11041_9789813272248_tp.

indd 1 5/7/18 11:34 AM

December 12, 2014 15:40 BC: 9334 – Bose, Spin and Fermi Systems 9334-book9x6 page 2

This page intentionally left blank

11041_9789813272248_tp.indd 2 5/7/18 11:34 AM
 Published 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

British Library Cataloguing-in-Publication Data

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

INTRODUCTION TO SOLID STATE PHYSICS

Copyright © 2019 by World Scientific Publishing Co. Pte. Ltd.
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 978-981-3272-24-8

For any available supplementary material, please visit

https://www.worldscientific.com/worldscibooks/10.1142/11041#t=suppl

Desk Editor: Rhaimie Wahap

Printed in Singapore

Rhaimie - 11041 - Introduction to Solid State Physics.indd 1 17-07-18 11:19:43 AM

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page v

Preface

Almost all natural materials solidify at a sufficiently low temperature. At equi-

librium, these solids are crystals built of elementary units that repeat themselves
periodically. This book introduces the reader to the physics of the solid state, and
is focused on the properties of periodic crystals. It presents just a part of the much
broader field of condensed matter physics, which encompasses also the study of
noncrystalline materials, e.g glasses and soft materials, as liquids, liquid crystals,
polymers, and membranes; the book sets a base for further studies of these more
advanced topics.
The authors are indebted to many friends and colleagues whose comments and
observations improved considerably the representation of the various topics. Special
thanks are due to professors Yoseph Verbin and Yoram Kirsh, who followed closely
our progress, commenting on each step. We thank professors Assa Auerbach, Dror
Orgad, and Yaacov Kantor who reviewed critically our initial proposal for the con-
tents of the book, and professors Alexander Palevski and Efrat Shimshoni who read
each of the chapters and improved the contents and the representation in so many
ways. Shlomi Matityahu and especially Yaacov Kleeorin are acknowledged for their
invaluable help in the preparation of the exercises and many of the illustrations.
Professors Ron Lifshitz, Yuri Rosenberg, and Hagai Shaked shared with us their
expertise in central topics (quasicrystals, X-ray scattering and neutron scattering).
We are grateful to professors Elyahu Eisenberg, Ady Stern and Moshe Schechter for
their useful comments. Last but not least, the staff of the Open University of Israel,
under whose auspices the Hebrew version of this book was prepared, is gratefully
acknowledged.

A. Aharony and O. Entin-Wohlman

v
December 12, 2014 15:40 BC: 9334 – Bose, Spin and Fermi Systems 9334-book9x6 page 2

This page intentionally left blank

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page vii

Contents

Preface v

1. Introduction 1
1.1 What is the nature of the solid state? . . . . . . . . . . . . . . . . . 1
1.2 The spatial dimensions . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 The present book . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2. The crystalline structure of solids 15

2.1 Lattices and crystals in one dimension . . . . . . . . . . . . . . . . 15
2.2 Two-dimensional lattices . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Planar Bravais lattices with a base . . . . . . . . . . . . . . . . . . 23
2.4 The lattice vectors and the unit cell . . . . . . . . . . . . . . . . . 30
2.5 The cubic lattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.6 Dense packing in space . . . . . . . . . . . . . . . . . . . . . . . . 43
2.7 Classification of periodic lattices according to their symmetries . . 48
2.8 Quasicrystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.9 Epitaxial growth, thin films, superlattices and multi-layered struc-
tures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
2.10 Magnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
2.11 Answers for the problems in the text . . . . . . . . . . . . . . . . . 75
2.12 Problems for self-evaluation . . . . . . . . . . . . . . . . . . . . . . 91
2.13 Answers for the self-evaluation problems . . . . . . . . . . . . . . . 97

Appendix A The Platonic bodies 109

A.1 Answers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

3. Radiation scattering off crystals 113

3.1 Introduction: identifying crystalline structures . . . . . . . . . . . 113
3.2 Bragg’s law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
3.3 Scattering off a point-like Bravais lattice . . . . . . . . . . . . . . . 124

vii
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page viii

viii Introduction to Solid-State Physics

3.4 The reciprocal lattice . . . . . . . . . . . . . . . . . . . . . . . . . . 128

3.5 Brillouin zones . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
3.6 The reciprocal lattice and lattice planes . . . . . . . . . . . . . . . 136
3.7 Experimental methods . . . . . . . . . . . . . . . . . . . . . . . . . 139
3.8 Fourier series and reciprocal lattice . . . . . . . . . . . . . . . . . . 145
3.9 The structure factor . . . . . . . . . . . . . . . . . . . . . . . . . . 147
3.10 The temperature dependence; the Debye-Waller factor . . . . . . . 159
3.11 Scattering off quasicrystals . . . . . . . . . . . . . . . . . . . . . . . 164
3.12 Scattering of electrons off surfaces . . . . . . . . . . . . . . . . . . . 166
3.13 Neutron scattering off magnetic crystals . . . . . . . . . . . . . . . 169
3.14 Answers for the problems in the text . . . . . . . . . . . . . . . . . 176
3.15 Problems for self-evaluation . . . . . . . . . . . . . . . . . . . . . . 193
3.16 Answers for the self-evaluation problems . . . . . . . . . . . . . . . 197

Appendix B Fourier series 207

B.1 Answers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213

4. Crystal bonding 215

4.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
4.2 The ionic bond . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
4.3 Thecovalent bond . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
4.4 Molecular crystals–van der Waals bond . . . . . . . . . . . . . . . . 254
4.5 The hydrogen bond . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
4.6 Metallic crystals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
4.7 Magnetism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
4.8 Answers for the problems in the text . . . . . . . . . . . . . . . . . 277
4.9 Problems for self-evaluation . . . . . . . . . . . . . . . . . . . . . . 290
4.10 Answers for the self-evaluation problems . . . . . . . . . . . . . . . 293

Appendix C Topics in quantum mechanics 301

C.1 Answer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307

5. Lattice vibrations 309

5.1 Classical equations of motion in one dimension . . . . . . . . . . . 309
5.2 Classical equations of motion (arbitrary dimensions) . . . . . . . . 322
5.3 The eigenmodes of a finite lattice . . . . . . . . . . . . . . . . . . . 333
5.4 Density of states, specific heat of phonons . . . . . . . . . . . . . . 336
5.5 Crystal stability and the Debye-Waller factor . . . . . . . . . . . . 351
5.6 Impurities, localized modes, scattering of sound waves . . . . . . . 355
5.7 Lattice momentum and inelastic scattering of phonons . . . . . . . 360
5.8 Phenomena related to anharmonic effects . . . . . . . . . . . . . . 363
5.9 Magnons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page ix

Contents ix

5.10 Answers for the problems in the text . . . . . . . . . . . . . . . . . 371

5.11 Problems for self-evaluation . . . . . . . . . . . . . . . . . . . . . . 393
5.12 Answers for the self-evaluation problems . . . . . . . . . . . . . . . 396

Appendix D The continuum limit 411

D.1 Answer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413

6. Electrons in solids 415

6.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 415
6.2 The Drude theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 416
6.3 The Sommerfeld theory . . . . . . . . . . . . . . . . . . . . . . . . 424
6.4 Electrons in a periodic potential: the Bloch theorem . . . . . . . . 436
6.5 The Kronig-Penney potential . . . . . . . . . . . . . . . . . . . . . 442
6.6 Nearly-free electrons . . . . . . . . . . . . . . . . . . . . . . . . . . 448
6.7 The tight-binding approximation . . . . . . . . . . . . . . . . . . . 461
6.8 Metals, insulators, and semiconductors . . . . . . . . . . . . . . . . 472
6.9 Density of states, charge-carriers’ density, specific heat . . . . . . . 484
6.10 The semiclassical equations of motion for electrons in a periodic
potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494
6.11 Electrons in a magnetic field . . . . . . . . . . . . . . . . . . . . . . 500
6.12 A metal or an insulator? . . . . . . . . . . . . . . . . . . . . . . . . 512
6.13 Answers for the problems in the text . . . . . . . . . . . . . . . . . 514
6.14 Problems for self-evaluation . . . . . . . . . . . . . . . . . . . . . . 546
6.15 Answers for the self-evaluation problems . . . . . . . . . . . . . . . 551

Appendix E The Schrödinger equation for electrons in a magnetic field 563

E.1 Answers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 570

7. Selected topics 575

7.1 Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 575
7.2 Soft condensed-matter . . . . . . . . . . . . . . . . . . . . . . . . . 576
7.3 Mesoscopic physics . . . . . . . . . . . . . . . . . . . . . . . . . . . 586
7.4 Answers for the problems in the text . . . . . . . . . . . . . . . . . 605
7.5 Problems for self-evaluation . . . . . . . . . . . . . . . . . . . . . . 612
7.6 Answers for the self-evaluation problems . . . . . . . . . . . . . . . 614

Index 621
December 12, 2014 15:40 BC: 9334 – Bose, Spin and Fermi Systems 9334-book9x6 page 2

This page intentionally left blank

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 1

Chapter 1

Introduction

This chapter begins, in Sec. 1.1, with a survey of the different phases of matter.
The particles in a gas are located at random positions; in a liquid (or in a glass)
there is a larger statistical probability to find pairs of particles for which the two
members are at a certain distance from one another, and in the crystalline solids
the particles are arranged on a periodic lattice. The information on the particles’
locations is encapsulated in the correlation function that quantifies the probability
to find pairs of particles at a given distance in-between them. This book explains
the fundamental concepts needed to describe the physics of such many-body sys-
tems, with examples focusing mainly on periodic solids (i.e., crystals), that are the
energetically-favored structures at low temperatures. The understanding of crys-
talline structures explains the wide range of solid materials’ properties, and paves
the way to the development of advanced technologies.
Carbon appears in Nature in many versatile periodic structures, that exist in
three spatial dimensions, but also in two dimensions (planar formations), in one di-
mension (chain-like networks) and also in zero dimensions (large molecules). These
structures are described in Sec. 1.2, exemplifying many of the structures encoun-
tered in the subsequent Chapters.
Section 1.3 outlines the contents of the book’s chapters: the explanation of
crystalline structures, the description of experimental methods for identifying them,
the analysis of the binding energies determining those structures, the discussion of
the atoms’ vibrations in the crystal, and the derivation of the electronic properties
of solids.

1.1 What is the nature of the solid state?

The physics of the solid state is concerned with systems comprising a very large
number of atoms or molecules; attractive forces cause them to occupy a finite volume
in space, in a rigid formation. The physical laws that govern these forces are based
on the theories of mechanics, electricity, and quantum mechanics. The large
number of atoms requires also the application of the laws of statistical mechanics.

1
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 2

2 Introduction to Solid-State Physics

The same microscopic laws lead to the formation of various arrangements of atoms
(or molecules) in the (macroscopic) solid, and thus to a variety of physical properties.
The theory of the solid state aims to explain the way in which the same physical laws
give rise to widely-disparate physical properties. For instance, why are certain solids
good electrical conductors (i.e., metals), while others are electrically insulating, or
inversely, superconductors? Why are there transparent materials that allow the
passage of light (or sound), and materials that reflect the light (or sound) waves
impinging on them? Why are certain materials rigid, and others are soft? How
does it happen that certain solids possess magnetic properties and others possess
ferroelectric (i.e., comprising electric dipole moments) ones?
The theory of solid-state physics not only explains materials found in Nature,
but also enables the development of novel materials that can be used in new
technologies. Conspicuous examples from the previous century include the ad-
vent of transistors, of integrated circuits and microelectronics, of magnetic mem-
ories, and more. The history of solid-state physics begins with the discoveries of
von Laue and Bragg (the father and the son) related to the periodic diffraction pat-
terns created by X-ray scattering off solid crystals (Chapter 3). Other important
milestones are the explanation of the properties of electrons moving in the periodic
potential of crystals, given by Bloch (Chapter 6), the discovery of the transistor by
Bardeen, Brattain, and Shockley, the understanding of magnetism by Néel, Ander-
son, and van Vleck (Chapters 2, 4, and 6), the discovery of superconductivity by
Kammerlingh-Onnes and its explanation by Bardeen, Cooper and Schrieffer, the
theory of Josephson, and the experiments of Giaver et al. on tunneling between
superconductors. More recent developments, taking place in the last 40 years,
include the invention of the scanning tunneling microscope, the discovery of super-
conductivity at high temperatures, of systems possessing giant magneto-resistances,
of quasicrystals, of the graphene films, of topological phases, and of the quantum
Hall effect. Several of these important topics are surveyed in this book. Many of
the physical examples given in the following Chapters are related to materials and
properties that have been discovered only recently.
Phases of matter. Very crudely speaking, matter can assume one of the
three states or phases: at high temperatures materials are in the gaseous state,
in which the atoms or the molecules are moving freely in the vessel containing
them, filling it randomly and homogeneously. The knowledge of the location of a
certain atom conveys no information about the locations of the others; the average
density of the gas is determined by the ratio of the number of atoms and the vessel’s
volume. (At much higher temperatures there appears the plasma state, in which
atoms are disintegrated into nuclei and electrons.) As the temperature is reduced, it
eventually reaches the boiling temperature, where the material separates into a
denser portion, called liquid, and another portion, the vapor of that material, which
behaves as a gas. This is the gas-liquid phase transition. Due to gravity, the
denser fluid fills the lower part of the vessel. The surface tension of the liquid causes
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 3

Introduction 3

the formation of a film separating the fluid from the gas above it. The liquid is not
rigid, and adjusts itself to the shape of the vessel; it is easily deformed. As in the
gaseous phase, the location of a certain atom in the liquid state has no one-to-one
correspondence with the other atoms’ locations; however, there exists a probabilistic
relation among these locations. As the temperature is decreased further it reaches
the melting temperature, at which the liquid-solid phase transition takes
place. At low temperatures almost all materials become solids. The solid state
is a rigid configuration of the atoms or molecules, where their motion is restricted;
at most they vibrate slightly around their equilibrium positions. A well-known
example for transitions among the various phases is provided by water: at ambient
pressure, water (which is a liquid) freezes and solidifies into ice below 0◦ Celsius,
and turns into aqueous vapour above 100◦ Celsius. (The two temperatures are
defined as the melting and boiling temperatures, and build the temperature scale
of Celsius). These temperatures vary with pressure; e.g., under higher pressures it
is harder for the molecules to turn into a gas, causing the boiling temperature to
rise. Below we mention more phases of matter.
Forces among atoms (or molecules). The various forces operating among
the atoms (or molecules) in the solid state are discussed in Chapter 4. Ubiquitously,
the potential energy, V , that gives rise to the force between a pair of particles,
depends on the distance between the particles , r, as illustrated in Fig. 1.1. At short
distances, the derivative (with respect to r) of the potential energy is negative and
hence the force is repulsive, preventing the particles from approaching too closely
one another. At larger distances the derivative is positive and the force is attractive.
The force vanishes at the minimum of the potential energy, r = rm . Classically,
(i.e., ignoring possible oscillations around the minimum) the point r = rm marks
the equilibrium configuration of the pair.
A simple description of a large ensemble of particles is based on the assumption
that the force operating between each pair of them is derived form the potential
energy displayed in Fig. 1.1. Initially the particles are located at random positions,
keeping the density as dictated by the number of particles divided by the volume of
the container. An initial velocity, of a random magnitude and a random direction, is
assigned to each particle; these velocities are chosen from a probability distribution
that depends on the temperature and is found according to the laws of statistical
mechanics. One then solves the Newtonian equations of motion to obtain (numer-
ically) the particles’ locations at later times. This method is named “molecular
dynamics”. Allowing the computer code to run a long enough time, the system
presumably achieves its equilibrium configuration. Figure 1.2 shows three instanta-
neous snapshots obtained by this procedure, pertaining to three different densities.
Each particle is represented by a sphere of a finite radius, to emphasize the repulsion
among the particles that prevents them from approaching each other (the spheres
cannot penetrate one into the other). At equilibrium the instantaneous snapshots
taken at different times are qualitatively the same; they are statistically similar.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 4

4 Introduction to Solid-State Physics

Fig. 1.1: Schematic plot of the potential energy of a pair of particles as a function
of the distance in-between them (details in Chapter 4).

Fig. 1.2: Instantaneous snapshots taken from molecular-dynamics simula-

tions of 500 particles, whose pair potential-energy is that of Fig. 1.1.
The panels differ by the density of the particles: 0.01 (right panel), 0.5
(mid panel), and 1 (left panel). The density varies following variations
of the vessel volume. Calculations by C. R. Iacovella and S. C. Glotzer.
Taken from http://web.archive.org/web/20071007200029/http://matdl.org/
matdlwiki/index.php/softmatter:Radial_Distribution_Function, by permis-
sion of Prof. S. C. Glotzer, University of Michigan.

The correlation function. The different panels in Fig. 1.2 correspond (from
right to left) to the phases of gas, liquid, and solid. A quantitative way to distinguish
among these phases is based on the radial correlation function. This function is
computed as follows. Locating the origin at the center of a certain particle in each
instantaneous snapshot, one finds the number of all other particles whose centers
are within a sphere of a radius r around the origin. (The entire sphere is located
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 5

Introduction 5

within the vessel containing the particles.) Then one finds the number of particles
in a slightly larger sphere, whose radius is r + Δr (also contained in the vessel).
The difference between the two numbers, ΔN , is the number of particles in the
spherical shell of volume 4πr2 Δr, and the density in that shell is ΔN/(4πr2 Δr).
This density is the probability density to find two particles separated by
a distance r. This procedure is repeated, applied on spherical shells centered
around other particles, on various instantaneous snapshots, and for different initial
velocities (but with the same spatial density and the same velocity distribution).
One then averages the results over all snapshots, to obtain the radial correlation
function
 ΔN 
g(r) = , (1.1)
4πr2 Δr
where the angular brackets denote the average. In a similar fashion, the correlation
function of a two-dimensional system is derived from the particles’ density in a cir-
cular layer around a center of a certain particle in the plane, g(r) = ΔN/(2πrΔr).
Chapter 3 describes radiation scattering off material samples; it is shown there
that the intensity of the scattered radiation is directly related to the Fourier trans-
form of the correlation function.
Gas. Figure 1.3 displays the correlation functions computed for the snapshots
shown in Fig. 1.2. In all cases the correlation function is rather small at short
distances, as the particles are prohibited from approaching each other too closely
(due to the repulsion among them). The correlation function in the left panel of Fig.
1.3 has a single peak at a distance that approximately corresponds to the minimum
of the potential energy of pairs of particles (see Fig. 1.1). At larger distances,
the probability to find another particle does not vary with r and remains constant.
This is typical for a gas. The peak indicates that momentarily, the probability to
find pairs of particles separated by the optimal distance is higher, owing to the fact
that this yields a lower potential energy. Nonetheless, the correlation function does
not convey any information about the relative locations of all other particles. As
mentioned, one cannot predict the relative locations of all other particles (situated
at distances far larger than rm ) from the location of a certain particle.
Liquid. In the mid panel of Fig. 1.2 the particles are much denser (as compared
to the gaseous phase) and their motion is rather confined in space, though it is
not completely blocked. The correlation function in the mid panel of Fig. 1.3
displays several broad peaks: these imply that around each particle there are several
spherical layers in which the density of atoms exceeds its average value. Such a
correlation function corresponds to a liquid. As opposed to the gaseous state, for
which there are no correlations among the particles’ locations, in the liquid state
there are different probabilities for various separations between pairs of particles.
The leftmost peak in the correlation function of the liquid resembles that of the
gas; the next one represents particles that are neighbors to those featuring in the
first peak, and so on.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 6

6 Introduction to Solid-State Physics

Glass. When the motion of particles as those exhibited in the mid panel of Fig.
1.2 is stopped abruptly (by reducing precipitously their velocities, an operation
equivalent to a sudden reduction of the temperature), the particles “freeze” in their
instantaneous positions, and attain a random pattern, typical to glasses and to
amorphous materials.
Ordered solid. As opposed to the mid panel of Fig. 1.2, most of the particles
in the right panel there are arranged in a periodic pattern: the surroundings
of all particles are identical and the entire structure repeats itself when translated
from one particle to the other. Ordered rows of spheres touching one another
can be discerned in the right panel in Fig. 1.2; generally, each sphere (particle) is
surrounded by six others, located at equal distances away from it. The mathematical
definitions of periodic lattices (also called crystals) are discussed in Chapter
2. It is shown there that the periodic arrangement displayed in the right panel
of Fig. 1.2 is the densest one possible for adjacent spheres arranged on a plane
(see Fig. 2.3), and the densest spatial ordering is built of such planes (the upper row
in Fig. 2.24). As each neighbor of a certain sphere in the plane is also surrounded by
six others, and so on, the entire planar structure is uniquely defined; the locations
of all spheres can be predicted given the positions of a few of them around a central
one (ignoring the small oscillations of each sphere around its equilibrium position).
The same is true in three dimensions, where each particle is surrounded by 12
neighbor particles. Consequently, the correlation function in the right panel of Fig.
1.3 contains many peaks; the intervals in-between them correspond to the spherical
shells in the ordered structure. The peaks become narrower as the sample volume
is increased and the statistics is improved. This is the description of a crystalline
solid, the main issue of the present textbook. In particular, it is found in Chapter
3 that each periodic structure has its own correlation function, that can be uniquely
identified from the data collected by scattering radiation off the material; such
experiments measure the Fourier transform of this correlation function.

Fig. 1.3: The radial correlation function (normalized such that it approaches 1 at
large distances; the radial distance is 1 for r = rm ), for the instantaneous snapshots
depicted in Fig. 1.2. Taken from the same source as in Fig. 1.2.

Order and disorder. The right panel in Fig. 1.2 displays a well-organized
pattern, in which the particles are located on a periodic crystal. In contrast, there
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 7

Introduction 7

is no definite order of the particles’ locations shown in the other two panels there.
At low temperatures the total energy, E, of the periodically-ordered structure is the
lowest. Any deviation away from this order increases the energy. As the temperature
is raised, the particles move faster, and more arrangements of them become possible.
This, in turn, gives rise to an increase in the entropy, S; the entropy is related
to the number of possible configurations of the system at a given energy, and it
increases as that number grows. At any temperature T the system is in the state
which has the minimal free energy, F = E −T S. As the temperature is raised, the
entropy term becomes more dominant and can reduce the free energy, even though
the energy is raised. Therefore, following the increase in the entropy, the amount
of disorder in the system increases as well. In this way one may understand the
transitions from solid to liquid, and from liquid to gas. At times these transitions
are referred to as order-disorder phase transitions.
Quantum fluctuations (i.e., random changes or oscillations), like those of a har-
monic oscillator even in its quantum ground state, can also lead to disorder. These
fluctuations are quite significant in helium (because of the low mass of the helium
atom), and consequently at atmospheric pressure helium remains in the liquid state
(and even in the viscosity-free superfluid state, a unique phase by itself) down
to very low temperatures. Hydrogen is lighter than helium, but since the bonds
among hydrogen atoms are stronger than those in helium (whose electronic “shells”
are full, forbidding the passage of electrons from one atom to the other), hydrogen
does solidify at about 13◦ K, or −260◦ C. The various bonds among atoms and
the way they affect the specific structure of the material are discussed in Chapter
4. Interestingly enough, in two dimensions the oscillations of the atoms around
their equilibrium locations destroy the long-range periodic structure; the periodic
crystalline structure is then replaced by topological phases (see Sec. 5.5; these
unique phases are currently at the forefront of the research in condensed matter).
Periodic solids vs. glasses. When the motion of all particles is halted,
the resulting pattern is that of a glass; it resembles the one in the mid panel of
Fig. 1.2. This is an extreme example of a non-periodic solid. Window glass is
created upon suddenly cooling liquid silicone oxide, SiO2 . Cooling abruptly this
material causes the particles (atoms or molecules) to freeze in their random posi-
tions, acquiring the same amorphous formation they had in the liquid state. At low
temperatures, the free energy of this state is higher than that of the periodically-
ordered solid. However, in order to reach the ideal periodic pattern corresponding
to the lower temperature, the particles have to re-arrange themselves. Numerical
molecular-dynamics studies indicate that this is not easily accomplished: each par-
ticle is captured at a local minimum of the free energy, and needs to overcome a
potential barrier in order to move away from it. The material therefore remains in
a state of a free energy higher than the “correct” (i.e., thermodynamic) one of that
temperature; this is a meta-stable state. The lifetime of this state, that is, the
duration of time until the true thermodynamic state is reached, is extremely long
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 8

8 Introduction to Solid-State Physics

at low temperatures (for instance, that of window glass is hundreds of years), and
therefore this state appears as a stable one during typical experimentally-accessible
times. Figure 1.4(a) exhibits the amorphous arrangement of silicone and oxygen
ions in a two-dimensional sample of glass. This structure differs qualitatively from
the periodic structure displayed in Fig. 1.4(b). Periodic crystals of SiO2 are ob-
tained by melting the glass, and then cooling it slowly enough to allow the particles
ample time to reach their optimal arrangement. Certain ceramic materials, as well
as alloys of metals (e.g., steel) also attain amorphous patterns.
Quasicrystals. There are alloys (e.g., stainless steel) that arrange themselves
in a periodic pattern of their constituents; others freeze in a non-periodic structure.
But surprisingly enough, when the cooling rate is neither too slow nor too abrupt,
the resulting solid is found in an intermediate phase, named quasicrystalline. The
order of the ions in this phase is not periodic; however, their pattern does possess
unique symmetries – quasicrystal show certain phenomena also found in periodic
crystals (see Secs. 2.8 and 3.11).
Condensed matter. Solids can be crystalline, as in the left panel of Fig.
1.2 and in Fig. 1.4(b), or amorphous, as in the mid panel of Fig. 1.2 and in
Fig. 1.4(a). In addition to the liquid and solid states, materials can assume other
phases, for instance that of liquid crystals, which are built of two-dimensional
parallel layers, with freely-moving molecules on each of them; polymers, i.e., very
long, practically one-dimensional molecules, moving freely in space; membranes,
that are two-dimensional surfaces oscillating along the normal to their plane, and
more. These phases that comprise properties of liquids and solids, are included in
the (relatively novel) scientific discipline called condensed-matter physics. The
present book discusses mainly periodic solids, like the one shown in the left panel
of Fig. 1.2; this is part of condensed-matter physics (see also Sec. 7.2).
The periodic table. Many of the specific examples discussed in this book
refer to crystals based on atoms, i.e., on elements. The disparities and similarities
among the various elements follow from their respective locations in the periodic
table, displayed in Fig. 1.5. The entries in this table contain the atomic number
of the element (below the letter indicating its name) and its atomic weight (above
that letter). The different hues indicate the melting temperature of the crystal
built of each element. As seen, the melting temperatures of the noble gases (in the
right column of the table) are rather low. In contrast, the melting temperature of
diamond, built of carbon atoms, is very high. The melting temperature is higher
when the binding energy of the crystal (the energy released as the material is
transformed from the gaseous state to the liquid one) is larger. Binding energies
are calculated in Chapter 4; the different values obtained for them explain the
disparity in the melting temperatures of various elemental crystals. The periodic
table reappears also in other contexts: Fig. 2.22 of Chapter 2 displays the lattice
structures of the crystalline elements; Fig. 6.24 of Chapter 6 identifies their electrical
properties (i.e., metals, insulators, etc.). Beside atomic crystals, there exist more
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 9

Introduction 9

(a) (b)

Fig. 1.4: Amorphous (a) and periodic (b) arrangements of silicone ions (full circles)
and oxygen ones (empty circles) in two-dimensional SiO2 . Top panels: theoretical
models; lower panels: scanning electron microscope (see Sec. 3.1) measurements
of a single layer absorbed on a graphite substrate (see the next section). Each
silicone atom is connected to an additional oxygen atom, located below the plane
of the picture, so that each Si is coupled to two O’s. [P. Y. Huang, S. Kurasch, A.
Srivastava, V. Skakalova, J. Kotakoski, A. V. Krasheninnkov, R. Hovden, Q. Mao,
J. C. Meyer, J. Smet, and U. Keiser, Direct imaging of a two-dimensional silica
glass on graphene, Nano Lett. 12, 1081 (2012).]

complex crystals composed of several atoms; an example is table salt, comprising

ions of sodium and ions of chlorine, depicted in the right panel of Fig. 2.17. The
structures of such crystals are related to the effective radii of the constituent ions
(and these in turn are determined by the wave functions of the electrons in each
ion); these are shown in Fig. 4.6 of Chapter 4.

1.2 The spatial dimensions

Solid phases. The periodic crystalline phase is quite versatile: certain materials
may appear in various structural formations, that have divers properties. These
are called polymorphs, and are different phases of the same solid material.
For instance, to date there are sixteen(!) disparate periodic structures of ice, some
of them discovered only recently. The beautiful patterns of snow flakes are related
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 10

10 Introduction to Solid-State Physics

Fig. 1.5: The periodic table. The various hues indicate the melting temperatures
of the elements, according to the scale above (darker hues correspond to higher
melting temperatures). Created by Atomic PC, see http:/www.blackcatsystems.
com/software/periodic-table-of-the-elements-software.html.

to several of these phases (see Fig. 4.29 of Chapter 4). Temperature or pressure
variations, (i.e., changes in the energy and in the entropy) lead to transitions among
the various phases, and hence to modifications in the material properties; e.g.,
from an insulator to a conductor and even to a superconductor (whose electrical
resistance vanishes at low enough temperatures), from a non magnetic substance to
a ferromagnet that has a finite magnetic moment (like that of the compass needle),
from a dielectric material to a ferroelectric one, which possesses an electric dipole
moment, and so on.
Carbon. Carbon appears in Nature in more than forty different structures; the
polymorphs of elements (like carbon) are called allotropes. Several of these are
surveyed in this section. In three dimensions, as explained in Chapter 2, there
are mainly two carbon allotropes: (1) the graphite (see the right panel of Fig. 2.9).
This structure comprises parallel planes, each contains a periodic array of hexagons
built of the carbon atoms (see Sec. 2.3). (2) When prepared under high pressures,
carbon has the diamond structure, shown in the left panel of Fig. 2.20. In this
configuration, each carbon atom is surrounded by four others, as displayed in Fig.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 11

Introduction 11

4.14. Graphite is soft, and is a good electrical conductor, while diamond is hard
and is an insulator. Carbon attains more conformations, that can be characterized
by a smaller number of spatial dimensions. Chapter 4 details the covalent bonds
among the carbon atoms, which explain these various forms.
Graphene. In two dimensions, the carbon atoms form the structure shown
in Fig. 1.6 [see also Figs. 2.8(a) and 2.10]. This planar structure is identical to
one of the planes of graphite (shown in the right panel of Fig. 2.9). Graphene was
originally produced by peeling single-atom layer off graphite, using a sticky tape;
its unique properties are currently subjected to intensive research.

Fig. 1.6: Schematic illustration of the carbons in graphene.

One-dimensional allotropes of carbon. Carbon possesses several one-

dimensional periodic structures. Figures 1.7(a) and (b) depict two long molecules
of carbon. The cumulene, Fig. 1.7(a), is built of a periodic arrangement of carbon
atoms, with the same spacing between each two neighboring atoms. In the polyyne
[Fig. 1.7(b)] on the other hand, there are two types of bonds, marked by a single
line and by a triple line. These bonds alternate along the molecule, and therefore
the structure repeats itself after each second carbon. The formal definitions and
the descriptions of these arrangements are detailed in Chapter 2, and the electronic
configurations leading to them are analyzed in Chapter 4.

(a) (b)

Fig. 1.7: One-dimensional chains of carbon. (a) Cumulene; (b) polyyne.

One-dimensional allotropes of carbon can be also prepared by cutting graphene

into narrow ribbons or nano ribbons, of widths that comprise only a small
number of atoms. Figures 1.8(a) and (b) display two options for this cutting. The
structure on the cut (on the left side) in Fig. 1.8(a) is called armchair, and that
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 12

12 Introduction to Solid-State Physics

in Fig. 1.8(b) (again on the left hand-side) is called zigzag. Each carbon on the
cut has a “superfluous” electron, that can bind chemically a hydrogen atom.

(a) (b)

Fig. 1.8: Carbon nano ribbons. [P. S. E. Yeo, K. P. Loh, and C. K. Gan, Strain de-
pendence of the heat transport properties of graphene nano ribbons, Nanotechnology
23, 495702 (2012).]

Carbon nanotubes. When a ribbon of graphene (see Fig. 1.8) is folded, and
the two opposite edges are attached together (instead of coupling there hydrogen
atoms), a nanotube, Fig. 1.9(a), is created. This tube can be quite long. Nanotubes
draw currently vast interest due to their very unique transport properties.
Fullerene. Other big carbon molecules may assume a spherical shape, like
the molecule C60 shown in Fig. 1.9(b). It is called Buckyball, Buckminster-
fullerene, or fullerene, owing to its similarity to the spherical dome designed by
Buckminster-Fuller for the 1967 exhibition in Montreal. There exist also “half”
molecules of this type, connected at the end of a nanotube.
The spatial dimension. It is customary to define the spatial dimension of a
sample as the number of independent directions in space along which it is spanned
over long distances. The ideal periodic solid is infinitely long along each of its
dimensions. According to this definition, graphite and diamond are (spatially)
three dimensional, graphene is two dimensional, while the cumulene, polyyne, nano
ribbons and nanotubes are one-dimensional; the fullerene is just zero dimensional.
The properties of a solid are significantly affected by its spatial dimensions.
Nanotechnology. The present day highly-advanced facilities enable the pro-
duction of zero-, one-, and two-dimensional samples in the lab, and to exploit them
for various applications. At ultra-low temperatures and extremely small length
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 13

Introduction 13

(a) (b)

Fig. 1.9: (a) A carbon nanotube; (b) the Buckminsterfullerene molecule, or Bucky-
ball. The carbon atoms are situated at the junctions, so that each has three carbon
neighbors like in graphene.

scales, the electronic properties of such systems are dominated by quantum me-
chanics. Those samples are therefore termed quantum dots, quantum wires, and
two-dimensional electron gas, 2DEG, respectively. The branch of condensed-
matter physics centered on materials of low dimensions, in particular on small sam-
ples (of length scale in-between the macroscopic and the microscopic ones), is termed
“mesoscopic physics”. This field encompasses both basic research, focusing on
the physical properties (which are determined by quantum mechanics), and ap-
plied research, aiming to develop new technologies. As the size of these systems is
typically about nanometers (1 nanometer= 10−9 meter) this particular branch of
condensed-matter physics is called nanotechnology.

1.3 The present book

As mentioned, this book is centered primarily on the properties of periodic crys-

tals. The (almost) perfectly-periodic crystal is the structure that most materials
assume at very low temperatures. The discussion of its properties introduces the
basic concepts, required to grasp and appreciate more advanced topics. Here is a
terse survey of the following Chapters. Chapter 2 finds that only a finite num-
ber of crystalline formations is possible. Their comprehension allows for the
analysis of experiments in which X-rays, electrons or neutrons are scattered
off the solid material; in this way various crystalline patterns are uniquely identi-
fied (see Chapter 3). Moreover, such experiments convey valuable information on
the internal structure of each unit of the crystal. The fact that there is just a
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 14

14 Introduction to Solid-State Physics

finite number of crystalline conformations enables the calculation of the binding

energy (the energy released by creating the periodic solid from the gaseous state)
of each structure, given the forces operating among its constituents (Chapter 4).
It is thus possible to predict the crystalline formation of each material by calcu-
lating its lowest energy, the one achieved in the state where the constituents are
located on the junctions of the periodic structure. One has though to allow for
the thermal motion (at finite temperatures) and the quantum fluctuations of
the basic units around this state of minimal energy. These motions are at times
detrimental to the crystalline structure (like in helium or at two dimensions), and
cause eventually (as the temperature is raised) the transition to the liquid state.
They are also the source of the thermal and acoustic properties of solids (Chapter
5). Chapter 6 investigates the electronic properties of solid crystals. These are
determined mainly by the quantum-mechanical properties of electrons moving in a
periodic potential. The quantum-mechanics considerations allow one to derive the
electrical properties of various solids, and to distinguish among metals (conductors),
insulators, and semiconductors. Chapter 7 surveys certain advanced topics.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 15

Chapter 2

The crystalline structure of solids

The focal point of this chapter is the geometry of periodic crystals. The first
two sections present one- and two- dimensional periodic crystals, with evaluations of
their packing densities. Section 2.3 brings the formal definitions of Bravais lattices
and crystals comprising a lattice and a base, and lists several examples, mainly
two-dimensional ones (for instance, graphene). The formal description of periodic
lattices is continued in Sec. 2.4 with the definitions of lattice vectors and unit
cells, e.g., the primitive and the Wigner-Seitz unit cells. The next two sections
survey the ubiquitous three-dimensional periodic lattices: the cubic crystals and the
hexagonal-closed-packed ones, and Sec. 2.7 explains the classification of the lattices
according to their symmetry properties. It is proven that there is a finite number
of possible Bravais lattices, and hence each crystal can be assigned to a certain
Bravais lattice. The remaining sections are devoted to several advanced topics,
which are still actively investigated. Quasicrystals are described in Sec. 2.8, epitaxial
adsorption, thin films, superlattices and multi-layered structures, are reviewed in
Sec. 2.9; Sec. 2.10 surveys tersely magnetic crystals. Appendix A lists the five
platonic solids; these bodies are referred to throughout the main text, serving as
demonstrations of various symmetries of solids.

2.1 Lattices and crystals in one dimension

Periodic solids. The main part of this chapter is focused on the symmetry
properties of perfectly-periodic solids. Metallic alloys that occasionally appear
in non-periodic quasicrystalline structures are briefly discussed in Sec. 2.8. The
discussion ignores the vibrations of the atoms in the crystal (those exist even
at zero temperature due to the quantum uncertainty principle) and is centered on
the geometric pattern formed by the average locations of the atoms (or ions) in the
material. Lattice vibrations are studied in Chapter 5.
Ideally, the atoms (or the molecules) are arranged in a periodic structure,
termed a “crystal”. The basic unit of the crystal is the “base”. The base may
consist of a single atom (or an ion) or of a group of atoms (ions). The entire

15
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 16

16 Introduction to Solid-State Physics

material is a periodic array of such basic units. A basic unit volume, called the
“unit cell”, is assigned to each base. These unit cells are adjoint to each other and
fill periodically the entire space. Representing each base by a single point (e.g., the
center of mass of its constituents) produces an ensemble of an infinite number of
points, termed a “lattice”. Examples of lattices and crystals are presented below,
followed by formal definitions of the concepts introduced above. The last three
sections survey several advanced topics, based on these concepts.
One-dimensional crystals. The simplest example is that of one-
dimensional lattices. As mentioned, there exists nowadays a vast experimental
endeavor to study one-dimensional systems, in the context of mesoscopic physics
and nanotechnology. The upper panel in Fig. 2.1 displays a periodic array of
identical points which form the most elementary one-dimensional periodic lattice.
Such a system can be derived for instance from Fig. 1.7 displaying the cumulene,
when each carbon atom there is replaced by a single point located at its center.
Formally speaking, a lattice is an infinite collection of points and the periodic
lattice is a special case of such an ensemble. In the case of Fig. 2.1, the “surround-
ings” of each site in the lattice is identical to that of any other site. Alternatively,
when one “resides” on any of the lattice sites and “looks” around, one observes
exactly the same view. A lattice in which all points are surrounded by
identical environments is a Bravais lattice.
Lattice vectors and lattice constants. The second row in the upper panel of
Fig. 2.1 repeats the very same lattice, but with certain additions. The arrow there
represents a vector a, that connects two neighboring lattice sites. Such a vector is
a lattice vector. The length of the lattice vector, denoted a, is termed the lattice
constant. It is identical to the distance in-between two nearest-neighbor sites. An
important property of the one-dimensional Bravais lattice is that each point on it
can be reached from any other by an integral number of lattice vectors.
In other words, the nth site on the lattice is located at Rn = na (relative to an origin
assigned to an arbitrary lattice site), or alternatively, the original lattice cannot be
distinguished from the one obtained upon displacing all lattice sites by the lattice
vector a (assuming an infinitely long lattice). This property identifies the lattice as
a periodic one. It can also serve as a definition of a Bravais lattice.
The unit cell. In that second row of upper panel of Fig. 2.1 there appear
also two rectangles, of length a and of an arbitrary width. (In the one-dimensional
case discussed here, each “rectangle” is in fact a straight segment of zero width;
the finite widths are guides to the eye.) One may assign a lattice point to each
“rectangle” (e.g., the point located at the center of the right rectangle, or the point
located at the left end of the left rectangle). When each of the lattice sites is
covered by one and only one of these rectangles, then the collection of rectangles
covers the entire axis. Each of these rectangles (or segments) may be declared as
the primitive unit cell of the lattice. The primitive unit cell is the shortest
segment (or the smallest area in two dimensions, or the smallest volume
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 17

The crystalline structure of solids 17

Fig. 2.1: Examples of one-dimensional crystals. The arrows represent the lattice
vector that transforms the crystal to itself. The points reached by the arrows form
the Bravais lattice. The rectangles represent possible unit cells.

in three dimensions) that contains the base of the crystal (a single point
in the example above). Displacing the primitive unit cell by all integral
multiples of the lattice vector covers the entire axis (or plane, or space).
The primitive unit cell is not uniquely defined: any segment (“rectangle”)
of length a that covers a single lattice site can serve as the primitive unit cell.
Lattices with a base. All the structures displayed in Figs. 2.1 and 2.2, except
the one in the first panel of 2.1, are not Bravais lattices, since the surroundings
of the sites in them are not identical. For example, the lattice featuring in the
second panel of Fig. 2.1 contains “small” points (A), placed in-between “large” ones
(B). These points may represent, e.g., two types of atoms of different radii. The
surrounding of the A atoms differs from that of the B ones (the nearest neighbors
of the A’s are B’s, and vice versa). On the other hand, the surroundings of the
small atoms are all identical, and so are also those of the large atoms. Hence, this
crystal can be described by a basic unit containing a pair of points, like those within
the rectangle. In that case the system is a crystal, the pair of atoms is its base,
and the rectangle containing them serves as the primitive unit cell of the crystal.
(That is, it is the smallest possible segment which includes solely a single base, and
upon being displaced by the lattice vector –shown by the arrow–covers the entire
axis.) In summary: the base is the cluster of sites within each unit cell, and the cell
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 18

18 Introduction to Solid-State Physics

Fig. 2.2: One-dimensional lattices with larger unit cells, and a lattice built of circles
(or spheres) arranged densely on a straight line.

itself (a segment in the one-dimensional crystal) is the volume including the base.
The crystal consists of periodically-repeated bases. The length of the primitive unit
cell in this example is twice that of the previous one shown in the upper panel of
Fig. 2.1; so is also the length of the lattice vector (the arrow there).
When the two points (atoms) become identical, the crystal illustrated in the
second panel of Fig. 2.1 is the same as the first panel there. The unit cell containing
two atoms, whose length is 2a, may still faithfully describe the lattice. However,
that cell is no longer primitive, as there exists a smaller cell (like the one in the
upper panel of Fig. 2.1) that contains a single atom, and that covers the entire axis
when displaced by multiples of the lattice vector. It is occasionally convenient to
use unit cells that are not primitive; the physical properties are independent of the
choice of the unit cell (as long as its displacements by multiples of the lattice vector
covers the entire axis).
Similarly, the lowest panel in Fig. 2.1 displays a crystal whose base includes
three different atoms, with a unit cell and lattice vector of length 3a. In the first
three panels of Fig. 2.2, the lattice sites are shifted as compared to their locations
in the “simple” case in the first panel of Fig. 2.1. For instance, the lattice exhibited
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 19

The crystalline structure of solids 19

in the first panel in Fig. 2.2 represents the one-dimensional crystal of carbon, the
polyyne [Fig. 1.7(b)], for which the distances among neighboring carbon atoms
assume two distinct values. Again, the surrounding viewed from each lattice site is
not identical to that seen from all others, resulting in a crystal that contains bases
of two atoms (first two panels) or four (third panel in Fig. 2.2).

