient solid-state
dye-sensitised solar cells’, J. Amer. Chem. Soc. 133, 18042–18045.
Burschka J., Pellet N., Moon S. J., Humphry-Baker R., Gao P., Nazeeruddin M. K. and
Grätzel M. (2013), ‘Sequential deposition as a route to high-performance perovskite-
sensitized solar cells’, Nature 499, 316–319.
Casanova D., Rotzinger F. P. and Grätzel M. (2010), ‘Computational study of promising
organic dyes for high-performance sensitized solar cells’, J. Chem. Theory Comput. 6,
1219–1227.
Cass M. J., Qiu F. L., Walker A. B., Fisher A. C. and Peter L. M. (2003), ‘Influence of
grain morphology on electron transport in dye sensitised nanocrystalline solar cells’,
J. Phys. Chem. B 107, 113–119.
Chondroudis K. and Mitzi D. B. (1999), ‘Electroluminescence from an organic-inorganic
perovskite incorporating a quaterthiophene dye within lead halide perovskite layers’,
Chem. Mater. 11, 3028–3030.
Chung I., Lee B., He J., Chang R. P. H. and Kanatzidis M. G. (2012), ‘All-solid-state dye-
sensitized solar cells with high efficiency’, Nature 485, 486–489.
Crossland E. J. W., Nakita N., Sivaram V., Leijtens T., Alexander-Webber J. A. and Snaith
H. J. (2013), ‘Mesoporous TiO2 single crystals delivering enhanced mobility and opto-
electronic device performance’, Nature 495, 215–219.
Dare-Edwards M. P., Goodenough J. B., Hamnett A., Seddon K. R. and Wright R. D. (1980),
‘Sensitization of semiconducting electrodes with ruthenium-based dyes’, Faraday Dis-
cussions 70, 285–298.
Desilvestro J., Grätzel M., Kavan L., Moser J. and Augustynski J. (1985), ‘Highly efficient
sensitization of titanium dioxide’, J. Amer. Chem. Soc. 107, 2988–2990.
De Wild-Scholten M. J. and Veltkamp A. C. (2007), ‘Environmental life cycle analysis
of dye sensitised solar devices: status and outlook’, 22nd EU-PVSEC, https://www.
ecn.nl/publicaties/ECN-M-07-081.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 446

446 M. Grätzel & J. R. Durrant

Dung D. H., Serpone N. and Grätzel M. (1984), ‘Integrated systems for water cleavage by
visible-light sensitization of TiO2 particles by surface derivatization with ruthenium
complexes’, Helvetica Chim. Acta 67, 1012–1018.
Durr M., Bamedi A., Yasuda A. and Nelles G. (2004), ‘Tandem dye-sensitised solar cell for
improved power conversion efficiencies’, Appl. Phys. Lett. 84, 3397–3399.
Durr M., Schmid A., Obermaier M., Rosselli S., Yasuda A. and Nelles G. (2005), ‘Low-
temperature fabrication of dye-sensitised solar cells by transfer of composite porous
layers’, Nature Materials 4, 607–611.
Durrant J. R., Haque S. A. and Palomares E. (2006), ‘Photochemical energy conversion:
from molecular dyads to solar cells’, Chem. Comm. 31, 3279–3289.
Dyesol (2013), ‘Dyesol exceeds stringent PV durability test by 400%’, May 28, www.
dyesol.com / posts / cat / corporate-news / post / dyesol-exceeds-stringent-pv-durability-
test-by-400-per-cent/.
Etgar L., Gao P., Xue Z., Peng Q., Chandiran A. K., Liu B., Nazeeruddin M. K. and Grätzel M.
(2012), ‘Mesoscopic CH3 NH3 PbI3 /TiO2 heterojunction solar cells’, J. Amer. Chem.
Soc. 134, 17396–17399.
Fabregat-Santiago F., Bisquert J., Garcia-Belmonte G., Boschloo G. and Hagfeldt A.
(2005), ‘Influence of electrolyte in transport and recombination in dye-sensitised
solar cells studied by impedance spectroscopy’, Solar Energy Mater. Solar Cells 87,
117–131.
Feldt S. M., Gibson E. A., Gabrielsson E., Sun L., Boschloo G. and Hagfeldt A. (2010),
‘Design of organic dyes and cobalt polypyridine redox mediators for high-efficiency
dye-sensitized solar cells’, J. Amer. Chem. Soc. 132, 16714–16724.
Ferrere S., Zaban A. and Gregg B. A. (1997), ‘Dye sensitization of nanocrystalline tin oxide
by perylene derivatives’, J. Phys. Chem. B 101, 4490–4493.
Frank A. J., Kopidakis N. and van de Lagemaat J. (2004), ‘Electrons in nanostructured TiO2
solar cells: transport, recombination and photovoltaic properties’, Coordination Chem.
Rev. 248, 1165–1179.
Gerischer H. and Tributsch H. (1968), ‘Elektrochemische Untersuchungen über den Mecha-
nismus der Sensibilisierung und Übersensibilisierung an ZnO-Einkristallen’, Ber. Bun-
senges. Phys. Chem. 72, 437–445.
Grätzel M. (2006), ‘Photovoltaic performance and long-term stability of dye-sensitised
mesosocopic solar cells’, Comptes Rend. Chimie 9, 578–583.
Grätzel M. (2009), ‘Recent advances in sensitised mesoscopic solar cells’, Acc. Chem. Res.
42, 1788–1798.
Grätzel M., Janssen R. A. J., Mitzi D. B. and Sargent E. H. (2012), ‘Materials interface
engineering for solution-processed photovoltaics’, Nature 488, 304–312.
Green A. N. M., Palomares E., Haque S. A., Kroon J. M. and Durrant J. R. (2005), ‘Charge
transport versus recombination in dye-sensitised solar cells employing nanocrystalline
TiO2 and SnO2 films’, J. Phys. Chem. B 109, 12525–12533.
Green M. A., Emery K., Hishikawa Y. and Warta W. (2011), ‘Solar efficiency tables (Ver-
sion 37)’, Progr. Photovoltaics 19, 84–92.
Hagfeldt A., Boschloo G., Sun L., Kloo L. and Pettersson H. (2010), ‘Dye-sensitized solar
cells’, Chem. Rev. 110, 6595–6663.
Hagfeldt A. and Grätzel M. (2000), ‘Molecular photovoltaics’, Acc. Chem. Res. 33,
269–277.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 447

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 447

Han L., Islam A., Chen H., Malapaka C., Chiranjeevi B., Zhang S., Yang X. and Yanagida M.
(2012), ‘High-efficiency dye-sensitised solar cell with a novel co-adsorbent’, Energy
Env. Science 5, 6057–6060.
Haque S. A., Palomares E., Cho B. M., Green A. N. M., Hirata N. and Klug D. R. (2005),
‘Charge separation versus recombination in dye-sensitised nanocrystalline solar cells:
the minimization of kinetic redundancy’, J. Amer. Chem. Soc. 127, 3456–3462.
Hara K., Sato T., Katoh R., Furube A., Yoshihara T. and Murai M. (2005), ‘Novel conjugated
organic dyes for efficient dye-sensitised solar cells’, Adv. Functional Mat. 15, 246–252.
Harikisun R. and Desilvestro H. (2011), ‘Long-term stability of dye solar cells’, Solar
Energy 85, 1179–1188.
Hengerer R., Kavan L., Krtil P. and Grätzel M. (2000), ‘Orientation dependence of
charge-transfer processes on TiO2 (anatase) single crystals’, J. Electrochem. Soc. 147,
1467–1472.
Henry C. H. (1980), ‘Limiting efficiencies of ideal single and multiple energy-gap terrestrial
solar cells’, J. Appl. Phys. 51, 4494–4500.
Hinsch A., Kroon J. M., Kern R., Uhlendorf I., Holzbock J. and Meyer A. (2001), ‘Long-term
stability of dye-sensitised solar cells’, Progr. Photovoltaics 9, 425–438.
Hirasawa M., Ishihara T., Goto T., Uchida K. and Miura N. (1994), ‘Magnetoabsorption
of the lower exciton in perovskite–type compound CH3 NH3 PbI3 ’, Physica B 291,
427–430.
Horiuchi T., Miura H., Sumioka K. and Uchida S. (2004), ‘High efficiency of dye-sensitised
solar cells based on metal-free indoline dyes’, J. Amer. Chem. Soc. 126, 12218–12219.
Hoyer P. and Weller H. (1995), ‘Potential-dependent electron injection in nanoporous col-
loidal ZnO films’, J. Phys. Chem. 99, 14096–14100.
Hsieh C.-P., Lu H.-P., Chiu C.-L., Lee C.-W., Chuang S.-H., Mai C.-L., Yen W.-N., Hsu S.-J.,
Diau E. W.-G. and Yeh C.-Y. (2010), ‘Synthesis and characterization of porphyrin sen-
sitizers with various electron-donating substituents for highly efficient dye-sensitized
solar cells’, J. Mater. Chem. 20, 1127–1134.
Huynh W. U., Dittmer J. J. and Alivisatos A. P. (2002), ‘Hybrid nanorod-polymer solar
cells’, Science 295, 2425–2427.
Im J.-H., Lee C. R., Lee J. W., Park S. W. and Park N. G. (2011), ‘6.5% efficient perovskite
quantum-dot-sensitized solar cell’, Nanoscale 3, 4088–4093.
Imahori H., Umeyama T. and Ito S. (2009), ‘Large π-aromatic molecules as potential sensi-
tizers for highly efficient dye-sensitized solar cells’, Acc. Chem. Res. 42, 1809–1818.
Ince M., Cardinali F., Raugossi M., Yum J., Gouloumis A., Martinez-Diaz M., Torre G.,
Nazeeruddin M., Grätzel M. and Torres T. (2012), ‘Phthalocyanines for molecular
photovoltaics’, Abstracts of Papers, 243rd ACS National Meeting and Exposition, San
Diego, CA, 25–29 March 2012, FUEL-522.
Ito S., Miura H., Uchida S., Takata M., Sumioka K., Liska P., Comte P., Pechy P. and
Grätzel M. (2008), ‘High conversion efficiency organic dye-sensitised solar cells with
a novel indoline dye’, Chem. Commun. 41, 5194–5196.
Ito S., Zakeeruddin S. M., Humphry-Baker R., Liska P., Charvet R. and Comte P. (2006),
‘High-efficiency organic-dye-sensitised solar cells controlled by nanocrystalline TiO2
electrode thickness’, Adv. Mater. 18, 1202–1205.
Jiu J. T., Wang F. M., Isoda S. and Adachi M. (2005), ‘Highly efficient dye-sensitised solar
cells based on single crystalline TiO2 nanorod film’, Chem. Lett. 34, 1506–1507.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 448

448 M. Grätzel & J. R. Durrant

Kagan C. R., Mitzi D. B. and Dimitrakopoulos C. D. (1999), ‘Organic-inorganic hybrid

materials as semiconducting channels in thin-film field-effect transistors’, Science 286,
945–947.
Kalyanasundaram K. and Graetzel M. (2010), ‘Artificial photosynthesis: biomimetic
approaches to solar energy conversion and storage’, Current Opinion Biotechnol. 21,
298–310.
Kim H. S., Lee C. R., Im J. H., Lee K. B., Moehl T., Marchioro A., Moon S. J., Humphry-
Baker R., Yum J. H., Moser J. E., Grätzel M. and Park N. G. (2012), ‘Lead iodide
perovskite sensitized all-solid-state submicron thin film mesoscopic solar cell with
efficiency exceeding 9%’, Nature Sci. Reports 2, 591.
Kinoshita,T., Dy J. T., Uchida S., Kubo T. and Segawa H. (2013), ‘Wideband dye-sensitized
solar cells employing a phosphine-coordinated ruthenium sensitizer’, Nat. Photonics
7, 535–539.
Kinoshita T., Dy J. T., Uchida S., Kubo T. and Segawa H. (2013), ‘Wideband dye-sensitized
solar cells employing a phosphine-coordinated ruthenium sensitizer’, Nature Photonics
7, 535–539.
Kojima A., Teshima K., Shirai Y. and Miyasaka T. (2009), ‘Organometal halide per-
ovskites as visible-light sensitizers for photovoltaic cells’, J. Amer. Chem. Soc. 131,
6050–6051.
Koops S. E., O’Regan B. C., Piers R. F., Barnes P. R. F. and Durrant J. R. (2009), ‘Parameters
influencing the efficiency of electron injection in dye-sensitised solar cells’, J. Amer.
Chem. Soc. 131, 4808–4818.
Kruger J., Plass R., Grätzel M., Cameron P. J. and Peter L. M. (2003), ‘Charge transport and
back reaction in solid-state dye-sensitised solar cells: a study using intensity-modulated
photovoltage and photocurrent spectroscopy’, J. Phys. Chem. B 107, 7536–7539.
Kubo W., Kitamura T., Hanabusa K., Wada Y. and Yanagida S. (2002), ‘Quasi-solid-state
dye-sensitised solar cells using room temperature molten salts and a low molecular
weight gelator’, Chem. Comm. 4, 374–375.
Kubo W., Sakamoto A., Kitamura T., Wada Y. and Yanagida S. (2004), ‘Dye-sensitised solar
cells: improvement of spectral response by tandem structure’, J. Photochem. Photobiol.
A — Chemistry 164, 33–39.
Lee M. M., Teuscher J., Miyasaka T., Murakami T. N. and Snaith H. J. (2012), ‘Effi-
cient hybrid solar cells based on meso-superstructured organometal halide perovskites’,
Science 338, 643–647.
Liska P., Thampi K. R., Grätzel M., Bremaud D., Rudmann D. and Upadhyaya H. M.
(2006), ‘Nanocrystalline dye-sensitised solar cell/copper indium gallium selenide thin-
film tandem showing greater than 15% conversion efficiency’, Appl. Phys. Lett. 88,
203103–203106.
Listorti A., O’Regan B. and Durrant J. R. (2011), ‘Electron transfer dynamics in dye-
sensitized solar cells’, Chem. Mater. 23, 3381–3399.
Mann S. A., de Wild-Scholten M. J., Fthenakis V. M., van Sark W. G. J. H. M. and
Sinke W. C. (2013), ‘The energy payback time of advanced crystalline PV modules
in 2020: a prospective study’, Progr. Photovoltaics, http://onlinelibrary.wiley.com/doi/
10.1002/pip.2363/full.
McEvoy A. J. and Grätzel M. (1994), ‘Sensitization in photochemistry and photovoltaics’,
Solar Energy Mat. Solar Cells 32, 221–227.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 449

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 449

McEvoy A. J. and Grätzel M. (2013), ‘Nanotechnology in dye sensitized solar cells’, in

Garcia-Martinez J. (ed.), Nanotechnology for Energy Challenges, Wiley, Weinheim.
Mishra A., Fischer M. K. R. and Bauerle P. (2009), ‘Metal-free organic dyes for dye-
sensitized solar cells: from structure:property relationships to design rules’, Angew.
Chemie Int. Ed. 48, 2474–2499.
Moller C. K. (1958), ‘Crystal structure and photoconductivity of caesium plumbohalides’,
Nature 182, 1436.
Moser J. (1887), ‘Notiz über Verstärkung photoelectrischer Ströme durch optische Sensi-
bilisierung’, Monatsch. Chem. 8, 373.
Namba S. and Hishiki Y. (1965), ‘Color sensitization of zinc oxide with cyanine dyes’,
J. Phys. Chem. 69, 774–779.
Nanu M., Schoonman J. and Goossens A. (2005), ‘Solar energy conversion in TiO2 / CuInS2
nanocomposites’, Adv. Functional Mater. 15, 95–100.
Nazeeruddin M. K., Amirnasr M., Comte P., Mackay J. R., McQuillan A. J. and Houriet R.
(2000), ‘Adsorption studies of counterions carried by the sensitiser cis-dithiocyanato
(2,2 -bipyridyl-4,4 -dicarboxylate) ruthenium(II) on nanocrystaline TiO2 films’, Lang-
muir 16, 8525–8528.
Nazeeruddin M. K., Kay A., Rodicio I., Humphry-Baker R., Muller E. and Liska P. (1993),
‘Conversion of light to electricity by cis-X2 -bis(2,2 -bipyridyl-4,4 -dicarboxylate)
ruthenium(II) charge-transfer sensitisers (X = Cl− , Br− , I− , CN− and SCN− ) on
nanocrystalline TiO2 electrodes’, J. Amer. Chem. Soc. 115, 6382–6390.
Nazeeruddin M. K., Pechy P., Renouard T., Zakeeruddin S. M., Humphry-Baker R. and
Comte P. (2001), ‘Engineering of efficient panchromatic sensitisers for nanocrystalline
TiO2 -based solar cells’, J. Amer. Chem. Soc. 123, 1613–1624.
Nelson J. and Chandler R. E. (2004), ‘Random walk models of charge transfer and transport
in dye-sensitised systems’, Coordination Chem. Rev. 248, 1181–1194.
Ning, Z. J., Fu, Y. and Tian, H. (2010), ‘Improvement of dye-sensitized solar cells: what we
know and what we need to know’, Energy Envir. Sci. 3, 1170–1181.
Noda K. (2012), Lecture, ‘Dye-sensitized Solar Cell Development at Sony’, PV Expo,
Tokyo, 29 February, Japan.
Nogueira A. F., Longo C. and de Paoli M. A. (2004), ‘Polymers in dye-sensitised solar cells:
overview and perspectives’, Coordination Chem. Rev. 248, 1455–1468.
Noh J. H., Im S. H., Heo J. H., Mandal T. N. and Seok S. I. (2013), ‘Chemical management
for colorful, efficient, and stable inorganic-organic hybrid nanostructured solar cells’,
Nano Lett. 13, 1764–1769.
Nozik A. J. (2003), ‘Quantum dot solar cells’, in Marti A. and Luque A. (eds), Next Gener-
ation Photovoltaics: High Efficiency Through Full Spectrum Utilization, CRC Press,
Boca Raton.
O’Regan B. and Grätzel M. (1991), ‘A low-cost, high-efficiency solar-cell based on dye-
sensitised colloidal TiO2 films’, Nature 353, 737–740.
O’Regan B. and Schwartz D. T. (1998), ‘Large enhancement in photocurrent efficiency
caused by UV illumination of the dye-sensitised heterojunction TiO2 /RuLL’NCS/
CuSCN: initiation and potential mechanisms’, Chem. Mater. 10, 1501–1509.
O’Regan B. C. and Durrant J. R. (2009), ‘Kinetic and energetic paradigms for dye-
sensitized solar cells: moving from the ideal to the real’, Acc. Chem. Res. 42, 1799–
1808.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 450

450 M. Grätzel & J. R. Durrant

Pattantyus-Abraham A. G., Kramer I. J., Barkhouse A. R., Wang Y., Konstantatos G.,
Debnath R., Levina L., Raabe I., Nazeeruddin M. K., Grätzel M. and Sargent E.
H. (2010), ‘Depleted-heterojunction colloidal quantum dot solar cells’, ACS-Nano 6,
3374–3380.
Peter L. (2009) “‘Sticky electrons” transport and interfacial transfer of electrons in the
dye-sensitized solar cell’, Accounts Chem. Res. 42, 1839–1847.
Peter L. M. and Wijayantha K. G. U. (2000), ‘Electron transport and back reaction in dye-
sensitised nanocrystalline photovoltaic cells’, Electrochim. Acta 45, 4543–4551.
Plass R., Pelet S., Krueger J., Grätzel M. and Bach U. (2002), ‘Quantum dot sensitization
of organic–inorganic hybrid solar cells’, J. Phys. Chem. B 106, 7578–7580.
Qin H., Wenger S., Xu M., Gao F., Jing X., Wang P., Zakeeruddin S.-M. and Grätzel M.
(2008), ‘An organic sensitiser with a fused dithienothiophene unit for efficient and
stable dye-sensitised solar cells’, J. Amer. Chem. Soc. 130, 9202–9203.
Qin, Y. C. and Peng, Q. (2012), ‘Ruthenium sensitizers and their applications in dye
sensitized solar cells’, Int. J. Photoenergy, http://www.hindawi.com/journals/ijp/
2012/291579/.
Reddy P. Y., Giribabu L., Lyness C., Snaith H. J., Vijaykumar C., Chandrasekharam M.,
Lakshmikantam M., Yum J.-H., Kalyanasundaram K., Grätzel M. and Nazeerud-
din M. K. (2007), ‘Efficient sensitization of nanocrystalline TiO2 films by a near-
IR-absorbing unsymmetrical zinc phthalocyanine’, Angewandte Chem. Int. Ed. 46,
373–376.
Redmond G. and Fitzmaurice D. (1993), ‘Spectroscopic determination of flat-band poten-
tials for polycrystalline TiO2 electrodes in nonaqueous solvents’, J. Phys. Chem. 97,
1426–1430.
Rensmo H., Keis K., Lindstrom H., Sodergren S., Solbrand A. and Hagfeldt A. (1997),
‘High light-to-energy conversion efficiencies for solar cells based on nanostructured
ZnO electrodes’, J. Phys. Chem. B 101, 2598–2601.
Rhee Y. M. and Head-Gordon M. (2007), ‘Scaled second-order perturbation correction
to configuration interaction singles: efficient and reliable excitation energy methods’,
J. Phys. Chem. 111, 5314–5326.
Rothenberger G., Fitzmaurice D. and Grätzel M. (1992), ‘Optical electrochemistry. 3.
Spectroscopy of conduction-band electrons in transparent metal-oxide semiconductor
films — optical determination of the flat-band potential of colloidal titanium dioxide
films’, J. Phys. Chem. 96, 5983–5986.
Sargent E. H. (2012), ‘Colloidal quantum dot solar cells’, Nat. Photonics 6, 133–135.
Sauvage F., Chhor S., Marchioro A., Moser J.-E. and Grätzel M. (2011), ‘Butyronitrile-based
electrolyte for dye-sensitised solar cells’, J. Amer. Chem. Soc. 133, 13103–13109.
Sayama K., Sugihara H. and Arakawa H. (1998), ‘Photoelectrochemical properties of
a porous Nb2 O5 electrode sensitised by a ruthenium dye’, Chem. Mater. 10,
3825–3832.
Schaller R. D. and Klimov V. I. (2004), ‘High efficiency carrier multiplication in
PbSe nanocrystals: implications for solar energy conversion’, Phys. Rev. Lett. 92,
86601–86604.
Schmidt-Mende L., Bach U., Humphry-Baker R., Horiuchi T., Miura H. and Ito S. (2005a),
‘Organic dye for highly efficient solid-state dye-sensitised solar cells’, Adv. Materials
17, 813–815.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 451

Dye- and Perovskite-Sensitised Mesoscopic Solar Cells 451

Schmidt-Mende L., Kroeze J. E., Durrant J. R., Nazeeruddin M. K. and Grätzel M. (2005b),
‘Effect of hydrocarbon chain length of amphiphilic ruthenium dyes on solid-state dye-
sensitised photovoltaics’, Nano Lett. 5, 1315–1320.
Semonin O. E., Luther J. M., Choi S., Chen H.-Y., Gao J., Nozik A. J. and Beard M. C.
(2011), ‘Peak external quantum efficiency exceeding 100% via MEG in a quantum dot
solar cell’, Science 334, 1530–1533.
Schmickler W. (1996), Interfacial Electrochemistry, Oxford University Press, Oxford.
Shklover V., Ovchinnikov Y. E., Braginsky L. S., Zakeeruddin S. M. and Grätzel M. (1998),
‘Structure of organic/inorganic interface in assembled materials comprising molecu-
lar components. Crystal structure of the sensitiser bis[(4,4 -carboxy-2,2 -bipyridine)
(thiocyanato)]ruthenium(II)’, Chem. Mater. 10, 2533–2541.
Snaith H. J., Moule A. J., Klein C., Meerholz K., Friend R. H. and Grätzel M. (2007), ‘Effi-
ciency enhancements in solid-state hybrid solar cells via reduced charge recombination
and increased light capture’, Nano Letters 7, 3372–3376.
Snaith H. J., Zakeeruddin S. M., Schmidt-Mende L., Klein C. and Grätzel M. (2005), ‘Ion-
coordinating sensitiser in solid-state hybrid solar cells’, Angew. Chem. Int. Ed. 44,
6413–6417.
Stathatos E., Lianos R., Zakeeruddin S. M., Liska P. and Grätzel M. (2003), ‘A quasi-solid-
state dye-sensitised solar cell based on a sol–gel nanocomposite electrolyte containing
ionic liquid’, Chem. Mater. 15, 1825–1829.
Tennakone K., Kumara G. R. R. A., Kumarasinghe A. R., Wijayantha K. G. U. and
Sirimanne P. M. (1995), ‘A dye-sensitised nanoporous solid-state photovoltaic cell’,
Semiconductor Sci. Technol. 10, 1689–1693.
Toyoda T. (2006), ‘Outdoor tests of large area dye-sensitised solar cell modules by Aisin
Seiki, Inc.’, Re2009 Renewable Energy Congress, Makura, Japan, October.
Tsubomura H., Matsumura M., Nomura Y. and Amamiya T. (1976), ‘Dye-sensitised zinc
oxide–aqueous electrolyte–platinum photocell’, Nature 261, 402–403.
Vittadini A., Selloni A., Rotzinger F. P. and Grätzel M. (1998), ‘Structure and energetics
of water adsorbed at TiO2 anatase (101) and (001) surfaces’, Phys. Rev. Lett. 81,
2954–2957.
Wang P., Klein C., Humphry-Baker R., Zakeeruddin S. M. and Grätzel M. (2005a), ‘A high
molar extinction coefficient sensitiser for stable dye-sensitised solar cells’, J. Amer.
Chem. Soc. 127, 808–809.
Wang P., Klein C., Humphry-Baker R., Zakeeruddin S. M. and Grätzel M. (2005b), ‘Stable
8% efficient nanocrystalline dye-sensitised solar cell based on an electrolyte of low
volatility’, Appl. Phys. Lett. 86, 1235081–1235083.
Wang P., Zakeeruddin S. M., Humphry-Baker R., Moser J. E. and Grätzel M. (2003a),
‘Molecular-scale interface engineering of TiO2 nanocrystals: improving the efficiency
and stability of dye-sensitised solar cells’, Adv. Mater. 15, 2101–2103.
Wang P., Zakeeruddin S. M., Moser J. E., Nazeeruddin M. K., Sekiguchi T. and Grätzel M.
(2003b), ‘A stable quasi-solid-state dye-sensitised solar cell with an amphiphilic ruthe-
nium sensitiser and polymer gel electrolyte’, Nat. Materials 2, 402–407.
Walter M. G., Rudine A. B. and Wamser C. C. (2010), ‘Porphyrins and phthalocyanines in
solar photovoltaic cells’, J. Porphyrins and Phthalocyanines 14, 759–792.
Wells H. L. (1893), ‘Uber die Casium- und Kalium-Blei halogenide’, Z. Anorg. Chem. 3,
195–210.
September 17, 2014 11:3 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch09 page 452

452 M. Grätzel & J. R. Durrant

West W. (1874), Vogel Centennial Symp., Photogr. Sci. Eng. 18, 35.
Wenger S., Seyrling S., Tiwari A. N. and Graetzel M. (2009), Fabrication and performance
of a monolithic dye-sensitized TiO2 /Cu(In,Ga)Se2 thin film tandem solar cell, Appl.
Phys. Lett. 94, 173508/1–173508/3.
Yella A., Lee H.-W., Tsao H. N, Yi C., Chandiran A. K., Nazeeruddin M. K., Diau E. W.-G.,
Yeh C.-Y., Zakeeruddin S. M. and Grätzel M. (2011), ‘Porphyrin-sensitised solar cells
with cobalt (II/III)-based redox electrolyte exceed 12% efficiency’, Science 334, 629–
634.
Yum J.-H., Baranoff E., Kessler F., Moehl T., Ahmad S., Bessho T., Marchioro A., Ghadiri E.,
Moser J.-E., Yi C., Nazeeruddin M. K. and Grätzel M. (2012), ‘A cobalt complex redox
shuttle for dye-sensitised solar cells with high open circuit potentials’, Nature Commun.
3, Art. 631.
Yum J.-H., Hagberg D. P., Moon S.-J., Karlsson K. M., Marinado T., Sun L., Hagfeldt A.,
Nazeeruddin M. K. and Grätzel M. (2009), ‘A light-resistant organic sensitiser for
solar-cell applications’, Angew. Chem. Int. Ed. 48, 1576–1580.
Zaban A., Ferrere S. and Gregg B. A. (1998), ‘Relative energetics at the semiconductor
sensitizing dye electrolyte interface’, J. Phys. Chem. B 102, 452–460.
Zhang G., Bala H., Cheng Y., Shi D., Lv X., Yu Q. and Wang P. (2009), ‘High efficiency
and stable dye-sensitised solar cells with an organic chromophore featuring a binary
π-conjugated spacer’, Chem. Comm. 16, 2198–2200.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 453

CHAPTER 10

QUANTUM WELL SOLAR CELLS

JENNY NELSON and NED EKINS-DAUKES
Blackett Laboratory, Imperial College London, SW7 2BZ, UK
[email protected]

No steam or gas ever drives anything until it is confined. No Niagara

is ever turned into light and power until it is tunneled.
Harry Emerson Fosdick.

10.1 Introduction
The quantum well solar cell (QWSC) was proposed by Barnham and Duggan
(1990) as a new type of multiple-bandgap, single-junction solar cell. The principle
is similar to that of the monolithic tandem cell: several bandgaps are used to absorb
different spectral ranges efficiently. However, rather than using two junctions made
from different semiconductors, the QWSC uses ultra-thin layers of different mate-
rials in a monolithic, two-terminal arrangement. The technology of quantum wells
(QWs) is borrowed from the optoelectronics industry. The 1980s and 1990s saw
developments in the use of low-dimensional semiconductor structures in optical
modulators, photodetectors, lasers and other devices (Weisbuch and Vinter 1991;
Zory, 1993). These applications exploit the special optical and electronic properties
available with highly confined carrier populations. Before 1989, low-dimensional
systems had been considered in connection with solar cells only in the context of
improved carrier transport through the neutral regions. In the QWSC, QWs are used
in the active region of the solar cell to enhance solar photon absorption and boost
the photocurrent. The QWs can be considered approximately as an incremental
current source in parallel with a conventional homojunction solar cell.
The use of low-dimensional semiconductor structures in solar cells was first
proposed in the early 1980s, where QW and superlattice arrays were introduced
into the doped regions of a tandem device based on p-n junctions (Wanlass and
Blakeslee 1982; Chaffin et al., 1984; Goradia et al., 1985). However, poor collection
due to inefficient carrier diffusion within a doped QW structure led to performances
inferior to those of bulk alloys, and the idea was abandoned.

453
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 454

454 J. Nelson & N. Ekins-Daukes

The usual and most effective geometry is to insert an array of QWs called
a multiple QW (MQW) into the depletion region (rather than the doped region)
of a p-n- or p-i -n-junction solar cell. Quantum wells absorb more solar photons
and so enhance the photocurrent. However, because of their lower bandgap, QWs
also enhance the dark recombination current that opposes the photocurrent. If
the increase in photocurrent exceeds the increase in dark current under operating
conditions, then the power-conversion efficiency of the cell will be increased.
The possibility of a higher limiting efficiency than is possible with a simple
single-junction cell is the main reason for interest in the QWSC, but QW structures
have also been considered as a means of improving the performance of practical
solar cells. For instance, the ability to control the bandgap through QW width is
useful in various applications, and the response of QW cells to increases in temper-
ature makes them attractive for solar concentration. The properties of QWSCs have
been studied, theoretically and experimentally, by a number of research groups.
Significant milestones in the development of the QWSC include the demonstration
of increased photocurrent from photogeneration in the quantum wells of a QWSC
(Barnham et al., 1991; Freundlich et al., 1994) and an understanding of the carrier
escape and current transport processes in forward bias (Nelson et al., 1993; Kitatani
et al., 1995). In parallel, the opportunity for a fundamental efficiency advantage was
debated (Corkish and Green, 1993; Araujo et al., 1994; Ragay et al., 1994; Luque
et al., 2001) and full device models were developed (Renaud et al., 1994; Anderson,
1995; Nelson, 1995; Ramey and Khoie, 2003). More recently, developments in so-
called ‘strain-balanced’ QW structures enabled high efficiency single-junction and
tandem devices. This development made QWSC technology potentially viable for
applications in space and terrestrial concentrator systems. Quantum well solar cell
technology is now being commercialised by JDS Uniphase Corporation. Progress
in epitaxial growth of highly strained layers has enabled strain-balanced superlat-
tice solar cells to be achieved (Wanget al., 2012).
This chapter covers the following areas: QW technology and materials issues;
the physics of QWs and how they differ from bulk materials in the processes
important for solar energy conversion; experimental results showing how these
properties affect the performance of the solar cell; the question of the limits to
efficiency; and practical applications of QWSCs.

10.2 Device design, materials and technology

10.2.1 Device design
The typical QWSC is a p-i -n junction containing a number of QWs in the intrinsic
(i ) region, as shown schematically in Fig. 10.1. The i region acts as a spacer layer
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 455

Quantum Well Solar Cells 455

p
n
n
EF
p qV Eb
EF

Figure 10.1 Schematic diagram of energy band vs. distance for a p-i-n QWSC at forward
bias V . Quantum wells are placed in the undoped i-layer, where the electric field helps to
separate photogenerated electron–hole pairs. Under illumination, the charges are separated
p
by the electric field and the electron and hole quasi-Fermi levels E F and E Fn are split by qV,
as shown.

to extend the field-bearing region. The rationale is that QWs should be placed in
this region where charge separation and collection is most efficient.
The QWs are thin layers of a second, narrower gap semiconductor between
barrier layers of the host material. III–V semiconductors are normally used for
both QW and host material. Typically the QWs are 60–150 Å wide, separated by
barriers of 50 Å or more. Some 50 QWs can be placed in an i -region 0.5–1 µm
thick. In practice the i -layer thickness is limited by the background level of charged
impurities, since these can cause the electric field to fall to zero within the i -layer
and render the remaining, neutral part of the junction useless (Zachariou et al.,
1996; Barnham et al., 1991).
The remaining parts of the cell — the p-layer, n-layer, substrate, window lay-
ers, contacts and optical coatings — are modelled on conventional III–V solar cell
designs. The emitter (top layer) may be somewhat narrower in order to admit as
much light as possible to the active i -region. Otherwise, the same design considera-
tions apply: choice of polarity ( p-n or n- p); techniques to enhance minority-carrier
collection such as surface passivation, window layers and graded emitters; choice
of materials for dopant, antireflective (AR) coat and substrate. The original goal
was to use the incremental photocurrent provided by the QWs to enhance the
efficiency of a well-designed homojunction cell.
An additional advantage is the possibility of controlling the QW width to tune
the absorption edge of one or more cells in a multijunction device.

10.2.2 Materials
Quantum well solar cells have been studied in several III–V materials combi-
nations. A short summary of some key examples follows, expressed as Barrier/
QW material: AlGaAs/GaAs (Barnham et al., 1991), GaAs/InGaAs (Barnes
et al., 1996), InP/InGaAs (Zachariou et al., 1998), InGaP/GaAs (Barnham et al.,
1996), InP/InAsP (Monier et al., 1998), InGaAsP/InGaAs (Rohr et al., 2006) and
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 456

456 J. Nelson & N. Ekins-Daukes

Figure 10.2 Plot of bandgap vs. lattice constant for III–V semiconductors.

GaAs/GaAsN (Freundlich et al., 2007). In choosing a compatible pair of materials

the lattice constant of the two should match approximately with each other and
with a suitable substrate. In addition, some strain-balanced combinations of QW
and barrier material where a lattice mismatch is accommodated through regions of
alternating strain have been investigated, notably GaAsP/InGaAs (Ekins-Daukes
et al., 1999) and Inx GaAs/In y GaAs (Connolly and Rohr, 2003).
As shown in Fig. 10.2, a wide range of lattice constants and bandgaps, spanning
the infrared (IR) to ultraviolet (UV) parts of the spectrum, can be achieved using
alloys of III–V semiconductor material. The AlGaAs/GaAs system has been most
used as a test system to investigate the basic physical principles of the QWSC,
because this combination has been widely used in optoelectronics and is well
understood. AlGaAs/GaAs QW solar cells are not in themselves interesting as
high-efficiency devices, since the bandgap of AlGaAs is higher than is optimal
for solar energy conversion, and performance can always be improved simply by
reducing the amount of Al.
The technologically relevant systems are gallium arsenide (GaAs, E g =
1.42 eV) or indium phosphide (InP, E g = 1.34 eV) p-n cells to which QWs have
been added. Both materials have bandgaps close to the optimum for solar energy
conversion. GaAs is currently a suitable material for high-efficiency multijunction
PV systems, since an efficient, high-bandgap (1.9 eV) cell composed of InGaP can
be grown lattice-matched to GaAs. These semiconductors can form the basis for
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 457

Quantum Well Solar Cells 457

the industry-standard InGaP/GaAs/Ge triple-junction solar cell that is presently

used in terrestrial concentrator systems and to power space satellites. However,
this material system is constrained by the bandgap of Ge, which is too low. The
optimal would be 1 eV but there is presently a lack of high-quality compound
semiconductors with this bandgap lattice that is matched to GaAs.
InP is attractive on account of its radiation hardness and, unlike GaAs, it has
a number of alloys suitable for forming low-gap solar cells. However, it suffers
from a relative lack of the lattice-matched high-gap alloys that are necessary for a
triple-junction solar cell.
The obvious material for QWs in GaAs is the alloy In x Ga1−x As, where
0 < x < 1. GaAs/In x Ga1−x As structures are widely used in optoelectronics in IR
photodetectors and lasers. However, they tend to suffer from strain-induced disloca-
tions arising from the difference in the lattice constants of GaAs and Inx Ga1−x As,
which increases with increasing indium fraction x. These dislocations degrade
the device quality and impose a limit on the number and depth of QWs that may
be added to a p-i -n structure (Griffin et al., 1996). One means to overcome this
limitation is to compensate the compressive strain introduced by the QWs with a
tensile barrier material that is different from the p and n host material. While each
bilayer remains strained, the QWs exert no net force on overlying layers, enabling
thick stacks of strained semiconductor material to be grown (Ekins-Daukes et al.,
2002). The GaAsP/InGaAs material system is one such strain-balanced alloy com-
bination that enables materials with absorption thresholds below that of GaAs to
be achieved, using strained semiconductors yet without introducing dislocations
(Ekins-Daukes et al., 1999). The value of strain-balanced structures was also recog-
nised by early pioneers of QW superlattice structures in solar cells (Chaffin et al.,
1984).
To achieve an efficient four-junction solar cell, a 1 eV threshold absorber is
required and achieving this on a GaAs substrate requires the use of dilute nitride
materials. In this respect, GaAs/GaAsN QW structures have been shown to achieve
an absorption threshold of 1.1 eV (Freundlich et al., 2007).
For InP, the best QW material is In0.53 Ga0.47 As. At this composition Inx Ga1−x
As is lattice-matched to InP, eliminating strain problems. Since In0.53 Ga0.47 As has
a much narrower bandgap than InP, the QWs are deep, extending absorption out to
around 1600 nm, so they offer significant photocurrent enhancements by utilising
IR solar radiation. In addition InP/In x As1−x P QW structures have shown radiation
resistance equal or superior to that of an InP homojunction (Walters et al., 2000).
For thermophotovoltaics, where a lower bandgap is needed because of the
lower source temperature, InGaAsP/In x Ga1−x As QW cells have been studied (see
Section 10.6), and InGaAsSb/GaSb is likely to be a promising system.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 458

458 J. Nelson & N. Ekins-Daukes

Also, QWSCs have been fabricated in the GaN/InGaN material system,

extending the spectral response of a GaN-based solar cell to 520 nm (Farrell et al.,
2011).

10.2.3 Fabrication technology

III–V QW structures are grown by epitaxial techniques such as molecular beam
epitaxy (MBE) and metal organic vapour phase epitaxy (MOVPE). Layers of con-
trolled composition are deposited, one atomic layer at a time, on a monocrystalline
substrate of the same or another lattice-matched III–V material. These methods are
also used to grow compound heterostructures for windowed and tandem solar cells
and can reach growth rates of several microns per hour. When growing precise QW
heterostructures, the growth rate is slowed, but this is the only significant difference
relative to the growth of conventional III–V solar cells. The manufacturing cost
of III–V concentrator solar cells is spread roughly equally between the substrate
cost, epitaxy cost and cell-processing cost. Growth on inexpensive substrates (e.g.
silicon) or epitaxial lift-off and wafer reuse (Bauhuis et al., 2010) can help reduce
the substrate cost. Epitaxy and processing costs may be reduced through high
growth rate deposition methods to be developed specifically for III–V photovoltaic
devices, as well as efficient, high-throughput metallisation technologies. In paral-
lel with the microelectronics industry, when the market grows sufficiently large,
transitioning to larger substrate sizes (6-inch GaAs or 12-inch Si substrates) will
help reduce all these manufacturing costs. Since cell efficiency directly impacts
the levelised cost of electricity, surprisingly small increases in efficiency can offset
higher manufacturing cost. A quantum well cell technology must therefore offer
ample efficiency premium to become commercially attractive.

10.2.4 Other materials

The possibilities of using QW structures for photoconversion in materials other than
III–V semiconductors have been, to date, largely unexplored. The first attempt to
grow a II–IV QWSC using the materials combination CdTe/CdMnTe (Scott et al.,
1997) was hindered by problems with series resistance. II–IV quantum dot (QD)
structures, developed mainly for their electroluminescent properties, have been
widely applied as n-type components of hybrid organic/inorganic bulk heterojunc-
tion solar cells, but these do not exploit the quantum properties of the semiconductor
nanoparticles and are reviewed elsewhere (Moulé et al., 2012; McGehee, 2009).
More recently, chalcopyrite QWSCs have been considered and while interdiffusion
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 459

Quantum Well Solar Cells 459

of elements has been observed, this is considered not to be a major limitation for
a functioning device (Afshar et al., 2011).
Si-based QWs and superlattices have also been investigated in an attempt to
fabricate an all-silicon tandem solar cell (Green, 2003), exploiting the 1.1 eV bulk
silicon junction as a bottom cell. Three-dimensional confinement is required to
achieve a 2 eV gap in Si for the top junction. For this purpose, arrays of Si QDs
confined by SiO and SiN barriers have been fabricated with an absorption threshold
approaching this value, although in common with II–VI QD structures, efficient
current transport through the QD array remains a challenging problem (Conibeer
et al., 2010).

10.3 Physics of QWs

Quantum wells possess a distinct electronic structure due to the confinement of
charge carriers into thin semiconductor layers. This special electronic structure
affects the physical processes of photon absorption, carrier relaxation and transport,
which are of central importance to photovoltaic energy conversion. In this section,
we will discuss the electronic structure of the QW and its effect on each of these
microscopic processes. In Section 10.4, we will relate the basic physics of QWs to
the performance of the photovoltaic device.

10.3.1 The quantum well

A QW is formed when a layer of a narrow-gap semiconductor a few nanometres
thick is grown epitaxially between layers of the wider-gap semiconductor. It is
important that the two materials be electronically similar (i.e. that the atomic part
of the crystal wave function is similar, and that the two materials have the same
crystal symmetry), that the interface be clean and that the same crystal planes form
the interface in either material. Then the different potential of the QW may be
treated as a perturbation to the crystal potential and delocalised carrier states can
extend across the two materials (Bastard, 1988).
If the conduction band edge is lower in energy and the valence band higher
in energy in the well material than in the barrier, then electrons and holes are both
confined in the well material. This is known as a Type I QW (Fig. 10.3). If only
one carrier type is confined in the well, the QW is Type II. Many QWs together
form an MQW, and if the barriers are thin enough for neighbouring wells to be
electronically coupled the structure is known as a superlattice (SL). Only Type I
QWs have so far been studied for solar cells, although SLs have been proposed as a
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 460

460 J. Nelson & N. Ekins-Daukes

energy
conduction band

Eb
Eg

valence band
well material
barrier material

distance

Figure 10.3 Band profile of a Type I quantum well.

means of improving carrier transport in high-resistivity InP solar cells (Varonides

and Berger, 1997).

10.3.2 Density of states

The QW forms a quasi-two-dimensional system. Confinement of electrons and
holes in the growth (say, z) direction leads to quantisation of the z component of
their momentum and kinetic energy. The quantised energy E n of the nth level is
related to the z component of the wave vector kn through
h̄ 2 kn2
En =
2m ∗
where m ∗ is the effective mass of the carrier in the growth direction. The carriers
are confined to a set of sub-bands of minimum energy Un , but are free to move in
the (xy) plane of the well where the symmetry of the crystal is maintained. Hence
a carrier in the nth sub-band has total energy
h̄ 2 k 2
E(k) = E n + (10.1)
2m ∗

where k is the total wave vector, k is the component in the xy plane (such that
k2 = k2 + kn2 ), and m ∗ is the effective mass of the carrier in this plane.
In the envelope function approximation, the shift V (z) in the conduction or
valence band edge due to the QW is considered as a perturbation to the periodic
crystal potential, and the wave functions as crystal eigenfunctions modulated by
an ‘envelope function’. The confined state energies E n and envelope functions
Fn (z) are solutions to an ‘effective mass’ equation, which resembles Schrödinger’s
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 461

Quantum Well Solar Cells 461

equation for a one-dimensional potential well. They are analogous to the energy
levels and wave functions of a one-dimensional quantum system. For a QW of
width L and depth V
h̄ 2 d 2 Fn (z)
− + E(z)Fn (z) = E n Fn (z)
2m ∗ dz 2
where
E(z) = 0, −L/2 ≤ z ≤ L/2
(10.2)
E(z) = V, z > |L/2|
This equation holds for both electrons in the conduction band and holes in the
valence band, but with different values of m ∗ and V . Energies E n are measured up
from the bottom of the QW in the conduction band for electrons, and down from
the top of the valence band for holes. The well depth V depends on the composition
of the barrier and well materials and on how the difference in bandgap is divided
between the valence and conduction bands. The effective mass m ∗ for each carrier
type is in general different for well and barrier.
In III–V semiconductors two different types of hole, heavy and light, need to
be considered. In the bulk crystal, heavy and light holes are carriers with different
effective mass associated with two degenerate crystal bands. For a QW in unstrained
material, heavy and light holes occupy the same potential well in the valence
band, but with different sets of confined-state energies on account of their different
effective masses. In a strained QW, the well depths for heavy and light holes can
be different. The number N of confined states contained in the QW for each carrier
type is given by

√ 
L 2m ∗ V
N = int +1 (10.3)
π h̄

where int(x) means the integer part of x. N increases with increasing well width and
depth, and carrier effective mass. The well is normally narrow enough to admit only
a few confined states. At energies E > V the carriers are no longer confined and a
continuum of states becomes available, as in the bulk material. These continuum
states will not be considered here.
In accordance with the uncertainty principle, the lowest energy level is always
shifted away from the bottom of the well, by an amount that increases with increas-
ing quantum confinement. This means that the ground-state energy, and hence the
absorption edge, can be controlled simply by varying the well width.
The corresponding envelope functions have well-defined parity and penetrate
further into the barrier as energy is increased. Energy levels and envelope functions
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 462

462 J. Nelson & N. Ekins-Daukes

0.2 e3
e2

0 e1
-1 -0.5 0 0.5 1

-0.2

-0.4
Energy/eV

-0.6

-0.8

-1

-1.2

-1.4
hh2 hh1
hh3
hh4
-1.6
Well widths

Figure 10.4 Calculated energy levels and envelope functions for a 100 Å GaAs QW in
Al0.3Ga0.7As. The relative energies of confined states and bandgaps are to scale, and the
bottom of the conduction band is taken as the zero of energy. Quantum number is measured
up from the bottom of the well for electrons, and down from the top of the well for holes.

for a typical Alx Ga1−x As /GaAs QW are shown in Fig. 10.4. In the SL configura-
tion, neighbouring wells are coupled, extended-state envelope functions span the
entire SL, and the previously discrete energy levels of the QW broaden into bands.
These effects improve carrier transport in the growth direction.
The density of states function can be constructed from the energy spectrum
in the usual way. For a QW of width L the density of states per unit volume V is
given by

2  m∗ 
N
D(E) = δ[E − E(k)] = (E − E n ) (10.4)
V
k
π h̄ 2 L n=1

where δ is the Dirac delta function and  is the Heaviside function. As illustrated in
Fig. 10.5, D(E) has the staircase structure characteristic of quasi-two-dimensional
systems.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 463

Quantum Well Solar Cells 463

D(E)

finite well bulk

infinite well

E1 E2 E3 E

Figure 10.5 Schematic density of states function D(E) for a finite QW, compared with
that for an infinitely deep QW and for the well material in the bulk. The first three confined
state energies, E 1 , E 2 and E 3, are shown.

This allows us to calculate the concentrations n (of electrons) and p (of holes)
in the QW, assuming a local quasi-thermal equilibrium. For electrons with density
of states function Dcb (E) in the conduction band
 ∞
n= Dcb (E(k)) f F D (E(k), T, 
µe )d E (10.5)
E cb

where f FD (E) is the Fermi–Dirac distribution function, E c is the conduction band-

edge energy, µ̃e the quasi-Fermi level for electrons in the QW, and T the effective
electron temperature. When E cb  µ̃e , n is given by

n = n iq exp[(
µe − E i + θn )/kT ] (10.6)

where n iq is the intrinsic carrier density of the QW material in the bulk, E i

is the intrinsic potential energy — the level at which the Fermi level would
lie in a perfectly intrinsic material — and θn is a measure of the shift in n
due to quantum confinement. This is analogous to the expression for n, namely
µe − E i )/kT ], in a non-degenerate bulk semiconductor.
n i exp[(
For the remainder of the discussion, we will assume that the QW is described
by a quasi-two-dimensional density of states and by a local quasi-Fermi level that
is not necessarily continuous with that in the barrier material.
In operating conditions, QWs placed in the space-charge region of a p-n
junction will be subject to a (small) electric field. The field tilts the QW band
profile, as shown in Fig. 10.6, distorts the confined-state functions and shifts their
energies.The energy of the lowest confined state is reduced. Strictly speaking, in
the presence of the field these wave functions are no longer confined — carriers
penetrate further into the barrier on the side of decreasing potential energy and
can tunnel out. For solar cells, the electric field is small enough for the ‘flat-band’
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 464

464 J. Nelson & N. Ekins-Daukes

electric field

q L

L
Figure 10.6 Band profile for a QW subject to an electric field in the growth direction. As
the field is increased the right-hand barrier is reduced, increasing the probability of electron
escape by thermionic emission or tunnelling.

approximation to the band structure to be adequate. However, it is relevant that

tunnelling through the barrier is possible.

10.3.3 Photogeneration
In a solar cell, photon absorption across the bandgap is important. Fermi’s golden
rule gives the absorption coefficient α in terms of the confined-state energies and
overlap integrals. For transitions between an initial valence-band state |i , of energy
E i and a final, conduction-band state | f , of energy E f , under the influence of an
electromagnetic field of angular frequency ω and polarisation e we have (Bastard,
1988)
A
α(E) = | f |e.p|i |2 δ[E f − E i − E]( f F D (E i ) − f F D (E f )) (10.7)
ω
i, f

where E is the photon energy h̄ω, p is the quantum mechanical momentum operator
and A is a sample-dependent optical constant. In the usual case where the light is
incident normal to the plane of the QW, the matrix element is proportional to the
overlap integral Mlm between the initial valence sub-band, l, and final conduction
band, m, envelope functions

Mlm = Fel (z)Fhm (z)dz (10.8)

This means that optical transitions are allowed only between sub-bands of the
same parity (l and m both even or both odd), and are strong only when l = m.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 465

Quantum Well Solar Cells 465

In addition, coulombic bound states (excitons) are formed at an energy just below
the minimum for each optically allowed sub-band-to-sub-band transition. The exci-
tons appear as strong peaks in the spectrum, even at room temperature, because of
their higher binding energy in two-dimensional systems. Including only the prin-
cipal (1s) exciton and summing Eq. (10.7) over initial and final state energies for
the lth electron – mth hole sub-band pair, we have

αlm (E) = αlh/ hh Mlm [ flm δ(E − Elm − Blm ) + (E − E lm )] (10.9)

where Elm is the electron-hole transition energy before Coulombic effects are
included, Blm and f lm are the exciton binding energy and oscillator strength, respec-
tively, and the constants αlh/ hh represent the absorption coefficient on the first step
edge. In III–V semiconductors, optical transitions occur between both electron–
heavy hole (hh) and electron–light hole (lh) states. The total absorption is the sum
of contributions from all such transitions.
 
α(E) = αel hh m (E) + αel lh m (E) (10.10)
l,m l,m

where each electron–hole sub-band pair contributes a step function and a set of
excitons to the total absorption spectrum. The absorption coefficient for a typical
Alx Ga1−x As /GaAs QW is shown in Fig. 10.7.
The QW absorption spectrum thus reflects the step-like form of the density
of states, modified by strong excitonic peaks. (Because of the strong exciton, the
QW spectrum may have a steeper absorption edge than the equivalent bulk alloy,

6
GaAs
5
–1

Al0.33Ga 0.67As/GaAs MQW

Absorption coefficient/ m

Al0.33Ga 0.67As
4

0
1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Photon energy/eV

Figure 10.7 Calculated absorption coefficient for a 100 Å Al0.33 Ga0.67 As/GaAs QW com-
pared with the absorption of bulk GaAs and bulk Al0.33 Ga0.67 As. (For the QW, the absorp-
tion coefficient is per unit thickness of well material, not including barrier thickness.)
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 466

466 J. Nelson & N. Ekins-Daukes

which could be useful for certain PV applications.) The absorption edge or effective
bandgap E a is blue-shifted from the absorption edge E g of the well material in the
bulk by the joint confinement energies E 11h of the lowest electron and heavy hole
sub-bands less the corresponding exciton binding energy B11h .

E a = E g + E 11h − B11h (10.11)

The effective bandgap E a is most strongly influenced by QW width and varies

from the bandgap E g of the well material for very wide wells, to the bandgap E b
of the barrier — or host — material for very narrow wells. This tunability of the
absorption edge is one of the most important features of the QWSC.
At photon energies above E b , photogenerated carriers are no longer confined
in the QW and the simple quantum mechanical model of absorption becomes
unhelpful. In this range the absorption spectrum of the QW begins to resemble that
of the bulk material.

10.3.4 Transport and recombination

As in any semiconductor device, the electron and hole, once excited, may be trans-
ported away from the point of creation, or recombine with each other or with trap
states in the bandgap. In the steady state, these processes are described by the
continuity equations for electrons and holes:
1 di e
+g −r = 0
q dz
and
1 di h
− +g −r = 0
q dz
where r is the volume recombination rate, g the volume generation rate, i e the elec-
tron current density and i h the hole current density. The materials parameters that
are normally used to quantify these processes in a bulk crystalline semiconductor
device — the recombination lifetimes and diffusion constants — are properties
of the bulk material and only have meaning in a material many times the thick-
ness of a QW. Level quantisation affects not only the generation term through the
QW absorption, discussed above, it also affects the rate of recombination and the
mechanism of transport in the direction of the built-in field. We shall discuss these
effects next.

10.3.5 Recombination
The processes that govern recombination in bulk materials apply to QWs. For
III–Vs the most important, in practice, is non-radiative recombination through
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 467

Quantum Well Solar Cells 467

traps. For a single trap state in the bandgap, the Shockley–Read–Hall recombination
rate is given by

np − n 2i
rnr = (10.12)
τe ( p + pt ) + τh (n + n t )
where pt , n t are the equilibrium populations of trap states occupied by holes and
electrons, and τ p , τn are the respective carrier trapping times. This formulation
should be appropriate to a QW provided that n and p are defined using the quasi-
Fermi level of the carriers in the QW (Eq. (10.6)). The lifetime parameters are
properties of the material and so as a first approximation to a QW we may take
the same values as for the well material in the bulk. However, the accumulation of
defects at the QW interface may affect the location and density of trap states, and
quantum confinement may reduce the trapping times.
In the limit of ideal material, radiative recombination is the process that deter-
mines solar cell efficiency. The excess radiative recombination in the biased device
(i.e. in addition to the recombination that balances thermal generation in equilib-
rium) then constitutes the dark current. In any volume element δV the radiative
recombination rate rrad depends on the local absorption spectrum α(E) and the
local quasi-Fermi level separation µ̃F , according to

rrad δV = α(E) j (E, T,  µ F )d EδV (10.13)

where the emitted flux density j is given by the generalised Planck equation (Wür-
fel, 1982; Tiedje et al., 1984)

2nr2 E2
j (E, T, µ̃ F ) = (E− µ̃
(10.14)
3 2
h c e F )/ kT − 1

and n r is the local refractive index, h is Planck’s constant and c the speed of light.

10.3.6 Transport in the growth direction

In a homojunction solar cell, electron and hole currents are normally described by
the drift–diffusion equations. The electron current is given by (Sze, 1981; Hovel,
1975)
dn
i e = i diff + i drift = q De + qu e nE (10.15)
dz
where De is the electron diffusion coefficient, E the electric field strength and
u n the electron mobility. In the region of a QW, carriers with energy above the
barrier band edge will be scattered into and out of the QW by much the same
processes as in the bulk. One approximation to transport through the QW is to
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 468

468 J. Nelson & N. Ekins-Daukes

apply the drift–diffusion equations to the composite structure and represent the
QW by a region of increased carrier population. In this approach it is assumed that
the QW and host are in quasi-thermal equilibrium and that the quasi-Fermi level
is continuous across the interfaces of the QW.
However, this treatment neglects the fact that carriers with energies below the
barrier band edge are essentially trapped in the QW. These carriers may escape
from the QW by quantum mechanical tunnelling or thermally assisted tunnelling
when the QW is subject to an electric field in the growth direction, and other car-
riers may be captured into the QW by scattering from higher energy states. Both
of these processes are quantum mechanical and cannot effectively be described
with the drift–diffusion equations. An alternative approach is to consider two pop-
ulations for each carrier: a mobile population with energy greater than the barrier
band edge, and a confined population in the QW. The confined population does not
actually contribute to the current, but it affects the current by influencing the rate
at which carriers are added to the mobile population through escape, or removed
from it through capture. Equation (10.15) for i e should therefore include terms for
escape and capture:

i e = i diff + i drift + i esc − i capture (10.16)

The escape current is important whenever carriers are generated in the QW

and an electric field is present perpendicular to the plane of the QW. It can be
calculated semiclassically from

h̄k z
i esc = Dcb (E(k)) f F D (E(k), T, 
µe )T r (k z ) ∗ d 3 k (10.17)
me

where Tr(k z ) is the probability of transmission through the barrier at wave vector
k z . The integral should be carried out with respect to wave vector rather than energy,
because of the anisotropy of the density of states. For shallow QWs in short-circuit
and low-forward-bias conditions, the escape current appears to be equivalent to the
sheet-generation rate in the QW (Nelson et al., 1993). In the operating conditions
of a QW, when barrier states are also populated, Eq. (10.17) should be modified to
allow for a finite probability of population of the final states, and a term added for
capture of carriers into the QW from states above the barrier. This is a notoriously
difficult problem, involving both localised and continuum states, and has not at the
time of writing been resolved for the QWSC.
Once given the expressions for carrier-pair generation, recombination and
current, carrier continuity may be invoked to complete the set of equations for the
QW device. Discretising the continuity equations for electrons in a QW of width L
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 469

Quantum Well Solar Cells 469

we have, in the steady state

i e
= q(g − r ) (10.18)
L
where i e represents the change in electron current across the QW, and g and r the
volume rates of pair generation and recombination in the QW. At each interface,
i e should be continuous. Thus a set of current and continuity equations can, in
principle, be constructed to describe the behaviour of the QWSC, much as for
conventional solar cells. To do this, the correct descriptions of escape and capture
must be known so that the carrier current can be related to the quasi-Fermi levels
in the QW.