2.2 Two-dimensional lattices

Dense packing. Section 2.1 describes a lattice as an ensemble of points. Alterna-

tively, one may consider the lattice in the first panel of Fig. 2.1 as a collection of
identical segments (“rectangles”, as those shown there) contiguous to one another.
Replacing each lattice site by a rectangle results in the densest possible pack-
ing for these rectangles. A similar conclusion is reached when each lattice site is
enclosed in a sphere: the densest packing of spheres whose centers are confined to
lie on a straight line is obtained for a periodic array, such that each sphere touches
its two neighbors, see the last row of Fig. 2.2. In this example, the lattice built of
the spheres’ centers is one-dimensional, but the base assigned to each lattice point
is a three-dimensional sphere. There is only a single dense-packed array in one
dimension.

(a) (b)

Fig. 2.3: The densest possible packing of identical circles (a) and of identical spheres
(b) lying on the plane.

In higher dimensions there are more ways to densely-pack spheres. Figure 2.3(a)
displays a dense arrangement of identical circles (e.g., coins) on the plane. Encom-
pass the dark circle at the center by identical circles, each being tangential to the
dark circle and to the circle introduced before; the dark circle is then surrounded
by exactly six circles (the dashed-grey ones in the figure). Adding more circles such
that each is tangential to as many circles as possible results in the array displayed
in Fig. 2.3(a). This structure is repeated until filling the entire plane. The pack-
ing ratio of this structure is defined as the ratio of the total area of the
circles to the total area of the plane. The structure in Fig. 2.3(a) forms the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 20

20 Introduction to Solid-State Physics

densest packing that can be attained by identical circles on the plane, i.e., it is
the array with the highest packing ratio. The proof is outlined below: one shows
that there is only a finite number of periodic lattices in two dimensions and then
calculates the packing ratios of them all, to find that the pattern in Fig. 2.3(a) is
the densest. The number of circles that are nearest-neighbors to any given
circle, usually denoted by z, is termed the “coordination number”; it is 6
for the structures of Fig. 2.3.
Another example for dense packing is realized in the following experiment: ping-
pong balls are thrown on a swimming pool. As long as the number of balls is small,
they land randomly on the water surface. However, as their number increases, the
balls come closer to each other. Eventually they arrange themselves in the pattern
displayed in Fig. 2.3(b) (observed from above). Such a layer of ping-pong balls
is exploited in the northern countries in out-of-doors swimming pools to prevent
the water there from freezing in winter. When a person jumps into such a covered
pool (the northern people do so, after taking a hot sauna followed by a splash in
the snow), the balls are scattered at all directions; but once the swimmer has gone
somewhere else the balls re-arrange themselves in the dense-packing pattern. This
indicates that this is their optimal arrangement. Though each unit in this “crystal”
is a three-dimensional sphere, their centers all lie on a plane, and therefore the two
lattices displayed in Fig. 2.3 are considered as two-dimensional.

Fig. 2.4: Close-packed identical circles on the plane. The points at the centers of
the circles form the triangular lattice. The arrows represent the two lattice vectors,
a1 and a2 . Displacing all points by each of the lattice vectors transforms the entire
lattice back to itself. The length of each lattice vector equals the lattice constant:
it is the distance between neighboring lattice sites, a. In the example of Fig. 2.3(a),
this distance is the diameter of each circle. The rhombus spanned by the two lattice
vectors represents a possible choice of a primitive unit cell.

The triangular lattice. Leaving the center of each circle (or sphere) in Fig.
2.3 as is, and decreasing the radii to an infinitesimal value, results in the points
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 21

The crystalline structure of solids 21

depicted in Fig. 2.4. This collection of points is an example of a periodic

lattice in the plane. The lattice portrayed in Fig. 2.4 is termed the “triangu-
lar lattice”: connecting together three nearest-neighbor sites gives an equilateral
triangle (indicated by the dashed lines); the entire lattice is built from equilateral
rhombi (each comprising two equilateral triangles) contiguous to each other and
repeating themselves periodically. Such a rhombus (see Fig. 2.4) can serve as a
primitive unit cell. At times this lattice is called the “hexagonal lattice” for
reasons clarified below.
Crystal growth. A physical way to build the triangular lattice is based on
a collection of particles, with a central force operating between each pair of them.
The force is repulsive when the two particles are close together, and is attractive
when they are farther away from one another. Such a force results from the poten-
tial energy plotted in Fig. 1.1. This potential energy characterizes various types of
crystal bonds (see Chapter 4). The total potential energy of the particles forming
the crystal is often dominated by the energies related to the forces between nearest
neighbors; the forces exerted by farther-away particles are less effective. At equilib-
rium, the distance between the centers of neighboring particles is roughly the one
that corresponds to the minimum of this two-particle potential energy. Intuitively,
it is plausible that when many such particles are distributed on the plane, the max-
imal gain in the potential energy of the whole system is reached when each particle
has as many nearest neighbors as possible (i.e., the largest feasible coordination
number), each residing at a distance corresponding to the minimal value of the
pair-wise potential energy. This gain is realized for the structure shown in Fig. 2.4,
where the center of each particle is located at a vertex of the triangular lattice and
has six nearest neighbors. It is impossible to construct a two-dimensional periodic
lattice made of identical particles with a coordination number larger than 6.
The process outlined in Fig. 2.3(a), that begins with a single particle to which
more and more particles are added, imitates a method for crystal-growing based
on a gradual adsorption of identical particles on a planar substrate. Once
the first particle (the dark circle) is attached to the substrate, a second one is added.
The two particles move on the plane, attracted to one another by the force operating
in-between them, until they form a little cluster in which they are separated by
the distance corresponding to their minimal potential energy. A third particle
arriving on the plane moves towards this cluster, to form with its two predecessors
an equilateral triangular cluster (the dashed triangle in Fig. 2.4), where each pair
of particles is separated by the distance at which the potential energy is minimal.
More particles join in a similar manner, thus building the triangular lattice. The
triangular lattice is formed in the plane by identical particles exerting on each other
short-range forces such that each one adheres to its neighbors. Crystals comprising
more types of atoms (for instance, table salt, built of positive sodium and negative
chloride ions) for which long-range forces are also effective (i.e., the Coulomb force
among the ions) grow in a different pattern.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 22

22 Introduction to Solid-State Physics

In reality, the situation is more complicated (see below), as clusters of various

perimeters can grow around different initial locations, and do not necessarily join
together smoothly. This may result in an aperiodic metastable structure; an ex-
ample is illustrated by the atoms B and C in Fig. 2.23. A metastable state
occurs when the system is at a local minimum of its energy, but not
in the global minimum. To reach the latter, the system must overcome po-
tential barriers, a process not easily accomplished at low temperatures. Warming
up the substrate on which the crystal is growing helps to overcome these barriers,
and allows small clusters to merge into larger ones. Ideally, the particles reach the
structure of a single triangular lattice.
There are other methods to grow crystals. Crystals grown under controlled
conditions are needed for certain precise measurements or for building up artificial
compounds which do not exist in Nature. Other ways to grow crystals are not
detailed here (but see Sec. 2.9).
The unit cell. Figure 2.5 illustrates the concept of a unit cell, mentioned in
the context of one-dimensional crystals. Filling the area around each circle in Fig.
2.3(a) up to the tangents common to that circle and all of its nearest neighbors,
results in a periodic collection of hexagons that tiles the entire plane. Each hexagon
represents a unit cell “belonging” to the circle inside it (or to the lattice site at
the center of that circle). Each of the hexagons in Fig. 2.5 contains all the
points in the plane that are closer to the lattice site at its center than
to any other center. This unit cell is termed the Wigner-Seitz unit cell. Its
hexagonal shape explains why the triangular lattice is occasionally referred to as
the “hexagonal lattice”. The Wigner-Seitz cell is a primitive unit cell; it contains a
single lattice site (which is the base of this crystal); its displacements by the lattice
vectors (displayed in Fig. 2.5) tile the entire plane. The Wigner-Seitz unit cell is
not the only possible primitive cell: for instance, the plane can be covered entirely
by identical rhombi as the one displayed in Fig. 2.4. Since both the rhombi and
the hexagons tile the whole plane, and each contains a single lattice point, their
areas must be equal (see problem 2.5). The rhombus can be displaced so that the
lattice site lies at its center (rather than at its left lower corner); this gives rise to
yet another optional primitive unit cell.

Problem 2.1.
a. What is the packing ratio of the planar structure in Fig. 2.3(a)?
b. Each circle there is replaced by a sphere, and the spheres are densely arranged
on a plane as in Fig. 2.3(b). Another plane, tangential to the spheres, is added
from above. What is the resulting Wigner-Seitz cell? What is the ratio of the total
volume of the spheres to that enclosed between the two planes? (this is the ratio of
a sphere’s volume to the volume of the unit cell). This ratio is termed the volume
packing-ratio of the system.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 23

The crystalline structure of solids 23

Fig. 2.5: Hexagonal primitive unit cells (Wigner-Seitz cells) representing the trian-
gular lattice of Fig. 2.4. The lattice points of Fig. 2.4, that are also the centers of
the circles in Fig. 2.3(a), are located at the centers of the hexagons. The arrows
are the lattice vectors (as in Fig. 2.4).

Problem 2.2.
Figure 2.6(a) portrays an array of mutually-tangential circles, each touching solely
four others at the end points of two perpendicular diameters. The lattice formed by
the circles’ centers (marked by dots) is called the “square lattice”.
a. Enclose each circle in a Wigner-Seitz cell, so that the cells completely cover the
plane. What is the shape of this cell?
b. What is the packing ratio of this structure?
c. What is the volume packing-ratio for spheres arranged on a square lattice [Fig.
2.6(b), when the spheres are caged between planes from below and from above]?

(a) (b)
Fig. 2.6: The square lattice, built of adjacent circles (a) or contiguous spheres (b).
The arrows represent the lattice vectors.

2.3 Planar Bravais lattices with a base

Bravais lattices. The triangular and the square lattices are two members of the
family of Bravais lattices. As mentioned, a Bravais lattice is an array of
sites, each of them surrounded by identical environments. Bravais lattices
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 24

24 Introduction to Solid-State Physics

are classified according to their symmetry properties; it turns out that there is a
finite number of them. There are solely four planar Bravais lattices, and seven
three-dimensional ones; those are illustrated in Fig. 2.27 and Fig. 2.28.
Crystals with a base. The vertices of the triangular lattice (Fig. 2.4) and of
the square lattice (Fig. 2.6) form Bravais lattices. Nonetheless, not any periodic
array of sites is a Bravais lattice. Figure 2.7 exhibits two periodic arrays
featuring large and small circles. These can represent, for instance, the (positive
and negative) ions of sodium and chloride in a planar cross section of the table-salt
crystal, in which each positive sodium ion is surrounded by four negative chloride
ones, and vice versa. (Since the chloride ion is negatively charged and the sodium
one is positively charged, and as positive and negative charges attract one another,
this arrangement has the lowest possible energy, see Chapter 4.) Thus, the crystal
looks differently when “viewed” from the center of the large circle or from that
of the small one. If the center of each circle is considered as a site on a square
lattice, the emerging lattice is not a Bravais lattice according the definition given
above. It is still possible, though, to describe the structures in Fig. 2.7 in terms of
arrays of sites from which one does “observe” the same surrounding. For example,
the environments of each of the centers of the large circles are identical. One may
therefore join together these centers, as shown by the dashed lines, and obtain a
Bravais lattice. This is a square
√ lattice, rotated by 45◦ relative to the original one
and its lattice constant is 2 larger. Alternatively, the surroundings viewed from
each of the vertices of the square bounded by solid lines are also identical to each
other, and therefore these vertices form a Bravais lattice as well, similar to the
previous one but displaced relative to it.
Any of the squares in Fig. 2.7 can serve as the primitive unit cell of the Bravais
lattice corresponding to this crystal, again indicating the redundancy in the choice of
the unit cell. Repeated Translations of each square by the lattice vectors (marked
in the figure) tile the whole plane by such cells. The square that connects the
centers of the small circles is also a proper primitive unit cell. Each square in this
Bravais lattice contains two circles, a large one and a small one. This is obvious
regarding the square bounded by solid lines which contains the centers of two circles,
large and small. The one marked by dashed lines contains a small circle at its
center, and four quarters of the large circle, and hence it also packs the areas of the
large and small circles. In contrast with the planar structures discussed in Sec. 2.2,
in which the unit cell represents a single lattice point, the unit cell of the lattices
here contains two lattice points (the centers of the small and the large circles). As
mentioned, a unit comprising the constituents contained in the primitive unit cell
(e.g., a small circle and a large one) is termed the base of the periodic array. The
base consists of the constituents that repeat themselves, while the primitive unit
cell is the area that is repeated ad infinitum and covers the whole plane. There is
one-to-one correspondence between a certain choice of the primitive unit cell and
the base contained in it.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 25

The crystalline structure of solids 25

           

           

           

           

(a) (b)

Fig. 2.7: Examples of a planar crystal built of two types of circles. The crystal may
represent a planar cross section of table salt, NaCl. The two panels differ by the
ratio of the circles’ radii, and consequently also by their packing ratios, but they
do share certain symmetry properties. The big circles in (a) are not tangential to
each other while in (b) they are; therefore the packing of (b) is denser. The squares
represent possible unit cells of the corresponding Bravais lattice: the surroundings
of each vertex of the squares are identical. The unit cell contains a single base
comprising a big circle and a small one. The dashed arrows are lattice vectors and
the solid vector is r2 , connecting two circles in the base that belongs to the dashed
unit cell.

Any periodic material comprises a Bravais lattice of points; each of

them represents a base of the material. This collection of bases is called
a “crystal”. The crystal is built of a lattice with a base.
A practical way to consider the lattice displayed in Fig. 2.7 is to note that the
large circles alone (deleting the small ones), form a square lattice with a unit cell
given by the dashed square; accounting only for the small circles (and obliterating
the large ones) again yields a square lattice, identical to the previous one. In the
crystal portrayed in Fig. 2.7, the lattice of the small circles is shifted relative to
that of the large ones by the vector r2 , the solid arrow in the figure. Hence, the
crystal consists of two square sublattices, displaced relative to one another. A
similar description pertains to any crystal with a base: the crystal always consists
of several identical sublattices, each corresponding to one of the elements contained
in its base. Each of these sublattices is a Bravais lattice, identical to the Bravais
lattice of the entire crystal.

Problem 2.3.
The small circles in Fig. 2.7(a) are tangential to the large ones, but the latter are
not tangential to each other.
a. Find the ratio of the total area covered by the two types of circles and the area
of the whole plane, i.e., the planar packing-ratio ρ of this structure, denoting the
radii of the small circles and those of the large ones by r< and r> , respectively.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 26

26 Introduction to Solid-State Physics

b. What is the ratio of the two radii, x = r< /r> , needed to obtain the densest
packing [when the large circles are also touching each other, as in Fig. 2.7(b)]?
c. What is the packing ratio upon further decreasing the radii of the small circles?
Plot the packing ratio as a function of the radii ratio, for 0 < x = (r< /r> ) < 1.
d. Express the lattice constant, for each of those cases, in terms of the two radii.

The hexagonal lattice–graphene. Interestingly enough, even lattices com-

prising identical atoms are not necessarily Bravais lattices. Examples are
portrayed in Fig. 2.2: the atoms are identical, but their surroundings are not. A
two-dimensional realization is shown in Fig. 2.8(a), which exhibits the hexagonal
lattice, also known as “honeycomb” for obvious reasons. As mentioned in Chap-
ter 1, this lattice is currently intensively studied, since it describes the structure of
graphene for which each point on the lattice represents a carbon atom (Fig. 1.6).

(a) (b)

Fig. 2.8: The hexagonal lattice. (a) The lattice sites at the corners of adjacent
hexagons. The figure displays two types of lattice sites, A and B, whose surround-
ings are different. The rhombus represents an optional unit cell, containing a base
of two sites connected by the vector r2 . (b) The A−type points are drawn as larger
dots and the thin solid lines demonstrate that these points build a triangular lattice,
with the same unit cell as in the other panel. The smaller dots are the B−type
lattice sites. These too are located on a triangular lattice shifted relative to the
former one.

Graphite. As mentioned in Chapter 1, carbon appears in numerous crystal

structures (or polymorphs) due to the multitude of possibilities for chemical bonds
among its atoms. The most famous is the diamond; under common circumstances
the stablest three-dimensional structure of carbon is graphite (diamond is formed
in Nature only under strict pressure and temperature conditions, and hence is rather
rare). The graphite is made of parallel planar layers inside each the atoms are
ordered in an hexagonal array as in Fig. 2.8(a), see also the right panel in Fig. 2.9.
This three-dimensional structure is also called the hexagonal lattice. Graphite
can be found in pencils; the layers are only loosely coupled, and the process of
writing “peels” two-dimensional layers off the black material onto the paper. In
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 27

The crystalline structure of solids 27

2004 Geim and Novoselov succeeded to peel two-dimensional layers of graphite and
thus opened the frontal research field of single layers, i.e., of graphene.
Figure 2.10 displays a graphene specimen, hanging freely in space. Though
topologically a plane, the graphene may attain a wavy form or fold itself in various
shapes. As mentioned, the two-dimensional graphene possesses certain spectacular
physical properties, e.g., high rigidity of structure and high electrical conductivity.
These unique features, and the rare possibility to easily produce usable samples,
have invoked the extensive interest in this material.

Fig. 2.9: Right panel: the hexagonal lattice of graphite. Each layer in the lattice is
an hexagonal lattice, identical to that displayed in Fig. 2.8. The atoms in the lower
and upper planes lie precisely one on top of the other, while the hexagonal lattice
forming the layer in-between is displaced with respect to those layers. The thick
lines show the unit cell. The central panel displays the arrangement of the atoms
in the middle layer [note that Fig. 2.8(b) shows the atoms in the lower layer]. The
A−type atoms occupy the same sites in the two planes, but the B−type ones are
displaced so that they are located above the centers of the hexagons in the layer
below (or below the centers in the layer above). Left panel: the lattices of the two
planes are shown together (viewed from above). The thick lines pertain to the lower
layer and the thin ones refer to the middle plane.

Graphene as a crystal with a base. The planar hexagonal lattice, Fig.

2.8(a), can be characterized by the two points A and B. The surrounding observed
from A differs from that seen from B. For instance, “glancing” to the right of A
reveals a B−point at a higher location (on the perpendicular axis), and a vacancy
at the mid point of the hexagon below. As opposed, on the right of B there is a
lattice point at a lower location (on the perpendicular axis) and a vacancy at the
center of the hexagon above it. In other words, A resides at the lower edge of the
hexagon, and hence the empty site at the center of this hexagon is above it, while B
is found at the upper edge of the hexagon, and the empty site of its center is below
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 28

28 Introduction to Solid-State Physics

it. The environment viewed from A is the same as seen from all the lower corners of
the hexagons, while all the upper corners share the same surroundings as point B.
Since these two environments are not identical, the hexagonal lattice is definitely
not a Bravais lattice, but it can be described as a crystal, i.e., as a Bravais lattice
with a base that contains two sites. Figure 2.8(a) displays a rhombic unit cell
which consists of A points at its corners and a B point inside it. Thus, A and B
form together the base of this crystal. Denoting the lattice vectors by a1 and a2 as
in the figure, the vector joining the two base sites is r2 = (a1 + a2 )/3. (The length
of this vector is 2/3 of the height of the equilateral triangle spanned by the two
lattice vectors, since the second point is located at the middle of this triangle–the
locus of its heights. The length of the longer diagonal of the rhombus, a1 + a2 ,
equals two such heights.) To emphasize the distinction between the two types of
points, Fig. 2.8(b) shows them as large circles (for A) and small ones (for B). The
A points create a triangular lattice, with the same unit cell and the same lattice
vectors introduced in Fig. 2.8(a). Likewise, all B points form an identical triangular
lattice, displaced relative to the former one by the vector r2 . The hexagonal lattice
can therefore be considered as a combination of two triangular sublattices, one
comprising the A sites, the other the B ones. Alternatively, the planar hexagonal
lattice is a triangular Bravais lattice with a base containing two identical atoms (to
avoid confusion: the lattice is triangular but it can be described by a rhombic unit,
as in Fig. 2.4).

Fig. 2.10: A graphene sample, hanging freely in space. Courtesy of A. Geim.

The lattice structure of graphite. The three-dimensional lattice of graphite

is illustrated in the right panel of Fig. 2.9. The atoms in the lower and in the
higher layers are exactly above one another; in the mid layer atoms A are above the
atoms A in the layer below, but atoms B are displaced compared to their locations
in the lower plane, and are each above a vacancy at the center of an hexagon in that
plane. In this configuration, the B atoms are located at the same distance from the
six atoms forming the hexagon in the lower layer and from the similar six atoms
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 29

The crystalline structure of solids 29

in the plane above, instead of being close to only a single atom in each such plane
(had the two hexagonal lattices been placed precisely one above the other). Though
the distance from each of the 12 neighbors is longer than in the other option, their
large number compensates for this difference and yields a lower potential energy.
The central panel of Fig. 2.9 exhibits the atomic array in the mid layer. The large
circles, representing the A atoms, are still located as in Fig. 2.8(b) (which describes
the lower layer of the graphite), but the B atoms are displaced by the vector r2
relative to their locations there. Glancing at the rhombus that represents a unit
cell in Fig. 2.8(b) reveals that the rhombus above it [in the mid layer, shown by
the thick lines in the central panel of Fig. 2.9 and by the dashed lines in the right
panel there] also contains two atoms, A and B, but the latter, instead of being at
the mid point of the left triangle of the rhombus as before, is now at the mid point
of the triangle on the right side.
Other materials described by the hexagonal lattice. Two comments are
called for. First, the lattice displayed in Fig. 2.8(b) describes as well crystals
comprising two different atoms, e.g., boron and nitrogen (BN), situated at the
locations of the large and small circles in the figure. This material is also at the
forefront of contemporary research. A somewhat simpler example are the crystals
of silicene and germanene, in which silicon or germanium atoms replace the carbon
ones forming the graphene. Much like the graphene, these materials also possess
intriguing physical properties, nowadays under active investigations. In those, the
two atoms of the unit cell are located in two different parallel layers (Fig. 2.11).
The dissimilarities between these crystals and graphene is discussed in Chapter 4.

(a) (b)

Fig. 2.11: Silicene lattice; (a) viewed from above, (b) a side view, both showing
the different heights of the two sublattices [A. O’Hare, F. V. Kusmartsev, and
K. I. Kugel, A Stable “Flat” Form of Two-Dimensional Crystals: Could Graphene,
Silicene, Germanene Be Minigap Semiconductors?, Nano Lett. 12, 1045 (2012).]

Second, both graphene and BN are examples of the more general case, shown in
Fig. 2.12, of a triangular lattice with three types of sites, marked A, B, and C.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 30

30 Introduction to Solid-State Physics

The square lattice featuring in Fig. 2.7 consists of two identical sublattices
displaced relative to one another; the triangular lattice can be described by three
sublattices. The unit cell of the lattice of Fig. 2.12 is still identical to those pre-
sented in Fig. 2.8 and the mid panel of Fig. 2.9, but its base contains three types
of sites. The lattices portrayed in Fig. 2.8(b) and 2.9 can be derived from the one
in Fig. 2.12 provided that all C or all B points are deleted, respectively. Hence,
the two different layers in the graphite lattice are only particular cases of Fig. 2.12.
Discarding all A points results in a hexagonal lattice identical to the formers, but
shifted relative to them. It is possible to build three-dimensional hexagonal lattices
by arranging differently these three planar structures, one on the top of the other.

Fig. 2.12: The triangular lattice with three types of sites.

Problem 2.4.
A circle of radius r> is drawn around each “large” point in Fig. 2.8(b), and a circle
of radius r< is drawn around each “small” one. Derive the packing ratio of the
structure as a function of x = r< /r> , in two cases: (i) when the large circles touch
one another; (ii) when the large circles touch the small ones. For which x is the
packing ratio maximal? Compare it with the one of the square lattice. Obtain the
lattice constant as a function of the two radii. What is the result for graphene?

2.4 The lattice vectors and the unit cell

Translations in the lattice. The periodicity of the triangular lattice can be

expressed in terms of the lattice vectors, marked by arrows in Figs. 2.4 and 2.5. Each
vector connects two neighboring lattice sites. In the infinite lattice, any translation
by one of these vectors (or by an integral multiple of it) transforms the lattice into
another one, completely identical to the original lattice. In formal terms, a three-
dimensional Bravais lattice is the ensemble of points in which each pair
can be connected by the vectors
R = n1 a1 + n2 a2 + n3 a3 , (2.1)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 31

The crystalline structure of solids 31

where n1 , n2 , and n3 are arbitrary integers (positive, negative, or zero),

and a1 , a2 , and a3 are three independent displacement vectors (i.e., they
are not parallel to each other and are not in the same plane), which
are called lattice vectors. Alternatively, any of the lattice sites may be chosen
as the origin, and then any other lattice point is given by Eq. (2.1), with the
proper integer coefficients. The Bravais lattice is defined in Secs. 2.1 and 2.3 as
the set of points with identical environments. This indeed pertains for all the
points connected by the vectors given in Eq. (2.1). Hence, the two definitions
are equivalent. Lattice vectors of planar two-dimensional lattices are represented
by the pairs of vectors emerging from the same site in Fig. 2.4 or 2.5. A three-
dimensional example is portrayed in Fig. 2.13: the three lattice vectors a1 , a2 , and
a3 form a parallelogram. When each of the vertices of the parallelogram is a lattice
site, and when the entire lattice is built of such identical parallelograms adjacent to
each other (along the common faces), then each lattice point can be reached from
the origin by a displacement vector of the form Eq. (2.1). The lattice so created is
a Bravais one, as the surroundings of all the points are identical.

Fig. 2.13: A unit cell built of the three lattice vectors a1 , a2 , and a3 , with a lattice
point located at each vertex. The angle α is between a2 and a3 , β is the angle
between a3 and a1 , and γ is the one between a1 and a2 .

The unit cell and the primitive unit cell. Since the lattice consists of the
collection of parallelograms, the one in Fig. 2.13 can be declared as the unit cell.
Each parallelogram can be identified with a single lattice site, situated, e.g., at the
origin (from which the three vectors emerge). In other words: the density of the
material described by this lattice is the ratio of the mass of the base represented by
the lattice site, and the volume of the primitive unit cell, given by
V = |a1 · [a2 × a3 ]| (2.2)
(check!). The three lattice vectors cannot be coplanar, lest the volume vanishes.
The area of the primitive unit cell in a two-dimensional lattice, described by two
lattice vectors, is
S = |a1 × a2 | (2.3)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 32

32 Introduction to Solid-State Physics

(check!). The choice of the lattice vectors is not unique. Figure 2.14 displays four
examples of pairs of lattice vectors in the plane. Linear combinations of the pair a1 ,
a2 , or of the pair a5 , a6 , or of the pair a7 , a8 , with integer coefficients [according to
Eq. (2.1)] displaces any point on the lattice to another one on it. In contrast, linear
combinations of the pair a3 , a4 , with integer coefficients, cannot reach all lattice
sites. For instance, a linear combination with non-integer coefficients is needed to
reach the point in the center of the parallelogram built of these two vectors. The
parallelograms built on the former pairs have the same area, which is the specific
area corresponding to a single lattice point (check!); the area of the parallelogram
built on the fourth pair is twice that area, and hence this parallelogram contains
two lattice sites (e.g., the one at the lower left corner and the one at the mid point).
There are other pairs of lattice vectors that can span the whole planar lattice; there
are also many possibilities to choose the lattice vectors of a three-dimensional lattice.
It is therefore customary to define primitive lattice vectors: for the vector a1
one picks up the shortest vector connecting two lattice sites. The vector
a2 is then chosen as the shortest one which is not parallel to the former,
and a3 is the shortest vector between nearest neighbors which does not
lie in the plane spanned by its two predecessors. When the vectors a1 and
a2 have equal lengths it is possible to interchange their notations to a2 and a1 ,
as happened in all the examples presented above. In the case of Fig. 2.14, the
primitive lattice vectors are a1 and a2 , which are normal to one another and have
the same length.

Fig. 2.14: Possible pairs of lattice vectors (and unit cells) in the plane.

The choice of the primitive unit cell is not unique. In Fig. 2.5 the unit
cell is chosen as the hexagon encompassing a single circle, i.e., as the Wigner-Seitz
cell. This hexagon includes all the points in the plane which are closer to the center
of the hexagon than to any other lattice site; it is encased by the mid orthogonals of
the segments connecting the central lattice site with its six neighbors. The extension
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 33

The crystalline structure of solids 33

to three dimensions is clearcut: one chooses a lattice site and connects it with other
lattice sites. One then bisects each segment by a perpendicular plane. The three-
dimensional Wigner-Seitz cell is the volume confined by these planes,
that contains the initial lattice site. Figure 2.13 offers a different definition
of the unit cell: the primitive unit cell is a cuboid, or a parallelepiped, built of
the three primitive lattice vectors of the lattice. In Fig. 2.4, this unit cell is
a rhombus, with edges of length equal to the distance between nearest-neighbor
lattice points, like the one shown at the lower left corner of the figure. The whole
plane can be described as a collection of such rhombi, or as a collection of hexagons
(those are shown in Fig. 2.5). Since each such cell, be the rhombus or the hexagon,
represents a single lattice site, and their lot tessellates the plane, one expects their
areas to be identical (problem 2.5). Generally, the three lattice vectors can attain
different lengths, as in Fig. 2.13. These are termed “lattice constants” (extending
the definition given for the one-dimensional case, Sec. 2.1). For the triangular
lattice, the lengths are equal and there is a single lattice constant. Whether to use
the primitive unit cell or the Wigner-Seitz one is a matter of convenience. Both
represent faithfully the lattice (and they are not the only possible unit cells). When
non-primitive unit cells are exploited, the definitions of the lattice constants have
to be readjusted.

Problem 2.5.
a. Calculate the area of the Wigner-Seitz unit cell of the triangular lattice (Fig.
2.5) and that of the rhombus in Fig. 2.4, in terms of the lattice constant a. Are
they indeed equal?
b. Determine the corresponding unit cells (the Wigner-Seitz one and the one built
on the lattice vectors) of the square lattice, discussed in problem 2.2. Find their
areas.
c. Given the lattice vectors of the square lattice, whose sites are occupied by two
types of circles, what are the lattice vectors of the crystal in Fig. 2.7? What is the
area of the primitive unit cell spanned by them?

Crystal with a base. When the base of a crystal includes more than one
lattice site, the choice of the origin relative to the locations of the base points is not
unique. Figure 2.7 illustrates two choices of a unit cell, each containing the base
points at different locations; one choice is mapped on the other by a displacement
of the entire lattice. The same is true for the unit cell shown in Fig. 2.8(b) and the
mid panel in Fig. 2.9. Quite generally, locating the lattice site that represents a
certain unit cell at the axes’ origin, the locations of the base points relative to
the origin (i.e., within the unit cell) are denoted by

ri = xi1 a1 + xi2 a2 + xi3 a3 , (2.4)

where i = 1, . . . , nB enumerates the nB base sites, with |xiα | < 1. In the unit
cells marked by dashed lines in Fig. 2.7, the two base points are at r1 = 0 and
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 34

34 Introduction to Solid-State Physics

r2 = (a1 + a2 )/2, where the lattice vectors a1 and a2 are marked by the dashed
arrows there.

Problem 2.6.
Prove that the unit cell spanned by the vectors a3 and a4 in Fig. 2.14 describes a
square lattice with a base comprising two sites. Determine the locations of the base
points, in terms of these vectors. Find the area of the unit cell.

Problem 2.7.
Find the lattice vectors of the unit cell marked by solid lines in Fig. 2.8, in cartesian
coordinates (expressed in terms of the length in-between nearest-neighbors on the
original hexagonal lattice). Give the locations of the base points in terms of these
vectors. What is the answer for the unit cell in the mid panel of Fig. 2.9?

2.5 The cubic lattices

The simple-cubic lattice. Most of the crystals found in Nature are three-
dimensional. The simplest way to construct three-dimensional lattices is to begin
with a two-dimensional one, and to lay above it an identical two-dimensional lattice,
such that each lattice point in the second layer is located precisely above a lattice
point in the lower plane. An example is the square lattice of spheres shown in Fig.
2.6; one places successive layers identical to the initial one, such that the spheres
in each layer are located above those in the layer below. The resulting structure
is displayed in Fig. 2.15(a). The lattice built of the centers of all these spheres
has orthogonal lattice vectors of identical length; they form the edges of a cube.
In cartesian coordinates these lattice vectors are a1 = ax̂, a2 = aŷ, and a3 = aẑ,
where a is the lattice constant (whose length is the cube’s edge and the diameter of
the spheres), and the unit vectors x̂, ŷ, and ẑ point along the cartesian axes. This
cube may be chosen as the primitive unit cell of the lattice [see Fig. 2.15(b)], which
is duly called the simple-cubic lattice, abbreviated as SC.

Problem 2.8.
a. Find the volume of the (primitive) unit cell of a SC lattice built of spheres whose
diameter is a. Find the volume packing-ratio and the coordination number.
b. What is the shape of the Wigner-Seitz cell of this lattice?
c. Propose two sets of lattice vectors that span all lattice points. What is the volume
of the unit cell constructed of these lattice vectors?

Cesium-chloride type crystals. As found in problem 2.2, the packing ratio

of circles arranged on a square lattice, 0.5236, is not the optimal one; the densest
packing of circles on the plane, 0.6046, is reached for circles on a triangular lattice.
Likewise, the packing ratio of the simple-cubic lattice (see Fig. 2.15) is not the
optimal one. (The reader is encouraged to try and identify, for instance by exper-
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 35

The crystalline structure of solids 35

(a) (b)

Fig. 2.15: (a) A sample built of eight such unit cells, each hosting a sphere.
(b) The unit cell of the simple cubic lattice.

imenting with balls, the lattice which possesses the densest packing of spheres in
space; the answer is below.) Around the center of the unit cell of the SC there
is an empty zone, that is responsible for the low packing ratio. This ratio can be
increased by, e.g., adding a small sphere, of radius r< , at the center of each cube,
i.e., around the point r2 = (a/2)(x̂ + ŷ + ẑ) = (a1 + a2 + a3 )/2. This is similar
to its planar analog, depicted in Fig. 2.7. This small sphere can be enlarged until
it touches the
√ eight surrounding large spheres. Since the length of the cube’s di-
agonal is a 3, and the radius of the large sphere, r> , is a/2 (the spheres on each
edge of the cube √ are contiguous), the large spheres touch each other as long as
2(r> + r< ) ≤ a 3. (The diagonal has to host √ two radii of the large spheres, and a
diameter of the small one.) Hence, r< ≤ ( 3 − 1)r> = 0.732r> . As the radii of the
large and small spheres are different, the lattice so formed in not a Bravais lattice,
but a simple-cubic lattice with a base, that contains one large sphere (e.g.,
at the corner of the cube) and one small sphere (e.g., at the center of the cube).
Exploiting the original lattice vectors, the vector connecting the two spheres in the
unit cell is r2 = (a1 + a2 + a3 )/2. This unit cell is shown in the right panel of Fig.
2.16. Disregarding the radii ratio of the spheres there, this unit cell belongs to the
cesium chloride (CsCl) crystal; all other crystals sharing this unit cell are hence
termed cesium-chloride type crystals. The bigger negative Cl ions occupy the
corners of the cell, while the small positive Cs ion resides at its center.
Occasionally it is expedient to describe such a unit cell by its projection on the
lower planar face, as done in the central panel of Fig. 2.16. Except for the ions
residing on this plane (marked by the dark circles), all other ions are represented by
circles which contain their height above the plane, in units of the lattice constant
along the normal to the plane (in the present case, 1/2). The cesium-chloride crystal
is in fact built of two simple-cubic sublattices, shifted relative to one another by r2 .
As mentioned, a crystal with a base may be described by sublattices, each
containing different atoms or ions of the base.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 36

36 Introduction to Solid-State Physics

Fig. 2.16: Right panel: the unit cell of a crystal of the cesium-chloride (CsCl) type.
(The radii of the spheres are not in scale; they are reduced for clarity.) Central
panel: the projection of the unit cell on the lower planar face. The dark circles at
the corners represent the four chloride ions on that face (on the XY plane) and the
empty circle is the cesium ion, located at height a/2 above (the height, 1/2 in units
of the lattice constant, is marked within the circle). Left panel: the unit cell of the
body-centered cubic lattice (BCC). The lattice vectors [Eq. (2.5)] are marked by
the arrows.

The body-centered cubic lattice (BCC). A special case of the crystal dis-
played in the right panel of Fig. 2.16 is realized when the ion (or atom) located at
the center of the cube is identical to those occupying the corners; this is shown in
the left panel of Fig. 2.16. The lattice there is the body-centered cubic lattice,
abbreviated as BCC. The surrounding of the central site in this structure is iden-
tical to those of the sites at the corners, and therefore the BCC lattice is a Bravais
lattice; the cubic unit cell in the figure is not the primitive unit cell of this lattice
(as it contains two lattice sites, e.g., at the center of the cube and in one of its
corners). The lattice vectors of this Bravais lattice are the three vectors connecting
a point in the lattice with three of its nearest-neighbors. A lattice point placed
at the origin is surrounded by eight nearest neighbors, at (a/2)(±x̂ ± ŷ ± ẑ) (in
cartesian coordinates). A possible choice for the primitive lattice vectors is hence
a1 = (a/2)(−x̂ + ŷ + ẑ) , a2 = (a/2)(x̂ − ŷ + ẑ) , a3 = (a/2)(x̂ + ŷ − ẑ) .
(2.5)
These are portrayed by the arrows in the left panel of Fig. 2.16. Though the cubic
unit is not the primitive unit cell of the BCC lattice, it is often convenient to use
it (with a base including two identical atoms) rather than the primitive one built
of the three vectors in Eq. (2.5). This is because the cubic unit cell reflects the
symmetry under rotations by 90◦ , which are hard to observe from the primitive cell.
It is also more convenient for carrying out various calculations due to the efficiency
of the cartesian coordinates.
The preceding discussion implies that had identical spheres been placed at each
point of the BCC lattice, and increased until each one touches its eight neighbors,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 37

The crystalline structure of solids 37

then the spheres at the corners do not touch one another √ (because the radius of
each sphere, r, equals a quarter of the diagonal: r = a 3/4, and thus is smaller
than a/2). The volume of the cube is a3 and it contains two spheres (one at the
center, and 1/8 of a sphere at each corner);
√ the total occupied volume is 2 × 4πr3 /3.
Hence, the volume packing-ratio is π 3/8 ≈ 0.680. This packing is denser than
that of the SC lattice.
Problem 2.9.
What is the packing ratio of the cesium-chloride type crystals as a function of the
radii ratio x = r< /r> ? What is the maximal packing ratio of this structure?
Problem 2.10.
a. Find the volume of the cell built on the primitive lattice vectors, Eq. (2.5). What
is its relation to the afore-mentioned volume of the cubic unit cell?
b. Another optional unit cell for the BCC lattice consists of the vectors a1 = ax̂,
a2 = aŷ, a3 = r2 = a(x̂ + ŷ + ẑ)/2. Find the number of atoms in this unit cell, and
its volume. Are these primitive lattice vectors? Is it a primitive cell?
Table salt and the face-centered cubic lattice (FCC). When the sites in
the SC lattice are occupied alternately by two types of atoms (or ions), as done
in the planar analog, Fig. 2.7, the outcome is the three-dimensional structure of
table salt, sodium chloride (NaCl), shown in the right panel of Fig. 2.17. The
same structure pertains (in Nature) to crystals in which the sodium is replaced by
some other alkali atoms from the first column of the periodic table, for instance
lithium (Li) or potassium (K), and also when the negative halogenic ion is replaced
by others from the column before the last one in the same table, e.g., fluorine (F) or
iodine (I). Quite generally, when a lattice site of the cubic lattice accommodates a
small ion (for example, sodium), then all its six nearest neighbor sites (two along the
positive and negative directions of each cartesian axis) contain the large ion, (e.g.,
chloride), and vice versa. This structure has the minimal potential energy for these
ions, for a certain range of the radii ratios (see Chapter 4). The crystal displayed in
Fig. 2.7 is a planar cross section of the structure in the right panel of Fig. 2.17, for
instance, the upper plane in that figure. Note though that neighboring planes are
not placed precisely one on top of the other; rather, above each sodium ion there is
a chloride one, and vice versa. The neighboring planes are shifted relative to each
other. As in Fig. 2.7, the ions in the right panel of Fig. 2.17 do not share the same
environments, and consequently this structure is not a Bravais lattice. The base
of this crystal contains two neighboring ions, represented by a small sphere and a
large one; see the atoms in the ellipse on the right panel. Alternatively, the crystal
consists of two identical simple cubic sublattices displaced relative to one another
along one of the cartesian axes, e.g., by the vector r2 = (a/2)ẑ connecting the two
ions in the base. As found in the two-dimensional example, each such base can
be represented by a single lattice point. For instance, the crystal can be described
by the lattice built solely of the large spheres, with each lattice site (at the center
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 38

38 Introduction to Solid-State Physics

of the sphere) representing the entire base. The resulting lattice is shown in the
central and left panels of Fig. 2.17 (projected on the plane of the lower face).

Fig. 2.17: Right panel: the table-salt crystal. The ellipse encompasses the base
(comprising a chloride ion–the larger sphere–and a sodium one–the smaller sphere).
Central panel: the face-centered cubic lattice (FCC), the Bravais lattice of table
salt. Each lattice site represents the base (see the right panel). The arrows show
the cubic lattice vectors. Left panel: a projection of the cubic unit cell of the FCC
lattice on the plane of the lower face.

The lattice shown in the central panel of Fig. 2.17 can be described by using
the cube there as the unit cell that repeats itself. Though this is not the primitive
unit cell of this lattice, the description is most convenient because the cubic unit
cell reflects the symmetry of the lattice under rotations by 90◦ , for instance around
each of the cartesian axes. When the cube is the unit cell, the lattice vectors, each
of length a, are directed along the cartesian axes and are normal to each other
precisely as for the cubic lattice in Fig. 2.15. As opposed to the unit cell of the
SC lattice, though, the present unit cell contains four lattice sites. Each of the
square faces has a point at its center, leading to the term “face-centered cubic
lattice”, abbreviated as FCC. To see that each cubic cell contains four lattice
points, one may slightly shift the cube away from the origin towards the negative
sides of all three axes, without moving the lattice points themselves. The displaced
cube contains the point at the origin, and its three nearest neighbors located at

r1 = (a/2)(x̂ + ŷ) , r2 = (a/2)(ẑ + x̂) , r3 = (a/2)(ŷ + ẑ) . (2.6)

Another method to obtain the very same result leaves the cube untouched, like in
the figure, but divides the “mass” of the ion at each lattice site among the unit
cells sharing that site. A lattice point at the corner of the cube is common to eight
cubes, while a lattice point at the center of each face resides in two cubes (at the
two sides of this face). Hence each corner “contributes” a “mass” of 1/8, and each
lattice point on the face has a “mass” of 1/2. The cube possesses eight corners and
six faces; hence the total “mass” in the cell is 8 × (1/8) + 6 × (1/2) = 4. The volume
of the cube is a3 , and the specific volume (per unit base) is a3 /4. Recalling that
each lattice point represents the base of table salt, which consists of two ions, one
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 39

The crystalline structure of solids 39

concludes that the base of the cubic unit cell of table salt includes eight ions, four
sodiums and four chlorides.
The FCC lattice is a Bravais lattice. Each FCC lattice site is surrounded
by the same environment of 12 nearest neighbors. For instance, the lattice site at
the origin of the cartesian coordinate system (say the one at the lowest left corner in
the mid panel of Fig. 2.17) has four nearest neighbors on each of the planes XY, ZX,
and YZ, leading to a total of 12 nearest neighbors. The same is true for any other
lattice site, e.g., the one at the center of the face. This large number of nearest
neighbors implies that the packing of spheres in the FCC lattice is very dense.
Indeed, the densest packing of identical spheres in three dimensions, as shown in
the following, is on an FCC lattice.
The primitive unit cell of the FCC lattice. As mentioned, the cubic unit
cell is not the primitive unit cell of the FCC lattice. Primitive lattice vectors are
the shortest lattice vectors, those that connect an arbitrary lattice site with three
of its nearest neighbors. The twelve nearest neighbors√ in the FCC lattice are all at
the same distance (i.e., half of the face diagonal, a/ 2). Hence, the lattice vectors
may be chosen as any triplet of non-planar vectors that connect the origin with its
nearest neighbors. An example for such a choice and the primitive cell built on
them, is given in Eq. (2.6) and shown in Fig. 2.18.