10.4 Performance characteristics of QWSCs

10.4.1 Quantum efficiency
Quantum wells enhance the photocurrent by extending absorption. The relevant
quantity for PV is the spectral response (SR) or zero-bias external quantum effi-
ciency of the cell, which is the probability that, for each incident photon of given
wavelength, an electron will be collected at the contacts. Figure 10.8 shows how
QWs extend the SR to longer wavelengths for an AlGaAs p-i -n cell with and with-
out GaAs QWs. The SR at photon energies above the baseline cell bandgap E b is
barely affected by the QWs. This shows that QWs do not degrade the collection of
minority carriers. Indeed, the SR is improved slightly at energies close to E b on
account of the higher absorption of the QW material at these wavelengths.
The SR at photon energies below E b is entirely due to absorption in the QWs
and reflects features of the QW absorption spectrum. This behaviour is straightfor-
ward to model. For QWs placed in the depletion region of the cell, we can ignore
diffusion and make the usual assumption that all minority carriers reaching the
edges of the depletion region by drift are collected. Then for N QWs of width L,
the SR is related to absorption through

SR(E) = (1 − r )ηesc (1 − exp[−NαQW (E)L]) E < E b (10.19)

where r is the front-surface reflectivity and ηesc the probability of carrier escape
from the QW. The fraction (1 − ηesc ) of the carriers that do not escape can be
assumed to recombine in the QW. In fact, for QWs in any of the systems mentioned
at room temperature, ηesc is unity. This has been confirmed by experimental studies
of the dependence on the field and temperature of the photocurrent from single
QW devices (Nelson et al., 1993; Barnes, 1994; Zachariou et al., 1998), which
show that carriers escape from the QWs at room temperature by thermally assisted
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 470

470 J. Nelson & N. Ekins-Daukes

0.50

p-i-n control cell

0.40 50 QW cell
Spectral response

0.30

0.20

0.10

0.00
400 500 600 700 800 900
Wavelength/nm

Figure 10.8 Measured spectral response (external quantum efficiency) for an Al0.33 Ga0.67
As p-i-n test device with and without 50 87 Å QWs in the i-region. These devices were not
AR coated.

tunnelling. Equation 10.19 reproduces the short-circuit SR fairly reliably for a

well-characterised QW (Paxman et al., 1993; Barnes et al., 1996; Connolly et al.,
1995). It is implicit in Eq. 10.19 that, once a carrier has escaped from the QW, it
will be collected.
The incremental photocurrent due to the QWs in an incident spectrum of
photon flux jinc(E) is thus

i ph = SR(E) jinc(E) d E (10.20)

where SR(E) is the incremental SR due to the QWs, which is given by Eq. (10.19)
for E < E b and the difference between the SR of well and host material for E > E b .
To maximise the effect of the QWs on photocurrent we need an SR which is
as high as possible for E > E a . In principle this can be achieved by increasing the
number of QWs and reducing the optical depth of the top layer (the p-region in a
p-i -n structure) by reducing its thickness or compositional grading, as shown in
Fig. 10.9.
In practice it is not always straightforward to increase the number of QWs.
A wide i -region requires a relatively low level of charged impurities (<1015 cm−3 )
in order to maintain the electric field (Barnham et al., 1991; Zachariou et al., 1996).
Figure 10.10 shows how the QW SR is affected by high background doping, in
this case arising from diffusion of the zinc dopant into the i -region.
For GaAs/In x Ga1−x As QWs, strain limits the spacing of the QWs. The GaAs
barrier layers need to exceed a critical thickness several times the Inx Ga1−x As
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 471

Quantum Well Solar Cells 471

1.00

0.80
Spectral response

test p-i-n cell

0.60
test QW cell
less p, more i
0.40 AR coat
back reflector
data
0.20

0.00
400 500 600 700 800 900
Wavelength/nm

Figure 10.9 Calculated spectral response SR for an Al0.33 Ga0.67 As p-i-n device and a
series of Al0.33 Ga0.67 As/GaAs QW devices, showing the effect of: a) reducing the p-layer
thickness and increasing the i-layer thickness; b) adding an AR coat; c) adding a front-
surface window and back-surface reflector. The measured SR for a cell of design (b) is
compared with the calculation.

1.00
Model: High background doping
Model: Low background doping
0.80 Experimental data
Spectral response

0.60

0.40

0.20

0.00
400 600 800 1000 1200 1400 1600
Wavelength/nm

Figure 10.10 Measured and calculated SR for an InP p-i-n device containing 20
In0.53 Ga0.47 As QWs. The calculated curves show that an unintentional background doping
in the i−layer of around 2 × 1016 cm−3 is required to explain the low SR. Diffusion of zinc
dopants increases the space charge in the i-region so that at zero bias the electric field falls
to zero near the centre of the i-region and only around half of the carriers generated in the
QWs are collected.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 472

472 J. Nelson & N. Ekins-Daukes

Figure 10.11 Measured SR for a series of GaAs/Inx Ga1−x As devices, each with 10 QW,
processed in different ways. The low SR with no light trapping is substantially increased by
adding a plane or a rough gold mirror at the back surface. The rough mirror, a gold epoxy
resin, achieves light trapping effectively through large-angle internal scattering.

QW width, in order to accommodate the lattice mismatch without dislocations

(Barnes et al., 1996; Griffin et al., 1996). In this system, light trapping is essential
to increase the SR to a useful level. Figure 10.11 illustrates the effect of light
trapping on the SR.
An alternative strategy is to use In x Ga1−x As of low indium content rather than
pure GaAs in the n- and p-regions, so partly compensating the strain introduced
by the QWs (Mazzer, 1997). This reduces the critical spacing, and allows more
QWs to be inserted into an i -region of given width.

10.4.2 Dark current density

The dark current i Dk (V ) of a solar cell at forward bias V is the sum of contributions
from radiative and non-radiative processes:

i Dk (V ) = i nr (V ) + i rad (V ) (10.21)

Since the generation rate g = 0 in the dark, i Dk is found by integrating all the
contributions to the recombination rate r over the volume of the cell. Radiative
recombination within the interior of the device is partly cancelled by absorption,
and the radiative current density i rad is given by q times the net photon flux escaping
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 473

Quantum Well Solar Cells 473

through the device surface. Thus i rad is obtained by integrating Eq. (10.13) over
surface elements S and solid angle

 ∞ a
i rad = q (E, θ, s) j (E, T, 
µ F )d
.dS.dE
0 S

where a(E, θ , s) is the probability that a photon of energy E will be emitted

from the point s on the surface at an angle θ . This equation should apply to the
radiative currents from a QW device provided that the appropriate values of α(E)
and  µF are used. In an ideal solar cell, radiative processes are dominant and  µF
is assumed to be constant and equal to the cell voltage V .
In real solar cells at typical operating biases, non-radiative processes are dom-
inant, and the non-radiative current density i nr (V ) is found by integrating the
non-radiative recombination rate rnr (z) over the volume of the cell:

i nr (V ) = rnr (V, z)dz (10.22)

In III–V devices, rnr is well modelled by the Shockley–Read–Hall expression

(Eq. (10.12)), and will be greatest in the depleted space-charge layer (SCL) of the
junction, where n and p are similar. Calculating i nr in the depletion approximation
(also known as the exhaustion layer approximation), we find for a p-n or p-i -n
homojunction cell
qn i W sinh(q V /2kT ) π
i nr (V ) = (10.23)
τnr (Vbi − V )q/kT 2
where W is the width of the SCL, τnr the ambipolar carrier lifetime and Vbi the
built-in bias of the junction (Sah et al., 1957).
Since QWs are placed in the SCL and have a much higher intrinsic carrier
density than the host material, we may expect them to dominate the dark current.
We can approximate the incremental recombination current i nr by treating the
QWSC in the depletion approximation in the same way as the p-i -n cell above.
(The depletion approximation appears to be valid so long as the quasi-Fermi levels
energy remain below the QW band-edges.) Then, for a single QW cell, we find
(Nelson et al., 1995)
qn iq L sinh(q V /2kT )
i nr = (10.24)
τnrq cosh(qϕq /kT )
where n iq is the intrinsic carrier density in the QW, τnrq is the mean non-radiative
recombination time in the QW and φq the value of the intrinsic potential at the
QW relative to the mid-point of the quasi-Fermi potentials for electrons and holes,
which depends on the location of the QW relative to the point where n = p and
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 474

474 J. Nelson & N. Ekins-Daukes

vanishes for a QW located in the centre of a symmetrically doped junction. In this

case, Eq. (10.24) reaches its maximum value and i nr takes the usual ideality factor
of 2 for recombination–generation currents. Comparing Eqs (10.23) and (10.24)
we could expect the dark current to increase by a factor of (n iq L/n i W ) for a single
QW added to the centre of the i -region. Typically, this factor may be one or two
orders of magnitude, which could reduce the open-circuit voltage by 10–20%.
As Fig. 10.12 shows, dark currents for QW and p-i -n solar cells in a range of
materials show that i Dk values do indeed increase when QWs are added. However,
when compared to theoretical calculations, the measured i Dk is smaller than pre-
dicted by around one order of magnitude, and has a different bias dependence or
ideality factor (Nelson et al., 1994, 1995). This means that the open-circuit voltage
is slightly larger than expected. Figure 10.13 shows dark currents calculated by
solving the carrier transport equations self-consistently to determine the n and p
profiles, and then integrating the SRH recombination rate across the device. A local
quasi-thermal equilibrium is assumed, so that electron and hole quasi-Fermi levels
are continuous across the QW interfaces.
Several possible explanations for the discrepancy are suggested by Eq. (10.24).
One is that the recombination lifetime for carriers in the QW is longer than for the
well material in the bulk. This is rather unlikely since material quality should, if
anything, be worse in the QW configuration on account of the interfaces. Another
is that background doping could shift the point where non-radiative recombination

10000

1000

100
Dark current / A m -2

85A MQW
10 50A SQW
AlGaAs p-i-n
1
140A SQW
0.4 0.6 0.8 1 1.2 1.4
0.1 GaAs p-i-n
GaAs / InGaAs
0.01

0.001

0.0001
Bias / V

Figure 10.12 Measured dark current densities for a series of QW and homogenous p-i-n
devices. The curves show that adding GaAs QWs to an Al x Ga1−x As p-i-n cell increases
the dark current, as does adding Inx Ga1−x As QWs to a GaAs device. The effect on the dark
current is larger for wider QWs, and for a greater number of QWs.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 475

Quantum Well Solar Cells 475

is maximal away from the centre of the i -region, and so reduce the recombination
rate at a centred QW. This would also explain the change in ideality factor, as
shown in Nelson et al. (1995, 1999). But it would not explain why the dark current
is lower than expected in multi-quantum well devices, where we would expect the
regular MQW array to sample all n/ p ratios equally.
A possible explanation is that n and p are smaller than expected in the QW,
i.e. the quasi-Fermi levels are closer to the centre of the bandgap than in the
neighbouring host material. This is supported by complementary measurements
of the radiative recombination current in single QW devices (Nelson et al., 1997,
1998), which show that the emission spectrum from a single QW is some tens of
meV smaller (and the radiative current a factor of 2–4 smaller) than expected if the
quasi-Fermi level separation were constant across the well/barrier interface.
This explanation suggests that the QWs are not in quasi-thermal equilibrium
with the barrier material. Quasi-equilibrium is appropriate when current is dom-
inated by low-field carrier drift and diffusion. These conditions are usually satis-
fied for a homojunction cell, where variations in carrier population are smooth.
In QWSCs in the region of the QW interfaces, as discussed above, current may
be dominated by carrier escape from the QWs through thermionic emission and
thermally assisted tunnelling. These currents may greatly exceed drift–diffusion
currents in the direction of increasing kinetic energy, and therefore the notion of
quasi-thermal equilibrium may not be valid here.
This explanation was proposed by Corkish and Honsberg (1997). They draw
on studies of bulk heterojunctions which show that a step in the minority-carrier
Fermi level may result at a heteroface in conditions where transport is limited by
thermionic emission or transport across a space-charge region. They show that a
moderately high level of background doping in the QW could give rise to reduced
quasi-Fermi level separation and lower dark currents. This is a promising idea and
more detailed modelling incorporating carrier escape and capture may well provide
a quantitative explanation of the observed dark currents.
Optimisation of the QWSC clearly requires minimisation of the incremental
recombination current. One way to do this would be to choose a material for the
QW with long non-radiative recombination times. In GaAs, for example, minority
carrier lifetimes are long compared to those in Alx Ga1−x As, and it is possible
that the recombination current from an Alx Ga1−x As /GaAs MQW may be lower
than that from the Alx Ga1−x As alloy of equivalent effective bandgap. Also the
dependence of recombination rate on QW position means that it should be possible
to optimise the design of QWSCs by locating QWs away from the centre of the
SCR, in the regions where n  p or p  n. (Nelson et al., 1999). This effect results
from the asymmetry in carrier populations and the form of the Shockley–Read–Hall
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 476

476 J. Nelson & N. Ekins-Daukes

recombination rate; it does not require that the quasi-Fermi level in the QW be
reduced.

10.4.3 Current–voltage characteristic

As explained in Section 10.3, QWs are expected to increase both the photocurrent
and the dark current of a single-bandgap solar cell. Making the assumption that
the dark current is not changed by illumination (the superposition assumption), it
is usual to write the current–voltage characteristic as

i (V ) = i sc − i Dk (V ) (10.25)

where i sc is the short-circuit photocurrent density. The efficiency is given by the

ratio of the maximum of the current–voltage product, i (Vmp ) × Vmp , to the incident
light power. For a QWSC:

i (V ) = i sc + i sc − i Dk (V ) − i Dk (V ) (10.26)

Therefore, QWs can benefit the power conversion efficiency only when the
increase in photocurrent i sc exceeds the increase in recombination current i Dk

100000
Model: low background doping
Model: high background doping
Experimental data
1000
Dark current / A m-2

10

0.1

0.001
0.4 0.8 1.2 1.6
Bias / V

Figure 10.13 Measured and modelled dark currents for an Al0.34 Ga0.66 As/GaAs single
QW device. The black line shows the dark current expected when the background doping
level Ni in the i region is small (∼1014 cm−3 ) and the QW is located close to the plane where
electron and hole densities are equal. For high background doping (Ni ≈ 2 × 1016 cm−3 ),
p  n at the QW and non-radiative recombination is reduced. The lower ideality factor
reflects the asymmetry in the electron and hole densities.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 477

Quantum Well Solar Cells 477

at the operating voltage. For small changes this means

i sc − i Dk (V ) > 0 (10.27)

The advantage increases with the number of QWs since, while i sc increases
approximately linearly with the number of QWs, the decrease in open-circuit volt-
age due to i Dk changes only logarithmically.
Figures 10.14 and 10.15 show I –V characteristics for an Al0.3 Ga0.7 As p-i -n
cell with and without 30 GaAs QWs and for a GaAs p-i -n cell with and without 10

60
AlGaAs p-i-n cell
50
AlGaAs/ GaAs QW cell
Current density/A m–2

40

30

20

10

-10

-20
0 0.2 0.4 0.6 0.8 1 1.2
Bias/V

Figure 10.14 Measured current–voltage characteristics for an Al0.3 Ga0.7 As p-i-n device
with and without 30 GaAs QWs, in a white light source approximating to a 3200 K black-
body spectrum. Note that these devices were not AR-coated, hence the low current densities.

600

500
–2
Current density/A m

400

300

200
GaAs p-i-n cell
100 GaAs/InGaAs QW cell

-100
0 0.2 0.4 0.6 0.8 1 1.2
Bias/V

Figure 10.15 Measured current–voltage characteristics for a GaAs p-i-n cell with and
without 10 InGaAs QWs, using the same light source as for Fig. 10.14.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 478

478 J. Nelson & N. Ekins-Daukes

In0.16 Ga0.84 As QWs. In both cases, introducing the QWs has increased i sc and
reduced Voc . The latter effect results from the increased dark current, which is
evident from Eq. (10.25).
In the case of the Al0.3 Ga0.7 As host cell, where the host bandgap exceeds
the optimum for solar energy conversion, the net effect of QWs is to increase
the cell efficiency. This is as expected, since QWs added to a wide-gap host cell
reduce its effective bandgap towards the optimum. In the case of the GaAs host,
the efficiency decreases, which is again the result expected simply from arguments
about the optimum bandgap for photoconversion: the addition of a lower-bandgap
material to GaAs will reduce the effective bandgap for absorption, and from detailed
balance arguments this is expected to reduce the efficiency of the solar cell.
However, the results for the GaAs and InP host materials are complicated by
material quality issues. In GaAs the number of QWs that may be added before
strain degrades the device quality is too few to produce an adequate photocurrent
enhancement. In InP, problems of high background doping have made it impossi-
ble to prepare good quality devices for comparison. However, In x Ga1−x As QWs
have been observed to increase the efficiency of a less than perfect InP p-i -n cell
(Zachariou et al., 1996). Strain-balanced GaAsP/InGaAs represents the cleanest
material system in which QWSCs have been grown showing a linear increase in
photocurrent with approximately constant voltage (Bushnell, 2005). As discussed
further in Section 10.5, this material becomes dominated by radiative recombina-
tion at high solar concentration and for a short period set a world-record power-
conversion efficiency, discussed further in Section 10.7.2. However, the record was
subsequently reclaimed by a GaAs homojunction (Green, 2011), so it is therefore
not yet possible to decide on the effect, in practice, of QWs on solar cell efficiency.
However, we can learn something from the effect of QWs on the open-circuit
voltage Voc of test devices. Detailed balance arguments (below) imply that Voc
should be controlled by the absorption edge E a . Therefore we would expect a
decrease E a in the absorption edge to cause a decrease in Voc of the same magni-
tude, and we have seen above that the decrease in E a caused by the introduction of
QWs is accompanied by a reduction in Voc . However, measurements (Fig. 10.16)
show that Voc is less sensitive than expected to the effective bandgap E a of the well
material (Barnham et al., 1996). This is reasonable since it is the host material that
controls carrier injection currents, and hence the population of carriers available
for recombination.

10.5 Developments in QWSC design and performance

In most applications, the aim of adding quantum wells is to increase the
photocurrent of the host homostructure cell. In most cases it is not possible to grow
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 479

Quantum Well Solar Cells 479

Figure 10.16 Open-circuit voltage against effective bandgap for a series of Al x Ga1−x As/
GaAs QWSCs and Al x Ga1−x As p-i-n cells of different Al fraction. Note how, for the QW
devices, Voc is higher than expected either from the measured dependence of Voc on Al
fraction for Alx Ga1−x As devices (dashed line) or from the theoretical dependence of Voc
on the absorption edge of the QW, the effective bandgap, (dotted line) expected from detailed
balance arguments.

an optically thick QW stack, since background impurity levels limit the width of
depleted i-region that can be grown. As a result, rear surface reflectors, such as
a distributed Bragg reflector (DBR), have been employed to increase the optical
path length in the material (Bushnell et al., 2003). These have resulted in quantum
efficiency levels in excess of 50% in the QW region.
Apart from increasing photogeneration, the effect of introducing a high-quality
reflector into the solar cell enables re-absorption of radiative recombination to take
place, an effect known as photon recycling (Marti et al., 1997). In a high-quality
material such as strain-balanced GaAsP/InGaAs, recombination is dominated by
radiative processes at high current levels (equivalent to 200× solar concentration),
and the introduction of reflecting stacks in devices made from such materials was
observed to reduce the electrical dark current of the solar cell (Johnson et al., 2007).
The observed reduction in recombination rate corresponds to a 0.3% increase in
absolute efficiency, yet in principle a gain of 2% in absolute efficiency is possible
through photon recycling.
In addition to trapping light inside the QWSC, it is also possible to manipulate
the emission profile for efficiency gain. In a bulk semiconductor and unstrained
quantum well, radiative recombination is roughly isotropic, but under compres-
sive biaxial strain, the radiative recombination becomes directional, favouring the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 480

480 J. Nelson & N. Ekins-Daukes

transverse electric (TE) polarised direction propagating normal to the QW plane

(Adams, 1986). Since a solar cell need only emit light into the optical modes
through which it receives light (Markvart, 2008), manipulating the radiation pattern
of radiative recombination can lead to substantial efficiency gains of 9% absolute
(Hirst et al., 2011). However, in practice the incomplete suppression of trans-
verse propagating light in QW samples reduces the size of the effect, delivering
a modest efficiency enhancement of 0.12% absolute for a compressively strained
InGaAs/GaAsP QW solar cell (Adams et al., 2011).
In a multijunction solar cell there is less scope for photon recycling since
photons cannot easily be confined in a single junction, but downward-propagating
radiative emission can be absorbed by an underlying junction. This radiative cou-
pling of sub-cells can lead to a significant recovery of efficiency in a multijunction
solar cell with sub-optimal threshold energies (Marti et al., 1996). In most bulk
cells, radiative coupling efficiencies have been low, on account of re-absorption
losses in the base region of the cell, but in a QW device the base is transparent
to the QW emission, resulting in an efficient transfer of energy from one sub-cell
to the underlying device. Recently, photon coupling efficiencies in excess of 50%
have been measured between sub-cells (Lee et al., 2012).

10.6 Limits to efficiency

A generalised detailed balance theory due to Araujo and Marti (Araujo and Marti,
1994; Araujo et al., 1994) concludes that no solar cell can achieve a higher limiting
efficiency than a homojunction cell of optimum bandgap. This conclusion rests
on two important assumptions: that the quasi-Fermi level separation is constant
throughout the device and equal to the applied bias V , and that each absorbed
photon delivers exactly one electron to the external circuit. The first condition
means that, according to detailed balance, the probability of photon emission from
the device is equal to the probability of photon absorption. The second means that
both probabilities have value unity for photon energies greater than the bandgap.
Then for the homogenous cell of bandgap E b , the current–voltage characteristic is
given by
 ∞  ∞
i (V ) = F
jinc(E, Ts , 0)dE − Fe j (E, Tc , q V )dE (10.28)
Eg Eg

where jinc is the incident photon flux from the Sun, radiating at temperature Ts , j
is the radiative flux within the cell at temperature Tc with quasi-Fermi level sep-
aration qV (Eq. (10.14)), and F
and Fe are geometrical factors giving the solid
angles for solar photon absorption and emission. For a black-body Sun at 5800 K,
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 481

Quantum Well Solar Cells 481

the available power i (V )V has a maximum at a bandgap of around 1.3 eV. When
applied to the case of a QWSC of effective bandgap E a this approach gives for the
incremental photocurrent and dark current
 Eg
i ph = F
b(E, Ts )a(E) d E
Ea
 Eg
i Dk (V ) = Fe j (E, Tc , 
µF )a(E) d E (10.29)
Ea

where a(E) is the probability of photon absorption in the MQW, i.e. its spectral
response. Now for the optimum QWSC a(E) = 1 for all E > E a and, if the quasi-
Fermi level separation  µF in the QW is equal to qV, then Eq. (10.26) becomes
identical to Eq. (10.28), and the optimum QWSC will be identical to the optimum
single-bandgap homojunction cell.
There has been some debate about whether the detailed balance theory applies
to the QWSC in practice (Corkish and Honsberg, 1997; Anderson, 1995; Araujo
et al., 1994). Measurements of radiative recombination currents from biased single-
QW test cells suggest that  µF is smaller in the QW than in the surrounding host
material. Irreversible carrier escape from the QW under the small electric field
present at the operating point has been suggested as a reason for this (Nelson
et al., 1995; Corkish and Honsberg, 1997). Interestingly, the same effect persists
under illuminated conditions (Ekins-Daukes et al., 2003), lending weight to the
hypothesis of the breakdown of quasi-thermal equilibrium at quantum well length
scales. Microscopic non-equilibrium theories for the QWSC are beginning to reveal
the carrier dynamics present in these systems (Aeberhard, 2010).
Another interesting idea is the possibility of exploiting ‘hot’ carrier effects
(Ross and Nozik, 1982) in QWSCs. At high carrier densities the relaxation of
excited carriers to the band edge can be slowed down by quantum confinement in
a QW. The carrier populations then appear to have a higher effective temperature
than the lattice, and recombination is reduced. Retarded relaxation has already been
observed in QW photoelectrodes (Rosenwaks et al., 1993), with the potential to
reach efficiencies in excess of 50% (Le Bris and Guillemoles, 2010). Some attempts
have been made to design hot-carrier superlattice solar cells (Hanna et al., 1997).

10.7 Applications
Because QWSCs are as costly to produce as high-efficiency III–V homojunc-
tion cells, we may expect them to be interesting only in those applications
where III–Vs are preferred. At the present time that means space, concentrator
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 482

482 J. Nelson & N. Ekins-Daukes

and thermophotovoltaic systems. Finally, we mention certain applications where

QWSCs are particularly promising.

10.7.1 Tandem cells

The efficiency of a monolithic tandem cell is highly sensitive to the combination of
bandgaps and to the requirement of current matching between the wide and narrow
bandgap components. Although the bandgap of bulk ternary semiconductors such
as Alx Ga1−x As can be adjusted simply by varying the composition (e.g. the Al
fraction x), non-radiative recombination increases rapidly with increasing x and
degrades collection efficiency. Alx Ga1−x As/GaAs QWSCs offer the alternative
possibility of controlling the bandgap through the width of the GaAs QWs. Com-
pared with wide-gap bulk alloys such as Alx Ga1−x As and InGaP, QW structures in
Alx Ga1−x As/GaAs, InGaP/GaAs and other combinations offer the advantages of:
(i) bandgap tunability through the QW width and (ii) control of the current through
the number of QWs.
At the time when the p-i -n AlGaAs/GaAs QW configuration was successfully
demonstrated (Barnham et al., 1991), the AlGaAs/GaAs tandem solar cell marked
the highest efficiency attained for photovoltaic power conversion. Recognising that
recombination will occur primarily in the lower-bandgap GaAs QWs, where recom-
bination lifetimes are longer than in Al x Ga1−x As, it was considered that a QWSC
would have superior practical performance to the Al x Ga1−x As homojunction cell
of the same effective bandgap (Connolly, 1998).
Around the same time, the InGaP/GaAs multijunction solar cell was being
developed and constrained by the relatively high bandgap of GaAs. The use of
GaAs/InGaAs strained QWs provided a means for lowering the bandgap and there-
fore increasing the current from the sub-cell but with a loss in voltage (Freundlich
et al., 1998). The advent of strain-balanced GaAsP/InGaAs enabled much of the
sub-cell voltage to be preserved (Ekins-Daukes et al., 1999).
Later, when triple-junction tandem cells started to outperform double-junction
tandems, QW structures became of interest for the high-gap cell in structures such
as the (high-gap semiconductor)/GaAs/Ge device.

10.7.2 Concentrator cells

In a homojunction cell, efficiency decreases at high light concentrations when the
increased temperature causes the bandgap to shrink and the open-circuit voltage,
which is directly related to the bandgap, to fall. In a QWSC, although the bandgaps
of the well and host material still reduce with increasing temperature, the effect on
Voc is less marked. Figure 10.17 compares the efficiency and temperature depen-
dence of Voc for a pair of QWSCs and homojunction cells. Although the mechanism
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 483

Quantum Well Solar Cells 483

8
7
6
Efficiency/%

5
4
3
2
1
0
0 10 20 30 40 50 60 70 80 90 100 110
Temperature/C

Figure 10.17 Temperature dependence of the efficiency of an InP/Inx Ga1−x As QW

cell (full line) in comparison with a homogenous InP p-i-n device (squares) and an
InP/Inx Ga1−x As heterostructure device with an Inx Ga1−x As i-region (triangles). The mea-
surements were made in a 3000 K black-body spectrum and scaled by correcting the pho-
tocurrent to the standard terrestrial AM 1.5 spectrum using the measured spectral response.

is not fully understood, clearly the efficiency of carrier escape from the QWs will
increase, or remain at unity, as T is increased. Faster carrier escape is likely to
reduce the probability of recombination in the QWs, and so offset the effect of the
decreasing QW bandgap.
Very high peak performance can be obtained from QWSCs under concen-
trated sunlight. Figure 10.18 shows the increase in efficiency of a GaAsP/InGaAs
quantum well solar cell measured under concentrated sunlight, retaining the tem-
perature stable at 25◦ C. A peak efficiency of 28.3% at 535 Suns was measured
under AM 1.5 D.

10.7.3 Thermophotovoltaics
In thermophotovoltaics, low-bandgap photovoltaic cells are used to produce elec-
tricity from the long-wavelength radiation emitted by a hot (2000–3000◦C) source.
The source is usually provided through fossil fuel combustion in a combined heat
and power system. Often a selective emitter is used to reabsorb the very low energy
photons and reemit them at higher energies to prevent heating. The reshaped spec-
trum is concentrated around certain bands characteristic of the emitter. For such
a spectrum, control of the bandgap of the PV cell is essential for good power
conversion efficiency. The flexibility of bandgap makes QWSCs of great interest
for TPV. It is also possible that Auger recombination, a longstanding problem in
low-bandgap solar cells, is suppressed in the QW device.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 484

484 J. Nelson & N. Ekins-Daukes

Figure 10.18 Single-junction 1.2 mm2 GaAsP/InGaAs quantum well solar cell efficiency
measured as a function of concentration. Data courtesy of Quantasol Ltd, measured at the
Fraunhofer Institute for Solar Energy Systems, May 2009.

Quantum well solar cells in InGaAsP/In x Ga1−x As have already been shown
to produce a higher Voc than the comparable Inx Ga1−x As homojunction cell (Grif-
fin et al., 1997; Connolly and Rohr, 2003). For low-temperature emitters, an
In x GaAs/In y GaAs strain-balanced combination grown on InP has demonstrated
absorption thresholds up to 1.95 µm (Rohr et al., 2006).

10.8 Conclusions
We have reviewed the use of novel QW semiconductor heterostructures in solar
cells. Quantum well structures are of interest as a means of enhancing the photocur-
rent and efficiency of crystalline solar cells. Photocurrent enhancement has been
demonstrated in a range of materials and is well understood. Efficiency enhance-
ment has been observed in materials whose bandgap is larger than the optimum for
solar energy conversion. In materials of bandgap close to the optimum, experimen-
tal tests on QW cells of equivalent quality to homojunction cells have not yet been
possible. Nevertheless there is some evidence that the effect of QWs in increasing
recombination within the device is smaller than expected from arguments based on
a quasi-thermal equilibrium distribution of carriers. If this is true under operating
conditions, then higher efficiencies may also be available with optimum-bandgap
cells.
Quantum well structures have the advantages over homojunction cells that the
effective bandgap can be controlled by tuning the width of the QW, rather than
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 485

Quantum Well Solar Cells 485

by varying the composition of a bulk alloy. This means that QWs may produce
better cells of better material quality than bulk alloys when particular bandgaps are
required. This is relevant for high-efficiency tandem cells and for thermophoto-
voltaic cells, and QW structures are being researched for both these applications.
A further important advantage is that QW structures have a better response to
temperature and consequently are expected to perform better under concentrated
light.
Some of the major challenges that remain are: to find and verify a theoretical
explanation for the observed dark currents and Voc behaviour; to establish whether
the suppressed recombination behaviour observed in the dark occurs under solar
cell operating conditions; and to prepare an optimum-bandgap QWSC of equivalent
quality and superior efficiency to a GaAs homojunction solar cell. More generally,
work on QW structures has stimulated a range of new ideas about the role of
quantum nanostructures in photovoltaics and the limits to efficiency of solar cells.

References
Aeberhard E. (2010), ‘Spectral properties of photogenerated carriers in quantum well solar
cells’, Solar Energy Mater. Solar Cells 94, 1897–1902.
Adams A. (1986), ‘Band-structure engineering for low-threshold high-efficiency semicon-
ductor lasers’, Electron Lett. 22, 249–250.
Adams J. G. J., Browne B. C., Ballard I. M., Connolly J. P., Chan N. L. A., Ioannides A.,
Elder W., Stavrinou P., Barnham K. W. J. and Ekins-Daukes N. J. (2011), ‘Recent
results for single-junction and tandem quantum well solar cells’, Progr. Photovoltaics
19, 865–877.
Afshar M., Sadewasser S., Albert J., Lehmann S., Abou-Ras D., Marron D. F., Rockett
A. A., Räsänen E. and Lux-Steiner M. C. (2011), ‘Chalcopyrite semiconductors for
quantum well solar cells’, Adv. Energy Mater. 1, 1109–1115.
Anderson N. G. (1995), ‘Ideal theory of quantum-well solar cells’, J. Appl. Phys. 78,
1850–1861.
Araujo G. L. and Marti A. (1994), ‘Absolute limiting efficiencies for photovoltaic energy
conversion’, Solar Energy Mater. Solar Cells 33, 213–240.
Araujo G. L., Marti A., Ragay F. W. and Wolter J. H. (1994), ‘Efficiency of multiple quantum
well solar cells’, Proc. 12th. European Photovoltaic Solar Energy Conf., H. S. Stephens
& Associates, Bedford, pp. 1481–1484.
Barnes J. M. (1994), ‘An experimental and theoretical study of GaAs/InGaAs quantum well
solar cells and carrier escape from quantum wells’, PhD Thesis, University of London.
Barnes J., Nelson J., Barnham K. W. J., Roberts J. S., Pate M. A., Grey R., Dosanjh S. S.,
Mazzer M. and Ghiraldo F. (1996), ‘Characterization of GaAs/ InGaAs quantum wells
using photocurrent spectroscopy’, J. Appl. Phys. 79, 7775–7777.
Barnham K., Ballard I., Barnes J., Connolly J., Griffin P., Kluftinger B., Nelson J., Tsui
E. and Zachariou A. (1997), ‘Quantum well solar cells’, Appl. Surf. Sci. 113/114,
722–733.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 486

486 J. Nelson & N. Ekins-Daukes

Barnham K., Connolly J., Griffin P., Haarpaintner G., Nelson J., Tsui E., Zachariou A.,
Osborne J., Button C., Hill G., Hopkinson M., Pate M., Roberts J. and Foxon T. (1996),
‘Voltage enhancement in quantum well solar cells’, J. Appl. Phys. 80, 1201–1206.
Barnham K. W. J., Braun B., Nelson J., Paxman M., Button C., Roberts J. S. and Foxon C. T.
(1991), ‘Short-circuit current and energy efficiency enhancement in a low-dimensional
structure photovoltaic device’, Appl. Phys. Lett. 59, 135–137.
Barnham K. W. J. and Duggan G. (1990), ‘A new approach to high-efficiency multi-bandgap
solar cells’, J. Appl. Phys. 67, 3490–3493.
Bastard G. (1988), Wave Mechanics Applied to Semiconductor Heterostructures, Editions
de Physique, Les Ulis.
Bauhuis G. J., Mulder P., Haverkamp E. J., Schermer J. J., Bongers E., Oomen G., Koestler
W. and Strobl G. (2010), ‘Wafer reuse for repeated growth of III–V solar cells’, Progr.
Photovoltaics 18, 155–159.
Bushnell D. B., Ekins-Daukes N. J., Barnham K. W. J., Connolly J. P., Roberts J. S., Hill G.,
Airey R. and Mazzer M. (2003), ‘Short-circuit current enhancement in Bragg stack
multi-quantum-well solar cells for multijunction space cell applications’, Solar Energy
Mater. Solar Cells 75, 299–305.
Bushnell D. B., Tibbits T. N. D., Barnham K. W. J., Connolly J. P., Mazzer M., Ekins-
Daukes N. J., Roberts J. S., Hill G. and Airey R. (2005), ‘Effect of well number on the
performance of quantum-well solar cells’, J. Appl. Phys. 97, 124908–124908-4.
Chaffin R. J., Osbourn G. C., Dawson L. R. and Biefeld R. M. (1984), ‘Strained superlattice
quantum well multijunction photovoltaic cell’, Proc. IEEE Photovoltaic Specialists
Conf., IEEE, New York, NY, pp. 743–746.
Conibeer G., Green M. A., Konig D., Perez-Wurfl I., Huang S., Hao X., Di D., Shi L.,
Shrestha S., Puthen-Veetil B., So Y., Zhang B. and Wan Z. (2010), Progr. Photovoltaics
19, 813–824.
Connolly J. P., Barnham K. W. J., Nelson J., Griffin P., Haarpaintner G., Roberts C., Pate
M. and Roberts J. S. (1995), ‘Optimisation of high efficiency Al x Ga1−x As MQW
solar cells’, Proc. Int. Solar Energy Society 1995 Solar World Congress, Harare, Zim-
babwe, http://www.researchgate.net/publication/235764284_Optimisation_of_high_
efficiency_AlxGa1-xAs_MQW_solar_cells.
Connolly J. and Rohr C. (2003), ‘Quantum well cells for thermophotovoltaics’, Semicond
Sci. Tech. 18, S216–S220.
Connolly J. P. (1998), Private communication.
Corkish R. and Green M. (1993), ‘Recombination of carriers in quantum-well solar-cells’,
Conf. Record 23rd. IEEE Photovoltaic Specialists Conf., IEEE, New York, pp. 675–680.
Corkish R. and Honsberg C. B. (1997), ‘Dark currents in double-heterostructure and
quantum-well solar cells’, Conf. Record 26th. IEEE Photovoltaic Specialists Conf.,
IEEE, New York, pp. 923–926.
Ekins-Daukes N. J., Ballard I., Calder C. D. J., Barnham K. W. J., Hill G. and Roberts J. S.
(2003), ‘Photovoltaic efficiency enhancement through thermal up-conversion’, Appl.
Phys. Lett. 82, 1974–1976.
Ekins-Daukes N. J., Barnham K. W. J., Connolly J. P., Roberts J. S., Clark J. C., Hill G. and
Mazzer M. (1999), ‘Strain-balanced GaAsP/InGaAs quantum well solar cells’, Appl.
Phys. Lett. 75, 4195–4197.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 487

Quantum Well Solar Cells 487

Ekins-Daukes N. J., Kawaguchi K. and Zhang J. (2002), ‘Strain-balanced criteria for multi-
ple quantum well structures and its signature in X-ray rocking curves’, Cryst. Growth
Des. 2, 287–292.
Farrell R. M., Neufeld C. J., Cruz S. C., Lang J. R., Iza M., Keller S., Nakamura S., Denbaars
S. P., Mishra E. K. and Speck J. S. (2011), ‘High quantum efficiency InGaN/GaN
multiple quantum well solar cells with spectral response extending out to 520 nm’,
Appl. Phys. Lett.98, 201107–201110.
Freundlich A., Fotkatzikis A., Bhusal L., Williams L., Alemu A., Zhu W., Coaquira J. A.
H., Feltrin A. and Radhakrishnan G. (2007), ‘Chemical beam epitaxy of GaAsN/GaAs
multiquantum well solar cell’, J. Vac. Sci. Technol. B 25, 987–990.
Freundlich A., Rossignol V., Vilela M. F. and Renaud P. (1994), ‘InP-based quantum well
solar cells grown by chemical beam epitaxy’, Conf. Record IEEE First World Conf. on
Photovoltaic Energy Conversion, IEEE, New York, pp. 1886–1889.
Freundlich A. and Serdiukova I. (1998), ‘Multi-quantum well tandem solar cells with effi-
ciencies exceeding 30% AM0’, Proc. 2nd World Conf. Photovoltaic Energy Conver-
sion, Vienna, p. 3707.
Goradia C., Clark R. and Brinker D. (1985), ‘A proposed GaAs based superlattice solar cell
structure with high efficiency and high radiation tolerance’, Proc. IEEE Photovoltaic
Specialists Conf., IEEE, New York, NY, pp. 776–781.
Green M. A. (2003), Third Generation Photovoltaics, Springer, New York.
Green M. A., Emery K., Hishikawa Y. and Warta W. (2011), ‘Solar cell efficiency tables
version 37’, Progr. Photovoltaics 19, 84–92.
Griffin P., Ballard I., Barnham K., Nelson J. and Zachariou A. (1997), ‘Advantages of
quantum well solar cells for TPV’, in Coutts T. J., Allman C. S. and Benner J. P.,
(eds), Thermophotovoltaic Generation of Electricity, American Institute of Physics,
New York.
Griffin P., Barnes J., Barnham K. W. J., Haarpaintner G., Mazzer M., Zanotti-Fregonara C.,
Grunbaum E., Olson C., Rohr C., David J. P. R., Roberts J. S., Grey R. and Pate M.
A. (1996), ‘Effect of strain relaxation on forward bias dark currents in GaAs/InGaAs
multiquantum well p-i-n diodes’, J. Appl. Phys. 80, 5815–5820.
Hanna M. C., Lu Z. H. and Nozik A. J. (1997), ‘Hot carrier solar cells’, in Future Generation
Photovoltaic Technologies — First NREL Conf., McConnell R. D. (ed.), American
Institute of Physics, New York, pp. 309–316.
Hirst L. C. and Ekins-Daukes N. J. (2011), ‘Fundamental losses in solar cells’, Progr.
Photovoltaics 19, 286–293.
Hovel H. J. (1975), Semiconductor and Semimetals, Volume 11 — Solar Cells, Academic
Press, London.
Johnson D. C., Ballard I. M., Barnham K. W. J., Connolly J. P., Mazzer M., Bessiere A.,
Calder C., Hill G. and Roberts J. S. (2007), ‘Observation of photon recycling in strain-
balanced quantum well solar cells’, Appl. Phys. Lett. 90, 213505–213508.
Le Bris A. and Guillemoles J.-F. (2010), ‘Hot carrier solar cells: achievable efficiency
accounting for heat losses in the absorber and through contacts’, Appl. Phys. Lett. 97,
113506–113506-3.
Lee K.-H., Barnham K. W. J., Connolly J. P., Browne B. C., Airey R. J., Roberts J. S.,
Führer M., Tibbits T. N. D. and Ekins-Daukes N. J. (2012), ‘Demonstration of photon
coupling in dual multiple-quantum-well solar cells’, IEEE J. Photovoltaics 2, 68–74.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 488

488 J. Nelson & N. Ekins-Daukes

Luque A., Marti A. and Cuadra L. (2001), ‘Thermodynamic consistency of sub-bandgap

absorbing solar cell proposals’, IEEE Trans. Electron Dev. 48, 2118–2124.
Kitatani T., Yazawa Y., Minemura J. and Tamura K. (1995), ‘Vertical transport-properties
of photogenerated carrier in InGaAs/GaAs strained multiple-quantum wells’, Jpn. J.
Appl. Phys. 34, 1358–1361.
Markvart T. (2008), ‘The thermodynamics of optical étendue’, J. Opt. A: Pure Appl. Opt.
10, 015008.
Marti A. and Araujo G. L. (1996), ‘Limiting efficiencies for photovoltaic energy conversion
in multigap systems’, Solar Energy Mater. Solar Cells 43, 203–222.
Marti A., Balenzategui J. L. and Reyna R. F. (1997), ‘Photon recycling and Shockley’s diode
equation’, J. Appl. Phys. 82, 4067–4075.
Mazzer M. (1997), Private communication.
McGehee M. D. (2009), ‘Nanostructured organic–inorganic hybrid solar cells’, MRS Bulletin
34, 95–100.
Monier C., Vilela M., Serdiukova I. and Freundlich A. (1998), ‘Photocurrent and photolumi-
nescence spectroscopy of InAs x P1−x /InP strained quantum wells grown by chemical
beam epitaxy’, J. Cryst. Growth 188, 332–337.
Moulé A. J., Chang L., Thambidurai C., Vidu R. and Stroeve P. (2012), ‘Hybrid solar cells:
basic principles and the role of ligands’, J. Mater. Chem. 22, 2351–2368.
Nelson J. (1995), ‘Multiple quantum well structures for photovoltaic energy conversion’,
Physics of Thin Films 21, Francombe M. H. and Vossen J. L. (eds), pp. 311–368.
Nelson J., Barnes J., Ekins-Daukes N., Barnham K. W. J., Kluftinger B., Tsui E. S.-M.,
Foxon C. T., Cheng T. S. and Roberts J. S. (1998), ‘Reduced radiative currents from
GaAs/InGaAs and AlGaAs/GaAs p-i-n quantum well devices’, in Conf. Record 24th.
IEEE Int. Symposium on Compound Semiconductors, IEEE, New York, pp. 413–416.
Nelson J., Barnes J., Ekins-Daukes N., Kluftinger B., Tsui E., Barnham K., Foxon C. T.,
Cheng T. and Roberts J. S. (1997), ‘Observation of suppressed radiative recombination
in single quantum well p-i-n photodiodes’, J. Appl. Phys. 82, 6240–6246.
Nelson J., Barnham K., Ballard I., Connolly J. P., Roberts J. S. and Pate M. (1999), ‘Effect of
QW location on quantum well photodiode dark currents’, J. Appl. Phys. 86, 5898–5905.
Nelson J., Barnham K., Connolly J. and Haarpaintner G. (1994), ‘Quantum well solar cell
dark currents — a theoretical approach’, Proc. 12th. European Photovoltaic Solar
Energy Conf., H. S. Stephens & Associates, Bedford, pp. 1370–1373.
Nelson J., Kluftinger B., Tsui E. and Barnham K. (1995), ‘Quasi-Fermi level separation
in quantum well solar cells’, Proc. 13th. European Photovoltaic Solar Energy Conf.,
H. S. Stephens & Associates, Bedford, pp. 150–153.
Nelson J., Paxman M., Barnham K. W. J., Roberts J. S. and Button C. (1993), ‘Steady
state carrier escape from single quantum wells’, IEEE J. Quantum Electron. 29,
1460–1467.
Paxman M., Nelson J., Barnham K. W. J., Braun B., Connolly J. P., Button C., Roberts J.
S. and Foxon C. T. (1993), ‘Modelling the spectral response of the quantum well solar
cell’, J. Appl. Phys. 74, 614–621.
Pearsall T. P. (1989), ‘Optical properties of Ge–Si alloys and superlattices’, J. Luminescence
44, 367–380.
Ragay F. W., Wolter J. H., Marti A. and Araujo G. L. (1994), ‘Experimental analysis of the
efficiency of multiple quantum well solar cells’, Proc. 12th. European Photovoltaic
Solar Energy Conf., H. S. Stephens & Associates, Bedford, pp. 1429–1433.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch10 page 489

Quantum Well Solar Cells 489

Ramey S. and Khoie R. (2003), ‘Modeling of multiple-quantum-well solar cells including

capture, escape, and recombination of photoexcited carriers in quantum wells’, IEEE
Trans. Electron Devices 50, 1179–1188.
Renaud P., Vilela M. F., Freundlich A., Bensaoula A. and Medelci N. (1994), ‘Modeling p-i
(multi quantum well)-n solar cells: a contribution for a near optimum design’, Proc.
1994 IEEE 1st. World Conf. Photovoltaic Energy Conversion, IEEE, New York, NY,
pp. 1787–1790.
Rohr C., Abbott P., Ballard I., Connolly J. P., Barnham K. W. J., Mazzer M., Button C.,
Nasi L., Hill G., Roberts J. S., Clarke G., and Ginige R. (2006), ‘InP-based lattice-
matched InGaAsP and strain-compensated InGaAs/InGaAs quantum well cells for
thermophotovoltaic applications’, J. App. Phys. 100, 114510.
Rosenwaks Y., Hanna M. C., Levi D. H., Szmyd D. M., Ahrenkiel R. K. and Nozik A. J.
(1993), ‘Hot-carrier cooling in GaAs — quantum wells versus bulk’, Phys. Rev. B. 48,
14675–14678.
Ross R. T. and Nozik A. J. (1982), ‘Efficiency of hot-carrier solar-energy converters’,
J. Appl. Phys. 53, 3813–3818.
Sah C.-T., Noyce R. N. and Shockley W. (1957), ‘Carrier generation and recombination
in p-n junctions and p-n junction characteristics’, Proc. Inst. Radio Engineers 45,
1228–1243.
Scott C. G., Sands D., Yousaf M., Abolhassani N., Ashenford D. E., Aperathitis E.,
Hatzopoulos Z. and Panayotatos P. (1997), ‘P −i −n solar cell efficiency enhancement
by use of MQW structures in the i-layer’, Proc. 14th. European Photovoltaic Solar
Energy Conf., H. S. Stephens & Associates, Bedford, pp. 2499–2502.
Sze S. M. (1981), Physics of Semiconductor Devices, Wiley, New York.
Tiedje T., Yablonovitch E., Cody G. D. and Brooks B. G. (1984), ‘Limiting efficiency of
silicon solar cells’, IEEE Trans. Electron Devices 31, 711–716.
Varonides A. C. and Berger A. W. (1997), Proc. 14th. European Photovoltaic Solar Energy
Conf., H. S. Stephens & Associates, Bedford, pp. 1712–1715.
Wang Y. P., Ma S. J., Watanabe K., Sugiyama M. and Nakano Y. (2012), ‘Management
of highly strained heterointerface in InGaAs/GaAsP strain-balanced superlattice for
photovoltaic application’, J. Cryst. Growth 352, 194–198.
Walters R. J., Summers G. P., Messenger S. R., Freundlich A., Monier C. and Newman F.
(2000), ‘Radiation hard multi-quantum well InP/InAsP solar cells for space applica-
tions’, Progr. Photovoltaics 8, 349–354.
Wanlass M. W. and Blakeslee A. E. (1982), ‘Superlative cascade solar cell’, Proc. IEEE
Photovoltaic Specialists Conf., IEEE, New York, NY, p. 584.
Weisbuch C. and Vinter B. (1991), Quantum Semiconductor Structures, Academic Press,
San Diego.
Würfel P. (1982), ‘The chemical potential of radiation’, J. Phys. C. 15, 3967–3985.
Zachariou A., Barnes J., Barnham K. W. J., Nelson J., Tsui E. S.-M., Epler J. and Pate M.
(1998), ‘A carrier escape study from InP/InGaAs single quantum well solar cells’,
J. Appl. Phys. 83, 877–881.
Zachariou A., Barnham K. W. J., Griffin P., Nelson J., Button C., Hopkinson M., Pate M.
and Epler J. (1996), ‘A new approach to p-doping and the observation of efficiency
enhancement in InP/InGaAs quantum well solar cells’, Conf. Record 25th. IEEE Pho-
tovoltaic Specialists Conf., IEEE, New York, pp. 113–117.
Zory P. S. (ed., 1993), Quantum Well Lasers, Academic Press, London.
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 491

CHAPTER 11

CONCENTRATOR SYSTEMS
IGNACIO LUQUE-HEREDIA
Compañía Española de Alta Eficiencia Fotovoltaica BSQ Solar, SL
Madrid 28040, Spain
[email protected]

and

ANTONIO LUQUE
Instituto de Energía Solar, Universidad Politécnica de Madrid
Madrid 28040, Spain
[email protected]

Porque el que no sabe mas que las palabras sin saber el fundamento que la regla tiene:
siguen se le muchos daños/y hallandose en ellos/no sabe ni alcança de donde le vienen.
(Because much harm will come to him who only knows the words; without knowing the
foundation of the rule, he does not know and cannot reach where they come from.)
Pedro de Medina, Regimiento de Navegación, fo. 26, 1563.

11.1 Introduction
Historically, solar cells have been considered to be expensive. A potential way of
reducing their cost is casting onto them a higher light intensity than is available
naturally. For this solar concentrators are used. Concentrators are optical elements
that collect the Sun’s energy in a certain area and redirect it onto the solar cells.
Obviously the collecting optical element has to be cheaper per unit area than the
solar cell, a necessary although not sufficient condition to render the concentrated
light system less expensive than an unconcentrated one.
The rapid reduction of the cost of silicon solar cells and the irruption of thin-
film cells in the market has weakened the convincing strengths of this approach.
However, another factor that has revitalised the interest in concentrated photo-
voltaics (CPV) has appeared, namely the fabulous potential of multijunction (MJ)
cells for high efficiency under concentration. This fact is perspicuous in the very
widely used National Renewable Energy Laboratory (NREL) chart in Fig. 11.1.
In this we see that the efficiency of triple-junction cells under concentration is not

491
 September 17, 2014 11:4
492

Clean Electricity From Photovoltaics (2nd Edition)

I. Luque-Heredia & A. Luque

9in x 6in
b1809-ch11 page 492
Figure 11.1 Champion laboratory cell efficiencies for different PV technologies. Courtesy of National Renewable Energy Laboratory, Golden,
CO (www.nrel.gov/ncpv/images/ efficiency_chart.jpg).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 493

Concentrator Systems 493

matched by any other photovoltaic technology. In the last decade their efficiency
has increased at the rate of about 1% (absolute) per year.
The increase of efficiency by concentration has solid theoretical grounds that
are explained later in this chapter. One of the authors has expressed his opinion
that 50% cell efficiency will eventually be achievable (Luque, 2011). This promis-
ing potential overcomes the fact that CPV only uses the direct sunbeam and not
the diffuse component of the radiation, which is harnessed by conventional photo-
voltaics (PV). At the same time the earlier motivation of concentration — reducing
the number and hence the cost of the cells required to produce a given power out-
put — becomes convincing in this context. Only under concentrated sunlight can
the sophisticated and expensive multijunction cells be exploited. But we think that
these high efficiencies have the potential of becoming cheaper than any other PV
technology and, furthermore, than any other energy technology.
The level of irradiance (luminous power flux) at which concentrating cells
operate is very variable. Recalling that for the purpose of solar cell rating, the
standard solar irradiance at the Earth’s surface is 1 kW m−2 , the level of irradiance
in static concentrator cells is in the range 1.5–5 kW m−2, while silicon tracking
concentrators range today from 10 to 500 kW m−2, and for multijunction cells
irradiances between 100 and 1500 kW m−2 are used or envisaged. It is very common
to refer to the irradiance level in ‘Suns’, meaning the number of times the actual
irradiance is higher than the standard solar irradiance. Thus a cell operating at
1500 kW m−2 is said to operate at 1500 Suns.
In this chapter, we shall look briefly first at the early history of CPV, then at
the basic operation of solar cells under concentration to explain the grounds of
the increase of the efficiency with concentration and also the limits of this: why
concentration cannot be increased indefinitely and lead to more efficient operation.
Then we outline the modern multijunction structures and their behaviour.
As regards concentrators, we start with their description. Then we examine
methods for their optical design. The theoretical grounds of concentrator optics are
also presented and the limits on increasing the concentration factor are described.
Following that, we analyse how concentrator cells are mounted and cooled, includ-
ing a quantitative analysis of the cooling.
Concentrating systems usually (but not invariably) need tracking mechanisms
to keep the sunlight focussed on the cells. These constitute an important part of
the cost of a CPV system, so we look at them in some detail in this chapter. Static
concentrators also exist, although they permit only very moderate concentration.
They also collect diffuse sunlight, to different extents. Such systems have not so far
been commercialised and will not be treated in this chapter. For more information,
the reader is directed to Luque (1986 and 1989).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 494

494 I. Luque-Heredia & A. Luque

Figure 11.2 Photovoltaic concentrator panel fabricated in the late 1970s at Sandia National
Laboratories, Albuquerque, NM.

Next we consider the performance and cost of CPV and the potential for cost
reduction. Finally we discuss what future we foresee for tracking concentrator
technology.

11.2 The early development of CPV

Figure 11.2 shows the first modern photovoltaic concentrating panel, developed
at Sandia National Laboratories in the late 1970s (Burgess and Pritchard, 1978).
Each elementary concentrator is formed by a point focus Fresnel lens (see Sec-
tion 11.4.1) that casts the radiation onto a circular cell of about 5 cm in diameter.
The concentration ratio is about 40× and total rated power about 1 kWp. A set of
15 × 8 such elementary concentrators is attached to a beam that rotates around a
horizontal axis (elevation tracking).This beam is placed on the top of a pedestal and
also rotates as a whole in azimuth. Jointly, this is the azimuth–elevation two-axis
tracking mechanism. We shall refer to this tracking configuration as the pedestal
type. The tracking electronics of this system were based on a sensor that provided
an error signal when the Sun moved out of aim. Since the panel movement is very
slow — one turn per day — the power spent tracking the Sun is negligible, usu-
ally less than 1% of the energy produced. In this first prototype, the cooling was
provided by cool water fed into the cell holder.
This concept has inspired several research prototypes, among them the Ramón
Areces panel, installed in the late 1970s at the Instituto de Energía Solar of the Uni-
versidad Politécnica de Madrid (IES-UPM) (Sala et al., 1979). In later prototypes
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 495

Concentrator Systems 495

(including the Ramón Areces), cooling of the cells was achieved by an extruded
aluminium multifin heat sink. A closed housing was also provided to protect the
cells from the environment. In some cases, parquets of several lenses were fabri-
cated to facilitate the assembling.
The passive-cooling modified Sandia Labs design was adopted by several
manufacturers in those days, Martin Marietta being the most remarkable. This
organisation installed about 350 kW in Saudi Arabia in the early 1980s (Salim
and Eugenio, 1990). Some other US companies, for instance Alpha Solarco and
Amonix (Garboushian et al., 1996) also continued this concept to develop large
panels in the 15–25 kW range.
While many other designs were considered at that time, the next original
concept that came into reality was the ENTECH concentrator. A plant of 300 kW
and several other smaller ENTECH arrays were installed (O’Neill et al., 1991).
In this concentrator, still commercialised today (mainly for space applications),
the cells are series-connected in a linear row located under an arched Fresnel lens
of linear focus, as shown in Fig. 11.3. The performance of such arched lenses is
strikingly insensitive to their position. The concentration ratio can be up to 20×,
and in some cases screen-printed cells are used. Cooling is again effected with
extruded heat sinks, and a ‘housing’ contains the whole module.
The concentrator array comprises a set of linear cell/lens modules in an ele-
vated east–west oriented frame. Each module can rotate separately to follow the
hour angle, and the frame as a whole can rotate in elevation on two supporting
poles situated at its east and west ends. This is the declination–hour angle tracking
mechanism. Its tracking control is similar to that of the Sandia prototype.
A concept developed jointly by BP Solar and IES-UPM, in an EU joint project
in which the University of Reading and ZSW-Stuttgart also participated, was the
EUCLIDES concentrator (Sala et al., 1996) shown in Fig. 11.4. The rationale

Figure 11.3 ENTECH concentrator array. This has axial focus curved lenses, tracking the
hour angle in a frame that tracks the Sun’s elevation angle.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 496

496 I. Luque-Heredia & A. Luque

Figure 11.4 View of the one-axis tracking EUCLIDES concentrator. The cells, encapsu-
lated inside modules, are the darker strip that can be seen in the figure (there is another strip
underneath). There are two separated asymmetric parabolic mirrors illuminating two rows
of cells.

behind this design was to have a high voltage output, able to feed an inverter without
the use of an intermediate transformer. The output needed for this purpose was 600–
650 V DC (direct current), this requiring many cells (∼1400) to be connected in
series, which resulted in very long arrays (84 m in the commercial version). The
arrays should therefore be horizontal and have a one-axis horizontal tracking. This
implies linear mirror optics, because linear lenses change their focal distance when
the Sun is at inclined angles.
Subject to these specifications, the concept of cell housing was no longer valid.
Instead, cells were encapsulated in receiving modules inspired by flat module
encapsulation, which provided excellent environmental protection. The receiver
module was formed of an aluminium tray, on which the cells were stuck on a layer
of material that had to be both an electric insulator and a good thermal conductor.
The cells were electrically connected in series and covered with glass. The inside
of this module was filled with a transparent resin.
The concentrating optics were made of mirrors instead of the Fresnel lenses
used previously in all (successful) photovoltaic concentration designs. The mirrors
were parabolic in outline, and their profiles were optimised using non-imaging
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 497

Concentrator Systems 497

optics (Luque, 1980), so as to allow for the highest manufacturing errors for a given
level of concentration (geometrical concentration of 32×). The focal distance,
position of the axis, useful arch within the infinite parabola, and the receiver (cell)
angle, were also optimised.
Since this mirror profile was not available in the market, and because of the
precise profile that was required (which we were not sure could be achieved with
glass), a new fabrication technology was developed using a weather-protected 3M
silvered layer laminated onto a thin aluminium sheet that was subsequently formed
to the desired shape with high precision and then stuck to a supporting aluminium
frame.
Cell cooling was provided in the EUCLIDES module by means of an
aluminium-finned heat sink, the aluminium sheets being held together by a core
stuck onto the aluminium tray of the modules. Cooling by natural convection is
more effective in vertical fins than in the inclined fins used in the two-axis tracking
configuration, so less aluminium is required for the same cooling performance.
The structure, which held two rows of cells and mirrors as shown in Fig. 11.4,
consisted of a long horizontal reticulated beam rotating on a large centrally placed
wheel resting on two smaller wheels. This configuration also provided the tracking
mechanisms. Two additional passive supports provided the vertical reaction to the
weight at the beam ends. The tracking control was provided by a microprocessor
that calculated the correct aiming of the system at any moment of the year, based on
astronomic data. However, in order to account for inaccurate module positioning,
a feedback based on system power output maximisation was occasionally used,
generating a table of corrections for the astronomic calculations.
The EUCLIDES concentrator used crystalline silicon cells of the laser-grooved
buried contact (LGBG) type, a concept that had been developed by the University
of New South Wales. These cells (also called SATURN, after the name of the indus-
trialisation project of BP Solar) had proven high efficiencies. These cells were, at
the time, very convenient for use in concentrator systems because, while their 1 Sun
version was more expensive than those used in conventional c-Si modules, when
used as 125×125 mm2 concentrator cells and designed for 30×, they could be sold
at 10–12 each, thus becoming economically very attractive to the manufacturer.
In their concentrator version, LGBG cells showed a highly homogeneous voltage
along their metallisation and a low surface recombination, which allowed them to
reach an average of 18.5% efficiency at 30× and up to 20% in small (1 cm2 ) cells
at 100× (Bruton et al., 1994).
The first prototype of this concentrator technology was installed in Madrid
at IES-UPM in 1995. It proved to have an efficiency of 10.8% close to noon on a
typical summer day which, extrapolated to the then-proposed concentrator standard
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 498

498 I. Luque-Heredia & A. Luque

Figure 11.5 Seven of the fourteen arrays constituting the EUCLIDES demonstration plant
installed in Tenerife by the ITER, the IES and BP Solar.

test conditions — direct radiation of 800 W m−2 and cell temperature of 25◦ C —
resulted in an efficiency of 14.4% (Luque et al., 1997). This was achieved at a
lower cost than a typical flat-plate PV power plant of the same rating at that time.
Following the Madrid prototype, a 480 kW EUCLIDES demonstration plant
was built in Tenerife (see Fig. 11.5) through the joint initiative of BP Solar, the
Instituto de Tecnología y Energías Renovables (ITER) in Tenerife and IES-UPM.
This became the biggest CPV plant of its time. However, several problems that
appeared in the series production of the receiver, some overestimation of the con-
centrator benefits, and finally the merger of BP with Amoco (which resulted in the
consolidation of their respective solar subsidiaries BP Solar and Solarex), led to
the abandonment of concentrator projects by these companies.
Other smaller CPV prototypes were developed using combinations of the
above designs. For instance, a point-focus concentrator in a declination–hour angle
frame configuration like that of ENTECH was developed by Midway Laboratories.
Other less fully developed ideas have also been published. For instance, among
the most original suggested was a large reticulated platform resting on wheels or
even floating on water to provide the azimuth tracking. On this platform, rows of
arrays would be installed, each with its own elevation tracking mechanism (Alarcón
et al., 1982).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 499

Concentrator Systems 499

These developments and others, which appeared in the last two decades of
the past century, constituted the ‘first wave’ of photovoltaic concentrators, set-
ting the foundations of CPV and paving the way for current CPV start-ups and
developments.
After this overview of the historical background of CPV, which has also intro-
duced the different elements of a concentrator system by describing several real
examples, we proceed in the next sections to further analyse them one by one and
in greater detail.