Fig. 2.18: Thin lines: the cubic unit cell of the FCC lattice, comprising four lattice
sites. The arrows are the primitive lattice vectors, which together with the dashed
lines form the primitive unit cell of this Bravais lattice.

Problem 2.11.
a. Find the volume of the primitive unit cell in Fig. 2.18.
b. Around each point in the FCC lattice a ball is blown up until all balls are
touching each other and cannot be grown any further. This is the densest packing
of this lattice. What is the packing ratio?
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 40

40 Introduction to Solid-State Physics

Wigner-Seitz cells. As explained, the Wigner-Seitz unit cell contains all points
in space which are closer to a given lattice point than to any other lattice point.
Figure 2.19 displays the Wigner-Seitz cells of the FCC and the BCC lattices. In the
FCC lattice the 12 nearest neighbors are located at (a/2)(±ŷ ± ẑ), (a/2)(±x̂ ± ŷ),
and (a/2)(±ẑ ± x̂). The Wigner-Seitz cell is confined by planes perpendicular to
each of these vectors, passing through their mid points. Consider for instance the
neighbor marked by a black sphere, Fig. 2.19(a). The plane normal to the line
connecting it with the central site is marked by the square within the thick lines.
Next consider the four neighbors connected by a rectangle (joined together by the
dash thick lines). The planes related to those intersect the former one on the thick
lines, which by symmetry, create a square. (Symmetry dictates that all edges and
all angles of the area so created should be identical.) Since this argument holds
for all other neighbors, it follows that the Wigner-Seitz cell of the FCC lattice
is a polyhedron built of 12 identical square faces. This body is termed rhombic
dodecahedron. Albeit being built of 12 identical squares, it is not one of the Plato
solids, since its vertices are not identical (see Appendix A). The Wigner-Seitz cell
of the BCC lattice consists of two types of faces, eight are equilateral hexagons
(separating the central lattice site from its 8 nearest neighbors at the corners of
the cube) and six are squares (located on the faces of the cubic unit cell), which
separate the central lattice site from its next six neighbors. By symmetry, the edges
of the hexagons and of the squares are all equal.

(a) (b)
Fig. 2.19: Wigner-Seitz unit cells of the FCC (a) and the BCC (b) lattices. Each
face of the cell is a plane normal to the line joining the site at the centre of the cube
with one of its nearest neighbors (marked by spheres), and cutting it at its center.

Problem 2.12.
Find the length of the edge of the Wigner-Seitz cell of the FCC lattice.
The zinc-blendecrystal. In a certain similarity to the cesium-chloride crystal,
the crystal of table salt may be constructed from an FCC lattice of chloride ions,
to which the (smaller) sodium ones are added at points r2 = (a/2)ẑ away from
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 41

The crystalline structure of solids 41

each Cl. In the case of table salt, the sodium ion “enters” into the cavity formed
in-between the six chloride ions “surrounding” it, precisely as the planar structure
of table salt is described in conjunction with Fig. 2.7 (where each sodium ion is
situated in the cavity created among four chloride ions). It is possible to compute the
packing ratio for crystals of the NaCl type, for various values of the radii ratio [see
problem s.2.7). Quite a number of crystals in Nature are described by sublattices of
the FCC structure, with various fillings of the cavities of this structure. The right
panel of Fig. 2.20 shows one of the structures of zinc sulphide sublattice, ZnS, whose
German name is zinc-blende. The larger spheres represent the sulphur ions, and
the smaller ones describe the zinc ones, each belonging to an FCC sublattice. The
two sublattices are displaced relative to one another, by the vector r2 = (a/4)(x̂ +
ŷ + ẑ) (a quarter of the cube’s diagonal). The projected cubic unit cell on the
lower-face plane appears in the central panel of Fig. 2.20. Note that the zinc ion
is situated in the space between four sulphur ions, but these are not located on the
same plane; e.g., the ions at the origin and at the three points given by Eq. (2.6).
Connecting all of these four lattice points to each other (the thick dashed lines)
creates a tetrahedron, whose center (the locus of its three heights) is occupied
by the zinc ion. (The tetrahedron is the simplest of Platonic solids; it consists of
four equilateral triangular faces, see Appendix A.) The zinc-blende structure is of
particular importance for semiconductors: it describes gallium arsenide (GaAs),
aluminium arsenide (AlAs), and indium arsenide (InAs), all used in many electronic
devices.
The diamond. As in the example of Fig. 2.16, the crystal of the zinc-blende
type attains a unique configuration when the two sublattices contain the same atom.
The lattice so formed is exhibited in the left panel of Fig. 2.20; it is the lattice of
diamond. In this structure each carbon atom is surrounded by four nearest neigh-
bors (i.e., each carbon atom resides at the mid point of a tetrahedron built of four
other carbon atoms), precisely like in the zinc-blende structure. These structures
are explained in Chapter 4. The spatial orientations of the various tetrahedra differ
for different carbon atoms in the unit cell. This can be observed on the right panel
of the figure: the environment viewed from a “dark” lattice point is not the same as
the one observed from a “bright” lattice site; the tetrahedra are oriented differently.
Nonetheless, all environments of the darker points are identical, and so are also
those of the brighter ones. The lattice of the diamond, therefore, must be described
by two sublattices, i.e., by an FCC lattice with a base that contains two atoms,
much like zinc-blende.
Perovskites. The three structures of ionic crystals examined so far comprise
solely two types of ions: the cesium-chloride-type, the table-salt-type, and the zinc-
sulphide-type. The number of positive nearest neighbors of each negative ion (and
vice versa) in these structures decreases from eight to six to four. This disparity
is explained in Chapter 4 in terms of the different ratios of the ions’ radii. The
unit cells become increasingly complex when the crystal is made of more than
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 42

42 Introduction to Solid-State Physics

Fig. 2.20: Right panel: the structure of zinc-blende. The larger (lighter) spheres
represent sulphur, and the smaller (darker) ones represent the zinc. Each zinc ion
has four sulphur ions for nearest neighbors (connected to it by the dashed lines),
and vice versa. The thick dashed lines create the tetrahedron caging a zinc ion.
Central panel: the projection of the cubic unit cell on the lower-face plane (the
dashed circles represent the zinc ions). Left panel: the cubic unit cell of diamond.

two kinds of ions. Examples of such crystals, important from both basic-research
and technological points of view, are the perovskites. Figure 2.21(a) portrays the
perovskite structure of the chemical formula ABO3 (where A and B stand for certain
positive ions), for instance, SrTiO3 . The strontium ions reside in the corners of
the cubes and the titanium ones are located at the centers, precisely as in the
structure of cesium chloride. The difference is that these two ions are positive, Ti4+
and Sr2+ ; the negative oxygen ions, O2− , are located at the centers of the faces,
and form an octahedron shown by the thick lines in Fig. 2.21(a); this octahedron
cages the B ion (or Ti) situated in the center of the cell. [The octahedron, built
of eight identical equilateral triangles, is one of the Platonic solids, see Appendix
A.] Another example of a perovskite crystal is BaTiO3 , much explored because
it possesses a ferroelectric phase transition: at low temperatures the positive ions
move relative to the negative ones, thus creating an electric dipole (see the end of
Sec. 2.7). Perovskites are generally cubic at high temperatures, as in Fig. 2.21(a),
but transform into a tetragonal or an orthorhombic structure at low temperatures,
see Fig. 2.28.
Generalized perovskites: Figure 2.21(b) displays the structure of lanthanum
cuprate, La2 CuO4 ; it is an example of the A2 BO4 structure. This is not a cubic
crystal, since the vertical edge of the unit cell is longer than the other two that
form the basis (which are equal). The lattice vectors are orthogonal to each other.
A lattice with mutually orthogonal lattice vectors, of which only two are of the
same length, is termed tetragonal. A certain similarity does exist between the
lanthanum cuprate and the perovskite structures: in both there are planes of AO
or of BO2 . However, in the structure shown in Fig. 2.21(b) there are two planes
of the first type in-between two planes of the second, as opposed to a single such
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 43

The crystalline structure of solids 43

(a) (b)

Fig. 2.21: (a) The perovskite structure of ABO3 , with, e.g., A=Sr, La, Ba, and
B=Ti, Al, Mn; (b) the tetragonal structure of lanthanum cuprate, La2 CuO4 . The
smaller spheres represent the oxygen ions, and the others are lanthanum or copper
ions, as marked in the figure. The lattice constants are a1 = a2 = a and a3 = c.

plane in the structure illustrated in Fig. 2.21(a). Each copper ion in the lanthanum
cuprate is surrounded by an octahedron of oxygens. The structure displayed in Fig.
2.21(b) is known as a “generalized perovskite”. The lanthanum-cuprate com-
pound is famous as being the first “high-transition-temperature-superconductor”;
it has been found that by replacing a fraction of the lanthanum ions by barium or
strontium ones the material becomes a superconductor below 35◦ K. At the time of
this discovery, this transition temperature was considerably higher than the super-
conducting transition temperatures known till then. It was later found that one
may add more planes in-between the copper ones, and so increase the transition
temperature further, even up to 138◦ K. The mechanism responsible for these high
transition temperatures is still a topic of intensive research.

Problem 2.13.
Determine the Bravais lattice of the strontium titanate, Fig. 2.21(a); find the prim-
itive unit cell and the base of this crystal. What is the relation of the base to the
chemical formula of the compound?

Problem 2.14.
Is the tetragonal cell of the lanthanum cuprate shown in Fig. 2.21(b) the primitive
one? If not, what is the primitive cell and what is the base? Find the lattice vectors.

2.6 Dense packing in space

Lattice structures of chemical elements. Figure 2.22 displays the periodic

table of the elements in Nature. The drawing in each entry indicates the lattice
structure of the crystal built on the corresponding element (in the solid phase of
the material, at low enough temperatures). It is seen that many of the metals are
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 44

44 Introduction to Solid-State Physics

arranged in the FCC or the BCC lattices. Another ubiquitous arrangement is the
hexagonal closed-packed crystal, abbreviated as HCP (or at times, HEX); see
below. For instance, the stable form of carbon is hexagonal, corresponding to the
structure of graphite displayed in the right panel of Fig. 2.9. However, there is only a
minute energy difference between this structure and that of diamond which appears
in the table as the stable FCC structure of silicon (Si) and germanium (Ge), both
located in the table below carbon (and therefore expected to have similar chemical
properties).

Fig. 2.22: The lattices formed by the elements in the periodic table: the figure in
each square indicates the structure (face-centered cubic FCC, body-centered cubic
BCC, hexagonal HEX, and more).

Close-packing in space. As discussed in Sec. 2.2, the triangular lattice

corresponds to a dense packing of circles on a plane, e.g., the ping-pong balls in
a swimming pool, see Fig. 2.3. The volume packing-ratio of this planar layer of
spheres is 0.6046, as found in problem 2.1(b). The simple hexagonal lattice
is formed when such triangular-lattice layers are placed one above the other, with
straight lines of lattice points along the direction normal to the layers. When the
spheres are also adjacent to those in the neighboring layers, the resulting volume
packing-ratio is identical to that of a single layer. This ratio decreases when the
spacing in-between the layers is extended (and the spheres in neighboring layers are
not touching each other).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 45

The crystalline structure of solids 45

The packing ratio of the simple hexagonal lattice is not the optimal one. A
larger ratio can be obtained by the following procedure. One begins with the dense
packing on the plane, i.e., from the infinite planar triangular arrangement of spheres
as in Fig. 2.3, and adds gradually more spheres. The first sphere to be added has
to initiate a new layer. This sphere is equally attracted to each of the spheres
in the first layer (as they are all identical), and therefore it “aims” to be close
to as many spheres as possible. It therefore lands in the valley created at the
middle of the triangle formed by neighboring spheres in the first layer (and not
precisely above a single sphere there, as happens in the simple hexagonal structure).
Joining the center of the newcomer with the centers of those three spheres creates
a tetrahedron. The tetrahedra repeat themselves all over the second layer; they
form a three-dimensional structure of close-packed spheres, much as the equilateral
triangles are a basis for the close-packing in two dimensions.
Each sphere in the first layer is surrounded by six valleys. However, once one of
the valleys is occupied, the two next to it remain empty (as all spheres are of equal
size, if two of them reside in neighboring valleys they have to partially “penetrate”
into one another). As a result, only three valleys around each sphere of the first
layer are occupied, for example in the arrangement of the dashed circles in Fig.
2.23 (marked by B). In this way a triangular lattice of dashed circles is formed. On
the other hand, when the valley next to the one exploited before is occupied, the
triangular lattice that emerges is the one marked by C in the figure (thin circles).
When the lattice is grown by adding atoms one by one, domains of the type B and
domains of the type C can be formed in the second layer, and they cannot be joined
together. These metastable states can be removed by carrying out carefully the
growth procedure (see Sec. 2.2); at times several cycles of heating and cooling are
needed in order to reach an arrangement based on one of these two options.

Fig. 2.23: Possible locations of spheres in the second layer above a triangular lattice.
The letter A represents the first layer (thick blue circles), and the letters B (dashed
red circles) and C (thin green circles) represent the two possible arrangements of
the second layer.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 46

46 Introduction to Solid-State Physics

There are also two options in constructing the third layer. Denoting the first
layer by A and the second by B, the spheres in the third layer can locate themselves
exactly above those in the first layer, as in Fig. 2.24(b), and then the third layer
is A. The periodicity of the structure so formed, ABABAB. . ., [depicted also in
the lower part of Fig. 2.24(b)] is of two layers in the direction normal to the
plane. This structure, called hexagonal close-packed and abbreviated HCP,
has the densest packing ratio of spheres in space. Alternatively, the centers of the
spheres in the third layer can place themselves above the empty valleys in the two
preceding layers as in Fig. 2.24(a), and then the third layer is marked by C, and
the structure is ABCABCABC. . . This structure is identical to the FCC lattice,
and it also appears under the name cubic close-packed, CCP. The lower panel
of Fig. 2.24(a) displays the cubic unit cell of the FCC lattice, as does also the
mid panel in Fig. 2.17. Note that the green spheres are all lying on the plane
spanned by three corners of the cube (two opposite ones in the base of the cube
and an opposite one in the layer above, connected by thin dashed lines). These
green balls are part of a triangular lattice. Likewise, the red balls, again lying on a
plane passing through three corners of the cube, are also part of a triangular lattice
(parallel to the previous one), and each of them resides in a valley formed in the
green plane. The blue sites on each side of these two planes create the third layer
in the ABCABC. . . arrangement.

(a) (b)

Fig. 2.24: (a) The FCC≡CCP structure; (b) the HCP structure. Each color
marks a dense planar layer.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 47

The crystalline structure of solids 47

The hexagonal close-packed lattice. The hexagonal prism in the lower part
of Fig. 2.24(b) can serve as the unit cell of the HCP lattice, but it is not the
primitive unit cell. The primitive unit cell of the triangular lattice, portrayed in
Fig. 2.4, is a rhombus; in the HCP structure, the layer above is also a triangular
lattice and consequently the primitive unit cell of this lattice is the prism with a
rhombohedral base, indicated in the figure by thick lines. The two√planar lattice
vectors in Figs. 2.4 and 2.5, i.e., a1 = (a, 0, 0) and a2 = (a/2, a 3/2, 0) (both
are of length a, with an angle of 60◦ in-between them) are reinforced by a third
vector normal to the plane, a3 = (0, 0, c), of length c. The prism contains two
lattice sites, marked in the figure by A1 and √ A2. The base points in the unit cell
are located at the origin and at r2 = (a/2, a 3/6, c/2) (just above the mid point
of the equilateral triangle below). In the dense hexagonal structure, the spheres
in the mid layer touch the spheres in the lower and upper layers. Adding the
 |r2 | = 2r = a (where
constraint  2r = a is the diameter of each sphere) yields
c = 2a 2/3 ≈ 1.633a (c/2 = a 2/3 is the height of the tetrahedron formed by the
sphere inside the unit cell and the three others in the basis touching it; the edge of
this tetrahedron is of length a. Check!) Nonetheless, in many hexagonal materials
c > 1.633a.
 The notation HCP is kept solely for the dense structure for which
c = 2a 2/3 ≈ 1.633a.

Problem 2.15.
a. Calculate the packing ratio of the HCP lattice.
b. Compare with that of the FCC lattice (problem 2.11). Are you surprised?

Problem 2.16.
Triangular lattices of spheres are arranged in layers, forming the structure
ABCBABCBA. . .. Determine the primitive unit of this lattice and calculate its
packing ratio.

The two lattices, HCP and FCC, share the same density, 0.74, which is the
largest packing ratio of identical spheres in space (see problem 2.15). They also
share the same potential energy if only that of nearest-neighbor pairs is accounted
for. As can be seen in the periodic table, Fig. 2.22, these two structures are rather
ubiquitous. The factors that determine which of them is more stable hinge on
delicate considerations, like the interaction between farther-away neighbors (Chap-
ter 4). The subtle disparity in the arrangements of successive planes leads at
times to more complex structures, e.g., ABCBABCBABCBA. . . (problem 2.16),
ABCACBABCACB. . . (with periodicity of six planes), and more. All these struc-
tures share the same densest packing ratio. Nature produces at times also random
arrays, with no periodicity along the normal to the planes.
Kepler’s conjecture. The question “what is the largest packing ratio of iden-
tical spheres in space?” had been already asked by the astronomer Johannes Kepler
in 1611 (following the one raised a century before regarding the best arrangement of
cannon balls in the ships sailing to America). Kepler conjectured that the maximal
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 48

48 Introduction to Solid-State Physics

packing ratio is 0.74, but was not able to prove it mathematically. In 1831 Gauss
succeeded to show that when the spheres are arranged periodically then indeed
the densest structures are the HCP and FCC lattices. All the periodic struc-
tures which are possible in space are sorted out below; a simple computation of
the corresponding packing ratios reproduces the afore-mentioned result. However,
the question whether there exist denser non-periodic structures remained open; it
has been even included by Hilbert in his 1900 list of unsolved mathematical prob-
lems. This question can be explored by exploiting formidable computer codes that
pursue all possibilities. In 1998 Thomas Hales argued that he had proven the Ke-
pler’s conjecture a formal proof was published in 2014 by him and his collaborators.
Non-periodic structures of identical spheres or other bodies are quite essential for
explaining the physics of granular materials. Granular systems are created when
grains are poured at random into a vessel (e.g., wheat grains poured into a silo, or
medication tablets into their container). The research of granular material is going
on for quite a number of years, but there are still many open questions, challenging
physicists and engineers.
Hexagonal crystals with a base – wurtzite. Examples of crystals built of
two types of ions have been encountered in the discussion of the cubic lattices.
For instance, the structure of table salt or of zinc-blende (Figs. 2.17 and 2.20)
contains two interwoven FCC lattices, of the sodium (zinc) ions, and of the chloride
(sulphur) ones. It turns out that ZnS assumes also the structure called wurtzite,
exhibited in Fig. 2.25. As for the zinc-blende, this structure is constructed such
that the zinc ions fill the tetrahedral cavities among the sulphur ions; in the case
of wurtzite, though, the sulphur ions occupy the sites of an HCP lattice. This
structure comprises two HCP lattices, one of the zinc ions and the other of the
sulphur ones, displaced relative to one another along the c−axis of the unit cell
by 5/8 the height of the unit cell. Due to the above-mentioned minute differences
between the CCP≡FCC and the HCP lattices, ZnS appears in Nature in the two
polymorphic structures, the particular one is chosen by subtle details, like long-
range interactions, the growth method, and the ambience conditions (temperature
and pressure).

Problem 2.17.
Find the primitive unit cell of the wurtzite structure. What is the base of this
crystal?

2.7 Classification of periodic lattices according to their symmetries

The symmetry groups of crystals. Each lattice (or crystal) possesses certain
operations that transform it to itself. Such an operation is called a “symmetry
operation”. For instance, a translation by a vector of the type given in Eq. (2.1)
leaves the lattice as is, and therefore it is a symmetry operation of the lattice.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 49

The crystalline structure of solids 49

(a) (b)

Fig. 2.25: (a) The structure of wurtzite. The larger points indicate the sulphur,
and the small ones the zinc. Each type of ions resides on an HCP lattice. (b)
The projection on the base plane of the hexagonal unit cell (the sulphur ions–large
dark circles–are at height 0 or 1/2, and the zinc ions–bright small circles–are at 1/8
height and 5/8 height of the unit cell).

The collection of the Translations by all linear combinations with integer coefficients
of the lattice vectors forms an algebraic group: the sum of any two such combi-
nations, or their product by an integer, are also vectors of the same character, and
hence the symmetry operations they represent belong to the same group as well.
(In fact, these properties are the formal definition of a group.) This group is called
the translation group of the lattice.
Likewise, most of the lattices reproduce themselves after being rotated. For
example, the planar triangular lattice is not changed upon being rotated by 60◦
around an axis normal to the plane and passing through one of the lattice sites.
Each of the cubic lattices is invariant (remains itself) when rotated by 90◦ around
an axis that coincides with one of the lattice vectors of its cubic unit cell, or by
120◦ around an axis which lies on a diagonal of the cubic unit cell. There are other
symmetry operations as well (see below). In this section it is established that all
the Bravais lattices can be classified according to their symmetries. There is only a
finite number of classes of lattices, where the lattices in each class share the same
symmetry operations that transform them back to themselves. In Chapter 3 it is
found that X-ray diffraction patterns of each lattice are invariant under the same
symmetry operations as those of the lattice itself. As the number of lattices’ classes
is finite, their structure can be identified by inspecting the symmetry operations of
the diffraction patterns.
Since a lattice is invariant under a certain symmetry operation, it follows that
applying two symmetry operations leaves it unchanged as well, and hence the com-
bined operation is also a symmetry one. This property implies that the ensemble of
all the symmetry operations of a certain lattice is an algebraic group, termed the
“symmetry group of the lattice”. It is customary to distinguish between two
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 50

50 Introduction to Solid-State Physics

types of symmetry operations and symmetry groups: a point group of symme-

try operations is the collection of operations which transform the lattice
to itself, but leave one lattice point intact. A space group includes also
symmetry operations in which all lattice points change their locations,
e.g., after a translation. The more symmetric is a lattice the larger is its
symmetry group.
The rotation angles allowed for a periodic lattice. The periodicity of the
lattice implies that there are only a few angles by which the lattice can be rotated
and still come back to itself. Consider the straight line of lattice points, CAA’D in
Fig. 2.26, where the distances between any neighboring sites is the lattice constant
of that direction, a (and hence they are all equal). Assume that there exists a
rotation axis passing through the point A which is normal to the plane of the page.
The lattice is rotated around this axis by an angle φ and comes back to itself. As
a result of the rotation, the point C moves to the point B (the points B and C are
both located at a distance a from A, AB=AC=a). The existence of this rotation
axis implies that the lattice is invariant also under the inverse rotation, by the angle
−φ. Such a rotation around an axis passing through the point A’ and parallel to its
predecessor, brings the point D to the point B’, with A’D=A’B’=a. The points B
and B’ create a straight line, parallel to the original one, CD. Since both A and A’
are on the lattice, the distance between them is necessarily an integer product, m,
of the lattice constant (along this direction), i.e., AA’=ma (m = 1 in the figure).
As the line BB’ is parallel to the line CD, the distance between the points B and
B’ must be an integer product of a as well, with a different integer coefficient m ,
i.e., b =BB’= m a (m = 2 in the figure). The length difference between AA’ and
BB’ is twice the projection of BA on CA; a simple trigonometric calculation shows
that difference to be (m − m)a = 2a cos φ. Hence,
cos φ = (m − m)/2 . (2.7)

As m and m are integers and | cos φ| ≤ 1, the allowed values for the angle φ are
obtained for m − m = −2, −1, 0, 1, 2, leading to φ = π, 2π/3, π/2, π/3, 0 (or
φ = 180◦ , 120◦ , 90◦ , 60◦ , 0).
Usually the allowed rotation angles are denoted as φ = 2π/n, and then the
integer n can be solely 1, 2, 3, 4, or 6. In this notation the rotation axis is termed
“rotation axis of order n”. (n = 1 represents a rotation by 360◦ , which is
identical to no rotation at all.) This proof forbids any other rotations, e.g., of
order 5 or 10.
Additional symmetry operations. In addition to Translations and rota-
tions, a Bravais lattice may be mapped back on itself by the following symmetry
operations: (i) reflection by a planar mirror; for instance, when the mirror is
normal to the x̂−axis at the origin, the reflection transforms the point (x, y, z) into
the point (−x, y, z); (ii) inversion through a point; each vector r originating at
that point is transformed into the vector −r; (iii) complex rotation; interweav-
ing a rotation and a reflection, or a rotation and a translation (like the motion of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 51

The crystalline structure of solids 51

Fig. 2.26: Rotations of a lattice. The rotation axes, passing through the points A
and A’, are normal to the plane of the page. The lattice constant along the lines
CD or BB’ is a.

a screw). Some of these operations belong to the point group (e.g., the inversion
through a lattice point, or the reflection through a mirror passing through a lattice
site), and other belong only to the space group (e.g., the inversionthrough a point
which is not a lattice site, or a complex rotation). Appendix A describes the Pla-
tonic bodies, whose symmetry groups include numerous elements. As mentioned
there, several of these symmetry groups are shared by periodic lattices. An example
are the symmetry groups of the cube and of the octahedron which in many aspects
are largely identical to those of the cubic lattices.
Classification of one-dimensional lattices. In one dimension there is only
a single Bravais lattice, built of a periodic array of points located at r = na, where
n is an integer and a is the lattice vector [see Fig. 2.1]. This lattice is mapped upon
itself by rotations of order 1 or 2 around an axis normal to the one-dimensional
lattice in two- or three-dimensional spaces, passing through a lattice site or through
the mid point between two lattice sites. The symmetry group of this lattice also
includes a reflection by a mirror perpendicular to the lattice at a lattice site or at
half way between two nearest neighbors, and is also invariant to inversions via these
locations. (These two operations are identical in one dimension.)
Classification of two-dimensional lattices. The square lattice in Fig. 2.6
is invariant under rotations of orders 2 and 4 (e.g., around an axis perpendicular to
the plane and passing through a lattice site or through the center of each square.
Each rotation axis of order 4 is a rotation axis of order 2 as well, but not the other
way around: an axis normal to the plane at the mid point between two nearest
neighbors is only a rotation axis of order 2). The triangular lattice (or the
hexagonal lattice) in Fig. 2.4 comes back to itself after rotations of order 2, 3, or 6.
These symmetry properties are independent of the length of the lattice constants,
and therefore they characterize all lattices of this structure. The rectangular
lattice for which the lattice vectors are normal to one another but are of disparate
lengths has only rotation axes of order 2. The most general planar lattice, the
oblique lattice shown in Fig. 2.27, has two lattice vectors of different lengths, and
the angle between them is not a right one. This lattice has also only rotation axes
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 52

52 Introduction to Solid-State Physics

of order 2. The rectangular lattice is more symmetric and indeed it is also invariant
under a reflection through a mirror normal to one of the axes at a lattice site or
at the mid-point between two nearest neighbors. These four lattices, illustrated in
Fig. 2.27, are the only possible Bravais lattices in the plane, according to
the constraint Eq. (2.7). The primitive lattice vectors a1 and a2 belonging to
each lattice, the angle γ in-between them, as well as the unit cell, are indicated in
the figure.
The rectangular-centered lattice. As mentioned in Sec. 2.5, it is at times
expedient to describe a lattice exploiting a non-primitive unit cell. The planar
example is the rectangular-centered lattice, for which each rectangle’s center is
also occupied by a lattice site. The right lower panel of Fig. 2.27 depicts both the
rectangular unit cell (with a base comprising two points, marked by dashed lines)
and the primitive one, which resembles that of the oblique lattice. With the former
cell, one may take advantage of the simpler cartesian coordinate system. Figure 2.27
lists separately the rectangular-centered lattice, since it is more symmetric than the
oblique one, and shares certain symmetry operations with the rectangular lattice:
e.g., its symmetry group includes reflections through mirrors normal to the plane
and passing through one of the edges (in contrast with the oblique lattice). For
this reason, the rectangular-centered lattice is occasionally presented as yet another
planar Bravais lattice, increasing the total number of planar Bravais lattices to five.
The point groups of the lattices in Fig. 2.27 are as follows (in all cases, the
rotation axes and the mirror planes are perpendicular to the lattice plane). The
oblique lattice possesses solely rotations of order 2 around axes passing through
lattice sites, through the mid point of each edge, and through the center of each
cell. The rectangular lattice has in addition reflection planes normal to each of
the axes, at each lattice point, at the mid point of each edge, and at the center of
each rectangle. Increasing further the symmetry, the square lattice possesses also
rotation axes of order 4 through each lattice site and through the center of each
square, and mirror planes passing through the diagonals of each square. Finally, the
most symmetric triangular lattice is characterized by rotations of order 2 around
the mid point of each edge, rotations of order 3 around the center of each triangle,
and rotations of orders 2, 3, and 6 around each lattice point. The symmetry group
of the triangular lattice also contains mirror planes through each edge, normal to
edge at its mid point.

Problem 2.18.
Why is the rectangular-centered lattice not mentioned in the preceding paragraph?

Classification of three-dimensional lattices. Figure 2.28 displays the al-

lowed Bravais lattices in space. The lengths of the lattice vectors and the angles
in-between them are defined in Fig. 2.13. Seven of the lattices have a primitive
unit cell that contains a single lattice site. In three of them the lattice vectors
are normal to each other: in the cubic lattice (all lattice vectors are of the same
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 53

The crystalline structure of solids 53

Fig. 2.27: The four Bravais lattices in the plane: the square, the triangular, the
oblique, and the rectangular. Also are shown the lattice vectors, the angle in-
between them, and the primitive unit cell. The primitive lattice vectors of the
fifth lattice, the rectangular centered, are like those of the oblique lattice, with an
arbitrary angle γ. The lengths of this lattice are such that cos γ = a1 /(2a2 ), and
the vectors a1 and 2a2 − a1 form a right angle. Indeed, these two vectors create a
rectangle, shown by the dashed lines. This rectangle represents the unit cell of a
rectangular lattice, with a base containing one point at the corner and one at the
center.

length), in the tetragonal lattice (two of the vectors have the same length) and
in the orthorhombic lattice (all lengths are different). These three types possess
rotation axes of order 2, and the first two have in addition axes of order 4. The
symmetry group of the cubic lattice also includes axes of order 3 (around the cube’s
diagonal). Note that all the symmetry operations of the cubic lattice are included
in the point group of the rhombic dodecahedron, shown in Fig. 2.19(a). There is
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 54

54 Introduction to Solid-State Physics

also the three-dimensional hexagonal lattice (displayed in Fig. 2.9 in connec-

tion with graphite and in Fig. 2.24 in the context of the HCP lattice), which has
rotation axes of orders 2, 3, and 6; the monoclinic lattice for which there exist
only axes of order 2; the trigonal lattice (called at times rhombohedral) whose
lattice vectors are all equal in length and the angles in-between them are equal to
each other as well; it therefore has a rotation axis of order 3 around the diagonal
of the unit cell (which forms identical angles with all three lattice vectors); and the
triclinic lattice, which has no rotations at all, except for the trivial one, of order
1.
Several of these lattices appear also as structures with a base, e.g., the BCC,
the FCC, and the diamond lattices. These have the same symmetries as the SC
lattice. The structure of lanthanum cuprate, displayed in Fig. 2.21(b), is an example
of a body-centered tetragonal lattice. Figure 2.28 shows also other structures
with a base, that have the same symmetries as one of the seven basic lattices. All of
them are Bravais lattices, though their symmetries are more easily observed when
considered as crystals with a base. Together with the structures with bases, there
are fourteen Bravais lattices in space.
Problem 2.19.
The list of spatial Bravais lattices does not include the base-centered tetragonal lat-
tice and the face-centered tetragonal lattice. Why?
Problem 2.20.
Show that the lattices formed by the primitive unit cells of the BCC and FCC
lattices, depicted in Fig. 2.16 and 2.18, are particular cases of the tetragonal lattice.
Determine the angles in-between the lattice vectors in each of them.
Problem 2.21.
Prove that any Bravais lattice is mapped back upon itself under an inversion through
a lattice point.
Problem 2.22.
a. List the symmetry operations of the tetragonal lattice.
b. Which members of this group disappear once the lattice is distorted into a mon-
oclinic structure?
c. Which operations are added to the group when the tetragonal lattice is distorted
to be a cubic one?
Problem 2.23.
A circle is placed at each lattice point of the (planar) oblique lattice, such that all
circles are touching each other. Find the planar packing ratio of this arrangement,
and show that (in the appropriate limits) the result reproduces the packing ratios of
all other lattices in Fig. 2.27. Compute the maximal packing ratio. To which lattice
does it belong? [Hint: analyze first the angle in the range 60◦ ≤ γ ≤ 120◦ and then
show that other angles can be mapped into this range.]
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 55

The crystalline structure of solids 55

Fig. 2.28: The seven spatial Bravais lattices (with a single lattice site in the prim-
itive unit cell) and the seven Bravais lattices with a base (each sharing the same
symmetries with one of the formers). The letter P indicates “primitive‘’, B is “body-
centered”, F is “face-centered”, and C is “based-centered”. The lengths of the three
lattice vectors are denoted by a1 , a2 , and a3 , and the angles among them are defined
in Fig. 2.13.

Subgroups of the symmetry group. The symmetry of a crystal with a base,

or a lattice in which each site is occupied by an atom or a molecule with a non-
spherical internal structure, is lower than the symmetry of the point-like lattice.
For example, when each lattice site of the one-dimensional lattice [Fig. 2.1] is
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 56

56 Introduction to Solid-State Physics

occupied by a structure of the form of the digit 0, this lattice still returns to itself
under reflections, under rotations of order two, and under inversions. All these
symmetries are lost when the site structure is replaced by one of the form of the
letter E. If the structure placed at each lattice site is of the form of the letter W
then the symmetry with respect to inversions still remains, but the rotation axis is
only of order 1. Thus, any solid material is characterized by its spatial symmetry
group, which is usually a subgroup of the symmetry group that characterizes the
corresponding point-like Bravais lattice. Therefore complex crystals possess many
more symmetry groups (as compared to the point-like Bravais lattices). In three-
dimensional space there exist 32 point groups and 230 space groups. They are
marked in various forms. The symmetry groups and their different notations are
well-documented, see e.g., A. J. C. Wilson, International Tables for Crystallography
(Kluwer Academic, Dordrecht, 1995).

Problem 2.24.
Identify all the symmetry operations of the one-dimensional crystals shown in Figs.
2.1 and 2.2.

Problem 2.25.
Suppose that each of the unit cells of the planar lattices displayed in Fig. 2.27 is oc-
cupied by a base comprising an atom at the coordinate origin of the cell and another
atom at the point r2 = (a1 + a2 )/3. Determine the modifications in the symmetry
groups, as compared to the symmetry groups of the original Bravais lattices.

Phase transitions. The crystal structure of a material is ultimately deter-

mined by the competition between the energy E and the entropy S. Recall that
the free energy is given by F = E − T S, and that any system “aspires” to be in
the state of minimal free energy. [Enough to mention that the Boltzmann prob-
ability to find the system at temperature T in a state where it has the energy E
is proportional (up to a normalization given by the sum over all probabilities) to
g(E) exp[−E/(kB T )], where g(E) is the degeneracy of the energy E (the number of
different states with the same energy), and kB is the Boltzmann constant. When
the entropy is written in the form g(E) = exp[S/kB ], this probability is propor-
tional to exp[−(E − T S)/(kB T )] = exp[−F/(kB T )], and consequently there is a
higher probability for the system to be found in a state where the free energy F
is lower.] The second term in the free energy, −ST , is dominant at high temper-
atures, and then states with higher entropy, i.e., with higher symmetry, are more
favorable. As the temperature T is reduced, the significance of this second term is
diminished; consequently the symmetry can be reduced and the system may attain
more ordered states and thus gain energy (the energy is usually smaller as the sym-
metry is reduced). Therefore, many materials undergo phase transitions among
various crystalline structures as the temperature is varied. These transitions occur
at temperatures termed “transition temperatures” and may be either of first
order, when the material “jumps” discontinuously from one structure to another,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 57

The crystalline structure of solids 57

or of second order, when the structure changes continuously as the temperature

drops below the transition temperature. It is common to refer to such transitions
as symmetry-breaking transitions. Thus, e.g., it is possible that a material
crosses over (as the temperature is reduced) from its gaseous state to a liquid one,
then to a solid phase with cubic symmetry, from it to a tetragonal structure, which
is less symmetric, and then to an orthorhombic structure, which is even less sym-
metric. This is the case with the two perovskite structures described in Fig. 2.21.
An hexagonal lattice may reduce its symmetry by transforming into a trigonal or
an orthorhombic lattice.
Ferroelectricity. A conspicuous example for a symmetry-breaking phase tran-
sition is found in ferroelectric materials, e.g., barium titanate, BaTiO3 . At high
temperatures this material has the perovskite structure, see Fig. 2.21(a). The bar-
ium ions are located at the corners of the cube, marked in the figure by A. At high
temperatures the structure is cubic, like the one shown in the right panel of the
figure, and therefore the centers of mass of the positive and the negative ions (in
each unit cell) are found exactly at the same point: for the cubic cell in Fig. 2.29,
there are +4 charges in the sites at the center of the cube, eight 1/8’s of +2 charges
in the sites at the corners, and six 1/2’s of −2 charges in the sites at the centers
of the faces. Altogether, there are +6 charges in the unit cell, and −6 charges that
can be assigned to the center of the cube. Because the centers of mass coincide,
this lattice lacks an electric dipole [which exists only when the center of mass of
the positive charges and the one of negative charges are separated by a finite dis-
tance; the moment is equal to the product of either the total positive charge or the
total negative one (equal up to a sign) with that distance]. Below the transition
temperature, the titanium and the barium ions move along one of the axes (chosen
at random), whereas all other ions remain in their locations. As a result of this
displacement, the symmetry is reduced from cubic to tetragonal, and an electric
dipole moment is created in the displacement direction, see the left panel of Fig.
2.29. In this new state the center of mass of the positive charges is shifted relative
to the center of the cube, along one of the cartesian axes. If the crystal is placed in
an electric field parallel to one of the axes during the cooling process, the ions move
so that the dipole moment created below the transition temperature is parallel to
the electric field; it is so oriented also after this field is switched off. Reversing the
electric field direction inverts the direction of the displacement, and thus also the
direction of the dipole moment.
This new phase is called “ferroelectric”. The “ferro” part of this term origi-
nates from the similarity between this phenomenon and ferromagnetism (Sec. 2.10);
it indicates that each unit cell carries the same dipole moment, all dipoles point-
ing at the same direction (like the magnetic dipole moments in a ferromagnet).
The “electric” part of the name characterizes the response to an electric field; the
dipole moment of the crystal follows the direction of an applied electric field. This
property of ferroelectric crystals enables them to keep and process information in
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 58

58 Introduction to Solid-State Physics

Fig. 2.29: The phase transition of barium titanate from the cubic structure (on
the right) to the ferroelectric tetragonal structure (on the left). The barium and
the titanium ions move slightly away from their locations in the perovskite lattice
shown in Fig. 2.21(a).

small samples. For this reason ferroelectric materials are very useful for various
technological applications.

Problem 2.26.
Based on Figs. 2.1 and 2.2, propose a one-dimensional model for the ferroelectric
phase transition.

2.8 Quasicrystals

The discovery of Shechtman –quasicrystals. The allowed rotation axes in

a Bravais lattice are proven to be solely of order 1, 2, 3, 4, or 6, and therefore
the lattice can be mapped back on itself only after rotations by 360◦ , by 180◦ ,
by 120◦ , by 90◦ , or by 60◦ . Chapter 3 discusses the experimental techniques to
identify the crystal structure of a given material. A particularly useful method
is based on the diffraction picture created by the interference of radiation waves
scattered from the periodic array of the crystal. The diffraction pattern reflects
the symmetry properties of the crystal; specifically, it contains the same rotation
axes that exist in the crystal. While investigating the diffraction patterns of alloys
grown from a mixture of aluminium and manganese, Shechtman had observed in
1982 “circles” around the center of the picture, comprising ten points, and invariant
under rotations by 36◦ . Such a symmetry operation, of a rotation axis of order 10,
contradicts Eq. (2.7), and is therefore forbidden according to the analysis above.
Indeed, the scientific community had doubted this result, hindering its publication
till 1984. The discovery was fully recognized in 2011, when Shechtman was awarded
the Nobel prize in chemistry. By now, many more materials of such symmetries
have been discovered, as well as crystals possessing other “forbidden” rotation axes,
like 8 or 12, and it is completely accepted nowadays that Shechtman discovered a
new phase in condensed matter. It became clear that the “new” symmetries
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 59

The crystalline structure of solids 59

are possible since the atoms in such materials are not arranged periodically; these
are not Bravais lattices. They are called “quasicrystals”, and their intriguing
properties are currently being investigated by many research groups.
The diffraction pattern conveys information on the Fourier transform of the
density probability in the sample, see Chapter 3. The density of the material
itself is obtained by computing the inverse transform of the diffraction pattern.
This technique, applied to the data of Shechtman, produces complex structures in
which the atoms are arranged according to certain rules, but are not periodically
ordered. Alternatively, one may explore the spatial structure of the atoms in
quasicrystals by measuring the properties of their surfaces. It is possible to observe
the surface directly by using a scanning electron microscope (see Chapter 3), and
it is also feasible to adsorb certain atoms on the surface and then monitor their
locations. The atoms place themselves at minima of the potential energy of the
surface; measuring these locations enables one to deduce the structure of the surface
itself. The right panel of Fig. 2.30 shows a picture, taken by a microscope, of silver
atoms adsorbed on the surface of a Al-Pd-Mn quasicrystal (made of aluminium,
manganese, and palladium). The left panel of Fig. 2.30 portrays the map of the
potential energy of the surface, as deduced from the microscope picture in the right
panel. Darker areas represent lower potential energy. The figure emphasizes a place
of strong attraction, which indeed adsorbed a silver atom. The left panel contains
many points whose local surrounding is invariant under rotations of order 5.

Fig. 2.30: Right panel: a picture taken by a scanning electron microscope of silver
atoms adsorbed on the surface of the quasicrystal Al-Pd-Mn. The left panel shows
the corresponding potential-energy map. Taken from: B. Unal, V. Fournée, K. J.
Schnitzenbaumer, C. Ghosh, C. J. Jenks, A. R. Ross, T. A. Lograsso, J. W. Evans,
and P. A. Thiel, Phys. Rev. B 75, 064205 (2007).

A new definition of a “crystal”. Following Shechtman’s discovery, it be-

came clear that quasicrystals are not periodic and hence cannot be described by
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 60

60 Introduction to Solid-State Physics

Bravais lattices, as this description requires a single unit cell that repeats itself.
Nonetheless, these structures do possess certain symmetry properties similar, in a
sense, to those of lattices, and do have a certain order. In addition, the diffrac-
tion patterns of quasicrystals comprise discrete radiation points – like those of a
periodic lattice (Chapter 3). The discovery of Shechtman caused the International
Union of Crystallography to modify the definition of a “crystal”, omitting the word
“periodic”. As of 1992, a crystal is a material whose diffraction pattern
comprises discrete points.
Fibonacci sequence. Perhaps the simplest example of a quasicrystals is pro-
vided by the Fibonacci sequence and the related one-dimensional lattice.
The sequence was proposed in 1202 by Leonardo da Pisa, alias Fibonacci, as a
model for the growth of hare populations. In this model, L (for long) represents a
couple of mature hares, and the letter S (for short) is a pair of babies. Each month
a mature couple gives birth to a young pair, and therefore each L is replaced by LS
(this notation indicates an ordering in which S appears to the right of L). During
that period a young pair grows up, and therefore S is replaced by L. The rows in Fig.
2.31 show the time evolution of the population according to these rules; each row
represents a generation. The row reached after an infinite number of generations is
the Fibonacci sequence.

Fig. 2.31: Successive generations of hare populations. (The order within each row
matters!)

Fibonacci series. Each row in Fig. 2.31 is constructed by merging together the
preceding two rows, placing the previous row to the left of the row before. Hence,
the number of letters (hares) in the nth row obeys
Fn = Fn−1 + Fn−2 . (2.8)
Beginning with S, like in the figure, then F0 = 1 and F1 = 1; as a result Fn =
1, 1, 2, 3, 5, 8, 13, 21, 34, . . .. This sequence of numbers is the “Fibonacci series”,
or Fibonacci numbers. For very large n it can be shown that the ratio of numbers in
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 61

The crystalline structure of solids 61

two successive generations, Fn /Fn−1 , approaches a constant, τ ; this constant is the

positive
√ root of the quadratic equation τ = 1 + 1/τ . It is called the golden ratio,
τ = ( 5 + 1)/2 = 1 + 2 cos(2π/5) = 1.618 . . . (see problem 2.27). This irrational
number appears in numerous contexts in physics and mathematics.
Fibonacci lattice. The letters in each row of Fig. 2.31 can be replaced by
linear segments: the letter L is replaced by a segment of length τ , and the letter
S by a shorter segment, of length 1. The row before the last in Fig. 2.31 is then
replaced by the one-dimensional lattice shown in the lower part of Fig. 2.32, and
the line before it by the lattice in the upper part. (The lengths are different in the
two parts of Fig. 2.32; for reasons to be clarified below they are chosen such that
the two lattices have the same length.)

Fig. 2.32: Fibonacci lattice, built of linear segments according to the Fibonacci
rules. The long segment is denoted a1 , and the shorter one is a2 .

By inspection, the sites in Fig. 2.32 do not form a Bravais lattice. It is impos-
sible to reach them all by integer multiples of a single lattice vector. On the other
hand, each point can be reached after an integer number of L steps and another
integer number of S steps. The location of a point on the lattice (relative to the first
one) can be written as xn = m1 a1 + m2 a2 , where m1 and m2 are certain integers,
and the vectors a1 and a2 (see Fig. 2.32) represent the steps, L and S, respectively.
Although the “lattice” is one dimensional (the points are located on a line), it is
described by two different lattice vectors, whose lengths’ ratio, τ , is irrational.
It is paramount to note that, contrary to the aforementioned lattices, the coeffi-
cients m1 and m2 cannot assume any integer value! These values are determined
by the Fibonacci rule, which limits their choice. For instance, the combinations
SS or LLL never appear in the rows in Fig. 2.31. The distance from the origin
(chosen to be one of the sites) of the point xn is |xn | = m1 τ + m2 . Were m1 and
m2 allowed to be any integer, then |xn | would have approached arbitrarily close any
numerical value. For example, 518τ − 835 ≈ 3.14161 is equal to π up to the fourth
digit (the reader is encouraged to approximate another arbitrary real number in a
similar way). It follows that by allowing the coefficients to be arbitrary integers,
one could have obtained lattice sites as close as one wishes to any point on the
line, leading to a very dense arrangement, in contrast with the lattices in Fig. 2.32.
There, the distance between nearest neighbors is never less than 1. Although the
two lattices in Fig. 2.32 are not periodic, the locations of the sites are determined
by the Fibonacci rule, and so are inter-related. This regularity causes the diffraction
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 62

62 Introduction to Solid-State Physics

patterns of the Fibonacci lattice to contain only discrete points. This lattice is thus
a one-dimensional quasicrystals.

Problem 2.27.
a. Prove that the ratio tn = Fn /Fn−1 approaches the golden ratio for large n.
b. Find the ratio between the number of longer segments to the number of shorter
ones in an infinite Fibonacci lattice.

Inflation symmetry. As mentioned, the Fibonacci lattice is not periodic.

However, the symmetry with respect to discrete Translations is replaced by another
symmetry: this lattice is self-similar under inflation transformations. Stated
differently, when lengths are measured by rulers with different unit lengths, the
lattice looks the same, independent of the units’ lengths. This property can be
grasped by mapping the lattice in the lower line of Fig. 2.32 back onto the previous
generation. Each pair of segments represented by LS, whose total length is τ + 1,
is replaced by a single segment in the upper line of the figure, which is termed L’.
Similarly, each L which is not followed by S, is copied to the upper line with no
change in length, and is denoted S’. It is then seen that the upper row in Fig. 2.32
represents faithfully the previous generation (the third from the end line in Fig.
2.31) provided that the new long segments are L’ and new short ones are S’. This
mapping procedure yields that the length of L’ is 1 + τ , and that of S’ is τ . The
ratio, L’/S’= (τ + 1)/τ is τ , as dictated by the quadratic equation determining the
golden ratio. Thus the lattice in the upper line of Fig. 2.32 is identical to the one in
the lower row, provided that the unit of length is multiplied by τ ! Such a symmetry
under inflation characterized many fractal structures, see Chapter 7.
Fibonacci lattice as a projection of a square lattice. The description
of the one-dimensional Fibonacci lattice by two lattice vectors is not accidental.
This lattice can be built by a projection of a square lattice. Figure 2.33 portrays
the projection of a square lattice on a straight line, which represents the physical
space. One draws a straight line tilted by the angle α relative the x̂− axis in a
fictitious square lattice, with tan(α) = 1/τ . All lattice points located within a
finite-width band that encompasses the straight line and is parallel to it (shown in
the figure by the two dashed lines) are projected on the straight line. One may
choose the band width, w, such that 3a sin(α) < w < 2a cos(α). For instance, w =
a[cos(α) + sin(α)] (the final result is independent of this arbitrary choice). In this
case the collection of lattice points projected on the straight line forms a staircase
structure, as seen in the figure. The left inequality ensures that no successive three
horizontal segments appear on the same stair, while the right inequality ensures
that the height of each stair does not exceed the length of a single step. Denoting
the coordinates of these lattice sites by rn = m1 a1 + m2 a2 , where a1 and a2 are
the lattice vectors of the square lattice, the projected points on the straight line
are located at xn = a[m1 cos(α) + m2 sin(α)]. The distance between two successive
points on the straight line assumes only one of two values, corresponding to the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 63

The crystalline structure of solids 63

projections of the two lattice vectors, the vertical and the horizontal, of the square
lattice. For example, the distance between the points x1 and x2 is L= a cos(α),
and the distance between the points x2 and x3 is S= a sin(α). The ratio of these
is τ . The conditions on w eliminate configurations of the types SS or LLL, and
therefore it can be proven that the order in which the segments L and S appear on
the thick line is identical to the order in the Fibonacci sequence. In particular, the
arrangement of the segments in Fig. 2.33 is identical to the one shown in the last
row of Fig. 2.31, beginning at the fourth letter there (the starting point is arbitrary
and is related to the choice of the width w of the band).

(a) (b)

Fig. 2.33: Projecting a square lattice on a one-dimensional quasicrystal. (a) The

lattice points in-between the two dashed lines are projected on the thick line. (b)
A magnification showing the points xn on the thick line.

Problem 2.28.
Prove that the distances of the points on the thick line in Fig. 2.33 from the origin
are given by

xn = S[n + ||n/τ ||/τ ] , (2.9)

where ||y|| is the integer closest to y. This expression enables the extension of a
Fibonacci lattice to negative n’s. How does one prove that this is indeed a Fibonacci
lattice?

Problem 2.29.
a. Suppose that the thick line in Fig. 2.33 passes through one of the square-lattice
sites. How many more lattice sites are then on this line?
b. Suppose that tan(α) in Fig. 2.33 is a rational number, i.e., tan(α) = p/q where
p and q are integers with no common factor. Describe the lattice that will appear
on the straight line.
c. It is possible to approximate the irrational number τ ≈ 1.61803 by a sequence of
rational numbers, 1.6, 1.61, 1.618. Which lattice is obtained by projecting from two
dimensions onto a straight line for each of these approximations?
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 64

64 Introduction to Solid-State Physics

The one-dimensional quasicrystals is described by two lattice vectors, whose

lengths’ ratio is irrational. This is a general property of quasicrystals: they
can be described by several lattice vectors whose number is larger than
the spatial dimension; the corresponding lattices can be obtained from
a projection of a periodic lattice of a higher dimension.

(a) (b)

Fig. 2.34: Attempts to tile the plane by pentagons.

Penrose tiles. A two-dimensional version of a quasicrystals was already an-

alyzed by Kepler in the seventeenth century, and by Penrose in the last century.
Penrose explored the question of covering the plane by tiles, i.e., tiling the
plane. As seen in Fig. 2.27, Bravais lattices can completely tesselate the plane, by
using identical oblique, square, triangular, rectangular, or hexagonal tiles. In con-
trast, this cannot be accomplished by identical pentagons: attempting to encircle a
point by identical pentagons, with the corner angle 3π/5 = 108◦ (as the pentagon
can be divided into three triangles, the sum of the five corner angles is 3π) leaves
a space of 36◦ , see Fig. 2.34(a). However, Fig. 2.35(b) illustrates that the plane
can be covered by pentagons, provided that rhombus-shaped tiles of opening angle
36◦ are added. That is, the plane can be tiled by two types of tiles. Penrose
demonstrated that the plane can be covered by two types of rhombi with identical
edges but different opening angles, 36◦ and 72◦ , displayed in Fig. 2.35(a). As shown
there, at certain points there appears symmetry with respect to rotations of order
5 (e.g., where 5 “big” rhombi meet at a point). Assigning a unit length to the edge
of the rhombi, the length of the long diagonal of the wide rhombus is τ , and that
of the short diagonal of the narrow rhombus is 1/τ (check!). Interestingly enough,
there are cuts of the structure of Penrose, along special directions, whose distances
create the Fibonacci lattice, see Fig. 2.35(b). It can be also proven that the ratio of
the numbers of wide and narrow rhombi equals 1/τ , as in the one-dimensional Fi-
bonacci lattice. Though the tiles of Penrose cover the plane, and albeit the fact that
rotating by 72◦ the structure in Fig. 2.35(a) brings it to a qualitatively similar one,
i.e, it is statistically symmetric, the structure is not periodic. It does not come
precisely back to itself after the rotation, and it is not described by a single unit
cell. However, the two-dimensional structure of Penrose can be obtained
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 65

The crystalline structure of solids 65

by a two-dimensional projection of a five-dimensional cubic lattice. The

quasicrystalline materials discovered so far can all be described by two or three
different unit cells, which fill completely the space. As shown by Steinhardt and
Levin in 1984, an extension of the Penrose structure to three dimensions, with two
different unit cells (containing the ratio τ as well) reproduces the diffraction pattern
found by Shechtman. This requires six basic translation vectors (in place of the
three of a Bravais lattice), indicating a projection from a six-dimensional periodic
lattice.

(a) (b)

Fig. 2.35: (a) Penrose tiling. (b) Parallel lines placed at distances that form the
Fibonacci sequence. Taken from http://www.physics.princeton.edu/~steinh/
QuasiIntro.ppt, courtesy of P. Steinhardt.

Problem 2.30.
a. Propose five vectors in the plane, such that each point in the lattice of Penrose
can be described as a linear combination of them, with integer coefficients.
b. Show that these five vectors can be obtained from a projection of the lattice
vectors of a five-dimensional cubic lattice on a plane normal to one of the (five)
diagonals.

On the symmetries of quasicrystals and Platonic bodies. Appendix

A lists the five Platonic bodies, symmetric polyhedra with equilateral faces. The
tetrahedron, cube, and octahedron have already been mentioned in the context of
periodic lattices. The bigger Platonic bodies, the icosahedron and the dodec-
ahedron, are not found in periodic lattices, but they do appear in quasicrystals.
The original findings of Shechtman were identified as representing icosahedral qua-
sicrystals: the diffraction patterns display an icosahedral symmetry point, and they
can be described in terms of six lattice vectors pointing at the vertices of an icosa-
hedron. There exist also descriptions of the Al-Mn alloys as packings of icosahedra,
with Al and Mn atoms occupying their corners.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 66

66 Introduction to Solid-State Physics

2.9 Epitaxial growth, thin films, superlattices and multi-layered

structures

Crystal growth. Certain materials are found in Nature in their crystalline forms;
crystals are also created by slowly cooling liquids, or upon crystalization of salts from
a solution. Frequently, however, it is required to grow crystals under controllable
conditions in the laboratory. As mentioned in Secs. 2.2 and 2.6, crystals can be
grown by adsorption. Another ubiquitous technique is the epitaxial growth from
molecular beams, abbreviated as MBE (molecular beam epitaxy). The name
refers to a regulated growth procedure of atomic (or molecular) slabs, where the
layers adjust themselves to the periodicity of the substrate on which they are being
deposited. Another method is the deposition off the vapor of a solid material
(vaporized upon heating by a laser beam in a vacuum chamber); this procedure is
called PDE (pulsed laser deposition).
Epitaxial growth on a crystalline substrate. As different materials possess
different lattice symmetries and disparate lattice constants, there are numerous op-
tions for choosing the adequate substrate to grow upon a certain material. Cutting
a crystal along a specific direction creates planar surfaces with symmetry properties
and lattice constants that are determined by the cut direction. For instance, differ-
ent cuts of the FCC lattice (Figs. 2.17 and 2.24) yield planes with sites that form a
triangular or a square lattice, with lattice constants dictated by the direction of the
cut (check!). Thus, one has to pre-design not only the substrate material, but also
the direction of the cut on which the crystal is to be grown. Modern technologies
enable the growth of thin films, comprising a single layer of the adsorbed atoms,
or a small number of such layers. These technologies allow for the investigation of
two-dimensional slabs (recall that these are lying on a substrate and hence are not
exactly two dimensional, in contrast with graphene, Fig. 2.10). It is also possible
to grow one-dimensional systems, e.g., by an adsorption on a substrate crossed by a
dent, followed by a removal of all deposited material except that sunk in the dent.
One-dimensional epitaxial growth. When the growing crystal comprises
planes with symmetry identical to that of the substrate, and when the lattice
constants match, the growth is accomplished smoothly. However, such a perfect
harmony is rarely encountered. Consider, e.g., the process described in Sec. 2.6,
in the simpler version of a one-dimensional substrate. It begins with a straight
line of circles, on top of which additional circles, of a different radius, are to be
adsorbed, as shown in Fig.√ 2.36. In the right and mid panels of Fig. 2.36 the
radius of the new circles is 2 times larger than the radius of the substrate circles.
When the attraction between the new circles and those in the substrate is larger
than the attraction in-between themselves, they “prefer” to land in the “valleys” of
the substrate, as in the mid panel of Fig. 2.36. The result is a periodic lattice with a
lattice constant twice that of the substrate. This new lattice constant is also that of
the joint lattice of the two layers together. This is a commensurate arrangement.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 67

The crystalline structure of solids 67

A similar periodic lattice is obtained also when the radii ratio is a rational number.
For instance, in the left panel of Fig. 2.36 the radii ratio is 3/2. The adsorbed
circles can still touch one another and yet sink into one of three “valleys”. The
constant of this periodic lattice is thrice that of the substrate.
On the other hand, when the attraction among the new circles is stronger than
the one between them and the circles on the substrate they “prefer” to touch each
other, even at the price of distancing themselves from the substrate. The result is
presented in the right panel of Fig. 2.36. The requirement that the circles touch each
other causes all new circles (except the first one) not to be exactly inside a dent;
i.e., the length of the new unit cell of the joint lattice is infinite: the distance of any
point from another cannot be covered by an integer number of finite-length lattice
vectors. This outcome is expected whenever the radii ratio is irrational and the
adsorbed circles are strongly coupled; the resulting structure is incommensurate.
Recall that this feature appears also in quasicrystals: there again the ratio of length
scales is irrational, the lattice periodicity is lost, and the crystal is described by an
infinite unit cell. Indeed, the atoms’ distribution on any symmetry axis of the
quasicrystals is determined by a finite number of lattice vectors (two or more), with
lattice constants whose ratio is irrational. A similar description pertains to the
incommensurate arrangement: each layer is characterized by its own lattice vector
and hence each lattice point can be reached by a linear combination of the two
lattice vectors, with arbitrary integer coefficients.

Fig. 2.36: Adsorption on a one-dimensional substrate. Right panel: Incommensu-

rate adsorption. Mid and left panels: commensurate adsorption, with a double and
a triple step of that of the substrate lattice.

Problem 2.31.
Determine the length of the unit cell of the joint two-layer lattice, when the ratio
between the radius of the circles in the upper layer and the radius of those in the
substrate is r2 /r1 = p/q, where p and q are integers with no common factor. What
is the base?
Lock-in of commensurate arrangements. √ Figure 2.37(a) displays an in-
commensurate arrangement of circles of radius 3 on a one-dimensional substrate
made of circles of radius 1. Like in the right panel of Fig. 2.36, the unit cell of
the joint structure is infinite. Nonetheless, note that the fifth circle from the left
end in the second layer is very close to the eighth valley in the substrate. It then
may happen that the system gains energy if that circle moves slightly to the right,
into the valley, even though it leaves a narrow space relative to its neighbor. This
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 68

68 Introduction to Solid-State Physics

configuration is shown in Fig. 2.37(b). Once this happens, the surrounding of the
fifth circle in the second layer coincides with that of the first circle there, and the
structure is commensurate, i.e., it is a Bravais lattice with a unit cell of length 7 (in
units of the diameter of the circles in the substrate). Such a transformation, from
an incommensurate pattern to a commensurate one, is called lock-in, and generally
occurs when there is a good rational approximation to the irrational ratio of the
circles’ radii. In the example
√ discussed here, the distance between the centers of
the first and fifth circles, 4 3 ≈ √ 6.9282, is replaced from by 7. This replacement
amounts to the approximation of 3 by the √ rational number 7/4 = 1.75. A similar
lock-in transformation takes place when 3 is approximated by any other rational
number, e.g., 173/100.

(a) (b)
Fig. 2.37: (a) Incommensurate adsorption. (b) Commensurate “lock-in” of the
adsorption in (a).

Two-dimensional epitaxial growth. In two dimensions there are many more

options for commensurate and incommensurate arrangements. An example is the
growth of a noble gas, like krypton, on a graphite substrate. Recall that the planes
building graphite are hexagonal, like the graphene lattice (Figs. 2.8 and 2.9). The
way the krypton atoms are organized is dictated by the competition between two
interactions. Each krypton atom is attracted to the minimum of the potential
energy, located at the center of each hexagon. Even when this interaction is the
only one operating, the krypton atoms still face a problem, as their radius is larger
than the distance between the centers of neighboring hexagons. Therefore, these
atoms cannot occupy all hexagons. In this scenario, the adsorbed atoms create
a new triangular lattice, that covers only one third of the hexagons (marked in
red and by the letter α in Fig. 2.38). As seen in the upper part of the figure,
there are three different configurations to cover the substrate, giving rise to regions
with different arrangements of the krypton layer (over hexagons of different colors);
this is similar to the two possibilities to arrange the second layer of spheres in the
close-pack configuration shown in Fig. 2.23.
So far the interaction between pairs of krypton atoms, due to which they would
have preferred to arrange themselves in a close-pack configuration, has been ignored.
With this interaction alone, the lattice constant of that configuration is determined
by the minimum of the potential energy. Once the attraction in-between the krypton
atoms is considerably stronger than that between them and the substrate, they tend
to arrange themselves in the pattern displayed in the lower part of Fig. 2.38. There,
the krypton structure is indeed incommensurate with the substrate: each layer has
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 69

The crystalline structure of solids 69

Fig. 2.38: Adsorption of krypton atoms on a graphite substrate. The three colors
represent three optional locations of the adsorbed atoms. Top: commensurate ad-
sorption; bottom: incommensurate adsorption. [R. J. Birgeneau and P. M. Horn,
Science 232, 329 (1986).]

its own lattice constant, and the ratio between those constants is irrational (the
krypton atoms are not exactly spheres, and the radii in the figure are adjusted to
the various possible structures). Intermediate configurations are also possible: when
the two interactions are comparable in magnitude, it may happen that after several
cycles of the upper layer, one of the krypton atoms reaches close to the center of a
certain hexagon; it then can gain energy by sliding into it, [as in Fig. 2.37(b)] on
the expense of the energy it loses by increasing the distances to its neighbors, or on
the expense of creating a small space in-between two dense regions (i.e., it creates a
“domain wall”). Such a lock-in transformation of the new layer produces a joint
lattice of the two layers, with a larger unit cell. The lattice constant of this new
lattice is the distance between these spaces.

Problem 2.32.
Determine the joint unit cell of the two layers in the commensurate lattice shown
in the upper part of Fig. 2.38. What is the base?

Superlattices. It is possible to grow “sandwiches” comprising several layers in

a controllable manner. Figure 2.39 exhibits several layers of lanthanum manganate
and strontium manganate, in a periodic array; each period contains a single layer
of one of these perovskites; along the normal axis the structure is built of alternat-
ing layers of the two materials. Their lattice constants are close to one another,
and therefore it is relatively easy to build a periodic array. It is an example of a
“superlattice”. Superlattices are structures in which different crystalline
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 70

70 Introduction to Solid-State Physics

layers of finite thicknesses repeat themselves periodically. The material in

the figure is La1−x Srx MnO3 , with x = 1/2; it is an anti ferromagnetic insulator
(see Sec. 2.10). At lower strontium concentrations, x < 1/2, the material is a
metallic ferromagnet. Due to the disparate properties (i.e., elastic, electric, or mag-
netic properties) of the parent compounds, it is feasible to engineer novel materials
with pre-designed characteristics, for instance, high rigidity, or an ability to store
magnetic information.

Fig. 2.39: A superlattice. Layers of lanthanum manganate and strontium man-

ganate; the oxygens are located on the corners of octahedra, each containing a
manganese ion [like in Fig. 2.21(a)]. The purple and green spheres represent La
and Sr ions, respectively [T. S. Santos, S. J. May, J. L. Robertson, and A. Bhat-
tacharya, Tuning between the metallic antiferromagnetic and ferromagnetic phases
of La1−x Srx MnO3 near x = 0.5 by digital synthesis, Phys. Rev. B 80, 155114
(2009)].

Problem 2.33.
Determine the unit cell of the structure in Fig. 2.39. How is the answer modified
for the periodic structure with x = 1/3?

Planar surfaces between two three-dimensional lattices. At times, the

contact layer in-between the substrate and the growing crystal above it has unique
properties by itself. A conspicuous recent example is the contact surface between
the two perovskites [Fig. 2.21(a)], strontium titanate and lanthanum aluminate, as
shown in Fig. 2.40. The lattice constants of these two materials are very close, and
consequently they are smoothly coupled. Since each layer of aluminium and oxygen
is negatively charged (as deduced from a summation over the various charges),
a surplus of electrons is accumulated on the contact layer, giving rise to unique
electric and magnetic properties (even superconductivity!), which can be modified
by varying the number of layers of LaAlO3 . This surface layer is currently under
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 71

The crystalline structure of solids 71

active research in many laboratories.

Fig. 2.40: Growth of LaAlO3 on a SrTiO3 substrate [M. Huijben, A. Brinkman,

G. Koster, G. Rijnders, H. Hilgenkamp, and D. H. A. Blank, Structure–Property
Relation of SrTiO3 /LaAlO3 Interfaces, Adv. Materials 21, 1665 (2009)].

Heterostructures. Another type of pre-designed materials are heterostruc-

tures. As mentioned, graphite is built of successive layers of graphene. Other planar
materials of identical structure are boron nitride and silicene, as well as MoS2 (in
which the sulphur ions are located above and below the lattice points of one of the
sublattices in the hexagonal lattice), WSe2 , and more. These are all characterized
by strong bonds within each plane, while the coupling between the planes is signif-
icantly weaker (see Chapter 4). Figure 2.41 is taken from a paper that surveys the
various possibilities for growing multi-layer structures, which consist of successive
layers of two-dimensional lattices. The growth process is compared to a game with
lego blocks–each layer is represented by lego blocks of a certain color. Of particular
potential importance are structures comprising hexagonal lattices like graphene.

2.10 Magnetism

Magnetic phase transitions. As mentioned, crystalline materials can undergo

phase transitions, such that below a certain transition temperature they acquire
new physical properties. Those are often related to a symmetry change of the crystal
(e.g., the ferroelectric phase mentioned at the end of Sec. 2.7). Here are discussed
phase transitions in the context of magnetism.
Atoms and ions often possess magnetic moments stemming from the angu-
lar momentum associated with the electrons’ motion, and/or that of the particles
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 72

72 Introduction to Solid-State Physics

Fig. 2.41: A multi-layer structure built of planar hexagonal lattices of different

materials [A. K. Geim and I. V. Griegorieva, Van der Waals heterostructures, Nature
499, 419 (2013)].

forming the nuclei, and from the spin of all these particles. The electronic mag-
netic moment is measured in units of the Bohr magneton, μB = e/(2m0 c) =
9.247×10−21 erg/Gauss, where e and m0 are the charge and the mass of the electron,
respectively,  is the Planck constant, and c is the speed of light. In an elementary
picture, non-metallic magnetic materials are described by assigning to each lattice
site i a magnetic moment represented by the vector μi . (This picture does not
pertain to metals, in which the conduction electrons possess magnetic moments as
well, and these magnetic moments are not localized.) At temperatures low enough
as compared to the transition temperature into the magnetic phase, the magnitude
of these moments is generally of the order of a Bohr magneton for each magnetic
atom (or ion) in the crystal. Several forces operate among the moments. The well-
known one is the electromagnetic force between dipole moments, that decays slowly
with the distance between them. In addition, there are also exchange forces,
originating from the quantum-mechanical treatment (see Chapter 4). The inter-
play among these forces may result in a geometrical arrangement of the magnetic
moments, when the temperature drops below a certain transition temperature.
Magnetic patterns. In the simplest case, the interactions among the mag-
netic moments cause them to coordinate themselves in parallel with each other, in
a ferromagnetic arrangement (line B in Fig. 2.42). Consequently, the material
acquires a macroscopic magnetic dipole moment. However, at times the inter-
actions between neighboring moments cause them to be antiparallel, and the result
is the antiferromagnetic order shown in row C of Fig. 2.42. Interactions among
farther away neighbors give rise to more complicated patterns, e.g., those portrayed
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 73

The crystalline structure of solids 73

in the other rows of the figure. Such interactions can also lead to incommensu-
rate arrangements of the magnetic moments (relative to the structural lattice).
For instance, the moments may rotate periodically along a certain direction in the
lattice, with a periodicity that is incommensurate with the lattice constants.
Spin glasses. The random directions of the magnetic moments in row A of
Fig. 2.42 represent two possible configurations. At higher temperatures the system
passes through states in which the moments are oriented at all possible directions,
similar to atoms in a gas which may pass through all points in space; then row
A reflects an instantaneous pattern of these moments. The temporal average of
each moment vanishes, and there is no global magnetic order. The only order in
the system is that of the lattice of atoms (or ions) building the material, with
a unit cell that contains a single atom (or ion). Upon decreasing the temperature
abruptly to a very low one, it may happen that the moments “freeze” in the random
arrangement they had at the higher temperature, and then row A describes a steady
state of the system, which does not correspond to a minimum of the free energy,
but is metastable. This configuration is called “spin glass”, and is analogous
to the “mundane” glasses, in which the atoms freeze in non-periodic arrangements
(see Chapter 1). At times, a spin glass is created when the magnetic interactions
compete with each other: a fraction of them prefers parallel arrangements of the
moments, while other attempt to make them antiparallel. The competition gives
rise to a frustration of the moments, and results in moments oriented at random
(see below). This unique situation is an example of a random system, a field
which is still under active research.

Fig. 2.42: Various magnetic patterns. From top to bottom: spin glass, ferromagnet,
simple antiferromagnet, ferrimagnet, and a structure combing a complex antiferro-
magnetic order with a ferromagnetic one.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 74

74 Introduction to Solid-State Physics

The magnetic unit cell. The ferromagnetic order in line B of Fig. 2.42 has the
same periodicity as the underlying lattice, since the moments are identical. Hence
the unit cell contains a single atom, as in the original lattice. Though the unit
cell is not changed, the symmetry group does, because the ferromagnetic lattice
has no rotational symmetry, as opposed to the underlying structural lattice. The
symmetry with respect to reflections through mirrors perpendicular to the lattice
still persists. On the other hand, the periodicity as well as the symmetry groups of
the structures in rows C, D, and E, differ from those of the original lattice, and the
unit cell contains more atoms (or ions). The structure of the simple antiferromagnet
(line C), and also that in line D (which shows a ferrimagnet: a magnetic pattern
with a ferromagnetic character – the total magnetic moment does not vanish –
combined with an antiferromagnetic feature) is similar to the one that appears in
the second panel of Fig. 2.1, or to the structure of table salt displayed in Fig. 2.7
(or in the right panel of 2.17): the unit cell is doubled as compared to the unit cell
of the simple lattice. The symmetry of the simple antiferromagnet C is higher than
the one of the structure in D.

Problem 2.34.
Determine the magnetic unit cells of each of the structures C, D, and E, in Fig.
2.42. List the symmetry operations corresponding to each of them.

Problem 2.35.
The magnetic moments of the atoms in a material whose structural lattice is a simple
cubic one, of a lattice constant a, are all directed along the same axis, and are given
by the expression μ(R) = μ0 cos(Q · R). The vector Q is Q = (1, 1, 1)(π/a), and
R is a vector to a lattice site. Determine the structure of the magnetic lattice and
find the primitive unit cell. What would have been the unit cell were the vector Q
given by (1, 1, 1)(τ /a), where τ is the golden ratio?

Frustration and three-partite patterns. An important comment is called

for: the magnetic cell of the antiferromagnet shown in line C of Fig. 2.42 contains
two moments, and the magnetic lattice can be viewed as a combination of two
intercalated simple sublattices. The same is true for the example described in
problem 2.35: the lattice there resembles the one of table salt, with the two ions
corresponding to the two moments (say, Cl for the positive moment, and Na for
the negative one). The magnetic lattice again consists of two sublattices, each
containing identical moments. Such a lattice is called bi-partite. Though this
configuration pertains to many other lattices, it is not possible for the triangular
lattice; as seen in Fig. 2.43 (and also in Fig. 2.12), the triangular lattice is built of
three identical sublattices (hence it is tripartite); each site is surrounded by three
neighbors that belong to one of the other two sublattices. When a magnetic moment
that can point only along one of two allowed directions (towards “up” or “down”, as
in Fig. 2.42) is assigned to each site, and when the magnetic interaction “prefers”
nearest-neighbor moments to be antiparallel, the result is a frustrated situation. The
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 75

The crystalline structure of solids 75

moments simply cannot be arranged such that all magnetic bonds are “satisfied”
(i.e., connect antiparallel moments). There are quite a number of materials in
Nature that have triangular or hexagonal symmetry and they all “suffer” from this
phenomenon. However, had the moments been allowed to point at any direction in
the plane, then there is an optimal arrangement: the magnetic moments in each of
the three sublattices point along directions that form an angle of 120◦ with each
other (see the arrows in Fig. 2.43). This arrangement still leaves a certain amount
of frustration, as the moments are not exactly antiparallel. This point is revisited
in Chapter 4.

Fig. 2.43: Dividing the triangular lattice into three sublattices; the arrows represent
the directions of the magnetic moments in each sublattice in the energetically-
optimal state for interactions that prefer an antiparallel arrangement for each pair
of nearest neighbors.

Problem 2.36.
Determine the primitive unit cell of the lattice in Fig. 2.43 and find its base.

2.11 Answers for the problems in the text

Answer 2.1.
a. The packing ratio is the total area of all circles divided by the area of the plane.
Since the lattice is periodic, this ratio is also the area of a single circle divided by
the area of the unit cell, for example, the hexagonal Wigner-Seitz cell depicted in
Fig. 2.5. Each of the hexagons there consists of six equilateral triangles. The height
of each triangle is√the radius of the circle it encloses, r, and therefore the area of
the triangle is r2 /√ 3. The hexagon
√ contains a single circle, and hence the packing
ratio is πr2 /(6r2 / 3) = π/(2 3) = 0.9069. Another way to obtain the same result:
consider the equilateral triangle created by the two lattice vectors displayed in Fig.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 76

76 Introduction to Solid-State Physics

2.5, and the √ line joining√their tips. The edge of the triangle is a = 2r, and hence
its area is a 3/4 = r2 3. The triangle contains three sectors of a circle, spanned
2

by 60◦ . The area √ of each such

2
√ sector is πr /6, and consequently the packing ratio
2 2
is 3(πr /6)/(r 3) = π/(2 3).
b. The corresponding Wigner-Seitz unit cell is a prism whose base is the hexagon
mentioned in part (a), and whose height is 2r. √ Hence the √volume of the parallelo-
2 3
gram containing a single sphere is√(2r)(6r / √3) = 12r / 3. The volume packing
ratio is therefore (4πr3 /3)/(12r3 / 3) = π/(3 3) = 0.6046.

Answer 2.2.
a. The planar Wigner-Seitz cell can be obtained as follows. One connects an
arbitrary lattice site to all its neighbors; the lines so formed are bisected at their
mid points by other lines, normal to the formers. The Wigner-Seitz cell is confined
by the cutting normal lines. In the present case it is a square; whose edge length is
the diameter of each circle, a = 2r (r is the circle radius), Fig. 2.44(a).
b. Each such square contains a single circle. Therefore, the packing ratio is
πr2 /(2r)2 = π/4 ≈ 0.7854.
c. The Wigner-Seitz cell is a cube of edge 2r, Fig. 2.44(b), and consequently the
packing ratio is (4πr3 /3)/(2r)3 = π/6 ≈ 0.5236.

(a) (b)

Fig. 2.44

Answer 2.3.
a. The length of the diagonal of the dashed unit cell is 2(r< + r> ), and hence
the length of √ the √edge of the cell, which equals the lattice constant a, is a =
2(r> + r< )/ 2 = 2(r> + r< ). The structure portrayed in Fig. 2.7(a) √ exists as
long as this edge is longer than the diameter
√ of the large circle, a = 2(r > + r< ) >
2r> , i.e., as long√as x = r< /r> > 2 − 1. In this range the packing ratio is
2 2
ρ = π(r> + r< )/[ 2(r> + r< )]2 = π(x2 + 1)/[2(x + 1)2 ]. √
b. The large√ circles are tangential to each other when a = 2(r> + r< ) = 2r> , i.e.,
when x = 2 − 1.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 77

The crystalline structure of solids 77

√
c. For x < 2 − 1, the smaller circles are located in the spaces in-between the
larger ones, and are not tangential to them. Hence, the lattice constant is still
2 2
a = 2r> , and the packing ratio is ρ = π(r> + r< )/[2r> ]2 = π(x2 + 1)/4. Figure
2.45(a) displays the dependence of the packing ratio on x. In the x = 0 limit (and
also in the x = 1 one) the system becomes a simple square lattice and therefore the
result is ρ = π/4 ≈ 0.7854, the same as the√answer to problem 2.2. The maximum,
ρ ≈ 0.92, is attained at the mid point x = 2 − 1.
d. The lattice constant is a, as calculated above.

(a) (b)

Fig. 2.45: (a) The packing ratio for answer 2.3. (b) The packing ratio for answer
2.4.

Answer 2.4.
Denote the lattice constant (the distance between the centers of two large circles)
by a. The distance between the centers of two nearest-neighbors, a small circle
and a large one, is 2/3 of the height of the triangle
√ built of the centers of three
large neighboring
√ circles, and is therefore a/ 3. The area of the rhombic unit
cell is a2 3/2, and it contains a large circle and a small one. The packing ratio
is the sum of the areas
√ of the two circles, divided by the area of the rhombus,
2 2
ρ = π(r> + r< )/[a2 3/2]. The large circles do not touch each other when 2r> < a;
it is assumed that in this case the large circle is tangential to the smaller
√ one,
and therefore the √ distance in-between them is such that r> + r< = a/ 3.√Then,
ρ = 2π(x2 + 1)/[3 3(x + 1)2 ]. The condition 2r√ > < a implies that x > 2 3/3 −
1 ≈ 0.1547. On the other hand, when x < 2 3/3 − 1, then the large √ circles
touch each other, and consequently a = 2r> . Then ρ = π(x2 + 1)/[2 3]. The
result resembles qualitatively the one obtained
√ in problem 2.3: the packing ratio is
maximal, with the value 0.9286 for x√= 2 3/3 − 1, and it attains the value of the
ordinary triangular√lattice, ρ = π/[2 3] ≈ 0.9069 when x = 0. For graphene x = 1
and then ρ = π/[3 3] ≈ 0.6046. This number is exactly two thirds of the packing
ratio of the ordinary triangular lattice, as can indeed be expected: the hexagonal
lattice is obtained from the triangular one when one third of its sites are obliterated.
The packing ratio as a function of x is depicted in Fig. 2.45(b).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 78

78 Introduction to Solid-State Physics

Answer 2.5.
a. The length
√ of the rhombus edge in Fig. 2.4 is the √ lattice constant, a, and its
height is a 3/2. Hence the area of the rhombus is a2 3/2. The hexagonal unit cell
in Fig. 2.5 consists of six equilateral
√ triangles; the height of each is a/2 and √ hence
the edge length is x = a/ 3. The area of the hexagon is thus 6ax/4 = a2 3/2.
Indeed, the two answers are identical.
b. In the case of the square lattice, the two unit cells are squares of edge a, and
therefore their areas are both a2 .
c. Denote the lattice vectors of the simple square lattice (e.g., in Fig. 2.6 or in
Fig. 2.14) by a1 and a2 . The lattice vectors of the crystal displayed in Fig. 2.7
(the dashed lines there) are then given by a1 = a1 + a2 and a2 = √ a1 − a2 . If the
original lattice constant is a then the “new” lattice constant is a 2, and the area
of the “new” unit cell is 2a2 , as it should, since it contains two lattice sites or two
“atoms”.

Answer 2.6.
This unit cell includes two identical lattice points, one located at the origin, r1 = 0,
and the other at r2 = (a1 + a2 )/2. The area of the unit cell is 2a2 , where a is
the lattice constant of the original lattice. This indeed is the specific area that
represents two lattice sites.

Answer 2.7.
Denote the lattice constant of the triangular lattice √
by a. In cartesian coordinates,
the lattice vectors are a1 = √ (a, 0) and a2 = (a, √a 3)/2. The distance between
nearest neighbors is R = a/ 3 and thus a = R 3. In the cell marked by the
solid lines in Fig. 2.8, one base point is at the origin, r1 = 0, and the other is at
r2 = (a1 + a2 )/3. For the unit cell shown in the mid panel of Fig. 2.9, one point is
located at the origin, but the other is at r2 = 2(a1 + a2 )/3.

Answer 2.8.
a. The volume is a3 . The volume packing-ratio is π/6 ≈ 0.524 (as in problem
2.2). The coordination number is 6: two nearest-neighbors along each of the three
axes.
b. The Wigner-Seitz cell is a cube, similar to the previous unit cell but shifted
such that the center of the cube is at the center of the sphere contained in it and
tangential to all its faces (as in problem 2.2).
c. Use the cartesian lattice vectors shown in Fig. 2.15(b); then another set of lattice
vectors is a1 , a2 , and (a1 + a3 ). These build a parallelogram whose volume is again
a3 , and it also represents a single lattice site. Another option, of the same volume,
is the set of vectors a1 = a1 , a2 = (a1 + a2 ), a3 = (a1 + a2 + a3 ). In this example
an arbitrary lattice site is given by R = n1 a1 + n2 (a1 + a2 ) + n3 (a1 + a2 + a3 ),
n1 , n2 , and n3 are arbitrary integers. This ensemble of points is identical to the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 79

The crystalline structure of solids 79

one obtained from Eq. (2.1), with different integral coefficients: n1 = n1 + n2 + n3 ,
n2 = n2 + n3 , n3 = n3 .

Answer 2.9.
The cube in the CsCl lattice contains eight 1/8’s of the big spheres, at the corners,
and a smaller sphere at its center. Hence, the total volume of the spheres within
3 3
the cube is 4π(r> + r< )/3, while the volume of the cube is a3 . The packing ratio
3 3
is therefore ρ = 4π(r> + r< )/(3a3 ). The discussion around √ Fig. 2.16 is confined to
the case in which the sphere at the center is small, x < ( √3 − 1). Then the lattice
constant is a = 2r> , yielding ρ = π(x3 +1)/6. When x > ( 3−1), the larger spheres
are not tangential to each other, and therefore a > 2r> . However, in this range the
sphere at the center of the cube √ is tangential to its big neighbors,
√ and therefore the
length of the cube diagonal is a 3 = 2(r> +r< ). Hence, ρ = π 3(x +1)/[2(x+1)3 )].
3

Figure 2.46 shows the packing ratio as a function of the radii ratio x. For x = 0, the
packing ratio is that of the SC lattice, ρ √ = π/6 ≈ 0.5236, and for x = 1 the packing
ratio is that of the BCC lattice, ρ = π 3/8 ≈ 0.68. The maximal √ packing ratio,
0.729, is reached when all spheres are tangential to each other, x = ( 3−1) ≈ 0.732.

Fig. 2.46

Answer 2.10.
a. The volume of this unit cell is given by Eq. (2.2),
 
 −a/2 a/2 a/2 
 
V = |a1 · [a2 × a3 ]| =  a/2 −a/2 a/2  = a3 /2 . (2.10)
 a/2 a/2 −a/2 

This answer can be also derived by realizing that the cubic unit cell includes two
lattice sites, and therefore its volume is twice that of the primitive unit cell.
b. This cell includes a single lattice site, and therefore its volume is a3 /2 and it is
a primitive unit cell. One may get this result directly: this cell is a parallelogram
whose base is a square of edge a and its height equals a/2. The lattice vectors
are not primitive, as they are not the shortest ones connecting a lattice site to its
neighbors.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 80

80 Introduction to Solid-State Physics

Answer 2.11.
a. a3 /4. This answer can be reached by two routes, either by inserting Eq. (2.6)
into Eq. (2.2), or by noting that the volume of the cubic unit cell of the FCC
lattice, which includes four lattice sites, is a3 . √
b. The cubic unit cell contains four spheres of radius r = a 2/4 (obtained from the
diagonal
√ of the cube face), and hence the volume packing ratio is 4 × (4πr3 /3)/a3 =
π 2/6 ≈ 0.740.