11.3 Concentrator solar cells

We devote this section to a theoretical overview of single-junction concentrator
solar cells. They are easier to understand and contain the fundamentals of any
concentrator solar cell of today. Actually, MJ solar cells are just series-connected
stacks of several single-junction solar cells. Many of the statements in this section
are therefore valid for MJ solar cells, which are discussed by Masafumi Yamaguchi
in Chapter 7.
Concentrator solar cells differ from conventional solar cells in that they must
be able to extract more current per unit area. Furthermore, as they must be no more
than a small part of the concentrator cost, they can afford to be more expensive in
order to be more efficient.
Under increasing luminous flux, the current from a solar cell usually increases
proportionally. The open-circuit voltage will increase according to the formula
Voc = Vth ln(Ith /I0 , derived from the single-diode model of the solar cell. In this
formula Iph is the photogenerated current, which is very close to the short-circuit
one; I0 is the reverse-bias saturation current; and Vth = βkT/q is the thermal
voltage. Here β is a quality factor used in the single-diode model. This tends to
be one in concentrator cells so that Vth = 0.0257 at 25◦ C. In consequence Voc
increases logarithmically with the current and thus with the luminous flux. Some
smaller deviations may be observed at very high fluxes due to inhomogeneous
distribution of the current, but they are not relevant at the operation conditions.
The fill factor can be written as (Luque, 1989):


 
Vth Vth Voc Iph RS
ηfill = 1 − 1− ln − (11.1)
Voc Voc Vth Voc

Excluding the term in R — the series resistance — ηfill slowly increases with Voc
and therefore with the flux intensity. However, at high currents, when the ohmic
drop increases, the fill factor decreases and this becomes the main factor causing
an efficiency decrease. Obviously this implies that the series resistance must be
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 500

500 I. Luque-Heredia & A. Luque

Figure 11.6 Evolution of efficiency with concentration factor in recent record-breaking

multijunction cells. Labels give information on the research lab or company developing the
cell, and the number of junctions; all of them are monolithic lattice-matched. The typical
evolution of efficiency with concentration that increases with flux intensity until loss due
to ohmic drop becomes significant and can be seen, particularly in the 3J Spectrolab cell.
Courtesy of Prof. Algora, personal communication.

very small in concentrator cells. This is perhaps the most important difference as
compared with the non-concentrator solar cells.
As a consequence of all this, the solar cell efficiency ηmp


Iph
ηmp = Voc ηfill = SL Voc ηfill (11.2)
PL
increases with concentration until the moment when the ohmic drop causes a reduc-
tion, as seen in Fig. 11.6 (single-junction cell labelled 1J IES-UPM). Here SL is the
cell photosensitivity (approximately constant) and PL is the luminous flux. A very
easy rule is that the highest efficiency is achieved when Iph RS ≈ Vth (Luque, 1989).
This behaviour, although ruled by a more complex expression, is also present (for
the same fundamental reasons) in MJ solar cells, as can also be seen in Fig. 11.6.
A very important feature of the concentrator solar cells is their thermal
behaviour. At higher temperatures the bandgaps decrease and therefore the cur-
rent tends to increase, but this is a very small effect — the relative increase of the
short-circuit current in silicon is only 3 × 10−4 K−1 . The most important thermal
effect in a solar cell is the voltage reduction, which is governed by the expression
dV E g /q − V − I RS
=− (11.3)
dT T
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 501

Concentrator Systems 501

where E g is the bandgap. In Si the temperature coefficient of the voltage drop

is about −2.1 mV K−1 whereas in GaAs it is only −1.4 mV K−1. This means
that cells experience a serious performance reduction when they operate at
high temperatures, which is smaller under concentration when the voltages are
bigger.
Most concentrator cells have a classical structure, with an extended p-n junc-
tion on the front face. The series resistance is reduced by using a carefully designed
grid. For the same surface-covering fraction, the series resistance is decreased if the
fingers are deep and narrow (less than 10 µm). That is why lift-off photolithogra-
phy is used to delineate the grid of concentration cells. The optimum grid-covering
factor is higher than in non-concentrator cells. Furthermore, the ohmic losses also
decrease if the cell is of small size, but in some cases, in particular in silicon cells,
going below 1 cm2 leads to reduced efficiency owing to increased edge recombina-
tion. In GaAs cells — of much shorter carrier lifetime — cells of 1 mm2 are more
effective.
In general, screen-printing technology — used for most non-concentrator sil-
icon cells — works poorly at concentrations above 10×. On the contrary, the
LGBG cell technology chosen for the EUCLIDES concept could be operated at
concentrations of 40×, as has been explained above.
Nevertheless, in cells of classical structure there is a fundamental limitation
on the series resistance that is independent of the grid and of the cell size: this is the
base series resistance. This component can only be reduced by reducing the base
resistivity and thickness. In silicon cells of conventional structure, this limits the
concentration factor to some 120× (Terrón et al., 1992). In III–V semiconductors
the base is usually very narrow and is situated on a substrate that is, to a large extent,
electrically inactive. This allows it to be of very low resistivity and therefore these
cells can operate properly at much higher concentrations, may be at 5000×, but
certainly at 1000×. It was at IES-UPM that the ability of III–V cells to operate
efficiently at more than 1000× was first proved (Algora et al., 2001).
Swanson and co-workers developed a new type of silicon cell (Sinton et al.,
1985) — the back point contact (BPC) cell — with a low-doped base of high
lifetime and p + and n + dot-shaped emitters that are the positive and the negative
contacting areas of the solar cells, respectively, all located closely intercalated at the
cell back face. In such cells no current flows from top to bottom, as in conventional
ones (but carriers of both types do, annihilating the overall current), and because
of this such cells do not have a base component of the series resistance. Thus they
can operate at higher concentration. These cells do not have any grid on the front
face and all the electric contacts are made to the rear face through interdigitated
dense grids.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 502

502 I. Luque-Heredia & A. Luque

In BPC cells the limit to the concentration efficiency is not determined by

series resistance but by the super-linear increase of Auger recombination at high
luminous flux caused by the steep profile of carriers produced by intense illumi-
nation. This leads to a sub-linear current–irradiance characteristic that precludes
operation at very high concentration. Best efficiencies are at some 200×. New
cells have been developed based on the same principle, with closely intercalated
p+ and n + emitters in the front face, covered by interdigitated grids in wedge form
so that they deflect the light towards the uncovered silicon areas. Such cells are in
principle able to operate at higher concentration, over 500× (Luque et al., 2004a).
However, the fabricated cells were imperfect and did not lead to the expected high
efficiencies (Luque et al., 2004b).
Despite Swanson’s cell invention originally being targeted at integration in
concentrator systems (and in fact it was used until 2010 by well-known CPV
manufacturers such as Amonix in the USA and Guascor Fotón in Spain), the BPC
cell, after having been reengineered to make it cost-effective in flat PV modules,
has been successfully commercialised as such by SunPower.

11.3.1 Beyond the Shockley–Queisser limit: multijunction solar cells

One of the main reasons behind today’s interest in CPV has been the strik-
ing increase in efficiency over the past 20 years of MJ cells based on III–V
semiconductors. These cells were originally devised for space applications —
powering of satellites — but proposals for their use in large-scale terrestrial power
concentrator plants are about as old. The potential of high-efficiency photovoltaic
devices to make the promise of concentrators to provide cheap solar electricity real,
has triggered the quest for other concepts that, independently or complementing MJ
cells, are today being developed in the laboratory, and may become tomorrow’s
concentrator cells. All these new concepts, treated in detail by Martin Green in
Chapter 2 of this book, are sometimes grouped together as third-generation cells,
which have as their common property their capacity to circumvent the so-called
Shockley–Queisser efficiency limit (Luque and Martí, 2011).
In 1961, William Shockley and Hans Queisser (SQ) published an elegant paper
establishing the efficiency limit of a solar cell (Shockley and Queisser, 1961).
They considered the solar cell as a system of two levels, the valence band (VB)
and the conduction band (CB), and assumed that each photon with energy above
the bandgap pumps one electron from the valence band to the conduction band.
They also assumed infinite mobility so that the CB and VB quasi-Fermi levels are
horizontal and their difference everywhere is the cell voltage (times the electron
charge). They also assumed that selective contacts (heavily doped n and p regions)
are formed so that it is possible to extract electrons and holes separately. They
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 503

Concentrator Systems 503

further assumed that all non-radiative recombination processes are suppressed and
that radiative recombination is only produced by radiation that escapes through the
cell surfaces. Thus they established a balance equation of the photons entering and
escaping the semiconductor and the electron-hole pairs produced and extracted.
No mention is made of the properties of the semiconductors — excepting the
bandgap — or of the need of any p-n junction.
The calculations were repeated by Araújo and Martí (1994) for the case of solar
radiation impinging on the solar cell isotropically rather than within the natural
cone of direct solar radiation. As we will see in the next section, this condition
corresponds to the maximum possible sunlight concentration at the Earth’s surface.
In this case the maximum possible efficiency is 40.7% for black-body radiation
at 6000/300 K (Sun/ambient) temperatures, as shown in Fig. 11.7, together with
cases for other spectra (direct and global) and concentration levels.
However, if the design principles of the solar cells are not considered, and
only thermodynamic considerations are taken into account, the limiting efficiency
of a solar energy converter was calculated by Landsberg and Tonge (1980) and is

Figure 11.7 Shockley–Queisser efficiency limit for an ideal solar cell versus bandgap
energy for: a) unconcentrated 6000 K black-body radiation (1595.9 Wm−2 ); b) fully con-
centrated 6000 K black-body radiation (7349.0 × 104 W m−2 ); c) unconcentrated AM 1.5-
Direct13 (767.2 W m−2 ); d) AM 1.5 Global13 (962.5 W m−2 ). Reprinted with permission
from Araújo and Martí (1994), © Elsevier.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 504

504 I. Luque-Heredia & A. Luque

given by


4
4 Tamb 1 Tamb
ηLandsberg = 1 − + (11.4)
3 TSun 3 TSun
Assuming the Sun to be a reservoir of photons at 6000 K (TSun ) and the Earth a cold
reservoir at 300 K (Tamb ), this results in a maximum 93.33% maximum conversion
efficiency (Section 2.4 of this book gives a detailed treatment).
The main reason a solar cell, even if ideal as in the SQ assumptions, is so
far from the Landsberg efficiency is that it does not use the whole solar spectrum
but only those photons of energy higher than the bandgap. A second reason is that
entropy is generated by the cooling of the electron-hole pairs produced by photons
with energy higher than the bandgap.
The straightforward way to overcome the fundamental limitation of the SQ
single-bandgap cell is to use several solar cells of different bandgaps to convert
photons of different energies. A simple configuration to achieve this is vertical
stacking of the cells so the uppermost cell has the highest bandgap, and lets the
photons with energy less than its bandgap pass through to the cells below. The last
cell in the stack is the one with the narrowest bandgap. In this way the entire solar
spectrum can be used and the upper cells can act as a filter for the lower ones,
each being illuminated by a narrower range of photon energies, thus reducing
the production of entropy. For an infinite number of ideal cells, as per the SQ
definition, the maximum efficiency of this arrangement is 86.8% under maximum
concentration, for a 6000/300 K Sun/ambient temperature.
The most common practical realisation of this concept is the so-called mono-
lithic MJ cell, consisting in a stack of solar cells of different semiconductor mate-
rials, along with tunnel diodes to interconnect them in series, all grown on a single
substrate. III–V compound semiconductors of a variety of materials can be grown
epitaxially with very high quality, if all the materials have a similar lattice con-
stant (although the so-called metamorphic cells allow for some lattice mismatch to
achieve a better bandgap combination). Figure 11.8 shows a schematic of a stack
of three cells built on a Ge substrate; the figure caption shows the roles of the dif-
ferent layers. The so-called window layers are intended to prevent recombination
at the surface or in the tunnel junctions. The fraction of Group III elements in each
layer is adjusted to obtain the proper lattice constant and, if possible, the desired
bandgap. As noted in the figure, the data are illustrative and each manufacturer has
their own recipe.
As noted above, the different cells in a monolithic MJ device are series-
connected, and this introduces an additional constraint, because the photogenerated
current of the stack will be the smallest of the currents generated by each cell.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 505

Concentrator Systems 505

Figure 11.8 Schematic cross-section of a monolithic two-terminal series-connected three-

junction solar cell. An n-on- p configuration is illustrated. Doping indicated by n ++ , n +
and n (or p++ , p+ , p) corresponds to electron (or hole) concentrations of the order of
1019 − 1020 , 1018 and 1017 , respectively. Typical materials, bandgaps and layer thicknesses
for the realisation of this device structure as a GaInP/GaAs/Ge cell are indicated. Note that
not all layers in an actual device (e.g. tunnel-junction cladding layers) are included in the
illustration. The figure is not to scale. Reprinted with permission from Friedman et al.
(2011), © Wiley.

The best situation is when all the cells produce the same photogenerated current.
This restricts the efficiency of a series-connected stack with respect to that of an
unconstrained stack. For an infinite number of cells, the efficiency limit is the same
(Tobías and Luque, 2002). However, in the case of three cells (the configurationthat
is commercially available today), the efficiency limit for the unconstrained stack
is 63.6% compared with 63.0% for the series-connected stack, where conditions
of maximum concentration (about 46000 Suns) and Sun/ambient temperatures of
5762 K/298.15 K are assumed in both cases. The Sun temperature considered here
is that deriving from the extra-terrestrial spectrum and the ambient temperature
is 25◦ C. These are some of the commonly used standard measuring conditions.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 506

506 I. Luque-Heredia & A. Luque

Optimal bandgaps for the series-connected case are 1.75, 1.10, 0.58 eV, but opti-
mal bandgaps at lower concentration are higher, and efficiency is lower.
Two main strategies are followed for the growth of the epitaxial cells: the
lattice-matched (LM) approach and the metamorphic (MM) approach. In the first
case, the lattices of the different cells are as closely matched as possible. That is
why a small proportion of In (1%) is added to the GaAs, allowing close matching
with the Ge substrate. The drawback is that this procedure does not admit optimal
bandgaps for ternaries of the elements under consideration (Ga, In, P and As). This
is evident in Fig. 11.9 (Guter et al., 2009), where, for the lattice-matched (LM)
cell, the bandgaps of the two upper cells are a bit too large. In this scheme, the
upper III–V cells give too little current as compared with the Ge cell.
In the metamorphic approach the bandgaps are better adjusted, but at the
expense of permitting the formation of threading dislocations, as shown in

Figure 11.9 Detailed balance calculations for the efficiency of different triple-junction
solar cell structures under the AM 1.5D ASTM G173–03 spectrum at 500 kW m−2 and
298 K. The grey haze, whose contours are projected in the three Cartesian planes, rep-
resents bandgap combinations, which allow efficiencies from 60.5% to 59.0% and hence
mark the optimum. Five specific triple-junction solar cell structures are shown: the lattice-
matched Ga0.5 In0.5 P/Ga0.99 In0.01 As/Ge (LM), two metamorphic GaInP/GaInAs/Ge (1.8,
1.29, 0.66 eV for MM1) and (1.67, 1.18, 0.66 eV for MM2), as well as two inverted meta-
morphic GaInP/GaInAs/GaInAs (1.83, 1.40, 1.00 eV for Inv1) and (1.83, 1.34, 0.89 eV for
Inv2) devices. Reprinted with permission from Guter et al. (2009), © AIP.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 507

Concentrator Systems 507

Figure 11.10 a) Lattice-matched and metamorphic three-junction cells grown on Ge. Dis-
locations are shown in the stressed layers; b) inverted metamorphic cell grown on GaAs,
detached and bonded to carrier; c) bifacial epitaxy metamorphic cell. Reprinted with per-
mission from Wojtczuk et al. (2010), © IEEE.

Fig. 11.10. This reduces the cell quality and may balance out the advantages of
better bandgap matching.
Very good results have been obtained with the present structures, as detailed by
Yamaguchi in Chapter 7. Several three-junction cells have reached efficiencies of
well over 40%; in late 2012, the world record was held by Solar Junction’s LM three-
junction InGaP/GaAs/GaInNAs cell, which is 44% efficient at 947 Suns. As for
four-junction cells, in autumn 2013 the record was held by Fraunhofer/Helmholtz
Centre’s cell, reported as 44.7% efficient under 297 Suns. These efficiency mea-
surements have been made at certified laboratories. However, it is worth noting
that uncertainty in the measurements is given as ±2.5 absolute percentage points
(Green et al., 2010).
Further increase in the efficiency of MJ cells may come through the use of
quaternary compounds that provide an additional degree of freedom to tune lattice
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 508

508 I. Luque-Heredia & A. Luque

match and bandgap simultaneously, or through the use of nanotechnology to tune

the semiconductor bandgap, such as in quantum well (QW) structures, which are
discussed in Chapter 10. The most obvious way of increasing the potential MJ cell
efficiency is to add more junctions. This is not straightforward, but there are several
studies on how to proceed. The work by King and co-workers of Spectrolab (King
et al., 2009) provides recommended reading on the subject. They have, in fact,
already presented a five-junction cell, that achieved an efficiency record under no
concentration of 37.8% (Optics.org, 2013).
Even if their detailed analysis is beyond the scope of this chapter, it has already
been mentioned that other high-efficiency concepts have been proposed that, hav-
ing the potential of exceeding the SQ limit, could also lead the way to cells that
could be stacked as multijunctions, further boosting the overall efficiency. Basically
these new concepts seek both a better matching of bandgaps and photon energies,
as well as the reduction of entropy production from excess photonic energy over
the bandgap. Three of these new concepts were qualified as ‘Revolutionary Photo-
voltaic Devices: 50% Efficient Solar Cells’ in a book edited by the US Department
of Energy (Lewis et al., 2005), these being the multiple exciton generation (MEG)
solar cell (Kolodinski et al., 1993), the intermediate-band (IB) solar cell (Luque
and Martí, 1997; 2012) and the hot-carrier solar cell (Ross and Nozik, 1982). Their
maximum theoretical efficiencies under the SQ assumptions would be, respectively,
85.4%, 63.2% and 85.4% (Luque and Martí, 2011) for 6000 K/300 K Sun/ambient
temperature. Martin Green discusses these and other high-efficiency concepts in
Chapter 2.

11.4 Optics for photovoltaic concentrators

Introducing the theoretical framework of the optics of the photovoltaic concentra-
tor, as well as its relatively modern recent design toolbox, is a good first step before
going on to describe the most common practical realisations.
Geometric optics has usually been applied to the design of image-forming
optical systems, and in this ambit it has achieved a very high degree of sophisti-
cation. The problem of collection and concentration of luminous power was also
traditionally one of the fields of application of geometric optics. However, it was
not until the early 1970s that specific design methods were developed for this type
of problem. These methods demonstrated that the most efficient light concentra-
tion systems could be obtained when image formation is not considered as a design
constraint, allowing thus for an extra degree of freedom.
Geometric optics results from the solutions obtained in Maxwell’s equations
for the electromagnetic field when wavelength tends to zero, i.e. spatial variations
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 509

Concentrator Systems 509

of the field are big compared with the wavelength. Then the concept of the light ray
is defined as the curve that is always normal to the wavefronts. The light ray (light
path) between any two points will have to satisfy Fermat’s principle, this being a
variational formulation stating that the length of the optical path followed will be
stationary (dL = 0), i.e. a maximal, a minimal or a saddle point, where the optical
path length between two points is defined as
 A
L= n(x, y, z)dl (11.5)
B

where n(x, y, z) is the refractive index of the medium at point (x, y, z) and dl is
the differential length along the light’s path between points A and B. All the laws
of geometric optics can be derived from Fermat’s principle, including Snell’s laws
of reflection and refraction. In fact, minimal path length is the case when a light
ray passes from one medium into another (refraction), or is reflected by a planar
mirror (this was already known by the Greeks, who formulated the reflection law
exactly through Fermat’s principle). However, it can also be a maximum, as occurs
in gravitational lensing.
An alternative formulation of geometrical optics results from the Hamiltonian
equations. A ray, specified as passing through a point (x, y, z) with a direction
given by the unitary vector v̄ of its direction cosines, is represented as a six-
vector (x, y, z, p, q, r ) where ( p, q, r ) = n(x, y, z)v̄ is the vector of the so-called
optical direction cosines of the ray. The Hamiltonian formulation states that the
rays follow trajectories that are solutions of the following system of first-order
differential equations
dx ∂H dp ∂H
= =−
ds ∂p ds ∂x
dy ∂H dq ∂H
= =−
ds ∂q ds ∂y
dz ∂H dr ∂H
= =− (11.6)
ds ∂r ds ∂z

where H = n 2 (x, y, z) − p2 − q 2 − r 2 is the Hamiltonian function, and s is a

parameter with no specific physical meaning. Valid solutions must result in H = 0
as required by the definition of ( p, q, r ) above. The Hamiltonian formulation can
be derived from Fermat’s variational principle through its equivalent in the form
of Lagrange equations, and then applying the Legendre transformation.
This Hamiltonian formulation invites us to consider geometric optics’ rays
as points in the six-dimensional space (x, y, z, p, q, r ), where rays are curves
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 510

510 I. Luque-Heredia & A. Luque

belonging to the five-dimension variety p 2 + q 2 + r 2 = n 2 (x, y, z). This five-

dimension space is called the extended phase space. If M is a set of points in the
extended phase space, and (M) is the set of the (points contained in the) rays
passing through the points in M, then M is a ray manifold if there are no two
points in M that belong to the same ray, i.e. two points that represent the same ray
in two different time instants. It is usual to define a ray manifold M through its
intersection with a reference surface  R of the space x − y − z that must intersect
the rays of M only once. This reference surface will then define a four-dimension
variety termed the phase space.
If M4D is a four (dimensional) parameter ray manifold then the etendue of the
manifold is defined as the integral:

E(M4D ) = dx dy d p dq + dx dz d p dr + dy dz dq dr (11.7)
M4D ( R )

which measures how many rays a manifold has. The theorem of conservation of the
etendue is fundamental to non-imaging optics design and it states that the etendue
is an invariant of any ray manifold as it propagates through an optical system.
In other words, the etendue is independent of the reference surface used for its
calculation.
In concentrator optics design, the bundle of rays that come from a certain light
source, the Sun in the case of solar energy, and impinge on the entry surface of the
optical concentrator is called the input manifold and is denoted by Mi . The manifold
of rays that traverse from the exit surface of the concentrator to the photovoltaic
cell, performing as receiver is denoted by Mo . The collected manifold Mc of rays
is defined as those rays that are connected by the concentrator and belong both to
the input and exit manifold. Also Mo is a subset of the manifold M R composed by
all the rays that can reach the receiver.
Design problems in non-imaging optics can be of several different types. If
the design is intended to couple the input and output manifolds perfectly, the
concentrator obtained will be such that (Mi ) = (Mo ) = (Mc ), and it is
then said to be ideal. If (Mc ) = (Mo ) = (M R ) then the concentrator is
said to be maximal provided it is able to illuminate the receiver isotropically.
A concentrator design with the properties of being both ideal and maximal is known
as optimum. On other occasions the target may be to just have Mi included in Mo ,
(Mc ) = (Mi ) ⊂ (Mo ), in which case we can impose an additional design
condition, which could be that Mc produces a prescribed irradiance distribution
on the receiver, e.g. homogeneous irradiance in photovoltaic cells improves their
conversion efficiency.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 511

Concentrator Systems 511

There are some ray manifolds that are especially useful. For example, the so-
called finite source is that of the rays that connect two parallel concentric planar
discs, both within the same homogeneous optical medium. When one of the discs
is moved away from the other towards infinity, then we talk of an infinite source
and the manifold consists of all the rays that impinge on the other disc with an
incidence angle (with respect to the entry surface normal) less or equal to a certain
constant angle α, named the source’s acceptance angle. It can be proved that the
etendue of an infinite source E inf is

E inf = Aπn 2 sin2 α (11.8)

where A is the disc’s surface and n is the refractive index of the surrounding
medium.
The geometrical concentration is defined as the ratio of the area of the entry
surface of the concentrating system to that of the receiver. As we have already
been doing, in a concentrator design the geometric concentration is usually written
followed by an ×, e.g. a 1000× design is a concentrator with an entry surface 1,000
times bigger than the surface of the cell.
In order to better quantify geometrical concentration, let Mi be the ray manifold
of an infinite source with acceptance angle α, impinging on the entry surface of a
concentrator system surrounded by a medium of refractive index n. Also let Mo
be the output manifold, characterised by a homogeneous angular spread β, and
n o the refractive index of the medium surrounding the receiver. Then if the entry
and receiver surface areas are AE and AR respectively, due to the conservation
of etendue the input and output manifolds, E Mi and E Mo , must be equal if the
concentrator perfectly couples these two manifolds (i.e. it is an ideal concentrator):

E Mi = AE n 2i π sin2 α = AR n 2o π sin2 β = E Mo ≤ E M R = AR n 2o π (11.9)

where the last inequality relates the input and output etendue to the maximum
possible etendue at the receiver when this is illuminated isotropically (β = π/2)
(i.e. it is a maximal concentrator), from which a maximum limit for geometric
concentration Cg can be derived:

AE n2
Cg = ≤ 2 o2 (11.10)
AR n i sin α
This is usually referred to as the thermodynamic limit of concentration, because it
can also be derived by equalling the temperature of an ideal black absorber to that of
the Sun considered a black-body radiator with TS = 5777 K (this is closer to reality
than the 6000 K often used for mnemonics) in accordance with the second law of
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 512

512 I. Luque-Heredia & A. Luque

thermodynamics. In the most common case of the entry surface to the concentrator
being surrounded by air (n i = 1), Eq. (11.10) turns into
n 2o
Cg ≤ (11.11)
sin2 α
This can also be expressed as

CAP = Cg sin2 α ≤ n 2o (11.12)

where the square of the refractive index of the medium surrounding the receiver
becomes the upper bound to the concentration acceptance product (CAP). This
parameter is frequently used today as a figure of merit of concentrator designs,
showing how close they get to ideality.
If we have a four-parametric ray manifold M, then δ M is defined as its edge-
ray manifold and is the boundary, in topological terms, of that manifold. δ M is
itself a three-dimensional subset of manifold M. One of the most powerful tools of
non-imaging optics design is the so-called edge-ray theorem, stating that if we want
to perfectly couple two manifolds Mi and Mo then it is enough to match their δMi
and δ Mo subsets. This can be expressed as (Mi ) = (Mo ) ⇔ (Mi ) = (Mo ) as
the inverse relationship is also true: if Mi and Mo are coupled by an optical system
then their edge-ray manifolds will also be coupled. This means that the design
process can focus just on edge-ray manifolds, thus reducing by one the dimension
of the problem of coupling four-dimensional manifolds. Even though this theorem
was assumed to be true and has been in use since the beginning of non-imaging
optics, it was not to be proved till the mid-1980s by IES-UPM’s Miñano (1985a,
b and c).
The problem of coupling two four-dimensional manifolds is usually too com-
plex, and in order to make it more attainable the usual simplification consists in
imposing an axis or plane of symmetry, on both the ray manifolds and the optical
system developed, and then working out the design in one of the planar sections.
Then it is a couple of two-dimensional manifolds, m i and m o subsets of Mi and
Mo , that are to be matched. For example, in the case of assuming a rotationally
symmetric optical system the design is carried out in one of its meridian planes —
those that contain the axis of symmetry — where ideal solutions perfectly matching
m i and m o can be obtained. Afterwards the three-dimensional optical system will
be obtained by rotating the two-dimensional resulting design around the symmetry
axis. However, there will be no guarantee that non-meridian rays in the Mi and
Mo manifolds will also be ideally coupled, and the assessment of the behaviour
of these other rays usually is completed using ray-tracing software applications.
Ray-tracing CAD tools, essentially able to compute Snell’s laws for very large sets
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 513

Concentrator Systems 513

of light rays as they proceed through different optical media, are very much used
in non-imaging optics to simulate and characterise the performance of resulting
designs.
An important characterisation parameter of CPV optical designs is the total
transmission ratio, defined as
P(Mc )
η= (11.13)
P(Mi )
where P(M) is the power transmitted by the M ray-manifold. Then the optical
concentration Co of the system is defined as

Co = C g η (11.14)

When designing for infinite sources with a certain acceptance angle, trans-
mission can be plotted as a function of the rays’ input angle θ so that we obtain the
transmission angle curve
d P[Mc (θ, θ + dθ )]
T (θ ) = (11.15)
d P[Mi (θ, θ + dθ )]
where d P[M(θ, θ +dθ )] is the infinitesimal power transmitted by rays in manifold
M in the angular differential interval (θ, θ + dθ ). The maximum of T (θ ) is termed
as the optical efficiency, ηo , usually occurring at normal incidence, i.e. θ = 0.
Several different design methods have been developed in the field of non-
imaging optics. The earliest was proposed by Welford and Winston (1978) and
used by them to develop all the family of the so-called CPCs (compound parabolic
concentrators) based on the application of the edge-ray theorem in two dimensions
(Fig. 11.11). The primary prototype of this family of concentrators consists of two
symmetric parabolic mirrors, each focusing edge rays entering the concentrator
with the maximum acceptance angle at the other’s lower rim, which coincides
with the receiver’s edge. In its two-dimensional version, this design performs as
an optimum concentrator. Several other versions of the CPC were later developed,
such as the dielectric-filled CPC (Ning et al., 1987) using lossless total internal
reflection instead of mirrors (which incur losses of ∼10%), or CPC designs for
finite sources or for non-maximal concentrations.
Welford and Winston attempted the extension of their initial edge-ray design
method to three-dimensional systems with their flow-line method, the most remark-
able product of which was the hyperboloid concentrator, also called the trumpet
concentrator (Winston and Welford, 1979a, b). Later, the Poisson bracket method
was developed by Miñano (1985a, b). While this method was specifically con-
ceived for the design of three-dimensional concentrators, it usually requires the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 514

514 I. Luque-Heredia & A. Luque

Figure 11.11 Winston’s CPC consisting of two symmetric parabolas with an inner mirrored
surface, and with their focal points at each end of the receiver. While the two-dimensional
version performs as an optimum concentrator, the plot on the right shows the transmission
performance of a three-dimensional rotationally symmetric design, with a design acceptance
angle of 16◦ where transmission is brought down to zero in approximately 1◦ .

use of variable refractive index media, this being impractical in most cases. The
main interest of the Poisson bracket method is that it is able to produce ideal three-
dimensional concentrators, thus ultimately proving their theoretical existence. It
also produced a family of ideal two-dimensional concentrators, the compound tri-
angular concentrator (CTC), with two different regions of constant refractive index
(Miñano, 1985c) (see Fig. 11.12).
The Simultaneous Multiple Surface (SMS) method was developed in the early
1990s at IES-UPM (Miñano and González, 1992) as a two-dimensional method,
and has to date produced a long list of different designs. Quite curiously, it was born
while trying to solve a conjecture. It was known that a single refractive surface can
exactly focus a plane wavefront into a point; these sort of surfaces are the so-called
Cartesian ovals. The conjecture, based once again on the edge-ray theorem, was
that two plane wavefronts could be exactly focused into two points, the receiver
edges, by means of consecutive refractions or reflections in two optical surfaces.
In fact the conjecture can be extended into the imaging optics world, by rephrasing
it as a number n of spherical wavefronts generated at object points in the near field
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 515

Concentrator Systems 515

Figure 11.12 a) Miñano’s compound triangular two-dimensional concentrator designed

using the Poisson bracket method, which requires no curved lines. It is filled with a medium
of refractive index n b > 1 plus narrow strips of refractive index n s > n b . These strips are
transparent or act as mirrors depending on the angle of incidence of the rays, and the
direction of the collected rays can be varied by modifying the refractive index of the strips;
b) Winston’s trumpet concentrator.

(or again plane wavefronts for points at infinity) that are to be focused into n image
points, again by sequential refractions or reflections in n optical interfaces.
No formal proof of the two-surface conjecture has yet been produced, but the
SMS method was developed as an iterative method by which these two sought
optical surfaces could be generated. Several designs have been produced by this
method, generally grouped by families named RR, XR, RX, XX, RXI, where the R
stands for refraction, X for reflection, and I for total internal refraction, which are
listed in the same order as incoming light rays will encounter their two constituent
optical interfaces (Fig. 11.13).

11.4.1 State-of-the-art of concentrator optics

Some of the canonical non-imaging optics developments described above have
important practical limitations, starting for example with the famous CPC. Despite
its optimality in two dimensions, and high optical efficiency in three dimensions
(over 90%), if designed for low acceptance angles, as required by Sun-tracking
concentration systems, its length-to-width ratio becomes unfeasibly high (e.g. 920
for an acceptance angle equal to the Sun’s half-subtended angle).
Despite the powerful theoretical design toolbox developed, in many cases
the optics used in photovoltaic concentration are simply derived from elementary
geometric optics, basically using the refraction and reflection laws. Design, when
not trivial, is undertaken by ray-tracing codes based on these laws.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 516

516 I. Luque-Heredia & A. Luque

Figure 11.13 a) The RXI (Refraction + Reflection & Total Internal Reflection) concentrator
designed using the Simultaneous Multiple Surface (SMS) method. These light rays refract
at the front surface, then reflect in the mirrored back surface and are finally reflected towards
the cell in the inner side of the front surface. This reflection is mostly total internal reflection,
except for the central part in which a reflective spot is placed to redirect rays arriving at high
angles; b) represents an RX SMS design. In a) and b), the entering edge rays arrive at both
ends of the receiver and at the receiver; thus obeying the edge-ray theorem.

In particular, among the simple options, it is known from elementary geometry

that, in two dimensions, a parabola focuses all incident rays parallel to its axis onto
its focus. Where the rays will fall when they come from small angles off the axis due
to optical aberrations is also a simple matter to determine. In general, a receiver at
the focus is designed to collect all these rays, and this puts a limit on the maximum
achievable concentration. Lenses may also be used in an elementary way, although
thick lenses have usually (but not always) been avoided in concentrator optics
because of their high cost.
Fresnel lenses are generally used instead of bulk lenses. These are usually
made of highly transparent plastics such as polymethyl methacrylate (PMMA)
that are manufactured by moulding or similar techniques. They consist of a set of
small prisms that deflect incident rays according to Snell’s laws. In thin continuous
lenses, only the tangent orientations (derivatives) at the entering and exiting faces
of the lens are relevant to the optical design. The same is true of Fresnel lenses, so
the derivatives of the continuous lens may be replicated. However, in the case of
continuous lenses there is a link between position and derivative, while in the case
of Fresnel lenses every prism may be designed independently, thus providing an
extra degree of freedom (see Fig. 11.14). In photovoltaic applications, this extra
degree of freedom can be used, for example, to improve the homogeneity of the
irradiance incident on the cell, which as we have said is beneficial in photovoltaic
conversion efficiency. In fact, Fresnel lens designs for solar applications do also
commonly embody the principles of non-imaging optics (Leutz and Suzuki, 2001),
with precursor designs dating back to the early 1980s (Lorenzo and Luque, 1981).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 517

Concentrator Systems 517

Figure 11.14 Schematic of a Fresnel lens showing its facets. The slopes of a plano-convex
lens (above) are projected into a thinner plate.

Figure 11.15 Generations of injection-moulded Fresnel lenses from CPV manufacturer

Daido Steel. Top left: First-generation flat design (400×, 77.3% of peak efficiency, 2001).
Top right: Second-generation half-dome design (400×, 81.5% of peak efficiency, 2002). Bot-
tom left: Third-generation full-dome design made by collapsible moulding die (400×, 85.4%
of peak efficiency). Bottom right: Fourth-generation full-dome design made by collapsible
moulding die (556×, 91% of theoretical efficiency).

For instance, dome- shaped Fresnel lenses as shown in Fig. 11.15 exhibit much
better characteristics.
What we usually desire of the optical design is to achieve the highest possible
concentration with the highest optical efficiency. In addition we wish this efficiency
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 518

518 I. Luque-Heredia & A. Luque

to be maintained even under aiming errors or manufacturing imperfections, i.e.

we want the lens to have the widest possible acceptance angle. Thus we would
be seeking a high CAP concentrator according to the definition of Eq. (11.12).
Moreover, we usually want the concentrator to produce uniform illumination of
the cell, i.e. homogeneous on it and zero outside. The latter is a hard condition that
can be only partly met.
The theorem of conservation of the etendue explained above establishes that
the highest concentration can be achieved only if the cell is isotropically illuminated
by the concentrator. Going back to Eq. (11.11) this concentration is
no
Cg = ≈ 46747n o (11.16)
sin2 φs

where n o is the index of refraction of the medium surrounding the cell and φS is
the Sun’s angular semi-diameter (about 0.265◦ ).
Even theoretically such concentrations cannot be achieved with the lenses or
mirrors we have described above, because they are not able to provide the necessary
isotropic illumination of the receiver. This isotropy can be increased through the
use of two-stage optics, with a primary optics element (POE) that directly receives
sunlight at its entry surface, and a secondary optics element (SOE), that receives the
concentrated light from the primary, and further conducts it into the PV cell sited
at its exit surface. Apart from increasing concentration in the receiver by enabling
light isotropy gain, secondaries are usually designed to increase homogeneity of
the light flux in the PV cell or to increase the acceptance angle of the overall
concentrating system. In the design terms discussed previously, the theoretical
requirement of a secondary is that it is able to cast ray manifolds from a finite
source, the output of the primary, into the receiver.
While not practically feasible as a POE — where small design acceptance
angles are involved — non-imaging CPCs or their derivatives, such as the dielectric-
filled CPCs or trumpet concentrators, can be used as efficient secondary optics as
acceptance angles for this stage are much bigger. Other simpler devices, such as
the truncated cone (for round cells) or truncated pyramid (for square cells), have
also been used. In these devices the inner walls are reflective — either made from
aluminium or covered with some reflective film — and redirect into the cell those
rays that enter and would fall beyond the cell’s perimeter, and increasing concen-
tration and acceptance angle. However, solid dielectric secondaries can achieve
higher concentrations for the same acceptance angles thanks to the n 2o factor in
Eq. (11.16) being higher than unity. Another successful secondary design along
these lines is the dielectric truncated pyramid, where lossless total internal reflec-
tion is used to provide two to four times additional concentration and at the same
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 519

Concentrator Systems 519

time, through the so-called kaleidoscopic effect, produces high homogeneity in the
light flux in the cell.
One of the ways of providing some tolerance in manufacturing, assembly and
Sun-tracking is to design for the collection of rays from a region of the sky larger
than the solar disk, i.e. to increase the optical design’s acceptance angle. The max-
imum concentration achievable by this means can be calculated from Eq. (11.16)
using an angle greater than the Sun’s semi-diameter for φs . Table 11.1 (Luque,
1989) shows the limit of concentration for several families of concentrators for the
case of collecting only the Sun’s semi-diameter, 0.265◦ , and for the more prac-
tical case of an allowance of 1◦ . The concentrator’s primary entry surface (also
called its aperture) is specified through the f-number, which is defined as the ratio
of aperture diameter to the focal distance. This table can be used to estimate the
angular allowance required by a certain design. For instance, the above-mentioned
ENTECH concentrator with arched Fresnel lenses had a geometrical concentration
of 20× (see Table 11.1) and the value in Table 11.1 for an angular tolerance of
1◦ is 26×. Simple proportion gives us an approximate angular allowance for this
concentrator of 1.3◦ . In the table we can observe that linear concentrators achieve
much lower concentrations than point-focus ones. This is because there is no con-
centrating effect in one of the dimensions. For two-dimensional concentrators, the
maximum thermodynamic concentration is only the square root of what appears
in Eq. (11.16). Note that the highest concentration for the Sun’s angular spread is
54× (limited by chromatic aberration) for arched Fresnel lenses, 108× for linear
parabolic mirrors and 376× for point-focus Fresnel lenses. In some cases con-
centrations higher than those shown in Table 11.1 with high efficiency have been
announced; and this is surprising to us as the results in this table are the result of
fundamental limitations.
Table 11.1 also shows how the highest concentrations are achieved through
the integration of secondaries. However, it must be noted that full isotropy in the
receiver is not practical, if incidence angles beyond 65◦ Fresnel reflection on the
cells is regarded as too high. So then again applying the conservation of etendue
principle, the maximum practical gain of a secondary is:

n 2o sin 65◦
C g sec ≤ (11.17)
sin2 αsec

where αsec is the incidence angle at the entry surface of the secondary, which can
be expressed as a function of the f-number as tan αsec = 1/2f. Thus we obtain

C g sec ≤ 0.9n 2o (1 + 4 f 2 ) (11.18)

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 520

520 I. Luque-Heredia & A. Luque

Table 11.1 Maximum concentration for concentrators accepting all incident rays.

Angular Angular
acceptance 0.265◦ acceptance 1◦
Concentrator type C g (Cgsec ) Aperture Cg (C gsec ) Aperture

Circular reflective parabolic dish 11678 f/0.50 821 f/0.50

Parabolic dish with secondary 67666 (16.1) f/1.51 4752 (16.3) f/1.51
(n o = 1.49)
Square flat Fresnel lens–square 376 f/1.35 73 f/1.19
cell
Square flat Fresnel lens with 12356 (42.1) f/2.12 1856 (38.4) f/1.98
axisymmetric secondary
(n o = 1.49)
Linear, flat Fresnel lens 22 f/1.74 10 f/1.38
Linear, arched Fresnel lens 54 26
Linear, parabolic reflector 108 f/0.50 29 f/0.50

Source: Luque (1989). C g is the overall geometrical concentration (primary optics aperture-to-cell area
ratio); Cgsec is the secondary geometrical concentration (aperture of the secondary-to-cell area ratio),
where the refractive index of the secondary is 1.49, which is that of PMMA. Aperture diameter or width
(for the primary) is the (freely selected) focal distance divided by the number in the table (f-number).

so that, for example in the case of the square flat lens with a secondary with
n o = 1.49 in Table 11.1, we find Cg,sec ≤ 37.9, i.e. lower than that achieving max-
imum overall concentration.
Figure 11.16 shows the results of the analysis of the transmission angle curve,
as defined above, of different common secondaries when placed below a Fresnel
lens with f/1 (Victoria et al., 2009). It can be seen that the rotational CPC provides
the highest acceptance angle.
Another important requirement for concentrator optics is homogeneity of illu-
mination. Strongly inhomogeneous illumination produces much higher than aver-
age local concentrations, leading to losses in the fill factor of the cell and reducing
efficiency. One solution is the above-mentioned kaleidoscope SOE, consisting of
a truncated glass pyramid in which the irradiance is expected to be homogenised
by multiple total internal reflection of the entering rays. However, this solution
substantially reduces the acceptance angle. An interesting solution has recently
been presented by Benítez et al. (2010): the Fresnel–Koehler (FK) concentrator,
which uses the principle of designing the SOE to image the POE onto the cell (see
Fig. 11.17). As the POE is homogeneously illuminated by the Sun it produces a
homogeneous light profile. The POE can be accurately imaged onto the cell with
the SOE. This produces a nice squared homogeneous illuminated area that fits with
the square shape of the cell. The FK SOE has the aspect of a dome and presents an
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 521

Concentrator Systems 521

Figure 11.16 Angular transmission curves for the different SOE (secondary optics element)
studied. Lens-to-cell geometrical concentration 1000× and f/1. Reproduced with permission
from Victoria et al. (2009), © OSA.

inactive optical region near the cell that is helpful when sticking the concentrator
to the cell (if this zone is active it gives rise to high losses due to the brimming
glue). As matter of fact in the figure, the FK concentrator is formed of four inde-
pendent POE–SOE couples, all four manufactured together, so this design aspect
is irrelevant for the user. Their acceptance angle is reasonably high, in the range
of 1.1◦ for 1000×.
The optics introduces changes in the spectrum impinging on the cell so that
its optimum design will not be that used today in champion cells designed for a
standard spectrum. Furthermore, chromatic aberrations may cause the illumina-
tion profile at the cell to be different for different spectral ranges. An additional
advantage of the FK concentrator is that it is almost free of chromatic aberrations.
Reflective optics is also used, although much less often. Figure 11.18 shows
the solution of the pioneering CPV company SolFocus, based on non-imaging
optics principles (Winston and Gordon, 2005). Its acceptance angle is rather high
(∼1.4◦ for 500 Suns) but the efficiency is relatively low owing to metal reflec-
tions (governed by the metal absorption) and two Fresnel reflections. Note that the
element over the cell is a homogenising kaleidoscope. Mirrors can bend rays more
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 522

522 I. Luque-Heredia & A. Luque

Figure 11.17 Above: View of the Fresnel–Koehler POE and SOE. This is formed of four
POE in a single Fresnel lens, each one imaged onto the cell by the four SOE elements. The
four SOE elements are cast in the same block. Below: Irradiance distribution on the cell for
the FK concentrator with C g = 625×, f/1, no antireflective (AR) coating on the SOE, when
the Sun is on axis and the solar spectrum is restricted to: a) the top sub-cell range (360–690
nm); and b) the middle sub-cell range (690–900 nm). Reproduced with permission from
Benítez et al. (2010), © OSA.

Figure 11.18 Cassegrainian optics in the SolFocus CPV module. Two reflections, first
in a parabolic dish and then in a hyperbolic mirror with same focus, sends light down to
the receiver block where a TIR kaleidoscope homogenises light reaching the cell below.
Courtesy of SolFocus, Inc.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 523

Concentrator Systems 523

Figure 11.19 Side-cut view of a common concentrator module design, based on a parquet
of Fresnel lenses as primaries.

than lenses; therefore this concentrator is about three times more compact than a
refractive system for the same POE entry aperture.

11.5 Photovoltaic concentration modules

In their most common configuration, cells and optics are put in a box to form a
concentrator module, usually made of metal except for the front cover, the module’s
entry surface, which will be a covering glass. Figure 11.19 shows the main parts
of a generic concentrator module. Primary plastic lenses can then be stuck to the
inner face of the covering glass. Parquets of plastic Fresnel lenses can be supplied
by some specialised manufacturers such as 3M, and directly assembled in place
of a covering glass, but care must be taken over surface scratching, especially in
sandy and windy environments. A common alternative to this possible problem is
to have the Fresnel lenses made of transparent silicone rubber polymerised directly
on the inner face of the covering glass, the so-called silicone-on-glass technique
that was first used in the Ramón Areces concentrator referred to above (Lorenzo
and Sala, 1979).
The cells have to be carefully attached to the supporting structure in such a way
that the wires issuing from the solar cell can be easily connected to external wires
that connect to the next cell. The arrangement must provide electric insulation of all
the cells, which must then be floating with respect to the module box, thus enabling
its grounding for safety purposes, and at the same time it should provide a good
thermal connection to the module’s back plate. To this end, an etched copper layer
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 524

524 I. Luque-Heredia & A. Luque

on an electrical insulating substrate, possibly alumina or AlN with high thermal

conductivity, is used. This must be stuck close to the metal plate that closes the
module at the back. Furthermore, a bypass diode is connected to the cell in an
antiparallel configuration to enable current flow to bypass a particular cell which
may for any reason be shadowed. This ensures that the module will operate even
under partial shading and prevents the appearance of strong reverse bias in the
cells, which is potentially destructive.
For single-substrate structures, removing heat from the cells is not a difficult
problem if the cells are relatively small. In today’s high-concentration modules
cooling is usually passive and cooling fins are seldom used. The amount of heat to
be removed is the fraction of the power incident on the module that is not converted
into electricity. This heat is partially radiated by the two faces of the module and
partially removed by convection, as in the flat module case. The temperature drop
in the thin insulating layer and in the metal layers is relatively small. The main
concern is the removal of heat from the cell itself where the power density is high
(although at 1000 Suns it is still ten times less that in today’s high-power light-
emitting diodes [LEDs]). The typical operating temperature of the cells does not
exceed 80–90◦C, that is, about 50◦ C over the ambient. This is roughly twice the
increase in a flat module.
Cells in modules are usually interconnected in different ways depending on
the cell size, which varies between 1 mm2 and 1 cm2 . Usually, cells are grouped
in parallel (if the cells are small) and then these groups are connected in series.
Greater paralleling in the cell interconnection will result in a higher acceptance
angle.
Something has to be done to keep the cells aligned with the primary optics,
given the difference in the expansion coefficients of glass and most metals. For
example, in order to address this problem, the so-called all-glass technology was
developed by engineers of the Ioffe Institute in St. Petersburg (Andreev et al.,
2003). In this design the module’s front cover was of the silicone-on-glass type
while the back plane was also made of sheet glass. This technology was later
transferred to German CPV start-up Concentrix, which successfully industrialised
it (Fig. 11.20). Beside the problem of optical misalignment, due to differences in
thermal expansion coefficients thermal stress is to be prevented and in this respect
expansion joints should be allowed and sealants must be chosen as sufficiently
elastic to avoid mechanical stress.
Moreover, the inside of the module must remain clean for 20 years with-
out requiring maintenance or any sort of disassembling. This may get somewhat
complicated when some sort of ventilation holes may be needed in the module
enclosure to help dew to dry up in the morning. Different solutions are used by
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 525

Concentrator Systems 525

Figure 11.20 Concentrix’s 500× module, following the all-glass design principle (initially
the side panels were also made of glass and one could look inside the module and check if
it was roughly focused). On the right is an inside view of the module where interconnected
cell substrates can be seen receiving the concentrated sunspot.

every manufacturer and details are usually kept secret. Some have opted for making
a vacuum inside the enclosure, which is hermetically sealed, while others go for
an open approach in which venting holes are only made dustproof through metal
cloth coverings, and rain water is allowed inside provided the cell encapsulation
and wiring is made waterproof. In between these two approaches, manufactur-
ers frequently choose venting solutions based on holes covered with breathable
waterproof Gore-Tex membranes.
Finally, it is interesting to describe, following Sala’s analysis (Sala and Antón,
2011), the impact of cell size on several other module design parameters and what
the range of its optimum value is. For the most usual case of Fresnel primaries,
optical efficiency decreases with f-number and usually this value is set above unity.
Assuming a minimum f-number of one, then we can establish an approximate
relation between cell size and focal distance
AE F2
AC = = (11.19)
Cg Cg

where AC is the cell size; AE is the primary lens surface or, in optical design terms,
the entry aperture surface; F the focal distance; and Cg the geometric concentra-
tion. In other words, the smaller the cell the less the module volume, this meaning a
reduction in the cost of the housing materials and in the module’s requirements for
storage and transportation space, and frequently in the cost of the tracker compo-
nents as well (not only because of a possible reduction in module weight, but also
because the centre of gravity of the array of modules stays closer to the tracking
drive, reducing its load rating). Also, smaller cell size will result in lesser series
resistance and consequently I 2 R ohmic losses, and as will be seen below a smaller
size also enables better heat dissipation. Furthermore, the smaller cells attain a
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 526

526 I. Luque-Heredia & A. Luque

Figure 11.21 Side view of one of the early designs of Daido Steel, a 550× module inte-
grating domed Fresnel primaries where one enclosure side panel has been removed to show
the module’s inside.

higher wafer-filling factor (i.e. the ratio of the total surface of cells made in a wafer
to the total wafer surface). But there are also some disadvantages in going to too-
small cells, for example a lesser fraction of active area per cell, or higher material
losses when dicing the cell wafers, and usually higher complexity in the assembly
process. Figure 11.21 shows a side inner view of one of the early module designs
of the Japanese manufacturer Daido Steel, while Fig. 11.22 shows some examples
of current and recent commercial high concentration PV (HCPV) modules.
Sala’s costing exercise assumed a fixed geometric concentration of 1000× and
20% efficient modules with a design based on a flat Fresnel lens as primary, and
a silo-type secondary. It focuses on two production regimes of 10 MW yr−1 and
30 MW yr−1, which are considered likely in the start-up phase of a CPV production
facility. By exploring cell sizes from 1 mm2 to 100 mm2 the curve in Fig. 11.23
is obtained, which shows a module cost minimum for cells of size between 10–
20 mm2 in area, beyond which module cost increases slowly due to decreasing
efficiency deriving from higher thermal losses.

11.5.1 Cooling of solar cells

Despite the simplicity of cell cooling solutions in most of commercially produced
CPV modules, the analysis of this deserves some attention in the design stage, given
that proper cooling is essential to obtain good system performance. The heat sink
is, in most cases, passive, meaning that the heat is dissipated by natural convection
to the air. This may be achieved by multi-finned structures or, as happens in most
of today’s high-concentration modules, by a single flat aluminium plate placed at
the back of the cells.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 527

Concentrator Systems 527

Figure 11.22 Different commercial CPV modules: a) the US’s SolFocus; b) Spain’s Iso-
fotón; c) Taiwan’s Arima Eco. Courtesy of SolFocus, Inc. Isofotón, SA, Arima Eco Energy
Technologies Corp.

Figure 11.23 Estimated cost estimate of 1000× module based on Fresnel POE and silo
SOE, rated at a DNI of 850 W m−2 in a manufacturing scenario of 30 MW yr−1 . The black
curve assumes fixed efficiency while the red curve takes into account efficiency variations
due to cell size, showing the impact of cooling degradation as the cells grow larger. Courtesy
of Prof. Sala, personal communication.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 528

528 I. Luque-Heredia & A. Luque

The heat is produced at the cell itself, mainly where light absorption occurs.
This tends to be at the cell front. Thus most of the heat goes through the cell volume
to the heat sink or back plate, and then to the interface with air of this metallic part,
where it is transferred to the ambient. Temperature gradients are the driving force
for this heat flow, so some temperature rise over the ambient cannot be avoided.
In many parts of the path, the heat is transferred by conduction so the heat flow
density is proportional to the gradient of the temperature, the thermal conductivity
κ being the constant of proportionality. This law, formally equivalent to Ohm’s law
of electric conduction, justifies the use of an electric equivalence between heat and
current flow. In this equivalence the heat flow plays the role of the electric current
and the temperature that of the electric potential. For a rod of cross-section A and
thickness d the thermal resistance (K W−1 ) is then

Rth = d/κ A (11.20)

Table 11.2 gives the thermal conductivity of some common materials, and
shows the importance of the encapsulation materials. Copper and aluminium are
good conductors but other metals are poorer. However, there are metals (like iron,
κ = 0.5 W cm−1 K−1 ) with a linear expansion coefficient better matched to that
of solar cell semiconductors than Cu, and with a reasonable thermal conductivity
so that they can be used as the stage on which the cell is bonded. Cells are either
soldered to these metal substrates or attached with heat- and electricity-conducting
glues. In order not to shunt the cell, it has to be electrically isolated from the metal
substrate performing as heat sink by an insulator that should also be a good thermal
conductor. In this respect several solutions are applied, usually coming from the
power electronics world, such as the use of IMS (insulated metal substrate) or DCB
(direct bonded copper) substrates that are essentially printed circuit boards (PCBs)

Table 11.2 Thermal conductivity of some common materials.

Thermal Thermal
conductivity/ conductivity/
Material W cm−1 K−1 Material W cm−1 K−1

Si 1.45 Solder 0.5

GaAs 0.8 Air 0.000252
Cu 3.85 Glass 0.007
4Al 2.1 Al2 O3 0.375
Fe 0.5 BeO 2.2
Epoxy (thermal) 0.027 AlN 1.8

Source: Sala, 1989. Reproduced with permission. © IOP.

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 529

Concentrator Systems 529

having a ceramic dielectric layer in between which has good thermal-conducting

properties. Alumina (Al2 O3 ), aluminium nitride (AlN) and beryllium oxide (BeO)
have been used as dielectrics, or in the case of an IMS the dielectric is usually
some heat-conductive epoxy. Air has been included because it can form voids or
gaps between mechanically attached parts, this having a negative impact due to a
serious increase in thermal resistance.
Finally, the heat flow will find an interface with a fluid, usually air. This air
may remove heat either by natural convection (the surrounding air being at rest),
or by forced convection (when the air is in movement, perhaps because of natural
wind). To model convection a heat transfer coefficient h (W cm−2 K−1 ) is used.
The modelling of h is a complex matter. Its value varies from 5×10−4 W cm−2 K−1
for natural convection from a flat plate into air, to (10−15) × 10−4 W cm−2 K−1
for the case of added vertical fins with wind speeds of 1–2 m s−1 .
Cooling by water is much more effective, with h values that may lie in the range
(400−3000) × 10−4 W cm−2 K−1 . Water cooling is often used for experiments,
but tends to be avoided in present-day concentrator designs, as it usually increases
complexity and introduces reliability issues.
In the case of the back plate of the module housing acting as heat sink, solving
the differential heat equation can prove that heat dissipation decreases with distance
to the cell (Napoli et al., 1977; Sala, 1989), to the point that for a certain value of
plate thickness w, there is a maximum effective radius Rmax beyond which more
heat sink material is of no use:

Rmax = 0.66/α = 0.66(wκ/h eq )1/2 (11.21)

where h eq = h + 8εκ R Tamb

3 is the equivalent thermal conductance both due to

convection and radiation at the back plate (radiation through two surfaces and
convection through the outer one) with κ R in the radiation term, the right addend,
being Wien’s constant and ε being the surface emissivity.
The temperature difference between the coldest point of the back plate and
the receiver bottom is
AE
Ti − To = −B(n)ηop ln(α Rc ) (11.22)
πwκ
where B(n) is the direct irradiance, Rc is the cell radius, and as above, A E is the
primary optics entry surface to the concentrator, and ηop is the optical efficiency.
The thermal resistance of the back plate working as heat distributor will be Rth =
ln(α Rc )/πwκ, which decreases with cell size. In general, for a fixed temperature
difference, the smaller the cell radius and the entry surface, the less the plate
thickness that will be required to maintain it. In the case of the back plate acting
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 530

530 I. Luque-Heredia & A. Luque

as the only heat conductor and dissipater, its surface will be the same as that of
the entry surface of the concentrator and in this respect we could impose that
this surface is limited by the maximum effective radius as in Eq. (11.21), so that
A E = π Rc2 and Eq. (11.22) becomes

ln(α Rc )
Ti − To = −0.44B(n)ηop (11.23)
h eq

The cell temperature can be obtained simply by adding to the drop in

Eq. (11.22) or its special case, Eq. (11.23), the contribution of the layers inter-
spersed between the light absorption surface of the cell and the upper surface of
the heat-sink back plate, e.g. the cell itself, the solder, the insulating layer, that act
as conductors with their own specific thermal resistance. According to Eq. (11.20)
its drop can be written as
 
0.44  di Rc2
Ti − To = 2 2 B(n)ηop − ln(α Rc ) (11.24)
α Rc κi κw
i

where κi is the thermal conductivity of the i th layer between cell and back plate
and di its thicknesses. However, for well-manufactured receivers most of the tem-
perature gap occurs in the back plate and the contribution of the upper layers is
small, so that their thermal properties can be to some extent disregarded. The use
of conductive glue instead of solder can create an exception.
A worked example with a real 500× module design allows us to see the
importance of the different elements of the encapsulation. Assume we are using
a triple-junction solar cell of area 0.5 cm2 , 150 µm thick, bonded with 50 µm of
solder to a copper tab 500 µm thick, which is then laminated to a thermal conductive
epoxy layer of 500 µm thickness, and then to 4 mm thick aluminium substrate slab
which is screwed to the module’s back plate, also made of aluminium and 2mm
thick. The surface of the Al slab is 16 cm2 and the Cu tabs and thermally conductive
epoxy printed in its surface cover 40% of this surface. Finally, in-between the Al
slab and the module’s back plate, thermal conductivity is enhanced by a 50 µm
thick layer of thermal grease. Taking h = 5 × 10−4 W cm−2 K−1 for the air–
back-plate interface and the emissivity of Al as 0.1, we find using Eq. (11.21)
that Rmax = 17.15 cm, which is greater than the radius of the 250 cm2 back-plate
dissipating surface, whether this is squared or circular. This means that all of is
being effectively used for heat dissipation. Assuming a direct normal irradiance
(DNI) value of 900 W m−2 , and an optical efficiency of 80%, we can calculate
the temperature difference between ambient temperature and the receiver bottom
using Eq. (11.22) as Ti − To = 56C.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 531

Concentrator Systems 531

Table 11.3 Thermal drops in layers between

the cell and the heat-sink back plate.