Answer 2.12.
One may obtain the result by examining the dashed square in Fig. 2.19(a). The
mid points of the edges of this square are each located on the mid orthogonals
of the lines connecting the vertices of this square with the origin at the center of
the cube. The square joining these mid points is built of four diagonals of squares
which are the faces √of the Wigner-Seitz cell. As the edge of the dashed square is a,
each diagonal is a/ 2, and consequently the length of the edge of the face of the
Wigner-Seitz cell is a/2. Another method is based on the rectangle marked √by the
thick dashed lines. The length of the smaller edge of this rectangle is a/ 2, and
it also equals the length of the diagonal of the face of the Wigner-Seitz cell. As a
result, the length of the edge of that face is a/2.

Answer 2.13.
The Bravais lattice of strontium titanate is a simple cubic, with a lattice constant
equal to the distance between neighboring strontium ions, a. The surroundings of
each of the strontium ions which form the lattice are identical. The surroundings of
each of the titanium ions, which also create a simple-cubic lattice, are identical as
well. On the other hand, the oxygen ions do not all share the same environment. It
is only when an oxygen ion is displaced by a lattice vector of the afore-mentioned
cubic lattice that it comes to a location with identical surrounding. Hence, the
primitive unit cell is the cube shown in Fig. 2.21(a). Upon slightly shifting the
cube along one of its diagonals (e.g., along the negative direction of all axes), it is
found that in the displaced cube there remain a single strontium ion (at the origin),
one titanium ion (at the center of the cube, much like in cesium-chloride lattice),
and three oxygen ions, at the points (a/2, a/2, 0), (a/2, 0, a/2), and (0, a/2, a/2).
Therefore, the base of the crystal contains exactly the chemical formula of the
material, SrTiO3 .

Answer 2.14.
By counting, the tetragonal cell in Fig. 2.21(b) contains twice the chemical formula
of the material: two copper ions, four lanthanums, and eight oxygens. The advan-
tage of the tetragonal cell is the right angles between its lattice vectors. Since the
cell contains a copper ion also at its center, the lattice is terms “body-centered
tetragonal”, for obvious reasons. Nonetheless, the surrounding of the copper ion
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 81

The crystalline structure of solids 81

at the center of the cell is identical to that of the copper at the low left corner;
it is therefore possible to build a primitive unit cell that contains half of the ions
appearing in the figure, i.e., the chemical formula. The primitives unit cell consists
of the two vectors (a, 0, 0) and (0, a, 0), which create the square of copper ions in the
basis of the cell, and of a third vector, (a/2, a/2, c/2), which connects the copper
ion at the center with the origin. The base of the crystal contains a copper ion at
the origin, two lanthanum ions at (a/2, a/2, c/6) and (0, 0, c/3), and four oxygen
ions at (a/2, 0, 0), (0, a/2, 0), (0, 0, c/6), and (a/2, a/2, c/3). This base generalizes
the one discussed in problem 2.10(b). The lattice vectors building the primitive √ cell
2 2 2
are the√ shortest possible ones, provided that (a + c )/4 > a , i.e., c > a 3. When
c < a 3, the shortest lattice vectors are similar to those in the left panel of Fig.
2.16 and in Eq. (2.5): a1 = (a/2)(−x̂ + ŷ) + (c/2)ẑ, a2 = (a/2)(x̂ − ŷ) + (c/2)ẑ,
a3 = (a/2)(x̂ + ŷ) − (c/2)ẑ.

Answer 2.15.
a. Consider the unit cell of the HCP lattice, indicated by the thick lines in Fig.
2.24(b). Recall that the lattice constant of the planar triangular lattice is a = 2r,
where r is the radius of√the spheres in the dense planar packing. The volume of
the unit cell is then ca2 3/2, where c is the height of the cell. The cell contains
two lattice sites; the total volume of the spheres centered on these two sites is
thus 8πr3 /3. (This can also be found by “dividing” each sphere at the corner of
the unit cell among all neighboring cells; in the lower plane and in the upper one
there are altogether four spheres “contributing” 1/6 of their volume and four others
“contributing” 1/12 of their volume. Their total contribution is then 4 × (1/6) +
4 × (1/12) = 1 of the sphere volume, to which the volume of√the “internal” sphere
√
3 2
is added.) Therefore, the packing
 ratio is ρ = (8πr /3)/(ca 3/2)
√ = 2πa/(3 3c).
In the densest lattice c = 2a 2/3 ≈ 1.633a, and then ρ = π/(3 √ 2) ≈ 0.74.
b. The packing ratio of the FCC is ρ = (16πr3 /3)/a3 = π/(3 2) ≈ 0.74, identical
to the result above. This is not surprising, since in both lattices each layer lies in
the “valleys” of its predecessor, and the density is not affected by shifting the layer
among these “valleys”.

Answer 2.16.
Since the arrangement marked by A repeats itself only after four layers, the unit cell
is a prism whose basis is the rhombic unit cell of the planar triangular
 lattice, but
its height is twice that of the simple HCP structure, i.e., c = 4a 2/3 ≈ 3.266a.
The distance between neighboring layers is determined by the condition that the
spheres are tangential to each other; it thus remains the same. Indeed, the height
of the prism is doubled, but so is also the number of spheres (four in place of two),
and the packing ratio is the same as before, ρ ≈ 0.74.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 82

82 Introduction to Solid-State Physics

Answer 2.17.
The unit cell is identical to that
√ of the HCP lattice, but the base contains two
zinc ions, at the points
√ (a/2, a 3/6, c/8) and (0, 0, 5c/8) and two sulphur ions, at
(0, 0, 0) and (a/2, a 3/6, c/2).

Answer 2.18.
This lattice is identical to the square lattice, rotated relative to the former by 45◦ .
If the lattice constant of the original
√ square lattice is a, then the lattice constant of
the square-centered lattice is a/ 2 (half the diagonal of a square of edge a).

Answer 2.19.
The base-centered tetragonal lattice consists of layers of square-centered lattices
placed one on top of the other. As seen in problem 2.18, the square-centered lattice
is identical
√ to a square lattice rotated by 45◦ whose edge length is smaller by a factor
of 2. Hence, the based-centered tetragonal lattice is identical to the tetragonal
lattice. Quite similarly, the face-centered tetragonal lattice is identical to a body-
centered tetragonal lattice also rotated by 45◦ . The lattice sites at the centers of
the faces become the sites at the centers of the body-centered tetragonal lattice.

Answer 2.20.
In both cases a1 = a2 = a3 and α = β = γ, and consequently the two primitive
unit cells construct trigonal lattices. The angle in-between the two lattice vectors
can be derived by using the relation a1 · a2 = a1 a2 cos γ. Equations (2.5) and (2.6)
thus yield α = β = γ = 109.47◦ for the BCC lattice, and α = β = γ = 60◦ for
the FCC one. This latter result can be directly observed in Fig. 2.18: each face of
the primitive unit cell is a rhombus comprising two equilateral triangles; the length
of the shorter diagonal equals the length of the triangles’ edge (and the distance
between nearest-neighbor lattice sites).

Answer 2.21.
Equation (2.1) which defines a Bravais lattice, gives the location of a lattice site as
R = n1 a1 + n2 a2 + n3 a3 , where the coefficients are arbitrary integers. Inversion
through the origin transforms this point into −R = −n1 a1 − n2 a2 − n3 a3 . The
coefficients here are integers as well, and hence this point is also a lattice point.

Answer 2.22.
a. Denote a rotation axis of order n by An , a mirror plane by m, and an inversion
center by i. The tetragonal lattice has one rotation axis of order 4, normal to the
square face of its unit cell and threading it at the center, and four axes of order
2, perpendicular to the rectangular faces of the unit cell, again passing through
their centers. In addition, this lattice possesses three mirror planes, parallel to the
faces and passing through the mid points between them, and two mirror planes
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 83

The crystalline structure of solids 83

connecting the diagonals of opposite faces. Finally, this lattice has an inversion
center at the center of the unit cell. Thus, the symmetry group of the tetragonal
lattice includes the elements i, 1A4 , 4A2 , 5m.
b. The square basis of the unit cell is transformed into a rectangular one whose
edges are not all equal. Therefore the rotation of order 4 and the two mirror planes
passing through the diagonals of this basis are eliminated from the group.
c. In this case all faces are squares, and consequently the mirror planes through
the diagonals of those faces have to be added to the group, and also all rotations
around axes passing through the mid points of the faces are of order 4. In addition,
there are now the rotations A3 around each of the cube diagonals. Altogether, the
symmetry group of the cubic lattice contains three axes of order 4, six axes of order
2, four axes of order 3, nine mirror planes, and one inversion center.

Answer 2.23.
a. The general oblique lattice is displayed in Fig. 2.47(a), for the case where
|a2 | = a2 is much longer than |a1 | = a1 . The lattice then comprises separate
parallel rows of circles, tangential to each other within each row, but not touching
the circles in the neighboring rows. The area of the unit cell is a1 a2 sin γ, and the
area of each circle is πr2 = π(a1 /2)2 (r = a1 /2 is the radius of each circle). The
packing ratio is thus ρ = [πa21 /4]/[a1 a2 sin γ] = πa1 /(4a2 sin γ). When a2 < a1 , one
needs just to interchange a1 and a2 .

(a) (b)
Fig. 2.47

The figure shows that the packing ratio can be increased by pushing the rows
closer to each other. When 60◦ ≤ γ ≤ 120◦ , one may decrease a2 until the circles
in different rows touch each other, like in Fig. 2.47(b). In this case, a1 = a2 =
2r, and the packing ratio of the oblique oblique lattice for angles in this range
is ρ = π/(4 sin γ). In particular, this result reproduces the packing ratio of the
square lattice, ρ = π/4, for
√ γ = π/2, and the packing ratio of the triangular lattice,
ρ = π/(4 sin 60◦ ) = π/[2 3] when γ = π/3. One may increase the packing ratio
by decreasing sin γ. However, the figure shows that γ cannot be reduced below
γ = π/3 = 60◦ , because in that case the circles in two neighboring rows overlap.
For the same reason, γ cannot be increased beyond γ = 2π/3 = 120◦ . It follows
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 84

84 Introduction to Solid-State Physics

that the maximal packing ratio of the oblique lattice with angles in the range
60◦ ≤ γ ≤ 120◦ is achieved in these two limits, γ = π/3 = 60◦ or γ = 2π/3 = 120◦ .
In both cases there emerges the triangular √ lattice displayed in Fig. 2.48(c), with
the packing ratio ρ = π/(4 sin 60◦ ) = π/[2 3].

(c) (d)
Fig. 2.48

When the angle γ is in the range γ < 60◦ , a2 cannot be made equal to a1 because
then again the circles in neighboring rows overlap. The optimal configuration of this
case is shown in Fig. 2.48(d). The vector a2 is now the basis of an isosceles triangle
(with one of the equal-length sides being the vector a1 ), and hence a2 = 2a1 cos γ,
leading to ρ = π/(4 sin 2γ). This result pertains only to the range 30◦ ≤ γ < 60◦ ,
because when γ reaches the value 30◦ the lattice becomes the one shown in Fig.
2.47(e), and smaller values of γ again cause overlapping of circles. In this range of
γ values, the largest packing ratio is achieved at the ends of the range, where the
lattice becomes a triangular one. The only difference is that in Fig. 2.49(e) the
unit cell is not the primitive one: the primitive unit cell of the triangular lattice is
portrayed in Fig. 2.48(c).

(e) (f)
Fig. 2.49

The lattice in Fig. 2.48(d) can be described by a primitive unit cell like the one
shown in Fig. 2.49(f) (right side). This unit cell is a reflection image of the cell in
Fig. 2.47(b); the results obtained from that figure apply to this case as well.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 85

The crystalline structure of solids 85

Answer 2.24.
a. A rotation axis of order 2 normal to the lattice and passing through any lattice
site and through any mid point between nearest neighbors, a reflection through a
mirror plane placed normal to the lattice at these points, and an inversion with
respect to any of those points.
b. A rotation axis of order 2 normal to the lattice and passing through any lattice
point, a reflection through a mirror perpendicular to the lattice at any lattice point,
and an inversion with respect to any lattice point.
c. No symmetry operations except for translations.
d. A rotation axis of order 2 threading the lattice perpendicularly through a mid
point between two sites, a reflection through a mirror normal to the lattice at the
mid point between two neighboring sites, and an inversion with respect to any of
the mid points between two lattice sites.
e. A reflection through a mirror placed at any lattice site normal to the lattice.
f. A rotation of order 2 around an axis normal to the lattice plane and passing
through the points A or C, a reflection by a mirror normal to the lattice placed at
the points B or D, and an inversion with respect to the points A or C.

Answer 2.25.
Most of the symmetry operations do not exist. The one that remains in both the
triangular and the square lattices is a reflection through a mirror perpendicular to
the lattice plane passing through the vector (a1 + a2 ).

Answer 2.26.
When the two ions in the unit cell portrayed in the first panel Fig. 2.2 have opposite
signs, their displacement relative to the structure shown in the first panel of Fig.
2.1 creates a dipole moment, directed along the direction of the lattice. This dipole
moment is shown by the arrow in Fig. 2.50(a). A similar situation occurs in the
crystal displayed in the second panel of Fig. 2.2, where an electric dipole moment
appears along the direction of the relative motion of the ions, see Fig. 2.50(b).

(a) (b)

Fig. 2.50: (a) The electric dipole moment for the lattice in the first panel of Fig.
2.2; (b) The electric dipole moment for the lattice in the second panel there.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 86

86 Introduction to Solid-State Physics

Answer 2.27.
a. Divide Eq. (2.8) by Fn−1 , to obtain tn = 1 + 1/tn−1 . Assuming that the ratio
tn does approach a constant limit, tn → τ , the equation becomes τ = 1 + 1/τ , and
the solution is indeed the golden ratio. Subtracting the two equations from one
another, the result is |tn − τ | = |tn−1 − τ |/(tn−1 τ ). As the denominator on the right
hand-side is larger than 1, the difference |tn − τ | decreases at each generation and
eventually approaches zero.
b. Denote the number of long and short segments in the nth generation by NL (n)
and NS (n), respectively. The Fibonacci rule, mapping S onto L and L onto LS,
yields the relations NS (n + 1) = NL (n) and NL (n + 1) = NL (n) + NS (n). Dividing
these two equations by one another gives x(n + 1) = 1 + 1/x(n), where x(n) =
NL (n)/NS (n). In the n → ∞ limit, corresponding to the infinite lattice, both x(n)
and x(n + 1) equal x(∞), and consequently the equation of part (a) is reproduced,
with the solution x(∞) = τ .

Answer 2.28.
Assuming that the thick line passes through the lattice point (0,0), as in Fig. 2.51,
its equation is y = x/τ , and the equations of the dotted lines are y = (x/τ ) ± y0 ,
with 2y0 = w/ cos α being the projected distance between these two lines on the
ŷ−axis. For the choice w = a(cos α + sin α), y0 = a(1 + τ )/(2τ ) = aτ /2. Denoting
an arbitrary lattice site located in-between the two dotted lines by rn = m1 (n)a1 +
m2 (n)a2 , the n + 1 successive lattice site along the ŷ−axis has m2 (n + 1) = m2 (n)
and m1 (n+1) = m1 (n)+1, and the one along the x̂−axis has m2 (n+1) = m2 (n)+1
and m1 (n + 1) = m1 (n). In both options, m1 (n + 1) + m2 (n + 1) = m1 (n) + m2 (n) +
1, so that one may write m1 (n) + m2 (n) = n. The lattice point rn is indeed
between the two dotted lines provided that m1 a/τ − y0 < m2 a < m1 a/τ + y0 and
(m2 a − y0 )τ < m1 a < (m2 a + y0 )τ . These two inequalities are identical; since
τ = 1 + 1/τ , they yield m1 − 1/2 < n/τ < m1 + 1/2, leading to m1 = ||n/τ ||.
Hence, xn = m1 (n)L+m2 (n)S= nS+m1 (n)(L-S)=S[n + ||n/τ ||/τ ]. Alternatively,
m1 < n/τ + 1/2 < m1 + 1, and consequently m1 = [n/τ + 1/2], where [y] is the
integral part of the number y. In order to prove that this is indeed a Fibonacci
sequence, one exploits an inflation transformation. Figure 2.51(a) shows Fig. 2.33
with an additional band, in-between the solid (red) lines, whose width is w/τ . The
new band does not include any of the vertical steps of the original line, whose
projections gave the segments of length S. The lattice sites within this new band
give rise solely to projections of length L’=L+S or S’=L and the new sequence
(starting at the left edge) is L’S’L’S’L’L’S’L’L’S’, which is also a Fibonacci lattice,
with a length scale larger by τ .

Answer 2.29.
a. With the choice of the origin as given in the problem, the equation of the thick
line is y = xtanα. All lattice sites are either x = n1 a, or y = n2 a. It follows that
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 87

The crystalline structure of solids 87

(a) (b)

Fig. 2.51

the ratio x/y = n1 /n2 is rational; it never becomes tan(α) = 1/τ , and therefore
there are no more lattice sites on this line.
b. When the slope is a rational number, the thick line can pass through a lattice
site; this line then passes also through all lattice sites obeying n1 /n2 = p/q. Figure
2.51(b) shows the result for tan(α) = 2/3. As seen, the square-lattice sites that lie

the thick line are projected onto a periodic lattice, with a lattice constant a =
on
2 2
a p + q . For the band between the two dashed lines, each such unit cell contains
also the projections of two additional points, at x1 = a[2 ± sin(α) cos(α)]/[2 sin(α)];
the distances on the thick line are AB = 2a/ sin(α), CD = a cos(α). This lattice
can be described by the periodic sequence LSLLSLLSLLS. More complicated bases
are possible for other widths of the band.
c. Since 1.6=8/5, 1.61=161/100, and √ 1.618=809/500, √ the lattice constants of the
corresponding periodic lattices are 82 + 52 ≈ 9.43, 1612 + 1002 ≈ 189.5, and
√
809 + 500 ≈ 951, all in units of the square-lattice constant. Each unit cell
2 2

represents a base comprising an increasing number of lattice sites, with irrational

coordinates within the cell. In the infinite limit, the unit cell is identical to the
whole lattice.

Answer 2.30.
a. An example for such a basis appears in Fig. 2.52(a), which shows five equal-
length vectors, with a 72◦ −angle in-between them. Each of the edges in Fig. 2.35 is
parallel to one of these five vectors: when the origin is placed at the locus of the five
central rhombi in Fig. 2.52(b), their vertices are reached by these five vectors. Any
edge of the bright rhombus is parallel to an edge of a dark rhombus, and therefore
it is also equal to one of the proposed lattice vectors.
b. The d–dimensional cubic lattice can be described by a cartesian basis, in which
the nth lattice vector is a vector of d components, all zero except for the nth one,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 88

88 Introduction to Solid-State Physics

which is 1. In five dimensions these vectors are

a1 = (1, 0, 0, 0, 0) , a2 = (0, 1, 0, 0, 0) , a3 = (0, 0, 1, 0, 0) ,

a4 = (0, 0, 0, 1, 0) , a5 = (0, 0, 0, 0, 1) .
√
One of the diagonals of the five-dimensional cube is along û = (1, 1, 1, 1, 1)/ 5.
The five lattice
√ vectors are oriented at equal angles relative to this vector, i.e.,
an · û = 1/ 5, and therefore they are arranged symmetrically around this diagonal.
A rotation of order 5 brings these vectors back to themselves, a1 → a2 , a2 → a3 , etc.
(this is similar to the symmetry of the three-dimensional cubic lattice to rotations
by 120◦ around the diagonal of the cube). It follows that û is a rotation axis
of order 5 for the five-dimensional cubic lattice. Upon observing these five vectors
from the direction of û, one notes the equal angles between two neighboring vectors.
Therefore their projections on a plane normal to û appears as in Fig. 2.52(a).

(a) (b)

Fig. 2.52

Answer 2.31.
Assume that the joint unit cell contains m1 circles on the substrate, and m2 ones
in the layer above. The length of the unit cell is hence 2m1 r1 = 2m2 r2 (2r1 and
2r2 are the respective diameters). It follows that m1 /m2 = r2 /r1 = p/q, and the
smallest values obeying this relation are m1 = p and m2 = q. The length of the
unit cell is hence 2pr1 = 2qr2 . The base contains p circles on the substrate and q
ones in the layer above.

Answer 2.32.
The adsorbed atoms create a triangular lattice, with a rhombic unit cell whose edge
is of length 3a, where a is the distance between nearest neighbors in the graphite
substrate, see Fig. 2.53. Shifting slightly this rhombus down to the left shows that
it includes a single krypton atom (at the lower left corner) and also six carbon
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 89

The crystalline structure of solids 89

atoms, encompassing the inner hexagon. Thus the base comprises a krypton atom
and six carbon ones.

Fig. 2.53

Answer 2.33.
The tetragonal unit cell of the structure in Fig. 2.39 contains two cubic-perovskite
cells [Fig. 2.21(a)], one above the other. Along the direction normal to the layers,
this array includes a SrO plane, a MnO2 plane, a LaO plane, and another MnO2
plane, altogether a single strontium ion, a lanthanum ion, two manganese and six
oxygen ions (see Fig. 2.54). When x = 1/3, there should be two layers of LaO for
each layer of SrO, and hence the tetragonal unit cell contains a Sr ion, two La ones,
three Mn ions and nine oxygen ions.

Fig. 2.54

Answer 2.34.
The unit cells of the structures C and D contain two lattice sites, while there are five
lattice sites in the unit cell of E. The symmetry group of C includes rotations of order
2 around axes normal to the lattice and passing through the mid points between
nearest neighbors, and inversions through these points. It also includes reflections
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 90

90 Introduction to Solid-State Physics

through mirrors normal to the lattice and passing through the lattice points. The
symmetry group of D includes only the latter reflections. The symmetry group of
E includes solely reflections through mirrors perpendicular to the lattice at the mid
points between two downwards-oriented moments, as displayed in Fig. 2.55.

Fig. 2.55

Answer 2.35.
The magnetic moment at the origin R = 0 is +μ0 ; its sign alternates upon prop-
agating along each axis. Thus, the magnetic lattice resembles that of table salt:
the “chloride” sites represent the moment +μ0 , and the “sodium” ones the moment
−μ0 . It follows that the base of the magnetic lattice contains two sites, one with
the moment +μ0 , the other with −μ0 (like the pair in the ellipse of Fig. 2.17), and
the lattice is an FCC one. As opposed to the lattice of table salt, where the two
sublattices consist of different ions and therefore no symmetry operations can inter
relate them, in the antiferromagnet case one may map one sublattice on the other
by a combination of adequate symmetry operations. For instance, a reflection by a
mirror passing through the mid points of the cube’s faces. When Q = (1, 1, 1)(τ /a)
and R = (n1 , n2 , n3 )a, then μ(R) = μ0 cos[τ (n1 + n2 + n3 )]. The moment at the
origin is still +μ0 , but there is no other lattice point with this moment, as no integer
product of τ can be an (even) integer product of π; the ratio τ /π is irrational. Thus
the magnetic order in this case is not commensurate with the structural lattice, and
the magnetic unit cell is infinitely large.

Answer 2.36.
The unit cell is the same rhombus that has been the base for the triangular lattice,
Fig. 2.4, or for the hexagonal lattice, Fig. 2.8. However, in the present case it
contains a base of three sites, each of them belonging to a different sublattice–
see Fig. 2.12. Exploiting this unit cell, the base includes the sites at r1 = 0,
r2 = (a1 + a2 )/3, r3 = 2(a1 + a2 )/3.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 91

The crystalline structure of solids 91

2.12 Problems for self-evaluation

s.2.1.
Draw the Wigner-Seitz cells belonging to each of the lattices in Fig. 2.27.

s.2.2.
Figure 2.56 displays two lattices called Kagomé (right panel) and Cairo (left panel).
The first is built of equilateral triangles; the second consists of pentagons, each
comprising four identical edges and a shorter one, with three 120◦ −angles and two
right ones. Identify the corresponding primitive unit cells and the lattice vectors.
Write down the radius vectors of the points that belong to the base when the unit
cell includes more than a single lattice site (a junction in the figures).

Fig. 2.56

s.2.3.
Determine which of the structures in the following list are Bravais lattices. Indicate
the lattice vectors of those which are Bravais lattices, and describe the lattice as a
Bravais one with a base for those that are not; write down the lattice vectors and
the coordinates of the points in the base within the primitive unit cell.
a. A square lattice with an additional lattice site at the mid point of each edge.
b. A simple cubic lattice with additional lattice points at the centers of the bases
(the horizontal faces).
c. A simple cubic lattice with additional lattice points at the centers of the four
vertical faces.
d. A simple cubic lattice with additional points at the mid point of each edge.

s.2.4.
Which Bravais lattice is created of the points (n1 , n2 , n3 ) in cartesian coordinates
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 92

92 Introduction to Solid-State Physics

when:
a. ni are all even, or are all odd;
b. (n1 + n2 + n3 ) is always even;
c. (n1 + n2 + n3 ) is always odd.

s.2.5.
Figure 2.19 portrays the Wigner-Seitz cells of the BCC and FCC lattices. Prove
that the four-edge polygons on the faces of the Wigner-Seitz cell of the BCC are
all
√ squares, that the hexagons are all equilateral, and that the length of each edge is
a 2/4, where a is the edge length of the cubic cell.

s.2.6.
a. Sodium crosses from the BCC structure to a perfect HCP one at 23◦ K. As-
suming that the density does not change, find the new lattice constant [in units of
the BCC one, a(BCC) = 4.23Å].
b. Iron has the FCC structure at temperatures above 900◦ C, and a BCC one at
lower temperatures. Assuming that the two structures are built of spheres of equal
radii adjacent to each other, find the ratio of the densities of the two structures.
Given that the BCC iron density is 7900 Kg/m3 , determine the lattice constant
(the atomic mass of iron is 55.85, and the mass of the proton is 1.6726×10−27 Kg.)

s.2.7.
a. Find the packing ratio of a crystal of the table-salt type, in which the larger
“chloride” ions are adjacent to each other and the radii ratio between the “sodium”
and “chloride” ions is x. Determine the relevant range of the x values.
b. Find the packing ratio when the ions of the two types touch each other, but the
“chloride” ions do not.
c. Provide a drawing of the dependence of the packing ratio on the radii ratio.

s.2.8.
a. Find the packing ratio of a zinc-blende lattice, in which the bigger “sulphur”
ions are contiguous, and the ratio of the radii of the “zinc” and “sulphur” ions is
x. What is the relevant range of x values?
b. Find the packing ratio when the ions of the two types touch each other, but the
“sulphur” ions do not.
c. Provide a drawing of the dependence of the packing ratio on the radii ratio.
Where is the diamond in this figure? Compare the results with those obtained for
structures of the table-salt type and the CsCl one. Which packing is denser for
various values of x?
d. The distance between the center of the sulphur ion and the center of the zinc one
in zinc-blende is 0.234nm. Find the density of this material (the atomic weights of
zinc and sulphur are 65.38 and 32.064, respectively, and the mass of the proton is
1.6726×10−27 Kg).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 93

The crystalline structure of solids 93

s.2.9.
Figure 2.57 displays the double-perovskite structure, A2 BB’O6 , in which half of the
B ions in the original lattice [Fig. 2.21(a)] are alternatively replaced by other ions
B’ (the figure shows Sr2 MoFeO6 ]. The left part shows only the ions B and B’
(filled and empty circles, respectively), and the right depicts the octahedra of the
oxygens encapsulating each one of them. Determine the lattice corresponding to
this structure. Find the primitive unit cell and the base.

Fig. 2.57

s.2.10.
a. Determine the primitive unit cell of the graphite lattice displayed in Fig. 2.9;
find the lattice vectors of this cell and its base. Determine the coordinates of the
atoms that belong to it (in terms of the lattice vectors).
b. The planes created in Fig. 2.12 once the atoms of type A, B, or C are eliminated
are denoted by the letters A, B, and C, respectively. The discussion following Fig.
2.9 shows that the graphite lattice is described by the arrangement BCBCBC. . . ,
where C is the configuration shown in Fig. 2.8(b) and B is the one in the mid panel
of Fig. 2.9. Determine the same quantities as in part (a) for the configuration
ABCABCABC. . . of the graphene layers.

s.2.11.
Referring to Fig. 2.28, express the volume of the primitive unit cell of the following
lattices, in terms of the three lattice constants and the angles in-between them.
a. face-centered orthorhombic;
b. triclinic;
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 94

94 Introduction to Solid-State Physics

c. monoclinic;
d. trigonal.

s.2.12.
Theparameters of a three-dimensional crystal are a1 = a2 = a, a3 = c = a[1 +
Δ( 8/3 − 1)], α = β = 90◦ , γ = 30◦ (3 − Δ). The base contains two spheres, at the
origin and at r2 = [(1 − Δ/3)(a1 + a2 ) + a3 ]/2. The two spheres touch one another.
Draw the packing ratio of this crystal as a function of Δ for 0 ≤ Δ ≤ 1, and find
the maximal packing ratio. Which lattice structures correspond to the two extreme
values, Δ = 0 and Δ = 1?

s.2.13.
Determine the structure that is created by repeating the projection
√ procedure de-
scribed in Fig. 2.32, with a line whose slope is tan(α) = 1/ 2.

s.2.14.
Figure 2.58 illustrates the adsorption of spherical “atoms” on two planar cuts of the
FCC lattice of iron, so that the adsorbed atoms are located in the dips formed among
the atoms of the substrate. The small dashed circles represent the iron atoms, and
the full large ones the adsorbed atoms. Figures A and B show the adsorption on a
plane passing through the face of the cubic lattice shown in the mid panel of Fig.
2.17 (e.g., the upper face), while figures C and D refer to the adsorption on a plane
passing through the diagonals of two adjacent faces, originated at the same lattice
site [e.g., the green points in the lower part of Fig. 2.24(a)]. The plane is normal
to the principal diagonal of the cubic unit cell.
a. Determine the two-dimensional lattice of each of these planes (before the adsorp-
tion takes place), and the corresponding primitive lattice constant (in terms of the
original lattice constant of the FCC, a).
b. Determine the primitive unit cells of each of the lattices displayed in Fig. 2.58,
comprising the substrate and the adsorbed planes, the lattice constants and the base.
c. Find the maximal radius of the adsorbed atoms in each of the figures, which is
compatible with the structure shown there.
d. For which values of the radius of the adsorbed atoms are all the dips in the
substrate occupied? In that case, what are the answers to part (b)?

s.2.15.
Graphene can be described as a hexagonal lattice (Fig. 1.6) of adjacent spheres
of radius r0 . Assuming that spheres of radius r1 are adsorbed on that plane, that
are attracted stronger to the substrate atoms than to each other, find the densest
packing possible for the adsorbed atoms (within a single planar layer), when the
ratio x = r1 /r0 is between 1 and 3.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 95

The crystalline structure of solids 95

Fig. 2.58

s.2.16.
Figure 2.59 displays an antiferromagnetic arrangement of magnetic moments on a
square lattice. Determine the magnetic primitive unit cell of this lattice, and its
base.

Fig. 2.59
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 96

96 Introduction to Solid-State Physics

s.2.17.
Figure 2.21(b) displays the tetragonal structure of lanthanum cuprate, which exists
at high enough temperatures. Figure 2.60 exhibits a single plane of CuO2 in this
lattice. The big points represent the copper ions, and the small ones the oxygen ions.
When the temperature is reduced, the octahedra encompassing each of the copper ions
(in the three-dimensional lattice) rotate slightly, by the small angle ϕ: the octahedra
around the cooper ions connected by the solid line in Fig. 2.60 rotate clockwise
around that line, while the others rotate counterclockwise around the dashed lines.
As a result, the oxygen ions marked by the horizontal lines move below the plane,
and those marked by a vertical line shift to new locations above it. The octahedra
preserve their original shape, and therefore the distances among the oxygens are not
changed. The final result is a squeezing of the lattice along the normal to the rotation
axes, like in an accordion; it yields a transition between the tetragonal phase (which
exists at high temperatures) and another, new lower-symmetry phase.
a. Determine the primitive unit cell of the planar lattice in Fig. 2.60, and its base
at the higher temperatures, before the octahedra rotations take place.
b. What is the structure of the planar lattice after the phase transition, what are
the lattice constants, the primitive unit cell, and the base?
c. At an even lower temperature there is another phase transition, where the copper
ions arrange themselves in an antiferromagnetic order. All ions connected by solid
lines in Fig. 2.60 attain a positive magnetic moment +1, while those connected by
dashed lines attain a negative magnetic moment −1 (in the suitable units). Deter-
mine the planar magnetic unit cell.
d. Each of the copper oxide planes in Fig. 2.21(b) undergoes the above-mentioned
phase transitions, with the rotation axes and the magnetic moments displaced in
parallel between the planes. The copper ions in the middle plane of the tetragonal
cell in Fig. 2.21(b) remain above the center of the rhombus created in the plane
below due to the squeezing. Ignoring possible deformations of the lattice along the
normal to the basis, what is the three-dimensional lattice structure, the unit cell,
and its base at low temperatures?

Fig. 2.60
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 97

The crystalline structure of solids 97

2.13 Answers for the self-evaluation problems

Answer s.2.1.
The Wigner-Seitz cell of the triangular lattice is displayed in Fig. 2.5. The Wigner-
Seitz cells of the square lattice and the rectangular lattice are those composed of
their lattice vectors, with the origin displaced into the center of the original cell, see
problem 2.2. The Wigner-Seitz cell of the oblique lattice is presented in Fig. 2.61,
in thick lines. Each such line is normal to the dashed line connecting the site at the
center with one of its neighbors, bisecting it at the middle.

Fig. 2.61

Answer s.2.2.
The Kagomé (the Japanese term for artisan paper cutting) lattice is triangular,
with a base that contains two triangles, see the rhombus marked by thick lines in
the right panel of Fig. 2.62. The environments of the corners of the unit cell are
identical; hence it is the primitive unit cell. Shifting slightly the rhombus downwards
to the left leaves three lattice sites within the cell–these form the base of this lattice.
These points are the origin, a2 /2, and (a1 + a2 )/2, where the lattice vectors are the
two edges of the unit cell.
The Cairo lattice appears in several versions in which the angle between the longer
edge and the horizontal axis is α or (90◦ − α). The one illustrated in the figure
has α = 30◦ . Assuming that the length of the longer edge is 1, the cartesian
coordinates of all lattice points can be expressed in terms of α. As seen in the
left panel of Fig. 2.62, the unit cell of the Cairo lattice is a square (marked by
thick lines), independent of α: the surroundings of all corners of the square are
identical. The base includes the six lattice points as marked in the figure (and can
be seen by slightly displacing the unit cell towards the left). The lattice vectors are
a1 = 2 cos(α)(x̂ + ŷ) and a2 = 2 cos(α)(x̂ − ŷ), and the base points are the origin,
and the points xn1 a1 + xn2 a2 , with
(1 + w, 1 − w) (1 + w, 1 − w) (w, 1) (1, 2 − w) 3 + w, 3 − w)
(xn1 , xn2 ) = , , , , ,
4 2 2 2 4
where w = tan(α).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 98

98 Introduction to Solid-State Physics

Fig. 2.62

Answer s.2.3.
a. The environments of the two types of sites (the original ones and those at the
centers of the edges) are not identical; this is not a Bravais lattice. The primitive
unit cell is identical to that of the original square lattice, and the base consists of
three points, for instance, those encircled in Fig. 2.63(a). The lattice vectors are
a1 = (a, 0) and a2 = (0, a), where a is the lattice constant of the square lattice.
The base comprises the points at (0, 0), (a/2, 0), and (0, a/2).
b. This is a bravais lattice. Figure 2.63(b) shows the unit cell of the tetragonal
lattice which corresponds to this case. Denoting the cubic lattice constant by a, the
lattice vectors are a1 = (a/2, −a/2, 0), a2 = (a/2, a/2, 0), and a3 = (0, 0, a).
c. This is not a Bravais lattice. The surroundings of the points at the centers of the
vertical faces are not the same as those of the corners of the cube. The primitive
unit cell is the unit cell of the original cubic lattice, and the base includes three
points, e.g., (0, 0, 0), (a/2, 0, a/2), and (0, a/2, a/2).
d. This one is also not a Bravais lattice. The primitive cell is the original cubic one,
and the base includes four points, (0, 0, 0), (a/2, 0, 0), (0, a/2, 0), and (0, 0, a/2).

(a) (b)

Fig. 2.63
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 99

The crystalline structure of solids 99

Answer s.2.4.
a. In the first case one uses ni = 2ni and obtains Ri = 2(n1 , n2 , n3 ); this is
a simple cubic lattice with a lattice constant of length 2a, where a is the lattice
constant of the original lattice. In the second case one denotes ni = 2ni + 1 and
obtains Ri = (1, 1, 1) + 2(n1 , n2 , n3 ). This again is a simple cubic lattice, with the
same lattice constant 2a, whose origin is at the point (1,1,1).
b. n1 + n2 + n3 is even when all three ni are even, or when two of them are odd and
the third one is even. The first option corresponds to the first part of (a) (SC whose
lattice constant is 2a). The second yields three lattices characterized by the points
  
Rm m 1 2 3
i = R + 2(n1 , n2 , n3 ), where R = (1, 1, 0), R = (1, 0, 1), and R = (0, 1, 1).
These points are the centers of the faces of the unit cell. When all types of points
are connected together, the result is an FCC lattice.
c. The requirement that n1 + n2 + n3 is odd is equivalent to the requirement that
this sum is even, provided that the vector (1, 1, 1) is added to Rm . This is again an
FCC lattice, shifted relative to the one in (b).

Answer s.2.5.
The left panel in Fig. 2.64 shows the rectangle created by connecting together two
opposite edges in the cube. It is also displayed by the thick lines on the cubic unit
√
cell of the reciprocal lattice in the right panel. Its edges are AG=a, and AF=a√ 2.
The diagonal of this rectangle is the diagonal of the cube, and hence AO=a 3/2.
The line BC is normal to AO (passing through its mid point), which connects the
two lattice points A and O, and therefore it lies on the face of the Wigner-Seitz
cell. This face, in turn, touches the face of the cube at point C. From the similar-

ity between the√triangles ABC and AOD it √ follows that AB/AC=AD/AO= 2/3.
Hence, AC=3a 2/8, and CD=AD-AC=a 2/8. The latter is equal to half of the
edge of the square face of the cell, which lies on the face. Repeating the above con-
siderations for all planes joining two opposite edges of the cube shows that all faces
are indeed squared, and completes the proof. This also ensures that all hexagons
are equilateral.

Fig. 2.64
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 100

100 Introduction to Solid-State Physics

Answer s.2.6.
a. Each cell in the BCC contains two atoms, and consequently the atomic density
(number of atoms per unit volume) is ρ(BCC) = 2/[a(BCC)]3 . The solution to
problem 2.15 states
√ that the√ volume of 3the unit cell of the HCP, which is denser,
is V (HCP ) = a2 c 3/2 = √ 2[a(HCP )] . This cell contains two atoms and hence
ρ(HCP ) = 2/V (HCP ) = 2/[a(HCP )]3 . Comparing the two densities yields
a(HCP ) = a(BCC) × 2−1/6 ≈ 3.77Å.
b. The discussion following Eq. (2.5) implies that if the √ atom radius is r, then
the edge of the cubic cell of the BCC is a(BCC) = 4r/ 3. As each √ cell contains
two atoms, the atomic density is ρ(BCC)√ = 2/[a(BCC)]3 = 3 3/(32r3 ). The
solution to problem 2.11 gives a(F CC) =√2 2r, and thus the atomic density for the
FCC√is ρ(F CC) = 4/[a(F CC)]3 = 1/(4 2r3 ). Consequently, ρ(F CC)/ρ(BCC) =
8/(3 6) ≈ 1.08866. The mass of the iron atom is the proton’s mass times the
atomic mass of iron, i.e., M (F e) = 55.85 × 1.6726 × 10−27 Kg= 9.34 × 10−26 Kg.
It follows that ρ(BCC) = 7900(Kg/m3 )/M (F e) = 8.46 × 1028 /m3 ; the relation
ρ(BCC) = 2/[a(BCC)]3 implies a(BCC) = [2/ρ(BCC)]1/3 = 0.287 nm=2.87Å.

Answer s.2.7. √
a. Consider first the x = 0 limit. For the dense FCC lattice, a = 2 2r> , as in
the solution to problem 2.11. This relation persists as long as the small ion√enters
in-between the bigger ones on the cube edge, i.e., a ≥ 2(r> + r< ), or x ≤ 2 − 1.
In this range the packing ratio is
√
ρ = 4 × (4π/3)(r> 3 3
+ r< )/a3 = π(x3 + 1)/(3 2) .
b. When the two ions touch√ one another then √ a = 2(r< + r> ), while the diagonal
on the face is such that a 2 ≥ 4r> , i.e., x ≥ 2 − 1. In this range
ρ = 4 × (4π/3)(r>
3 3
+ r< )/a3 = 2π(x3 + 1)/[3(x + 1)3 ] .
c. Figure 2.65 shows the result. The packing ratio varies between the one of the
FCC, ρ ≈ 0.74, for x = 0, and that of SC, ρ ≈ 0.52 for x = 1. The packing
ratio√is maximal, ρ ≈ 0.793, when all ions are contiguous, i.e., when the ratio is
x = 2 − 1 ≈ 0.414.

Fig. 2.65
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 101

The crystalline structure of solids 101

Answer s.2.8. √
a. The value x = 0 corresponds to the dense FCC lattice, for which a = 2 2r> as
in the solution to problem 2.11. This relation persists as long as the small ion fits
into the tetrahedron shown in Fig. 2.20. As the distance between the centers of two
neighboring ions is√a quarter of the diagonal of√the cube, the bigger ions touch each
other as long as a 3/4 ≥ (r> + r< ), or x ≤ 6/2 − 1. In this range the packing
ratio is
√
ρ = 4 × (4π/3)(r<3
+ r>3
)/a3 = π(x3 + 1)/(3 2) .
√
b. When the ions touch one√ another, then a √3/4 = (r< + r> ), while the diagonal
on the face is such that a 2 ≥ 4r> , i.e., x ≥ 6/2 − 1. In this range
√
ρ = 4 × (4π/3)(r<3
+ r>3
)/a3 = π 3(x3 + 1)/[4(x + 1)3 ] .
c. Figure √ 2.66(a) displays the packing ratio. Diamond corresponds to x = 1 and
has ρ = π 3/16 ≈ 0.34. The √ maximal packing ratio, ρ ≈ 0.749, is reached when
all ions are contiguous, x = 6/2 − 1 ≈ 0.225. Figure 2.66(b) displays the packing
ratios of the three types of crystals. When x = 1 the cesium chloride crystal
is
√ the densest. As x is reduced, the√table-salt type is the densest in the range
6/2 − 1 ≤ x ≤ 41/3 − 1. When x ≤ 6/2 − 1, table salt and zinc-blende have the
same packing ratios.
d.√ From the results of part (b) for the distance between neighboring ions:
a 3/4 = (r< + r> ) = 0.234 nm. Hence, a = 0.540 nm, and the density is
4[m(Zn) + m(S)]/a3 = 4(65.38 + 32.064) × 1.6726 × 10−27 Kg/a3 = 4140 Kg/m3 .

(a) (b)

Fig. 2.66

Answer s.2.9.
The ions B and B’ by themselves have the structure of table salt (a cubic lattice on
which the locations of the two ions alternate). They therefore form an FCC lattice,
with a base comprising a B ion and a B’ ion. It remains to add all other ions.
Denoting the lattice constant of this FCC by a, the ions B and B’ can be located,
e.g., in the points (a/4, a/4, a/4) and (a/4, a/4, 5a/4), that is, at the centers of
two simple perovskite cells placed one on the top of the other. In these notations,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 102

102 Introduction to Solid-State Physics

the unit cell contains also two A ions, at (0, 0, 0) and (0, 0, a/2), and six oxygen
ions, at (a/4, a/4, 0), (a/4, 0, a/4), (0, a/4, a/4), (a/4, a/4, a/2), (a/4, 0, 5a/4), and
(0, a/4, 5a/4). All in all, the unit cell contains the chemical formula.

Answer s.2.10.
a. The unit cell is displayed in Fig. 2.9. The lattice vectors in the lower and
√ upper
planes are those shown in Fig. 2.8, i.e., a1 = (a, 0, 0) and a2 = (a/2, a 3/2, 0).
The third vector is normal to the planes, and its length is the distance in-between
them c, hence a3 = (0, 0, c). The base of the√crystal contains two carbon ions in
the lower plane, r1 = (0, 0, 0) and r2 = (a/2,
√ a 3/6, 0), and two carbon ions in the
mid plane, r3 = (0, 0, c/2) and r4 = (a, 2a 3/3, c/2). The two latter appear in the
planar unit cell shown in the mid panel of Fig. 2.9.
b. Figure 2.67 displays the rhombic unit cells of the three types of planes. In a
structure of the form . . .ABCABC. . . the unit cells are placed one on top of the
other. The lattice repeats itself in the fourth layer, whose distance from the first one
is c; the distance between two successive planes is √ thus c/3. In these new notations,
the lattice vectors are a1 = (a, 0, 0), a2 = (a/2, a 3/2, √ 0) and a3 = (0, 0, c).
√ Each
unit cell contains a base of six atoms, at r1 = (a/2, a √3/6, 0), r2 = (a, 2a 3/3, 0)
from the first plane (A), r3 = (0, 0, c/3), r√ 4 = (a, 2a 3/3, c/3) from the second
plane (B), and r5 = (0, 0, 2c/3), r6 = (a/2, a 3/6, 2c/3) from the third plane (C).

Fig. 2.67

Answer s.2.11.
In all answers below, the angles are defined according to Fig. 2.13.
a. The volume of the orthorhombic cell is abc. It contains four lattice sites, and
hence the volume of the primitive unit cell (which contains a single site) is abc/4.
b. Choose the triclinic lattice vectors along the x̂−axis, in the XY plane, and
along an arbitrary third direction. Then, a1 = a(1, 0, 0), a2 = b(cos γ, sin γ, 0),
a3 = c(cos β, (cos α − cos β cos γ)/ sin γ, z), where z 2 = 1 − cos2 β − (cos α −
cos β cos γ)2 /(sin γ)2 [this is obtained from the equation a3 · a3 = c2 ]. The
y−component of a3 is derived from a2 · a3 = bc cos α. Exploiting Eq. (2.2) yields

V = abcz sin γ = abc 1 − cos2 α − cos2 β − cos2 γ + 2 cos α cos β cos γ .

c. Using α = γ = 90◦ yields that the volume of the monoclinic unit cell is abc sin β.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 103

The crystalline structure of solids 103

d. The trigonal lattice is a special case of the √triclinic one, with a = b = c and
α = β = γ, and hence, according to (b), V = a3 1 − 3 cos2 α + 2 cos3 α.

Answer s.2.12.
The volume of the unit cell is V = a2 c sin(γ). The diameter of each sphere is
the distance between the centers of the two spheres in the base, (2r)2 = r22 =
[2(1 − Δ/3)2 (1 + cos(γ)) + (c/a)2 ]a2 /4. As in each unit cell there are two spheres,
the packing ratio is ρ = 8πr3 /(3V ). Figure 2.68 displays this packing ratio. One
finds that Δ = 0 corresponds to a BCC lattice, while Δ = 1 yields the quintessential
HCP one. Indeed, the maximal ratio, ρ = 0.74, is obtained for the latter.

Fig. 2.68

Answer s.2.13.
Figure 2.69 √displays a projection of the square lattice on a solid line of a slope
tan(α) = 1/ 2. One observes that there are only two projections, of lengths S =
a sin(α) and L = a cos(α). Examining the segments on the solid line, it is seen that
their ordering (beginning with the left edge) is LSLSLLSLSLSLLSL . . .. This
is neither a Fibonacci sequence, nor a periodic one. It describes a quasicrystals
different than the Fibonacci one.

Fig. 2.69
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 104

104 Introduction to Solid-State Physics

Answer s.2.14.
a. The substrate
√ in parts A and B of Fig. 2.58 is a square lattice with a lattice
constant a/ 2 (half of the diagonal on the face of the cube). The unit cell is
represented by the full line in the right panel of the upper row in Fig. 2.70. The
substrate in parts C and D is a triangular lattice with the same lattice constant (the
diagonal of each face in the FCC lattice contains two diameters of the substrate
atoms). The unit cell is the rhombus in the right panel in the lower row of Fig.
2.70.
b. The lattice
√ adsorbed on the substrate in part A is a square one, with alattice
constant a 2 (the square in part A of Fig. 2.70). Each unit cell contains a base
comprising one adsorbed atom and four substrate atoms. The adsorbed lattice in
part B is a square one as well, but with a lattice constant a (that is, half of the
diagonal of the square lattice in part A, and rotated by 45◦ relative to the former).
The unit cell contains one adsorbed atom and two substrate ones. √ The unit cell of
the adsorbed lattice in part C is a rhombus whose edge length is 2a, and its angle
is 60◦ . It is therefore a triangular lattice with a base that contains one adsorbed
atom and four substrate ones. √ The adsorbed lattice in part D is again a triangular
one, whose lattice constant is 3a (two heights of the triangles in the substrate).
The unit cell is rotated by 30◦ relative to the formers, and is represented by the
rhombus in the corresponding panel of Fig. 2.70.
c. In each of the cases, the maximal radius of the spheres that can occupy the
displayed structures is half of the lattice constant.
d. When the radius of the adsorbed atoms is smaller than that of the substrate
atoms, they can occupy all the dips between the substrate atoms. Then the adsorbed
lattice is identical to the substrate lattice. The two lattices are shifted relative to
one another by the distance between the center of an atom in the substrate and the
center of the nearest dip. The unit cell is identical to that of the substrate as well,
but includes also an adsorbed atom in addition to a substrate atom.

Answer s.2.15.
The spheres that represent the atoms in the substrate are marked by dashed curves
in Fig. 2.71. The distance between nearest neighbors on this hexagonal lattice
is 2r0 . The centers of the hexagons, (the sites for √
the adsorbed atoms), create a
new triangular lattice, whose lattice constant√is 2 3r0 . Figure 2.71(b) shows a
sphere (in solid curve) whose radius is r1 = 3r0 around each such center. For
this choice of the radius, the spheres are adjacent. Hence,√when the radius of the
adsorbed atom is larger than this value, i.e., x = r/r1 ≥ 3, the adsorbed atoms
cannot√ assume the structure displayed in Fig. 2.71(b). On the the other hand,
x ≤ 3 allows an adsorbed atom in each hexagonal dip of the substrate; when
x ≤ 1 the adsorbed sphere hovers over the center of each hexagon, and is in the
same plane as the substrate atoms. For larger values of x, the adsorbed spheres
touch the six adsorbing atoms around each hexagon; the centers of the adsorbed
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 105

The crystalline structure of solids 105

Fig. 2.70

√ √
ones move up, until x = 3. For 3 ≤ x ≤ 3 the adsorbed spheres can occupy next
nearest-neighbor hexagons on the triangular lattice of the hexagons’ centers. They
then create a hexagonal lattice of a lattice constant 6r0 . Figure 2.71(a) displays
the adsorbed lattice for the limiting case x = 3. For x ≥ 3 this lattice cannot
be realized, and the adsorbed spheres arrange themselves√ on a lattice of an even
larger lattice constant. For the two limiting cases, x = 3 and x = 3, the adsorbed
atoms form a triangular lattice, and in both cases the packing ratio is the same, the
maximal one.

(a) (b)

Fig. 2.71
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 106

106 Introduction to Solid-State Physics

Answer s.2.16.
The unit cell is represented by the dashed lines in Fig. 2.72. It contains four sites,
two with “up” spins (in the ellipse shown by a solid curve), and two with “down”
ones (in the dashed ellipse).

Fig. 2.72

Answer s.2.17.
a. The structure of the high-temperature phase, before the phase transition takes
place, is similar to that discussed in part (a) of problem s.2.3: the lattice is a square
one, with a base comprising a copper ion and two oxygen ones.
b. Because of the twisting of the plane, each edge of the original square is squeezed
and its length becomes a cos ϕ; the diagonals are perpendicular to one another, but
the edges are not. The two diagonals of the square are not equal to one another
anymore; the square turns into a rhombus. √ The diagonal along the direction of
the rotation axis preserves its original length 2a, but the second one is shortened
√ 
and becomes 2 (a cos ϕ)2 − (a 2/2) = a 2 cos(2ϕ) (Pythagoras theorem for a
2

triangle built of two half diagonals and a squeezed edge of the original square). The
surroundings of the copper ions marked by stars in the upper panel of Fig. 2.73
are identical but are different from those of the ions that are not marked in this
way. The edges of the new unit cell, shown in that figure, are the diagonals  of
the original√cell after the deformation; hence their lengths are a1 = a 2 cos(2ϕ)
and a2 = a 2. As the diagonals are perpendicular to one another, the result is a
rectangular lattice, with a base containing two copper ions and four oxygen ones.
c. In the antiferromagnetic arrangement all copper ions marked by stars have the
same magnetic moment, and therefore the unit cell is identical to that in (b).
d. The lower panel of Fig. 2.73 exhibits a projection of the copper ions’ lattice on
the the plane of the basis at low temperatures. The unit cell of that plane is the
dashed rectangle, which contains two copper ions as in part (b). Along the normal
to this plane, at the height of the original unit cell, there is another plane identical
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 107

The crystalline structure of solids 107

to the basis; the new unit cell is thus an orthorhombic one, with a rectangular
basis and of the same height as that of the original tetragonal unit cell (ignoring
distortions along this direction). The copper ions in the plane located at half height
(which were before at the centers of the tetragonal unit cells) are now located at the
centers of the faces of the orthorhombic unit cell. The new lattice is hence a face-
centered orthorhombic one, and its base comprises four copper ions (at the origin
and at the centers of the three faces close to it). As each copper ion is surrounded by
the chemical formula of the material, the new base includes that chemical formula
four times: four copper ions, eight lanthanum ones, and sixteen oxygen ions.

Fig. 2.73
December 12, 2014 15:40 BC: 9334 – Bose, Spin and Fermi Systems 9334-book9x6 page 2

This page intentionally left blank

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 109

Appendix A

The Platonic bodies

Plato’s polyhedra. Symmetries have enchanted people since antiquity and sym-
metric bodies feature in many ancient art objects. Their namesake, Plato (circa
400 BC), had classified polyhedra: three-dimensional bodies with two-dimensional
faces, one-dimensional edges, and point-like corners (vertices). A perfect platonic
body is a concave polyhedron built of identical equilateral polygons with identical
edges and identical vertices; it thus has perfect symmetries. Neighboring faces of
a perfect polyhedron touch each other along one common edge; identical numbers
of polygons share each corner. As seen in problem (A.1), there exist solely five
bodies that comply with these rules. These are displayed in Fig. A.1; one can
verify that they all obey the Euler theorem, V − E + F = 2, where V , E, and F
are the numbers of corners, edges, and faces, respectively. There are many proofs
of this important theorem, which pertains in fact to any polyhedron.

Problem A.1.
Each face of a perfect polyhedron is bounded by S edges, and P faces join together
at each of its corners. Explain why the sum of the P angles on the faces that meet
at a certain corner must be smaller than 360◦ and why does this fact entail that
P (1 − 2/S) < 2. Exploit the result to prove that there exist only five platonic bodies.

Symmetries. Inspection of Fig. A.1 shows that a platonic body is invariant

under many symmetry operations which form its symmetry group. The symmetry
group of the octahedron coincides with that of the cube: the octahedron in Fig.
A.1 is not altered by any of the symmetry operations of the cube. In particular, the
octahedron consists of three squares and eight triangles which are invariant under
rotations of 90◦ and 120◦ , respectively. Each of the platonic polyhedra possesses a
dual polyhedron, obtained upon joining the centers of neighboring faces. The two
dual bodies share the same symmetries. The octahedron is dual to the cube, the
dodecahedron is dual to the icosahedron, and the tetrahedron is dual to itself.
The symmetry group of the platonic bodies can be derived by exploiting the
fact that each of the symmetry operations transforms the corners of the polyhedron

109
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 110

110 Introduction to Solid-State Physics

Fig. A.1: The platonic bodies, from left to right: the tetrahedron (4 equilateral
triangles, 6 edges, and 4 corners), the octahedron (8 equilateral triangles, 12 edges,
and 6 corners), the cube (6 squares, 12 edges, and 8 corners), the icosahedron (20
equilateral triangles, 30 edges, and 12 corners), and the dodecahedron (12 equilateral
pentagons, 30 edges, and 20 corners).

to themselves. For example, the tetrahedron has four vortices, and therefore there
are 4! = 24 permutations of them, which produce 24 elements in the tetrahedron’s
symmetry group. This argument is valid only for the tetrahedron, where each corner
is a neighbor of all others. For the bigger bodies that are not dual to themselves, not
all permutations are allowed: in some of them neighboring corners do not remain as
such after the permutation. In general, each of the symmetry operations consists of
a combination of rotations and reflections of the polyhedron. For instance, if each
of the faces of a polyhedron is bounded by S edges, and at each of its corners meet
P faces, then the line joining the center of the polyhedron with the center of each
face is a rotation axis of order S, and the line joining that center with any corner
is likewise a rotation axis of order P .

Problem A.2.
Describe the symmetry operations of the tetrahedron as rotations and reflections.

The symmetry group of the icosahedron. A glance at Fig. A.1 shows

that each triangular face is opposite to another triangular one, rotated by 180◦ .
Similarly, opposite to each pentagon face of the dodecahedron there is a parallel
one whose pentagon is rotated by 180◦ . In addition, opposite to each of the edges
of these polyhedra there is another, parallel, edge. The argument given above
implies that the symmetry group of the icosahedron contains four rotations of order
five (rotations by 2π /5, with = 1, 2, 3, and 4) around the axes connecting six
pairs of opposite corners, and two rotations of order three around each of the axes
that connect ten pairs of oppositely-located faces. There are also the second-order
rotations around each of the axes connecting the 15 midpoints of the oppositely-
located edges. Together with the unity, all these operations amount to a symmetry
group comprising 60 elements. The icosahedron and the dodecahedron do not
appear in periodic lattices, but do play a decisive role in quasicrystals (see Sec. 2.8).
The reason is that the point symmetry-group of the icosahedron contains rotations
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 111

The Platonic bodies 111

of order five, as indeed is observed in experiments carried out on quasicrystals.

Figure 2.19 and problem 3.11 mention the rhombic dodecahedron. Though
this body consists of identical square faces, its corners do not share the same en-
vironment: there are corners with three faces, and others that combine four faces.
Hence, the rhombic dodecahedron is not a prefect platonic body. Nonetheless, one
may convince oneself that the symmetry of the rhombic dodecahedron is the same
as that of the cubic lattice.

A.1 Answers

Answer A.1.
Joining one of the vertices of an equilateral polygon with the S − 3 vertices that are
not its neighbors partitions the polygon into S − 2 triangles. Therefore, the sum
over the angles of the polygon is (S − 2)180◦ . As the polygon possesses S corners,
the angle at each corner is (S −2)180◦ /S. The sum of the angles around each corner
is thus P (S − 2)180◦ /S and since the polygon is concave, this sum must be smaller
than 360◦ . It follows that P (1 − 2/S) < 2. When P (S − 2)180◦ /S = 360◦ the faces
lie on a plane, and when P (S − 2)180◦ /S > 360◦ , the body becomes convex. The
number of the edges S must exceed 3 (lest it will not be a polygon) and P must
exceed 3 in order to build a corner. For S = 3 the inequality P (1 − 2/S) < 2 is
obeyed solely for P = 3, P = 4, or P = 5, values that correspond to the tetrahedron,
the cube, and the icosahedron. For S = 4 and S = 5 the inequality is obeyed solely
for P = 3, and these values agree with the cube and the dodecahedron. There are
no other values of S and P that obey P (1 − 2/S) < 2; there are thus only five
platonic bodies.

Answer A.2.
The line joining each vertex of the tetrahedron with the center of the opposite
face is the height of the tetrahedron. Around each of these heights, there are two
rotations of order 3, for instance by the angles ±2π/3, altogether eight rotations.
On each of the four faces there are three heights, altogether 12 heights. Joining
each of the heights with the vertex opposite to the face results in a planar mirror;
a reflection through which leaves the tetrahedron unchanged. The eight rotations
correspond to permutations of the three vertices, and the 12 reflections correspond
to permutations of pairs of vertices. In addition there are three symmetry operations
transforming all four vertices simultaneously, e.g., ABCD→DABC. Each of these
operations can be accomplished in two steps; e.g., ABCD→ADBC→DABC. The
first step is a rotation of the triangle BCD, and the second is a reflection that
interchanges between AD and DA. Together with the identity element there are 24
symmetry operations.
December 12, 2014 15:40 BC: 9334 – Bose, Spin and Fermi Systems 9334-book9x6 page 2

This page intentionally left blank

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 113

Chapter 3

Radiation scattering off crystals

This chapter focuses on experimental methods to identify the crystalline structures

of various materials. The paramount one to be described below exploits scatter-
ing of certain types of radiation off crystals, notably X-rays, but also scattering of
neutron or electron beams. The radiation scattered by the periodic lattice creates
a diffraction pattern on a screen. The laws that determine the scattering angles
which give rise to constructive interference, named after Bragg and von Laue, are
examined, as well as ubiquitous tools and techniques, as monochromatic scattering
off powders or from a single crystal upon rotating it, or multi-chromatic radiation.
The intensity of the scattered radiation comprises distinct peaks (“Bragg peaks”)
that appear at particular scattering angles, for which the difference between the
impinging wave vector and the scattered one equals one of the vectors of the recip-
rocal lattice. Accordingly, the properties of reciprocal lattices are studied in detail.
These particular scattering angles correspond to diffraction from families of planes
in the lattice. The intensity of the scattered radiation is related to the structure
factor, i.e., the Fourier transform of the scatterers’ density in the material. This
is the density of the electrons for X-ray scattering, of the nuclei or the magnetic
moments for neutron scattering. Additional topics surveyed in this chapter include
scattering off quasicrystals, scattering of electrons from planar samples, and scat-
tering of neutrons off magnetic systems. The appendix contains a terse summary
of Fourier series on periodic lattices.

3.1 Introduction: identifying crystalline structures

Scanning tunneling microscope (STM). There are several experimental

methods to identify crystalline structures. A relatively modern one, in par-
ticular suitable for investigations of two-dimensional surfaces, is based on the
scanning tunneling microscope, STM. In this method, a very thin metallic tip
that hovers above a solid surface, is moved around. Upon applying a voltage drop
between the tip and the surface [generated by contacting the surface and the tip to
the terminals of a voltage source, see Fig. 3.1(a)] the current that flows between

113
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 114

114 Introduction to Solid-State Physics

them increases as the tip approaches closer the surface: the space in-between the
tip and the surface forms a quantum potential barrier, which the charge carriers
(i.e., the electrons) overcome by quantum tunneling. This potential barrier is
reduced as the tip approaches closer to the surface, leading to an increase in the
current. Measuring the current as a function of the location of the tip enables the
mapping of the surface. Alternatively, the current is kept constant but the tip is
connected to a spring that allows its vertical motion to follow the varying height of
the surface. In that case one measures the voltage that monitors the height. Figure
3.1(b) displays such a mapping of graphene (the graphene structure is described in
Sec. 2.3, see Figs. 2.8 and 2.10, and also Fig. 1.6).

(a) (b)

Fig. 3.1: (a) Illustration of the scanning tunneling microscope, STM. (b) Mapping
of graphene by STM (the brighter color corresponds to a larger current, and hence
to higher regions of the surface). The dark regions represent valleys at the centers
of the hexagons of the graphene lattice. Courtesy of E. Andrei, Rutgers Univ.

The atomic force microscope (AFM). A complementary version of the

STM is the atomic force microscope, AFM, which is based on a cantilever with
a very sharp needle-like edge (thickness of the order of nanometers), see Fig. 3.2.
The cantilever hovers over the surface; its tip is sensitive to the force exerted on
it by the surface, stemming, e.g., from the mechanical contact, or from the electric
or magnetic potentials of the surface. This force increases as the cantilever comes
closer to the surface, causing the latter to move up and down. These vibrations
are detected by a laser beam reflected from the upper side of the cantilever; the
reflection angle is in particular sensitive to the distance of the cantilever from the
surface.
Scattering of radiation off a crystal. The methods described above are use-
ful for the study of thin films or surfaces of three-dimensional crystals; they do
not convey much information about the internal structure of the bulk itself. Such
knowledge is gathered by scattering radiation off crystals. Each of the atoms
in the crystal scatters the impinging waves; the outgoing radiation comprises in-
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 115

Radiation scattering off crystals 115

Fig. 3.2: Atomic force microscope.

terference of all scattered waves, from all atoms. Conspicuous diffraction patterns
are obtained when the wave length of the radiation is of the order of the distances
between the scattering objects, i.e., of the order of several angstroms for solid crys-
tals. Electromagnetic waves of frequency ν and wave length λ consist of photons of
energy ε = hν = hc/λ, where h is the Planck constant and c is the speed of light.
Hence,

λ(Å) = 12.4/ε(KeV) , (3.1)

where the wave length is measured in angstroms (1Å = 10−8 cm), and the energy
of the photon is measured in kilo electron-volt, KeV (1 eV=1.602×10−19 Joule).
Energies of the order of several KeV’s yield wave lengths of the order of the distances
in the lattice; such energies correspond to X-rays. These rays are created, for
instance, when an electron in a lower atomic level is “kicked” out, and another
electron from an upper level moves “down” to occupy it, emitting in the process
a photon of energy equal to the difference between the two levels. The emitted
radiation contains waves of well-defined, discrete wavelengths. The quantum states
of electrons in the atom are characterized by the quantum numbers n, , and j,
which represent the energy, the orbital angular momentum, and the total angular
momentum (orbital and spin), respectively. The ground state of a single electron is
given by n = 1, = 0, and j = 1/2. This state is termed K, or 1s1/2 (the notation s
represents = 0). When this state is empty, it can be filled by an electron coming
from a higher energy state, upon emitting a photon of the X-ray radiation. This
transition can take place from excited levels with = 1, which are denoted by npj
(the notation p represents = 1). In particular, the transitions from the levels
2p3/2 , 2p1/2 , and 3p3/2 are called Kα1 , Kα2 , and Kβ1 , respectively.

Problem 3.1.
a. The ground-state energy of an electron in copper is −8979 eV, and the energies
of the excited states 2p3/2 , 2p1/2 , and 3p3/2 are −952 eV, −933 eV, and −76 eV,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 116

116 Introduction to Solid-State Physics

respectively. Determine the wave lengths corresponding to the three transitions Kα1 ,
Kα2 , and Kβ1 .
b. The energies of the electronic states in atoms are proportional to z 2 , where z
is the atomic number, i.e., the nucleus’ charge (in units of the electron charge).
[The Rydberg constant contains the factor e4 , where e is the electron’s charge. This
factor originates from the square of the coefficient of the Coulomb potential between
the nucleus and the electron. As the charge of the nucleus is ze, it follows that the
Coulomb potential is multiplied by z and consequently the levels’ energies are mul-
tiplied by z 2 .] Potential sources for X-rays are chromium, cobalt, or molybdenum.
What are the wave lengths of the three transitions for these sources?

Synchrotron radiation. “Standard” sources of X-rays are based on the ac-

celeration of electrons in vacuum tubes. Electrons emitted from a hot cathode
are accelerated in a high voltage drop, to impinge on the anode, where they re-
lease electrons from low-energy levels. In this way the X-ray radiation described in
problem 3.1 is created. The intensity of this radiation is not particularly strong.
Higher-intensity radiation, involving photons of energies of the order of several kilo
electron-volts, can be obtained e.g., from electrons colliding with solids; the
deceleration of these electrons results in the emission of photons. Another option is
synchrotron radiation: the radiation emitted from charged particles moving on a
circle. Such particles emit photons because of their acceleration. In both cases the
emitted spectra comprise a continuum of energies, but discrete wave lengths can
be achieved upon using suitable filters. Synchrotron radiation can be of intensity
much higher than the one emitted from vacuum tubes.
Scattering of electrons or neutrons. Diffraction patterns can also be
achieved by scattering massive particles off the crystal, e.g., neutrons or elec-
trons (found in nuclear reactors or in accelerators), exploiting the duality of massive
particles and waves in the quantum realm. The de Broglie wave vector and wave
length of a particle of mass m and momentum p are k = p/ and λ = h/|p|, respec-
tively; the energy of each particle is ε = h2 /(2mλ2 ). The wave length of electrons
is thus given by
λ(Å) = 12/[ε (eV)]1/2 , (3.2)
and that of the neutrons is
λ(Å) = 0.28/[ε (eV)]1/2 (3.3)
(check!). The corresponding energies are of the order of several electron-volts for
electrons, and parts of electron-volts for neutrons. Note though that the atomic
scattering cross-section for electrons is quite large, and therefore the electrons are
mostly scattered by the atoms close to the surface of the specimen, and do not
penetrate deep into the crystal. For this reason electrons’ scattering is used mainly
in explorations of surfaces (see Sec. 3.12). Neutrons, on the other hand, have
two unique properties that distinguish them from photons: (i) they are scattered
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 117

Radiation scattering off crystals 117

mostly by the nuclei and not by the electrons in the material, and thus are suitable
for extracting information on the nuclei’s locations; (ii) they possess a magnetic
moment and are electrically neutral, and hence are useful for investigations aimed
to study the orientations of the magnetic moments in the material (see Sec. 3.13).

3.2 Bragg’s law

Elastic scattering from a crystal. In Chapter 4 it is explained that the sum

of the forces between the atoms (or ions, or molecules) in the crystal vanishes at
equilibrium, when the atoms are located on the sites of a perfectly periodic lattice.
In reality the atoms oscillate around these lattice sites (see Chapter 5), in the
same way as a quantum harmonic oscillator vibrates around the minimum of its
potential energy, even in the ground state. These oscillations are ignored in most
of this Chapter: it is assumed that the atoms stay in their average locations, i.e.,
at the lattice sites. Figure 3.3 displays the two main configurations exploited in
experiments that monitor the scattering of radiation off crystals. As the atoms rest
in the lattice sites, there is no energy exchange between them and the radiation,
and hence the scattering is elastic, much like a classical ball that hits a wall
is reflected from it with no energy exchange. The kinetic energy of the scattered
particles (photons, neutrons, or electrons) is conserved, and the wave length of the
scattered wave is the same as that of the incoming one. Below are presented several
methods to derive the condition for constructive interference of elastically-scattered
radiation, all equivalent to each other.
Scattering off a single crystal. A scattering experiment can be carried out
on powder, which contains many tiny crystals (grains) of the material, or on a
single crystal. Figure 3.3(a) exhibits scattering off a single crystal. As found
below, when the scattered radiation contains constructive interference, it can be
associated with scattering from a family of parallel planes comprising atoms of
the crystal. For simplicity, assume that the surface of the sample is planar, and
contains one of those planes. The crystal is oriented such that the angle between the
impinging beam and that plane, denoted by θ, and the angle between the scattered
beam and the plane, are equal [see Fig. 3.3(a)]. The angle between the normal to
the plane and the impinging wave, 90◦ − θ, is called in optics the impinging angle.
The angle between the impinging beam (incoming from the left) and the outgoing
one [hitting the monitor in Fig. 3.3(a) or the counter in Fig. 3.3(b)] is 2θ, and
is called the scattering angle. In the geometry adopted here, the surface plane
of the sample is perpendicular to bisector of the angle in-between the incoming
and the outgoing waves. Usually, the sample is much smaller than the distances
between it and the radiation source and the monitor, and the wave packet is very
narrow (though wide compared to the distances between atoms within the crystal).
Then all the rays parallel to each other that hit the counter are focused there into a
single point, and hence the amplitude of the scattered wave, called the scattering
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 118

118 Introduction to Solid-State Physics

(a) (b)

Fig. 3.3: Illustration of radiation scattering. (a) The beam incident from left is
scattered from a single crystal, and then impinges on the radiation counter; the
latter can be moved on the surface of a sphere enclosing the crystal (represented
by the circle). The parallel lines inside the crystal represent the planes scattering
the radiation. (b) The beam scattered from a powder of the investigated material
impinges on a screen, where each circle corresponds to a particular scattering angle;
the various points on each circle result from different grains in the powder.

amplitude, is the sum over all their amplitudes. The intensity of the scattered
radiation is proportional to the absolute value squared of this amplitude. The
scattered waves interfere; a constructive interference is realized only for discrete
values of the wave length (when the scattering angle is fixed) or for discrete values
of the scattering angle (when the wave length is fixed and the data are taken at
various orientations). Each such constructive interference is related to scattering
off a certain family of planes in the crystal, and therefore the detection of those
discrete values allows for the identification of the structure of the scattering crystal
(see below). One may use monochromatic radiation, which has a single wave
length, moving the counter over the sphere shown in Fig. 3.3(a), and so measure
the intensity of the scattered radiation for various scattering angles. Alternatively,
one may locate the counter at a certain point and use multi-chromatic radiation
which comprises a continuum of wave lengths. Constructive interference is again
achieved only for certain wave lengths. The counter may be replaced by a screen on
which there appear discrete points whenever the interference is constructive. These
points create the diffraction pattern.
Scattering off powders. Figure 3.3(b) displays scattering off a powder: the
sample is chopped into tiny grains (but still much larger than the distances between
atoms in the crystal). The lattice planes in each grain are oriented differently
relative to the impinging beam, resulting in circular rings of radiation on the screen,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 119

Radiation scattering off crystals 119

at specific scattering angles obeying Bragg’s law. Figure 3.4 exhibits such rings with
the corresponding intensities of the scattered radiation.

Fig. 3.4: Diffraction pattern obtained from a powder. The circular plane is the
screen on which the scattered radiation is impinging; the bright peaks give the
average intensity of each ring. Each peak corresponds to a specific scattering angle
for which there is a constructive interference.

Bragg’s theory of scattering off planes. The diffraction patterns generated

by scattering off crystals were discovered by von Laue. More than a hundred years
after this discovery the science of crystallography, which explores the structures
of various materials by scattering off radiation, is still quite active. Figure 3.5
summarizes the hallmarks of the progress in this field up to 2014, a year which was
declared the “international year of crystallography”.
The analysis of the diffraction patterns generated by scattering off crystals was
accomplished by William Henry Bragg and by his son, William Lawrence
Bragg. They assumed that the material consists of parallel planes, as illustrated in
Fig. 3.3(a), and that the diffraction pattern is created by scattering off these planes.
This is not quite so, as each crystal is built of discrete elements (atoms, molecules,
or more complicate bases). The analysis carried out by the two Braggs pertains to
the planes on which those elements are situated. Nonetheless, calculations of the
scattering amplitude off discrete points in the lattice yield the same results (see
below).
Planes in the lattice. As mentioned in Sec. 2.9, planar cross-sections of a
three-dimensional lattice can assume various structures. In fact, a three-dimensional
lattice can be described as an ensemble of parallel planes, where each plane is a two-
dimensional lattice. For example, the graphite lattice in Fig. 2.9 is a collection of
planar hexagonal layers, (where each plane is the two-dimensional graphene lattice).
The cubic lattice in Fig. 2.15 is built of square-lattice planes, parallel to each other,
and the FCC and HCP lattices in Fig. 2.24 comprise planar layers, each being a
triangular lattice. The FCC lattice can also be described as built of parallel square-
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 120

120 Introduction to Solid-State Physics

Fig. 3.5: Hundred years of crystallography. [N. Jones, Nature 505, 602 (2014).]

lattice planes [e.g., the plane of the cube’s face in the central panel of Fig. 2.17].
A point in the three-dimensional lattice can hence be denoted by the serial number
of the plane on which it is located, and by its location within that plane. Figure
3.6 shows that a lattice can be described by different families of planes. Figure
3.6(a) displays five such families for the rectangular lattice (where each “plane” is
in fact a line); Fig. 3.6(b) portrays three families of planes for the cubic lattices. It
is seen that each site in the lattice belongs to a certain plane, in each of the families.
The families differ from each other by the distances among the planes, and by the
density of points on each plane.
A lattice may be partitioned into families of planes as follows. One chooses at
random three lattice points (that are not lying on a straight line). These three points
define a plane in space. As each pair of them is connected by a vector comprising
lattice vectors, two (out of three) of the lattice vectors allow one to reach an infinite
number of other lattice points, all located on the same plane. Choosing a point on
the plane and identifying the lattice point closest to it which is not on the same
plane, and then using the very same planar lattice vectors to reach other lattice
points results in the creation of a new plane, parallel to its predecessor. Repeating
this procedure eventually assigns all lattice points to such planes.
As an example, consider the planes of the FCC lattice in Fig. 3.6 marked by
(200). Assign to this lattice the cubic lattice vectors, a1 = ax̂, a2 = aŷ, and
a3 = aẑ, and choose the three points (0, 0, 0), (0, a, 0), and (0, 0, a). These points
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 121

Radiation scattering off crystals 121

(a) (b)

Fig. 3.6: Partition of lattices into families of planes. (a) The rectangular lattice;
(b) the three cubic lattices, SC, FCC, and BCC in descending order. The thick
arrows show vectors normal to each family of planes (see text).

are located on the YZ plane (the left plane in the figure). The lattice points on this
plane form a square lattice, rotated by 45◦ relative to the cartesian coordinates,
with the lattice vectors a2 = (a2 + a3 )/2 and a3 = (a2 − a3 )/2. Any other point
in this plane can be reached by a linear combination of these lattice vectors, with
integer coefficients. Noting that the distance of the point at R2 = (a/2, a/2, 0)
(located at the center of the cube near the origin) from the YZ plane is a/2, which
is the smallest possible distance of a lattice point from this plane, identifies the
nearest plane. All points on the plane parallel to first one, which are on the same
plane as R2 , are located at R2 + n2 a2 + n3 a3 , with integer coefficients. Similarly,
the points on the next plane are located at R3 + n2 a2 + n3 a3 , where R3 = (a, 0, 0).
The distance of the point at R3 to the second plane is a/2, and so forth. It is easy
to be convinced that the ensemble of points on all the planes is identical to that
of the points on the original lattice. A different choice of the initial three points
constructs another family of planes. The various families, as well as the notations
in Fig. 3.6, are sorted out below.

Problem 3.2.
a. Find the equation of the plane that contains the three points at the ends of the
edges of the lattice vectors in Eq. (2.5) and in the left panel Fig. 2.16,
a a a
R1 = (−x̂ + ŷ + ẑ) , R2 = (x̂ − ŷ + ẑ) , R3 = (x̂ + ŷ − ẑ) .
2 2 2
Determine the lattice vectors that describe the points of the BCC lattice located
on this plane. Find the structure of the planar lattice formed by these points, and
the area of the unit cell. At which points does this plane cut the axes of the cubic
lattice? Does this plane appear in Fig. 3.6?
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 122

122 Introduction to Solid-State Physics

b. Find the equation of the plane parallel to the one discussed in (a) that passes
through the origin. Find the distance between the two planes. Find the equation of
the plane closest to the one discussed in (a), which is parallel to it from the other
side (relative to the origin), that passes also through the points of the BCC lattice.
Write down the equations of all other planes parallel to these two, and that passes
through the lattice points.
c. Use the area of the unit cell of each plane and the distances among the planes
to compute the density of points of the lattice. Compare the result with the density
of points on the original BCC lattice, and demonstrate that the family of planes
indeed contains all lattice points.
Bragg’s law. Consider a beam incoming from a far-away source on the left
of the scatterer, and being scattered by a family of planes, e.g., the one shown in
Fig. 3.7(a). The detector is located far away on the right. Each horizontal line
in this figure represents a plane normal to the page, and connects lattice sites on
this plane. The grey areas (in the page plane) represent the impinging beam and
the scattered one, which are both oriented at an angle θ relative to each of the
planes. It is the same angle for the scattered and the impinging beam since – like
for particles colliding with a planar surface–the momentum of the radiation particles
along the tangent to the surface is not changed; momentum is exchanged only along
the normal to the surface. The two rays marked by thick arrows are scattered by
points located on neighboring planes. Denoting the distance between the planes by
d, the difference between the lengths of these two rays is 2d sin θ. [This is the sum
of the two bold segments in the figure, each being the side opposite to the angle θ in
a right triangle (the dashed lines), whose hypotenuse is d.] Wave mechanics implies
that for the two rays to interfere constructively the difference of their lengths should
be an integer number of wave lengths, λ, i.e.,
2d sin θ = nλ , (3.4)
where n is an integer. Equation (3.4) represents Bragg’s law. Monochromatic
rays (i.e., of a single wave length λ) that are scattered from a certain family of
planes (i.e., with a given value of d) interfere constructively only for discrete angles,
for which sin θ = nλ/(2d). One has to rotate the crystal or the counter in Fig.
3.3(a) until constructive interference is achieved. Alternatively, for a given family
of planes and a specific scattering angle, the interference is constructive only for
certain wave lengths λ = 2d sin θ/n. The Bragg condition, although derived for
continuous planes, is valid also for scattering off discrete points (see below).
Problem 3.3.
Show that condition (3.4) is obtained also when the two rays are scattered from any
pair of points in two neighboring planes, not only from the particular one displayed
in Fig. 3.7(a).
von Laue law. An alternative derivation of the condition for constructive
interference is carried out by introducing a vector d of length d that is normal to
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 123

Radiation scattering off crystals 123

(a) (b)

Fig. 3.7: (a) A beam scattered from planes in the lattice. (b) The wave vectors of
the impinging and the scattered waves, and their vectorial difference.

the planes in Fig. 3.7(a), and assigning to the incoming and to the scattered beams
the wave vectors k and k , respectively. The latter point along the propagation
directions of the waves, and their lengths (“wave vectors”) are |k| = |k| = k = 2π/λ.
The wave lengths of the rays are identical since the scattering is elastic, i.e., it
conserves energy. The wave vectors are related to the momentum of the photons (or
in general, of the particles scattered off the crystal), e.g., p = k. The momentum
of the scattered particles is not conserved; the difference between the momentum
of the impinging wave and that of the scattered one is absorbed by the crystal.
Examining Fig. 3.7(a), one notes that k · d = −k · d = kd sin θ. The product of
the wave number k and the length of the traversed path is the optical path; these
paths are marked by the thick segments in Fig. 3.7(a). The difference between the
optical paths of the two rays is the phase difference between the two waves. The
two waves interfere constructively when the phase difference is an integer product
of 2π. It follows that the condition for constructive interference, i.e., the Bragg law,
can take the form
(k − k) · d = 2πn . (3.5)
This form is known as the von Laue law.
As seen in Fig. 3.7(b), the vector k − k is normal to the planes, i.e., it is parallel
to the vector d. Hence, the scalar product on the left hand-side of Eq. (3.5) is the
product of the lengths of the two vectors, |k − k|d = 2πn. The smallest momentum
difference is obtained for n = 1,
min|k − k| = 2π/d . (3.6)
Bragg peaks. Note that the scattered beam is in the plane that contains the
impinging beam, which is normal to the crystalline planes. Therefore, a screen
located perpendicular to this scattering plane “captures” the scattered beam in dis-
crete points, at angles which obey the Bragg (or the von Laue) law, with various
values of the integer n. This resembles a well-known phenomenon of optics: each
peak in the diffraction pattern corresponds to a difference of the optical paths com-
prising an integer number of wave lengths. A plot of the intensity of the scattered
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 124

124 Introduction to Solid-State Physics

radiation as a function of the location of the corresponding point on the screen

(i.e., as a function of the scattering angle) thus contains a narrow peak for each
such point. These peaks are named after Bragg, and their very existence is a proof
that the scatterer is a crystal. When the scattering experiment is carried out as
illustrated in Fig. 3.3(b) or in Fig. 3.4, there appear circles on the screen, as the
scattering is accomplished from differently-oriented grains (see Sec. 3.7). An analy-
sis of the scattering off quasicrystals is presented in Sec. 3.11; though not periodic,
those also give rise to narrow peaks. Following their discovery, the definition of a
crystal was modified, to be a material characterized by narrow and discrete
Bragg peaks, whether it is periodic or else. The discussion in the following is
focused on periodic crystals.