Layer Thermal drop (◦ C)

Cell 0.67
Solder 0.36
Copper tab 0.03
Thermal epoxy 2.67
Aluminium slab 0.21
Thermal grease 1.87

Also using the respective thermal conductivities, thicknesses and areas of the
different layers, from the cell to the receiver’s bottom in contact with the back
plate, we would find the temperature jumps shown in Table 11.3.
The total thermal drop in the conducting layers adds up to 5.83◦C, which
added to the difference between the ambient temperature and the receiver makes a
total of 61.83◦ C, which is the temperature above ambient reached by the cell. This
example is good to show how most of thermal drop occurs in the module’s back
plate, performing as a heat sink. Care must be taken with the insulating layers —
the epoxy layer in our example — to ensure the required breakdown voltage that is
commonly specified in the 1–2 kV range is not exceeded. Here we have not taken
into account the convection in the Al substrate slab, which also reduces the overall
temperature drop somewhat.
In summary, cooling the cells is an important problem to tackle when facing
CPV module design, in which a combination of electric insulation and good thermal
contact has to be achieved. But in the end, it is helpful to know that simple passive
cooling means can provide the heat dissipation needed by the cells, even under
solar concentrations of 1000×.

11.6 Tracking systems for photovoltaic concentration

Most CPV systems use only direct solar radiation, and they must therefore perma-
nently track the Sun’s apparent daytime motion, and hence incorporate an automatic
Sun-tracking structure able to mount and position the concentrator optics in such a
way that direct sunlight is always focused on the cells. This Sun tracker is basically
composed of a structure with a sunlight-collecting surface on which to attach con-
centrator modules or systems, which is in some way coupled to a one- or two-axis
mechanical drive, and also some Sun-tracking control system which operates over
the drive axes, and maintains optimum aim of the collecting surface or aperture
towards the Sun.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 532

532 I. Luque-Heredia & A. Luque

Static mounts are only feasible today for low concentration factors below 5. In
the long term, static concentrators with higher ratios making use of luminescence
and photonic crystals might become available. However, all these issues are beyond
the scope of this work.
Line-focus reflective concentrators, such as troughs (the already-mentioned
mentioned EUCLIDES case), require only one-axis tracking to maintain the PV
receiver along the focus line. However, due to the daily variations in the Sun’s
elevation, the incidence of sunlight on the tracker’s aperture is usually somewhat
oblique, reducing the intercepted energy and causing the Sun’s image to move up
and down within the focal axis, producing further losses whenever it oversails the
receiver’s ends. Line-focus refractive concentrators (for example, those based on
linear Fresnel lenses, as in the ENTECH type) experience severe optical aberra-
tions when light incidence is not normal, thus requiring two-axis sun tracking, and
the same happens to most of the point-focus concentrators, excepting some low-
concentration factor devices with sufficient acceptance angle to admit the Sun’s
altitude variations.
Nearly all PV concentrators that are already commercialised or currently under
development use two-axis tracking, the so-called pedestal tracker, with its azimuth-
elevation axes, being the most common configuration, followed by the tilt-roll
tracker operating on the declination–hour angle axes shown in Fig. 11.24.
Strictly speaking, the main commitment to be fulfilled by a CPV Sun tracker is
permanently to align the pointing axis of the supported concentration system with
the local Sun vector, in this way producing maximum power output. The reasons
for the decrease of Sun-tracking performance can be classified into two main types
(Luque-Heredia et al., 2006): (1) those purely related to the precise pointing of the

Figure 11.24 The two most common Sun-tracking axes geometries used in solar trackers:
declination–hour angle (on the left) and azimuth-elevation tracking (on the right).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 533

Concentrator Systems 533

tracker at the Sun; and (2) those which provoke shrinkage of the overall acceptance
angle of the concentrator system, thus indirectly increasing the tracking accuracy
required. Among those related to the tracking accuracy are on the one hand the
exactness of the Sun’s positional coordinates generated by the control system,
expressed in terms of rotation angles of the tracking axes, either by Sun ephemeris-
based computations or derived from the feedback of Sun-pointing sensor readings,
or a combination of both, which are affected by numerous error sources. On the
other hand, there is the precision with which the tracker can be positioned at these
dictated orientations, i.e. the positioning resolution of the tracking drive and its
control system, which essentially depends on the performance of tracking speed
control and on the mechanical backlash introduced by the drive’s gearings.
Regarding acceptance angle losses caused by the tracking system, these are
due to the accuracy, which can be attained in the mounting and alignment of the
concentrator system atop of the tracker. This is basically a design problem having
to do with the fixtures provided for this purpose, their accurate assembly and
the means provided for in-field fine tuning, but also with the mounting protocols
devised to carry out these tasks. Also resulting in acceptance angle cuts is the
stiffness of the tracker, which is to say the deformations allowed in the different
elements of its structure under service conditions.
In a first iteration, characterisation of the service conditions for a CPV tracker
basically consist in determining the CPV array payload (modules’ weight) and fix-
ing the maximum wind load, i.e. wind speed, to be withstood during Sun-tracking
operation without any effect on the concentrator’s power output. To ensure service
under these maximum permanent and variable loads, a maximum flexure defor-
mation should be specified for the tracker structure.
As for the variable loads, the higher the maximum wind speed to be resisted
while maintaining productive operation, the heavier and more expensive is the
tracking structure required to keep deformation within the bounds required for
accurate tracking. This behaviour can be seen in Fig. 11.25 for one of Inspira’s
early tracking designs, a 9 m2 CPV pedestal tracker, in which normalised cost
vs. maximum service wind speed for a maximum 0.1◦ flexure is plotted (Luque-
Heredia et al., 2003).
A cost-effective approach here is to determine the maximum wind load from
the cross-correlation between wind speed and direct radiation, in the location or set
of locations in which the trackers are to be installed. A case example of this type of
analysis is presented in Fig. 11.26, worked out with one year of continuous wind
speed and direct normal irradiation hourly data (assuming a two-axis tracker), for
the Spanish city of Granada (Luque-Heredia et al., 2007). Above this threshold
stiffness specifications do not have to be met, service is not guaranteed, and the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 534

534 I. Luque-Heredia & A. Luque

Figure 11.25 Tracker cost vs. maximum service wind speed for a specific tracker design
dimensioned for three different aperture surfaces and 1000 units/yr production.

Figure 11.26 Wind speed vs. yearly DNI correlation for the determination of optimum
service conditions applied to Granada.

tracker would be better to switch to some low wind profile stow position to decrease
stress and increase operative lifetime. For example, as a general rule of thumb a
13 m s−1 maximum wind load both windward or leeward to the modules in any
of the Sun-tracking orientations of the tracker can be considered a reasonably
conservative value which has been proved to comprise a minimum of 95% of the
direct radiation measured by the 26 weather stations of the SOLMET network
distributed over the contiguous United States (Randall and Grandjean, 1982).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 535

Concentrator Systems 535

Regarding the CPV array payload, and again bearing in mind the targets set
for a possible CPV module design team in order to decrease tracker cost, it would
be worthwhile to decrease the module’s weight as much as possible; the weights of
present module technologies range from 40–140 g W−1. Moreover, module sizing
is also an important issue, and an optimum is to be sought in which the size of
modules does not require redundant framing from the tracker just to hold them,
but on the other hand in which module stiffness doesn’t develop into an excess of
self-weight.
The achievable tracking accuracy and the acceptance angle of the modules
determine the maximum deformation that can occur in the tracker structure without
affecting power output. The higher the tracking accuracy and the wider the nominal
acceptance angle of a module, the bigger the deformation tolerance which can be
set for the structure, and consequently the lower its cost.
Aside from the structural specifications derived from service conditions, basi-
cally determining the tracker’s stiffness, its ultimate limit state and the loads at
which this is reached is to be considered. This is usually a better-paved way in the
sense that there is a good set of international standard building codes. Even if these
do not yet consider the particular case of Sun-trackers, the CPV designer will be
able to interpret and adapt them, dimensioning structural ultimate resistance for
the recommended values of variable loads, namely wind and (depending on the
location) snow and earthquakes. Along with these static loads, dynamic effects
should also be taken into account in what respects the resonant frequencies of the
structure.
For a flat-plate tracker, acceptance angle is meaningless and thus ultimate limit
state dimensioning constitutes only structural specification; stiffness is only an
issue if excess bending can somehow stress or damage the PV modules. However,
in the case of CPV trackers it can easily happen that the stiffness specifications are
more stringent than the ultimate resistance ones, so the fulfilment of the former
implies that the latter are overridden. For this reason a CPV tracker, whichever
its tracking axes configuration, is frequently heavier than its conventional flat-
plate counterpart, and therefore somewhat more expensive. But on the other hand,
thanks to the fact that HCPV modules are already more than double the average PV
efficiency, tracker costs per unit Wp have the potential to outperform their flat-plate
counterparts.
Further exploring the sensitivity of tracker cost to flexure specifications,
Fig. 11.27 shows how for a 50 m2 CPV pedestal tracker designed by Inspira, weight
diminishes when maximum allowed flexure under maximum service conditions is
increased, apparently following a potential law, to the point that below 0.4◦ we
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 536

536 I. Luque-Heredia & A. Luque

Figure 11.27 Steel weight vs. maximum flexure under maximum service conditions for a
50 m2 pedestal CPV tracker design.

start to have weights which correspond to those of flat-plate trackers, flexure is no

longer the dimensioning factor and the ultimate limit state starts to command.

11.6.1 Sun-tracking accuracy

Field experience with photovoltaic concentrators consistently indicates closed-
loop, sensor-based tracking controllers as not being up to the stringent reliability
standards required by low-maintenance plants, which are necessary to sustain the
economic feasibility of these technologies.
Open-loop controllers, based on the computation of Sun ephemeris, imple-
mented as digital embedded systems, appear as a reliable and almost maintenance-
free alternative, able to attain the increasingly high accuracies demanded by the
rising trends in concentration ratio, and offering the possibility of including added
value content, such as remote control and monitoring capabilities, at reasonable cost
when approaching industrial volumes. However, an open-loop controller, even if
operating on the very precise Sun ephemeris equations available to date, is affected,
once connected in the field to its concentrator, by many sources of error which can
degrade its final tracking accuracy well below its ephemeris’ nominal value, even
to the point of missing the concentrator’s specifications.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 537

Concentrator Systems 537

Among these error sources the most significant have a deterministic nature
and mainly result from self-characterisation inaccuracies, i.e. tolerance deviations
appearing in the controller’s default assumptions regarding the tracker’s installa-
tion or manufacturing. Drifts in the internal timing required for the computation of
the Sun ephemeris are the other critical error source, which should be restrained.
Second-order error sources, such as gravitational bending in wide-aperture track-
ers, the effect of mismatch in multi-secondary axis trackers, or even ephemeris
inaccuracies due to the effect of local atmospheric refraction, might have to be
considered as well.
To suppress the tracking errors caused by the above-mentioned sources, feed-
back has to be integrated in the control strategy. This open-loop core strategy
blended with the fed-back closed-loop is sometimes referred to as the hybrid
approach, which in the past has been mainly connected with control theory (Luque-
Heredia et al., 2005a). Related to this approach but relying on the principles of
automated instrumentation calibration, and working in much the same way as the
pointing control of big telescopes, Inspira developed the SunDog sun-tracking
control unit (STCU) (Luque-Heredia et al., 2005b), whose operation is based on
the internal calculation of high-precision solar ephemeris (mean error ≈0.01◦ ).
To convert the solar coordinates generated by the ephemeris into tracking axes
rotation angles with respect to a specific reference orientation, the SunDog STCU
contains a non-linear calibration model characterised by a series of parameters that
determine the transformation in a precise manner. Those parameters are related to:
(1) the relative position of the tracking axes with respect to the Earth’s local sur-
face; (2) the orientation of the axes of rotation references; and (3) the orientation
of the pointing vector of the concentrator system with respect to the tracking axes.
The parameters of the model are least squares fitted internally by means of global
optimisation routines and from precise measurements of the Sun’s position, deter-
mined by maximising the concentrator’s power output. From then on the STCU
operates as an open-loop controller, with proven long-term accuracies of better
than 0.1◦ with 97% probability and below 0.05◦ with 60% probability, as shown in
Fig. 11.28. This was measured with Inspira’s SunSpear tracking accuracy sensor,
which is based on solid-state image sensors (Luque-Heredia et al., 2005c). This
tracking control technique was first used, in a different context, by Arno Penzias,
the discoverer of cosmic background radiation, for pointing early satellite receivers
by fitting an error model with position measurements of a precisely located radio
galaxy.
Also of the utmost importance to achieve Sun-tracking accuracy is the posi-
tioning resolution, i.e. the capacity to position the tracker at whatever command
position is produced by the controller, for example in the case of a SunDog STCU
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 538

538 I. Luque-Heredia & A. Luque

Figure 11.28 Example of daily Sun-tracking error statistics, probability density and dis-
tribution function, with the probability density of collimated sun beam over the tracking
accuracy sensor surface superimposed, showing the 0.1◦ and 0.2◦ tracking error rings.

the position resulting from the Sun ephemeris computation and the later conver-
sion of these Sun coordinates into tracking axes rotation angles by the calibration
model. These high resolutions can usually be achieved by a mixture of versatile
speed control and low backlash gearings, which is in the case of step-based tracking
control schemes will permit a tracking step that is as small as possible, which in
the last instance will serve to decrease tracking error variance.

11.7 High-concentration systems

After the overview of the technical elements behind CPV systems, we now continue
reviewing the timeline of CPV development from the point we left it in Section 11.2.
As we saw there, the origin of concentration as a means to reduce the costs of
photovoltaic systems through a reduction of the necessary cell area can be traced
back to the late 1970s, when the first 1 kW 40× prototype was produced by Sandia
Laboratories in the US. Several other similar early prototypes were developed and
tested in France, Italy and Spain, and since then the number of attempts to develop
CPV systems has been connected to the pace of development of the photovoltaics
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 539

Concentrator Systems 539

industry. Different corporations were created in the United States in the mid-1980s
around this concept, out of which, as we will see below, Amonix and Entech Solar
are still operating in 2013. Also, some important demonstration projects were
built, such as the above-mentioned EUCLIDES plant in Tenerife, which was until
2008 the biggest CPV plant in the world. However, concentration photovoltaic
projects were hardly ever developed further than their prototype stage, even when
they showed promising results in the laboratory, either because they did not reach
a commercialisation stage, or if they did, they could not develop the minimum
economies of scale necessary to demonstrate cost reductions.
The situation has changed radically since those early days of CPV. Because
of the booming growth experienced in the period 2000–2010, the photovoltaics
industry reached sufficient critical mass to boost business ideas based on alternative
technologies such as concentration.
With this fertile scenario, the rapid increase of multijunction cell efficiency
in recent years has played an important role to recall and reassert the promise
of the high cost-reduction potential of concentration systems. Likewise, the fact
that theoretically there is still plenty of room for efficiency increase by means
of increasing the number of junctions, and also in the medium term through the
introduction of other high- efficiency concepts now being researched, has taken
concentration systems, formerly based on more or less customised silicon cells
conceived to operate in the 20–100× range, to the so-called HCPV range. Here the
concentration factors range from 500× to 1000×, which is the level at which these
new high-efficiency cells can show a clear economic advantage. Using a simple
example, at a concentration ratio of 1000, a triple-junction cell cost of 100,000 $/m2
falls to 100 $/m2 , which at efficiency levels of 33% and normal direct irradiation
of 1000 W/m2 , results in cell costs of 0.3 $/W, as opposed to the 2.3 $/W cost of
conventional PV modules in 2011 (here, it must be noted, including also housing
costs).
Concentrated photovoltaics has evolved to become the platform from which
to launch high-efficiency cells onto the market, even though this market entry has
significantly slowed down due to the drastic price reductions in flat-plate silicon
modules that occurred in 2012/13. However, the tendency to achieve higher con-
centration ratios has been confronted with important technical challenges. As we
have seen, the higher the concentration factor, the smaller is the maximum possi-
ble acceptance angle of the optics. At a system level, additional acceptance angle
losses will also have to be subtracted from the nominal acceptance angles of sin-
gle concentrator units, these deriving from misalignments in the mounting of such
sub-systems in modules and subsequently of these on the tracking systems, and
also from structural flexure in the tracking system due to its payload or service
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 540

540 I. Luque-Heredia & A. Luque

wind load. In spite of these difficulties, moving to high concentration factors also
yields some advantages. For example, moving to the use of small cells of at most
1 cm2 size helps to make heat dissipation easier and at some point also facilitates
the adoption of the manufacturing processes used by optoelectronic industries that
have achieved very high learning rates, i.e. extent of cost reductions as cumulative
production increases.
In sum, a quickly growing PV industry, the potential foreseen in high-efficiency
cells, and the knowledge gathered from past CPV prototypes and projects, all
together favoured the development of concentration photovoltaics until the eco-
nomic downturn of 2012–2013. And as a clear signal in this respect, more than
30 new CPV start-ups appeared during the first decade of the twenty-first century.
In Table 11.4 we give a list of some of these start-ups, choosing mostly those that
started some industrial production and have operating installations of a relevant
size, or those that represent an example of a novel design.

11.7.1 CPV’s market entry

As we will see in Section 11.9 when reviewing CPV’s cost projections, the cumu-
lative production figures now required to prove CPV’s superior cost reduction rates
are relatively small compared with the actual volume of the PV market, and CPV is
therefore apparently more affordable. However, CPV’s traditional chicken-and-egg
problem still remains when attempting market entry. This in the sense that cus-
tomers of conventional PV (flat-plate silicon modules), i.e. PV plant developers,
the EPC (engineering, procurement and construction) companies that build these
plants, and most importantly, the banks that finance their construction, view CPV
as a new and still relatively expensive technology that lacks the necessary track
record to enter the mainstream PV market. But on the other hand, CPV needs to
enter the market and accumulate production volume, in order to progress along its
expectedly fast-learning curve and demonstrate its cost-reduction promise.
One of the landmarks in the history of CPV is ISFOC, the Instituto de Sis-
temas Fotovoltaicos de Concentración, which was created in the La Mancha town
of Puertollano, Spain in 2006 by the Consejería de Educación de Castilla La Man-
cha following a project developed and managed by IES-UPM. The main target of
ISFOC was to provide new CPV companies with an instrument to bridge the gap
to the market. Conscious that CPV systems technology was very much ready to
enter the market, but that some entity, not subject to strict market rules, had to help
as introducer, ISFOC is performing as a customer-like research centre that pur-
chases and operates CPV plants, through international calls for tenders. The idea
was to encourage CPV start-ups to get beyond prototypes, and act as suppliers for
real plants of significant size, in order to create a track record audited by ISFOC,
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 541

Concentrator Systems 541

Table 11.4 Main CPV start-ups and industrial projects worldwide.∗

Company Typea Cg kWb Comments

Amonix (USA) Point, Si, Fresnel 500× 950 Redesign underway;

(TP-R), 2-axis, moving to higher
Az&El, pedestal concentrations and
III–V, MJ cells
Arima Eco Point, III–V, Fresnel 476× 400 Subsidiary of Arima
(Taiwan) (TP-R), 2-axis, electronics company
Az&El, pedestal
BSQ Solar Point, III–V, Fresnel 820× 148
(Spain) (TP-TIR), 2-axis,
Az&El, pedestal
Concentrix Point, III–V, Fresnel 530× 1, 622 Acquired by Soitec
(Germany) (−), 2-axis, Az&El, (France) manufacturer
pedestal of electronics
substrates
Redesign underway;
introducing secondary
Daido Steel Point, III–V, Domed 550× 270 Business division within
(Japan) Fresnel (TP-TIR), specialty steel
2-axis, Az&El, manufacturer Daido
pedestal Steel
Redesign underway;
moving to higher
concentrations
Emcore (USA) Point, III–V, Fresnel 520× 830 Redesign underway;
(TP-R), 2-axis, moving to 1090×
Az&El, pedestal concentration and
two-axis tilt-roll tracker
Joint venture Suncore
with Chinese group
San’an Optoelectronics
Entech Solar Linear, Si, Arched 20× 640 Redesign underway;
(USA) Fresnel (-), 2-axis, moving to two-axis
Dec&Ha, tilt–roll pedestal tracker
Greenvolts Point, III–V, Mirror- 625× 1, 000 Redesign underway;
(USA) off axis parabolic, moving to Fresnel +
2-axis, Az&El, secondary and two-axis
turntable tilt-roll tracker
Acquired by Swiss group
ABB
(Continued)
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 542

542 I. Luque-Heredia & A. Luque

Table 11.4 (Continued)

Company Typea Cg kWb Comments

Guascor Fotón Point, Si, Fresnel 400× 12, 550 Subsidiary of engine
(Spain) (TP-R), 2 axis, manufacturer Guascor
Az&El, pedestal Redesign underway;
moving to higher
concentrations and
III–V, MJ cells
Renamed as Foton HC
Systems
Isofotón (Spain) Point, III–V, TIR-R 1000× 400 Business division within
(specific secondary), flat-plate PV
2 axis, Az&El, manufacturer Isofotón
pedestal Redesign underway;
moving to Fresnel +
secondary and lower
concentration
Opel (USA) Point, III–V, Fresnel 500× 470 Sold solar assets to
(TP-TIR), 2-axis, Northern States Metals
Az&El, pedestal in Dec. 2012
Semprius (USA) Point, III–V, Plano 1110× 8 Siemens acquires
convex (glass ball), minority share
2-axis, Az&El,
pedestal
Sharp (Japan) Point, III–V, Fresnel 700× 100
(−), 2-axis, Az&El,
pedestal
Skyline (USA) Linear, III–V, Mirrored 10× 27
Trough, 1-axis,
horizontal, NS
Silicon CPV Point, Si, Parquet of 120× 75 Also developing a III–V,
(UK) prisms, 2-axis, MJ 960× system
Az&El, pedestal
Sol3g (Spain) Point, III–V, Fresnel 476× 1000 Acquired by engineering
(TP-TIR), 2-axis, and construction
Az&El, pedestal company Abengoa
(Spain)
Redesign underway
Solar Systems Point, III–V, Mirrored 500× 1, 200 Acquired by engineering
(Australia) Parabolic dish, company Silex Systems
2-axis, Az&El, (Australia)
pedestal
(Continued)
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 543

Concentrator Systems 543

Table 11.4 (Continued)

Company Typea Cg kWb Comments

SolFocus (USA) Point, III–V, Mirror 650× 4, 010

Cassegrain
(TP-TIR), 2-axis,
Az&El, pedestal
Soliant (USA) Point, III–V, Fresnel 500× 50 Acquired by Emcore
(TP-TIR), 2-axis, (USA)
Dec&Ha, tilt-roll
Suntrix (China) Point III–V, Fresnel 595× 270
(−), 2-axis, Az.&El,
pedestal
Whitfield (UK) Point, Si, Fresnel (−), 70× 10 Redesign underway;
2-axis, Dec&Ha, moving to higher
tilt-roll concentrations and
III–V, MJ cells

a Focus geometry (linear, point); cell material (Si, III–V compound); type of primary optics; type
of secondary optics (TP: Truncated pyramid, TIR: Total internal reflection, R: Refractive); one- or
two-axis tracking; tracking geometry.
b Installed capacity in kW.
*This information was correct at the time of compilation (November 2011).

that could convince conventional customers of the excellence of their performance

and their reliability. At the same time winning bidders would go through a pro-
cess that would help them to refine their manufacturing, logistics, field works and
plant operation, again with the help of a proactive customer willing to obtain and
share the widest possible breadth of knowledge regarding the performance and
operation of CPV plants. It is worth noting that, believing in the greater potential
of high-concentration photovoltaic systems, ISFOC had as one of the eligibility
conditions that bidders’ technologies would implement concentration factors of
over 250×.
Two calls for tenders were completed by ISFOC, resulting in 2.7 MW of CPV
plants. In the first call, the winning contenders were SolFocus (500 kW), Concentrix
(500 kW) and Isofotón (400 kW), and these were in fact the very first large-scale
installations of these three CPV companies, which were completed in 2008 (see
Fig. 11.29). Also awarded with contracts in ISFOC’s second call for tenders were
Arima Eco (300 kW), Emcore (300 kW), Renovalia (300 kW) and Sol3g (400 kW).
Analyses of the first operating periods for these CPV plants were released by
ISFOC, focusing on the performance and operations and maintenance (O&M)
aspects, and also providing first recommendations regarding plant design and sizing
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 544

544 I. Luque-Heredia & A. Luque

Figure 11.29 One of the ISFOC CPV plants at Puertollano, in which three different tech-
nologies can be seen. In the foreground is Soitec (formerly Concentrix), in the background
SolFocus and on the right is Isofotón. Courtesy of ISFOC.

(Rubio et al., 2010a; Gil et al., 2010; Alamillo et al., 2010). Some of these results
will be reviewed in the next section.
One of CPV’s problems has been the wide variety of different designs pro-
posed, as usually happens with any nascent industry (recall the early days of avia-
tion, for example). This lack of standardisation hindered the development of a set
of specific CPV component providers (e.g. trackers, optics, etc.) that could attain
cost reductions through aggregated market volume. However, in recent times, as
can be inferred from the different system descriptions in Table 11.4, CPV system
designs are converging towards a set of common features, firstly confirming the
general trend towards higher concentration ratios, which is in most cases achieved
through the use of Fresnel primary optics plus some kind of secondary, frequently
a simple reflective truncated pyramid. Regarding the tracker, the two-axis azimuth-
elevation pedestal configuration is the most-chosen option, unless when targeting
the rooftop market, in which case the tilt-roll tracker has the advantage of a lower
profile. Without debating the pros and cons of this coalescing standardisation, it did
indeed help the appearance of dedicated suppliers to the CPV industry, whether
start-ups or already-established companies creating product lines specifically to
supply CPV manufacturers. Also a symptom of this standardisation is that design
debates have become more specific. For example, given that a Fresnel primary is
usually chosen, the question now is should it be made in PPMA acrylic plastic
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 545

Concentrator Systems 545

or should it be of the so-called silicone-on-glass (SOG) type, where the Fresnel

profile is stamped in a layer of silicone behind a protecting glass.

11.7.2 Crossing the valley of death

Returning to the manufacturers’ list in Table 11.4, it is interesting to focus our
attention on the first two companies in terms of cumulative installations to date,
and briefly explain some reasons behind their success. Way ahead of all the rest
of CPV manufacturers is Guascor Fotón, a spin-off of Spain’s manufacturer of
diesel and gas engines Guascor. Guascor Fotón arose from a technology transfer
agreement from US veteran CPV manufacturer Amonix. This led to a technology
industrialisation project by Guascor Fotón, which ranged from all the tracker and
module manufacturing and assembly down to the production of its high-efficiency
silicon cells. The first Guascor Fotón 25 kW CPV prototype system was installed in
the test field of IES-UPM, replacing its long-serving ancestor the ‘Ramón Areces’
1 kW system already mentioned. This project was well timed in that Guascor Fotón
had completed its manufacturing lines and was ready to start marketing the product
during the rise of the Spanish PV market, thanks to generous FITs (feed-in tariffs).
Guascor Fotón installed 11.5 MW of its CPV systems, all of them in Spain, between
2007 and 2008, which in 2011 still represented more than 50% of the global
cumulative CPV power installed (see Fig. 11.30). A well-timed and structured
project, a relatively well-proven technology, that of Amonix, even if not the newest
of its time, and most importantly a strongly incentivised market in a high DNI
country, which spurred such a high demand spike of PV as to admit also alternatives
to conventional flat plate as CPV systems — all these contributed to the success
of Guascor Fotón. The Spanish PV market boom (roughly 4 GW installed in two
years) passed, but Guascor Fotón remains, now renamed Foton FC and working
on a new generation of its product, in which silicon cells have been replaced by
multijunction cells and concentration has been driven upwards. A 31.5% module
efficiency had been achieved by the end of 2011.
The second in the list of top CPV installers in Table 11.4 is the Silicon Valley
start-up SolFocus. If the Guascor success was mostly market-driven thanks to the
Spanish national FIT — once successful technology transfer and industrialisation
had been achieved — SolFocus is a good example of the thrust of venture capital
(VC)-backed tech companies in one of the most dynamic entrepreneurial regions in
the world: Silicon Valley. A typical US garage start-up founded in 2005 in Saratoga,
California, by Gary Conley and Steve Horne, SolFocus was set to develop a CPV
system built around a concentrating optics design developed in cooperation with
Roland Winston, one of the inventors of non-imaging optics. By 2006, SolFocus
had received its first $3 million of VC seed funding and moved to the Xerox
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 546

546 I. Luque-Heredia & A. Luque

Figure 11.30 Built in 2008 by Spanish developer Parques Solares de Navarra, the Vil-
lafranca plant with 10 MW of Guascor Fotón CPV systems was the largest CPV installation
in the world until 2012, when the 30 MW Alamosa Solar Plant opened in Colorado. In the
upper-left corner of the picture, the systems in a darker tone are 2 MW of flat PV mounted
on two-axis trackers. Courtesy of Parques Solares de Navarra, SL.

Palo Alto Research Centre (PARC), a high-tech incubator with a distinguished

reputation for its contributions to information technology and hardware systems.
In just one year SolFocus left the PARC incubator and moved to headquarters in
Mountain View and were able to raise $150 million, mostly coming from venture
capital firm New Enterprise Associates (NEA), which will appoint the above-
mentioned Nobel laureate Penzias as chairman of SolFocus’ technical advisory
panel. After having built a few prototypes in the US, it will be through ISFOC that
SolFocus will build their first plant of a significant size, 500 kW distributed in two
locations of La Mancha, in this way quickly moving from product development
to manufacturing. United States VC funding, able to rapidly pour large amounts
of money into technology start-ups, demands in exchange a relatively high and
prompt return on its investment; thus SolFocus had to hurry up the closing of
its product development stage, which in other CPV companies had taken several
years, in order to get to the market as soon as possible. For example, conscious that
tracking technology was a critical element in their CPV system, instead of spending
time in its internal development, SolFocus opted to acquire the Spanish company
Inspira, founded by one of the authors (Luque-Heredia), which had been developing
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 547

Concentrator Systems 547

Figure 11.31 Victor Valley College 1 MW plant built by Solfocus in 2010 was the first one
set by this company in the United States. Courtesy of Solfocus, Inc.

tracker technology specifically designed for CPV’s accuracy specifications since

its participation in the EUCLIDES project.
In 2011 (the time of writing), SolFocus continues to raise VC funding and has
been able to build about 4 MW of CPV power plants (see Fig. 11.31), both in the
US and Europe (Spain, Italy and Greece), in very competitive PV markets against
a background of conventional PV achieving fast and pronounced cost reductions,
and during the world’s recent economic downturn, which has forced the reduction
of PV incentives.1

11.8 Rating and performance

The first issue that appears when considering the performance of today’s CPV
systems is their power rating. Several approaches have been proposed, regarding
both the standard test conditions (STC) used and the measurement procedures.
The STC under which flat PV modules are rated are well known: 1000 W m−2,
◦
25 C cell temperature and AM 1.5 G spectrum. These are not related to real

1 Sadly, SolFocus later fell victim to the fall in price of flat-panel silicon PV and in September 2013
was reportedly in negotiations to sell its assets and intellectual property (IP).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 548

548 I. Luque-Heredia & A. Luque

operation conditions — they are overly optimistic with respect to what is normally
found in real operation — but there is by now sufficient track record for PV plant
planners to know what the energy output will be for an installation rated under
standard conditions in a location with a known irradiance and temperature. There
are also well-accepted formulae that have been developed to transform electrical
output readings — current, voltage and power — from a PV module, made at what-
ever irradiance and cell temperature conditions, into those that would correspond
to STC.
A set of STC must also be established to rate CPV modules or systems. Having
those for flat PV modules as the market’s well-established reference, it seemed
reasonable to lower the irradiance value somewhat, given that CPV modules do not
capture diffuse irradiance as tracking flat-plate PV modules would do. However,
ultimately it is the energy production which is of most interest, and in order to
remain close to the way things are in the flat-plate PV market, it seems a good
approach to determine CPV’s STC as those that will result in a similar energy output
to that of an equivalent non-concentrating PV system, i.e. a two-axis tracking PV
system. This was conceptually the approach taken by Luque in the specifications for
ISFOC’s first call for tenders in 2007, where he estimated the STC to be 850 W m−2
and 60◦ C cell temperature. Obviously this estimate is site-specific, mostly through
the available DNI irradiance and the ambient temperature.
This approach was further analysed by ISFOC (Rubio et al., 2010b), who
showed that at Puertollano — the location of most of ISFOC’s installed CPV
power — STC defined as 850 W m−2 DNI and 60◦ C cell temperature resulted in
yearly cumulative production (kWh/kW) of CPV which was about 5% higher than
that of tracked PV in the same location. On the other hand, were we to choose
STC similar to those of flat-plate PV, i.e. 1000 W m−2 DNI and again 60◦ C cell
temperature, then the yearly cumulative CPV production would be almost 15%
below that of tracked PV. The best match in ISFOC’s analysis, in-between the
cumulative productions of CPV and tracked PV, when varying the STC DNI, is
obtained with 900 W m−2 DNI and 60◦ C cell temperature. Also a good match,
better than 5%, occurs when varying cell temperature instead, for example using
850 W m−2 and 25◦ C cell temperature. While the rationale behind this analysis
and approach to a definition of the CPV rating STC puts tracked PV and CPV
on the same level in energy production terms, some CPV manufacturers have
justifiably opted to raise DNI and lower cell temperature under which they rate
their products, while there is still no generally accepted STC, in order to somewhat
increase the nominal power of their products and therefore lower their /W, which
as has been said is still the market’s main metric. The obvious trade-off is that going
to more favourable STC would lower their kWh/kWp, eventually below those of
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 549

Concentrator Systems 549

tracked PV systems, but initial sales experiences showed this was less important in
a marketing strategy. The issue of the appropriate STC for CPV has been debated
by the International Electrotechnical Commission (IEC) when working on its IEC
62670 standard. Although a good case was made by the IEC TC82 Working Group
7 for 900 W m−2 and 25◦ C cell temperature to be the standard, the IEC has recently
ruled that the STC for CPV will remain the same as those for flat-plate PV as regards
irradiance and cell temperature (1000 W m−2 and 25◦ C) but that the standard solar
spectrum will be AM 1.5 DNI rather than AM 1.5 G (Solar Industry, 2013).

11.8.1 Rating procedures and instruments

Moving now to measurement procedures, it is obviously not practical to rely on
outdoor measurements when rating the module production of a manufacturing line,
and since the early days of the PV industry, solar simulators have been used to sim-
ulate natural sunlight for repeatable and accurate indoor rating of PV modules. The
same is needed for CPV modules and a specific solar simulator has been devel-
oped at IES-UPM for this technology by Domínguez et al. (2008). This simulator
concept is gradually being adopted by many of today’s CPV manufacturers. The
main difference between a conventional solar simulator for PV modules and that
designed for CPV is that the latter should produce collimated sunlight with a sub-
tended angle as close as possible to that of the Sun (±0.275◦ ), in order to reproduce
DNI. This has been achieved in the IES-UPM design by placing a Xenon flash light
at the focus of a large-area parabolic mirror (see Fig. 11.32).

Figure 11.32 On the left is a drawing of the complete solar simulator developed by IES-
UPM for the rating of CPV modules, providing an extended source of collimated light (with
a subtended angle equivalent to the Sun’s), through the reflection of the light of a Xenon
flash lamp in a 2 m accurate parabolic mirror, shown in the picture on the right. Courtesy
of Soldaduras Avanzadas, SL.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 550

550 I. Luque-Heredia & A. Luque

As is well known, one of the properties of a parabola is that every ray coming
from its focus is reflected back parallel to its axis, so that by correctly sizing the
diameter of the light source located at this focal point, the light reflected by the
mirror will have a uniform angular spread. For example, in the currently available
setup, a 2 m diameter parabolic mirror with a 6 m focal distance, at which a 6 cm
diameter bulb is placed, will ideally provide a ±0.275◦ spread beam in the area
covered by the parabolic mirror. The simulator operates as a multi-flash testing
system. The Xenon lamp is triggered to illuminate the CPV module that is biased
at a different voltage at each flash pulse, recording different current–voltage pairs
during the decay of the flash pulse. Each current measurement within a single
flash corresponds to a certain irradiance level, which can be determined through a
calibrated cell, so that at the end of the series of flashes, a family of I –V curves is
obtained at varying concentration factors. A computer controls the whole process
by triggering the lamp, setting the biasing voltage for each flash, and recording the
current–voltage pairs.
One of the main difficulties in developing this type of solar simulator was
to find a 2 m diameter parabolic dish, which due to the long focal distance
required demanded extremely accurate optical quality. This could only be found
at sky-high prices in the astronomical telescope industry. A parabolic profile was
therefore machined out of a bulk aluminium piece, and its mirroring provided by
an aluminium-based reflective coating. Errors in the profile are in the 10 µm range,
which is higher than the tolerance permitted in astronomical telescopes but good
enough for the CPV application. Scattering due to the mirror surface finishing has
the effect of widening the angular spread of the collimated beam up to ±0.43◦ ,
i.e. the angular size of the Sun plus some circumsolar radiation, which is still good
enough for most CPV modules.
The spectrum of the flash lamp varies over time during its discharge. This
can be monitored by means of isotypes, which are triple-junction solar cells with
a single cell electrically contacted. The result is that only one of the curves in
the family described above has the desired spectrum. With faster electronics it is
possible to draw the whole I –V curve in a single flash. Additional software can
provide spectrum-corrected I –V curves at different irradiances.
We will also describe a procedure devised for the power rating of complete
CPV plants, which is the one used by ISFOC in the acceptance tests it carries out
on its contractors’ plants and which constitute a contractual payment milestone
(ISFOC, 2006). First, the plant’s DC power is measured at the ISFOC-specified
STC (850 W m−2 , 60◦ C cell temperature) by tracing a set of I –V curves, using a
capacitive electronic load. This load essentially consists of a discharged capacitor
that is connected to the CPV array, so that it will start to charge itself as it receives the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 551

Concentrator Systems 551

array’s current, thus gradually increasing its voltage and decreasing the current,
sweeping in this way the I –V curve and recording a set of current and voltage
pairs. This type of electronic load was specifically developed at IES-UPM for the
measurement of high-power PV systems. Full 100 kW plant units can be measured
with it, in a rather inexpensive way (Antón, 2004).
At least 15 I –V curves are required by the ISFOC procedure: this repetition is
very important because the tracker is always slightly out of aim and the I –V curve
may vary by about 5% depending on the moment of the measurement. These must
be measured with DNIs over 700 W m−2 and wind speed values below 12 km/h.
Before and after tracing each of these I –V curves, the temperature in the back plate
of a few modules in the plant is measured and the average is recorded, together
with the DNI. Each of these I –V curves will be translated from its registered mea-
surement conditions of average back-plate temperature and DNI to STC, so that
the STC DC power rating of the plant, PDC , is the average of the maximum power
values of all I –V translated curves. The translation equations used by the ISFOC
procedure first calculate the cell temperature through the average temperature mea-
sured in the module’s back plates. A thermal resistance parameter ρT is introduced,
which is defined as dependent on the direct irradiance received by these modules:

Tcell − Tbackplate
ρT = (11.25)
B
where Tcell and Tbackplate are the cell and back-plate temperatures respectively, and
B is the measured DNI. Translation of current–voltage pairs from measurement
conditions to STC is obtained through

BSTC
ISTC = I
B
 
0.0257(TSTC − Tcell ) (I L1 − I )(I L2 − I )(I L3 − I )
VSTC =V +N ln
297 I L1 I L2 I L3


TSTC
+ [N(E g1 + E g2 + E g3 ) − VOC ] 1 − (11.26)
Tcell

where I, V are respectively the measured current and voltage, ISTC , VSTC the
current and voltage measured at STC, BSTC, TSTC the values for STC DNI and
cell temperature (all temperatures in Kelvin), and N is the number of array cells
connected in series. E g1 , E g2 , E g3 are the energy bandgaps, in eV, of each three
junctions in the cell, and I L1 , I L2 , I L3 the short-circuit currents of each three cell
junctions under the measurement conditions. When going through this rating pro-
cedure, ISFOC contractors could themselves provide the currents of the sub-cells
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 552

552 I. Luque-Heredia & A. Luque

(isotypes) measured at AM 1.5 D spectrum and duly validated, or by agreement

with ISFOC at any other common spectrum, and it was allowed that the relative
proportions among these currents would be maintained at any measurement con-
ditions. If the contractor cannot provide these data, then Isc = I L1 = I L2 = I L3 /3
is assumed, and also unless the contractor justifies other values, the bandgap ener-
gies are taken to be E g1 = 1.85 eV, E g2 = 1.42 eV and E g3 = 0.66 eV. These
translation equations, proposed by Luque, are based on the simplification that all
the CPV modules in the plant ultimately behave as its longest string of N cells that
are all considered identical.
Once the DC rating is completed, the AC (alternating current) rating of the
CPV system or plant can be obtained using its energy production, recorded by the
counter at the grid connection point, for a period of at least seven days. During
this period continuous DNI readings are stored together with the back-plate or
heat-sink temperature of a randomly selected set of modules in the plant. Within
this analysis time span, those periods in which the DNI goes below 600 W m−2
will not be considered. For this analysis period, the initially obtained DC power
rating of the plant will be translated to real conditions every five minutes using the
DNI and the averaged module temperature readings and the translation equations,
Eq. (11.26). Integration of these instantaneous power estimates will provide the
expected lossless energy output at the inverter’s input, so that we can then define
F as the ratio of the energy output measured by the counter to the estimated DC
energy output, and obtain the AC power rating, PAC , from the equation

PAC = PDC ηinv F (11.27)

where ηinv is the inverter efficiency.

Other technical specifications have been proposed to measure the power of
a CPV plant, such as the American Standard Test Method for ‘Rating Electrical
Performance of Concentrator Terrestrial Photovoltaic Modules and Systems under
Natural Sunlight’, ASTM E 2527-06.

11.8.2 Field performance of CPV plants

Turning now to the performance of CPV systems, we will again resort to the data
released by ISFOC regarding the operation of its plants (Rubio et al., 2010b).
Considering energy production, after one and a half years of data recording, the
output of the CPV plants in Puertollano is very similar to that of tracked PV
systems, which are also installed at the Puertollano site and provide real production
data for comparison (see Fig. 11.33). This was to be expected given that the CPV
systems are rated at 850 W m−2 DNI and 25◦C cell temperature, which as explained
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 553

Concentrator Systems 553

Figure 11.33 Comparison between cumulative energy of CPV, and tracked and fixed-mount
conventional (non-concentrating) PV systems monitored by ISFOC for 1.5 years. In this
case the CPV rating, 850 W m−2 and 25◦ C cell temperature, produces an energy output
very similar to that of tracked PV and as can also be seen the cumulative DNI evolves very
closely to the cumulative production, implying a performance ratio value for the case of
CPV very close to unity. Reprinted with permission from Rubio et al. (2010b), © AIP.

above was the STC best option found by ISFOC. It is worth noting that this is the
overall production of all the CPV plants at Puertollano, regardless of differences
in their technologies, and without filtering any outage periods due to research and
development (R&D) works, maintenance, etc.
It is also interesting to check energy production on a monthly basis. We can
see in Fig. 11.34 that energy production in Puertollano during the summer months
with the highest DNI is up to 25% better than that of tracked PV and as much as
65% better than fixed-mount PV. The good performance of CPV plants during the
high DNI months is due not only to intrinsic module efficiency increase, but also
to the more fluent operation of the inverters, which under clear skies do not suffer
from intermittent connections and disconnections due to passing clouds. This can
be seen more clearly in Fig. 11.35, where efficiency of CPV and tracked PV plants
are plotted as a function of the DNI: while CPV’s performance increases linearly
with DNI, silicon PV’s efficiency linearly worsens due to its higher temperature
coefficient.
The inverter can also be a source of efficiency losses in the high DNI months
if it is undersized, as can be seen in ISFOC’s analysis of one of its concentrator
systems that had a 6 kW inverter per tracker, which was found to be undersized
during the summer, thus capping system power and therefore efficiency. Thinking
engineering-wise this undersizing should be avoided; however, the sizing of the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 554

554 I. Luque-Heredia & A. Luque

Figure 11.34 Monthly cumulative energy of the same three types of installations as in
Fig. 11.33, showing how CPV outperforms tracked PV in the summer months. Reprinted
with permission from Rubio et al. (2010b), © AIP.

Figure 11.35 Monthly cumulative energy of the same three types of installations as in
Fig. 11.34, showing how CPV outperforms tracked PV in the summer months. Reprinted
with permission from Rubio et al. (2010b), © AIP.

inverter can get to be an important and rather subtle issue in a solar developer’s
financial spreadsheet, as certain market situations have proved to favour the so-
called repowering of PV plants, i.e. inverters being undersized by as much as 30%
(Franco de Saravia and Rossel, 2010). In CPV it has been a common practice
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 555

Concentrator Systems 555

from the beginning to install one inverter per concentrator system of power in
the 5–10 kW range. However, there is no special advantage in doing this (Alamillo
et al., 2010) and it is a fact that centralised inverters — i.e. a single inverter servicing
several concentrator systems of up to 500 kW — are as much as 35% cheaper than
the small 5 kW ones (Knoll, 2010). Some inverter manufacturers have reported that
power vs. operating voltage of concentrator fields, as these grow bigger, present
several maxima, thus requiring MPPT (maximum power point tracking) algorithms
that seek the global maximum. Usual P&O (peek & observe) algorithms will not
then be good enough to lock the global maximum, and voltage sweeps should be
programmed periodically.
Also among the ISFOC’s first reports on its CPV plants were those dealing
with reliability issues (Gil et al., 2010). Classification and statistics of defects,
detected during plant installation, prior to its commissioning and acceptance by
ISFOC, were undertaken for 2 MW of total installed power, distributed in 20 plants
of 100 kW, comprising five different CPV technologies. The mechanical tracker
was reported as the most flaw-prone during commissioning, as can be seen in
Table 11.5.
Note that the mechanical tracker is considered separately from the tracking
controller. As regards the mechanical tracker, ISFOC’s report also distinguishes
between the tracker’s steel structure, the tracking drive, and the foundation, finding
that 76% of mechanical tracker problems correspond to the structure and 23%
to the drive. Problems detected in the tracker structure are mainly due to faulty
transportation, which has either bent the steel profiles or damaged their coating,
with rust points appearing. Concentrated photovoltaics tracker transportation to the
field and assembly and leveling of modules on site are possibly areas with much

Table 11.5 Distribution of installation problems at

ISFOC.
CPV plant element No. of reported flaws (%)

Mechanical Tracker 905 36.1%

AC protections 530 21.1%
Wiring 524 20.9%
DC protections 253 10.1%
Earth connection 171 6.8%
Controller 90 3.6%
Modules 26 1.0%
Inverters 8 0.3%
Total 2507 100%
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 556

556 I. Luque-Heredia & A. Luque

room for improvement. Regarding the drive problems, most of these seemed to be
localised in the elevation axis, usually addressed by all tracker designs using power
jacks. Regarding the second most important source of problems, in the wiring or
electrical protection, these were usually related to defective layout and overlooking
electrical safety codes.
In 2011, ISFOC reported in respect of degradation and O&M issues 2.5 years
after the installation of its oldest plants (Rubio et al., 2011). After a threefold
analysis checking for efficiency reductions, they found no evidence of noticeable
degradation having happened during this time.
To end, we have referred to the work being done towards the CPV rating norm
within IEC, and it must also be mentioned that the first CPV-specific standard
already approved and published is IEC 62108 (IEC, 2007). Essentially this is a set
of accelerated aging tests devised to detect long-term reliability problems in the
CPV modules (Muñoz et al., 2010).

11.9 Cost considerations

Let us now talk about costs. In this respect even if CPV has got beyond its initial
prototyping stage and an early industry has definitely appeared, its costs, and even
its prices, are yet not well known, and manufacturing companies do not readily
publish them except in their commercial quotations. In general, for all CPV man-
ufacturers it is no big secret that their costs continue to be to a greater or lesser
extent above the price of equivalent tracked flat-plate PV. This cannot be surprising
given that flat-plate PV is well ahead of CPV in its learning curve, with Chinese
flat-plate PV manufacturers having achieved yearly manufacturing volumes of
several GW.
Probably more relevant than knowing the spread in the costs of the current early
stage CPV industry is to review some of the published cost projections. Cost anal-
ysis and prognosis is usually a difficult task, highly sensitive to the initial assump-
tions. However, reduction of costs is frequently one of the main targets behind
R&D projects, and cost projections produced in these, without pretending to attain
very high accuracies, do serve as tools to compare different technology alternatives.
Curiously, and possibly a symptom of the coming of age of CPV, the publication of
cost projections, traditionally made in the past by academic researchers involved
in CPV development, is gradually being handed over to the increasing number of
PV industry research firms; in fact, CPV has already given birth to specific pub-
lications of this kind such as the British CPV Today. However, the cost analysis
published by one of the authors a decade ago (Yamaguchi and Luque, 1999) and
summarised in Table 11.6, still seems to retain useful information, and the way in
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 557

Concentrator Systems 557

Table 11.6 HCPV system cost forecast by Yamaguchi and Luque (1999), reprinted with
permission, © IEEE.

Short-term Long-term
1Ja 2J No learning Learning

Substrate wafer ($ per cm2 wafer area) 8.50 8.50 8.50 2.37
Cells ($/cm−2 cell area) 13.4 15.85 15.85 4.43
Cells ($/m−2 module area) 134 159 159 44
Optics, heat sink & assembling ($/m−2 ) 131 131 131 69
Module ($/m−2 ) 265 290 290 113
Cell efficiency (%) 23.1 37 45 45
Module efficiency (%) 19.0 30.5 37.1 37.1
Module ($/W) 1.39 0.95 0.78 0.31
Area-related BOS ($ per m2 aperture 114 114 114 60
area)
Power-related BOS ($/W) 0.22 0.22 0.22 0.12
Plant price ($ per m2 aperture area) 526 589 607 271
Madrid DNI (kWh m−2 year−1 ) 1826 1826 1826 1826
Performance ratio∗ 0.606 0.606 0.606 0.606
Electricity costs in Madrid ($/kWh) 0.186 0.130 0.110 0.050
Electricity costs in EGLb ($/kWh) 0.131 0.091 0.077 0.035

a 1J: single-junction cell; 2J: double-junction cell.

which it was derived is clearer than that of present-day forecasts by market research
firms, having at least a didactic value.
The cost analysis was produced for 1000× concentration modules that used
III–V cells of 1 mm2 surface area and integrated different, highly compact optical
designs produced with the SMS method. This CPV technology was jointly devel-
oped by IES-UPM and Isofotón (Algora et al., 2000). The module design followed
the so-called ‘LED-like approach’ in that the assembly processes proposed resem-
bled those in the optoelectronics industry (Algora et al., 2006). The levelised cost
of electricity (LCOE) for an amortisation period of 30 years was calculated for the
average DNI of Madrid (1,826 kWh m−2 yr−1 ) and also for that of an extremely
good DNI location (EGL in Table 11.6 above) with 2,600 kWh m−2 yr−1 .
Single- and double-junction cells were the state of the art at the time (first and
second column in Table 11.6, with header 1J and 2J) and the authors, assuming
average prices in the wholesale electricity markets of 0.05 $/kWh, and customer
electricity prices of 0.1 $/kWh, concluded that CPV electricity could become cost-
effective for the customer at EGL locations. Overall system prices, including CPV
module and area-related BOS (balance of system) (which were calculated based
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 558

558 I. Luque-Heredia & A. Luque

on the EUCLIDES plant data) envisaged for the 1J and 2J cell CPV systems, were
respectively 1.62 $/W and 1.17 $/W and were calculated for a cumulative produc-
tion of 100 MW, with a 25% mark-up applied on top to account for installation
costs and commercial profit margin. The learning rate, LR, of a certain product
or industry is the cost reduction achieved every time the cumulative production
doubles (Nemet, 2005), and can be calculated from

−b
qt
Ct = C 0 where LR = (1 − 2−b ) (11.28)
q0
with C0 being the known cost of a product at some past time and Ct the projected
cost at some future time, q0 and qt the cumulative productions achieved at these two
times, and b the so-called learning coefficient; the value b = 0.175, appropriate
for the PV industry of the time, was assumed.
The short-term projections when extended into the long term, supposing mod-
ules with four-junction tandem cells (third and fourth columns in Table 11.6) for
which module efficiencies of 37% could be expected, are so reasonably given
that present-day three-junction cells are now achieving module efficiencies in the
32–35% range. One of the authors of this chapter has written a paper in which his
belief that 50% cells will eventually be achieved is substantiated (Luque, 2011).
In the long-term case, Table 11.6 shows two options (third and fourth columns),
firstly with no learning, i.e. no consideration of cost reduction due to production
scaling, and secondly integrating the cost reduction achieved when arriving at a
cumulative production of 1 GW with a 32% learning rate, which is that exhibited
by electronics and optoelectronics. A higher learning rate than that achieved by
conventional PV at that point, was used for this long-term estimate, considering on
the one hand that CPV is much less mature than PV and therefore could achieve
faster cost reductions in its initial industrialisation, and taking the optoelectron-
ics reference as that closer to the LED-like design approach to HCPV modules.
This last long-term forecast, assuming learning, is capable of matching wholesale
electricity prices after a relatively low cumulative volume by today’s standards.
An important positive surprise when looking at this table is the value assumed
for the performance ratio (PR). This is defined as the ratio between the energy
produced and the rated power installed times the ‘equivalent hours’ of operation of
the modules. The latter are the energy falling into a unit module surface divided by
the rating solar irradiance for the modules. In the easy case that the rating irradiance
in the modules is 1 kW m−2 the equivalent hours are exactly the kWh of solar energy
incident on 1 m2 of concentrator aperture. In Table 11.6 this rating is used and the
PR value is very low (0.606) as deduced from that measured in the EUCLIDES
prototype. However, the data registered at ISFOC for three-junction solar cells are
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 559

Concentrator Systems 559

much higher, 0.82 vs. 0.72 for the silicon flat module plants in the area. This better
behaviour is due to the smaller temperature effect in the three-junction solar cells
as compared with the silicon cells and this also explains the low value (0.606) for
concentrated silicon systems. Of course, the silicon tracking system collects global
normal irradiance (GNI) while the CPV system collects DNI, which in this site is
76% of the former. That is why the concentrator system produced only 87% of the
flat-panel’s output if the rating is the same for both (1 kW m−2 and 25◦ C) but the
ratio is not that of DNI/GNI but much better. With the measured PR (0.82) instead
of the one derived from Si cells experiments (0.606) the cost of CPV electricity is
substantially reduced.

11.9.1 The value of the concentration factor

Another interesting cost analysis exercise was undertaken by Algora (2007), in
which the impact of the concentration factor and cell efficiency on CPV sys-
tem costs was studied. The assumptions were very much the same of those
of Yamaguchi and Luque above, as this study was also undertaken with the
cooperation of IES-UPM and Isofotón in connection with the development of a
1000× concentration technology. Figure 11.36 shows the complete CPV system
price at plant level as a function of cell efficiency, with the concentration factor
as a variable parameter. The results of this analysis are presented for a cumulative
production of 10 MW, and again estimating cost reduction after cumulative pro-
duction of 1 GW with the same 32% learning rate as above. It can be seen how

Figure 11.36 CPV system cost estimated by Algora (2007) as a function of the concentra-
tion factor and the cell efficiency for two different scenarios of cumulative production.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 560

560 I. Luque-Heredia & A. Luque

Figure 11.37 Breakdown of costs of a triple-junction concentrator panel after 10 MW of

learning curve. Left: for 40% cells at 250 Suns, leading to an installed cost of 3.8 /W.
Right: for 33% cells at 1000 Suns, leading to an installed cost of 2.4 /W. Reprinted with
permission from Algora et al. (2008).

for example the estimated final price for an installation operating at 400× is about
1 /W higher than if the installation operates at 100 Suns, almost independent of
the cell efficiency after 10 MW of cumulative production.
Also in this analysis, to further stress the cost-reduction leverage of increasing
the concentration factor, a system cost breakdown is provided for two different
cases. The pie chart on the right in Fig. 11.37 represents the contribution of the
different CPV plant cost elements, when using 40% three-junction cells operating
at 250× (after 10 MW of cumulative production), resulting in a final cost of 3.8
/W at plant level, mostly due to the big impact (66%) of the expensive solar cells,
while on the left cheaper 33% two-junction solar cells operating at 1000× achieve,
also after 10 MW of cumulative production, a much lower CPV plant cost of 2.4
/W. In this way, Algora (2008) shows the importance of concentration factor as
a CPV system cost reduction factor, in particular in the starting phase when mass
production costs have not yet been achieved.