3.3 Scattering off a point-like Bravais lattice

Scattering off points. The discussion so far dealt with the condition of construc-
tive interference; the radiation intensity has not been calculated. The amplitude
of the scattered wave is the sum of the amplitudes of the waves scattered off various
elements in the crystal. Consider first the simplest case, in which each point R
on a Bravais lattice accommodates a single, point-like scatterer; more general
configurations are discussed below. The scattered waves emerge only from these
points. Assuming that the amplitude of the wave outgoing from the (arbitrarily
chosen) origin is a = a0 exp[iψ0 ] (where a0 is a real number), the amplitude of the
wave emerging from a lattice point at R is a(R) = a0 exp[iψ(R)]. The phase dif-
ference between the two waves is the difference between the optical paths. As seen
in Fig. 3.8, this difference is the sum of the optical path from the origin to point
A (which is k · R), and the optical path from B to the origin (which is −k · R).
Hence, ψ(R) = ψ0 + (k − k) · R, and the total amplitude of the scattered beam,
called the scattering amplitude, is
 
A = a0 eiψ0 ei(k −k)·R ≡ a0 eiψ0 Z(q) , (3.7)
R

where q = k − k is, up to the multiplicative factor , the momentum difference

between the impinging and scattered particles (e.g., photons). In Eq. (3.7),

Z(q) = eiq·R . (3.8)
R

Vectors in the reciprocal lattice. The intensity of the scattered radiation

is proportional to the absolute value squared of the total amplitude, i.e., to |A|2 =
a20 |Z(q)|2 . This intensity is maximal when all terms in the sum (3.8) have the same
phase, e.g., 0 (for any other constant phase, say ψ, the factor exp[iψ] can be taken
out of the sum, and then it does not change the absolute value). When this is the
case, the sum equals the number of lattice sites, Z(q) = N , and |A|2 = a20 N 2 . This
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 125

Radiation scattering off crystals 125

Fig. 3.8: Interference of two waves scattered from the points 0 and R in the lattice.

maximal value is reached when

eiq·R = 1 , (3.9)
for any lattice point R. Such an equality is realized only for a specific sequence of
vectors q. Indeed, a collection of vectors obeying Eq. (3.9) for any R can be found
for any Bravais lattice, as proven in Sec. 3.4. Denote the solutions of Eq. (3.9) by
{q = G}, where with each solution q = G there appears also the solution q = −G.
This collection of vectors builds a periodic Bravais lattice in “momentum” space,
which is called the reciprocal lattice. It follows that maximal scattering intensity
is reached when q equals one of the vectors of the reciprocal lattice. For a given
direction of the wave vector k of the impinging radiation, the peaks in the intensity
appear for the vectors k such that q = k −k = G is one of the vectors in reciprocal
lattice. These are the aforementioned Bragg peaks. When this is not the case,
i.e., q is not one of the vectors of the reciprocal lattice, the sum Z(q) contains
terms with various phases, and therefore possesses real and imaginary terms that
may be of opposite signs and so are partially cancelled. As a result, the intensity
of the scattered radiation is much smaller.
Scattering off large but finite lattices. When the sample in the scattering
experiment is very large, the Bragg peaks are quite narrow; when the size approaches
infinity their width tends to zero. This property is proven as follows. Inserting into
the sum (3.8) the expression for R in terms of the lattice vectors, R = n1 a1 +
n2 a2 + n3 a3 [with integer coefficients, Eqs. (2.1)], yields
   
Z(q) = eiq·R = eiu1 n1 eiu2 n2 eiu3 n3 , u i = q · ai , (3.10)
R n1 n2 n3

where the sums run over the entire lattice. Assume that there are Ni atoms in
the direction of ai and exploit periodic boundary conditions at the surfaces of the
specimen [i.e., the lattice repeats itself ad infinitum at each direction (assumptions
on the edges in a macroscopically-large sample have almost no effect)]. This type
of boundary conditions is named after Born and von Karman. The periodic
boundary conditions allow one to place the origin at an arbitrary lattice point.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 126

126 Introduction to Solid-State Physics

Hence, when Ni is odd, it can be taken at the middle of the sample, and then the
sum encompasses values in the range −(Ni − 1)/2 ≤ ni ≤ (Ni − 1)/2. (For even
Ni the range is −Ni /2 ≤ ni ≤ Ni /2 − 1; the final result is identical for both cases
provided that the sample is large enough. Any choice of the origin produces the
same result for the scattering intensity – check.) Since each of the sums in Eq.
(3.10) is a geometrical series, it yields
(Ni −1)/2
 1 − eiNi ui sin(Ni ui /2)
Zi (ui ) = exp[iui ni ] = exp[−i(Ni − 1)ui /2] = .
1−e iiu sin(ui /2)
−(Ni −1)/2
(3.11)
The function Si = |Zi | is depicted in Fig. 3.9 for two values of Ni . It has narrow
2

peaks, Si = Ni2 , whenever the denominator in Eq. (3.11) vanishes, i.e., for q · ai =
ui → 2πmi with an integer mi . This condition is identical to the one in Eq. (3.9),
for which the scattering intensity is maximal (check!). At the vicinity of each such
peak, the ratio on the right hand-side of Eq. (3.11) equals sin(Ni wi /2)/ sin(wi /2),
with wi = ui − 2πmi . This ratio equals Ni when wi = 0, and is approaching zero
when sin(Ni wi /2) = 0 and wi = 0. The first zero appears for wi = ±2π/Ni . The
ratio is sharply peaked around wi = 0, with height Ni and width (at its base) 4π/Ni .
Hence, the height increases with Ni while the width decreases; the area under the
peak, however, does not change and is close to 2π (like the area of a triangle of base
4π/Ni and height Ni ). The function remains relatively small for other values of ui
in-between the peaks, where the numerator in Eq. (3.11) oscillates wildly between 1
and −1, and the denominator remains finite. As the samples in the experiment are
always of a finite size, any experimentally-detected Bragg peak has a finite width.
Nonetheless, the width is much narrower as compared to the intervals between the
reciprocal-lattice vectors.
In three dimensions, Eq. (3.10) yields Z(q) = Z1 (u1 )Z2 (u2 )Z3 (u3 ), and there-
fore Bragg peaks appear only when all three equations are satisfied, q · ai =
ui = 2πmi , i = 1, 2, 3, namely when q · R = q · (n1 a1 + n2 a2 + n3 a3 ) =
2π(n1 m1 + n2 m2 + n3 m3 ), or exp[iq · R] = 1. The discussion following Eq. (3.9)
implies that q must be one of the reciprocal-lattice vectors.
Scattering off an infinite lattice. As in most cases the measured lattices are
rather large, one may well describe them as having an infinite size. To consider this
limit, recall first the properties of the Dirac delta-function, δ(r). Mathematically
speaking, the delta function is a distribution and not a function, but the difference
is of no importance for the present purposes. In one dimension, the delta function
is defined by the its property,
 
dxf (x)δ(x) = f (0) , for any continuous function f (x) and for any  > 0 .
−
(3.12)

When f (x) = 1 Eq. (3.12) gives dxδ(x) = 1 and therefore the integral over the
−
delta function must be 1 over any segment that includes the origin. As this equality
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 127

Radiation scattering off crystals 127

Fig. 3.9: The function Si = |Zi |2 for Ni = 7 (thick curve) and Ni = 10 (thin curve).

must hold for any , including  → 0, the delta function diverges at the origin and
vanishes elsewhere (such that its integral over a segment that does not include the
origin vanishes). Similarly, the three-dimensional delta function is defined by

d3 rf (r)δ(r) = f (0) , (3.13)

for any continuous function f (r), when the integration is over a volume which
contains the point r = 0.
Equation (B.27) in Appendix B demonstrates that for an infinite periodic lattice,
 (2π)3 
Z(q) = eiq·R → δ(q − G) , (3.14)
V
R G

where the sum over R contains all lattice points, and the sum over G includes all
points in the reciprocal lattice. This sum, the “three-dimensional Dirac’s comb”,
contains peaks of infinite heights and zero widths that occur at points where the
vector q = k − k equals one of the reciprocal-lattice vectors. In other words, Bragg
peaks, e.g., those shown in Fig. 3.9, become delta functions.
In summary, the scattering amplitude of point-like scatterers is sharply
peaked when the vector q = k − k coincides with one of the reciprocal-
lattice vectors. As each Bravais lattice has its own reciprocal lattice (see below),
monitoring the locations in momentum space where the Bragg peaks appear enables
the unequivocal identification of the lattice structure.
The scattering intensity. The scattering intensity is given by the absolute
value squared of the amplitude, Eq. (3.7), |A|2 = a20 |Z|2 . This intensity (in three
spatial dimensions) is proportional to the product of the three functions Si = |Zi |2 ,
Eq. (3.10), where each of them is
      
Si = |Zi (ui )|2 = eini ui e−ini ui = ei(ni −ni )ui = eini ui = Ni Zi .
ni ni ni ,ni ni ,n
i

(3.15)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 128

128 Introduction to Solid-State Physics

(The summation index ni has been replaced by ni = ni − ni in the last two steps.)
The resulting sum is equal to Zi (as the periodic boundary conditions allow a shift
of the origin), and is independent of ni . Hence,
|A|2 = a20 N Z , Z = Z 1 Z 2 Z 3 , N = N1 N2 N3 , (3.16)
(check!) Equation (3.14) (or its analogue for a finite large sample) implies that
the intensity of the scattered radiation is peaked when the difference of the wave
vectors equals one of the reciprocal-lattice vectors. Note, though, that by Eq. (3.14)
the intensity of all peaks is the same (as the coefficients of the delta functions
are identical). As shown in the following, this result is valid only for point-like
scatterers (one per unit cell). When the unit cell contains more than a single point-
like scatterer, the Bragg peaks do appear at the same locations as for the point-like
scatterers, but the intensity of each of them is different; it thus conveys information
about the scatterers within the unit cell.

3.4 The reciprocal lattice

The von Laue equations. As found in Sec. 3.3, the radiation intensity is peaked
when the momentum difference between the impinging particles and the scattered
ones equals a vector G, which obeys the equation
exp[iG · R] = 1 , (3.17)
where R is the radius vector to any of the lattice points. The vectors G form a
periodic lattice, called the “reciprocal lattice”, whose definition is Eq. (3.17).
For the vector G to obey the condition (3.17), it has to satisfy
G · R = 2πm , (3.18)
where m is an arbitrary integer and R is an arbitrary lattice site. Thus, Eq. (3.18)
is satisfied for any R if it is obeyed for each of the lattice vectors,
G · ai = 2πmi , (3.19)
where (in three dimensions) i =1, 2, 3 and mi is an integer. [Substituting R =
n1 a1 +n2 a2 +n3 a3 , Eq. (2.1), in Eq. (3.18) yields G·R = i ni G·ai = 2π i ni mi .]
Equations (3.19) are identical to those obtained in the discussion following Fig. 3.9.
Combining these equations with the requirement k − k = q = G that follows from
Eq. (3.9) results in (k − k) · ai = 2πmi ; these are the von Laue equations.
Equation (3.19) comprises three equations, each imposing the tip of the vector
G = k − k to lie on a plane normal to the lattice vector ai , such that the projection
of G on that lattice vector is 2πmi /ai . Constructive interference is achieved only
when all three equations are obeyed together, i.e, all three planes cut each other.
Such a joint cut exists only at a discrete number of points. In other words, three
equations for three unknowns (which are the components of the vector G = k − k)
have at most a single solution. For this reason there appear only discrete points on
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 129

Radiation scattering off crystals 129

the monitor, e.g., in the geometry of Fig. 3.3(a). It is worthwhile to note that for
a one-dimensional sample there is only a single plane, while for a two-dimensional
specimen there are two planes. In these cases the diffraction pattern can include
lines in place of points. This issue is revisited in Sec. 3.12, where scattering off
planes is analyzed.
The vectors of the reciprocal lattice. When two vectors, G1 and G2 , obey
Eq. (3.19), their sum, G1 + G2 , fulfills it as well (prove!). It thus suffices to find
three independent vectors obeying Eq. (3.19); then any linear combination of the
three vectors, with integer coefficients, also fulfills that equation. To determine all
these vectors, one needs to find the shortest independent vectors. (When along a
certain direction the vector is not the shortest one, then there exists another vector
pointing along this direction which also satisfies the equation but cannot be written
as a linear combination of the three vectors with integer coefficients!) Denote the
required vectors by b1 , b2 , and b3 . As it is possible to choose any three integer
coefficients, clearly the linear combinations of these vectors create an infinite lattice,
which is the reciprocal lattice.
The shortest vectors that obey the aforementioned equations are found by choos-
ing the smallest values of the integers in Eqs. (3.19), mi =1, 0. A sufficient condition
is then
bi · ai = 2πδij , i, j = 1, 2, 3 , (3.20)
where
1, i=j
δij =
0,i=j
is the Kronecker delta. As the vectors bi obey Eqs. (3.19), they are all vectors in
reciprocal lattice. Any other vector in that lattice is a linear combination
of them, i.e.,
G = hb1 + kb2 + b3 , (3.21)
where h, k, and are arbitrary integers. Indeed, inserting Eq. (3.21) in Eq.
(3.19) yields, e.g., G · ai = hb1 · a1 + kb2 · a1 + b3 · a1 = 2πh, as required by
Eq. (3.19). Moreover, Eq. (3.21) [upon comparing it with Eq. (2.1)] shows that
the collection of vectors G forms a Bravais lattice, with lattice vectors b1 , b2 , and
b3 . Since constructive interference is achieved only for q = k − k = G, and as
the lattice vectors are discrete, then only a discrete collection of scattering angles
or wave lengths gives rise to such interference.
Simple examples. In one-dimension there is only a single equation in (3.20),
ba = 2π, and therefore the length of the reciprocal-lattice vector is b = 2π/a. Note
that the dimensions of the vectors in reciprocal lattice are length−1 (and when
multiplied by , after the de Broglie’s rule, the units are those of momentum; this
is the reason for referring to the reciprocal lattice as momentum space). A similar
result is obtained for the two-dimensional rectangular lattice for which the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 130

130 Introduction to Solid-State Physics

reciprocal-lattice vectors are b1 = (2π/a1 )x̂ and b2 = (2π/a2 )ŷ, or for the three-
dimensional orthorhombic lattice. In the latter case, denoting the original
lattice vectors (of the “direct” lattice in real space) by a1 = a1 x̂, a2 = a2 ŷ, and
a3 = a3 ẑ, the solution of Eqs. (3.20) is
b1 = (2π/a1 )x̂ , b2 = (2π/a2 )ŷ , b3 = (2π/a3 )ẑ . (3.22)
Figure 3.10 displays the rectangular lattice and its reciprocal.

(a) (b)

Fig. 3.10: (a) The rectangular lattice, with lattice constants a1 < a2 . (b) The
reciprocal lattice of the rectangular lattice, with lattice constants b1 = 2π/a1 >
2π/a2 = b2 .

Three-dimensional lattices. The solution of Eqs. (3.20) for the general

three-dimensional lattice is more complex. The vector b1 is perpendicular to both
a2 and a3 ; one may conclude that it is parallel to the vectorial product a2 × a3 .
Similar reasons determine the orientations of the other two lattice vectors. The
lengths of all three is dictated by the additional equation bi · ai = 2π. Assuming
that b1 = βa2 × a3 , then b1 · a1 = β[a2 × a3 ] · a1 , and Eq. (3.20) yields β =
2π/(a1 · [a2 × a3 ]). The volume of the unit cell in the original lattice, from Eq.
(2.2), is V = |a1 · [a2 × a3 ]|. When the sign of the triple product is negative, one has
just to invert the signs of the lattice vectors. It is therefore customary to choose
the reciprocal lattice-vectors as
b1 = 2πa2 × a3 /V , b2 = 2πa3 × a1 /V , b3 = 2πa1 × a2 /V . (3.23)
The reciprocal-lattice vectors are at times defined without the 2π factor, i.e., by the
condition bi · aj = δij . Indeed, the reciprocal-lattice vectors of the orthorhombic
lattice, Eqs. (3.22), obey the more general relations (3.23) (check!).
The connection with the distance between planes. Interestingly enough,
the length of the vector b3 equals the product of 2π and the ratio of the area of
the “base” of the unit cell, S = |a1 × a2 |, and the volume of that cell. Hence,
|b3 | = 2π/d(001), where d(001) is the height of the unit cell, i.e., the distance
between the base plane and the plane parallel to it that contains the opposite face
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 131

Radiation scattering off crystals 131

of the unit cell (which passes through the tip of a3 , Fig. 2.13). Problem 3.2 gives an
example for such a distance. Indeed, the length of each vector in reciprocal lattice is
related to a certain distance among planes in the lattice; this explains the notation
(001) and similar notations that appear in Fig. 3.6.
Two-dimensional lattices. The vectors of the reciprocal lattice of a two-
dimensional lattice, whose base includes the vectors a1 and a2 , are found by
introducing a unit vector normal to the lattice plane, ẑ. The reciprocal-lattice
vectors are then given by b1 = 2πa2 ×ẑ/S and b2 = 2πẑ×a1 /S, where S = |a1 ×a2 |
is the area of the unit cell [Eq. (2.3)].
The triangular lattice. The reciprocal lattice of the triangular lattice is
√ to be a1 = ax̂
found as follows. √As in Fig. 2.4, the (direct) lattice vectors are chosen
and a2 = (x̂ + 3ŷ)a/2. [One may choose also a2 = √(−x̂ + 3ŷ)a/2, but the
physical results are unaltered.]√Hence, S = |a1 × a2 | =√ 3a2 /2, and consequently
b1 = 2πa2 × ẑ/S = 2π(x̂ − ŷ/ 3)/a and b2 = 4πŷ/(a 3). The triangular lattice
and its reciprocal are displayed in Figs. √ 3.11. The reciprocal lattice is triangular
too, with a lattice constant b = 4π/(a 3) (check!). From the scalar product of the
two lattice vectors one finds that the angle between them is 120◦ (as opposed to the
lattice vectors of the original lattice, where the corresponding angle is 60◦ ). As can
be expected, the reciprocal-lattice vectors are perpendicular to those of the original
lattice (or alternatively, rotated relative to them by ±30◦ ).

(a) (b)

Fig.
√ 3.11: (a) The triangular lattice, with lattice vectors a1 = ax̂ and a2 = (x̂ +
3ŷ)a/2. (b) The
√ reciprocal lattice of √
the triangular lattice, with lattice vectors
b1 = 2π(x̂ − ŷ/ 3)/a√and b2 = 4πŷ/(a 3). It is a triangular lattice whose lattice
constant is b = 4π/(a 3).

Problem 3.4.
a. Prove that the reciprocal-lattice vectors of an arbitrary two-dimensional lattice
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 132

132 Introduction to Solid-State Physics

are given by
b1 = 2π[|a2 |2 a1 − (a1 · a2 )a2 ]/S 2 ,
b2 = 2π[|a1 |2 a2 − (a1 · a2 )a1 ]/S 2 ,
where S is the area of the original unit cell.
b. What is the area of the unit cell of the planar reciprocal lattice?
Problem 3.5.
Find the reciprocal of the simple hexagonal lattice (which is built of triangular lattices
placed on top of each other).
Non-primitive cubic lattices. A more complex example is the face-centered
cubic lattice, FCC, whose lattice vectors are given in Eq. (2.6) and in Fig. 2.18.
For this lattice a1 · [a2 × a3 ] = −a3 /4 = −V . Inserting the vectors (2.6) into Eq.
(3.23) yields, e.g., b1 = 2π(a/2)2 [(x̂ + ẑ) × (ŷ + ẑ)]/(−V ) = (2π/a)(x̂ + ŷ − ẑ).
Hence,
2π 2π π
b1 = (x̂ + ŷ − ẑ), b2 = (−x̂ + ŷ + ẑ), b3 = (x̂ − ŷ + ẑ) . (3.24)
a a a
A comparison of these expressions with the ones in Eq. (2.5), shows that they
are the lattice vectors of a BCC lattice, whose cube edge is 4π/a. Hence, the
reciprocal of the FCC lattice is a BCC lattice.
Problem 3.6.
Prove that in three dimensions, the product of the volumes of the two unit cells, the
one of the original lattice and that of the reciprocal one, equals (2π)3 . What are the
analogous relations in one and two dimensions?
Problem 3.7.
What is the reciprocal of the reciprocal lattice?
Problem 3.8.
What is the reciprocal of the BCC lattice? Can one deduce this result without
carrying out the detailed calculation?
As found above, the reciprocal of the orthorhombic lattice is orthorhombic as
well, and so is the case also for the triangular and the hexagonal lattices. The
reciprocal of the cubic FCC and BCC lattices are also cubic (BCC and FCC,
respectively). In all these examples the symmetries of the original lattices are similar
to those of the reciprocal ones. This feature is common to all Bravais lattices as
can be proven by computing the reciprocal lattices of all 14 Bravais lattices (see
also problem s.3.2 at the end of the chapter). Since scattering experiments probe
directly the reciprocal lattice, this property enables the unique identification of the
corresponding Bravais lattice.
Ewald sphere. A geometric picture which clarifies the relations among the
vectors of the impinging and scattered radiations and those of the reciprocal lat-
tice, is displayed in Figs. 3.12. Figure 3.12(a) shows a planar cross section of the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 133

Radiation scattering off crystals 133

reciprocal lattice. Assume that the impinging wave vector k lies on the horizontal
axis in space, and locate its tip on a reciprocal-lattice site, O. A sphere of radius
equal the length of k, i.e., 2π/λ, is drawn around the other tip of k (placed at point
B and marked by ×). This is the Ewald sphere. As the outgoing wave has the
same wave length, its wave vector k connects the point B with another point on the
same sphere (marked by C). The discussions following Eqs. (3.9) and (3.19) imply
that the difference k − k is identical to a reciprocal-lattice vector G. The figure
demonstrates that this difference connects the point O and C. Hence, the scattered
wave that propagates along k produces a point on the screen only when it passes
through a reciprocal-lattice site on the sphere (C in the present example). In other
words, one plots the Ewald sphere for a certain impinging wave k, and identifies
the directions of the scattered radiation that can produce points on the screen, like
in Fig. 3.12(b). The points where these directions cut the sphere are sites of the
reciprocal lattice. Other wave lengths give rise to spheres of different radii, which
contain other sites of the reciprocal lattice. Hence, all points on the screen are
related to reciprocal-lattice sites, and their lot should reflect the symmetry of that
lattice. In particular, when the impinging ray is directed along one of the rotation
axes of the lattice (for instance, of order 4), then the points on the screen possess
this symmetry as well (in the present example, the diffraction pattern does not
change upon rotating the sphere by 90◦ along that axis).

(a) (b)

Fig. 3.12: The Ewald sphere. (a) The vectors of the impinging and scattered waves,
and the reciprocal lattice vector joining them, G = k − k. (b) The diffraction
pattern on the screen. Whenever the scattered wave passes through a reciprocal-
lattice site on the sphere there appears a Bragg peak on the screen (Courtesy of M.
Sawaya).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 134

134 Introduction to Solid-State Physics

3.5 Brillouin zones

As discussed, constructive interference is achieved only when the vectors of the

impinging and scattered waves obey Eq. (3.9), i.e.,
k = k + G , (3.25)
where G is an arbitrary vector of the reciprocal lattice. Squaring this equation and
using the relations k 2 = k 2 = (2π/λ)2 give 2k · G = −G2 , i.e.,
k · Ĝ = −G/2 , (3.26)

where Ĝ is a unit vector along G. The left hand-side is the projection of k on the
direction of G. As this projection is equal to half of the length of G (the sign is
irrelevant), it follows that when the vectors k and k emerge from the same point
O in reciprocal space and end at the opposite tips of G (see Fig. 3.13), then O lies
on a plane normal to G and bisecting it (only vectors emerging from such a point
and ending at the tip of G have projections on it whose length is G/2, Fig. 3.13).
Hence, scattering of nonzero intensity is achieved only when the vector of
the impinging wave (and hence also the vector of the scattered radiation)
emerges from a point on the plane normal to one of the reciprocal-lattice
vectors at its mid point and ends at the tip of that vector.

Fig. 3.13: The wave vectors of the impinging and scattered radiation in the
constructive-interference configuration. G is a vector of the reciprocal lattice, and
the point O from which the two wave vectors emerge is on the normal to it (that
cut it at the middle). Then the projection of each of the wave vectors on G is equal
to G/2, as dictated by Eq. (3.26). The point O may be located anywhere on the
plane normal to G at its mid point.

Once the coordinate origin is assigned to one of the reciprocal-lattice sites, then
reciprocal-lattice vectors emerge from it to all other sites in the reciprocal lattice.
The planes normal to the mid points of the vectors connecting the origin with its
nearest neighbors encompass the volume of the reciprocal lattice that contains all
points closer to the origin than to any other site there. This volume is analogous
to the Wigner-Seitz unit cell, introduced in Chapter 2; it is the unit cell of the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 135

Radiation scattering off crystals 135

reciprocal lattice, that exists in momentum space. This is the first Brillouin
zone. One may repeat this construction for the vectors which join the origin with
its next-nearest neighbors, and thus create the second Brillouin zone, and so forth.
Figure 3.14 displays several Brillouin zones of the two-dimensional square lattice.
Figure 3.14(a) portrays the sites of the reciprocal lattice, and Fig. 3.14(b) shows the
boundaries among the various Brillouin zones. In two dimensions these are lines,
each being perpendicular to one of the lattice vectors. As shown in the solution of
problem 2.2, the Wigner-Seitz cell of the square lattice is a square encompassing
a single lattice point, whose edge length is the lattice constant. Therefore, the
first Brillouin zone of the square lattice is a square whose edge’s length is 2π/a,
enclosing a single reciprocal-lattice site. It is the small square enclosing the origin
in Fig. 3.14(b), indicated by “1”, whose boundaries are lines perpendicular to
the reciprocal-lattice vectors connecting the origin with its four nearest neighbors,
bisecting them at the middle (G = ±b1 or G = ±b2 ). The outer borders of the
second Brillouin zone are the lines perpendicular to the reciprocal-lattice vectors
connecting the origin with the four next-nearest neighbors (G = ±b1 ± b2 ). As
seen, the latter form a square rotated relative to the axes by 45◦ , whose area is
twice that of the first Brillouin zone. Note though, that the second Brillouin zone
contains the lattice sites that are within the large square but are not in the first
zone. Hence, the actual area of the second Brillouin zone, which consists of the
four triangles marked in the figure by “2”, is identical to that of the first Brillouin
zone. Similarly, the lines normal to the vectors G = ±2b1 or G = ±2b2 , which
connect the origin with the third nearest-neighbors, are parallel to the axes, and
together with the lines defined previously contain eight triangles which form the
third Brillouin zone. Its area is equal to that of its two predecessors.

(a) (b)

Fig. 3.14: (a) The reciprocal of the square lattice. (b) Brillouin zones of the square
lattice. The numbers within each area indicate the serial number of the Brillouin
zone to which this area belongs. The straight lines are normal to the lattice vectors
connecting the origin with its neighbors, for instance G = ±2b1 , G = ±2b2 ,
G = ±b1 , G = ±b2 , or G = ±b1 ± b2 .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 136

136 Introduction to Solid-State Physics

Problem 3.9.
Draw the first three Brillouin zones of the triangular lattice.

The construction of the Brillouin zones of the simple cubic lattice resembles the
one carried out for the square lattice. This task becomes more laborious for the
more complicated Bravais lattices. For instance, as shown in Sec. 3.4, the reciprocal
of the BCC lattice is an FCC one, and vice versa. Hence, the first Brillouin zones
of these two lattices are those that appear in Figs. 2.19: the first Brillouin zone of
the BCC in Fig. 2.19(a), and the one of the FCC in Fig. 2.19(b). In both cases
the length of the edge of the cube is 4π/a.
In summary, constructive interference is achieved only when the scat-
tered wave vector emerges from a point on one of the planes that sepa-
rates different Brillouin zones, and ends at the tip of a reciprocal-lattice
vector to which this plane is perpendicular (cutting it at its middle).
The following chapters emphasize other properties of the first Brillouin zone: the
frequencies of the vibrations of the atoms around their equilibrium locations (Chap-
ter 5) or the electronic motion (Chapter 6) are characterized by wave numbers; it
suffices to explore those only for wave vectors in the first Brillouin zone (their values
in the other zones are identical to the ones in the first zone).

3.6 The reciprocal lattice and lattice planes

Planes perpendicular to reciprocal-lattice vectors. As mentioned, there are

several ways to specify the directions along which the scattered waves interfere con-
structively. The relation between these directions and families of lattice planes is
discussed in Sec. 3.2. Section 3.3 connects those directions with the reciprocal-
lattice vectors. These two descriptions are equivalent: each vector in the recip-
rocal lattice is related to a family of parallel lattice planes, and vice versa.
Though the various considerations yield the same result (luckily enough), each is
based on a different point of view. The experimental observations are conveniently
described in terms of scattering angles, but equally convenient is the identification
of each angle that gives rise to constructive interference with the corresponding
lattice vector, and, in addition, the family of lattice planes to which it is related.
The reciprocal-lattice vectors are given in Eq. (3.18). That equation may be
understood in the following manner: given the integer m and the reciprocal-lattice
vector G, then the tips of all vectors R that obey Eq. (3.18) lie on a plane perpen-
dicular to G; hence Eq. (3.18) is an equation of a plane. Indeed, two vectors, R and
R , that obey Eq. (3.18) also satisfy G · (R − R ) = 0 and therefore the difference
R − R is normal to G; it hence lies in a plane normal to G. This is true for all
vectors obeying Eq. (3.18), with the same values of m and of G. Other values of
m represent other planes, all normal to G, and consequently parallel to each other.
This collection of planes is the family of planes normal to G. Suppose that the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 137

Radiation scattering off crystals 137

origin is located on one of these planes, say the plane for which m = 0. The vector
R in any other plane can be written in the from R = d + R⊥ , where d||G and
R⊥ ⊥ G. Equation (3.18) becomes G · d = 2πm. Equation (3.25), k = k + G, is
identical to Eq. (3.5), making the reciprocal-lattice vectors identical to the vectors
of momentum difference as given by the laws of von Laue and Bragg (see Sec. 3.2).
The distance between neighboring planes. There are many reciprocal-
lattice vectors that are normal to a certain family of planes. Each of the thick arrows
in Fig. 3.6 represents the direction of a vector normal to each of the families of planes
displayed there. There are numerous reciprocal-lattice vectors pointing along this
direction, since nG (where n is an integer) is also a vector in the reciprocal lattice.
It is customary to characterize the family of planes by the shortest reciprocal-lattice
vector of this type, denoted G0 . Since Eq. (3.18) can be written in the form G0 ·R =
2πm, it is clear that successive values of m represent successive planes in that family.
The distance between the planes corresponds to m = 1, G0 d = G0 · R = 2π (recall
that d||G0 ), and therefore the distance between neighboring planes is
d = 2π/G0 . (3.27)
This result is equivalent to Eq. (3.6). Bragg’s law, Eq. (3.4), can hence take the
form
sin θ = nG0 λ/(4π) = Gλ/(4π) . (3.28)
Measurements of the angles for which the scattered waves interfere con-
structively yield the “allowed” values of the length of vectors in the
reciprocal lattice (that give rise to constructive interference). These
lengths are equivalent to the “allowed” distances between planes in the
families of planes that exist in the crystal. Calculations of angles correspond-
ing to Bragg peaks for various lattices are presented below, e.g., at the beginning
of Sec. 3.7.
Miller indices. Each of the vectors G in the reciprocal lattice is specified
by three integers, h, k, and , Eq. (3.21). The corresponding family of planes is
characterized by the shortest vector, G0 , which is also given by such three integers.
Consider now two neighboring planes in the family, of which one passes through the
origin and the other is a nearest neighbor parallel to it. The equation of this second
plane is G0 · r = (hb1 + kb2 + b3 ) · r = 2π [the discrete points R are replaced
by all (continuous) points on the plane, denoted by r]. This plane cuts the lattice
vectors of the original lattice. Denote the coordinates of these cuts by the vectors
x1 a1 , x2 a2 , and x3 a3 . Inserting each of these into the equation of the second plane,
and using Eq. (3.20), yields
x1 h = x2 k = x3 = 1 . (3.29)
The three numbers h, k, and (that belong to G0 ) therefore characterize the cuts
of the planes with the major directions in the lattice (i.e., along the
lattice vectors). In particular, the lattice vector a1 is cut by h planes that belong
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 138

138 Introduction to Solid-State Physics

to the family (for which x1 = 1/h, 2/h, . . . , 1), etc. Consequently, the family of
planes is identified by (hk ). The three numbers h, k, and are called the indices
of Miller. Examples for lattice planes are displayed in Fig. 3.6. In the upper part
of Fig. 3.6(a) appear lines (in a two-dimensional planar lattice these replace the
planes of the three-dimensional lattice). For instance, the lines in the upper left
part of Fig. 3.6(a) cut the axes at a1 and a2 , and therefore are marked by (11). The
lines in tFig. 3.6(a) are parallel to the lattice vector a2 , and therefore cut this axis
only at infinity; hence k = 1/x2 = 0. These lines cut the a1 −axis at x1 = 1, and
hence are marked as (10). Similar considerations apply to the other lines. When
one of the numbers, e.g., x2 , is negative, one adds a minus sign above the respective
index, for example, (4, −1) ⇔ (4, 1) (as in the lower left part of the rectangular
lattice in the figure). Figure 3.6(b) also displays a family of planes for the three
cubic lattices. In Sec. 3.2, and in particular in problem 3.2, several such planes are
identified. The reader is welcome to examine the Miller indices of those.
Problem 3.10.
The equations which describe a certain family of planes in the BCC lattice are
discussed in problem 3.2, where the distances between the planes are also calculated.
Identify the reciprocal-lattice vector that corresponds to this family, and verify that
the results obtained in the solution of that problem are identical to the ones expressed
in terms of this vector.
The density of sites in the plane. The specific volume of a primitive Bravais
lattice (the volume per each lattice point) is that of the unit cell, V . Since the
distance between neighboring planes is d = 2π/G0 , the planar density of points
(per unit area) in each plane is d/V = 2π/(V G0 ). Hence, as the reciprocal-lattice
vector becomes longer, the distance between nearest-neighbor planes is shortened,
and the density of points in each plane decreases. Similar conclusions pertain to the
densest lines in a planar lattice. Several such densities are computed in problem 3.2.
The reader is asked to confirm (or else) those results by exploiting the expressions
given here.
Multiplicity of plane. Different planes may be related to each other by sym-
metry operations, i.e., one is obtained from the other by applying one of the sym-
metry operations of the lattice. For instance, the plane (hk ) in the cubic lattice
is equivalent to the planes for which the indices are interchanged, e.g., (h k) or
(kh ). Likewise, each of the eight planes (±h, ±k, ± ) is also similar to that plane.
The family of all those equivalent planes is denoted {hk }. Symmetry considera-
tions imply that all planes in such a family yield the same scattering angles. The
number of triplets (hk ) which gives rise to the same scattering angles is called the
multiplicity of the {hk } family of planes.
Problem 3.11.
a. Find the multiplicity of the following families of planes in a cubic lattice: {00k},
{hk }, {hh }, {hhh}, {k 0}, {kk0}, where h, k, and are all different and nonzero.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 139

Radiation scattering off crystals 139

b. Plot all planes close to the origin in a cubic lattice, that belong to the {111}
family. What shape do they form?
c. Describe all planes close to the origin in a cubic lattice, that belong to the {110}
family. Determine the shape created by them.
d. Find the maximal multiplicity of different planes that belong to the same family.

Problem 3.12.
An alternative way to obtain the Miller indices is the following. Define a plane by
the points where it cuts the three lattice vectors, a1 /h, a2 /k, and a3 / , where h, k,
and are three integers. Prove that the vector G = hb1 + kb2 + b3 is normal to
this plane, and that the distance of the plane from the origin is 2π/G.

Problem 3.13.
A family of planes in the BCC lattice is related to the reciprocal-lattice vectors
G = hb1 + kb2 + b3 , where b1,2,3 are the lattice vectors found in problem 3.8.
Show that these vectors can be represented in terms of the reciprocal-lattice vectors
of the SC, e.g., those given in Eq. (3.22), b1 = (2π/a)x̂, b2 = (2π/a)ŷ, and
b3 = (2π/a)ẑ, G = h b1 + k  b2 +  b3 , provided that the new indices {h k   }
cannot assume all integral values. What are the limitations on these new indices?

The hexagonal lattice. The planes of the hexagonal lattice can be described
in terms of the three Miller indices from Eq. (3.21), with the reciprocal-lattice
vectors given in Eq. (3.18). The literature at times describes these planes with
four indices, (ijk ). The fourth index represents the cuts of these planes with the
ẑ−axis, like the third index in the previous description. The first three indices
represent the cuts of the planes with three axes in the plane, with angles of 120◦
in-between them (see Fig. 3.15). These three indices are not independent, e.g.,
k = −i − j. Nonetheless, the characterization with four indices is based on three
symmetric axes, and thus has certain advantages.

3.7 Experimental methods

The scattering angles and lattice planes. Inserting Eq. (3.27), d = 2π/G, into
the Bragg law, Eq. (3.4), and making use of Eq. (3.21) yield
sin θ = λ|hb1 + kb2 + b3 |/(4π) . (3.30)
The angle θ is between the wave vector of the impinging (or of the scattered)
radiation and the scattering planes. Recall that the scattering angle is defined
as 2θ (see Figs. 3.3 and 3.7). As the lattice vectors of the reciprocals of the 14
Bravais lattice are known (see Sec. 3.4 and problem s.3.2), one may use them in
Eq. (3.30), and obtain the scattering angles for various planes (hk ), for which the
interference is constructive. Alternatively, when the scattering angle is fixed but the
wave vector of the radiation is varied, one may identify the discrete wave lengths
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 140

140 Introduction to Solid-State Physics

(a) (b)
Fig. 3.15: (a) The four directions in the simple hexagonal lattice that define the four
Miller indices for this lattice. (b) The (2200) plane, that is parallel to the ẑ−axis
and to the a3 −axis, and cut the a1 and a2 axes at the middle of the corresponding
lattice vectors.

that give rise to constructive interference. A comparison of the experimentally

measured values (of the angles or of the wave lengths) with those calculated for
each of the lattices (whose number is finite) enables the unique determination of
the crystalline structure and its lattice constants. Examples for the calculation of
the scattering angles, or alternatively, the determination of the crystal structure
from its diffraction pattern, are discussed below.
Scattering off a simple-cubic lattice. The reciprocal lattice vectors of the
SC lattice are G = (2π/a)(hx̂ + kŷ + ẑ) and therefore
λ2 2
sin2 θ = (h + k 2 + 2 ) . (3.31)
4a2
For a monochromatic wave, of a single wave length λ, and for a given lattice (whose
lattice constant is a), constructive interference is achieved only for discrete values
of θ, that correspond to certain triplets h, k, and . For instance, consider the
range of values h2 + k 2 + 2 ≤ 24. (This range is chosen arbitrarily to demonstrate
the disparity among the cubic crystals; it is relevant for wave lengths such that
λ < 2a/5, i.e., sin2 θ ≤ 1.) In this range, the combination (h2 + k 2 + 2 ) can attain
all integers between 1 and 24, except for 7, 15, and 23 [see problem 3.14)]. In
practice, one prepares a list of the measured scattering angles. For a SC lattice,
the smallest angle is given by h2 + k 2 + 2 = 1, i.e., sin θmin = λ/(2a), and the other
angles are then given by
sin2 θ/ sin2 θmin = h2 + k 2 + 2
, (3.32)
where the right hand-side assumes only the aforementioned allowed values. When
the measurement yields the allowed values (and numbers higher than 24) for the
ratio on the left hand-side, then one may conclude that the measured crystal is a
SC. The value of the minimal θ yields the lattice constant a. For the treatment of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 141

Radiation scattering off crystals 141

other lattices, see problem s.3.2. As shown in Sec. 3.9, in other lattices of a cubic
symmetry (e.g., BCC, FCC, or diamond) the scattering intensity from certain
families of planes may vanish because of destructive interference (among the rays
scattered from different base points); this in fact enables the identification of the
internal structure of the unit cell.
Scattering off powders. The experimental technique to monitor all scattering
angles in a single measurement is based on scattering off powders. In this method,
illustrated in Fig. 3.3(b), a monochromatic radiation, of a single wave length,
is scattered off a powder that consists of many grains directed at random. Each grain
is large enough to provide a good diffraction picture, but is also sufficiently small so
that there are many grains in the sample. For this reason, it is safe to assume that
there are enough grains to scatter the impinging beam along all allowed directions;
this means that the points that appear on the screen in Fig. 3.3(b) are due to
all allowed angles according to Eq. (3.30). Since the lattice vectors in each grain
are oriented at random relative to the direction of the beam (and to the directions
of the other grains), the points on the screen are at different locations on circles
determined by those angles. As a result, the pattern displayed on the screen is that
of continuous circles, as in Fig. 3.3(b) or in Fig. 3.4. In other words, given the
direction of the wave vector of the impinging radiation, k, the wave vector of the
scattered radiation, k , lies on a cone, whose axis is along k and its aperture angle
is one of the allowed values of the scattering angle 2θ. The circles in Fig. 3.4 are
created by the cut of the cone with the screen. In the configuration of Fig. 3.3(b),
where only the forward-scattered radiation is examined, the circles appear only for
2θ < π/2. This condition limits the values of G which are available for the analysis.
A less stringent condition pertains to the geometry of Fig. 3.3(a): there 2θ < π.
Multiplicity and intensity. Equation (3.31) implies that for a cubic lattice,
the scattering angle depends solely on the combination h2 +k 2 + 2 . That is, the same
scattering angle corresponds to all planes belonging to the family {hk }, as defined
in the previous section. When the scattering is off a powder, all possible planes
contribute to it; as a result, the intensity for each scattering angle is proportional
to the number of planes belonging to the family (assuming point-like scatterers).
For instance, the {100} family contains six planes, the {110} one contains 12, and
the {211} family comprises 24 equivalent planes. It is therefore expected that the
scattering intensities at angles corresponding to these indices have the ratios 1:2:4.
When the combination h2 + k 2 + 2 increases there might appear triplets of different
Miller indices that give the same scattering angles; for example, the triplets (221)
and (300). All reciprocal-lattice vectors whose tips lie on a sphere whose equation
is |G|2 = (2π/a)2 (h2 + k 2 + 2 ) and its center is at the origin, give rise to scattering
at the same angle.

Problem 3.14.
Determine the scattering angles and their multiplicities for a SC lattice whose lattice
constant is trice the wave length of the radiation. [Consider only forward scattering,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 142

142 Introduction to Solid-State Physics

like in Fig. 3.3(b)].

Other factors that affect the scattering intensities. Beside the multiplic-
ity, the intensity of the scattered radiation at each of the Bragg peaks is sensitive
to other factors.

• As mentioned at the end of Sec. 3.3, the internal structure of the unit
cell (rather than scattering off point-like scatterers) affects the intensity.
This issue is discussed in Sec. 3.9.
• In a finite specimen the number of grains is finite and therefore it is not
obvious that the full multiplicity is reached for each angle. At times, the
intensity is multiplied by a factor that takes into account the possible pref-
erence of certain grains’ directions (this may happen for grains of specific
geometrical shapes, a fact which restricts their random arrangement in the
powder).
• Each of the grains has a finite size, and therefore the Bragg peaks attain
a finite width (as explained in Sec. 3.3, e.g., Fig. 3.9). The intensities to
be compared with each other are determined by their integrals under each
peak (and not only their maximal heights).
• There are two possible polarization directions for the electromagnetic
fields of the X-rays, in the plane perpendicular to k. Assuming that the
radiation is unpolarized, then the intensity has to be averaged over the two
directions.
• As the radiation penetrates the sample, its intensity gradually decays due
to inelastic scattering of the photons (or the scattered electrons) by the
material electrons. (The photons collide with the latter, transfer them to
higher energy levels and lose part of their energy in the process; therefore
their wave length is modified.) This effect adds a multiplicative factor to
the intensity of the elastically-scattered radiation that decays exponentially
with the path each photon traverses in the material (which also depends
upon the angle θ).

It follows that the extraction of the “pure” intensity to be compared with the
calculated one is not a simple task and requires special experimental dexterity.
This section focuses on the analysis of the scattering angles; these allow for the
determination of the primitive lattice of the measured material. The factors that
determine the intensity of the scattering, which carry the information on the internal
structure (i.e., the base of the unit cell), are considered in the following.

Problem 3.15.
a. Derive equations that replace Eq. (3.30) for the orthorhombic lattice and for the
tetragonal one.
b. X-rays, obtained from the transitions Kα1 and Kβ1 of copper (problem 3.1), are
scattered off a powder of a tetragonal lattice whose lattice constants are a1 = a2 =
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 143

Radiation scattering off crystals 143

2Å, a3 = 3Å. Find the smallest eight scattering angles for which the interference is
constructive.

Problem 3.16.
a. The smallest scattering angle measured in a scattering experiment off a powder
of a SC lattice is 40◦ . How many scattering circles are detected? At which angles?
Assume that each unit cell contains a single point-like scatterer, and that the scat-
tering intensity of all of them is the same [Eq. (3.14)]. Find the scattering intensity
in each ring (accounting only for the multiplicities, and not for the other factors
that affect the intensity).
b. As the temperature decreases, the lattice described in part (a) becomes tetragonal,
with the ratio a3 /a2 = 1.1 (the lattice constants a1 = a2 = a remain the same).
How are modified the scattering angles and the intensities?
c. Describe qualitatively the scattering angles and the intensities when there is an
additional transition, to the orthorhombic structure.