11.9.2 Metrics and cost breakdown

The two most common metrics that help to assess the financial feasibility of a PV
project, which we have already seen used in the previous cost analyses, are on the
one hand the installed cost in /W and on the other the LCOE in /kWh. The /W
is a relatively simple metric that is based solely on the production and installation
costs of the system, divided by its power rated under some consensus irradiance and
temperature conditions, usually STC. This is the main metric used by the PV market
to date, basically because it can be easily verified and it will ultimately appear in
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 561

Concentrator Systems 561

the invoice issued by the PV manufacturer to the plant developer or the appointed
EPC firm — i.e. it is what you will pay for the PV system if you want to build a
plant, whether to operate it for its whole lifetime or to resell just after it is built.
The LCOE is defined as the cost of a unit of electricity in current monetary
units. It incorporates the /W of installed costs described above, and adds the
operation and maintenance costs over the life cycle of the plant or system, also
known as the total life-cycle costs (TLCC). These costs are divided by the energy
generated, which is a function of the DNI irradiation in the plant’s location, so that
the overall expression is
TLCC
LCOE = N Qn
(11.29)
n=1 (1+d)n

where Q n is the energy output in year n, N is the number of years of the analysis
period, usually the total number of productive years, and d is the discount rate used
to discount back future annual energy output to a present value. The discount rate
is the rate of return that could be earned on an investment in the financial markets
with similar risk.
The LCOE should be the most relevant metric for PV plant operators, for
example the so-called IPPs (Independent Power Producers) that enter into PPAs
(power purchase agreements) with electric utilities. This metric also has strong
advocates among present CPV manufacturers, who are usually elusive in public
regarding their installed costs ( /W), presumably because they yet do not bear
favourable comparison with those of conventional flat-plate PV, but frequently
present aggressive LCOE estimates and projections, and propose that LCOE is
standardised as the main commercial metric, instead of the /W of installed costs.
However, while it makes sense to base the financial assessment of investments
in PV installations on their LCOE values, in general the PV market does not yet
permit CPV manufacturers to base their marketing strategies on LCOE values, for
two main reasons. The first is that there is still no confidence in the O&M cost
figures and the long-term predicted power outputs of a new technology such as
CPV, but it is these figures that are used to compute the LCOE; this problem is
very much linked to that already mentioned: CPV’s lack of track record. In other
words, potential CPV customers frequently prefer to know about installed system
prices in order to easily compare with other PV technologies, and afterwards, if
they are in for a long-term investment, they try and make the LCOE calculation by
themselves with all sorts of safety margins.
The second reason is that frequently, where PV plants are financed by banks
through project finance, as happens in the FIT-based markets in Europe, there is
only a small set of strong EPC companies that are considered eligible to receive
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 562

562 I. Luque-Heredia & A. Luque

funding (i.e. bankable) to build any plant. The EPC role will be to build a turnkey
plant and hand it over to the developer — the one who has received the fund-
ing — for its operation, and therefore EPC’s involvement in the plant ends after
two or three years, and EPC companies are not too much concerned about met-
rics such as LCOE to assess the long-term profitability of the plant; their main
target is to maximise profit in the plant construction (within quality margins)
and for this purpose it is /W of systems bought which is really of interest
to them.
To get a feeling of HCPV costs in recent years, it is helpful to refer to mar-
ket research done by specialised firms such as the above-mentioned CPV Today
(Extance and Márquez, 2010). In this report, the main HCPV industry players
(defining HCPV as geometric concentrations above 300×) were surveyed to deter-
mine current and projected installed costs, as well as the breakdown of costs among
system components. As can be seen in Table 11.7, the surveyed HCPV companies
were grouped into three categories, according to their installed costs, and averages
were given — to keep each company’s data anonymous — for each group. The
lowest installed cost group offered quite a low average, falling in the range of
flat PV modules. However, the authors noted that this group included a set of
companies ranging from recent HCPV entrants who claim new approaches which
can attain pronounced cost reductions (but which remain to be proven) to the most
experienced HCPV vendors, who may be attaining the production volumes and
experience necessary to bring costs down, but may be selling at a loss to enter the
PV market and develop volume.
Probably more realistic is to take the survey’s overall average of 3.7 /W,
which can be compared with the turnkey price of a tracked PV system which in
summer 2011 could be assumed as around 2.8 /W.
CPV Today also presented the averaged cost breakdown for HCPV systems,
as shown in Fig. 11.38a, derived from their survey. The assembly costs in the pie

Table 11.7 Costs of HCPV (over 300×) according to CPV Today’s market survey
(Extance and Márquez, 2010).

Projected installed
Installed cost 2010 cost 2015
Segment ( /W, arithmetic mean) ( /W, arithmetic mean)

HCPV overall 3.68 1.83

HCPV high installed cost tier 5.43 2.60
HCPV mid installed cost tier 3.41 1.92
HCPV low installed cost tier 2.28 1.17
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 563

Concentrator Systems 563

Figure 11.38 Two different cost breakdowns for CPV. a) An average one produced by CPV
Today after polling a wide set of CPV manufacturers; b) that presented by Guascor Fotón
derived from their experience for a 5 MW CPV plant with their technology. Reprinted with
permission from Extance and Márquez (2010), © FC Business Intelligence. Courtesy of
Guascor Fotón, SA.

chart were those related to the module integrating cells and optics. Figure 11.38b
shows the CPV cost breakdown of Guascor’s technology (Aurtenetxe, 2010) for a
5 MW PV plant, this company being the only that had built plants of this size at
that time.

11.10 Conclusions
By the end of 2011, when this chapter was written, CPV technology had entered
a phase of incipient commercialisation. Dozens of companies were involved in
its development and the most promising of them, in our opinion, were the HCPV
technologies, fuelled by the fantastic efficiency achievements of MJ solar cells.
Most companies currently using high-efficiency silicon technology were moving
to adapt MJ cells to their products. The rapid fall in flat-plate silicon modules in
2012 and 2013 has affected the commercial development of CPV, as it has for some
of the other PV technologies described in this book.
It is predicted that the efficiency of MJ cells will continue to increase from their
present 44% towards the 50% efficiency to be achieved with five or six junctions
or perhaps with third-generation approaches, such as the intermediate-band solar
cell (Luque and Martí, 1997).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 564

564 I. Luque-Heredia & A. Luque

Trends towards high concentration are underway from the 500 Suns or less of
many present-day solutions towards the 1000 Suns or more to be expected in the
near future. This increase will have a very important benefit to the cost of CPV
systems in the short term, while the production volume of MJ cells is too small to
provoke a reduction-by-experience of the cell cost.
Concentrator optics is well understood today and the main challenges are
solved at least theoretically: good efficiency, large angular aperture and good illu-
mination homogeneity. Challenges to come are maintaining these properties with
the more demanding 1000× technology under way. In spite of these challenge,
improvements are to be expected that will lead to practical optical efficiencies of
90% as compared with today’s typical values of 80%.
Simple solutions suffice for heat management. Passive cooling without fins
is generally adopted and, despite the high concentration, over 1000 Suns, the cell
temperature is usually in the range of 80–90◦C, which high-performance cells can
easily cope with (Vázquez et al., 2007; Núñez et al., 2010).
Field experience is still small, but the ∼10 MW commercial plant in Vil-
lafranca, (Spain), the largest in the world today (still with Si cells), and the 30
MW Alamosa Solar Plant in Colorado (with MJ cells) show that the time of no
more than prototypes has gone. The establishment of the Institute for Photovoltaic
Concentration Systems (ISFOC) in Spain by 2006, and the 3 MW international
call-for-tenders launched there, have been important steps to help companies to
develop MJ cell technology. Module manufacturing has matured, and from the
∼20% efficiency at the module level by about 2008, when the plants for ISFOC
were delivered, we are already at over 30% module efficiency at the time of writing
(summer 2011). A module efficiency of 45% is foreseen (Luque, 2011).
Trackers are an important element of the systems cost, constituting maybe
a third of the total. Therefore this has been carefully studied and the conditions
for producing them at a cost per unit similar to those used for unconcentrated
arrays with tracking are clear: excellent control and mechanisms and wide-aperture
optics. Moreover, the high efficiency of MJ technology reduces by a factor of two
to three the cost per watt of the tracker as compared with that for unconcentrated
arrays.
Very good news is the high performance ratio of the MJ concentrator cells,
over 82% as compared to 72% for unconcentrated silicon cells and about 60% for
Si concentrated cells. This is attributable to the better temperature performance
of MJ solar cells. This may have a very important influence on the LCOE for
concentrators.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 565

Concentrator Systems 565

Ultimately CPV will develop if the prices are below those for unconcentrated
approaches. For this they must climb the barrier of the low-cost conventional uncon-
centrated photovoltaics, fuelled by 50 GW of manufacturing experience. Yet aca-
demic studies of CPV in the late 1990s predicted very low costs in the long term,
comparable or below most electricity sources, and by looking at these studies again
today, we think these goals can be achieved, if the above-mentioned barrier can
be overcome. If this is the case, CPV would become the leading technology for
large PV plants, which in its turn is the largest and fastest growing PV application
today.

References
Alamillo C., de la Rubia O., Gil E., Hipólito A., Martín A., Rubio F., Pachón J. L. and
Banda P. (2010), ‘Analysis of inverter configuration on CPV Plants’, Proc. 25th Euro-
pean Photovoltaic Solar Energy Conf., Valencia, pp. 4733–4736.
Alarcón J., Bassy A., Blazquez L., Luque A., Sala G. and Lorenzo E. (1982), ‘Central
fotovoltaica flotante’, Mundo Electrónico 117, 95–99.
Algora C. (2007), ‘Very-high-concentration challenges of III–V multijunction solar cells’,
in Luque A. and Andreev V., Concentrator Photovoltaics, Springer Verlag, Berlin,
pp. 89–112.
Algora C., Miñano J. C., Benítez P., Rey-Stolle I., Álvarez J. L., Díaz V., Hernández M.,
Ortiz E., Muñoz F., Peña R., Mohedano R., Luque A., Smekens G., de Villers T.,
Andreev V., Khostikov V., Rumiantsev V., Schvartz H., Nather H., Viehmann K. and
Saveliev S. (2000), ‘Ultra compact high flux GaAs cell photovoltaic concentrator’,
Proc. 16th European Photovoltaic Solar Energy Conf., Glasgow, pp. 2241–2244.
Algora C., Ortiz E., Díaz V., Peña R., Andreev V. M., Khvostikov V. P. and Rumyantsev R. D.
(2001), ‘A GaAs solar cell with an efficiency of 26.2% at 1000 suns and 25.0% at 2000
suns’, IEEE Trans. Electron Dev. 48, 840–844.
Algora C., Rey-Stolle I., Galiana B., González J. R., Baudrit M. and García I. (2006), ‘Strate-
gic options for a LED-like Approach in III–V concentrator photovoltaics’, Conf. Record
4th World Conf. Photovoltaic Energy Conversion, Waikoloa, Hawaii, pp. 741–744.
Algora C., Rey-Stolle I., García I., Galiana B., Baudrit M., Espinet P., Barrigón E., Datas
A., González J. R. and Bautista J. (2008), ‘A dual junction solar cell with an efficiency
of 32.6% at 1000× and 31.0% at 3000× ’, Proc. 5th Int. Conf. on Solar Concentrators,
Palm Desert, CA.
Andreev V., Ionova E., Rumyantsev V., Sadchikov N. and Shvarts M. (2003), ‘Concentrator
PV modules of ‘all-glass’ design with modified structure’, Conf. Record 3rd World
Conf. Photovoltaic Energy Conversion, Osaka, pp. 803–876.
Antón I (2004), ‘Métodos y equipos para la caracterización de sistemas fotovoltaicos de
concentración’, PhD Thesis, Universidad Politécnica de Madrid.
Araújo G. L. and Martí A. (1994), ‘Absolute limiting efficiencies for photovoltaic energy
conversion’, Solar Energy Mater. Solar Cells 90, 1068–1088.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 566

566 I. Luque-Heredia & A. Luque

Aurtenetxe A. (2010), ‘Current and future HCPV competitiveness in utility scale installa-
tions’, 3rd Concentrated Photovoltaics Summit, Sevilla, 18 November.
Benítez P., Miñano J. C., Zamora P., Mohedano R., Cvetkovic A., Buljan M., Chaves J. and
Hernández M. (2010), ‘High performance Fresnel-based photovoltaic concentrator’,
Optics Express 18, A25–A40.
Bruton T. M., Heasman K. C., Nagle. J. P., Cunningham D. W. and Mason N. B. (1994),
‘Large area high efficiency silicon solar cells made by laser grooved buried grid
process’, Proc. 12th European Photovoltaic Solar Energy Conf., Amsterdam. H. S.
Stephens & Associates, Bedford, pp. 761–765.
Burgess E. L. and Pritchard D. A. (1978), ‘Performance of a one kilowatt concentrator photo-
voltaic array utilizing active cooling’, Conf. Record 13th IEEE Photovoltaic Specialists
Conf., Washington D.C. IEEE Press, Piscataway, NJ, pp. 1121–1124.
Domínguez C., Antón I. and Sala G. (2008), ‘Solar simulator for concentrator photovoltaic
systems’, Optics Express 19, 14894–14901.
Extance A. and Márquez C. (2010), The Concentrated Photovoltaics Industry Report 2010,
FC Business Intelligence, London.
Franco de Saravia C. and Rossel A. D. (2010), ‘Power Boost’, Photon International 12,
44–52.
Friedman D. J., Olson J. M. and Kurtz S. (2011), ‘High efficiency III–V multijunction solar
cells’, in Luque A. and Hegedus S. (eds.), Handbook of Photovoltaic Science and
Engineering, John Wiley & Sons, Chichester, UK.
Garboushian V., Roubideaux D. and Yoon S. (1996), ‘An evaluation of integrated high-
concentration photovoltaics for large-scale grid connected applications’, Conf. Record
25th IEEE Photovoltaic Specialists Conf., Washington D.C. IEEE Press, Piscataway,
NJ, pp. 1373–1376.
Gil E., Sánchez D., de la Rubia O., Alamillo C., Rubio F. and Banda P. (2010). ‘Field
technical inspection of CPV power plants for the 25th EU PVSEC/ WCPEC-5’, Conf.
Record 25th European Photovoltaic Solar Energy Conf., Valencia, pp. 4483–4486.
Green M. A., Emery K., King D. L., Hisikawa Y. and Warta W. (2010), ‘Solar efficiency
tables (version 36)’, Progr. Photovoltaics 18, 346–352.
Guter W., Schöne J., Philipps S. P., Steiner M., Siefer G., Wekkeli A., Welser E., Oliva E.,
Bett A. W. and Dimroth F. (2009), ‘Current-matched triple-junction solar cell reach-
ing 41.1% conversion efficiency under concentrated sunlight’, Appl. Phys. Lett. 94,
223504.
IEC (2007), IEC 62108: 2007 Concentrator photovoltaic (CPV) modules and assemblies —
design qualification and type approval, International Electrotechnical Commission,
Geneva, Switzerland.
ISFOC (2006), ‘Specifications of general conditions for the call for tenders for con-
centration photovoltaic solar plants for the Institute of Concentration Photovoltaic
Systems’, Instituto de Sistemas Fotovoltaicos de Concentracion, Puertollano, Spain,
pp. 34–39.
King R. R., Boca A., Hong W., Liu X.-Q., Bhusari D., Larrabee D., Edmondson K. M.,
Law D. C., Fetzer C. M., Mesropian S. and Karam N. H. (2009), Proc. 13th European
Photovoltaic Solar Energy Conf., Hamburg, pp. 55–61.
Knoll B. (2010), ‘Still falling: inverter prices in the German spot market continue to get
cheaper’, Photon International 12, 108–109.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 567

Concentrator Systems 567

Kolodinski S., Werner J. H., Wittchen T. and Queisser H. J. (1993), ‘Quantum efficiencies
exceeding unity due to impact ionization in silicon solar cells’, Appl. Phys. Lett. 63,
2405–2407.
Landsberg P. T. and Tonge G. (1980), ‘Thermodynamic energy conversion efficiencies’,
J. Appl. Phys. 51, R1–R20.
Lewis N. S., Crabtree G., Nozik A. J., Wasielewski M. R. and Alivisatos P. (2005), ‘Basic
research needs for solar energy utilization’, Office of Science — US Department of
Energy, http://science.energy.gov/∼/media/bes/pdf/reports/files/seu_rpt.pdf.
Leutz R. and Suzuki A. (2001), Non Imaging Fresnel Lenses, Springer Verlag,
Berlin/Heidelberg.
Lorenzo E. and Sala G. (1979), ‘Hybrid silicone–glass Fresnel lens as a concentrator for
photovoltaic applications’, in Sun II: Proc. Silver Jubilee Congress, Pergamon Press,
Atlanta, GA, pp. 536–539.
Lorenzo E. and Luque A. (1981) ‘Fresnel lens analysis for solar energy applications’, Appl.
Optics 20, 2941–2945.
Luque A. (1980), ‘Quasi-optimum pseudo-lambertian reflecting concentrator: an analysis’,
Appl. Optics 19, 2398–2402.
Luque A. (1986), ‘Non-imaging optics in photovoltaic concentration’, Phys. in Technology
17, 118–124.
Luque A. (1989, ed.), Solar Cells and Optics for Photovoltaic Concentration, Adam Hilger,
Bristol.
Luque A. (2011), ‘Will we exceed 50% efficiency in photovoltaics?’, J. Appl. Phys. 110,
031301–031301-19.
Luque A., Gidon P., Pirot M., Anton I., Caballero L. J., Tobias I., Canizo C. del and
Jausseaud C. (2004a), ‘Performance of front contact silicon solar cells under con-
centration’, Progr. Photovoltaics 12, 517–528.
Luque A. and Martí A. (1997), ‘Increasing the efficiency of ideal solar cells by photon
induced transitions at intermediate levels’, Phys. Rev. Lett. 78, 5014–5017.
Luque A. and Martí A. (2011), ‘Theoretical limits of photovoltaic conversion and new
generation solar cell’, in Luque A. and Hegedus S. (eds.), Handbook of Photovoltaic
Science and Engineering, John Wiley & Sons, Chichester, UK.
Luque A. and Martí A. (2012), ‘Understanding intermediate-band solar cells’, Nature Pho-
tonics 6, 146–152.
Luque A., Sala G., Arboiro J. C., Bruton T., Cunningham D. and Mason N. (1997), ‘Some
results of the EUCLIDES photovoltaic concentrator prototype’, Progr. Photovoltaics
5, 195–212.
Luque A., Tobias I., Gidon P., Pirot M., Canizo C. del, Anton I. and Jausseaud C. (2004b),
‘Two-dimensional modeling of front contact silicon solar cells’, Progr. Photovoltaics
12, 503–516.
Luque-Heredia I., Cervantes R. and Quéméré G. (2005c), ‘A sun tracking error monitor
for photovoltaic concentrators’, Conf. Record 4th World Conf. on Photovoltaic Energy
Conversion, Hawaii, 706–709.
Luque-Heredia I., Martín C., Mañanes M. T., Moreno J. M., Auger J. L., Bodin V.,
Alonso J., Díaz V. and Sala G. (2003), ‘A subdegree precision sun tracker for 1000×
micro-concentrator modules’, Conf. Record 3rd World Conf. on Photovoltaic Energy
Conversion, Osaka, Japan, pp. 857–860.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 568

568 I. Luque-Heredia & A. Luque

Luque-Heredia I., Moreno J. M., Magalhães P. H., Cervantes R., Quéméré G. and Laurent O.
(2007), ‘Inspira’s CPV sun tracking’, in Luque A. and Andreev V. (eds), Concentrator
Photovoltaics, Springer Verlag, Berlin, pp. 221–252.
Luque-Heredia I., Moreno J. M., Quéméré G., Cervantes R. and Magalhães P. H. (2005a),
‘CPV Sun tracking at Inspira’, Conf. Record 3rd Int. Conf. on Solar Electric Concen-
trators for the Production of Electricity or Hydrogen, Scottsdale, AZ.
Luque-Heredia I., Moreno J. M., Quéméré G., Cervantes R. and Magalhães P. H. (2005b),
‘SunDog STCU TM generic sun tracking unit for concentration technologies’, Conf.
Record 20th European Photovoltaic Solar Energy Conf., Barcelona, pp. 2047–2050.
Luque-Heredia I., Quéméré G., Magalhães P. H., Fraile de Lerma A., Hermanns L., de
Alarcón E. and Luque A. (2006), ‘Modelling structural flexure effects in CPV sun
trackers’, Conf. Record 21st European Photovoltaic Solar Energy Conf. and Exhibition,
Dresden, Germany, pp. 2105–2110.
Miñano J. C. (1985a), ‘Two-dimensional non-imaging concentrators with inhomogeneous
media: a new look’, J. Opt. Soc. Amer. A 2, 1826–1831.
Miñano J. C. (1985b), ‘Design of three-dimensional optical concentrators with inhomoge-
neous media’, J. Opt. Soc. Amer. A 3, 1345–1353.
Miñano J. C. (1985c), ‘New family of 2-D non imaging concentrators: The compound
triangular concentrators’, Appl. Optics 24, 3872–3876.
Miñano J. C. and González J. C. (1992), ‘New method of designing of non-imaging con-
centrators’, Appl. Optics 31, 3051–3060.
Miñano J. C., González J. C. and Benítez P. (1995), ‘RXI: a high gain, compact, non-imaging
concentrator’, Appl. Optics 34, 7850–7856.
Muñoz E., Vidal P. G., Nofuentes G., Hontoria L., Pérez P., Terrados J., Almonacid G. and
Aguilera J. (2010), ‘CPV standardization: an overview’, Renewable and Sustainable
Energy Rev. 14, 518–523.
Napoli L. S., Swartz G. A., Liu S. G., Klein N., Fairbanks D. and Tamatus D. (1977), ‘High
level concentration of sunlight on silicon solar cells’, RCA Review 38, 76–108.
Nemet G. F. (2005), ‘Beyond the learning curve: factors influencing cost reductions in
photovoltaics’, Energy Policy 34, 3218–3232.
Ning X., Winston R. and O’Gallagher J. (1987), ‘Dielectric totally internally reflecting
concentrators’, Appl. Optics 26, 300–305.
Núñez N., Vázquez M., Gonzalez J. R., Algora C. and Espinet P. (2010). ‘Novel accelerated
testing method for III–V concentrator solar cells’, Microelectronics Reliability 50,
1880–1883.
O’Neill M., McDanal A., Walters R. and Perry J. (1991), ‘Recent development in lin-
ear Fresnel lens concentrator technology including the 300 kW 3M/Austin Sys-
tem and the 20 kW PVUSA system and the concentrator initiative’, Conf. Record
22nd. IEEE Photovoltaic Specialists Conf., Las Vegas. IEEE Press, Piscataway, NJ,
pp. 523–528.
Optics.org (2013), ‘Spectrolab beats unconcentrated PV efficiency record’, http://optics.
org/news/4/4/16, 15 April.
Randall D. E. and Grandjean N. R. (1982), ‘Correlation of insolation and wind data for
SOLMET stations’, SAND82-0094, Sandia National Laboratories.
Ross R. T. and Nozik A. J. (1982), ‘Efficiency of hot-carrier solar energy converters’, J. Appl.
Phys. 53, 3813–3818.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 569

Concentrator Systems 569

Rubio F., Martínez M., Hipólito A., Martín A. and Banda P. (2010a), ‘Status of CPV technol-
ogy’, Proc. 25th European Photovoltaic Solar Energy Conf., Valencia, pp. 1008–1011.
Rubio F., Martínez M., Martín A., Hipólito A. and Banda, P. (2010b), ‘Evaluation parameters
for CPV production’, 6th Int. Conf. on Concentrating Photovoltaic Systems, American
Institute of Physics, Freiburg, Germany, pp. 252–255.
Rubio F., Martínez M., Sánchez D., Aranda R. (2011), ‘Results of three years of CPV
demonstration plants in ISFOC’, Conf. Record 37th IEEE Photovoltaic Specialists
Conf., Seattle> IEEE Press, Piscataway, NJ, pp. 3543–3546.
Sala G. (1989), ‘Cooling of solar cells’, Ch. 8 in Luque A., ed. Solar Cells and Optics in
Photovoltaic Concentration, Adam Hilger, Bristol.
Sala G. and Anton I. (2011), ‘Photovoltaic Concentrators’, in Luque A. and Hegedus S. (eds),
Handbook of Photovoltaic Science and Engineering, John Wiley & Sons, Chichester,
UK.
Sala G., Araújo G. L., Luque A., Ruiz J. M., Coello A., Lorenzo E., Chenlo F., Sanz J.
and Alonso A. (1979), ‘The Ramón Areces concentration photovoltaic array’, Sun
II: Proc. ISES International Solar Energy Society Silver Jubilee Congress, Pergamon
Press, Atlanta, GA, pp. 1737–1741.
Sala G., Arboiro J. C., Luque A., Zamorano J. C., Miñano J. C. and Dramsch C. (1996),
‘The EUCLIDES prototype: an efficient parabolic trough for PV concentration’, Conf.
Record 25th IEEE Photovoltaic Specialists Conf., Washington D. C. IEEE Press, Pis-
cataway, NJ, pp. 1207–1210.
Salim A. and Eugenio N. (1990), ‘A comprehensive report on the performance of the longest
operating 350 kW concentrator photovoltaic power system’, Solar Cells 29, 1–24.
Sinton R., Kwark Y. and Swanson R. M. (1985), ‘23-percent efficient Si point contact
concentrator solar-cell’, IEEE Trans. Electron Devices 32, 2553.
Shockley W. and Queisser H. J. (1961), ‘Detailed balance limit of efficiency of p-n junction
solar cells’, J. Appl. Phys. 32, 510–519.
Solar Industry (2013), ‘CPV industry converges on standard rating conditions’, 7 March.
Terrón M. J., Davies P. A., Oliván J., Alonso J. and Luque A. (1992), ‘Deep emitter silicon
solar cells for concentrated sunlight and operation with light confining cavities’, Proc.
11th European Photovoltaic Solar Energy Conf., Montreux, Switzerland. Harwood
Academic Publishers, Chur, pp. 233–236.
Tobías I. and Luque A. (2002), ‘Ideal efficiency of monolithic, series-connected multijunc-
tion solar cells’, Progr. Photovoltaics 10, 323–329.
Vázquez, M., Algora C., Rey-Stolle I. and González J. R. (2007), ‘III–V concentrator solar
cell reliability prediction based on quantitative LED reliability data’, Progr. Photo-
voltaics 15, 477–491.
Victoria M., Dominguez C., Antón I. and Sala G. (2009), ‘Comparative analysis of different
secondary optical elements for aspheric primary lenses’, Optics Express 17, 6487–
6492.
Welford E. T. and Winston R. (1978), The Optics of Non-Imaging Concentrators, Academic
Press, New York.
Winston R. and Gordon J. R. M. (2005), ‘Planar concentrators near the étendue limit’, Optics
Lett. 30, 2617–2619.
Winston R. and Welford E. T. (1979a), ‘Geometrical vector flux and some new non-imaging
concentrators’, J. Opt. Soc. Amer. 69, 532–536.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch11 page 570

570 I. Luque-Heredia & A. Luque

Winston R. and Welford E. T. (1979b), ‘Ideal flux concentrators as shapes that do not
disturb the geometrical vector flux field: a new derivation of the compound parabolic
concentrator’, J. Opt. Soc. Amer., 69, 536–539.
Wojtczuk S., Chiu P., Zhang X., Derkacs D., Harris C., Pulver D. and Timmons M. (2010),
‘InGaP/GaAs/lnGaAs 41% concentrator cells using bi-facial epigrowth’, 35th IEEE
Photovoltaic Specialists Conf., IEEE, Honolulu, Hawaii, pp.1259–1264.
Yamaguchi M. and Luque A. (1999), ‘High efficiency and high concentration in photo-
voltaics’, IEEE Trans. Electron Devices 46, pp. 2139–2144.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 571

CHAPTER 12

PHOTOVOLTAIC MODULES, SYSTEMS

AND APPLICATIONS
NICOLA M. PEARSALL
Northumbria Photovoltaics Applications Centre
Northumbria University, Newcastle, NE1 8ST, UK
[email protected]

The best way to predict the future is to invent it.

Alan Kay, Apple Computers.

12.1 Introduction
The electricity from photovoltaic (PV) cells can be used for many different appli-
cations, from power supplies for small consumer products to large power stations
feeding electricity into the grid. Previous chapters in this book have discussed
different cell technologies and the optimisation of cell structures to achieve high
efficiency of conversion from light to electricity. In this chapter, we will address
the aspects that allow us to take those photovoltaic cells and incorporate them into
a system delivering a required service.
The chapter concentrates on the use of the most common types of photovoltaic
cells available commercially, described mainly in Chapters 3 to 6, and on typical
system applications, including both stand-alone and grid-connected options. The
stand-alone system can be defined as one where the PV array, together with other
system components, is designed to be the only power source for the load. By
contrast, the grid-connected PV system operates in parallel with the normal grid
distribution system and feeds any electricity not required for local loads into the
grid. This chapter considers the design and use of flat-plate PV systems. Readers
are referred to Chapter 11 for a discussion of the issues involved in the design of
PV systems incorporating concentration of sunlight.
In the next section, we discuss the construction and performance of photo-
voltaic modules. Individual solar cells must be connected to provide an appropriate
electrical output and then encapsulated so as to protect the cells from environmen-
tal damage, particularly from moisture. The design of the module depends on the
application for which it is to be used and an expansion of those applications in

571
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 572

572 N. M. Pearsall

recent years has led to a range of alternative module designs, including the use
of variable transparency and different electrical configurations. Finally, in Sec-
tion 12.2, we discuss module testing, including the establishment of rated output
and long-term performance.
In Section 12.3, we turn to the design of PV arrays, including electrical config-
uration, optimum tilt angle and orientation, protection from shading and mounting
aspects. The variation in performance expected from different array configurations
will be discussed.
The next Section 12.4 discusses the PV system, commencing with the rest
of the system components, usually referred to as the balance-of-system (BOS)
equipment. The BOS portion of the system differs substantially according to the
application and the use of the electricity produced by the PV array. In this section
we discuss the requirements of equipment to be included in a PV system, testing
and standardisation, issues of power conditioning and sizing of the PV system to
meet the required application. Finally, we will discuss applications of PV systems
including current implementation, some aspects of integration into the electrical
network and some environmental considerations.

12.2 Photovoltaic modules

In order to provide useful power for any application, the individual solar cells
described in previous chapters must be connected together to give the appropriate
current and voltage levels and they must also be protected from damage by the
environment in which they operate. This electrically connected, environmentally
protected unit is termed a photovoltaic module. It may be supplied with or without
a metal frame that allows it to be fixed to an appropriate support. Figures 12.1a and
b show simple schematic diagrams of typical module constructions for crystalline
silicon and thin-film cells, respectively.
Due to the difference in fabrication process, module designs for crystalline
and thin-film cells, whilst following the same basic principles in terms of electri-
cal connections and environmental protection, differ in several aspects of module
construction and design. Indeed, thin-film cells are usually fabricated in modu-
lar form, requiring only the encapsulation step after completion of the thin-film
deposition processes. For simplicity, the crystalline silicon solar cell will be con-
sidered initially in each sub-section, since it currently represents the largest share
of the PV market. Variations introduced by the use of thin-film cells will then be
identified.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 573

Photovoltaic Modules, Systems and Applications 573

Glass

Encapsulant

Tedlar

(a)

Superstrate

Metal back contact Note: layers are not

drawn to scale
Solar cell structure
Transparent front contact

(b)
Figure 12.1 Schematic of module construction for: a) crystalline silicon cells — exploded
view showing the different layers which make up the module; b) thin-film cells.

12.2.1 Electrical connection of the cells

The electrical output of a single cell is dependent on the design of the device and
the semiconductor material(s) chosen, but is not usually large enough to meet most
applications. In order to provide the appropriate quantity of electrical power, a
number of cells must be electrically connected. There are two basic connection
methods: series connection, in which the top contact of each cell is connected to
the back contact of the next cell in the sequence, and parallel connection, in which
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 574

574 N. M. Pearsall

Current = I cell
Voltage = 3 x Vcell

V oc Voltage 3 x V oc
0
Current

Single cell
3 cells in series

I sc

Figure 12.2 Series connection of cells, with resulting current–voltage characteristic.

all the top contacts are connected together, as are all the bottom contacts. In both
cases, this results in just two electrical connection points for the group of cells.

Series connection
Figure 12.2 shows the series connection of three individual cells as an example;
the resultant group of connected cells is commonly referred to as a series string.
The current output of the string is equivalent to the current of a single cell and
the voltage output is the sum of the voltages from all the cells in the string (in this
example, since there are three cells connected, the voltage output is equal to 3Vcell ).
It is important to have well-matched cells in the series string, particularly with
respect to current. If one cell produces a significantly lower current than the other
cells (under the same illumination and temperature conditions), then the string will
operate at that lower current level and the remaining cells will not be operating at
their maximum power points. This could also happen in the case of partial shading
of a string (further discussed in Section 12.3.5).

Parallel connection
Figure 12.3 shows the parallel connection of three individual cells as an example.
In this case, the current from the cell group is equivalent to the sum of the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 575

Photovoltaic Modules, Systems and Applications 575

Current = 3 xI cell
Voltage = Vcell

Voltage V oc
0

I sc
Current

Single cell
3 cells in parallel

3 x I sc

Figure 12.3 Parallel connection of cells, with resulting current–voltage characteristic.

current from each cell (here, 3Icell ), but the voltage remains equivalent to that of a
single cell.
As before, it is important to have the cells well matched in order to gain
maximum output, but this time the voltage is the important parameter since all
cells must be at the same operating voltage. If the voltage at the maximum power
point is substantially different for one of the cells, then this will force all the cells
to operate off their maximum power point and the output power will be reduced
below the optimum.

Typical module configurations

The electrical connections within a module can be arranged in any desired com-
bination of series and parallel connections, remembering the importance of the
matching of the units in any series or parallel string. This means, for example, that
parallel connection of series strings should be made using similar strings with the
same number and type of cells. The series/parallel configuration will determine
the current and voltage values obtained from the module under given illumination
and load conditions.
The initial crystalline silicon module designs were for use in stand-alone appli-
cations for the charging of batteries and, thus, the electrical output was required
to be appropriate for battery charging under a range of sunlight conditions. This
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 576

576 N. M. Pearsall

was most readily achieved by the series connection of about 36 crystalline silicon
cells. This produces a rated voltage at maximum power point voltage of around
18 V. Although the voltage reduces with reducing light intensity and increasing
temperature, this configuration allows the module voltage to remain over 12 V,
and therefore capable of charging a 12 V battery, for a wide range of operating
conditions.
Many more PV modules are now used in grid-connected applications, where
the output voltage is not constrained by battery requirements. However, most crys-
talline silicon power modules use between 36 and 72 series-connected cells since
this provides a useful voltage and reasonable module size, which can be handled
and transported. In some cases, a 72-cell module will be configured as two strings
of 36 cells each, either then connected in parallel or with separate terminals so
as to allow flexibility in module connection. In general, larger modules are used
for large ground-mounted systems and architectural applications, whilst smaller
modules are used for small systems to allow greater flexibility in electrical design.
The physical arrangement of cells in the module is generally designed to min-
imise the area, consistent with the minimum spacing between the cells required to
prevent electrical interactions and at the edges of the module to provide environ-
mental protection. Therefore, most modules are rectangular in shape and a typical
configuration is four rows of nine cells per row to obtain a 36-cell module. Typi-
cal module sizes are 0.5–2 m2 in area and the module should also be suitable for
transportation and generally light enough to be lifted by one or two people for ease
of installation.
In the case of thin-film modules, the same design principle is adopted in terms
of voltage enhancement by series connection, although the number of cells in the
string will differ since the voltage per cell is dependent on the cell technology. This
is accomplished by the electrical connection of the cells during fabrication using
a series of patterning steps. As new materials are developed, the module structure
will be adapted for the characteristics of that material (see, for example, Chapter 9
for a discussion of dye-sensitised solar cells). Some modifications of the structure
and manufacturing process can also be expected for developments of existing cell
types (e.g. rear-contact silicon cells).

Module I–V characteristic

The typical I –V characteristic of a photovoltaic cell was described in Chapter 1.
The module I –V characteristic is of a similar shape and can be described by the
same equations, where now the parameters of open-circuit voltage, short-circuit
current, fill factor, maximum power point, reverse saturation current, diode factor,
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 577

Photovoltaic Modules, Systems and Applications 577

4.5
60
4

3.5 50

3
40
Current (A)

Power (W)
2.5

2 30

1.5 20
Current–voltage curve
1 Power curve
10
0.5

0 0
0 5 10 15 20 25
Voltage (V)

Figure 12.4 I –V characteristic of a crystalline silicon module, with the variation of power
with voltage also shown. This illustrates the position of the maximum power point. In this
case, the module has a peak power output of around 65 W.

series and shunt resistances refer to the whole module and are dependent on the
type, number and electrical connection method of the cells.
Figure 12.4 shows an I –V characteristic together with the power curve, to
illustrate the position of the maximum power point. Owing to mismatch between
the characteristics of the component cells and to an increased overall series resis-
tance, the module will typically have a reduced fill factor as compared with its
constituent cells.

Module rating and efficiency

As with the individual cells, the module output varies with illumination and temper-
ature conditions and therefore these must be defined when considering the power
rating of the module. Module testing uses the same standard test conditions (STC) as
are used for the measurement of cells, these being a light intensity of 1000 W m−2,
a spectral content corresponding to a standard AM 1.5 global spectrum and an
operating temperature of 25◦ C (IEC, 2007).
In the ideal case, the module rating would simply be the sum of the rating of
the individual cells but there are, of course, additional losses that must be taken into
account. The most important of these is the mismatch between the cells, whereby
differences in performance will mean that the maximum power point operation of
the module as a whole does not coincide with the maximum power point operation
of some or all of the cells in the module. The mismatch losses can vary depending
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 578

578 N. M. Pearsall

on the operating conditions and whether differences in cell performance are light or
temperature induced. Where possible — for example, for crystalline silicon cells —
manufacturers usually sort their cells into batches according to their performance
and use cells from the same batch to construct the modules. In this way, mismatch
losses are minimised.
The module efficiency is related to the total area of the module in the same
way that the efficiency of a cell is related to the total area of the cell. Because
it is necessary to have the cells physically separated, the module area is always
larger than the sum of the cell areas and therefore the module efficiency is always
lower than the cell efficiency. The resulting reduction in efficiency depends on
the configuration of the module and is defined by the packing density (ratio of
cell area to module area). Typically, a fully packed crystalline silicon module will
have a packing density of around 85% and so, if it uses 17% efficient cells, the
module efficiency will be around 14.5%. While most modules have maximum
packing density in order to produce the highest output possible per unit area, some
architectural applications require the modules to have a higher level of transparency,
for example to provide a combination of electrical generation, shading and daylight
access. In this case, the cells are more widely separated and the resulting module
efficiency is reduced accordingly.
For thin-film cells, the reduction in efficiency is lower because the strip cells
are separated only by the contact strip. In this case, the mismatch between cell
performances can be more important since it is not possible to sort and select the
cells as for the crystalline devices. Since the mismatch arises from variations in
the production process across the surface of the module, it is important to control
the uniformity of all processes.
The performance of the module is also a function of its operating temperature
and hence the rated efficiency is quoted at a standard temperature of 25◦C. The mod-
ule voltage reduces with increasing temperature and, although the current increases
slightly, the overall effect is that the efficiency reduces as the temperature rises. The
amount of the change depends on the cell type and structure, with crystalline silicon
cells typically losing about 0.4–0.5% of their output per degree Celsius rise in tem-
perature. Higher bandgap cells have a lower temperature coefficient; for example,
thin-film silicon reduces by only about 0.2% per degree Celsius. However, thin-film
silicon modules also exhibit a thermal dependence due to annealing of the light-
induced (Staebler–Wronski) degradation and this acts in the opposite direction.
So, their overall temperature coefficient can be zero or even slightly positive over
some temperature ranges. This depends on cell structure and operating conditions.
The operating temperature varies as a function of the climatic conditions of
ambient temperature and incident sunlight and also depends on the module design
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 579

Photovoltaic Modules, Systems and Applications 579

and mounting. Both these latter factors affect the ability of the module to lose heat
and hence determine the operating temperature under given climatic conditions.
A measure of the effect of module design is given by the nominal operating cell
temperature (NOCT) of the module. This is the operating temperature of the cell
junction under normal incidence sunlight at an intensity of 800 W m−2 , when the
module is held at open circuit and mounted on an open rack, the ambient temper-
ature is 20◦ C and the wind speed is 1 m s−1 (IEC, 2007).

12.2.2 Module structure

The structure of the PV module is dictated by several requirements. These include
the electrical output (which determines the number of cells incorporated and the
electrical connections), the transfer of as much light as possible to the cells, the cell
temperature (which should generally be kept as low as possible) and the protection
of the cells from exposure to the environment. The electrical connections have
already been discussed, so in this section we will concentrate on physical protection
from the environment and maintenance of cell operating conditions. Figure 12.5
shows a typical polycrystalline silicon PV module.

Figure 12.5 Typical polycrystalline silicon module (photograph courtesy of Sharp Elec-
tronics Europe GmbH). The square cells allow a high packing density.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 580

580 N. M. Pearsall

In modern crystalline silicon modules, the front surface is usually glass, tough-
ened to provide physical strength and with a low iron content to allow transmission
of short wavelengths in the solar spectrum. The rear of the module can be made from
a number of materials. One of the most common is Tedlar (see Fig. 12.1), although
other plastic materials can also be used. If a level of transparency is required, then
it is possible to use either a translucent Tedlar sheet or more commonly a second
sheet of glass. The glass-glass structure is popular for architectural applications,
especially for incorporation into glazed façades or roofs.
The glass-Tedlar module is usually fabricated by a lamination technique. The
electrically connected cells are sandwiched between two sheets of encapsulant, for
example EVA (ethylene vinyl acetate), and positioned on the glass sheet that will
form the front surface of the module. The rear plastic sheet is then added and the
whole structure is placed in the laminator. Air is removed and then reintroduced
above a flexible sealing membrane to provide pressure on the module to expel air
from the structure. The module is heated and the encapsulant flows and surrounds
the cells. Additional encapsulant material is included at the module perimeter to
ensure complete sealing of the module edges.
In the thin-film module, the glass substrate on which the cell is deposited
is often used as the front surface of the finished module. The module is then
laminated in the same way as for crystalline modules although only a single layer
of encapsulant is required. Lower temperatures are often used to avoid damage to
the cells. If the substrate used for the cell fabrication is positioned at the bottom
of the module, a second glass layer is used to give the transparent front surface
required. In some cases, the thin-film solar cells are deposited on a flexible substrate
(metal foil or plastic). In all cases, particular care must be taken with edge sealing
since all thin-film cells tend to be badly affected by the ingress of moisture. During
manufacture, a clear gap must be left around the edge of the deposited cell area for
proper sealing of the module.
The electrical connections to the module are made via a junction box, usually
fixed to the rear of the module, or by flying leads. These typically exit the module
through the rear Tedlar sheet. In glass-glass modules, the leads may exit through one
edge of the module to avoid drilling holes in the glass sheet. The points at which the
electrical connections are brought out of the module are sealed to prevent moisture
ingress.
The module will exhibit the highest efficiency when the maximum amount of
the light falling on the module is incident on the cells. Light which is incident on the
spaces between cells or at the module edge is either reflected or converted to heat.
Most power modules use the minimum cell spacing, which is 2–3 mm between the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 581

Photovoltaic Modules, Systems and Applications 581

cell edges. This gap is to prevent any problems with electrical shorting between
cells.
The shape of monocrystalline silicon cells is usually described as pseudo-
square, where the cell is cut from a circular wafer and is square apart from the
cut-off corners. Polycrystalline silicon cells are often truly square, depending on
the manufacturing technique of the material. Thin-film cells are deposited in strips,
usually of around 1 cm in width and running the length of the module, although
dimensions can vary depending on cell properties.
In good sunlight conditions and with an ambient temperature of 25–30 ◦ C,
the module usually operates at temperatures in the region of 50–80 ◦ C; the actual
value depends on the module mounting. While these operating temperatures are
not excessive, the difference in thermal expansion of the various components must
be taken into account. Also, allowance must be made for the higher temperatures
experienced during manufacture, albeit for a much shorter time. The cell stringing
allows for some differential expansion in the length of ribbon between each cell.
The electrical connection is also made in two or three places on each cell to allow
for any problems with thermal expansion and other stresses during manufacture or
operation.
The ideal module would also provide good heat transfer in order to keep the cell
temperature as low as possible. However, the encapsulant is required to provide
electrical isolation and physical protection, so a high heat transfer coefficient is
not always possible. The operating temperature is also influenced by the module
encapsulation: glass-glass structures usually run at a higher temperature than glass-
Tedlar modules under similar conditions. The colour of the rear Tedlar film also has
some influence. For example, a module with a white Tedlar backing will reject more
heat than one with a black Tedlar backing, allowing it to operate at higher efficiency.
The module is often provided with a metal frame in order to make it straightfor-
ward to fix to a support structure, although this is less usual for building integrated
applications, where use can be made of traditional mounting methods for glazing.
The frame also provides some protection of the module edges in operation.

12.2.3 Module testing

The electrical output of the module is tested under STC as described earlier. The
measurement under STC provides the module rating in peak watts (Wp) and defines
the module efficiency. The testing method requires control of module temperature,
light spectrum and illumination uniformity.
It is also important to assess the effectiveness of the module construction in
protecting the cells from the environment, since this determines the lifetime of
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 582

582 N. M. Pearsall

the module in operation. Testing conditions have been defined for accelerated life
testing. These include thermal cycling, hail impact, humidity-freeze and electrical
isolation tests and are detailed in IEC standard 61215 for crystalline silicon modules
(IEC, 2005) and standard 61646 for thin-film modules (IEC, 2008). Whether a
module meets the standard is determined by setting maximum limits for change in
output and visual faults after each test. As the applications for photovoltaic systems
change, the operating conditions and lifetime requirements also change. Therefore,
the performance standards are regularly reviewed, updated and added to in order
to meet the requirements of the evolving industry.

12.3 The photovoltaic array

A PV array consists of a number of PV modules, mounted in the same plane and
electrically connected to give the required electrical output for the application.
The PV array can be of any size from tens of watts to tens of megawatts or more,
although the larger systems are often divided into several electrically independent
sub-arrays each feeding into their own power conditioning system.

12.3.1 Electrical connection of modules

As with the connection of cells to form modules, a number of modules can be
connected in a series string to increase the voltage output of the array, or in parallel
to increase the current level, or in a combination of the two. The exact configuration
depends on the current and voltage requirements of the load circuitry fed by the
system output. Matching of interconnected modules in respect of their outputs
can maximise the efficiency of the array, in the same way as matching cell output
maximises module efficiency.
If there is a shaded module in a series-connected string of modules, it can
act as a load to the string in the same way as a shaded cell does in an individual
module. As with the cell, damage can occur due to heating by the current flowing
through the module. The severity of the problem varies according to the number of
modules in the string (and hence the potential voltage drop across the module) and
the likelihood of partial shading of the string (which depends on system design and
location). Where the shading situation may cause damage to the module, bypass
diodes can be included. The bypass diode is connected in parallel with the module
and, in the case of the module being shaded, current flows through the diode rather
than through the module.
The use of bypass diodes adds some expense and reduces the output of the
string by a small amount, because of the voltage that is dropped across the diode.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 583

Photovoltaic Modules, Systems and Applications 583

However, it is now common for bypass diodes to be incorporated into the module
structure at the manufacturing stage. Several diodes may be used, each protecting
different sections of the module, with the diode typically being connected across
twelve to eighteen cells. This integration reduces the need for extra wiring, although
it makes it more difficult to replace the diode if it fails. Where the likelihood of
shading is very low and the system voltage is also low, modules without bypass
diodes are sometimes used.
In systems where shading may reduce the output of one of the strings sub-
stantially below that of the others, it can also be advantageous to include a
blocking diode connected in series with each string. This prevents the current
from the remainder of the array being fed through the shaded string and causing
damage.

12.3.2 Mounting structure

The main purpose of the mounting structure is to hold the modules in the required
position without undue stress. The structure may also provide a route for the elec-
trical wiring and may be freestanding or part of another structure (e.g. a building).
At its simplest, the mounting structure is a metal framework, securely fixed into
the ground. It must be capable of withstanding appropriate environmental stresses,
such as wind loading, for the location. As well as the mechanical issues, the mount-
ing has an influence on the operating temperature of the system, in determining
how easily heat can be dissipated by the module.

12.3.3 Tilt angle and orientation

The orientation of the module with respect to the direction of the Sun determines
the intensity of the sunlight falling on the module surface. Two main parameters
are defined to describe this. The first is the tilt angle, which is the angle between the
plane of the module and the horizontal plane. The second parameter is the azimuth
angle, which is the angle between the normal to the plane of the module projected
onto the horizontal plane and due south (or sometimes due north depending on the
definition used). Correction of the direct normal irradiance to that on any surface
can be determined using the cosine of the angle between the plane that is normal
to the direct beam from the Sun and the module plane.
The optimum array orientation will depend on the latitude of the site, the
prevailing weather conditions and the loads to be met. It is generally accepted that,
for low latitudes, the maximum annual output is obtained when the array tilt angle
is roughly equal to the latitude angle and the array faces due south (in the northern
hemisphere) or due north (in the southern hemisphere). For higher latitudes, such
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 584

584 N. M. Pearsall

as those in northern Europe, the maximum output is usually obtained for tilt angles
of the latitude angle minus about 10–15◦ , due to the effect of both day length
and sunlight levels in the summer. The optimum tilt angle is also affected by the
proportion of diffuse radiation in the sunlight, since diffuse light is only weakly
directional. For locations with a high proportion of diffuse sunlight, the effect of
tilt angle is therefore reduced.
However, although this condition will give the maximum output over the year,
there can be considerable variation in output with season. This is particularly true in
high-latitude locations, where the day length varies significantly between summer
and winter. Therefore, if a constant or reasonably constant load is to be met or,
particularly, if the winter load is higher than the summer load, then the best tilt
angle may be higher in order to boost winter output.
Prevailing weather conditions can also influence the optimisation of the array
orientation if they affect the sunlight levels available at certain times of the day.
Alternatively, the load to be met may also vary during the day and the array can
be designed to match the output with this variable demand by varying the azimuth
angle. For example, for air-conditioning loads peaking in the afternoon, it may
be advantageous to orient the array to the west of south to boost the afternoon
output.
Notwithstanding the ability to tailor the output profile by altering the tilt and
azimuth angles, the overall array performance does not vary substantially for small
differences in array orientation. Figure 12.6 shows the percentage variation in
annual insolation levels for the location of London as tilt angle is varied between 0
and 90 degrees and azimuth angle is varied between −45◦ (south east) and +45◦
(south west). The maximum insolation level is obtained for a south-facing surface
at a tilt angle of about 35◦ , as would be expected for a latitude angle of about 51◦ N.
However, the insolation level varies by less than 10% with changing azimuth angle
at this tilt angle. A similarly low variation is observed for south-facing surfaces for
+/−30◦ from the optimum tilt angle.
The final point to consider when deciding on array orientation is incorporation
into the support structure. For building-integrated applications, system orientation
is also dictated by the nature of the roof or façade in which it is to be incorporated.
It may be necessary to trade off the additional output from the optimum orientation
against the additional costs incurred to accomplish this. Aesthetic issues must also
be considered.

12.3.4 Sun-tracking/concentrator systems

The previous section assumed a fixed array with no change of orientation during
operation. This is the usual configuration for a flat-plate array. However, some
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 585

Photovoltaic Modules, Systems and Applications 585

100

95
Percentage of maximum solar radiation

95 –100
90 –95
90 85 –90
80 –85
75 –80
85 70 –75
65–70

80

75

45
70 30
15
0
65 Azimuth angle
–15
0 5 10
15 20 25 30 35 –30
40 45 50
55 60 65 –45
70 75 80
Tilt angle (degrees) 85 90

Figure 12.6 Percentage variation of annual global sunlight levels as a function of tilt angle
and azimuth angle. The calculations were carried out for the location of London using
Meteonorm Version 4.0.

flat-plate (non-concentrating) arrays are designed to track the path of the Sun. This
can account fully for the Sun’s movements during the day and by season by tracking
in two axes, or partially by tracking only in one axis, from east to west.
The output gain achieved by tracking for a flat-plate system depends on the
latitude, the climatic conditions (especially the ratio of direct to diffuse radiation),
the type of photovoltaic module used and the local terrain. Because the tracked
system receives more sunlight across the day, it will generally operate at a higher
temperature than the fixed array and so the efficiency will be reduced by an amount
determined by the array mounting and the temperature dependence of the type of
module being used. However, this is usually more than offset by the increased light
level received. In terms of climate, higher gains are seen where there is a high
proportion of direct sunlight.
Huld et al. (2008) have predicted the gain from two-axis tracking for sites
in Europe, compared with arrays fixed at the optimum angle for the site, in both
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 586

586 N. M. Pearsall

cases based on crystalline silicon modules. They show that, on an annual basis,
the potential gain is largest in northern latitudes (>60◦ N) and is in the region of
40–60%. This is because of the long days in the summer, when the movement of the
Sun takes it behind the fixed-plate system at the beginning and end of the day. Gains
of 30–40% are observed in southern Europe (latitudes <45◦N) due to the high direct
component in the sunlight at these locations. Lower gains of 20–30% would be
expected in central Europe owing to higher proportions of diffuse sunlight. These
values assume no restrictions due to the local terrain (e.g. mountains) preventing
tracking throughout the day.
Of course, a tracking system incurs extra costs both for the initial installation
and for maintenance, due to the required movement. Large installations will require
multiple tracking pedestals, with the number dictated by the weight of the sub-array
that can be accommodated by each pedestal. In general, a large installation of a two-
axis tracked system would also require more land area than a fixed system of the
same capacity, due to the need to avoid shading of the sub-arrays by neighbouring
ones as they move. These extra costs need to be considered against the additional
output in selecting whether to use a fixed or tracking approach for a flat-plate
system.
For concentrator systems, described in Chapter 11, the system must track the
Sun to maintain the concentrated light falling on the cell. The accuracy of tracking,
and hence the cost of the tracking system, increases as the concentration ratio
increases.

12.3.5 Shading
Shading of any part of the array will reduce its output, but this reduction will vary
in magnitude depending on the electrical configuration of the array. Clearly, the
output of any individual cell or module that is shaded will be reduced according
to the reduction of light intensity falling on it. However, if this shaded cell or
module is electrically connected to other cells and modules that are unshaded, their
performance may also be reduced since this is essentially a mismatch situation.
For example, if a single module in a series string is partially shaded, its current
output will be reduced and this will then dictate the operating current of the whole
string. If several modules are shaded, the string voltage may be reduced to the point
where the open-circuit voltage of that string is below the operating point of the rest
of the array, and then that string will not contribute to the array output. If this is
likely to occur, it is often useful to include a blocking diode for string protection,
as discussed earlier.
Thus, the reduction in output from shading of an array can be significantly
greater than the reduction in illuminated area, since it results from the loss of
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 587

Photovoltaic Modules, Systems and Applications 587

output from:

• Shaded cells and modules.

• Illuminated modules in any severely shaded strings that cannot maintain oper-
ating voltage.
• The remainder of the array because the strings are not operating at their indi-
vidual maximum power points.

For some systems, such as those in a city environment, it may be impossible to

avoid all shading without severely restricting the size of the array and hence losing
output at other times. In these cases, good system design, including the optimum
interconnection of modules in relation to the shading pattern, the use of string
or module inverters and, where appropriate, the use of protection devices such as
blocking diodes, can minimise the reduction in system output for the most prevalent
shading conditions.

12.4 The photovoltaic system

A PV system consists of a number of interconnected components designed to
accomplish a desired task, which may be to feed electricity into the main distribu-
tion grid, to pump water from a well, to power a small calculator or one of many
more possible uses of solar-generated electricity. The design of the system depends
on the task it must perform and the location and other site conditions under which
it must operate. In this section, we consider the components of a PV system, vari-
ations in design according to the purpose of the system, system sizing and aspects
of system operation and maintenance.

12.4.1 System design

There are two main system configurations — stand-alone and grid-connected. As
its name implies, the stand-alone PV system operates independently of any other
power supply and it usually supplies electricity to a dedicated load or loads. It often
includes a storage facility (e.g. battery bank) to allow electricity to be provided
during the night or at times of poor sunlight levels. Stand-alone systems are also
often referred to as autonomous systems since their operation is independent of
other power sources. By contrast, the grid-connected PV system operates in par-
allel with the conventional electricity distribution system. It can be used to feed
electricity into the grid distribution system or to power local loads, which can also
be fed from the grid when there is insufficient power from the PV system.
It is also possible to add one or more alternative power supplies (e.g. diesel
generator, wind turbine) to the system to meet some of the load requirements. These
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 588

588 N. M. Pearsall

DC
loads
PV array
Charge Battery
controller bank
= AC
loads
~
Inverter

Figure 12.7 Schematic diagram of a stand-alone photovoltaic system.

DC
loads

Charge
controller
PV array

Battery
~ bank = AC
~ loads
=
Motor Rectifier Inverter
generator

Figure 12.8 Schematic diagram of grid-connected photovoltaic system.

systems are then known as ‘hybrid’ systems. Hybrid systems can be used in both
stand-alone and grid-connected applications but are more common in the former
because, provided the power supplies have been chosen to be complementary, they
allow reduction of the storage requirement without increased probability of not
being able to meet the load, as discussed in Section 12.4.6. Figures 12.7 to 12.9
show schematic diagrams of the three main system types.

12.4.2 System components

The main system components are the PV array (which includes modules, wiring and
mounting structure), power conditioning and control equipment, storage equipment
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 589

Photovoltaic Modules, Systems and Applications 589

DC
loads

Charge
controller
PV array

Battery
~ bank = AC
~ loads
=
Motor Rectifier Inverter
generator

Figure 12.9 Schematic diagram of hybrid system incorporating a photovoltaic array and a
motor generator (e.g. diesel or wind).

(if required) and load equipment. It is particularly important to include the load
equipment for a stand-alone system because the system design and sizing must take
the load into consideration. By convention, the array components are split into the
photovoltaic part (the PV modules themselves) and the balance-of-system (BOS)
components. In the remainder of this section, we briefly discuss the most common
system components and their role in the system operation, with some examples
of typical performance. There are many different options for BOS equipment,
depending on the detail of the system, and it is only possible to give a general
overview here.

The photovoltaic array

The PV array is made up of the PV modules themselves and the support structure
required to position and protect the modules. The array has already been discussed
in Section 12.3.

Power conditioning
Most systems incorporate electrical conditioning equipment to ensure that the
system operates under optimum conditions. In the case of the array, the highest
output is obtained for operation at the maximum power point. Since the voltage
and current at maximum power point vary with both solar irradiance level and
temperature, it is usual to include control equipment to follow the maximum power
point of the array, commonly known as the maximum power point tracker (MPPT).
The MPPT is an electrical circuit which can control the effective load resistance
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 590

590 N. M. Pearsall

that the PV array sees and thus control the point on the I –V characteristic at which
the system operates.
One of the most common tracking methods is the perturb and observe (P&O)
approach. At regular intervals, the power level is checked for a small change in
operating voltage on either side of the current operating point. If a gain in power is
observed in one direction, then the MPPT moves the operating point in that direction
until it reaches the maximum value. It is also possible to calculate the expected
maximum power point based on measurement of open-circuit voltage or knowledge
of the operating conditions of the module. These methods can achieve a more rapid
determination of the maximum power point but require additional measurements
to be made. For grid-connected systems, the MPPT is often incorporated into the
inverter for ease of operation, although it is possible to obtain the MPPT as an
independent unit.
Although many stand-alone systems use an MPPT approach, some systems
operate at constant voltage, so as to allow battery charging or direct current (DC)
loads to be met at a specific voltage level. It may be necessary to include a DC–DC
converter to change the voltage level of the output of the array to that required
for input to the load. For systems with storage batteries, it is also usual to include
charge control circuitry, in order to control the rate of charge and so prevent damage
to the batteries and extend their lifetime.

Inverter
If the PV system is to supply alternating current (AC) loads or feed power directly
into the AC distribution grid, then an inverter must be included to convert the DC
output of the PV array to the AC output required. As with PV systems, inverters
can be broadly divided into two types, these being stand-alone and grid-connected
(sometimes referred to as line-tied).
The stand-alone inverter is capable of operating independently from the util-
ity grid and uses an internal frequency generator to obtain the correct output fre-
quency (usually 50/60 Hz). By contrast, the grid-connected inverter must integrate
smoothly with the electricity supplied by the grid in terms of both voltage and
frequency and therefore uses the grid as a reference. Some modern inverters have
the capability to change from stand-alone to grid-connected mode. The output volt-
age of the inverter is chosen according to the load requirements, e.g. 220–230 V
single-phase for European domestic appliances. However, if the electricity from
the PV system is to be fed directly into the supply of a large office building, for
example, a 415 V three-phase output may be chosen. For large PV installations
feeding directly into the distribution grid, higher voltages may also be chosen. The
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 591

Photovoltaic Modules, Systems and Applications 591

output voltage of the inverter is selected to match the grid voltage in the location
at the network level required.
The input voltage to the inverter depends on the design of the PV array, the
output characteristics required and the inverter type. Stand-alone systems com-
monly operate at 12, 24 or 48 V, since the system voltage is determined by the
storage system, whereas grid-connected inverters usually operate at significantly
higher voltages (between 150 and 1000 V, depending on the system size).
The shape of the output waveform is important because some loads can over-
heat or be damaged if a square wave output is used. True sine wave or quasi-sine
wave (or modified sine wave) outputs are generally more costly but are much
more widely applicable. Most modern stand-alone inverters provide a modified
sine wave output, whilst grid-connected inverters should have a sine wave output
with a very low harmonic content. Limits on the harmonic distortion are defined
in the regulations for grid connection.
While most systems use an inverter connected to a group of modules, it is
also possible to obtain modules with an integrated inverter, positioned on the rear
of a module and converting the electrical output from that single module. This
module–inverter combination is sometimes referred to as an AC module. There are
some advantages in system design, such as the use of AC wiring for most of the
power transmission and reduced losses for non-uniform systems (e.g. where there is
shading) since each module has individual maximum power point tracking. Indeed,
the system designer now has a wide choice of inverter configurations, from a single
inverter per module via inverters connected to each string to a single central inverter
per system. A recent development has been the use of individual maximum power
point trackers, attached to one or two modules, with the output from a collection of
modules then being sent to a single inverter for DC/AC conversion. The choice of
system configuration depends on the uniformity of irradiance (particularly whether
there are shading problems), the overall efficiency of the different options and the
overall cost.
Inverters for PV systems are designed to have high conversion efficiency and,
although this varies with inverter design, values in the range of 95–97% for maxi-
mum efficiency are now common. The efficiency varies with the operating point of
the inverter, but usually reaches its maximum between 30 and 50% of rated capac-
ity and shows only a small decrease as the power level increases. However, the
efficiency generally reduces substantially at power levels below about 10% of full
power. Because of the variation of efficiency with operating conditions, it is also
usual for European inverter manufacturers to quote the Euro-efficiency value of
their unit. This is calculated by considering the efficiency at several input levels and
producing a weighted average according to a defined set of operating conditions.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 592

592 N. M. Pearsall

Recently, it has been argued that this method should be refined to also take account
of the maximum power point voltage range over which the inverter operates
(Bletterie, 2011), although this modification has not yet been widely adopted.
In locations in the middle and north of Europe, the performance at low light
levels (and hence low power levels) can have a significant effect on the overall
system efficiency. Thus, it is usual to size the inverter at about 80% of the array
capacity so that high inverter efficiencies are maintained at lower power levels.
This means that the very high power levels are sacrificed since they are out of the
range of operation of the inverter, but the balance of low and high power operation
is usually such that it is more advantageous to use a reduced inverter size. This may
not be the case for systems that experience a significant proportion of high power
levels due to cold, clear weather conditions. Although all locations experience low
light level conditions, at least for fixed flat plate systems, at the end of the day,
southern European locations operate more often under high irradiance conditions
and therefore it may be most beneficial to derate the inverter by 5–10% only. The
nature of the installation should also be taken into account. For example, the solar
irradiation received on an array mounted on a vertical façade will be less than
that for the same array mounted at the optimum tilt angle in the same location.
Therefore a smaller inverter would be advisable for the façade system.
When the inverter is grid-connected, it is essential to ensure that the system
will not feed electricity back into the grid when there is a sustained fault on the
grid distribution system. This problem is known as islanding, and safeguards are
required in order to provide protection for equipment and personnel involved in the
correction of the fault. Islanding is usually prevented by closing down the inverter
when the supply from the grid is outside certain limits. The allowable limits vary
from country to country but are usually around +/−2% in voltage and frequency.
Requirements for prevention of islanding for systems are detailed in the connection
regulations for each country.
In recent years, there has been a growing amount of embedded generation (i.e.
electricity generation systems sited towards the consumer end of the distribution
system) connected to the grid in some countries. This includes wind, photovoltaic
and other systems. Problems can occur if all this generation turns off when there is
a transient fault on the grid and in some countries grid operators are now requiring
a fault ride-through capability on the inverter to address this issue. This means that
the embedded generation will continue to operate for a short period to maintain
continuity of supply. If the grid fault persists beyond a specified time, the inverter
will then close down until the fault is cleared.
Although not yet fully realised, the inverter also has the capability to contribute
to the safety and stability of the grid by a variety of means. Most inverters currently
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 593

Photovoltaic Modules, Systems and Applications 593

operate at unity power factor, but there is the possibility to supply reactive power
when required by the grid, at the penalty of some loss of active power output.
The inverter also often provides system monitoring and could be incorporated
into a smart grid configuration, supplying information on the PV system status
and output. Some inverters also supply integral storage to allow the exchange of
electricity with the grid to be managed.

Storage
For many PV system applications, particularly stand-alone, electrical power is
required from the system during hours of darkness or periods of poor weather.
In this case, storage must be added to the system. Typically, this is in the form of
a battery bank of an appropriate size to meet the demand when the PV array is
unable to provide sufficient power.
The storage capacity must be chosen so as to provide sufficient security of
supply for the load and this will vary depending on how critical it is to maintain
the operation of the load. Of course, increasing storage capacity also increases the
system cost and so an appropriate balance between expenditure and performance
must be achieved.
There is a range of battery types and designs that can be used, although the most
common type is a low maintenance lead-acid battery, designed for medium depth
discharge. The design of lead-acid battery commonly used in vehicles is unsuitable
for PV applications since this only allows shallow discharge if the lifetime of the
battery is to be maintained. As previously mentioned, a charge controller should
be used so as to control the rate of charging and discharging of the battery and
to prevent complete discharge, which can seriously damage the battery. Attention
must also be paid to the location of the battery, since extremes of temperature will
also reduce operating efficiency and lifetime.
Generally, grid-connected systems do not include storage but this would be
one method to offset the variability of output from the PV system. As mentioned
in the previous section, some inverter manufacturers are beginning to offer integral
storage as an option. This adds cost but may be advantageous if the value of
deferring use of the electricity from the PV system is higher than that of exporting
it to the network. It is likely that storage will become a more common feature of
grid-connected systems as the level of penetration on the network increases.

Load equipment
The nature of the load equipment will determine the need for, and suitability of,
the power-conditioning equipment and the capacity of both the PV system and
the storage. The first consideration is whether the load or loads use DC or AC
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 594

594 N. M. Pearsall

electricity. In the former case, the loads can be operated directly from the PV
system or battery storage whereas AC loads will require an inverter to be included
in the system.
Where the system is grid-connected, loads are almost always AC but for
autonomous systems, a choice can be made. This choice will depend on the avail-
ability, cost and performance of the DC and AC versions of the load equipment. For
example, it is possible to obtain high-efficiency DC fluorescent lighting which, by
virtue of its superior performance compared with AC lighting, results in a smaller
capacity requirement for the PV system and hence, usually, reduced costs. In the
case of water pumping, the choice between DC and AC pumps depends on the
nature of the water supply (e.g. deep borehole or surface pump).
The requirements of the load in terms of voltage and current input range will
influence the type of power conditioning included in the system and the load profile
will determine the relative sizes of the PV system and the storage, if used. System
sizing in accordance with load details is discussed in more detail later in the chapter.

Cabling and switching equipment

The array cabling ensures that the electricity generated by the PV array is trans-
ferred efficiently to the load and it is important to make sure that this is specified
correctly for the voltage and current levels which may be experienced. Since many
systems operate at low voltages, the cabling on the DC side of the system should be
as short as possible to minimise the voltage drop in the wiring. Switches and fuses
used in the system should be rated for DC operation. In particular, DC sparks can
be sustained for long periods, leading to possible fire risk if unsuitable components
are used.
Because of their design and exposed situation, PV systems can be vulnerable
to lightning strikes, with the probability depending on the prevalence of lightning
at the location in question. This is usually addressed by ensuring correct grounding
of the system and including surge protection of sensitive electrical components. In
some cases (e.g. for arrays on buildings), lightning conductors can divert strikes
away from the array although this does not prevent induced currents.

12.4.3 System sizing

It is important to determine the correct system size, in terms of both peak output
and overall annual output, in order to ensure acceptable operation at minimum
cost. If the system is too large, it will be more expensive than necessary without
increasing performance levels substantially and therefore the system will be less
cost-effective than it could be. However, if too small a system is installed, the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 595

Photovoltaic Modules, Systems and Applications 595

availability of the system will be low and the customer will be dissatisfied with the
equipment. Again, the cost-effectiveness is reduced.
Although many of the same principles are included in the sizing process, the
approach differs somewhat for stand-alone and grid-connected systems. Consid-
ering stand-alone systems first, the initial step is to gather the relevant information
on the location and purpose of the system. Location information includes:

• Latitude and longitude.

• Weather data — monthly average sunlight levels, ambient and maximum tem-
peratures, rainfall, maximum wind speeds, other extreme weather conditions.
• Constraints on system installation — tilt angle, orientation, risk of shading.
Information on system purpose includes:

• Nature of load or loads.

• Likely load profile — daily, annual variation (if any).
• Required reliability — ability to cope with loss of load (for example, clinic
lighting requires a higher level of reliability than a lighting system for a domes-
tic house).
• Likelihood of increase of demand — many systems fail because they are sized
for an existing load, but demand increases soon after provision of the PV
supply.
If an autonomous system is required, the PV system must provide sufficient elec-
tricity to power the loads even under the worst conditions. Thus, system sizing
is usually carried out for the month that represents the worst conditions in terms
of the combination of high load levels and low sunlight conditions. This is not
necessarily the month that has the lowest sunshine or the highest load, but that for
which the combination represents the worst case.
For a given system design, the average electrical output in the sizing month
can be calculated from the average daily insolation level (usually expressed in
kWh m−2 ) taking into account the number of modules, their rated efficiency, the
efficiencies of all control and power conditioning equipment, the efficiency of any
storage system, mismatch losses, wiring losses and the operating temperature. For
an autonomous PV system, the average daily electrical output should match or
exceed the average daily load. If this is not the case, then the PV array size must
be increased.
The battery storage allows for variations in the load level during the day and
the provision of power at night. The battery bank must be sized to accommodate
the average daily need for electricity that cannot be directly supplied by the PV
system and so that this results in only the allowable discharge of the batteries.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 596

596 N. M. Pearsall

So far, we have considered only average values for load and sunlight levels.
The daily sunlight levels can vary substantially and the battery storage must also
allow for providing power in periods of unusually poor weather conditions. The
length of the period to be allowed for is determined by consideration of local
weather conditions (i.e. the probability of several days of poor weather) and the
importance of maintaining power to the load. Clearly, if the system is used for
medical purposes or communications, loss of power could have serious conse-
quences, whereas for other situations, such as powering domestic TV or lighting,
it is merely an inconvenience. Since an increase in the period for which supplies
can be maintained involves an increase in the size of the PV array and/or battery
bank and hence an increase in system cost, this aspect is an important part of the
sizing exercise. Supply companies refer to this by many different terms, including
reliability, availability and loss-of-load probability (LOLP).
Clearly, the sizes of the PV array and battery bank are linked, and an increase
in the size of one can often allow a decrease in the size of the other. The sizing oper-
ation is usually an iteration of the problem to find the most cost-effective solution,
taking into account the requirements and preferences of the user. A detailed sizing
can be performed using system-sizing software, commercially available from a
variety of sources, using detailed solar data for the location and allowing a more
robust solution than using daily or monthly averages.
For a grid-connected system, it is not usually necessary to meet a particular
load but only to contribute to the general electricity supply. Some systems are
designed to feed all their output into the electricity grid whilst others (e.g. most
building-integrated systems) are designed to meet some of the load in a local area
with the rest of the requirement being supplied by the grid. These latter systems
only feed power into the grid when their output exceeds the demand of the load. The
system sizing is therefore not often governed by the size of the load, but by other
constraints such as the area available for the system and the budget available for its
purchase and installation. Therefore, most sizing packages are used to determine
potential output and to compare different options of system location and design,
rather than optimising system size as such. Modern sizing software includes options
for different mounting configurations and for the inclusion of an allowance for
shading based on details of the array surroundings.
The accuracy of the output of any simulation will depend on the accuracy of
the input data, as with all such systems. However, since there is a natural variation
in solar irradiation levels depending on climatic conditions, this must also be taken
into account in the use of results from a simulation. If average insolation data are
used, as is most common, then an average output will be obtained as a result. This
is strictly speaking only the average value over the period represented by the input
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 597

Photovoltaic Modules, Systems and Applications 597

data rather than a prediction of what any future values will be. Thus it is possible
to obtain practical results from a system, which are significantly different from
the simulation results of the design process, simply because of normal climatic
fluctuations.

12.4.4 System operation

The output of any PV system depends mainly on the sunlight conditions but can also
be affected by temperature, shading and the accumulation of dirt on the modules.
The overall system performance is usually represented by the efficiency, which
is defined as the ratio of the electrical output delivered to the load (in kWh) to
the sunlight energy input (also in kWh) over the surface of the array in the same
period. In general, this overall efficiency results from several processes to which
individual efficiency values can be assigned, e.g. the conversion of sunlight to DC
electricity, the conversion of DC to AC by the inverter and so on.
The system yield is also a useful parameter. This expresses the annual output
(or that over another defined period) normalised for the rated capacity of the system
and is in units of kWh/kW. This allows comparison of systems in different locations.
For example, based on average annual irradiation levels, system yields in southern
Europe are typically in the range 1300–1500 kWh/kW, whereas in Germany a value
of around 1000 kWh/kW might be expected. Since the yield does not explicitly
include the sunlight level received over the period, account must be taken of whether
the level was above or below average if a measured value is to be used for a critical
assessment of system performance.
A useful parameter for assessing system performance is the performance ratio,
which is either given as a percentage or as a number between zero and one. Essen-
tially, this parameter expresses the performance of the system by comparison of
the system output with that from a reference system of the same design and rating
at the same location, but with no losses. Because the sunlight level is included
in the assumption of the output of the reference system, the performance ratio
is independent of sunlight conditions. Thus, it allows the comparison of system
design and performance in different locations. The performance ratio is mainly
affected by the operating characteristics of the inverter, including efficiency and
reliability, the effect of temperature on the PV module efficiency, the effectiveness
of the maximum power point tracking, electrical losses in the cabling, mismatch
between modules, shading and dirt accumulation on the modules.
The performance ratio (PR) can be calculated from the following formula:
PR = system output over period/(average daily irradiance × array rating
× number of days in period × monitoring fraction)
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 598

598 N. M. Pearsall

where all parameters are values for the same period, the system output is in kWh,
the average daily irradiance is in kWh m−2 and the array rating is in kW. The
array rating is calculated as the sum of the power ratings of the individual modules
making up the PV array. The monitoring fraction is the fraction of the period for
which monitoring data are available and have been used to determine the values of
the other parameters. The formula makes the assumption that average conditions
are experienced for the time when data are not collected and so care must be taken
with the use of PR values calculated for monitoring fractions less than 0.9.
Performance ratio values for grid-connected systems give a useful measure
of system design and operating quality with a value of between 0.8 and 0.9 being
the aim. This implies high inverter efficiency, minimum electrical losses between
the modules and the inverter, good thermal transfer to keep module temperature
to a reasonable level, and good system reliability. Performance ratio values cannot
normally be used to indicate the quality of a stand-alone system since the array may
not always be operating with maximum power point tracking and the requirement
to supply the load fully under worst-case conditions means that some potential
output is wasted under more favourable conditions. This can lead to a low PR
value even for a well-designed system. A different approach must be used for the
stand-alone system with availability of load (i.e. the provision of sufficient power
when it is needed to operate the load) being the most important aspect.

12.4.5 Operation and maintenance

Because of the lack of moving parts and simple connections, a PV system generally
requires little maintenance. However, it is necessary to ensure continued access to
sunlight, by cleaning the panels at appropriate intervals, refraining from building
any structures that could shade the panels and by cutting back any branches or
other vegetation that could cover the system. The electrical connections should
also be checked at regular intervals to eliminate any problems, such as corrosion
or loose connections. If included in the system, the battery bank may need regular
maintenance according to the type chosen.
The requirement for cleaning is often overestimated by those with little experi-
ence of PV systems. It is generally assumed that 3–5% of performance will be lost
if the system is only cleaned annually, with up to half of that loss being experienced
within a few weeks of cleaning. However, the losses incurred and thus the require-
ment for cleaning are very dependent on location and are best determined from
practical applications operating under similar conditions. For example, if there is
the possibility of dust or sandstorms causing accumulation on the modules, per-
haps in a desert area, then more frequent cleaning will be required. This can also
be the case for systems installed in industrial areas close to sources of airborne
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 599

Photovoltaic Modules, Systems and Applications 599

pollutants. For building integrated systems on houses in many parts of Europe, it

may not actually be necessary to clean the systems regularly, since the action of
rainwater on the inclined panels removes surface dust. A thorough cleaning every
few years would then be carried out to remove persistent dirt.
Operational problems can occur as a result of poor maintenance of the BOS
components (including loads and batteries) or allowing the array to become
obscured or damaged. This latter problem indicates a lack of understanding of
the operation of the system and there is a need for education of users to ensure
that they operate the system correctly. This is also demonstrated by system failures
arising from the addition of loads that were not included in the original system
sizing for a stand-alone system. In this case, the combination of the PV and storage
system cannot meet the increased demand and there is a danger of damage to the
batteries from deep discharging.
The costs of operation and maintenance will vary with application, since they
are dependent on the ease of access and the requirement for cleaning, the remote-
ness of the system and any replacements that may be required. However, they are
generally not more than a few percent of the system cost per annum. For large
applications, such as ground-mounted power plants, it is common to have a mon-
itoring and maintenance contract that ensures minimum downtime on the plant
operation and early identification of loss issues.

12.4.6 Photovoltaic applications

The wide range of applications in which photovoltaic systems are employed cannot
be covered in depth in this chapter, but we will summarise some of the main uses
for stand-alone systems before considering both building-integrated and ground-
mounted systems that are connected to the grid.

Stand-alone PV systems
Most stand-alone PV systems are used to power loads in locations that are remote
from the electricity grid supply, although they are also used in urban locations
where it may be inconvenient or expensive to use the grid supply (e.g. for low-
voltage lighting on bus shelters). Stand-alone systems generally range in size from
a single module powering a solar home system (SHS) to a few kilowatts of PV
supplying a local area grid network, although there are options for much larger
stand-alone systems for applications such as rural electrification.
The systems are autonomous and so must include some type of energy storage
to supply power in the absence of sunlight. Systems are usually categorised as
critical, meaning that there are issues relating to health or safety, or non-critical,
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 600

600 N. M. Pearsall

meaning that loss of load causes no more than inconvenience. In some cases,
systems powering applications with a high economic value (e.g. communications)
are also treated as critical applications from the design point of view. In general, a
critical system will have a higher storage capacity to allow for continued operation
through prolonged periods of low insolation compared with the average for that
location. The economics of storage dictate that, for larger systems and for those
where high reliability is paramount, a hybrid system is often used, allowing some
of the energy storage to be in the form of fuel for an internal combustion engine.
In locations where the seasonal availability of wind energy is complementary to
that of the solar irradiance, it can also be cost-effective to include a wind turbine
in the hybrid system. Other options would be biomass or small hydro-electricity
systems, depending on availability at the location.
In a small, non-critical system, such as an SHS, one or two PV modules
charge a battery during the day, and the power is used at night for a few high-
efficiency lights and a radio or small TV. A charge controller ensures that the bat-
tery is not overcharged or deep-discharged, to provide as long a battery lifetime as
possible. Battery capacity and lifetime can be substantially reduced by poor charg-
ing regimes, especially consistent discharging below the design level of charge. To
keep costs as low as possible, standard systems are sold to all users, though there
may a choice of the number of modules based on local needs and solar conditions.
System reliability largely depends on users modifying their usage of the loads to
ensure an appropriate battery charging regime, often using indicator lights on the
charge controller.
For systems that are part of safety-critical networks, for example those in
aircraft navigation aids or telecommunication systems, a very low LOLP must
be guaranteed, perhaps as little as one or two hours per year on average. The
stochastic variability of solar irradiance means that a large PV array feeding into a
large battery storage capacity must be provided to ensure that almost all periods of
poor weather can be accommodated, or additional charging must be provided from
a small internal combustion engine, usually a diesel, with a fuel store sufficient
to maintain this option with fuel deliveries under normal maintenance visits only.
The PV array and battery system should be sized so that the engine is run at full
power on a sufficiently regular basis to keep it in good condition, but so that the
overall cost is minimised, taking into account the fuel and equipment costs. The
design will vary with location in terms of both climate and cost of fuel.
Some stand-alone systems provide power for a local network for a small com-
munity. Again, the use of a hybrid system is common, although in this case the
alternative power source may be another renewable energy technology such as wind
or small hydropower to supplement the PV output. The optimum combination will
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 601

Photovoltaic Modules, Systems and Applications 601

take account of the required load profile and the generation profile of the various
power sources. It is possible to commence a local network with a small PV or PV
hybrid system and then add to it as consumer demand grows and finance allows.
The first commercial use of PV cells was in space, providing the electrical
power required by satellites for their operation, and this continues to be an impor-
tant application of the technology. These systems could be described as a special
case of a stand-alone system, where the space environment must be considered in
determining the system size needed to meet the load requirement. Particular atten-
tion must be paid to the PV array size and weight, because of launch constraints,
and to the performance of the solar cells after exposure to irradiation in the space
environment.
Stand-alone PV systems are used for a wide range of applications from lighting
for rural homes to telecommunication repeater stations, from providing the power
for parking meters to maintaining the vaccine cold chain by powering refrigerators
in rural clinics, from water pumping to navigation buoys. Some systems have only
the PV array as their power source, whilst others incorporate one or more other
power supplies. Whilst the uses differ greatly, the design approach is the same.
The load is carefully considered and then the system is sized so as to achieve the
required probability that it will meet the load given the climatic conditions at the
location. In this way, PV systems can make a remarkable contribution to providing
power in difficult locations and to people who have no other source of electricity.

Grid-connected PV systems
The grid-connected system, as its name implies, is a PV system that is installed
in a location where there is also grid electricity available and a direct connection
is made to allow power to be fed to or taken from the grid. We can divide grid-
connected PV systems into two main categories, which are generally described as
centralised and distributed. Centralised systems feed electricity directly into the
grid and are usually connected at medium voltage. These systems will be discussed
later in this section. We will first consider distributed systems, which generally feed
local loads first and then export any excess to the grid. These systems are usually
connected at the level of the electricity consumer (i.e. at low voltage). The most
common example of this kind of system is the one that is mounted on or integrated
into a building.

Building-integrated photovoltaic systems

The building-integrated photovoltaic (BIPV) system is an excellent application for
the use of photovoltaics in an urban environment since it takes advantage of the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 602

602 N. M. Pearsall

distributed nature of both sunlight and the electrical load. The potential benefits of
the BIPV system can be summarised as follows:

• In common with other PV systems and most renewable energy technologies,

it has a lower environmental impact than production of electricity from con-
ventional fuels.
• The electricity is generated at the point of use, so reducing the impacts and
costs of distribution.
• There is a possibility of offsetting some of the cost of the PV array by the
amount that would have been paid for the building material it has replaced.
• The system does not require additional land area, since building surfaces are
used to accommodate the array.

The PV modules can be integrated in several different ways, for example to

replace roofing tiles, in place of façade materials or as sunshades. Figure 12.10

Figure 12.10 Example of façade integration of photovoltaics. The photograph shows the
40 kWp PV façade on the Northumberland Building at the University of Northumbria.
The PV array is integrated into the rainscreen overcladding. This system was installed in
1994 and is an early example of façade integration. Photograph courtesy of University of
Northumbria.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 603

Photovoltaic Modules, Systems and Applications 603

Figure 12.11 Integrated PV façade using copper indium gallium diselenide thin-film
modules. Photograph courtesy of Würth Solar GmbH.

shows an early example of façade integration using crystalline silicon modules,

but there are many different ways of including the PV array in the building design.
Figure 12.11 shows a later example, this time with thin-film copper indium gallium
diselenide modules. It can be seen that the visual aspect of the two buildings is
significantly different.
A large proportion of current PV systems on buildings are not fully integrated,
but rather the array is mounted on the roof or façade, leaving the original exterior of
the building in place. This is generally less expensive than full integration, although
it is often not as aesthetically pleasing and does not present as many opportunities
for using the modules to perform other building-related functions.
The principle of the technical system design is similar to that for other PV
applications, but there are some additional aspects to be taken into account. In
contrast to the stand-alone systems described in the previous section, the BIPV
system is rarely sized to meet a particular load but usually makes a contribution to
the electricity requirement of the building as a whole. It may be designed to match
the general load profile or to provide higher output levels when, for instance, air
conditioning is required, but it does not need to be an autonomous system since
the building also has a grid supply.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 604

604 N. M. Pearsall

The area available for the BIPV array may be constrained by building design,
shading from surrounding structures or owner preference. Thus, the system size is
often dictated by the building’s structural details rather than its electrical loads. The
visual aspect of the system is also important and this will influence the choice of
module type, location and detailed integration method. Finally, the system design
must take into account ease of installation, maintenance and operation, and com-
pliance with local building regulations, both structural and electrical.
A fully integrated BIPV array performs at least two tasks, the generation
of electricity for use in the building and the protective functions of the external
building element, but arrays can also be designed to perform additional functions.
The most common function is shading of internal spaces, by louvre systems on
the exterior of the building, by designing the cladding so as to provide shading to
the windows at high Sun positions or by the use of semitransparent PV elements
for a roof or façade. Figure 12.12 shows an example of the use of semitransparent
modules in a glazed façade, where the cells provide both visual stimulation by
variation of the arrangement pattern and shading to reduce solar gain and glare.
In some cases the heat at the rear of the modules can also be used. For all
except the highest efficiency modules, less than 20% of the light falling on the
module is converted to electricity and, whilst a few percent is reflected, the rest
is absorbed as heat. This results in a module operating temperature that can be

Figure 12.12 Example of the use of photovoltaic modules to influence indoor lighting
patterns. The Solar Office at the Doxford International Business Park in Sunderland, UK,
has a 73 kWp array formed from semitransparent PV modules. The cell spacing is varied to
create the light effects in the inner atrium. Photograph courtesy of Akeler Developments Ltd.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 605

Photovoltaic Modules, Systems and Applications 605

25–50 ◦ C above ambient, depending on the details of integration. Reducing the

operating temperature by removing some of the heat is advantageous in terms of
increasing system efficiency and a double benefit can be obtained if the heat is
useful for another purpose.
Because of the rather large area of the module and the relatively modest tem-
perature differential between the module and ambient temperatures, it is not usually
cost-effective to use forced air or fluid flow to extract the heat unless there is a direct
use for that heated air or fluid. However, the heat can be used to assist natural venti-
lation within the building by taking in cold air at the bottom of the building. As this
air is heated behind the PV façade, it rises and pulls in more cold air to replace it.
One of the earliest examples of this type of application was the Mataró Library in
Spain (Lloret et al., 1997).
Even for a system where no use is made of the heat, care must be taken to
ensure that the PV array operating temperature remains at an acceptable level. For
ground-mounted PV arrays, there is free air movement around the array and so
some cooling is effected. This is not the case for a BIPV system, which forms part
of the building fabric. The design must include adequate ventilation around the
modules if significant losses in efficiency are to be avoided. It is somewhat easier
to ensure adequate ventilation for PV arrays added to the building rather than fully
integrated. In this case, a small gap (typically a few cm) is provided between the
existing roof or façade and the rear of the PV array so as to allow some natural
airflow.
One aspect of installation that must be carefully considered is the possibility
of shading of the modules, from surrounding buildings, trees and other vegetation
and, in some cases, parts of the same building (e.g. by ventilation outlets, offset
roofs, etc.). It can often be difficult to avoid shading completely and therefore the
design of the system needs to minimise the effect on the output by good choice
of string connections and splitting the system into sub-systems using a number of
inverters.
The PV system is connected in parallel with the local grid, with the conven-
tional electricity supply meeting any shortfall between the system electrical output
and the building demand. The system must conform to safety regulations for con-
nection and those relating to voltage, frequency and cut-off in the case of grid
faults, as discussed previously. Arrangements can usually be made to sell back any
excess production from the BIPV system to the electricity supply company. There
is a wide range of tariffs offered for this electricity, ranging from the replacement
generation cost (i.e. the cost for production of the same amount of electricity by
the electricity company, not including distribution costs and overheads) to several
times the normal electricity rate, where a scheme to promote BIPV exists.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 606

606 N. M. Pearsall

The wide range of possible designs and the variation in the cost of the building
material that is being replaced, and hence the offset of the PV system cost that this
provides, makes it difficult to generalise about the cost of BIPV systems. Some
systems are less costly than the material being replaced and so the electricity
produced is essentially free except for any small additional maintenance costs.
Most systems are replacing material that would be less costly than the PV array
and so there is some residual cost to assign to the electricity produced. Nevertheless,
the reducing costs of PV modules and other system equipment means that these
distributed systems are becoming economically attractive across a wide range of
locations, especially as the costs of conventionally generated electricity rise.

Centralised grid-connected PV systems

Since the inception of strong financial support schemes for PV across several
European countries and in the US and Asia, there has been increased interest in
centralised PV systems and the capacity of such systems has grown considerably.
These systems are generally ground-mounted (although there are a few that are
included on the roofs of large buildings or building complexes) and they are pre-
dominantly designed to feed output into the grid rather than meeting local loads.
In the last five years, the maximum system capacity has grown from around 50 MW
to around 250 MW, with even larger systems currently under construction.
The system design is modular in nature, with the overall system being divided
up into a number of similar sub-systems, typically of 0.1–1 MW in size, depend-
ing on the overall capacity of the system. This approach has several advantages.
From the electrical point of view, this allows operation at suitable voltages for
reduction of losses, usually between 500–1000 V, and means that multiple iden-
tical inverters can be used. These can be installed in a distributed manner across
the plant site to reduce cabling losses. From an installation point of view, the use
of sub-systems allows different configurations to be used in different parts of the
array field to allow for any physical constraints of the site. Finally, the modular
nature allows the plant to be built in sections and, if permitted, to connect each
section to the grid as it is completed. This reduces the time between the start of
plant construction and the first sale of exported electricity. The centralised PV sys-
tem is usually connected to the grid at medium voltage levels (typically tens of kV
depending on the location), rather than at the low voltage level typical of distributed
systems.
The modules are usually mounted on open rack support structures to allow
maximum ventilation and therefore minimum operating temperature. The strength
of the support structures must take account of wind or snow loading likely at the
location. The most common arrangement is to have rows of modules, tilted at the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 607

Photovoltaic Modules, Systems and Applications 607

Figure 12.13 Ground-mounted PV system at Douneika, Greece. The system is rated at 2

MW and consists of 9,274 polycrystalline silicon modules, tilted at 27◦ . The physical and
electrical layout takes account of the topography of the site. Photograph courtesy of Phoenix
Solar Greece.

optimum angle for solar capture, and spaced so that there is a low level of shading
between the rows. This requires a trade-off between the available land area and
the amount of shading at low solar elevations and the spacing can be expressed by
a parameter known as the Ground Cover Ratio. This is the ratio of the total area
of PV modules to the total land area of the site and is typically around 0.5–0.6,
although it can vary depending on the latitude and the details of the site. In some
cases, the spacing, tilt angle and even azimuth angle can be varied on different
parts of the site to take account of different topographies, the aim being to generate
the highest energy yield from the available area. Figure 12.13 shows an example
of a ground-mounted centralised PV system.
While most of the large grid-connected PV systems are fixed, some flat-plate
systems employ Sun-tracking on either one or two axes. For tracked systems, the
most common configuration is to mount a group of modules on a pedestal, to
give the required freedom of movement, similar to the configuration illustrated in
Fig. 11.29 in Chapter 11 for concentrator systems. The array field is then made
up of a multiple of pedestals up to the required system capacity. Tracking both
increases the output of the modules and modifies the output profile, keeping it
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 608

608 N. M. Pearsall

high across a longer portion of the day. As a disadvantage, a system using tracking
will require a larger area of land so as to keep the shading between the modules
to an acceptable level. The support structure and tracking system also make the
installation more costly. The performance advantage for a tracked system depends
on the climatic details at the location and is more pronounced for sites with a high
fraction of clear days across the year. The choice of whether or not to use flat-plate
tracking is therefore location-dependent. Some of the large centralised systems also
use concentrator PV technology (see Chapter 11), where the climatic conditions
are favourable.

Environmental impact
Whereas the stand-alone system is designed to meet a specific load, the motivation
for the grid-connected system is ultimately to reduce consumer energy costs and
to reduce the environmental impact of electricity production. There is an environ-
mental impact from the manufacture and disposal of PV modules and the rest of
the system components, but this is substantially lower per unit of electricity across
the system lifetime than for fossil-fuel-based generation (IPCC, 2012). Further-
more, it can be expected to reduce consistently over time, due to reducing energy
inputs in manufacturing and reducing carbon content in relation to those inputs, as
more renewable energy technologies are incorporated into the grid. The compari-
son of impact between PV systems and other renewable technologies depends on
assumptions made about the system design and location, so would require a fuller
treatment than is possible here. It can be concluded that PV systems, as part of a
portfolio of renewable technologies, can contribute significantly to the reduction
of carbon emissions resulting from electricity provision.
As the use of PV systems increases, it is also necessary to address the issue of
how to deal with the system components at the end of the system life. Photovoltaic
systems are generally straightforward to decommission and the industry, especially
in Europe, has now begun to put in place schemes for the collection and recycling
of modules at the end of their life. Within Europe, PV equipment comes under
the Waste Electrical and Electronic Equipment Directive (European Union, 2012)
and this requires the suppliers to provide a route for customers to return their used
equipment. It has been demonstrated that a large proportion of the PV module can
be recycled, with materials either being reused for manufacturing new modules
(e.g. refurbished crystalline silicon cells) or in other sectors (e.g. glass).

Network development
As the cost of PV electricity reduces towards the cost of conventionally generated
electricity in several countries, one of the technical challenges to be faced is how
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 609

Photovoltaic Modules, Systems and Applications 609

to deal with large amounts of PV electricity being fed into the electricity network,
whether from distributed or centralised systems. The output from the PV system is
variable, depending on weather conditions. The output profile can be modified by
introducing storage, whether local or central, and, if sufficient capacity is installed,
by curtailment of output power to match requirements. The use of the inverter to
provide both storage and reactive power has been discussed earlier in the chapter.
Clearly, the network also has to include generation capacity to meet demand when
there is no solar output.
The ability to control the supply of electricity via the grid where there is a
high penetration level of variable renewable technologies (e.g. wind, PV) requires
knowledge of likely outputs. This is leading to advances in forecasting techniques
for system output over short and medium timescales and in the communication
systems and models required to make use of those forecasts.

12.5 Conclusions
Photovoltaic cells have social and commercial value when they are used in a system
to provide a service, whether that is the provision of electricity to meet a specific
load or to contribute to the power network serving a multitude of loads. This chapter
has given a brief overview of the technical considerations that allow the cells to
provide such a service, and of the current applications of PV systems.
Photovoltaic cells may be incorporated directly into a product and add value
to that product to the extent that their use is commercially viable. In most cases,
however, the cells are contained in a PV module, interconnected to give a required
output depending on the application. The module structure also protects the cells
from damage in transportation, installation and use. This chapter has described
the construction of PV modules and their quality assurance testing, designed to
provide a product with an assured output, reliability and lifetime when operating
in varied climatic conditions. It is these developments in module performance that
have provided the basis for the expanding market for PV throughout the world.
While the current module construction is similar in concept to that developed over
30 years ago, consistent developments in manufacturing have reduced the material
and energy requirements in manufacture and reduced the electrical losses in moving
from the individual cell to the module. Modules have also become larger and more
varied to address different user requirements. As new cell types develop and new
applications are found, PV modules will surely continue to evolve. Nevertheless,
the requirements for lifetime and durability will remain.
This chapter has also discussed the range of other equipment needed in PV
systems to give optimal operation in both stand-alone and grid-connected systems.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 610

610 N. M. Pearsall

Here, it was possible to give only a brief overview of the equipment and the design
criteria, especially in relation to the development of inverters. As the applications
of PV systems become more widespread, especially in relation to connection to
the grid, so the BOS components are also being developed to address specific
market needs such as grid support. The interested reader is referred to the specialist
journal and conference papers for the latest information, as well as to manufacturer
information on the latest commercial products.
Photovoltaic systems are clearly well suited to the provision of electrical power
at locations that are remote from the electricity grid and where reliability is a key
requirement. This is especially true where only small amounts of power are needed
to bring major improvements in quality of life, e.g. for rural medical applications.
The main issues are the correct design of the system, the education of the user
to ensure the system is operated correctly, and the provision of financial mech-
anisms to allow the purchase of the systems (which is beyond the scope of this
chapter).
The great majority of the installed capacity of PV systems around the world is
grid-connected, with a market share estimated to be >98% at the end of 2011 (IEA
Photovoltaic Power Systems Programme, 2012). The reduction in PV module price
over the last few years and the simplicity of their installation and operation has
meant that small, distributed systems have become popular. The market promotion
schemes introduced by several countries have also encouraged the installation of
centralised PV systems and capacity has increased year on year, as described in
Chapter 13.
While PV systems only provide a very small percentage of networked elec-
tricity at the moment, there are a few locations, particularly in southern Germany,
where there is a high penetration of systems. Under the correct conditions, it has
been demonstrated that the load can be met solely from a combination of wind
and solar electricity. This situation brings challenges in relation to control and
variability of supply, but also provides the stimulus for the developments in sys-
tem hardware and output forecasting that will eventually allow PV to be a major
contributor to the world’s electricity supply.

References
Bletterie, B., Bründlinger R. and Lauss G. (2011), ‘On the characterisation of PV inverters’
efficiency — introduction to the concept of achievable efficiency’, Progr. Photovoltaics
19, 423–435.
European Union (2012), Directive 2012/19/EU of the European Parliament and of the
Council on Waste Electrical and Electronic Equipment (WEEE), Official Journal of
the European Union, 24 July.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch12 page 611

Photovoltaic Modules, Systems and Applications 611

Huld T., Šúri M. and Dunlop E. D. (2008), ‘Comparison of potential solar electricity output
from fixed-inclined and two-axis tracking photovoltaic systems in Europe’, Progr.
Photovoltaics 16, 47–59.
IEA Photovoltaic Power Systems Programme (2012), Trends in Photovoltaic Applications:
Survey Report of Selected IEA Countries between 1992 and 2011, Report IEA-PVPS
T1-21:2012.
IEC (2007), Solar Photovoltaic Energy Systems — Terms, Definitions and Symbols, IEC/TS
61836:2007.
IEC (2005), Crystalline Silicon Terrestrial Photovoltaic (PV) Modules — Design Qualifi-
cation and Type Approval, IEC 61215:2005.
IEC (2008), Thin-film Terrestrial Photovoltaic (PV) Modules — Design Qualification and
Type Approval, IEC 61646: 2008.
IPCC (2012), Renewable Energy Sources and Climate Change Mitigation, Special Report of
the Intergovernmental Panel on Climate Change, Cambridge University Press, p. 732.
Lloret A., Aceves O., Andreu J., Merten J., Puigdollers J., Chantant M., Eicker U. and
Sabata L. (1997), ‘Lessons learned in the electrical system design, installation and oper-
ation of the Mataró Public Library’, Proc. 14th European Photovoltaic Solar Energy
Conf., Barcelona, Spain. H. S. Stephens & Associates, Bedford, pp. 1659–1664.
May 2, 2013 14:6 BC: 8831 - Probability and Statistical Theory PST˙ws

This page intentionally left blank

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 613

CHAPTER 13

THE PHOTOVOLTAIC BUSINESS:

MANUFACTURERS AND MARKETS
ARNULF JÄGER-WALDAU
Senior Scientist, Renewable Energy Unit
Institute for Energy and Transport
EU Joint Research Centre, Ispra, Italy
[email protected]

Viewed in the light of the world’s growing power needs, these gadgets are toys.
But so was the first flea-power motor built by Michael Faraday over a century ago
— and it sired the whole gigantic electrical industry.
From Prospects for Solar Power, Harland Manchester, Reader’s Digest, June 1955.

13.1 Introduction
Over the last decade photovoltaics has been one of the fastest-growing industries,
with a compound annual growth rate (CAGR) of over 50%. Solar cell and module
production have increased by more than an order of magnitude during this decade
and despite the negative impacts of the on-going economic crisis since 2009, cumu-
lative installed capacity of photovoltaic (PV) systems surpassed 100 GW1 at the end
of 2012. In 2013, new installed photovoltaic capacity could for the first time surpass
new installed wind-power capacity. This is mainly driven by rapid market expan-
sion in Asia, which will be the largest market, surpassing Europe. Photovoltaics is
on the right track to become a major electricity source within the next decade.

13.1.1 Global production data

Estimates of global cell production2 in 2013 vary between 35 GW and 42 GW. The
uncertainty in these data is due to the highly competitive market environment, as

1 All production and installation figures cited in W in this chapter are actually Wp.
2 Solar-cell production capacities mean: in the case of wafer silicon-based solar cells, only the cells; in
the case of thin-films, the complete integrated module. Only those companies that actually produce the
active circuits (solar cells) are counted; companies that purchase these circuits and make cells are not
counted.

613
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 614

614 A. Jäger-Waldau

45
Rest of World
40 United States
Malaysia
35
Japan
Annual Production [GW]

Europe
30
Taiwan

25 PR China

20

15

10

0
2005 2006 2007 2008 2009 2010 2011 2012 2013e

Figure 13.1 World PV cell/module production from 2005 to 2012, and 2013 (estimated).
Data source: Photon International (2012), RTS Corporation (2013), PV News (2013) and
the author’s analysis.

well as the fact that some companies report shipment figures, while others report
sales and others still production figures. The data presented, collected from stock
market reports of listed companies, market reports and colleagues, were compared
to various data sources and this led to an estimate of 38.5 GW (Fig. 13.1), repre-
senting a moderate increase of about 10% compared with 2012; as manufacturing
plant utilisation began to improve in the second quarter of 2013 another moderate
increase of about 10% is expected for 2013.
Total PV production has increased almost by two orders of magnitude since
2000, and the CAGR over the last decade has been above 50%. The most rapid
growth in annual production over the last five years has occurred in Asia, where
China and Taiwan together now account for more than 70% of worldwide produc-
tion.
Publicly traded companies manufacturing products along the value chain,
installing photovoltaic electricity systems or offering related services, have
attracted a growing number of private and institutional investors. In 2012, world-
wide new investments in clean energy decreased by 11% compared with 2011 but
were still very considerable at $269 billion ( 207 billion3 ), including $30.5 bil-
lion ( 23.5 billion) corporate and government research and development spending
(Bloomberg New Energy Finance, 2013a; PEW Charitable Trusts, 2013). In 2012,

3 Exchange rate: 1 = 1.30 US$.

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 615

The Photovoltaic Business: Manufacturers and Markets 615

clean energy markets outside the Group of 20 (G20) grew by more than 50% to
surpass $20 billion ( 15.4 billion), while investments in the G20 countries dropped
by 16% to $218 billion ( 167.7 billion). Despite the overall decline in investments,
the decrease of renewable energy system prices more than compensated this and
allowed these investments to install a record 88 GW of new clean-energy genera-
tion capacity, bringing the total to 648 GW, capable of producing more than 1500
TWh of electricity per annum. This corresponds to 64% of the electricity generated
per annum by nuclear power plants worldwide.
For the third year in a row, solar power attracted the largest amount of new
investment into renewable energies. Despite a 9% decline in solar-energy invest-
ments, solar power attracted $137.7 billion ( 105.9 billion), which is 57.7% of all
new renewable energy investments (Bloomberg New Energy Finance, 2013b).
In contrast to Europe and the Americas, where new investments in renewable
energy decreased, new investments continued to increase in Asia/Oceania, reaching
$101 billion ( 77.7 billion) in 2012. Europe took the second place with $62.1 billion
( 47.8 billion), followed by the Americas with $50.3 billion ( 38.7 billion) (PEW
Charitable Trusts, 2013). With a 20% increase, China became the largest investor
in renewable energy with $65.1 billion ( 50.1 billion) followed by the USA with
$35.6 billion ( 27.4 billion) and Germany with $22.8 billion ( 17.5 billion). The
country with the biggest change in 2012 was South Africa, which moved up to
ninth place with $5.5 billion ( 4.2 billion).

13.1.2 Prices
The existing overcapacity in the solar industry has led to continuous downward
price pressure along the value chain and resulted in a reduction of spot market
prices for polysilicon, solar wafers and cells, as well as solar modules. Photovoltaic
module prices have decreased in price by 80% since 2008, and by 20% in 2012
alone (Bloomberg New Energy Finance, 2013b). These rapid price declines are
putting all solar companies under enormous pressure and access to fresh capital is
key to survival. It is believed that this situation will continue until at least 2015,
when the global PV market should exceed 50 GW of new installations. The increase
of polysilicon spot prices and the levelling of module prices since the beginning
of 2013 indicate that some production capacity has been taken out, and that prices
may stabilise for some time until they are back on the learning-curve slope. It
should be noted that PV system hardware costs are priced more or less the same
worldwide, and the so-called soft costs, which consist mainly of financing, permit
and labour costs, and installer/system integrator profits, are the main reason for the
significant differences which are still observed.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 616

616 A. Jäger-Waldau

The continuation of the difficult worldwide financial situation and the fact that
support schemes are changed with little notice, undermining long-term investor
confidence, make project financing more difficult as risk premiums are added. On
the other hand, the declining module and system prices have already opened new
markets that offer perspectives for further growth of the industry — at least for those
companies with the capability to expand and reduce their costs at the same pace.
Despite the problems of individual companies, business analysts are confident
that the industry fundamentals, as a whole, remain strong and that the overall
photovoltaics sector will continue to experience significant long-term growth. For
example, in its second Medium-Term Renewable Energy Market Report, published
in July 2013, the IEA increased its predicted installed capacity to 300 GW of
cumulative PV installations by 2018 (IEA, 2013a).
Market predictions for new PV system installations in 2013 have been
upgraded regularly during the year. They range between 31 GW from Solarbuzz,
33.1 GW in the Bloomberg conservative scenario, >35 GW from IHS research and
39.6 GW in the Bloomberg optimistic scenario (Bloomberg New Energy Finance,
2012; IHS Global Demand Tracker, 2013; Solarbuzz, 2013). The wide spread of
analyst numbers derives from the different methodologies, and whether the given
numbers represent constructed/installed systems or actual grid-connected systems.
For 2014, analysts expect a further increase to over 40 GW, mainly driven by grow-
ing Asian markets. Even in the optimistic forecasts, massive overcapacities in cell
and module manufacturing will continue to exist. Depending how capacities are
calculated, overcapacity estimates for 2013 range between 60 and 70 GW.
Despite a number of bankruptcies and companies idling production lines, or
even permanent closure of production lines, the number of new entrants into the
field, notably large semiconductor or energy-related companies, has overcompen-
sated this in the past. However, the rapid changes in the sector and the difficult
financing situation make a reasonable forecast of future capacity developments
very speculative. The consequence is the continuation of price pressure in an over-
supplied market. This will accelerate the consolidation of the photovoltaics industry
and spur even more mergers and acquisitions.
The existing overcapacity is a result of very ambitious investments dating
back to 2010, triggered by the more than 150% growth of the PV market in that
year. This led to a peak in capital spend on manufacturing equipment of about
$14 billion ( 10.8 billion) in 2011. Since then, equipment spending has declined
dramatically and will probably hit the bottom with around $1–2 billion ( 0.77–1.54
billion) invested in 2013 before a moderate recovery from 2014 onwards will be
possible. This development has had a serious effect on equipment manufacturers,
all of which now need a new strategy for the PV industry. Companies with no
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 617

The Photovoltaic Business: Manufacturers and Markets 617

significant business segment outside the PV supply chain were hit the hardest,
and some of them are struggling to survive the slump until the predicted recovery
kicks in.

13.1.3 Market share by cell/module type

Current solar-cell technologies are well established and provide a reliable product,
with sufficient efficiency and energy output for at least 25 years of lifetime. This
reliability, and the increasing likelihood of electricity interruption from grid over-
loads, as well as the rise in electricity prices from conventional energy sources,
add to the attractiveness of photovoltaic systems.
With over 85% of current production, wafer-based crystalline silicon remains
the dominant market technology, discussed by Martin Green in Chapter 3. Projected
silicon production in 2013 ranges between 290,000 metric tons (Bloomberg New
Energy Finance, 2013b) and 409,690 metric tons (Ikki, 2013a). It is estimated that
about 27,000 metric tons will be used by the electronics industry. The resulting
solar-cell production will depend on the amount of material used per Wp; the
current worldwide average is about 5.6 g/Wp.
Estimates of the thin-film share of the market by different analysts and market
institutes range from 9 to 12%. This is shared by the three main thin-film technolo-
gies: 3.5–4% a-Si/µ-Si, ∼5% CdTe and ∼2.5% CIGS, described in Chapters 4, 5
and 6, respectively.
As in other areas of technology, new products will enter the market, enabling
further cost reduction. Concentrated photovoltaics (CPV), described in Chapter 11,
is an emerging market. There are two main CPV options — either high concen-
tration >300 Suns (HCPV), or low to medium concentration with concentration
factors of 2 to about 300. In order to maximise the benefits of CPV, high direct
normal irradiation (DNI) is required and this is only found in a limited geographical
range — the Sun Belt of the Earth. Dye cells, discussed in Chapter 9, are entering
the market as well. The development of these technologies is accelerated by the
growth of the PV market as a whole, but economic conditions for their market
entry are getting more and more challenging as silicon module and system prices
are continuing to decrease.
It can be concluded that, in order to maintain the extremely high growth rate
of the photovoltaic industry, different technology pathways have to be pursued at
the same time. As the cost share of solar modules in a PV system has declined to
below 40% in a residential system and below 50% in a commercial system, the
non-technology costs and ‘soft costs’ have to be targeted for further significant
cost reductions.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 618

618 A. Jäger-Waldau

With an increasing share of PV electricity in the grid, the economics of inte-

gration become more and more important and it is urgent to focus attention on
issues such as:

• Development of new business models for the collection, sale and distribution
of photovoltaic electricity, e.g. development of bidding pools at electricity
exchanges, virtual power plants with other renewable power producers and
storage capacities.
• Adaptation of regulatory and legal procedures to ensure fair and guaranteed
access to the electricity grid and market.

The cost of PV-generated electricity has dropped to less than 0.05 /kWh, and
the main cost component is to transport the power from the module to where it is
needed. Therefore innovative and cost-effective overall electricity system solutions
for the integration of PV electricity are needed to realise the vision of PV as a major
electricity source for everybody everywhere. More effort is needed to optimise the
non-PV costs and greater public support, especially as regards regulatory measures,
is crucial.

13.2 The photovoltaic market

After the worldwide photovoltaic market more than doubled in 2010, the market
grew again by almost 30% in 2011, and then another 11% in 2012, despite difficult
economic conditions. The 2010 market volume of 20.9 GW includes those systems
in Italy, which were reported under the second ‘conto energia’ and installed, but
connected only in 2011. There is uncertainty about the actual installation figures
in China. The Chinese National Energy Administration reported a cumulative
installed capacity of 7 GW at the end of 2012, whereas most other market ana-
lysts cite figures between 8 and 8.5 GW (National Energy Administration, 2013).
The stronger than expected market in Germany and the marked increase of instal-
lations in Asia and the USA resulted in a new installed capacity of about 30 GW
in 2012; increases to about 37 GW and 45 GW are expected for 2013 and 2014,
respectively, as shown in Fig. 13.2. This comprises mostly the grid-connected
photovoltaic market.
To what extent the off-grid and consumer product markets are included is not
clear, but it is believed that a substantial part of these markets are not accounted
for as it is very difficult to track them. Figure 13.3 shows the cumulative capacity
installed in the period 2000 to 2014. With a cumulative installed capacity of over
69 GW, the European Union (Rest of Europe + Italy + Spain + Germany) had the
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 619

The Photovoltaic Business: Manufacturers and Markets 619

50

45 Rest of Europe
Annual Photovoltaic Installations/GWp

Italy
40 Spain
Germany
35 Rest of World
China
30 United States
Japan
25

20

15

10

0
2000 2005 2006 2007 2008 2009 2010 2011 2012 2013e 2014e

Figure 13.2 Annual photovoltaic installations from 2000 to 2012 and estimates for 2013
and 2014. Data sources: European Photovoltaic Industry Association, 2013; National Energy
Administration, 2013; Systèmes Solaires, 2013, and author’s analysis.

200
Cumulative Photovoltaic Installations/GWp

180 Rest of Europe

Italy
160 Spain
Germany
140 Rest of World
China
120
United States
Japan
100

80

60

40

20

0
2000 2005 2006 2007 2008 2009 2010 2011 2012 2013 e 2014e

Figure 13.3 Cumulative photovoltaic installations from 2000 to 2012 and estimates for
2013 and 2014. Data sources: European Photovoltaic Industry Association, 2013; National
Energy Administration, 2013; Systèmes Solaires, 2013, and author’s analysis.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 620

620 A. Jäger-Waldau

largest portion of the total worldwide 100 GW of installed capacity at the end of
2012.

13.2.1 Asia and Pacific Region

The Asia and Pacific Region continued its upward trend in PV installations. There
are a number of reasons for this, ranging from declining system prices, heightened
awareness and favourable policies to the sustained use of solar power for rural
electrification projects. Countries such as Australia, China, India, Indonesia, Japan,
Malaysia, South Korea, Taiwan, Thailand, the Philippines and Vietnam show a
very positive upward trend, thanks to increasing government commitment to the
promotion of solar energy and the creation of sustainable cities. The introduction
or expansion of feed-in tariffs (FiTs) is expected to be an additional big stimulant
for on-grid solar PV system installations for both distributed and centralised solar
power plants in countries such as Australia, Japan, Malaysia, Thailand, Taiwan and
South Korea.
In 2012, about 8.5 GW of new PV electricity generation systems was installed
in the region, which corresponds to a 60% growth as compared with 2011. The
largest market was China with 3.7 GW, followed by Japan with 2.3 GW and Aus-
tralia with over 1 GW. For 2013, a market increase to about 20 GW is expected,
driven by the major market growth in China (∼6–8 GW), India (>1 GW), Japan
(6.9–9.4 GW), and Malaysia and Thailand. For 2014, the market expectations
exceed 25 GW.

Australia
In 2012, slightly more than 1 GW of new solar photovoltaic electricity systems were
installed in Australia, bringing the cumulative installed capacity of grid-connected
PV systems to 2.45 GW. Photovoltaic electricity systems accounted for 36% of all
new electricity generation capacity installed in 2011. As in 2011, the 2012 market
was dominated by grid-connected residential systems, which accounted for more
than 90% of the total installed. The average PV system price for a grid-connected
system fell from 6 AUD/Wp (4.29 /Wp4 ) in 2010 to 3.9 AUD/Wp (3 /Wp5 )
in 2011 and 3.0 AUD/Wp (2.15 /Wp4 ) in early 2013. Due to this, the cost of
PV-generated electricity has reached, or is even below, the average residential
electricity price of 0.27 AUD/kWh (0.19 /kWh).

4 Average exchange rate 2010 and 2013: 1 = 1.40 AUD.

5 Average exchange rate 2011: 1 = 1.30 AUD.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 621

The Photovoltaic Business: Manufacturers and Markets 621

In 2012, PV electricity systems generated 2.37 TWh, which was about 1% of

total Australian electricity supply. The total renewable electricity share was 13.34%
and this should increase to 20% by 2020. For 2013, a PV market of 750 MW is
forecast.
Within two years, Australia has installed almost 2 GW of PV capacity, and
about 10% of residential buildings now have a PV system. At first, most installations
took advantage of the incentives under the Australian Government’s Renewable
Energy Target (RET) mechanisms and feed-in tariffs in some states or territories.
At the beginning of 2011, eight out of the eleven Australian federal states and
territories had introduced eleven different kinds of FiT schemes, mainly for systems
under 10 kWp. All except three of these schemes had built-in caps, which were
partly reached that year, so that in 2012 only six schemes were available for new
installations.

India
Market estimates for PV systems in India in 2012 vary between 750 and 1000 MW,
because some statistics cite the financial year and others the calendar year. Accord-
ing to the Ministry of New and Renewable Energy (MNRE), total capacity at the
end of financial year (FY) 2012/136 was 1.9 GW grid-connected and 125 MW off-
grid capacity. The Indian Jawaharlal Nehru National Solar Mission (JNNSM) was
launched in January 2010, and it was hoped that this would give impetus to the
grid-connected market. This Mission aims to make India a global leader in solar
energy and envisages an installed solar generation capacity of 20 GW by 2020,
100 GW by 2030 and 200 GW by 2050. After only a few MW had been installed
in 2010, installations in 2011 slowly picked up, but the majority of the JNNSM
projects will come online from 2015 onwards. Market expectations for 2013 vary
between 1.00 and 1.35 GW (Bridge to India, 2013; Mercom Capital Group, 2013).

Israel
Three years after the introduction of an FiT in 2008, Israel’s grid-connected PV
market saw about 60 MW of capacity newly connected in 2012. One of the main
drives behind the development of solar energy is energy security, and Israel plans
to have about 1 GW of solar PV systems installed by the end of 2014. In August
2012, about 215 MW had been built and a further 300 MW approved (Bloomberg
New Energy Finance, 2012). The FiTs depend on the system size segment and
have individual caps. Market expectations for 2013 are in the range 150–200 MW.

6 The Indian financial year ends 31 March.

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 622

622 A. Jäger-Waldau

Japan
The Japanese market experienced significant growth in 2012, increasing its domes-
tic shipments to 2.47 GW. Cumulative installed capacity increased by about 1.7 GW
to reach 6.6 GW at the end of 2012 (IEA, 2013b). By the end of February 2013,
more than 5.5 GW had received approval under the new national feed-in scheme,
which started in July 2012. The market outlook for Japan has been raised and is now
between 6.9 and 9.4 GW new installed capacity by the end of 2013 (Bloomberg
New Energy Finance, 2013d).
The consequence of the accident at the Fukujima Daiichi Nuclear Power Plant
in March 2011 was the reshaping of the country’s energy strategy. For PV power,
an official target of 28 GW was set for 2020. In July 2012, a Ministry for Economy,
Trade and Industry (METI) panel published its long-awaited plan to reform the
country’s power market. This aims to increase the share of renewable power supply
from 11% in 2011 to 25–35% by 2030.
Until 2010, residential rooftop PV systems represented about 95% of the
Japanese market. In 2011, due to a change in permit policy, large ground-mounted
systems, as well as large commercial and industrial rooftop systems, increased
their market share to about 20%.
In June 2012, METI finally issued the Ministerial Ordinances for the new FiTs
for renewable energy sources and these were adjusted in March 2013. The tariff
for commercial installations (total generated power) larger than 10kWp is
36 per
kWh for 20 years and
38 per kWh7 for residential installations (surplus power)
smaller than 10 kWp for 10 years, starting from 1 April 2013 (Ikki, 2013a).

People’s Republic of China

In 2012, the Chinese PV market grew to 3.7 GW, bringing the cumulative installed
capacity to about 7 GW (NEA, 2013). About 3.3 GW of this capacity was connected
to the grid at the end of 2012 (SERC, 2013). This is a 600% growth compared with
2010, but still represents only 16% of total Chinese PV production. In January 2013,
the Chinese National Energy Administration (NEA) announced an ambitious target
of 10 GW of new domestic PV installations during 2013 (NEA, 2013). A further
increase can be expected due to the recent change of the solar energy target for
2015, which was increased to 41 GW for PV and solar thermal electricity combined
by the NEA (Bloomberg New Energy Finance, 2013a).
According to the 12th Five-Year Plan, which was adopted on 14 March 2011,
China intends to cut its carbon footprint and become more energy efficient. The

7 Exchange rate: 1 = 130
.

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 623

The Photovoltaic Business: Manufacturers and Markets 623

Plan targets are 17% lower carbon dioxide emissions and 16% lower energy
consumption per unit of gross domestic product (GDP) by 2015. The total invest-
ment in the power sector under the Plan is expected to reach $803 billion ( 618
billion), divided into $416 billion ( 320 billion), or 52%, for power generation, and
$386 billion ( 298billion) to construct new transmission lines and other improve-
ments to China’s electrical grid.
In August 2012, the NEA released China’s new renewable energy five-year
plan for 2011–2015 (NEA, 2012). The new goal of the NEA calls for renewable
energy to supply 11.4% of the national energy mix by 2015. To achieve this goal,
renewable power generation capacity has to be increased to 424 GW. Hydro-power
is planned to be the main source, with 290 GW including 30 GW pumped storage,
followed by wind with 100 GW, solar with 21 GW (this target was later increased
to 41 GW) and biomass with 13 GW.
The plan estimated new investments in renewable energy of CNY 1.8 trillion
( 222 billion) between 2011 and 2015. China aimed to add a total of 160 GW of new
renewable energy capacity during the period 2011–15, in the form of 61 GW hydro,
70 GW wind, 21 GW solar8 (10 GW small distributed PV, 10 GW utility scale PV
and 1 GW solar thermal power), and 7.5 GW biomass. For 2020, the targets were
set as 200 GW for wind, 50 GW for solar (27 GW small distributed PV, 20 GW
utility-scale PV and 3 GW solar thermal power) and 30 GW for biomass.
These required investment figures to be in line with a World Bank report stating
that China needs additional investment of $64 billion ( 49.2 billion) annually over
the next two decades to implement an energy-smart growth strategy (World Bank,
2010). However, according to the report, the reductions in fuel costs through energy
savings could largely pay for these additional investment costs. At a discount rate of
10%, the annual net present value (NPV) of the fuel cost savings from 2010 to 2030
would amount to $145 billion ( 111.5 billion), which is about $70 billion ( 53.8 bil-
lion) more than the annual NPV of the additional investment costs required.
On 24 February 2012, the Chinese Ministry of Industry and Information Tech-
nology published its industrial restructuring and upgrading plan (2011–2015) for
the photovoltaic industry (MIIT, 2012). In this document, the Ministry stated that
by 2015 it expects to support only backbone enterprises, which should produce a
minimum of 50000 MT polysilicon, or 5 GW of solar cell or module production.
The Plan also projects a cost reduction of electricity generated with PV systems
to 0.8 CNY/kWh (0.098 /kWh9 ) by 2015 and 0.06 CNY/kWh (0.074 /kWh) by
2020.

8 This has been updated in the meantime.

9 Exchange rate 2012: 1 = 8.1 CNY.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 624

624 A. Jäger-Waldau

South Korea
In 2012, about 250 MW of new PV systems were installed in South Korea, bringing
the cumulative capacity to a total of 981 MW (IEA, 2013c). Since January 2012,
Korea’s Renewable Portfolio Standard (RPS) officially replaced its earlier FiTs. For
2013, the RPS set-aside quota for PV was set at 450 MW, and this should increase to
1.2 GW in 2016. The result is an annual target of 230 MW in 2013, 240 MW in 2014,
250 MW in 2015 and 260 MW in 2016. Under the RPS, income for power generated
from renewable energy sources is a combination of the wholesale system marginal
electricity price and the sale of Renewable Energy Certificates (RECs); certificate
sales in the second half of 2012 were around 40000 KRW/MWh (0.026–0.035
/MWh10 ). These RECs are multiplied by an REC multiplier, varying between
0.7 for ground-mounted free-field systems to 1.5 for building-adapted systems, to
reflect the different costs of the different systems.
The new RPS programme obliges power companies with at least 500 MW of
generating capacity to increase their renewable energy mix from not less than 2%
in 2012 to 10% by 2022. The renewable energy mix in the Korean RPS is defined
as the proportion of renewable electricity generation to the total non-renewable
electricity generation.

Taiwan
In June 2009, the Taiwan Legislative Yuan gave its final approval to a Renewable
Energy Development Act to bolster the development of Taiwan’s green energy
industry. The goal was to increase Taiwan’s renewable energy generation capacity
by 6.5 GW to a total of 10 GW within 20 years. Promotion of all types of renewable
energy was expanded: total installed capacity of 9952 MW (accounting for 14.8% of
total power generation installed capacity), with new installed capacity of 6600 MW,
has been planned by 2025 so that the goal set by the Act can be achieved five years
early. By 2030, the target will be further increased to 12.5 GW, accounting for 16.1%
of total power generation installed capacity and capable of generating 35.6 billion
kWh of electricity. This is equivalent to the annual electricity consumption of 8.9
million households (accounting for 78% of the number of households consuming
electricity nationwide).
Between 2009 and 2012, a total capacity of 194 MW was installed, bringing
the total installed capacity to 222 MW at the end of 2012 (Ministry of Economic
Affairs, 2013). The FiTs in the first half of 2013 for rooftop systems were 8.4

10 Exchange rate: 1 = 1, 420 KRW.

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 625

The Photovoltaic Business: Manufacturers and Markets 625

NT$/kWh (0.215 /kWh11 ) for systems up to 10 kW, 7.54 NT$/kWh (0.193 /kWh)
for systems between 10 and 100 kW, 7.12 NT$/kWh (0.183 /kWh) for systems
between 100 and 500 kW and 6.33 NT$/kWh (0.162 /kWh) for systems larger than
500 kW. Ground-mounted systems had a tariff of 5.98 NT$/kWh (0.153 /kWh).
For the second half of 2013, a tariff reduction of 2.5% to 6.1% was foreseen to
reflect declining system prices.
The installation targets for 2013 were increased twice and are currently at
175 MW. This is in line with Taiwan’s new Million Solar Rooftop Programme,
which aims to achieve installed capacity of 610 MW by 2015 and 3.1 GW by
2030. However, the increased installation target for 2013 represents only 3.5% of
Taiwan’s total 2012 PV production volume, reflecting the country’s strength as a
PV exporter.

Thailand
Thailand enacted a 15-year Renewable Energy Development Plan (REDP) in early
2009, setting the target of increasing the renewable energy share of the final energy
consumption of the country to 20% by 2022. Besides a range of tax incentives,
PV systems are eligible for a feed-in premium or ‘adder’ for a period of ten years.
The original 8 THB/kWh12 (0.182 /kWh) ‘adder’ facilities in the three south-
ern provinces, and those replacing diesel systems, are eligible for an additional
1.5 THB/kWh (0.034 /kWh. This was reduced to 6.5 THB/kWh (0.148 /kWh)
for those projects not approved before 28 June 2010. The original cap of 500 MW
was increased to 2 GW at the beginning of 2012, due to the high oversubscription
of the original target. In addition to the ‘adder’ programme, projects are now being
developed with power purchase agreements (PPAs).
At the end of 2012, grid-connected PV systems of about 360 MW total capacity
were operational, of which 210 MW were installed in that year (IEA, 2013c). In
September 2012, projects with around 1.8 GW capacity had signed PPAs, projects
with 76 MW already had a letter of intent (LOI) and another 925 MW of projects
were waiting for an LOI (Kruangam, 2013).

13.2.2 Emerging markets

Bangladesh
In 1997, the government of Bangladesh established the Infrastructure Development
Company Limited (IDCOL) to promote economic development in Bangladesh.

11 Exchange rate: 1 = 39 NT$.

12 Exchange rate: 1 = 44 THB.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 626

626 A. Jäger-Waldau

In 2003, IDCOL started its Solar Energy Programme to promote the dissemina-
tion of solar home systems (SHSs) in the remote rural areas of Bangladesh, with
financial support from the World Bank, the Global Environment Facility (GEF),
the German Kreditanstalt für Wiederaufbau (KfW), the German Technical Coop-
eration (GTZ), the Asian Development Bank and the Islamic Development Bank.
By April 2013, the programme had seen more than 2 million SHSs installed in
Bangladesh (World Bank, 2013). At the time of writing, the installation rate is
more than 60,000 units per month. In 2011, the Asian Development Bank agreed
to provide financial support to Bangladesh for implementing the installation of
500 MW within the framework of the Asian Solar Energy Initiative (The Daily
Star, 2011; UNB Connect, 2011).

Indonesia
The development of renewable energy in Indonesia is regulated in the context of
national energy policy by Presidential Regulation No. 5/2006 (The President of the
Republic of Indonesia, 2006). This decree states that 11% of the national primary
energy mix in 2025 should come from renewable energy sources and the target
for solar PV was set at 1000 MW by 2025. By the end of 2011, about 20 MW of
solar PV systems had been installed, mainly for rural electrification purposes. In
2012, the Indonesian Ministry of Energy and Mineral Resources (ESDM) drafted
a Roadmap that foresaw installations of 220 MW between 2012 and 2015 (ESDM,
2012). A new policy to promote solar energy through auction mechanisms was
published in June 2013 (Bloomberg New Energy Finance, 2013a); how this new
policy will influence the market remains to be seen.

Kazakhstan
The development of renewable energy was one of the priorities of Kazakhstan’s
State Programme for Accelerated Industrial and Innovative Development for 2010–
2014. The main goal of this programme is to develop a new and viable economic
sector for growth, innovation and job creation. In addition, it drives the develop-
ment of renewable energy sources for the electricity sector in Kazakhstan and is
regulated by the Law on Support to the Use of Renewable Energy Sources, adopted
in 2009 (CIS Countries Legislation Database, 2009). In February 2013, the Kazakh
government decided to install at least 77 MW of PV by 2020 (Government of Kaza-
khstan, 2013). In 2011, JSC NAC Kazatomprom, jointly with a French consortium
headed by Commissariat à l’Energie Atomique et aux Energies Alternatives (CEA),
started the project Kaz PV with the aim of producing PV modules based on Kazakh
silicon (Kaz Silicon, 2011). The first project step was concluded in January 2013,
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 627

The Photovoltaic Business: Manufacturers and Markets 627

when a new 60 W PV module production plant was opened in Kazakhstan’s capital

city, Astana. In May 2013, Zhambilskie Electricheskie Seti LLP signed a Memo-
randum of Understanding with NanoWin Thin Film Tech to build a 60 MW CIGS
factory and a 5 MW solar plant in the country (Nanowin, 2013).

Malaysia
The Malaysia Building Integrated Photovoltaic (BIPV) Technology Application
Project was initiated in 2000, and by the end of 2009, a cumulative capacity of
about 1 MW of grid-connected PV systems had been installed under this pro-
gramme. The Malaysian government officially launched their GREEN Technology
Policy to encourage and promote the use of renewable energy for Malaysia’s future
sustainable development in 2009. By 2015, about 1 GW is planned to come from
renewable energy sources, according to the Ministry of Energy, Green Technology
and Water.
In April 2011, renewable energy FiTs were approved by the Malaysian Par-
liament, with a target of 1.25 GW installed by 2020. For the period December
2011–June 2014, a total quota of 125 MW was allocated for PV. The 2013 tariffs
were set by the Sustainable Energy Development Authority between 0.782 and
1.555 MRY/kWh (0.195 to 0.389 /kWh13 ), depending on the type and system
size. In addition, there is a small bonus for local module or inverter use. An annual
reduction in these tariffs is foreseen. As of 30 April 2013, 28.92 MW of PV systems
were operational under the new FiT scheme and another 141.58 MW had received
approval, and were in various stages of project planning or installation (SEDA,
2013). First Solar (USA), Hanwha Q Cells (Korea/Germany), Sunpower (USA),
and recently Panasonic (Japan), have set up manufacturing plants in Malaysia, with
more than 3.8 GW of total production capacity.

The Philippines
The Philippines’ Renewable Energy Law was passed in December 2008 (Philippine
Department of Energy, 2008). Under this law, the country had to double the energy
it derives from renewable energy sources within ten years. In June 2011, the Energy
Secretary unveiled a new Renewable Energy Roadmap, which aimed to increase
the share of renewables to 50% of total energy consumption by 2030, and to boost
renewable energy capacity from the 2011 level of 5.4 GW to 15.4 GW by 2030.
In early 2011, the country’s Energy Regulator National Renewable Energy Board
recommended a target of 100 MW of solar installations to be constructed in the

13 Exchange rate: 1 = 4.0 MRY.

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 628

628 A. Jäger-Waldau

country over the following three years. An FiT of 17.95 PHP/kWh (0.299 /kWh)14
was suggested, to be paid from January 2012 onwards. For 2013 and 2014, an
annual degression of 6% was foreseen. The initial period of the programme is
scheduled to end on 31 December 2014. In July 2012, the Energy Regulatory
Commission decided to lower the tariff in view of lower system prices to 9.68
PHP/kWh (0.183 /kWh15 ) and confirmed the degression rate. By the end of 2012,
about 2 MW of the 20 MW of installed PV systems were grid-connected. SunPower
had two cell-manufacturing plants outside of Manila, but closed down Fab. No 1
in early 2012. Fab. No 2, with a nameplate capacity of 575 MW, remains in action.

Vietnam
The National Energy Development Strategy of Vietnam was approved in December
2007. This gave priority to the development of renewable energy and included the
following targets: increase the share of renewable energies from negligible to about
3% (58.6 GJ) of the total commercial primary energy in 2010, to 5% in 2020, 8%
(376.8 GJ) in 2025, and 11% (1.5 TJ) in 2050. By the end of 2011, about 5 MW
of PV systems had been installed, mainly in off-grid applications. In May 2011,
the Indo-Chinese Energy Company broke ground in the central coastal Province
of Quang Nam for the construction of a thin-film solar panel factory with an initial
capacity of 30 MW and a final capacity of 120 MW. However, the company applied
for permission to delay the project in June 2012, with no new launch date set. In
January 2013, WorldTech Transfer Investment and the UAE-based Global Sphere
began work on a solar-panel manufacturing plant in the central province of Thua
Thien-Hue (Global Sphere, 2013). This plant is located in the Phong Dien Industrial
Park and the first phase of the project (60 MW) should be operational by June 2015.
The capacity is planned to be increased to 250 MW in a second phase.

13.2.3 Europe and Turkey

Market conditions for PV vary substantially from country to country across Europe.
This is due to different energy policies and public support programmes for renew-
able energies, and especially for PV, as well as the varying degrees of liberalisation
of domestic electricity markets. As shown in Fig. 13.4, overall growth in PV capac-
ity has been impressive, increasing 373-fold from 185 MW in 2000 to 69 GW at
the end of 2012 (European Photovoltaic Industry Association, 2013; Systèmes
Solaires, 2013).

14 Exchange rate 2011: 1 = 60 PHP.

15 Exchange rate 2012: 1 = 53 PHP.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 629

The Photovoltaic Business: Manufacturers and Markets 629

100000

2000
10000
2010
2012
1000
Cumulative installed [MW]

100

10

0.1

0.01

0.001
DE IT ES FR BE CZ UK EL BG SK AT DK NL PT SI LU SE MT CY FI RO HU PL LT LV IE EE HR TR

Figure 13.4 Cumulative installed grid-connected PV capacity in the European Union +

Candidate Countries (CC). Note that the installed capacities do not correlate with solar
resources.

The net growth of all renewable energy power generation capacity in Europe
and Turkey between 2000 and 2012 was 178 GW. This compares with 121 GW
of new gas-fired capacity and a reduction in capacity of generation from coal
(−12.7 GW), oil (−7.4 GW) and nuclear (−14.7 GW). Wind (96.7 GW) and PV
(69 GW) accounted for more than 93% of the renewable capacity. This net growth
of 178 GW to a total of 316 GW of renewable capacity increased the total share of
renewable power capacity in Europe and Turkey from 22.5% in 2000 to 33.9% in
2012.
A total of about 45 GW of new power capacity was connected in the European
Union (EU) in 2012 and 12.5 GW was decommissioned, resulting in 32.5 GW
of new net capacity, as shown in Fig. 13.5 (European Wind Association, 2013;
Systèmes Solaires, 2013). Photovoltaic electricity generation capacity accounted
for 16.8 GW, or 51.7%, of the new net capacity. Wind power was second with
11.7 GW (36%), followed by 5 GW (15.4%) gas-fired power stations, 1.3 GW (4%)
biomass, 0.8 GW (2.5%) solar thermal power plants, 266 MW (0.8%) hydro and
61 MW (0.2%) other sources. The net installation capacity for coal-fired, oil-fired
and nuclear power plants was negative, with decreases of 2.3 GW, 3.2 GW and
1.2 GW, respectively. The renewable share of new power installations was more
than 69% and more than 95% of new net capacity in 2012.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 630

630 A. Jäger-Waldau

Oil
Installed Capacity
Nuclear Decommissioned Capacity

others

Hydro

CSP

Biomass

Coal

Gas

Wind

PV

-6 -4 -2 0 2 4 6 8 10 12 14 16
New Installed or Decomissioned Capacity in 2012 [GW]

Figure 13.5 New connected or decommissioned electricity generation capacity in the EU

in 2012.

In the following sub-sections, the market development in some of the EU

Member States, as well as of Switzerland and Turkey, is described.

Belgium
Belgium showed another strong PV market performance in 2012, with new system
installations of 882 MW bringing the cumulative installed capacity to 2694 MW.
Photovoltaic power supplied 14% of Belgium’s residential consumption, or 2.8%
of the country’s total electricity needs. However, most of this capacity was again
installed in Flanders, where green certificates have been in place for 20 years.
The certificate system was suspended at the end of July 2012 and replaced by a
new regime for PV systems commissioned after 1 August 2012 (Belgisch Staats-
blad, 2012, 2013): for systems smaller than 250 kWp, the tariff was reduced from
0.23 /kWh to 0.21 /kWh, and for larger systems from 0.15 /kWh to 0.09 /kWh
for systems installed before the end of 2012. At the same time, the duration for
which the certificates could be claimed was reduced from 20 to 10 years.
Since 1 January 2013, the right to receive green certificates has been deter-
mined by the duration of the amortisation period — for PV it was 15 years
in 1H 2013 and the net metering scheme for systems below 10 kW continued.
A technology-dependent banding factor, which is set twice a year by the Flemish
Energy Agency, was introduced, so that one certificate is no longer for 1 MWh,
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 631

The Photovoltaic Business: Manufacturers and Markets 631

but now depends on the type of installation (RES LEGAL Europe, 2013). Since 1
January 2013 the value of the certificates has been 93.
In Wallonia, the green certificate scheme called SOLWATT was replaced by
a new scheme called QUALIWATT from 1 April 2013 (Government of Wallonia,
2013). The main change was that green certificates with a value of 0.065 /kWh
will only be granted until the time that the PV installation is fully reimbursed,
and their maximum term will be 10 years. In addition, for systems below 10 kW,
net metering continues and a progressive pricing system has been introduced. The
value of green certificates in the Brussels region is 65 and for PV systems there
is a multiplier of 2.2. In addition, the possibility of net metering exists for systems
below 5 kWp so long as the generated electricity does not exceed the in-house
electricity demand of the owner.

Bulgaria
A new Renewable Energy Source (RES) Act was approved in May 2011. This fixed
the FiT levels and resulted in new installations of around 110 MW, increasing the
total installed capacity to 134 MW by the end of 2011. At the end of 2012, 933 MW
of PV systems had been cumulatively installed. In March 2012, the Bulgarian
Parliament voted to revise the RES Act (Bulgarian State Gazette, 2012a). The
most significant change was that the price at which electricity would be purchased
was no longer fixed at the date of completion of installation, but at the date the
usage permit was granted.
In September 2012, the Bulgarian State Energy and Water Regulatory Com-
mission (SEWRC) published prices for the retroactive grid-usage fee, in accor-
dance with the Energy Act amendments adopted by the Bulgarian Parliament in
July 2012 (Bulgarian State Gazette, 2012b) for access to the transmission and dis-
tribution grid. For PV systems commissioned after 1 April 2010 and in 2011, the
fee amounted to 20% of the FiT. For systems commissioned during the first half
of 2012, the fee was 39% of the FiT, for those commissioned between 1 July and
31 August 2012 it was 5%, and after 1 September 2012, 1% of the respective FiT
applied (SEWRC, 2012a).
Modified FiTs, mandating a 34–54% reduction, depending on system type,
came into force on 1 June 2012. A further reduction ranging between 5 and 35%
came into force in October 2012 (SEWRC, 2012b).

Denmark
Due to the introduction of a net-metering system and the country’s high electricity
prices of 0.295 /kWh, 378 MW of PV systems were installed in Denmark in
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 632

632 A. Jäger-Waldau

2012. Following this rapid development, the regime was changed in November
2012 (Denmark, 2012). Under the new rules, full net metering was only possible
within one hour from when PV electricity was produced, and the excess electricity
exported to the grid was reimbursed at the rate of 1.30 DKK/kWh (0.174 /kWh16 )
in 2013. To account for decreasing PV system prices, this rate will decrease to 1.17
DKK/kWh (0.157 /kWh) in 2014 and to ∼1.00 DKK (0.174 /kWh) in 2015.
After ten years, the rate will be further reduced to about 0.60 DKK/kWh (0.080
/kWh).

France
In 2012, 1.08 GW of new PV systems were installed in France, which led to an
increase of the cumulative installed capacity to over 4 GW, including about 300 MW
in the French Overseas Departments. New PV installations in mainland France
accounted for 35% of total new electricity production capacity commissioned in
2012. Of the total capacity, residential systems smaller than 3 kWp represented 16%
or 0.64 GW, systems up to 250 kWp accounted for 40% or 1.6 GW, and systems
larger than 250 kWp added 44% or 1.76 GW.
In 2013, France had three different support schemes for PV. For systems up to
100 kWp, there is the FiT (allocation of 200 MW for residential and 200 MW for
commercial applications); for rooftop systems between 100 and 250 kWp there is
a ‘simplified’ call for tender with a volume of 120 MW for 2013; and for systems
larger than 250 kWp (large rooftop and ground-mounted systems) an additional
call for tender with a volume of 400 MW was issued, which received more than
1.9 GW of project applications.
In 2012, four PV tenders were offered and the average electricity sale price
proposed by the bidders fell from 229 /MWh during the first round to 194 /MWh
in the fourth one. New FiTs were published in February 2013, foreseeing an adap-
tation every three months (Ministère de l‘Écologie, 2013). Photovoltaic systems
with defined European content were eligible for a bonus of 5 to 10%.

Germany
Photovoltaics installed in Germany in 2012 increased slightly compared with 2011,
from 7.5 GW to 7.6 GW (Bundesnetzagentur, 2013). German market growth is
directly correlated to the introduction of the Erneuerbare Energien Gesetz (Renew-
able Energy Sources Act) in 2000 (EEG, 2000). This introduced a guaranteed FiT
for electricity generated from solar PV systems for 20 years and had a built-in

16 Exchange rate: 1 = 7.46 DKK.

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 633

The Photovoltaic Business: Manufacturers and Markets 633

annual decrease, which was adjusted over time to reflect the rapid growth of the
market and the corresponding price reductions. As only estimates of the installed
capacity existed before 2009, a plant registrar was introduced from 1 January 2009
onwards.
The German market performed strongly throughout 2012 with peaks of 1.2 GW
installed in March, 1.8 GW in June and 1 GW in September. The total installed
capacity at the end of the year was 32.7 GW. Since May 2012, the FiT has been
adjusted on a monthly basis depending on the actual installations during the
previous quarter. The fact that the tariff for residential PV systems (July 2013:
0.151 /kWh) is now below the rate conventional electricity consumers are pay-
ing (0.287 /kWh) makes the increase of self-consumption more attractive and is
opening new possibilities for the introduction of local storage (Bundesnetzagentur,
2013; Strategies Unlimited, 2013). Since 1 May 2013, the Kreditanstalt für Wieder-
aufbau (KfW) has offered low interest loans with a single repayment bonus of up
to 30% and a maximum of 600 per kW of storage for PV systems up to 30 kWp
(KfW, 2013). The maximum repayment bonus is limited to 3000 per system.

Greece
In 2009, Greece introduced a generous FiT scheme but this had a slow start until the
market accelerated in 2011 and 2012. In 2012, 687 MW of new PV systems were
installed — more than 1.5 times the 439 MW that had been cumulatively installed
up to the end of 2011. In April 2013, the total installed capacity surpassed 2 GW
(over 1.9 GW in mainland Greece and over 115 MW in the Islands) (HEDNO S.A.,
2013; LAGIE, 2013). On 10 May 2013 the Greek Ministry of Environment, Energy
and Climate Change announced retroactive changes in the FiT for systems larger
than 100 kWp and new tariffs for all systems from 1 June 2013 onwards.

Italy
Italy connected more than 3.5 GW during 2012, bringing cumulative installed
capacity to 16.4 GW by the end of the year (Gestore Servizi Energetici, 2013a).
The Quinto Conto Energia (Fifth Energy Bill) was approved by the Italian Council
of Ministers on 5 July 2012 (Italy, 2012). This set new half-yearly reductions of the
tariffs, and the annual expenditure ceiling for new installations was increased from
500 million to 700 million. In addition, a new requirement to register systems
larger than 12 kWp was introduced. On 6 June 2013, Gestore Servizi Energetici
announced that the 6.7 billion ceiling of the bill had been reached with 18.2 GW,
out of which 17.1 GW were already operational, and that the Quinto Conto Energia
would cease within 30 days, as foreseen (Gestore Servizi Energetici, 2013b).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 634

634 A. Jäger-Waldau

Slovakia
After two years of rapid growth, the Slovakian market decreased by over 90% in
2012 and only 29 MW were newly installed. The total capacity of 517 MW is more
than three times the original 160 MW capacity target for 2020, published in the
National Renewable Energy Action Plan in 2010. From February 2011, support
was limited to applications for systems smaller than 100 kW, and from 1 July 2013
onward, support was limited to systems of up to 30 kW placed on buildings.

Spain
Spain is still third in Europe for cumulative PV capacity, with 4.5 GW installed
by the end of 2012. Most of this was installed in 2008, when the country was
the biggest market, with close to 2.7 GW in 2008 (European Photovoltaic Industry
Association, 2013). This was more than twice the expected capacity and was due to
an exceptional race to install systems before the Spanish Government introduced
a cap of 500 MW on the yearly installations in the autumn of 2008. Royal Decree
1758/2008 set considerably lower FiTs for new systems and limited the annual
market to 500 MW, with the proviso that two-thirds should be rooftop-mounted
and not free-field systems. These changes resulted in a steep drop in the number
of new installations. In 2012, new system installations with a capacity of 194 MW
increased total capacity to 4.5 GW. Photovoltaic-generated electricity contributed
7.8 TW, or 2.9%, of the Spanish power demand in 2012.
In January 2012, the Spanish government passed a further Royal Decree 1/12
(Government Gazette, 2012), which suspended the remuneration pre-assignment
procedures for new renewable power capacity, affecting about 550 MW of planned
solar PV installations. The justification given for this move was that Spain’s energy
system had by then amassed a 30 billion power-tariff deficit and it was argued
that the special regime for renewable energy was the main reason. However, the
Spanish government has prevented utilities from charging consumers for the true
cost of electricity for over a decade. Instead of allowing utilities to increase rates
every time electricity generation costs increased (due to rising coal or natural gas
costs, inflation or changes in energy or environmental policy), the government
allowed them to create schemes such as a deferral account whereby they could
recover shortfalls in any individual year from revenues generated in subsequent
years.
In January 2007, the European Commission opened an in-depth investigation
to examine the potential support to large and medium-sized companies and elec-
tricity distributors in Spain, in the form of artificially low regulated industrial tariffs
for electricity (EUR-Lex, 2007). In 2005, these regulated tariffs led to a deficit of
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 635

The Photovoltaic Business: Manufacturers and Markets 635

3.8 billion in the Spanish electricity system, which increased to almost 9 billion
in 2007, a time when payments under the special regime for renewable energy were
still limited.
Despite the Royal Decree 1/12, the tariff deficit increased further in 2012 and
reached 35.6 billion at the end of February 2013 (Comición National de Energía
2013). The question remains: is renewable electricity generation responsible for
this? Despite the fact that the premium payments amount to 8.4 billion in 2012
and 9.1 billion for 2013, the answer is not clear. An often-neglected aspect is the
fact that investment in renewable capacity has increased supply at the wholesale
market, thus decreasing the system marginal price by the order of merit effect. It is
argued that renewable energy ‘pays for itself’ because by bidding at the pool at
zero prices these units substantially decrease the system marginal price. Therefore
the cost of all electricity should be lower, if this price reduction is passed on to
the customer. An analysis for the entire special regime concluded that for 2010 the
decrease in the wholesale price of around 29 /MWh covered 70% of the FiT costs
(Ciarreta et al., 2012b).

Switzerland
About 200 MW of PV systems were installed in Switzerland in 2012, almost dou-
bling the total capacity to 411 MW. Prices for turnkey systems decreased by over
40% in 2012 (Photovoltaik Guide, 2013). In view of these decreasing prices, the
FiT was reduced three times in 2012.

United Kingdom
The United Kingdom introduced a new FiT scheme in 2010, which led to the
installation of approximately 55 MW that year and over 1 GW in 2012. This steep
increase was caused by the announcement of a fast-track review of large-scale
projects by the Department of Energy & Climate Change (DECC) in February
2011, which led to a rush to complete these projects during the first half of 2011
(DECC, 2011). A second rush occurred towards the end of the year to meet the
deadline of 12 December 2011, when the DECC planned to decrease the residential
tariff by about 50% as a result of another fast-track consultation. However, this
decision was contested in court and the tariffs were only changed on 1 April 2012.
The average reductions in April were 44–54% for systems smaller than 50 kWp
and 0–32% for systems above 50 kWp. In November 2012, a further reduction
of 3.5% for systems smaller than 50 kWp was imposed. No reduction was made
for larger systems, as almost no such systems had been installed between May
and July 2012. However, for larger systems it was possible to receive Renewables
Obligation Certificates (ROCs) until April 2013; for PV the rate was two ROCs
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 636

636 A. Jäger-Waldau

per MWh. Overall about 630 MW were installed, bringing the cumulative capacity
to 1.8 GW at the end of 2012.
During 2012, the Energy and Climate Change Minister, Greg Barker, repeat-
edly declared his desire to see the UK solar market reach 22 GW by 2020. During
the first six months of 2013, the UK added 802 MW of new solar PV installations,
representing the strongest first-half year ever for the UK PV industry. This was
comprised of 20 MW in the first quarter of 2013, followed by 282 MW in the
second quarter of 2013 (Colville, 2013).

Austria
In 2012 Austria installed about 230 W of new PV systems and more than doubled
its cumulative capacity to 417 MW. The Ökostrom-Eispeisetarifverordnung 2012
(eco-electricity decree) is the regulation that sets the prices for the purchase of
electricity generated by green power plants. An investment subsidy with a budget
of 36 million in 2013 is also in place. Regardless of the size of the systems, a
maximum of 5 kWp is supported, with 300 /kWp for add-on and ground-mounted
systems and 400 /kWp for building-integrated systems. Operators of PV systems
larger than 5 kWp can choose to opt for the so-called net-parity tariff (Netzparitäts-
Tarif) for a period of 13 years. Since 1 January 2013, this option has only been
available for systems on buildings.

Portugal
Despite high solar radiation, solar photovoltaic system installation in Portugal has
grown only very slowly, reaching a cumulative capacity of 229 MW at the end of
2012.

Ukraine
Ukraine again saw an impressive growth and almost doubled its capacity with over
180 MW new installed PV systems to 370 MW. In February 2013, the Ukrainian
Parliament had the first reading of a bill to simplify the access of households to
the feed-in scheme. Reduction in the feed-in tariffs by between 16% and 27%,
depending on the type of installation, was finally adopted in February 2013 and
went into force on 1 April 2013. Since then, tariffs have been adjusted on a monthly
basis. In 2013 another doubling of the installed capacity is possible.

Turkey
In March 2010, the Energy Ministry of Turkey unveiled its 2010–2014 Strategic
Energy Plan. One of the government’s priorities was to increase the proportion
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 637

The Photovoltaic Business: Manufacturers and Markets 637

of renewably generated energy to 30% of total energy generation by 2023. At the

beginning of 2011, the Turkish Parliament passed renewable energy legislation
that defined new guidelines for FiTs. The FiT is 0.133 $/kWh (0.10 /kWh) for
owners commissioning a PV system before the end of 2015. If components made
in Turkey are used, the tariff will increase by up to $0.067 ( 0.052), depending on
the material mix. The first licensing round, with a volume of 600 MW, closed in
June 2013 and was oversubscribed about 15 times, with close to 9 GW of projects
submitted to the Turkish Energy Regulatory Authority. How much of this will be
installed still has to be seen. At the end of 2012, around 15 MW of grid-connected
and stand-alone systems were estimated to be installed cumulatively.

13.2.4 Africa
Despite its vast solar resources and the fact that there are areas in Africa where
the solar potential is very considerable — with the same photovoltaic panel able to
produce twice as much electricity in Africa as in Central Europe on average — only
limited use of PV electricity generation is made in Africa. The main application
is for small solar home systems, and the market statistics for these are extremely
imprecise or non-existent. Therefore, all African countries are considered potential
or emerging markets and some of them are mentioned below.

Algeria
In 2011, the Algerian Ministry of Energy and Mines published its Renewable
Energy and Energy Efficiency Programme, which aimed to increase the share of
renewable energy used for electricity generation to 40% of domestic demand by
2030. The plan foresaw 800 MW of installations by 2020 and a total of 1.8 GW by
2030. It is estimated that about 5 MW of small decentralised systems were installed
at the end of 2012. For 2013, new installations of around 20 MW were planned.

Capo Verde
Capo Verde’s Renewable Energy Plan (2010–2020) aims to increase the use of
renewable energy to 50% of total energy consumption in 2020. The policy to
achieve this is to use PPAs. Law n1/2011 established the regulations regarding
independent energy production. In particular, it established the framework con-
ditions for the set-up of independent power producers using renewable energy
(15 years PPA), and for self-production at user level. It created the micro-generation
regime, regulated rural electrification projects and established the tax exemption
of all imported renewable energy (RE) equipment. At the end of 2012, 7.5 MW
of centralised grid-connected PV systems were installed. In addition, there are a
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 638

638 A. Jäger-Waldau

number of smaller off-grid and grid-connected systems. To realise the 2020 50%
renewable energy target, about 340 MW of PV systems will be required.

Ghana
In 2011 the Parliament of Ghana passed a Renewable Energy Bill with the aims of
increasing the proportion of renewable energy, particularly solar, wind, mini hydro
and waste-to-energy schemes, in the national energy supply mix, and contributing
to the mitigation of climate change (Parliament of Ghana, 2011). The bill set a goal
of renewable energy constituting 10% of national energy generation by 2020. By
the end of 2012, a few thousand solar home systems and a few of off-grid systems
with an estimated 5 MW had been installed in the country. In 2012, Episolar,
Canada, signed a PPA with Ghana’s second largest utility, the Electricity Company
of Ghana, for a 50 MW PV plant with an option to increase the overall project size
to 150 MW (Episolar, 2012). In December 2012, Blue Energy, UK, announced that
it had agreed with Ghana’s Public Utilities Regulatory Commission on a 20-year
PPA for the 155 MW PV plant in Nzema to be operational by 2015 (Blue Energy,
2012). These two projects are still in the development phase looking for strategic
investors. In May 2013, the Volta River Authority inaugurated its first solar power
plant at Navrongo with a capacity of 2 MW. The Volta River Authority plans to
install a total of 14 MW by 2014.

Kenya
In 2008, Kenya introduced FiTs for electricity from renewable energy sources, but
solar was only included in the revision dated 2010 (Ministry of Energy, Kenya,
2010). However, only a little over 560 kW of PV capacity was connected to the
grid in 2011. The majority of the 14 MW of PV systems were off-grid installations.
In 2011, Ubbink East Africa Ltd., a subsidiary of Ubbink B.V. (Doesburg, The
Netherlands) opened a solar PV manufacturing facility with an annual output of
30,000 modules in Naivasha, Kenya. This plant produces modules for smaller PV
systems such as solar home systems. The estimates for the PV market in Kenya are
average sales of 20,000–30,000 home systems and 80,000 solar lanterns per year.

Morocco
The Kingdom of Morocco’s solar plan was introduced in November 2009, with the
aim of establishing 2000 MW of solar power by 2020. To implement this plan, the
Moroccan Agency for Solar Energy was founded in 2010. Both solar electricity
technologies, concentrated solar thermal power (CSP) and PV, were to compete
openly. In 2007, the National Office of Electricity announced a smaller programme
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 639

The Photovoltaic Business: Manufacturers and Markets 639

for grid-connected distributed solar PV electricity, targeting 150 MW of solar PV

power. Various rural electrification programmes using PV systems have been on-
going for a long time. At the end of 2012, Morocco had about 20 MW of PV systems
installed, mainly under the Global Rural Electrification Programme Framework,
and about 1 to 2 MW grid-connected systems.

South Africa
South Africa has a fast-increasing electricity demand and vast solar resources. In
2008, the country enacted its National Energy Act, which called for a diversifica-
tion of energy sources, including renewables, as well as fuel switching to improve
energy efficiency (South Africa, 2008). In 2011, the Renewable Energy Indepen-
dent Power Producer Procurement Programme (IPP) was set up with rolling bidding
rounds. Two rounds took place in 2011 and 2012, and a third one closed in August
2013. The overall target is 3.725 GW, including 1.45 GW for solar PV. In the first
two bidding rounds 1048 MW of solar PV projects were allocated to preferred
bidders. The average bid price decreased between the first round (closure date: 4
November 2011) from 2.65 ZAR/kWh (0.252 /kWh17) to 1.65 ZAR/kWh (0.157
/kWh) in the second round (closure date: 5 March 2012). It is estimated that about
30–40 MW of PV systems were installed in South Africa by the end of 2012.

13.2.5 The Americas

Canada
In 2012, the Canadian market was about the same as in 2011, with 268 MW cumu-
lative installed PV capacity increasing to about 830 MW. This development was
driven by the introduction of an FiT in the Province of Ontario, enabled by Bill
150, Green Energy and Green Economy Act 2009. More than 77% of the total
capacity is installed in Ontario. On the federal level, an accelerated capital cost
allowance exists under the income tax regulations. On a province level, nine Cana-
dian provinces have net metering rules, with solar photovoltaic electricity as one of
the eligible technologies. Sales tax exemptions and renewable energy funds exist
in two provinces and micro grid regulations and minimum purchase prices each
exist in one province.
The Ontario FiT s were set in 2009, depending on system size and type, and
were then reduced in various steps. In June 2013, the Minister of Energy announced
an annual cap of 150 MW for the small FiT regime and 50 MW for the micro FiT
regime for the following four years, with a transition in 2013, which meant that

17 Exchange rate: 1 = 10.5 ZAR.

September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 640

640 A. Jäger-Waldau

systems larger 500 kW would no longer be eligible for the FiT. Following the WTO
ruling in May on the local content regulations, the Minister said that Ontario would
comply with it.

United States of America

With over 3.3 GW of newly installed PV capacity, the USA reached a cumulative
PV capacity of 7.7 GW (7.2 GW grid-connected) at the end of 2012. Utility PV
installations again more than doubled, compared with 2011 and became the largest
segment with 1.7 GW in the year. The top ten states — California, Arizona, New
Jersey, Nevada, North Carolina, Hawaii, Maryland, Texas, New York and Mas-
sachusetts — accounted for more than 88% of the market (SEIA, 2013). Market
growth around 30% is estimated for 2013. Photovoltaic projects with PPAs with
a total capacity of 10.5 GW are already under contract and over 3 GW of these
projects are already financed and under construction (SEIA, 2013). If one adds
those projects in an earlier development stage, the pipeline stands at almost 22 GW.
Many state and federal policies and programmes have been adopted to encour-
age the development of markets for PV and other renewable technologies. These
consist of direct legislative mandates (such as renewable content requirements) and
financial incentives (such as tax credits). One of the most comprehensive databases
about the different support schemes in the USA is maintained by the Solar Centre of
the State University of North Carolina. The Database of State Incentives for Renew-
able Energy (www.dsireusa.org) is a comprehensive source of information on state,
local, utility, and selected federal incentives that promote renewable energy.

Argentina
In 2006, Argentina passed its Electric Energy Law, which established that 8% of
electricity demand should be generated by renewable sources by 2016 (Argentina,
2006). It also introduced FiTs for wind, biomass, small-scale hydro, tidal, geother-
mal and solar for a period of 15 years. In July 2010, amongst other renewable
energy sources, the government awarded PPAs to six solar PV projects totalling
20 MW. By the end of 2012, about 17 MW (7 MW grid-connected) of PV systems
had been installed. According to the renewable energy country attractiveness indi-
cator, the Argentinean government has set a 3.3 GW target for PV installations by
2020 (Ernst & Young, 2011).

Brazil
At the end of 2012, Brazil had about 20 MW cumulative installed capacity, mainly
in rural areas. In April 2012, the board of the National Agency of Electric Energy
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 641

The Photovoltaic Business: Manufacturers and Markets 641

approved new rules to reduce barriers to install small-distributed generation capac-

ity. The rule applies to generators that use subsidised sources of energy (hydro,
solar, biomass, cogeneration and wind). Over 2 GW of PV projects applied for
approval. It is estimated that about 40 MW (6.6 MW grid-connected) had been
installed by the end of 2012. In its mid-term market report, the IEA forecast a
cumulative installed PV capacity of about 200 MW in 2013, and 1.2 GW by 2018
(IEA, 2013a).

Chile
In February 2012, the President of Chile announced a strategic energy plan to
reach 20% of non-conventional renewable energy by 2020. In the first quarter of
2012, the first MW-size PV system was installed in the northern Atacam desert. By
June 2013, Chile’s Environmental Assessment Service had approved over 4 GW
of projects, with an additional 2.2 GW still under review. At the end of 2012, it
was estimated that about 10 MW of PV systems (5 MW grid-connected) had been
installed. In June 2013, about 70 MW of PV systems were under construction and
it is expected that about 50 MW of new systems will be operational at the end of
2013.

Dominican Republic
In 2007, the Dominican Republic passed a law promoting the use of renewable
energy and setting a target of 25% renewable energy share in 2025 (Domini-
can Republic, 2007). At that time ∼1–2 MW of solar PV systems had been
installed in rural areas, which had increased to over 5 MW by 2011. In 2011,
the first PPA for 54 MW was signed between Grupo de Empresas Dominicanas de
Energía Renovable and Corporación Dominicana de Empresas Eléctricas Estatales
(CDEEE). The first phase (200 kW) of the project became operational in July 2012,
and the whole solar farm was connected to the grid early 2013. In 2012, CDEEE
signed two more PPAs with a total capacity of 116 MW.

Mexico
In 2008, Mexico enacted a Law for Renewable Energy Use and Financing Energy
Transition to promote the use of renewable energy (Mexico, 2008). In 2012, the
country passed its Climate Change Law, which foresaw a decrease in greenhouse
gas emissions of 30% below the business-as-usual case by 2020, and 50% by 2050
(Mexico, 2012). It further stipulated that renewable electricity should constitute
35% of total supply by 2024. By the end of 2012, about 52 MW of PV systems had
been installed (IEA, 2013c).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 642

642 A. Jäger-Waldau

Peru
In 2008, Peru passed its Legislative Decree 1002, which made the development
of renewable energy resources a national priority. The decree stated that at least
5% of electricity should be supplied from renewable sources by 2013. In February
2010, the Organismo Supervisor de la Inversión en Energía y Minería held the first
round of bidding and awarded four solar projects with a total capacity of 80 MW.
A second bidding round was held in 2011, with a quota of 24 MW for PV. By
the end of 2012, about 85 MW of PV systems had been installed. A doubling of
capacity is regarded as possible in 2013.

13.3 The photovoltaic industry

The photovoltaic industry consists of a long value chain from raw materials to PV
system installation and maintenance. So far, the main focus has been on solar cell
and module manufacturers, but in addition there is the so-called upstream industry
(e.g. materials, polysilicon production, wafer production, equipment manufactur-
ing) and the downstream industry (e.g. inverters, balance-of-system [BOS] com-
ponents, system development, project development, financing, installations and
integration into existing or future electricity infrastructure, plant operators, oper-
ation and maintenance). In the near future, it will probably be necessary to add
(super)-capacitor and battery manufacturers, as well as power electronics and IT
providers for demand and supply management, including meteorological forecast-
ing. The main focus in this section, however, will be on solar cell and module
manufacturers and polysilicon manufacturers.
In 2012, world solar cell production grew by 10% to about 38.5 GW.
The market for installed systems grew by almost 30% and values between 29
and 31 GW have been reported by various consultancies and institutions. This
mainly represents the grid-connected photovoltaic market. To what extent the
off-grid and consumer-product markets are included is unclear. The difference
of roughly 7–9 GW has therefore to be explained as a combination of unac-
counted off-grid installations (∼1–200 MW off-grid rural, ∼1–200 MW communi-
cation/signals, ∼100 MW off-grid commercial), consumer products (∼1–200 MW)
and cells/modules in stock.
In addition, the fact that some companies report shipment figures, some report
sales figures and others report production figures, adds to the uncertainty. An addi-
tional source of error is the fact that some companies produce fewer solar cells
than solar modules, but the reporting does not always differentiate between the
two, and there is a risk that cell production is double-counted, first at the cell
manufacturer and second at the ‘integrated’ cell/module manufacturer. Difficult
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 643

The Photovoltaic Business: Manufacturers and Markets 643

economic conditions contributed to decreased willingness to report confidential

company data. Nevertheless, the figures show a significant growth in production,
as well as an increasing installation market.
Despite the fact that a significant number of companies filed for insolvency,
scaled back or even cancelled their expansion projects over the past year, the num-
ber of new entrants into the field, notably large semiconductor or energy-related
companies, overcompensated. The announced production capacities — based on a
survey of more than 300 companies worldwide — increased again in 2012, and is
expected to continue to do so in 2013 and 2014, albeit at a slower pace. However,
rapid changes in the sector and the difficult financing situation make forecasting
for future capacity developments very speculative. It is believed that the current sit-
uation of overcapacity in solar cell manufacturing will continue until at least 2015,
when the global PV market should exceed 50 GW of new installations annually.

13.3.1 Technology mix

As silicon prices fell drastically after the temporary silicon shortage, wafer-based
silicon solar cells decreased in cost very rapidly and remain the main technol-
ogy, with close to 90% market share in 2012. Commercial module efficiencies
fall within a wide range between 12% and 21%, with monocrystalline modules
between 14% and 21%, and polycrystalline modules between 12% and 18%. The
massive manufacturing capacity increases for both technologies were followed by
the necessary capacity expansion for polysilicon production.
In 2005, production of thin-film solar modules reached more than 100 MW per
annum for the first time. Between 2005 and 2009, the compound annual growth rate
of thin-film solar module production was greater than that of the overall industry,
increasing the market share of thin-film products from 6% in 2005 to 10% in 2007
and 16–20% in 2009. Since then, the thin-film market share has decreased slowly
as the ramp-up of new production lines did not follow that of wafer-based silicon.
The majority of thin-film companies are silicon-based and use either
amorphous silicon or an amorphous/microcrystalline silicon structure. Next are
companies using Cu(In,Ga)(Se,S)2 as the absorber material, whereas only a few
companies use CdTe or dye and other materials. These technologies are covered
in Chapters 4 to 9.
Concentrated photovoltaics (CPV), discussed by Luque-Heredia and Luque
in Chapter 11, is an emerging technology that is growing fast, although from a low
starting point. About 60 companies are active in the field of CPV development, the
majority focusing on high-concentration concepts. Over half of these companies
are located either in the USA (primarily in California) or in Europe (primarily
Spain).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 644

644 A. Jäger-Waldau

Within CPV, there is a differentiation according to the concentration factor18

and whether the system uses a dish (Dish CPV) or lenses (Lens CPV) to concentrate
sunlight. The main parts of a CPV system are the cells, the optical elements and the
tracking devices. The recent growth in CPV is based on significant improvements
in all of these areas, as well as system integration. However, CPV is at the beginning
of an industry learning curve, with considerable potential for technical and cost
improvements. The most challenging task is to become cost-competitive with other
PV technologies quickly enough in order to grow to reach factory sizes, which can
count on economies of scale. Despite the small installed CPV capacity to date,
various consultancy companies predict that the CPV market will grow to 500 MW
by 2015, and 1 GW by 2020 (GlobalData, 2013; Strategies Unlimited, 2013).
The existing PV technology mix is a solid foundation for future growth of
the sector as a whole. No single technology can satisfy all the different consumer
needs, ranging from mobile and consumer applications, with the need for a few
watts, to multi-MW utility-scale power plants. The variety of technologies is insur-
ance against a roadblock for the implementation of solar photovoltaic electricity,
if material limitations or technical obstacles restrict the further growth or develop-
ment of a single technology pathway.

13.3.2 Solar cell manufacturers19

Worldwide, more than 350 companies produce solar cells. The solar cell indus-
try over the last decade has been very dynamic, and the changes that have been
happening since 2011 allow only for a snapshot picture of the current situation,
which might already have changed a few weeks later. Despite the fact that a few
dozen companies filed for insolvency, scaled back, idled or stopped production, the
number of newcomers and their planned capacities still exceeds the retired capacity.
The following sections give short descriptions of the 20 largest companies, in
terms of actual production/shipments in 2012. More information about additional
solar cell companies and details can be found in various market studies. The capac-
ity, production or shipment data are from the annual reports or financial statements
of the respective companies or the cited references.

18 High concentration > 300 Suns (HCPV), medium concentration 5 < x < 300 Suns (MCPV), low
concentration < 5 Suns (LCPV).
19 Solar cell production capacities mean: in the case of wafer silicon-based solar cells, the cell; in the
case of thin-films, the complete integrated module. Only those companies that actually produce the
active circuit (solar cell) are counted. Companies that purchase these circuits and make cells are not
counted.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 645

The Photovoltaic Business: Manufacturers and Markets 645

Yingli Green Energy Holding Company Ltd. (China)

Yingli Green Energy (www.yinglisolar.com) went public on 8 June 2007. The main
operating subsidiary, Baoding Tianwei Yingli New Energy Resources, is located in
the Baoding National High-New Tech Industrial Development Zone. The company
deals with the whole set, from solar wafers, cell manufacturing and module produc-
tion. According to the company, production capacity reached 1.85 GW at the end of
2011. In its 2012 annual report, the company reported a capacity of ingot, wafers,
cells and modules of 2.45 GW at the end of the year, and that in addition to their
own cells they also purchased cells from other companies. For 2012, total reported
shipments of solar modules were 2.3 GW. Solar cell production was estimated at
1950 MW for 2012.
In January 2009, Yingli acquired Cyber Power Group Ltd., a development-
stage enterprise designed to produce polysilicon. Through its principal operating
subsidiary, Fine Silicon, the company started trial production of solar-grade polysil-
icon in late 2009, and was still ramping up to full production capacity of 3000 metric
tons per year at the end of 2011. However, recent financial results indicate that the
company has written off its investment in Fine Silicon and according to other media
reports, production has now closed down.
In January 2010, the Ministry of Science and Technology of China approved an
application to establish a national-level key laboratory in the field of PV technology
development, the State Key Laboratory of PV Technology, at Yingli Green Energy’s
manufacturing base in Baoding. In October 2013, BTU International signed a two-
year agreement to support this laboratory, which has not yet been built.

First Solar LLC (USA/Germany/Malaysia)

First Solar LLC (www.firstsolar.com) is one of the few companies worldwide to
produce CdTe thin-film modules. The company currently has three manufactur-
ing sites, in Perrysburg (USA), Frankfurt/Oder (Germany) and Kulim (Malaysia),
with a combined capacity of 2.376 GW at the end of 2011. A second Frank-
furt/Oder plant, doubling the capacity there to 528 MW, became operational in May
2011 and expansion in Kulim increased production capacity there to 1.584 GW
at the end of 2011. In April 2012, however, the company announced a major
restructuring in response to changing market conditions (First Solar, 2012), clos-
ing all its manufacturing operations in Frankfurt/Oder at the end of 2012, and
idling four production lines in Kulim. In addition, the company put its fac-
tory in Meza (AZ) on hold, and sold its factory in Dong Nam Industrial Park,
Vietnam.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 646

646 A. Jäger-Waldau

In 2012, the company produced 1.875 GW of product. For Q1 of 2013, the

company reported production costs of 0.69 $/Wp (0.53 /Wp), including under-
utilisation and upgrading costs of 0.05 $/Wp (0.038 /Wp).

JA Solar Holding Company Ltd. (China)

JingAo Solar (www.jasolar.com) was established in May 2005 by the Hebei Jing-
long Industry and Commerce Group, the Australia Solar Energy Development Pty.
Ltd. and Australia PV Science and Engineering Company. Commercial operation
started in April 2006 and the company went public on 7 February 2007. According
to company reports, production capacity was 2.5 GW for cells, 1.8 GW for modules
and 1 GW for wafers at the end of 2012. For 2012, sales of 1.7 GW were reported.

Trina Solar Ltd. (China)

Trina Solar (www.trinasolar.com) was founded in 1997 and went public in Decem-
ber 2006. The company has integrated product lines, making ingots, wafers and
modules. In December 2005, a 30 MW monocrystalline silicon wafer product line
went into operation. According to the company, production capacity was 1.2 GW
for ingots and wafers and 2.4 GW for cells and modules at the end of 2012, and
shipments of 1.6 GW were reported for the year.
In January 2010, the company was selected by the Chinese Ministry of Sci-
ence and Technology to establish a State Key Laboratory to develop PV tech-
nologies within the Changzhou Trina PV Industrial Park. This laboratory has
been established as a national platform for driving PV technologies in China.
Its mandate includes research into PV-related materials, cell and module tech-
nologies and system-level performance. It will also serve as a platform to bring
together technical capabilities from the company’s strategic partners, including
customers and key PV component suppliers, as well as universities and research
institutions.

Hanwha (China/Germany/Malaysia/South Korea)

Hanwha Group (www.hanwha.com) acquired a 49.99% share in Solarfun Power
Holdings in 2010 and the name was changed to Hanwha SolarOne in January 2011.
The company produces silicon ingots, wafers, solar cells and solar modules. Its
first production line was completed at the end of 2004 and commercial production
started in November 2005. It went public in December 2006, and reported the
completion of its production capacity expansion to 360 MW in Q2 of 2008.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 647

The Photovoltaic Business: Manufacturers and Markets 647

In August 2012, the company acquired Q-Cells (Germany/Malaysia), which

had filed for insolvency in April 2012. Hanwha, with its two brands Hanwha Q
Cells and Hanwha SolarOne, has combined production capacity of 2.3 GW of
solar cells (1.3 GW China, 800 MW Malaysia, 200 MW Germany). In addition
Hanwha SolarOne has 800 MW of ingot and wafer production capacity. Ship-
ments of 830 MW for Hanwha Solar-One and an additional solar cell production
of 570 MW for Hanwha Q Cells were reported for 2012 (Ikki, 2013b).

Suntech Power Company Ltd. (China)

Suntech Power (www.suntech-power.com) was founded in January 2001 by Dr.
Zhengrong Shi and went public in December 2005. Suntech specialises in the
design, development, manufacturing and sale of photovoltaic cells, modules and
systems. In its preliminary financial statement, the company reported shipments of
1.8 GW for 2012. External reports give the cell production at 1.35 GW (PV News,
2013).
Due to financial problems, Wuxi Suntech Power, the Chinese subsidiary of
Suntech Power Holdings, filed a petition for insolvency restructuring with Wuxi
Municipal Intermediate People’s Court in Jiangsu Province, China, which was
accepted on 20 March 2013.

Motech Solar (Taiwan/China)

Motech Solar (www.motech.com.tw) is a wholly-owned subsidiary of Motech
Industries Inc., located in Tainan Science Industrial Park. The company started
mass production of polycrystalline solar cells at the end of 2000, with an annual
production capacity of 3.5 MW. Production increased from 3.5 MW in 2001 to
1 GW in 2011. In 2009, Motech started the construction of a factory in Kunshan,
China, which reached its nameplate capacity of 500 MW in 2011. Total production
capacity at the end of 2012 was given as 1.6 GW. For 2012, total shipments of
1.28 GW were reported.

Canadian Solar Inc. (China)

Canadian Solar Inc. (CSI) (www.canadiansolar.com) was founded in Canada in
2001 and was listed on NASDAQ in November 2006. It has established six wholly-
owned manufacturing subsidiaries in China, manufacturing ingots/wafers, solar
cells and solar modules. According to the company, CSI had 216 MW of ingot and
wafer capacity, 1.6 MW cell capacity and 2.4 GW module manufacturing capacity
(2.1 GW in China, 330 MW in Ontario, Canada) at the end of 2012. For 2012, the
company reported sales of 1.47 GW of modules, but no cell production figure was
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 648

648 A. Jäger-Waldau

given. This must, however, be lower, because the company states in its financial
reports that it buys cells from other manufacturers. External reports gave cell pro-
duction as 1.24 GW (PV News, 2013).

Gintech Energy Corporation (Taiwan)

Gintech (www.gintech.com.tw) was established in August 2005 and went public
in December 2006. Production at Factory Site A, Hsinchu Science Park, began in
2007 with an initial production capacity of 260 MW and increased to 1.17 GW at
the end of 2011, with further expansion to 1.5 GW in 2012. In 2012, the company
produced 1.1 GW (PV News, 2013).

JinkoSolar Holding Company Ltd. (China)

Jinko Solar (www.jinkosolar.com) was founded by HK Paker Technology Ltd. in
2006. Starting from upstream business, the company expanded operations across
the solar value chain, including recoverable silicon materials, silicon ingots and
wafers, solar cells and modules in 2009. The company went public in May 2010 and
is currently listed on the New York Stock Exchange. According to the company,
it had manufacturing capacities of 1.2 GW each for wafers, solar cells and solar
modules at the end of 2012. It reported sales of about 1.2 GW (912 MW modules,
84 MW cells and 190 MW wafers) for 2012.

Neo Solar Power Corporation (Taiwan)

Neo Solar Power (www.neosolarpower.com) was founded in 2005 by PowerChip
Semiconductor, Taiwan’s largest DRAM (dynamic random-access memory) com-
pany, and it went public in October 2007. The company manufactures both mono-
and multicrystalline silicon solar cells. Production capacity of silicon solar cells at
the end of 2012 was 1.3 GW. In 2013, the company produced about 958 MW (PV
News, 2013). In 2013, the company merged with DelSolar to become the largest
Taiwanese cell producer.

SunPower Corporation (USA/Philippines/Malaysia)

SunPower (http://us.sunpowercorp.com) was founded in California in 1988 by
Richard Swanson and Robert Lorenzini to commercialise proprietary, high-
efficiency, silicon solar cell technology. The company went public in November
2005. SunPower designs and manufactures high-performance silicon solar cells,
based on the interdigitated rear-contact design described in Section 3.4 of Chapter3.
The initial products, introduced in 1992, were high-concentration solar cells with
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 649

The Photovoltaic Business: Manufacturers and Markets 649

an efficiency of 26%. SunPower also manufactures a 22%-efficient solar cell, called

Pegasus, which is designed for non-concentrating applications.
SunPower conducts its main research and development (R&D) activity in
Sunnyvale, California, and has cell-manufacturing plants in the Philippines and
Malaysia. In 2011, the company had two cell-manufacturing plants outside of
Manila, but decided to close down Fab. No 1 in early 2012. It then consolidated
production at Fab. No 2 (with a nameplate capacity of 575 MW) in the Philippines
and its newer facility, Fab. No 3 in Malaysia, which is operated as a joint venture
with AU Optronics Corporation and had a capacity of 1.4 GW at the end of 2012
(PV Tech, 2012). The company has two solar module factories in the Philippines
(600 MW) and, since 2011, also in Mexico (500 MW). In addition, modules are
assembled for SunPower by third-party contract manufacturers in China, Mexico,
Poland and California. Total cell production in 2012 was reported at 925 MW.

Hareon Solar Technology Company Ltd. (China)

Haeron Solar (www.hareonsolar.com) was established as the Jiangyin Hareon
Technology in 2004 and changed its name to the Hareon Solar Technology in 2008.
It has five manufacturing facilities in Jiangsu and Anhui provinces: Jiangyin Hareon
Power, Altusvia Energy (Taicang), Hefei Hareon Solar Technology, Jiangyin Xin-
hui Solar Energy and Schott Solar Hareon. Solar cell production started in 2009,
with an initial capacity of 70 MW. According to the company, production capac-
ity was to be increased to over 2 GW for cells and 1 GW for modules in 2012.
Production of 900 MW was reported for 2012 (PV News, 2013).

Kyocera Corporation (Japan)

In 1975, Kyocera (http://global.kyocera.com/prdct/solar/) began research on solar
cells. The Shiga Yohkaichi Factory was established in 1980, and R&D and manu-
facture of solar cells and products started with mass production of multicrystalline
silicon solar cells in 1982. In 1993, Kyocera became the first Japanese company to
sell home PV generation systems.
Besides its solar-cell-manufacturing plants in Japan, Kyocera has had module-
manufacturing plants in China in a joint venture with the Tianjin Yiqing Group
(10% share) in Tianjin (since 2003), Tijuana, Mexico (since 2004) and Kadan,
Czech Republic (since 2005).
The company also markets systems that both generate electricity through solar
cells and exploit heat from the Sun for other purposes, such as heating water. Its
Sakura factory in Chiba Prefecture is involved in everything from R&D and system
planning to construction and servicing, and its factory in Shiga Prefecture is active
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 650

650 A. Jäger-Waldau

in R&D, as well as the manufacture of solar cells, modules, equipment parts and
devices which exploit heat. In 2012, Kyocera had a production of 800 MW and
planned to increase its capacity to 1 GW in 2013 (Ikki, 2013ab).

Renewable Energy Corporation AS (Norway/Singapore)

The Renewable Energy Corporation’s (REC) (www.recgroup.com) vision is to
become the most cost-efficient solar energy company in the world, with a presence
throughout the whole value chain. The company, located in Høvik, Norway, has
five business activities, ranging from silicon feedstock to solar system installations,
and is involved in all major aspects of the PV value chain through its various
subsidiaries. In 2011, the company closed down REC ScanCell, which was located
in Narvik, and had a production capacity of 180 MW. The next closure, announced
in March 2012, was the wafer factory in Glomfljord, which had 300 MW capacity
for multicrystalline wafers, while the 650 MW wafer plant at Herøya continued
operations. From 2012, production of solar cells and modules continued only at
REC Solar Singapore, which operates an integrated site for wafers, solar cells and
module production, with a capacity of 750 MW. In 2012, production was reported
at 750 MW.

ReneSola Ltd. (China)

Renesola (www.renesola.com) was established in 2005 and went public on the New
York Stock Exchange in 2008. The company produces polysilicon, wafers, solar
cells and solar modules. According to the company, it had a production capacity
of 2 GW of wafers, 1.2 GW of modules and 800 MW of solar cells at the end of
2012. Shipments for 2012 were reported as 1.5 GW wafers and 710 MW modules.

Changzhou EGing Photovoltaic Technology Company Ltd. (China)

EGing PV (www.egingpv.com) was founded in 2003 and works along the complete
photovoltaic industry value chain, from the production of monocrystalline furnaces,
quartz crucibles, 5–8 inch monocrystalline silicon ingots, supporting equipment
of squaring and wire sawing, monocrystalline silicon wafers, solar cells and solar
modules. According to the company, it had a production capacity of 1 GW across
the complete value chain of ingots, wafers, cells and modules at the end of 2011.
Sales of 600 MW were reported for 2012 (ENF, 2013).

Astronergy Solar (China)

Astronergy Solar (www.astronergy.com) was established as a member of the Chint
Group in October 2006. The first production line of 25 MW, for crystalline silicon
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 651

The Photovoltaic Business: Manufacturers and Markets 651

cells and modules, was installed in May 2007 and an increase in production capacity
to 100 MW was achieved in July 2008. Commercial production of micromorph
solar modules started in July 2009. Thin-film capacity was 30 MW in 2010 and
increased to 75 MW in early 2011. Total production capacity for 2012 was reported
as 800 MW, with 584 MW of production (ENF, 2013).

Solar Frontier (Japan)

Solar Frontier is a 100%-owned subsidiary of Showa Shell Sekiyu K.K. In 1986,
Showa started to import small modules for traffic signals, and entered module pro-
duction in Japan, cooperatively with Siemens (now Solar World). The company
developed CIS solar cells and completed the construction of its first factory with
20 MW capacity in October 2006. Commercial production started in FY 2007.
The second Myazaki factory (MP2) started manufacturing in 2009 with a pro-
duction capacity of 60 MW. In July 2008, the company announced they would
open a research centre ‘to strengthen research on CIS solar-powered cell tech-
nology, and to start collaborative research on mass production technology of the
solar modules with Ulvac, Inc.’ The aim was to start a new plant in 2011 with
a capacity of 900 MW. The ramp-up started in February 2011, and at the end of
that year overall capacity was 980 MW. In 2011, the company changed its name
to Solar Frontier, and production was reported as 450 MW (RTS Corporation,
2012). A moderate increase to above 500 MW was expected for 2012. Early in
2013, the company reported that its Kunitomi plant was operating at full capacity.
Production at MP2 was halted at the end of 2012 to make adjustments for the
production of new, differentiated products, and the plant resumed production in
July 2013.

Jiangxi Risun Solar Energy Company Ltd. (China)

Risun Solar Technologies was established in 2008. The company manufactures
mono- and multicrystalline solar cells and modules, with a stated production
capacity of 700 MW for solar cells and 300 MW for modules. An expansion to
3 GW was planned without a specified date. Production of 500 MW was reported
for 2012 (ENF, 2013).

13.3.3 Polysilicon supply

The rapid growth of the PV industry since 2000 led to the situation where, between
2004 and early 2008, the demand for polysilicon outstripped supply from the semi-
conductor industry. Prices for purified silicon started to rise sharply in 2007, and
prices for polysilicon peaked at around 500 $/kg in 2008, resulting in higher prices
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 652

652 A. Jäger-Waldau

for PV modules. This extreme price hike triggered a massive capacity expansion,
not only by established companies, but attracting many new entrants as well.
These massive production expansions, as well as the difficult global economic
situation, led to price decreases for polysilicon throughout 2009. Prices reduced to
about 50–55 $/kg at the end of 2009, with a slight upwards tendency throughout
2010 and early 2011, before they dropped significantly again. The lowest level on
the spot market was reached in December 2012 with prices below 16 $/kg (12.3
/kg) before they started to recover slightly, trading between 16.5 and 17.5 $/kg
(12.7–13.5 /kg) for most of 2013.
China was and remains a major polysilicon producer. In January 2011, the
Chinese Ministry of Industry and Information Technology tightened the rules for
future polysilicon factories to combat the oversupply problem. New factories must
be able to produce more than 3000 metric tons of polysilicon per year and also
meet certain efficiency, environmental and financing standards. Their total energy
consumption must be less than 200 kWh/kg of polysilicon, and China is aiming to
create large companies with at least 50000 metric tons annual capacity by 2015.
These two framing conditions, in addition to the enormous price pressure, are the
reasons why a significant number of Chinese manufacturers closed down their
polysilicon production in the first half of 2012. This is also why China imported
83000 metric tons of silicon in 2012, 32% more than 2011 (Bloomberg New Energy
Finance, 2013g).
Projected global silicon production capacities in 2013 vary between 290,000
metric tons (Bloomberg New Energy Finance, 2013e) and 409,690 metric tons
(Ikki, 2013a). It is estimated that about 27000 metric tons will be used by the
electronics industry. The possible solar cell production will, in addition, depend
on the material used per Wp; the current worldwide average is about 5.6 g/Wp.

13.3.4 Polysilicon manufacturers

Worldwide, more than 100 companies produce or have started up polysilicon pro-
duction. This section gives a short description of the ten largest companies in terms
of production in 2012. More information about additional polysilicon companies
and details can be found in various market reports.

Wacker Polysilicon (Germany)

Wacker Polysilicon AG (www.wacker.com) is one of the world’s leading manufac-
turers of hyper-pure polysilicon for the semiconductor and photovoltaic industry,
chlorosilanes and fumed silica. In 2011, Wacker increased its capacity to over
40000 metric tons and reported sales of 32000 metric tons. Their 15000 metric
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 653

The Photovoltaic Business: Manufacturers and Markets 653

ton factory in Nünchritz Saxony, started production in 2011. In 2010, the company
decided to build a polysilicon plant in Tennessee with 15000 ton capacity. The
groundbreaking of this new 18000 metric ton factory took place in April 2011,
and construction should be finished by the end of 2013. In addition, the company
also expanded its Burghausen capacity by 5000 metric tons in 2012 and, together
with a further expansion of the Nünchritz factory by 5000 metric tons, it plans to
have 70000 metric tons of production capacity in 2014. Total polysilicon sales are
reported as 38000 metric tons in 2012.

GCL-Poly Energy Holdings Ltd. (China)

GCL-Poly (gcl-poly.com.hk) was founded in March 2006 and construction of their
Xuzhou polysilicon plant (Jiangsu Zhongneng Polysilicon Technology Develop-
ment) began in July 2006. Phase I has a designated annual production capacity
of 1500 tons and the first shipments were made in October 2007. Full capacity
was reached in March 2008. At the end of 2011, polysilicon production capac-
ity had reached 65000 metric tons and 8 GW of wafers. For 2012, the company
reported production of 37055 metric tons of polysilicon and 5.6 GW wafers. It has
invested in the downstream business of solar: GCL Solar System Ltd. (SSL) is a
wholly-owned subsidiary of GCL-Poly Energy Holdings Ltd. and provides solar
system turnkey solutions for residential, governmental, commercial and solar farm
projects, including design, equipment supply, installation and financial services.
Another subsidiary is GCL Solar Power, which develops, operates and manages
solar farms.

OCI Company (South Korea)

OCI Company Ltd. (www.oci.co.kr) (formerly DC Chemical) is a global chemical
company with a product portfolio spanning the fields of inorganic chemicals, petro
and coal chemicals, fine chemicals, and renewable energy materials. In 2006, the
company started its polysilicon business and completed its 6500 metric ton P1
plant in December 2007. The 10500 metric ton P2 expansion was completed in
July 2009, and P3 with another 10000 metric tons brought the total capacity to
27000 metric tons at the end of 2010. The debottlenecking of P3 in 2011 increased
the capacity to 42000 tons at the end of the year. Further capacity expansions,
P4 (20000 tons) and P5 (24000 tons), were put on hold due to the rapid price
decline of polysilicon. Instead the company is pursuing a further debottlenecking
of its existing plants to increase capacity by 10000 metric tons during 2013. For
2012, production of 33000 metric tons was estimated from the company’s financial
figures.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 654

654 A. Jäger-Waldau

OCI has also invested in downstream business and holds 89.1% of OCI Solar
Power, which develops, owns and operates solar power plants in North America.
On 23 July 2012, the company signed a PPA with CSP Energy, Texas, for a 400 MW
solar farm in San Antonio, TX.

Hemlock Semiconductor Corporation (USA)

Hemlock Semiconductor Corporation (www.hscpoly.com) is based in Hemlock,
Michigan. The company is a joint venture of Dow Corning Corporation (63.25%)
and two Japanese firms, Shin-Etsu Handotai Co. Ltd. (24.5%) and Mitsubishi Mate-
rials Corporation (12.25%). The company is the leading provider of polycrystalline
silicon and other silicon-based products used in the semiconductor and solar indus-
try. In 2007, the company had an annual production capacity of 10000 tons of
polycrystalline silicon and production at the expanded Hemlock site (19000 tons)
started in June 2008. A further expansion at the Hemlock site, as well as a new
factory in Clarksville, Tennessee, was started in 2009. Total production capacity
was 46000 metric tons in 2011 and expansion to 56000 metric tons was scheduled
to be finalised in 2012. Production of 28000 metric tons was estimated for 2012.

Renewable Energy Corporation AS (Norway)

In 2005, Renewable Energy Corporation AS (REC) (www.recgroup.com) took
over Komatsu’s US subsidiary, Advanced Silicon Materials LLC (ASiMI), and
announced the formation of its silicon division business area, REC Silicon Division,
comprising the operations of ASiMI and REC Solar Grade Silicon LLC (SGS).
Production capacity at the end of 2012 was around 22400 metric tons and according
to the company, a total of 21405 metric tons of polysilicon, of which 18791 metric
tons were electronic-grade silicon, was produced in 2012.

SunEdison Inc. (USA)

SunEdison, formerly MEMC Electronic Materials Inc. (http://sunedisonsilicon.
com), has its headquarters in St. Peters, Missouri. It started operations in 1959 and
it makes semiconductor-grade wafers, granular polysilicon, ultra-high purity silane,
trichlorosilane, tetrafluorosilane and sodium aluminium tetrafluoride. In February
2011, the company and Samsung entered into a 50/50 equity joint venture to build
a polysilicon plant in Korea with an initial capacity of 10000 metric tons in 2013.
At the end of 2011, the company closed its 6000 metric ton Merano, Italy, factory
and reduced its capacity in Portland, USA, during 2012. Production capacity at the
end of 2012 was ∼10000 metric tons with estimated production of about 12000
metric tons.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 655

The Photovoltaic Business: Manufacturers and Markets 655

SunEdison is developing solar power projects as well as acting as a solar energy

provider. At the end of 2012, the company had 73 MW of photovoltaic power plants
under construction and 2.6 GW of projects in the pipeline.

Tokuyama Corporation (Japan)

Tokuyama (www.tokuyama.co.jp) had an annual production capacity of 9200 tons
in 2010. In February 2011, the company broke ground for a new 20000 ton facility
in Malaysia. The first phase, with 6200 metric ton capacity, started trial production
in March 2013, and the second phase, with 13800 metric ton capacity, should be
fully operational in 2015. For 2012, production of 7800 tons was estimated.

M. Setek Company Ltd. (Japan)

M. Setek is a subsidiary of Taiwan’s AU Optronics. The company is a manufac-
turer of semiconductor equipment and monocrystalline silicon wafers. It has two
plants in Japan (Sendai, Kouchi) and two in China (Hebei Lang Fang Songgong
Semiconductor Co. Ltd. in Beijing, and Hebei Ningjin Songgong Semiconductor
Co. Ltd. in Ningjin). In April 2007, polysilicon production started at the Soma
Factory in Fukushima Prefecture. For 2012, a production capacity of 7000 metric
tons was reported (Ikki, 2013b), and production was estimated at 5600 metric tons.

Kumgang Korea Chemical Company (South Korea)

Kumgang Korea Chemical Company (KCC) (www.kccworld.co.kr/eng) was estab-
lished by a merger in 2000 of Kumgang and the Korea Chemical Company In
February 2008, KCC announced its investment in the polysilicon industry and
began to manufacture high-purity polysilicon with its own technology at the pilot
plant of the Daejuk factory in July of the same year. It started to mass-produce
polysilicon in 2010, with an annual capacity of 6000 tons, and increased this to
12100 metric tons in 2012 (Ikk 2013a). Production for 2012 is given at around
5400 metric tons.

Daqo New Energy Company Ltd. (China)

Daqo New Energy (www.dqsolar.com) is a subsidiary of the Daqo Group and was
founded by Mega Stand International Ltd. in January 2008. The company built a
high-purity polysilicon factory with an annual output of 3300 tons in the first phase
of construction in Wanzhou. The first polysilicon production line, with an annual
output of 1500 tons, started operation in July 2008. Production capacity in 2009
was 3300 tons and reached more than 4300 metric tons at the end of 2011. In Q4
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 656

656 A. Jäger-Waldau

2012, expansion Phase 2 with 3000 metric tons came online, and according to the
company it will reach a total production capacity of 9300 metric tons at the end
of 2013. The company invested in the downstream business, ranging from wafers,
cells, modules and projects. At the end of 2011, it had a manufacturing capacity of
125 MW wafers and 100 MW modules. Polysilicon shipments of 3585 metric tons
were reported for 2012.

13.4 Outlook
The photovoltaic industry has changed dramatically in the last few years. China has
become the major manufacturing country for solar cells and modules, followed by
Taiwan, Germany and Japan. Amongst the 20 biggest photovoltaic manufacturers
in 2012, only three had production facilities in Europe, namely First Solar (USA,
Germany, Malaysia), Hanwha Q Cells (Germany and Malaysia) and REC (Norway
and Singapore). However, REC closed down its production in Norway in early 2012
and First Solar closed its factory in Germany later in the year.
It is important to remember that the PV industry consists of more than just
the cell and module production, and that the whole PV value chain is relevant.
Besides the information in this chapter, the whole upstream industry (e.g. materials,
equipment manufacturing), as well as the downstream industry (e.g. inverters, BOS
components, system development, installations) has to be looked at as well.
The implementation of the 100,000 roofs programme in Germany in 1990,
and the Japanese long-term strategy set in 1994, with a 2010 horizon, were the
beginning of an extraordinary PV market growth. Before the start of the Japanese
market implementation programme in 1997, annual growth rates of the PV markets
were in the range of 10%, mainly driven by communication, industrial and stand-
alone systems. Since 1990, PV production has increased by almost three orders
of magnitude, from 46 MW to about 38.5 GW in 2012. Statistically documented
cumulative installations worldwide accounted for almost 100 GW in 2012. This
represents mostly the grid-connected photovoltaic market. To what extent the off-
grid and consumer product markets are included is not clear, but it is believed that
a substantial part of these markets are not included in these figures as it is very
difficult to track them.
Even with the current economic difficulties, overall installations are increasing,
due to overall rising energy prices and the pressure to stabilise the climate. In the
long term, growth rates for photovoltaics will continue to be high, even if economic
conditions vary and lead to short-term slowdowns.
This view is shared by an increasing number of financial institutions, which
are turning towards renewables as a sustainable and stable long-term investment.
Increasing demand for energy is pushing the prices for fossil energy resources
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 657

The Photovoltaic Business: Manufacturers and Markets 657

higher and higher. In 2007, a number of analysts predicted that oil prices could
well hit 100 $/bbl by the end of that year or early 2008 (Bloomberg, 2007). After
the spike of oil prices in July 2008, at close to 150 $/bbl, prices decreased due to
the worldwide financial crisis and hit a low of around 37$/bbl in December 2008.
Since then, the oil price has rebounded and the IEA reported average prices for
oil imports around 110 $/bbl since the second quarter of 2011, with peaks around
120 $/bbl between February and April 2012 and September 2013. Oil demand has
increased from about 84 million bbl/day in Q1 2009 to around 91 million bbl/day
in Q3 2013, and a moderate further increase to 93 million bbl/day by Q4 2014 is
expected. Even though no significant changes are currently forecast by analysts
for 2014, the fundamental trend, that increasing demand for oil will drive its price
higher again, is intact and will be evident as soon as the global economy recovers.
Over the last 20 years, numerous studies of the potential growth of the PV
industry and the implementation of PV electricity generation systems have been
produced. In 1996, the Directorate-General for Energy of the European Commis-
sion (EC) published a study entitled Photovoltaics in 2010 (European Commission,
1996). The medium scenario of this study was used to formulate the EC’s White
Paper target of 1997, which was to have a cumulative installed capacity of 3 GW
in the EU by 2010 (European Commission, 1997). The most aggressive scenario in
this report predicted a cumulative installed PV capacity of 27.3 GW worldwide and
8.7 GW in the EU for 2010. This scenario was called ‘Extreme scenario’ and it was
assumed that a number of breakthroughs in technology and costs, as well as con-
tinuous market stimulation and elimination of market barriers, would be required
to achieve it. The reality check reveals that even the most aggressive scenario is
lower than what we expect from the current developments.
According to investment analysts and industry prognoses, solar energy systems
will continue to grow at high rates in the coming years. The different photovoltaic
industry associations, as well as Greenpeace, the European Renewable Energy
Council (EREC) and the IEA, have developed new scenarios for the future growth
of PV. Table 13.1 shows the various scenarios of the Greenpeace/EREC study,
as well as the different 2013 IEA World Energy Outlook scenarios and the IEA
PV Technology Roadmap of 2010. It is interesting to note that the 2015 capacity
values of only two scenarios — the Greenpeace (revolution) and IEA 450 ppm
Scenarios — were not reached at the end of 2013. With forecast new installations
of between 93 and 106 GW in 2014 and 2015, even the Greenpeace revolution
scenario is no longer fictional thinking (Bloomberg New Energy Finance, 2014).
The above-mentioned solar photovoltaic scenarios will only be possible if
solar cell and module manufacturing are continuously improved and novel design
concepts are realised, as with current technology the demand for some materials,
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 658

658 A. Jäger-Waldau

Table 13.1 Predicted cumulative solar electrical capacities until 2035. Data sources:
Bloomberg New Energy Finance, 2014; Greenpeace, 2012; IEA, 2010, 2013d.

2012 2013 2015 2020 2030 2035

Year [GW] [GW] [GW] [GW] [GW] [GW]

Actual installations 100 138

Bloomberg New Energy 231–244
Finance Market Outlook
Greenpeace∗ (reference 88 124 234 290
scenario)
Greenpeace∗ (revolution 234 674 1764 2420
scenario)
IEA Current Policy 133 352 545 680
Scenario∗∗
IEA New Policy Scenario 138 379 710 951
IEA 450 ppm Scenario∗∗ 144 422 950 1389
IEA PV Technology 76 210 872 1330
Roadmap∗∗∗

∗ 2035 values are extrapolated, as only 2030 and 2040 values are given
∗∗ 2015 and 2030 values are extrapolated, as only 2011, 2020 and 2035 values are given
∗∗∗ 2015 and 2035 values are extrapolated, as only 2010, 2020, 2030 and 2040 values are given

like silver, would dramatically increase the costs for these resources within the
next 30 years. Research to avoid such problems is underway and it can be expected
that these bottle-necks will be avoided. With 100 GW cumulative installed pho-
tovoltaic electricity generation capacity installed worldwide by the end of 2012,
photovoltaics still is a small contributor to the electricity supply, but its importance
for our future energy mix is finally acknowledged.

References
Argentina (2006), Boletín Oficial de la República de Argentina, Número 31.064
(www.diputados-catamarca.gov.ar/ley/BO020107.pdf).
Belgisch Staatsblad (2012), 182e Jaargang, No 237, 20 Juli 2012, Derde Editie, N.
2012—2122 [C − 2012/35855] 13 JULI 2012.—Decreet houdende wijziging van het
Energiedecreet van 8 mei 2009, wat betreft de milieuvriendelijke energieproductie (1),
p. 40448ff.
Belgisch Staatsblad (2013), 183e Jaargang, No 11, 16 Januari 2013, [C − 2013/11001],
21 DECEMBER 2012.—Koninklijk besluit tot wijziging van het koninklijk besluit
van 16 juli 2002 betreffende de instelling van mechanismen voor de bevordering van
elektriciteit opgewekt uit hernieuwbare energiebronnen, p. 1574ff.
Bloomberg (2007), ‘$100 oil price may be months away, say CIBC, Goldman’, 23 July 2007
(http://www.bloomberg.com/apps/news?pid=newsarchive&sid=ajxtV4oWcHk0).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 659

The Photovoltaic Business: Manufacturers and Markets 659

Bloomberg New Energy Finance (2012), ‘Analyst reaction—solar’, 17 August 2012.

Bloomberg New Energy Finance (2013a), ‘Press release’, 14 January 2013 (http://about.
bnef.com/press-releases/new-investment-in-clean-energy-fell-11-in-2012-2/).
Bloomberg New Energy Finance (2013b), ‘Scaling up financing to expand the renewables
portfolio’, presentation given by Michael Liebreich at the IEA Renewable Energy
Working Party, Paris, 9 April 2013.
Bloomberg New Energy Finance (2013c), ‘PV market outlook Q2 2013’, 13 May 2013.
Bloomberg New Energy Finance (2013d), ‘Analyst reaction—solar’, 31 May 2013.
Bloomberg New Energy Finance (2013e), ‘China said to prepare import duties on polysili-
con’, 16 May 2013.
Bloomberg New Energy Finance (2014), ‘Q2 2014 PV Market Outlook–A time for cautious
expansion’, 15 May 2014.
Blue Energy (2012), ‘Press release’, 8 December 2012.
Bridge to India (2013), India Solar Compass, 20 April 2013 (www.bridgetoindia.com).
Bulgarian State Gazette (2012a), Decree 147/2012 from 28 March 2012, 10 April 2012
(http://dv.parliament.bg).
Bulgarian State Gazette (2012b), ‘Law amending the Law on Energy, ORDINANCE No
271 from 3 July 2012’, 17 June 2012 (http://dv.parliament.bg).
Bundesgesetzblatt für die Republik Ősterreich (2012), Jahrgang 2012 Ausgegeben am 18
September 2012, Teil II, 307. Verordnung: Ökostrom-Einspeisetarifverordnung 2012
– ÖSET-VO 2012.
Bundesnetzagentur (German Federal Network Agency) (2013), Datenmeldungen für 2012
(www.bundesnetzagentur.de/DE/Sachgebiete/ElektrizitaetundGas/Unternehmen_
Institutionen/ErneuerbareEnergien/Photovoltaik/DatenMeldgn_EEG-VergSaetze/
DatenMeldgn_EEG-VergSaetze_node.html).
Ciarreta A., Espinosa M. P. and Pizarro-Irizar C. (2012a), ‘Efecto de la energía renovable
en el mercado diario de electricidad’, Escenario 2020, Cuadernos Económicos de ICE
83, 101–116.
Ciarreta A., Espinosa M. P. and Pizarro-Irizar C. (2012b), ‘The effect of renewable energy in
the Spanish electricity market’, Lecture Notes in Information Technology 9, 431–436.
CIS Countries Legislation Database (2009), ‘Law of the Republic of Kazakhstan, from 4
July 2009 of No. 165-IV ZRK, “About support of usage of renewable energy sources”’
(http://cis-legislation.com).
Colville F. (2013), ‘UK enters global PV top six with record-breaking half year – Part Two’,
Solar Business Focus, 24 September 2013 (www.solarpowerportal.co.uk).
Comición National de Energía (2012), Annual Report 2012, 18 April 2013 (www.cne.es).
The Daily Star (2011), ‘Target 500 MW solar project’, 15 May 2011 (www.thedailystar.net).
DECC (2011), ‘Huhne takes action on solar farm threat’, 7 February 2011
(https://www.gov.uk).
Denmark (2012), Klima-, Energi- og Bygningsministeriet, Bekendtgørelse om nettoafreg-
ning for egenproducenter af elektricitet, Lovtidende A, BEK nr 1068 af 16/11/2012
(www.lovtidende.dk/Forms/L0700.aspx?s 31=10&s19=1068+&s20=2012).
Dominican Republic (2007), Ley No. 57-07 sobre Incentivo al Desarrollo de Fuentes Ren-
ovables de Energía y de sus Regímenes Especiales (Dominican Republic).
EEG (Erneuerbare-Energien-Gesetz) (2000), Bundesgestzblatt Jahrgang 2000 Teil I, Nr. 13,
p. 305, 29 March 2000.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 660

660 A. Jäger-Waldau

ENF (2013), ‘China production database’, 2013 (www.enfsolar.com).

Episolar (2012), ‘Siginik Energy partners with the electricity company of Ghana to
develop largest solar installation in West Africa’, 4 June 2012 (www.marketwired.
com/press-release/siginik-energy-partners-with-electricity-company-ghana-develop-
largest-solar-installation-1664719.htm).
Ernst & Young (2011), ‘Renewable energy country attractiveness indicator’, November,
Issue 31.
ESDM (Ministry of Energy and Mineral Resources) (2012), ‘Development of new and
renewable energy and energy conservation in Indonesia’, presented at Global Workshop
on Clean Energy Development, Washington DC, 3–8 December 2012.
EUR-Lex (2007), ‘Tarifs réglementés de l’électricité en Espagne’, Official Journal of the
European Communities, C 43/50, 27 February 2007.
European Commission (1996), Photovoltaics in 2010, European Commission, Directorate-
General for Energy, Office for Official Publications of the European Communities,
ISBN 92-827-5347-6.
European Commission (1997), Energy for the Future: Renewable Sources of Energy: White
Paper for a Community Strategy and Action Plan, COM (1997) 599 Final (26/11/97).
European Photovoltaic Industry Association (2013), Global Market Outlook for Photo-
voltaics until 2017 (www.epia.org).
European Wind Association (2013), ‘Wind in power—2012 European statistics’, February
2013 (www.ewea.org).
Gestore Servizi Energetici (2013a), ‘Rapporto Statistico sul solare fotovoltaico 2012’, 8
May 2013.
Gestore Servizi Energetici (2013b), ‘Conto Energia: raggiunti i 6,7 miliardi di euro’, 6 June
2013 (www.gse.it).
Global Sphere (2013), ‘Construction begins on 300 million USD solar panel factory in Hue’,
17 January 2013 (www.global-sphere.com).
GlobalData (2013), ‘Concentrated Photovoltaic (CPV), 2012 Global Market Size, Average
Installation Price, Market Share, Regulations and Key Country Analysis to 2020’,
GlobalData Report Store, January 2013.
Government Gazette (2012), ‘Spanish Royal Decree 1/12’, 28 January 2012 (www.boe.es).
Government of Kazakhstan (2013), ‘Government of Kazakhstan targets 1,040 MW of renew-
able energy by 2020’, 6 February 2013 (www.kazakhembus.com).
Government of Wallonia (2013), ‘Dossier électricité: traduction juridique de l’accord
Solwatt et approbation de la tarification progressive et solidaire’, 30 May 2013
(http://nollet.wallonie.be).
Greenpeace (2012), The Energy [R]evolution 2012, July 2012, ISBN 978-90-73361-92-8.
HEDNO S.A. (Hellenic Electricity Distribution Network Operator) (2013), ‘Press release:
issuance of power generation informative report for the non-interconnected islands,
January 2013’, 5 March 2013.
IEA (International Energy Agency) (2010), PV Technology Roadmap.
IEA (International Energy Agency) (2013a), Medium-term Renewable Energy Market
Report 2013—Market Trends and Projections to 2018, IEA, Paris.
IEA (International Energy Agency) (2013b), ‘Photovoltaic power systems programme,
national survey report of PV power applications in Japan 2012’, 31 May 2013.
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 661

The Photovoltaic Business: Manufacturers and Markets 661

IEA (International Energy Agency) (2013c), ‘Photovoltaic power systems programme,

annual report 2012’, 30 April 2013.
IEA (International Energy Agency) (2013d), World Energy Outlook 2013, ISBN 978-92-
64-2013-9.
IHS Global Demand Tracker (2013), ‘Global PV installations to exceed 35 gigawatts in
2013; Asia & Americas more than compensate for ailing Europe’, April 2013.
Ikki O. (2013a), PV Activities in Japan, Volume 19, No 4, April 2013.
Ikki O. (2013b), Pv Activities in Japan, Volume 19, No 7, April 2013.
Italy (2012), Supplemento n. 143 alla Gazzetta n. 159 del 10/07/2012, Ministero dello
sviluppo economico, decreto ministeriale 5 luglio 2012; Attuazione dell’articolo 25
del decreto legislativo 3 marzo 2011, n. 28, recante incentivazione della produzione di
energia elettrica da impianti solari fotovoltaici (cd. Quinto Conto Energia).
Kaz Silicon (2011), ‘Production of photovoltaic modules based on Kazakhstan silicon’
(www.kazsilicon.kz/en/node/16).
KfW (Kreditanstalt für Wiederaufbau) (2013), ‘KfW-Programm Erneuerbare Energien
“Speicher”’, Programmnummer 275 (www.kfw.de).
Kruangam D. (2013), ‘Current status of solar energy policy and market in Thailand’, 9th
Workshop on the Future Directions of Photovoltaics, Tokyo Institute of Technology,
7–8 March 2013.
LAGIE (Operator of Electricity Market) (2013), ‘RES and CHP statistics, April 2013’, 27
May 2013.
Mercom Capital Group (2013), ‘Solar report’, 18 March 2013.
Mexico (2008), DECRETO por el que se expide la Ley para el Aprovechamiento de
nergías Renovables y el Financiamiento de la Transición Energética, Diario Ofi-
cial (Primera Sección) 88, 28 November 2008 (Mexico) (www.diputados.gob.mx/
LeyesBiblio/ref/laerfte/LAERFTE_orig_28nov08.pdf).
Mexico (2012), Se Expide la Ley General de Chambio Climático, Diario Oficial
(Segunda Sección) 1, 6 June 2012 (Mexico) (www.diputados.gob.mx/LeyesBiblio/
ref/lgcc/LGCCorig_06jun12.pdf).
Ministère de l‘Écologie, du Développement Durable et de l’Energie (2013), 6 February
2013 (www.developpement-durable.gouv.fr/Quels-sont-les-tarifs-d-achats).
Ministry of Economic Affairs (2013), ‘Bureau of Energy, Press Release’, 28 May 2013
(http://web3.moeaboe.gov.tw).
Ministry of Energy, Kenya (2010), ‘Feed-in-tariffs policy for wind, biomass, small hydros,
geothermal, biogas and solar’, 1st revision, January 2010 (http://kerea.org).
MIIT (Chinese Ministry of Industry and Information Technology) (2012), ‘Issuance of the
solar photovoltaic industry “Second Five-Year Development Plan” ’, 24 February 2012
(www.miit.gov.cn).
NanoWin Thin Film Technology (2013), ‘Zhambilskie Electricheskie Seti LLP’, 9 May
2013 (www.nanowin.com.tw/news.php?id=60).
NEA (National Energy Administration) (2012), ‘Press release’, 8 August 2012
(www.nea.gov.cn).
NEA (National Energy Administration) (2013), ‘China’, 8 January 2013 (www.nea.gov.cn).
PEW Charitable Trusts (2013), ‘Who’s winning the clean energy race?’
(www.pewenvironment.org).
September 17, 2014 11:4 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-ch13 page 662

662 A. Jäger-Waldau

PHOTON International (2012), ‘PHOTON International market survey reveals Chinese

companies dominate top 10’, March 2012 (www.pv-tech.org/news/china_dominates_
top_ten_global_solar_manufacturers).
Photovoltaik Guide (2013), ‘Schweiz: Preise von Photovoltaikanlagen auf marktübliches
Niveau gesunken’, press release, 5 April 2013
(www.photovoltaik-guide.de/schweiz-preise-von-photovoltaikanlagen-auf-
marktuebliches-niveau-gesunken-26098).
PV News (2013), published by Greentech Media, ISSN 0739-4829 (www.pv-
magazine.com).
RES LEGAL Europe (2013), ‘Legal sources on renewable energy’, April 2013 (www.res-
legal.eu/home).
RTS Corporation (2013), ‘PV activities in Japan’, Vol. 18 (www.rts-pv.com).
Parliament of Ghana (2011), ‘Renewable Energy Bill’ (www.parliament.gh).
Philippine Department of Energy (2008), ‘An act promoting the development, utilization
and commercialization of renewable energy resources and for other purposes’, 16
December 2008 (www.doe.gov.ph).
The President of the Republic of Indonesia (2006), ‘National Energy Policy, Presidential
Regulation 5/2006’, 25 January 2006 (http://faolex.fao.org).
SEDA (Sustainable Energy Development Authority) (2013), ‘Current governing regulations
of renewable energy for feed-in tariff’, Seminar on RE Standards and FiT Implemen-
tation, 21 May 2013.
SEIA (Solar Energy Industry Association) (2013), ‘US solar market insight, 2012 year in
review, executive summary’ (www.seia.org).
SERC (State Electricity Regulatory Commission, People’s Republic of China) (2013), 15
January 2013 (www.serc.gov.cn).
SEWRC (Bulgarian State Energy and Water Regulation Commission) (2012a), ‘Decision
No. C-33’, 14 September 2012 (www.dker.bg).
SEWRC (Bulgarian State Energy and Water Regulation Commission) (2012b), ‘Decision
No. C-28’, 29 August 2012 (http://dker.bg).
Solarbuzz (2013), ‘Marketbuzz 2013, press release’, 11 March 2013.
Strategies Unlimited (2013), ‘SPV market research product roadmap for 2013’, April 2013.
Systèmes Solaires (2013), ‘Photovoltaic energy barometer’, 9 April 2013.
UNB Connect (2011), ‘ADB assures fund for 500 MW solar system’, 4 June 2011
(www.unbconnect.com).
World Bank (2010), ‘Winds of change—East Asia’s sustainable energy future’, May 2010
(http://web.worldbank.org).
World Bank (2013), ‘Bangladesh—lighting up rural communities’, 15 April 2013
(www.worldbank.org).
September 17, 2014 11:7 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-app page 663

APPENDIX I

FUNDAMENTAL CONSTANTS

Quantity Symbol Value

Speed of light in vacuum c 2.998 × 108 m s−1

Planck constant h 6.626 × 10−34 J s
Boltzmann constant k 1.3807 × 10−23 J K−1
Permittivity of vacuum
o 8.854 × 10−12 F−1 m−1
Unsigned charge on an electron q 1.602 × 10−19 C
Stefan–Boltzmann constant σSB 5.6705 × 10−8 W m−2 K−4

663
September 17, 2014 11:7 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-app page 664

APPENDIX II

USEFUL QUANTITIES
AND CONVERSION FACTORS

1 eV = 1.6022 × 10−19 J = 96.488 kJ mo1−1 = 8065.6 cm−1 = 2.4180 × 1014 Hz

1 cm−1 = 1.9864 × 10−23 J = 11.963 J mo1−1 = 0.12398 meV = 2.9979 × 1010 Hz
1239.8/(λ/nm) = E/eV
kT at 298.15 K = 0.02569 eV
kT/q at 298.15 K = 0.02569 V
(ln 10)kT/q at 298.15 K = 0.05915 V
h/2π = 6.582 × 10−16 eV s
1 kWh = 3.6 × 106 J
1 thermochemical calorie = 4.184 J
1 horse power = 0.7457 kW
1 atm = 1.01325 × 105 Pa
1 torr = 1/760 atm
1 Å = 10−8 cm
1 mile = 1.6093 km
1 hectare = 10−2 km2 = 2.471 acre = 3.861 × 10−3 mile2
1 acre = 4.047 × 10−3 km2 = 0.4047 hectare = 1.5626 × 10−3 mile2
1 litre = 1 dm3 = 0.26417 US gallon = 0.21997 UK gallon
1 lb = 0.45359 kg
1 tonne (1 metric ton) = 103 kg
1 ton (UK) = 1 long ton (US) = 1.0160 tonne
1 ton (US) = 1 short ton (US) = 0.90718 tonne

664
September 17, 2014 11:7 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-app page 665

APPENDIX III

LIST OF SYMBOLS

Preferred
Symbol Unit Description

a absorptance (absorptivity)
A electron acceptor
Ae eV electron affinity
Ah eV hole affinity
ci mol dm−3 molar concentration of species i
D electron donor
D(E) cm−3 eV−1 density of electronic states of energy E
Di cm2 s−1 diffusion coefficient of species i
De,h cm2 s−1 diffusion coefficient of electron or hole
E eV internal energy
molecule−1
or kJ mol−1
Ec , Ev eV conduction or valence band-edge energy
Eg eV bandgap energy or minimum energy for electronic
molecule−1 excitation
or kJ mol−1
EF eV Fermi level
E Fn , E F p eV Fermi level in n or p material
E Fe , E Fh eV quasi-Fermi level of an electron or hole
g cm−3 s−1 or volume rate of generation of excited states
m−3 s−1
G eV Gibbs free energy
molecule−1
or kJ mol−1
H eV enthalpy
molecule−1
or kJ mol−1
H m cell width
H p,n m width of the p or n layer
HRP eV electronic coupling matrix element between reactant (R)
molecule−1 and product (P) states
i mA cm−2 current density
i ph mA cm−2 photocurrent density
ij mA cm−2 junction (dark) current density

665
September 17, 2014 11:7 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-app page 666

666 Appendices

Preferred
Symbol Unit Description

io mA cm−2 reverse-bias saturation current density

I A current
I W m−2 irradiance
Iλ W m−2 spectral irradiance per unit wavelength
nm−1
ISo W m−2 solar constant: annual average normal solar irradiance just
outside the Earth’s atmosphere
j ph photons m−2 flux of photons with λ ≤ λg
s−1
jgo photons m−2 incident flux of solar photons with λ ≤ λg
s−1
jSo photons m−2 incident solar photon flux of all wavelengths
s−1
jλ photons m−2 spectral photon flux with respect to wavelength
s−1
k wave vector
kA
c,v m6 s−1 band-to-band Auger recombination rate constant for the
conduction or valence band
kTA
c,v m6 s−1 trap-assisted Auger recombination rate constant for the
conduction or valence band
kT
e,h m3 s−1 Shockley–Hall–Reed trap recombination rate constant for
an electron or hole
knr s−1 rate constant for nonradiative decay
kr s−1 rate constant for radiative decay
L e,h m minority carrier diffusion length of an electron or hole
L εe,h m minority carrier drift length of an electron or hole
L W m−2 sr−1 radiance
Lλ W m−2 sr−1 spectral radiance with respect to wavelength
nm−1
m∗e , m∗h kg effective mass of an electron or hole
M W m−2 radiant emittance
MS W m−2 radiant emittance of the Sun
n cm−3 or electron concentration in the conduction band of a
m−3 semiconductor
no cm−3 or equilibrium electron concentration
m−3
nr refractive index
NA , ND cm−3 or density of acceptor or donor impurities
m−3
Nc , Nv cm−3 or effective density of states in conduction or valence band
m−3
September 17, 2014 11:7 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-app page 667

Appendices 667

Preferred
Symbol Unit Description

p atm or Pa pressure
p cm−3 or hole concentration in the valence band of a semiconductor
m−3
po cm−3 or equilibrium hole concentration
m−3
P W cm−2 or power density
W m−2
q C charge of an electron
Q C electrical charge
Q p , Qn m width of the p- or n-layer quasineutral region
r cm−3 s−1 or volume rate of recombination of excited
m−3 s−1
rλ 1 reflectivity at wavelength λ
R  resistance
R
 square−1 sheet resistance
RL  load resistance
Se , Sh m s−1 surface recombination velocity for an electron or hole
T K absolute temperature
T transmittance (transmissivity)
ue , uh cm2 s−1 electron or hole mobility
V−1
uE cm s−1 drift speed in electric field strength E
UOx,Rd V vs. ref. electrode potential for the reaction Ox + ne−
Rd vs.
reference electrode
Uo V vs. ref standard electrode potential (based on activities)
U o V vs. ref formal electrode potential (based on concentrations)
V m3 volume
V V voltage
Vb V band-bending potential
Vbo V equilibrium band-bending potential
Vt h V thermal voltage (β) kT/q (= 0.02569 V for β = 1 and
25◦ C)
α cm−1 or optical absorption coefficient
m−1
αλ cm−1 or optical absorption coefficient at wavelength
m−1
β diode quality factor
ε emissivity

o permittivity of a vacuum

op optical dielectric constant (
op = n r2 )

s static dielectric constant
E V m−1 electric field strength
September 17, 2014 11:7 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-app page 668

668 Appendices

Preferred
Symbol Unit Description

η power conversion efficiency

ηmp power conversion efficiency at maximum power
κ quantum mechanical transmission coefficient
λ nm wavelength
λg nm bandgap wavelength
µi eV chemical potential of species i
molecule−1
µ̃i eV electrochemical potential of species i
molecule−1
ν Hz frequency
ν̄ cm−1 wavenumber
ρ C cm−3 space-charge density
ρ  cm electrical resistivity
φ V inner electric potential (Galvani potential)
φ quantum yield
φb V barrier potential
φscl V space-charge layer potential drop
 eV work function
χ V surface (dipole) electric potential
ψ V outer (Volta) electric potential
 steradian (sr) solid angle
September 17, 2014 11:7 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-app page 669

APPENDIX IV

ACRONYMS AND ABBREVIATIONS

AM Air Mass
BHJ bulk heterojunction
BIPV Building Integrated Photovoltaics
BOS Balance Of System
BSF Back-Surface Field
cb (as a superscript or subscript) conduction band
CBE Chemical Beam Epitaxy
CLEFT Cleavage of Lateral Epitaxial Films for Transfer
CPV Concentrated Photovoltaics
CSP Concentrated Solar Power
CVD Chemical Vapour Deposition
CZ Czochralski (method of crystal growth)
DN(I) Direct Normal (Irradiance)
Dk (as a superscript or subscript) in the dark
DoS Density of States
EFG Edge-defined Film-fed Growth
EPC Engineering, Procurement and Construction
EQE External Quantum Efficiency
fb (as a superscript or subscript) flat band
FZ Floatzone (method of crystal growth)
G Global (irradiance)
HCPV High-Concentration Photovoltaics
IQE Internal Quantum Efficiency
ITO indium tin oxide
Lt (as a superscript or subscript) in the light
LCL lateral conduction layer
LGBC Laser-Grooved Buried Contact
LOLP Loss-Of-Load Probability
LPE Liquid Phase Epitaxy
MBE Molecular Beam Epitaxy
MIS Metal|Insulator|Semiconductor (junction or device)
MOCVD Metal Organic Chemical Vapour Deposition
mp (as a superscript or subscript) maximum power
MS Metal|Semiconductor (junction or device)
NOCT Nominal Operating Cell Temperature
oc (as a superscript or subscript) open circuit

669
September 17, 2014 11:7 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-app page 670

670 Appendices

Ox oxidant
PCE power conversion efficiency
PEDOT:PSS poly(3,4-ethylenedioxythiophene):poly(styrenesulphonate)
PTC Practical Test Conditions (for cell efficiency measurements):
(plane-of-array irradiance 1000 W m−2 AM1.5G, ambient temperature
25◦ C, wind speed air 1 m s−1 )
PPV poly( p-phenylenevinylene)
PV photovoltaic
QNR quasi-neutral region
RAPS Remote Area Power Supplies
Rd Reductant
sc (as a superscript or subscript) short circuit
sc1 (as a superscript or subscript) space-charge layer
SHE Standard Hydrogen Electrode
SHS Solar Home System
SIS Semiconductor|Insulator|Semiconductor junction
ss (as a superscript or subscript) surface state
STC Standard Test Conditions (for cell efficiency measurements):
plane-of-array irradiance AM1.5G 1000 W m−2 , cell temperature
25 ± 1/2 C
TCO Transparent Conducting Oxide
TPV thermophotovoltaic
vb (as a subscript or superscript) valence band
Wp peak watts
September 17, 2014 11:9 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-index page 671

INDEX

AC modules, 591 building-integrated photovoltaic (BIPV)

Advanced Silicon Materials LLC, 654 systems, 601–606
Agua Caliente Solar Project, 1 built-in voltage, 45
Alamosa Solar Plant, 546, 564 Bulgaria
AlGaAs/GaAs/Ge cells, 311 PV market, 631
AlInGaP/GaAs/Ge cells, 311 bulk heterojunctions, 17, 33, 46, 66,
AlInP window layer, 319 352–354, 357, 358, 361, 362, 364, 376,
AM1.5 solar spectra, 46 377, 384, 385, 389, 458, 475
Amonix, 495, 502, 539, 541, 545 bypass diodes, 582
ARCO Solar, 16, 246, 268
Argentina Canada
PV market, 640 Canadian Solar Inc., 647
Astronergy Solar, 650 PV market, 639
Auger recombination, 103, 104, 125, 278, Capo Verde
483, 502 PV market, 637
Australia Carnot efficiency, 49, 52
PV market, 620 CdSTe, 227, 236
Austria band bowing, 234–236
PV market, 636 CdSTe-alloy region, 225–227
azimuth angle, 583–585, 607 CdTe
azimuth-elevation tracking, 494, 532 phase diagram, 214, 215
CdTe solar cells
balance of system (BOS), 572, 589, 599, back contact, 234
610 buffer layer, 221
band tail, 148, 150, 294 capacitance–voltage analysis, 232
bandgap CdCl2 activation, 230
amorphous Si, 140, 148 CdCl2 removal, 231
microcrystalline Si, 140, 155 CdS layer, 223
Bangladesh CdSTe layer, 227
PV market, 625 CdTe deposition temperature, 233
barrier layer cells, 7 CdTe layer, 225
battery storage, 594–596 commercial development, 212
Belgium contact interface layer (CIFL), 231,
PV market, 630 234
black dye, 427 energy payback time, 239
black-body radiation, 46 environmental considerations, 238
Boeing Corporation, 246, 268 high-efficiency NREL cell, 221
Brazil history, 211
PV market, 640 minerals availability, 237

671
September 17, 2014 11:9 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-index page 672

672 Index

reliability, 237 compound triangular concentrator

site of junction, 236 (CTC), 514
superstrate materials, 219 cost reduction, 314
TCO layer, 220 declination–hour angle tracking, 495,
cell design 498, 532
substrate configuration, 214 early development, 494
superstrate configuration, 213 ENTECH concentrator, 495
centralised PV systems, 606, 607, 610 EUCLIDES concentrator, 496, 539
Centre for Solar Energy and Hydrogen plant power rating methodology, 550
Research (ZSW), 246, 270 Ramón Areces concentrator, 494, 545
Chile standard test conditions, 547
PV market, 641 trumpet concentrator, 513
China Cu x S/CdS solar cell, 210
Astronergy Solar, 650 Cu(In,Ga)3 Se5 , 265
Daqo New Energy, 655 Cu(In,Ga)(S,Se)2
EGing Photovoltaic Technology, 650 graded-gap devices, 293
GCL-Poly, 653 Cu(In,Ga)Se2
crystal structure, 247
Hanwha SolarOne, 646
defect physics, 250, 290
Hareon Solar Technology, 649
grain boundaries, 266
JA Solar Holding, 646
phase diagram, 249
Jiangxi Risun Solar Energy, 651
surface states, 265
Jinko Solar, 648
Cu(In,Ga)Se2 solar cells, 245
PV market, 618, 620, 622
band diagram, 271, 285
Suntech Power, 647
carrier lifetimes, 279
Trina Solar, 646
CdS buffer layer, 267, 270
Yingli Green Energy, 645 cell structure, 253
chlorophyll collection losses, 275
in OPV devices, 350 diode quality factor, 281–283, 285
CIGS, 245 fill factor, 285
tandem cell with DSSC, 432 heterojunction formation, 266
close-space vapour transport (CSVT), 210 history, 245
compound parabolic concentrator (CPC), influence of oxygen, 262
513 influence of sodium, 253–255, 260
compound triangular concentrator (CTC), interconnects, 268
514, 515 MoSe2 layer, 255, 275, 290
COMSAT Laboratories, 92, 93 open-circuit voltage, 275, 279
Concentrix, 541, 543, 544 optical losses, 273
congruent condensation, 216, 217 recombination losses, 282
congruent sublimation, 217 recombination processes, 277, 280
CPV (concentrated photovoltaics) selenisation, 258, 269
acceptance–concentration product, tunnelling recombination, 281
512 vacancy relaxation, 285–287
azimuth-elevation tracking, 494, 532 window layer, 268
compound parabolic concentrator ZnO window layer, 253, 267, 268,
(CPC), 513 270, 273, 274
September 17, 2014 11:9 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-index page 673

Index 673

Czochralski (CZ) crystal growth process, energy payback time, 144, 238, 239
90, 94, 106–109 Entech Solar, 539, 541
CZTS (kesterite) solar cells, 294 environmental impact of PV, 608
etendue, 46, 49, 51, 510, 511, 518, 519
Daqo New Energy, 655 Euro-efficiency value, 591
declination–hour angle tracking, 495, 498, EuroCIS, 16, 268
532 European Union
defect PV market, 618
charged, 151, 152, 156 excitons, 344
dangling bond, 148, 152 CT states, 375
density, 150, 152 exciton quenching, 345, 349, 394
equilibrium, 150–153 field-assisted dissociation, 374
neutral, 149, 151, 152 hot excitons, 376
Denmark triplet excitons, 395
PV market, 631 external quantum efficiency (EQE), 61,
Dexter energy transfer, 346 62, 274, 350
Doha Climate Change Conference, 6 a-Si cell, 156
doping, 151, 155 triple cell, 162
double-hetero (DH) structures, 309, 322, external radiative efficiency (ERE), 61, 65,
324 66
down-conversion, 74, 75, 81, 105
DSSC Förster energy transfer, 346
see dye-sensitised solar cells feed-in tariffs (FiTs), 2, 545, 620–622,
dye-sensitised solar cells (DSSCs) 625, 636
charge screening in, 419, 420 First Solar, 15, 81, 212, 627, 645, 656
device configuration, 415–417 floatzone (FZ) crystal growth process, 109
DSSC/CIGS tandem cell, 432 fluorene, 387
device fabrication, 417, 419 fluorenone, 387
dye regeneration, 419–421, 424 France
energetics of operation, 418–422 PV market, 632
historical background, 414 free carrier absorption (FCA), 176, 185
kinetics of operation, 422–426 fullerenes, 352, 353, 358, 360, 361, 364,
molecular engineering of interface, 370–372, 374–377, 383, 384, 389, 393,
439, 440 395, 400
organic dyes in, 436, 437
panchromatic sensitisers, 426–429 GaAs, 10, 11, 27, 29, 31, 33
quantum dots as sensitisers, 437, 438 cell history, 16
solid-state DSSC, 440, 441 in QW solar cells, 456
stability, 433–436 minority-carrier lifetimes, 317
tandem cells, 431–433 photoluminescence intensity, 319
ultrafast forward electron injection in, GaAsN
423 in QW solar cells, 456, 457
GaAsP
edge-ray theorem, 512, 513 in QW solar cells, 12, 313, 455, 457,
EGing Photovoltaic Technology, 650 478–484
electronic conjugation, 343 GaInP/GaAs tandem cells, 309
September 17, 2014 11:9 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-index page 674

674 Index

GaInP/GaInAs/Ge cells, 11, 505, 506 InP

GCL-Poly, 653 in QW solar cells, 456
Germany interface texture, 168, 169
PV market, 618, 632 etching, 171, 186
Wacker Polysilicon, 652, 653 imprinting, 172
Ghana multi-scale, 177
PV market, 638 nanoimprinting, 177
Gintech Energy Corporation, 648 periodic, 172
Greece intermediate-band devices, 73
PV market, 633 internal radiative efficiency (IRE), 61
grid-connected systems, 571, 587, 601, inverters, 593
606 iodide/triiodide redox couple, 416, 419,
network stability, 608, 610 421
Ground Cover Ratio, 607 slow recombination on TiO2 , 425
Guascor Fotón, 502, 542, 545, 546, 563 islanding, 592
Israel
Hanwha Q-Cells, 647, 656 PV market, 621
Hanwha SolarOne, 646 Italy
Hareon Solar Technology, 649 PV market, 618, 633
Hemlock Semiconductor Corporation, 654
heterojunction with intrinsic thin layer JA Solar Holding, 646
(HIT) cell, 100–103, 120, 125, 129 Japan
hole conductors, 435 Kyocera Corporation, 649
liquid electrolyte, 416 M. Setek Company, 655
solid-state, 18, 19, 417, 440 PV market, 620, 622
hot-carrier cells, 42, 71, 72, 80, 81, 481 Solar Frontier, 651
hybrid systems, 588, 600 Tokuyama Corporation, 655
Jiangxi Risun Solar Energy, 651
I–III–VI2 chalcopyrites Jinko Solar, 648
bandgaps, 247, 288
III–V semiconductors K–19 dye, 434
in QWSCs, 455 Kazakhstan
III–V solar cells, 33, 307 PV market, 626
III–V tandem cell technologies, 308 Kenya
India PV market, 638
PV market, 620, 621 kesterite solar cells,
Indonesia see CZTS solar cells
PV market, 626 Kumgang Korea Chemical Company, 655
InGaAs homojunction cells, 320 Kyocera Corporation, 649
InGaP/GaAs tandem cells, 11, 309 Kyoto Protocol, 6
use in space, 327
InGaP/GaAs/GaInNAs cells, 309 laser scribing, 190
InGaP/GaAs/Ge cells, 311, 327, 330 lattice mismatch, 319
InGaP/GaAs/InGaAs cells, 309, 331 light-induced degradation, 149, 156
InGaP/InGaAs/Ge cells, 309, 331 light management, 168, 174
September 17, 2014 11:9 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-index page 675

Index 675

limiting efficiency, 41 organic photovoltaic cells, 339

isotropic limit, 51 bilayer heterojunctions, 357
Landsberg, 50 bulk heterojunctions, 352, 357
maximal concentration limit, 51 device structure, 348
Shockley–Queisser limit, 42, 54 electronic disorder, 364
thermodynamic limit, 49 inverted structures, 348
Trivich–Flynn limit, 42, 52 morphological stability, 390
photodegradation, 386
M. Setek Company, 655 planar heterojunctions, 351
Malaysia polymer:fullerene devices, 353, 372,
PV market, 627 377, 389
maximum power point tracker, 589 polymer:nanocrystal devices, 394
Mexico tandem devices, 391
PV market, 641 thin-film processing, 346
Miller–Abrahams hopping model, 366 organic semiconductors
MINP cell, 9 electronic conjugation, 340
mobility edge, 368 geminate recombination, 376, 384
mobility gap, 148 photostability, 386
module testing, 581 spontaneous singlet fission, 394
Montrol–Scher model, 438 variable-range hopping, 366
Morocco outdoor performance, 190
PV market, 638
MoSe2 , 255, 275, 290 P3HT:PCBM, 358, 360, 372
Motech Solar, 647 panchromatic dye sensitisers, 426
multijunction solar cells, 140, 142, 307 passivation, 155
computer analysis, 311 PECVD (plasma-enhanced chemical
connection modes, 316 vapour deposition), 140, 147, 162, 187
multiple exciton generation (MEG), 72, PEDOT:PSS, 46, 359, 390
438 PERC cell, 9, 97
multiple quantum well cells, 74 PERF cell, 99
performance ratio, 597
N3 dye, 418, 427, 434 PERL cell, 9, 99
N719 dye, 418, 424, 434 perovskite solar cells, 18
n-i- p configuration, 142, 172 PERT cell, 99
National Renewable Energy Laboratory, Peru
246 PV market, 642
Neo Solar Power, 648 PESC cell, 9, 95
nominal operating cell temperature photon recycling, 479
(NOCT), 579 manufacturers, 644
Norway p-i-n junction, 142, 156
Renewable Energy Corporation, 654 Planckian temperature, 48
plastic electronics, 347
OCI Company Ltd., 653 polarons, 346, 375
Ohl, Russell, 8, 90 polymer:fullerene devices
OMeTAD, 440 see organic photovoltaic cells, 353
organic dyes, 436 polysilicon, 3
September 17, 2014 11:9 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-index page 676

676 Index

Portugal Auger recombination, 125

PV market, 636 black cell, 8, 93
power conditioning, 589 buried contact cell, 119
PV arrays, 582 cost breakdown, 122
array orientation, 583 early development, 8
PV modules EWT cell, 121
module configuration, 575 HIT cell, 100, 120
module design, 572 LDSE cell, 119
module testing, 581 LGBG type, 497
packing density, 578
MINP cell, 9
series and parallel connections, 573
multicrystalline, 87
MWT cell, 121
Qatar Solar Technologies, 3
PERC cell, 9, 97
quantum dots
PERF cell, 99
as sensitisers in DSSC, 437
quantum wells (QW) solar cells, 12, 453 PERL cell, 9, 99
applications, 481 PERT cell, 99
current–voltage characteristics, 476 PESC cell, 9, 95
dark current, 472 silicon thin-film cells, 127
device design, 454 silicon-on-ceramic cell, 128
fabrication, 458 violet cell, 8, 93
materials, 455 Si modules, 573, 580
spectral response, 469 Siemens Solar Industries, 246, 258
strain-balanced structures, 454, 457 silicon, 1
quantum wells, 459 amorphous, 139, 147, 164
density of states, 460 Czochralski (CZ) growth, 9, 106
floatzone (FZ) crystal growth, 109
radiation resistance, 327 high-purity, 105
rear point-contact solar cell, 96 metallurgical grade, 9, 105
Renewable Energy Corporation (REC), microcrystalline, 140, 154, 164
654 multicrystalline ingot production, 109
REC Silicon Division, 654 nanocrystalline oxide, 175
REC Solar Singapore, 650, 656
sheet and ribbon growth, 111
roll-to-roll manufacture, 398
SiC alloy, 154
roll-to-roll technology, 190
SiGe alloy, 140, 153
ruthenium bipyridyl dyes, 424
thin-film, 140
Sandia National Laboratories, 494 upgraded metallurgical grade, 106
screen printing, 115 Slovakia
Shockley’s p-n junction theory, 42 PV market, 634
Shockley–Queisser limiting efficiency, 42, Solar Frontier, 651
54 Solar Home System (SHS), 599, 600
Shockley–Queisser theory, 377, 381 Solar Power Corporation, 9
Shockley–Read–Hall recombination, 467 Solibro Research, 246
Si cells, 10, 12, 27, 87 South Africa
antireflection coating, 115 PV market, 639
September 17, 2014 11:9 Clean Electricity From Photovoltaics (2nd Edition) 9in x 6in b1809-index page 677

Index 677

South Korea tilt angle, 583

Kumgang Korea Chemical Company, TiO2 , 415, 421
655 energy levels in DSSC, 420
OCI Company Ltd., 653 mesoporous films of, 438
space applications, 601 nanoparticle fabrication, 417
Spain slow recombination of I− −
3 /I redox
PV market, 634 couple on, 425
spiro-OMeTAD, 440 TiO2 mesoscopic films
Staebler–Wronski effect, 149, 433 electron diffusion length in, 425
stand-alone systems, 571, 587, 599 electron transport in, 424
standard test conditions (STC) Tokuyama Corporation, 655
for CPV, 547 transparent conductive oxide (TCO), 142,
SunEdison, 654 144, 178, 185, 219, 220, 417
sunlight Cd2 SnO4 , 220
standard terrestrial reference spectra, ITO (indium tin oxide), 219
46 SnO2 :F, 220
thermodynamic properties, 46 ZnO:Al, 219
SUNPATH, 3 Trina Solar, 646
SunPower Corporation, 648 Trivich–Flynn limiting efficiency, 52, 59
Sunshine Programme, 324 tunnel junctions, 11, 309, 317, 321, 325
SunShot initiative, 3 Turkey
Suntech Power, 3, 647 PV market, 636
superlattices, 459
Switzerland Ukraine
PV market, 635 PV market, 636
system sizing, 594 UN Climate Change Conference, 6
system yield, 597 United Kingdom
PV market, 635
Taiwan United States of America
Gintech Energy Corporation, 648 First Solar, 645
Motech Solar, 647 Hemlock Semiconductor Corporation,
Neo Solar Power, 648 654
PV market, 624 PV market, 640
tandem cell, 142, 159 SunEdison, 654
organic devices, 391 SunPower Corporation, 648
Thailand
PV market, 625 Vietnam
the Philippines PV market, 628
PV market, 627 Villafranca solar plant, 546, 564
thermionic converters, 78 violet cells, 93
thermodynamic limiting efficiency, 49
thermoelectric converters, 80 Wacker Polysilicon, 652
thermophotonic converters, 77
thermophotovoltaics, 77, 483 Yingli Green Energy, 645