Measurement by rotating a single crystal. Another experimental method

exploits monochromatic radiation, of a single wave length; the scattering angle
is varied by rotating the crystal. Constructive interference is achieved only for
angles for which there exist suitable families of planes. The setup is displayed
in Fig. 3.16: the beam impinges normal to the axis around which the crystal is
rotated, and the scattered beam is collected on a cylindrical screen encompassing
the rotation axis. Denoting that axis by x̂, and the direction of the incoming beam
by ẑ, then the vector G representing a family of planes parallel to the rotation axis
is perpendicular to that axis. Thus the vectors k and G are in the YZ plane. It
then follows that the vector k = k + G lies in the same plane, and therefore there
appears a point on the screen whenever the scattering angle fulfils the Bragg law,
i.e., sin θ = λG/(4π). When the planes are inclined by an angle γ relative to the
rotation axis, then the vector G forms an angle 90◦ − γ with that axis; hence its
x−component is fixed, Gx = |G| sin γ, independent of the scattering angle. The
x−component of k = k + G is fixed as well, and all radiation points resulting from
these planes are on a plane parallel to the YZ planes. The distances between the
planes can be deduced from the scattering angles.
The von Laue method. A third experimental method, named after von
Laue, uses a single crystal irradiated by a beam comprising a continuum of wave
lengths, spanned over several angstroms. The scattered beam hits a screen, like
in Fig. 3.3(b). For an arbitrary impinging direction it may well happen that a
monochromatic beam will not obey the Bragg law. However, when the radiation
consists of many wave lengths, there is a finite probability that one (or several) of
them will give rise to constructive interference. Therefore, as the crystal is rotated
radiation points appear on the screen, due to various vectors in the reciprocal lat-
tice. These reflect the symmetry of the reciprocal lattice, and, in turn, that of the
original lattice. Consequently, one may ensure that the crystal is placed along a
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 144

144 Introduction to Solid-State Physics

Fig. 3.16: Diffraction pattern obtained by rotating a crystal placed on the axis of
a cylinder. The radiation impinges normal to the axis, and is scattered onto the
cylindrical screen.

direction that corresponds to a certain symmetry operation. This method is indeed

mainly exploited to identify the symmetries of the crystal and is not usually used
for precision measurements of the structure.

Fig. 3.17: Diffraction pattern of table salt. [D. Halliday, R. Resnick, and K. S.
Krane, Physics, vol. 2, 5th edition, John Wiley and Sons, 2002.]

Figure 3.17 displays a typical diffraction picture, obtained by scattering X-rays

off a table-salt crystal (which has an FCC structure) using the von Laue method.
The impinging ray is parallel to a rotation axis of 90◦ , and the diffraction pattern
reflects this symmetry. A detailed analysis of the radiation points for this structure
is presented in Sec. 3.9.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 145

Radiation scattering off crystals 145

3.8 Fourier series and reciprocal lattice

Periodic functions on the crystal. Leaving the physical discussion aside for a
little while, this section considers the properties of functions that have the pe-
riodicity of the lattice. Appendix B surveys some formal aspects of the Fourier
theorem; the implications of this theorem are exploited here. As mentioned in
Chapter 2, any physical property f (r) of the periodic crystal is periodic as well.
That is, it has identical values at points whose distance from each other is one of
the lattice vectors,
f (r + R) = f (r) , (3.33)
for any lattice vector R = n1 a1 + n2 a2 + n3 a3 . To apply this condition to a finite
crystal, it is assumed that the crystal obeys periodic boundary conditions (those of
Born and von Karman), and hence properties on a certain face of the specimen are
identical to those on the opposite face; this assumption ensures that Eq. (3.33) is
satisfied also near the surfaces [see also the discussion preceding Eq. (3.11)]. The
results obtained under these circumstances are valid for very large crystals, but
do not pertain to nanometric ones (for which one may not ignore the effect of the
surface). The periodic boundary conditions imply that the specimen repeats itself
time and again and therefore they are equivalent to its periodic continuation over
the entire space.
Fourier series. Consider the function Φk (r) = exp[ik · r], which represents a
plane wave in space, with wave vector k. For this function to satisfy Eq. (3.33),
it should be such that Φk (r) = Φk (r + R) = exp[ik · (r + R)] = Φk (r) exp[ik · R].
This function has the periodicity of the lattice provided that exp[ik · R] = 1 for
any lattice vector R. As defined in the context of Eq. (3.9), this condition requires
that the wave vector k coincides with one of the reciprocal-lattice vectors G [see
also Eq. (3.17)]. The generalized Fourier theorem states that the collection of
functions {ΦG (r) = exp[iG · r]} spans the space that comprises all functions
which have the periodicity of the lattice (3.33); i.e., each such function
can be written as a series,

f (r) = f(G)eiG·r , (3.34)
G

with coefficients f(G) determined by the function, and the sum runs over all
reciprocal-lattice vectors. This series is named after Fourier. The equation is
identical to Eq. (B.24) in Appendix B, where it is obtained in a different way.
The coefficients are given by Eq. (B.25); another possibility to derive them is by
multiplying both sides of Eq. (3.34) by exp[−iG · r] and then integrating over a
unit cell, of volume V ,
  
3
d rf (r)e −iG ·r
= 
f (G)

d3 rei(G−G )·r . (3.35)
V G V
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 146

146 Introduction to Solid-State Physics

When G = G the integrand on the right hand-side is unity, and the integral yields
V . In contrast, when G = G , the integral vanishes. This can be shown as follows.
Since the integrand is periodic and satisfies Eq. (3.33), the integral is unchanged
when the unit cell is shifted by an arbitrary vector d. This shift is equivalent to a
shift in the integration variable by d, without changing the integration range,
  
3 i(G−G )·r 3 i(G−G )·(r+d) i(G−G )·d 
d re = d re =e d3 rei(G−G )·r .
V V V
 
As d is arbitrary and G = G , the factor exp[i(G − G ) · d] is not equal to 1, and
the integral must vanish. Hence,


d3 rei(G−G )·r = V δG,G , (3.36)
V

where δG,G is the Kronecker delta function [mentioned after Eq. (3.20)]. It follows
that the sum on the right hand-side of Eq. (3.35) includes just the term G = G ,
i.e.,

1
f(G) = d3 rf (r)e−iG·r , (3.37)
V
just like Eq. (B.25). Thus, in order to obtain the coefficients of the Fourier
series of a function periodic on the crystal it suffices to compute the
integral (3.37) on a single unit cell.

Problem 3.17.
A function f (r) is periodic on a planar square lattice, whose lattice constant is a.
The function equals 1/b2 within the square 0 ≤ x ≤ b and 0 ≤ y ≤ b and zero
otherwise, where b ≤ a. Find the Fourier series of this function. What is the
answer in the limit b → 0?

Examples. The length of the reciprocal-lattice vector in one dimension is

b = 2π/a. It follows that the coefficients in the Fourier series of a one-dimensional
periodic function, f(q), are nonzero only for q = G = 2π /a, where is an arbitrary
integer, and Eq. (3.37) becomes
 a

f ( ) = (1/a) dxf (x)e−2πix/a . (3.38)
0

Thus, the Fourier series of the periodic function f (x) is

∞

f (x) = f( )e2πix/a . (3.39)
=−∞

These are precisely the expressions of the Fourier theorem in one dimension, Eqs.
(B.2) and (B.4) in Appendix B. A similar result is obtained for the rectangular
lattice in two dimensions, or for the orthorhombic lattice in three dimensions.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 147

Radiation scattering off crystals 147

3.9 The structure factor

Several scatterers in the unit cell. So far only scattering off point-like scatterers
is considered, for which all Bragg peaks have the same intensity, e.g., Eq. (3.14).
In reality, X-rays are scattered mostly by the electrons within the atoms while
neutrons are more sensitive to the nuclei density and/or to the magnetic moment
of each atom. Even when the unit cell contains a single atom, its electrons are
not confined to a single point and therefore the assumption of a point-like scatterer
is not entirely correct. Likewise, the nuclei occupy a finite volume and are not
precisely confined to the center of the atom. It follows that Eq. (3.7) has to be
modified. When the scatterers are all identical (e.g., they are all electrons) then
their scattering amplitudes are also the same; denote that amplitude by a0 exp[iψ0 ].
In addition to the summation over all unit cells, as carried out in Eq. (3.7), it is also
required to sum over all contributions to the amplitude A of the scattered wave,
coming from each of the scatterers within each unit cell,
 
A = a0 eiψ0 ei(k −k)·(R+ri ) , (3.40)
R i
where ri is the location of the i−th scatterer in the unit cell, relative to the lattice
point R that determines the origin there.
The structure factor. Equation (3.40) can be written as
 
A = a0 eiψ0 eiq·R eiq·ri = a0 eiψ0 Z(q)F (q) , (3.41)
R i
where q = k − k, and Z(q) = R exp[iq · R] is defined in Eq. (3.8). The new
function,

F (q) = eiq·ri , (3.42)
i
is a sum over the locations of the scatterers in the unit cell, and therefore carries
the information on the structure of the base contained in it. This function is called
the structure factor. In Sec. 3.3 it is shown that the function Z(q) is sharply
peaked when q approaches one of the reciprocal-lattice sites, i.e., when q → G.
These peaks turn into delta functions for the infinite lattice [Eq. (3.14)]. Hence,
when the unit cell possesses an internal structure that is not point-like,
the diffraction pattern is still peaked at the reciprocal-lattice points.
The structure factor, though, adds a dependency of the scattering intensity on the
reciprocal-lattice vector, |F (G)|2 , so that each Bragg peak has a different inten-
sity. As demonstrated below, there are even cases where the contributions of the
various constituents within the unit cell to the structure factor (3.42) for a cer-
tain reciprocal-lattice vector cancel each other, and the intensity approaches zero.
In such cases the waves scattered from the various elements in the unit cell inter-
fere destructively. Monitoring the scattering intensities at many -reciprocal-lattice
vectors provides information on |F (G)|2 , that enables one to deduce the scatter-
ers’ locations {ri } and thus the internal structure of the base. As the number of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 148

148 Introduction to Solid-State Physics

those scatterers increases more data points must be collected from a larger number
of reciprocal-lattice vectors. The spatial distribution of the electrons within each
atom induces a decay of |F (G)|2 as |G| becomes larger (see below). It follows that
measurements at large G s necessitate high intensities of the incoming radiation (to
ensure sufficient intensities for large reciprocal-lattice vectors).
Integral representation. A more general representation of the scattering
amplitude, that allows for a continuous volume-density of scatterers, is derived
by transforming the sums in Eq. (3.40) into a continuous integral over the entire
volume of the crystal. Suppose that an infinitesimal volume-element d3 r around an
arbitrary point r contains n(r)d3 r scatterers (e.g., electrons or nucleons). Equation
(3.40) then becomes


A = a0 eiψ0 d3 rn(r)ei(k −k)·r , (3.43)

where the integral is carried out over the crystal volume N V , and where exp[i(k −
k) · r] represents the phase difference between the wave scattered from the point
r and the one scattered from the origin, see Fig. 3.18. The function n(r) is the
scatterers’ density at the point r. When the unit cell contains a single atom that
contains z electrons, and when it is assumed that all electrons reside at the center
of the atom, then the density function is a sum over delta functions, [or a “comb”
of delta, Eq. (B.26)],

n(r) = zδ(r − R) . (3.44)
R

Substituting this in Eq. (3.43), and making use of the main property of the delta
function [Eq. (3.13)], yield A = za0 exp[iψ0 ] R exp[i(k − k) · R]; this is the
scattering amplitude of a point-like scatterer, Eq. (3.7). Similarly, substituting
n(r) = R,j zδ(r − R − rj ) reproduces Eq. (3.40).

Fig. 3.18: Scattering off an arbitrary point in a continuous specimen. The phase

difference between the wave scattered from the origin, 0, and the one scattered from
the scatterers located within a volume d3 r around the point r (both within the spec-
imen) is (k − k) · r and the total amplitude of these scatterers is a0 exp[iψ0 ]n(r)d3 r.

The scattering amplitude and Fourier series. In a periodic lattice the

density function n(r) follows the periodicity (of the lattice), and therefore it satisfies
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 149

Radiation scattering off crystals 149

Eq. (3.33),

n(r + R) = n(r) . (3.45)

Divide the integration volume in Eq. (3.43) into N unit cells of the sample, and
write down the integral as the sum of integrals over each of them, which contains
the lattice point R. The result is
  
  
d3 rn(r)ei(k −k)·r = d3 rn(r − R)ei(k −k)·(r−R) ei(k −k)·R
NV R V (R)



= Z(k − k) d3 rn(r)ei(k −k)·r ,
V (0)

where the first step uses Eq. (3.45), and in the second each member is multiplied
and divided by exp[i(k − k) · R]. By translating the origin in each of the integrals
within the square brackets, it is found that they are all equal to each other; this
results in the last step of the calculation. Substituting in Eq. (3.43) reproduces
then Eq. (3.41), with the structure factor given by

F (q) = d3 rn(r)eiq·r . (3.46)
V

As in Eq. (3.41), the factor Z(q) is nonzero only when q = k − k approaches a

reciprocal-lattice vector, q → G. It therefore suffices to find the structure factor at
reciprocal-lattice sites, F (G). Equation (3.45) implies that n(r) obeys Eq. (3.33)
and hence it satisfies Eq. (3.34) as well, n(r) = G n (G) exp[iG·r], with coefficients
given by Eq. (3.37),

1 F (−G)
(G) =
n d3 rn(r)e−iG·r ≡ . (3.47)
V V V
Therefore, the structure factor is proportional to the Fourier coefficient
of the scatterers’ density. When the scattering is off discrete points in the unit
cell, the integral in Eq. (3.47) becomes a discrete sum, precisely as in Eq. (3.42).
The correlation function. The intensity of the scattered radiation of
the Bragg peaks at k − k = G is given by the absolute value squared of the
amplitude, Eq. (3.43). Indeed, for k − k = q = G,
 

|A|2 = a20 d3 rn(r)eiG·r d3 r n(r )e−iG·r
N V  NV

= a20 d3 r d3 r n(r)n(r + r )e−iG·r .
NV NV

The periodic boundary conditions of Born and von Karman are used in the second
step to interchange the integration variable, r → r = r − r. This expression can
be written as

|A|2 = a20 (N V )2 Γ(G) , (3.48)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 150

150 Introduction to Solid-State Physics

where


Γ(G) = [1/(N V )] d3 rΓ(r) exp[−iG · r]
NV

is the coefficient in the Fourier series of the average correlation function, defined
by

1
Γ(r ) = d3 rn(r)n(r + r ) = n(0)n(r ) . (3.49)
N V NV
(Here the integrations are carried out over the entire sample; this enables the con-
sideration of samples with no inner periodicity, like liquids or gases.) The average
correlation function gives the average probability per unit volume to find a scat-
terer at r when it is known that there is a scatterer at r = 0. The angular brackets
represent an average over the origin location [the integral in Eq. (3.49) implies
a summation over all possible origins, and a division by their total number, and
therefore it represents an average over them]. The correlation function is discussed
in Chapter 1, and Fig. 1.2 illustrates the disparity among the correlation functions
of a gas, a liquid, and a solid. Measuring radiation scattered off various materials
yields the Fourier transform of the correlation function, and thus provides informa-
tion on the specific phase of the material and its structure. In particular, a periodic
solid (and also a quasi-periodic one) gives discrete Bragg peaks, as opposed to gases,
liquids, and glasses, which yield a continuous diffraction pattern with broad peaks.
The Fourier coefficient of the scatterers’ density is

1
(q) =
n d3 rn(r)e−iq·r , (3.50)
N V NV
and therefore |A|2 = a20 (N V )2 |
n(q)|2 . This result also follows from the relation

Γ(q) = | 2
n(q)| . The Fourier coefficients of the correlation function are the absolute
values squared of the coefficients of the scatterers’ density function. This can be
seen directly from Eq. (3.49), which states that the average correlation function
is a convolution of the density function with itself, Γ(r) = n(r) ⊗ n(r), see
Eq. (B.12).
Scattering off a single atom in the unit cell. Suppose that each unit cell
contains a single atom. This atom is located at the center of the Wigner-Seitz
unit cell; the density function n(r) is thus expected to decay away from that center.
For instance, in the ground state of the hydrogen atom, the quantum probability
density to find the electron at a distance r from the nucleus is n(r) = |ψ(r)|2 =
exp[−2r/aB ]/(πa3B ), where aB = 2 /(me2 ) = 0.529Å is the Bohr radius. Such an
exponential decay is obtained also for other atoms, with an atomic radius of the same
order of magnitude. Because of this strong decay outside the Wigner-Seitz cell, the
integral in Eq. (3.47) can be approximated by an integral carried out over the entire
space, and not only within the cell. (In more precise computations one begins from
the electrons’ distribution within the unit cell, takes into account the potentials of
the other atoms in the lattice and the boundary conditions on the faces of the cell–
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 151

Radiation scattering off crystals 151

and then calculates the integrals only within the unit cell.) When the atomic radius
is significantly smaller than the size of the unit cell, then r  aB  a ∼ 2π/G
(using the fact that the length of the reciprocal-lattice vector is of the order of
2π/a, where a is one of the lattice constants). Under these circumstances, the
oscillating term exp[−iG · r] is roughly 1, and consequently the integral in Eq.
(3.47) gives the total number of electrons in the atom, z. The structure factor
becomes independent of the reciprocal-lattice vector G, and the intensities at the
points of constructive interference are identical (up to the multiplicity). This is the
case of scattering off point-like scatterers, analyzed in Sec. 3.3. Nonetheless, as |G|
increases, the factor exp[−iG·r] oscillates more wildly within the atomic radius, and
the integral decreases (inverse signs of this factor cancel each other’s contribution).
These oscillations diminish the contribution to the integral from far-away regions,
and validate the integration over the entire space. As the scattering angle increases
[and with it the length of the vector G, see Fig. 3.12(a)], the intensity of the
radiation points collected on the screen decreases. Hence, when many points with
not-too-small intensities are needed for extracting information about the density
function, the intensity of the impinging radiation has to be increased. For instance,
good scattering data off complicate molecules, e.g., ribosome, located at lattice
points, requires the use of radiation emitted from a synchrotron.

Problem 3.18.
a. Prove that the structure factor of a charge distribution symmetric under reflec-
tion, n(r) = n(−r), is real.
b. Prove that for a spherically-symmetric electrons’ distribution in an atom
(or molecule), and when the integration is over the entire space, the structure
factor depends solely on the length of the vector G, and is given by F (G) =
∞
(4π/G) 0 dr r n(r) sin(Gr).
c. Find the structure factor for the ground state of the hydrogen atom. Determine
the way the answer is modified for the helium atom (ignoring the interaction between
the two electrons).

Scattering off a crystal with a base. When each unit cell contains a base
with nB atoms (or ions) [Eq. (2.4)], the electrons’ density in the unit cell is a sum,
nB
n(r) = i=1 ni (r − ri ), where the function ni (r − ri ) is the density of electrons
around the i−th atom, located at ri . Inserting this sum into Eq. (3.47), and
displacing the origin in each of the summands to the center of the i−th atom (the
integration variable in each of the summands is changed from r to r − ri ) yields
 
nB 
nB
F (G) = d3 r ni (r − ri )eiG·r = Fi (G)eiG·ri , (3.51)
V i=1 i=1

where Fi is the structure factor of the i−th atom,


Fi (G) = d3 rni (r)eiG·r , (3.52)
V
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 152

152 Introduction to Solid-State Physics

and the integration bounds were shifted to the Wigner-Seitz cell encompassing the
i−th atom. As in the preceding discussion, it is plausible that the density ni (r)
decays with the distance form the center of the i−th atom, and therefore the inte-
gration region can be extended to cover the entire space (though this approximation
is not really necessary here). Equation (3.51) contains the structure factors of all
atoms in the base of the crystal and their locations within the unit cell. As the
scattering intensity is related to the reciprocal-lattice vector G (which in turn de-
termines the structure factor) then when enough data points are collected, it is
possible in principle to identify the atoms in the base and their locations.
Scattering off a rectangular-centered lattice. A simple two-dimensional
example is that of the rectangular-centered lattice, described by the lattice
vectors a1 = a1 x̂ and a2 = a2 ŷ, with a base comprising two identical atoms at the
origin and at the point r2 = (a1 + a2 )/2 (these are not the Bravais lattice vectors
displayed in Fig. 2.27). The reciprocal of the rectangular lattice is rectangular as
well (Fig. 3.10), with G = 2π(hx̂/a1 + kŷ/a2 ). In this case, Eq. (3.52) yields
F (G) = F0 (G)(1 + exp[iG · r2 ]) = F0 (G)(1 + exp[iπ(h + k)]) ,
where F0 (G) is the structure factor of a single atom. Hence, F (G) = 2F0 (G) when
h + k is even, and F (G) = 0 when h + k is odd. The Bragg peaks appear only for
h + k even, e.g., (hk) = (±1 ± 1), (±2 0), (0 ± 2), (±1 ± 3), . . .. The collection of
these points forms a rectangular-centered lattice, with edges of lengths 4π/a1 and
4π/a2 (check!).
This result can be derived in a different way. The lattice vectors of the primitive
unit cell are a1 = a1 x̂ and a2 = (a1 x̂ + a2 ŷ)/2 (Fig. 2.27), and its area is a1 a2 /2.
The reciprocal-lattice vectors are b1 = 2π(x̂/a1 − ŷ/a2 ) and b2 = 4πŷ/a2 . Hence
G = h b1 + k  b2 = 2π[h x̂/a1 + (2k  − h )ŷ/a2 ], leading to h = h and k = 2k  − h .
Indeed, h + k is even, as found previously. This identity is quite general: the
measured structure factor is obviously independent of the choice of the unit cell.
Sometimes it is more convenient to compute it in the primitive unit cell, while there
are situations in which it is more expedient to find it in a bigger unit cell (which
is more symmetric). Other such examples appear in all cubic lattices (e.g., FCC,
BCC, diamond).
Scattering off hexagonal lattices. A somewhat more complex case is that of
the hexagonal lattice, which describes the planar crystals of graphene. As seen in
Fig. 2.8, this is a triangular lattice with two atoms in the unit cell, at the origin and
at r2 = (a1 +a2 )/3. Figure 3.11 shows that the reciprocal √ of the triangular lattice
√ is
also triangular, with the lattice vectors b1 = 2π(x̂− ŷ/ 3)/a and b2 = 4πŷ/(a 3),
and hence an arbitrary reciprocal-lattice vector is G = hb1 + kb2 . It follows that
F (G) = F0 (G)(1 + exp[iG · r2 ]) = F0 (G)(1 + exp[2iπ(h + k)/3])
√
(check!). The expression in the brackets has the values 2 or (1 ± i 3)/2, when h + k
is an integral multiple of 3 or else, respectively. Figure 3.19 displays the reciprocal
of the triangular lattice. The big points there are those for which h+k is an integral
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 153

Radiation scattering off crystals 153

multiple of 3. For instance, the two arrows in the figure are for (hk) = (12), (21).
These are the lattice vectors
√ of a triangular lattice containing
√ all big points, whose
lattice constant is b = b 3 = 4π/a [where b = 4π/(a 3) is the reciprocal-lattice
constant in Fig. 3.11]. The scattering intensity at these points is four times that of
the other sites in the reciprocal lattice (marked by small points in the figure). As
shown in Fig. 2.12, the triangular lattice is naturally divided into three sublattices.
The diffraction pattern from the hexagonal lattice breaks the symmetry among
those sublattices, by yielding higher scattering intensities on one of them. Notice
that the other two sublattices form a hexagonal lattice, such that the hexagonal
symmetry is conserved in the diffraction pattern. The points of the higher intensity
are at the centers of the hexagons. For a simple triangular lattice, the intensity is
the same at all points.

Fig. 3.19: Lattice points in the reciprocal of the hexagonal lattice. The scattering
intensity at the big points is larger. The intensities at all big (or all small) points are
identical. The arrows confine the enlarged unit cell of this lattice, which contains
one big point and two small ones.

Problem 3.19.
Assume that the two points in the unit cell of the hexagonal lattice represent two
different point-like atoms, whose structure factors are FA and FB . Determine the
structure factor of this crystal. Can the intensity at part of the reciprocal-lattice
sites vanish?

Problem 3.20.
Determine the structure factor of graphite, shown in Fig. 2.9 (see also problem
s.2.10). Are there any sites in the reciprocal lattice where the structure factor is
zero?

Scattering off cubic lattices with a base. An example for a three-

dimensional lattice with a base is the BCC lattice, the left panel in Fig.
2.16. Its base comprises two sites, at the origin and at the center of the cube,
a(x̂ + ŷ + ẑ)/2. The reciprocal-lattice vector is G0 = (2π/a)(hx̂ + kŷ + ẑ). The
structure factor (3.51) is thus proportional to {1 + exp[iπ(h + k + )]} and is nonzero
when h + k + is even, i.e., when all three indices are even, or when one of them is
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 154

154 Introduction to Solid-State Physics

even and the other two are odd. In both options the allowed values of h2 + k 2 + 2
are all even, and the structure of the crystal may be identified by measuring the
ratios sin2 θ/ sin2 θmin = (h2 + k 2 + 2 ), Eq. (3.32). Compared with the simple cubic
lattice, the diffraction pattern of the BCC one contains far less Bragg peaks.
The FCC lattice, Fig. 2.18, has four atoms in the unit cell, at the origin and in
the points (ŷ+ẑ)a/2, (x̂+ŷ)a/2, and (ẑ+x̂)a/2. A similar calculation shows that the
structure factor is proportional to {1+exp[iπ(h+k)]+exp[iπ(h+ )]+exp[iπ(k+ )]}.
Hence, constructive interference is realized only when the three integers h, k, and
are all even or are all odd. From the discussion around Eq. (3.31), the only values
of the sum h2 + k 2 + 2 that yield constructive interference off the FCC lattice, up
to 24, are 3, 4, 8, 11, 12, 16, 19, 20, and 24 (check!) In both cases, the list of angles
that yield constructive interference is much shorter than that of the SC lattice,
and one may deduce the type of the lattice from this list. Figure 3.20 summarizes
the values of h2 + k 2 + 2 that give Bragg peaks for the ubiquitous cubic lattices.
The disparity among the diffraction patterns enables the experimental distinction
between them.

Fig. 3.20: The values of h2 + k 2 + 2 that yield Bragg peaks for the SC, the BCC,
the FCC lattices, and the diamond lattice, mentioned in problem 3.25.

Problem 3.21.
Derive the structure factor of the FCC lattice directly, using its primitive lattice
vectors, Fig. 2.18.

Problem 3.22.
a. Determine the reciprocal-lattice vectors that belong to the planes of the highest
site densities which lead to constructive interference off the BCC lattice.
b. Find the planar lattice that describes these planes.
c. What is the densest line on each plane?
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 155

Radiation scattering off crystals 155

d. What are the answers for the FCC lattice?

Scattering off a crystal with several types of atoms. Consider the table

salt crystal, Fig. 2.17. As discussed in Sec. 2.5, table salt is described by the FCC
lattice, with a base comprising a chlorine ion at the origin and a sodium one at
the mid point of the cube’s edge, e.g., along the ŷ−axis. Instead of calculating the
structure factor for the primitive unit cell, it is more convenient to derive it for the
cubic unit cell, that contains four chlorine ions [at the origin, and at (x̂ + ŷ)a/2,
(ŷ + ẑ)a/2, and (ẑ + x̂)a/2] and four sodium ions (displaced away from each of these
four points by the vector aŷ/2). For the cubic lattice whose lattice constant is a,
the reciprocal-lattice vectors are given by G0 = (2π/a)(hx̂ + kŷ + ẑ). It follows
that Eq. (3.51) for table salt takes the form
F (hk ) = [1 + eiπ(h+k) + eiπ(h+) + eiπ(k+) ](FNa + FCl eiπk ) . (3.53)
The expression in the square brackets is the one of the FCC lattice with a single
element at each lattice site. The other factor in Eq. (3.53) is the structure factor of
the crystal, which contains contributions from the two ions forming the base. This
term yields (FNa + FCl ) for even k’s (and then all indices are even) and (FNa − FCl )
for the odd k’s (and then all indices are odd). These two types of points give
rise to disparate intensities. The chlorine and sodium ions are quite different and
therefore both types produce nonzero intensities. As opposed, potassium chloride,
KCl, yields a different result: each of the ions K+ and Cl− , possesses the same
number (18) of electrons, like the noble gas argon (see the periodic table in Fig.
2.22). Therefore, although the charges of their nuclei are not the same (and, as a
result, the radii are slightly different) their structure factors are quite close to one
another, and the radiation intensity at the odd points is rather small. A similar
conclusion pertains to CeSb: an inspection of the periodic table, together with
the assumption that three of the cerium electrons have been transferred to the
antimony, reveals that the electrons’ numbers on the two ions are 55 and 54; they
both are similar to the xenon atom, and their structure factors are also close to
one another. Notice that the similarity of the structure factors of potassium and
chlorine (as well as that of the Ce and Sb ions) pertains only for X-ray scattering.
For comparison, neutrons are scattered mainly from the nuclei; the scattering cross
sections of neutrons from Cl and K are significantly dissimilar. Thus, much can be
learnt from comparisons between the diffraction patterns of X-rays and neutrons,
taken from the same crystal.
Figure 3.21 displays experimental values of the intensity of the peaks obtained
by scattering X rays, with wavelength λ ≈ 1.545Å that come from the Kα1 line of
copper, off table salt (see problem 3.1). The data are deduced from an analysis of
the intensities of rings on a screen, as illustrated in Fig. 3.4. The digits on the
vertical axis, which denote the intensity (i.e., the heights of the Bragg peaks) are
determined up to a multiplicative factor, since the measured intensity is proportional
to, e.g., the duration of the irradiation, and also to other experimental details. As
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 156

156 Introduction to Solid-State Physics

the interpretation of the data follows from ratios of intensities, this multiplicative
factor is insignificant. Each peak is marked by the Miller indices that represent
the relevant reciprocal-lattice vector. The intensities indeed decay when the indices
become larger [see the discussion following Eq. (3.50)]. Nonetheless, the dependence
on the scattering angle is not always monotonic, indicating additional dependencies
of the structure factors on the reciprocal-lattice vectors; the intensities are also
affected by the multiplicities of each of the family of planes, and other factors. The
peaks have a finite width also due to the finite size of each grain in the powder (as
discussed in Sec. 3.3).

Fig. 3.21: Intensity of the scattering off a powder of table salt as a function of the
scattering angle. The scattering is of X-rays emitted from copper, with wave length
λ ≈ 1.545Å. [courtesy of Y. Rosenberg, Tel Aviv University.]

Problem 3.23.
a. Show that the scattering angles in Fig. 3.21 indeed correspond to an FCC lattice,
and deduce from them the lattice constant of the table-salt crystal.
b. Exploit the first two peaks in Fig. 3.21 to estimate the ratio FNa /FCl , assuming
that it is real and independent of h, k, and . What are the conditions for this
assumption to be true?
c. What can be done when this ratio is complex?
d. Assume that the electrons’ density in the i−th ion is approximately given by
ni (r) = zi exp[−2r/ai ]/(πa3i ), where zi is the number of electrons on this ion and
ai is its “radius”. Find the dependence of the structure factor on h, k, and ,
and obtain the ratio between the two first peaks. Given that aNa = 0.95Å and
aCl = 1.81Å, calculate the ratio zNa /zCl .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 157

Radiation scattering off crystals 157

Problem 3.24.
Show, by a calculation, that were the structure factors of potassium and chlorine
precisely identical in the KCl crystal (which is similar to the table-salt one) then its
diffraction pattern would have been identical to that of a simple cubic lattice with a
lattice constant a/2. Explain this result (without carrying out the calculation).

Problem 3.25.

Determine the structure factors of CsCl, of diamond, and of zinc-blende.

Problem 3.26.
a. Determine the structure factor of the HCP lattice.
b. Assuming that all scatterers are point-like, what are the values of the intensities
in the various Bragg peaks (ignoring the multiplicities)?
c. Which Miller
 indices give the smallest scattering angles for the ideal HCP lattice,
with a3 /a1 = 8/3? (Express the result in terms of the ratio λ/a1 .)
d. Determine the structure factor of Wurtzite.

Problem 3.27.
Determine the structure factors of the body-centered orthorhombic lattice, of the
face-centered one, and of the base-centered one.

Problem 3.28.
Which of the angles derived in problem 3.15 correspond to constructive interference
for the body-centered tetragonal lattice?

Scattering off large organic and biological molecules. The examples

studied so far are of crystals whose unit cells contain only a few ions, and there-
fore their structures can be deduced from a relatively small number of data points
of the intensities |F (G)|2 . When the unit cell contains many scatterers, the de-
ciphering of the experimental results becomes more complex and requires many
Bragg peaks. The problems at the end of the chapter present several examples of
more complicated crystals. The treatment of such crystals is carried out by various
methods. One possibility is to conjecture the structure of the unit cell and then to
calculate structure factors for each reciprocal-lattice vector. The electron density
in each cell is computed by numerically solving the Schrödinger equation, including
the electronic interactions. These numerically-computed structure factors are com-
pared with the experimental data; the initial conjecture is then improved and the
procedure repeats itself till a satisfactory agreement with the data is achieved.
Another method is based on an initial identification of the lattice symmetry (for
instance, by scattering multi-chromatic radiation which contains many wave lengths
off a single crystal, i.e., by using the von Laue technique); then the scattering angles
are deduced, for example by scattering monochromatic radiation off powders. The
results are then compared with the symmetries found in the first stage. In the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 158

158 Introduction to Solid-State Physics

next step, the scattering intensities at each of the Bragg peaks are analyzed. As
mentioned in Sec. 3.7, the intensities contain many multiplicative factors that have
to be included in the analysis. At the end of this procedure one obtains the values of

Γ(G) [Eq. (3.48)], for a finite numbers of vectors G. The inverse Fourier transform
of these values produces the correlation function of the crystal,

n(0)n(r) = Γ(r) = 
Γ(G)e iG·r
. (3.54)
G
This function provides valuable information on the relative distribution of the scat-
terers among pairs of sites in the crystal. Since this information is collected from
lattice vectors smaller than a certain one, Gmax , determined by the experimental
setup, and consequently is collected from wave lengths longer than 2π/|Gmax |, it
is reliable only for distances such that r > 2π/Gmax . The precision is enhanced by
increasing Gmax , for instance by using more intensive radiation.
Notice that the intensity at the Bragg peaks gives the absolute values of the
structure factors of the entire unit cell (each of them is a sum over atomic structure
factors with phases) and hence does not convey information about their phases.
The structure factors of atoms which have reflection symmetry are  real (see prob-
lem 3.18). Hence, for such crystals it is possible to obtain n 
(G) = Γ(G), and then
derive the distribution of the scatterers in the crystal, n(r) = G n (G) exp[iG · r].
This procedure is usually not applicable for organic chemistry or for molecular
biology, i.e., when the crystals contain
 big molecules lacking reflection symmetry.
In that case the function F (G) = n (−G) [Eq. (3.47)] is complex, and measure-
ments of the intensity do not provide information about its phase. This is called the
phase problem. These phases can be identified by various methods (for instance,
by replacing chemically part of the atoms in the unit cells by others, and exploiting
information obtained from the correlation functions). Once those phases are known,
the inverse Fourier transform of the structure factor yields the distribution of the
scatterers within the unit cell, n(r). The latter enables the identification of the
atoms or ions in the base, and consequently the structure of the molecules in each
unit cell.
This procedure is an important research tool in molecular biology: exper-
imental methods to explore the structure of large biological molecules are based
on scattering off crystals made of such molecules. The analysis of the diffraction
patterns yields the chemical and the geometrical structures of the molecules. The
structures of DNA and penicillin were discovered in this way. Another conspicuous
example concerns the ribosomes’ molecules that contain several tens of proteins.
For instance, the ribosome 50S contains two molecules of RNA, one comprising
≈ 2900 atoms and the other ≈ 120. Figure 3.22 displays a diffraction pattern
measured on a single crystal of the ribosome; the crystal was rotated while be-
ing irradiated by synchrotron radiation, of wave length λ = 1.51Å. The analy-
sis of this pattern reveals a triclinic lattice with relatively large lattice constants,
a = (182 ± 5)Å, b = (584 ± 10)Å, c = (186 ± 5)Å, β = 109◦ . The base comprises
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 159

Radiation scattering off crystals 159

two molecules, of molecular mass 1.6 × 106 (in units of the proton mass).

Fig. 3.22: Diffraction pattern of a ribosome molecule, taken from halobacterium

marismortui. [I. Makowski, F. Frolow, M. A. Saper, M. Shoham, H. G. Wittman,
and A. Yonath, Single Crystals of Large Ribosomal Particles from Halobacterium
marismortui Diffract to 6 Å, J. Mol. Biol. 193, 819 (1987).]

3.10 The temperature dependence; the Debye-Waller factor

Atomic vibrations in the crystal. The discussions in the previous chapters are
based on the tacit assumption that the centers of mass of the atoms (or the ions) rest
precisely on the lattice sites (or at the sites of the base). In fact, both the electrons
and the nuclei are constantly in motion around the centers of the mass. In order to
compare with the experimental data, the previous expressions have to be averaged
over these motions; the average over the electrons’ motion around the nuclei is
carried out in the previous section, e.g., in part (c) of problem 3.18. The average
over the nuclei’s motion is discussed in this section. This average is accomplished in
two steps: first it is done over the quantum states of each degree of freedom. Recall
that quantum mechanics yields the probability distribution for a degree of freedom
to attain a certain value. Each of these values appears with its proper weight – the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 160

160 Introduction to Solid-State Physics

absolute value squared of the quantum wave functions. Second, the various quantum
states are averaged upon with their Boltzmann weights. The Boltzmann weight of
a state of energy E (measured relative to the ground-state energy), at temperature
T , is proportional to exp[−E/kB T ], where kB is the Boltzmann constant. When
the energies of all states exceed by far the thermal energy kB T , then these states
can be ignored and it suffices to quantum-mechanically average over the degrees of
freedom in the ground state. Usually these energies are indeed high for the excited
states of the electrons in the atom. This is the reason for the assumption that
the “density” of the electrons is equal to the quantum probability to find them
in the ground state at each point in space (i.e., equal to the absolute value squared
of the wave function of their ground state).
The centers of mass of the atoms (i.e., their nuclei) reside at minima of the
potential energy, see Chapter 4. But quantum mechanics implies that microscopic
particles are never exactly at the sites of their minimal potential energy, and there-
fore their locations have to be averaged upon. Even in the ground state of the
harmonic oscillator the wave function differs from zero for many a value of the
particle location, indicating that there is a finite probability to find the particle
away from the minimum of the potential energy. Consider the vibrations of the
centers of mass of the atoms in the particular case of a single atom in each unit
cell. As opposed to electrons, the energies of these vibrations are not high com-
pared to kB T , and thus at finite temperatures the excited states have also to be
included in the average, each with its (temperature dependent) Boltzmann weight.
The instantaneous location of the center of mass of the atom can be written in the
form R = R + uR , where R is the location on the periodic lattice and uR is the
instantaneus distance away from that site. As a result, the scatterers’ density is
different in each cell, n(r) = R n0 (r − R − uR ), where n0 (r) is the density of the
scatterers around a single atom (placed at the origin). It follows that Eq. (3.50)
(G) has to be revised,
for n

1 
(G) =
n d3 rn0 (r − R − uR )e−iG·(r−R−uR ) e−iG·uR
NV V (R)
R
F (−G)  −iG·uR
= e . (3.55)
NV
R

In the first step the integral is separated into a sum of integrals on each of the unit
cells, displacing the origin in each term to the instantaneous location of the atom
center in the relevant unit cell (and exploiting the identity exp[−iG · R] = 1). As
the duration of the experiment is much longer than the period of the vibrations,
the measurement is sensitive only to the average value of these vibrations. It is
therefore plausible to replace each summand by its average, exp[−iG · uR ], where
the angular brackets denote the double average, the quantum average and thermal
one. Usually the probabilities for the displacements uR and −uR are identical,
and consequently exp[−iG · uR ] = cos(G · uR ) is real. As |G| increases, the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 161

Radiation scattering off crystals 161

expression within the average oscillates faster, and reduces the average. Since it
is smaller than 1, it is customary to denote exp[−iG · uR ] = exp[−W ]. This
quantity, which suppresses the structure factor, is called the Debye-Waller factor.
Assuming that the averages over all atoms are the same, one retrieves Eq. (3.47),
but with

F (G) ⇒ F (G)exp[−iG · uR ] ≡ F (G) exp[−W ] . (3.56)

Calculating the Debye-Waller factor. The Debye-Waller factor can be

estimated by assuming that the vibrations are small; the exponential can then
be expanded in a Taylor series: exp[−iG · uR ] ≈ 1 − iG · uR − (G · uR )2 /2 + . . . .
Assuming in addition that the vibrations of each component of uR are independent,
and that their average vanishes, the first term in the expansion disappears. The
second yields (G · uR )2  = G2 (uR )2 cos2 φ = G2 (uR )2 /3, where φ is the angle
between G and uR ; the average of cos2 φ is 1/3 when all directions are identical
(see problem 3.29 with G directed along one of the axes). Using the approximation
exp[−W ] ≈ 1 − W gives

W ≈ G2 (uR )2 /6 . (3.57)

The quantity W determines the Debye-Waller factor. Within the present ap-
proximation, this factor decays exponentially with the square of the length of the
reciprocal-lattice length, G2 . This decay hinders the collection of data from many
Bragg peaks: the peaks of the large G’s practically disappear, unless high-radiation
intensity is used.

Problem 3.29.
Prove that the average over all directions of an arbitrary vector in a d-dimensional
space, pointing along an arbitrary direction that makes an angle φ with one of the
axes, is cos2 φ = 1/d.

It remains to find the average of the amplitude squared, i.e., (uR )2 . This is
carried out in detail in Chapter 5, which focuses on the theory of lattice vibrations.
A qualitative answer can be obtained by assuming that the vibration of the atom
at R is independent of those of the other atoms, and that there exists solely a
single vibrational frequency, ω (as in the model of Einstein, discussed in Chapter
5). Assuming that (i) the vibration’s amplitude is small, and (ii) the symmetry
around the minimum (of the potential energy) is spherical, the potential that induces
the vibrations, in the harmonic approximation, is V = M ω 2 |uR |2 /2, where M is
the mass of the atom and ω is the angular frequency of the harmonic oscillation
around the minimum (K = M ω 2 is the second derivative of the potential at the
minimum, or the “spring constant”). Decomposing the vector uR into its cartesian
components, x, y, and z, then V = M ω 2 (x2 + y 2 + z 2 )/2, i.e., the sum of three
potentials of one-dimensional, independent harmonic oscillators. It remains to find
the standard deviation of such an oscillator, x2  = y 2  = z 2 .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 162

162 Introduction to Solid-State Physics

As mentioned, the averages are carried out in each quantum state of the os-
cillator, with weights given by the absolute value squared of the wave function of
the quantum state, and over the ensemble of quantum states with the appropriate
Boltzmann weights. There are two limits where the result is simple. At zero tem-
perature the oscillator is in its ground state, and then its average potential energy
is the quantum average of the standard deviation there, 3M ω 2 x2 /2 = 3ω/4. In-
deed, as the wave function in the ground state is ψ0 = C0 exp[−(x/x0 )2 /2], where
x0 = /(M ω), it follows that x2  = dxx2 |ψ0 |2 / dx|ψ0 |2 = x20 /2, and the stan-
dard deviation of uR represents the zero-point motion of the oscillator, which is
larger as the mass and/or the frequency are smaller. [As x = 0 in the oscillator
states, the square of the standard deviation is (x − x)2  = x2 .] In this limit,
2W = G2 /(2M ω).
At high temperatures, when ω  kB T , the thermodynamic equipartition
theorem can be used; it implies that the average energy of each degree of free-
dom is kB T /2. Thus, in this limit M ω 2 x2 /2 = kB T /2, which yields 2W =
G2 kB T /(M ω 2 ). The Debye-Waller factor is smaller than 1 even at zero temper-
ature, and becomes even smaller as the temperature increases.
Intermediate temperatures require a full calculation of both the quantum and
the thermal averages. The one-dimensional harmonic oscillator is described by
the potential energy V (x) = M ω 2 x2 /2. It is known that the quantum energy
levels of this oscillator are given by En = ω(n + 1/2), where n is a non negative
integer. The kinetic energy and the potential energy of a classical oscillator are
equal to one another. This equality persists also for the average quantum values
of these energies in the n−th energy level of the oscillator, i.e., EK n = V n =
En /2 = ω(n + 1/2)/2, hence the quantum average in the n−th state is x2  =
(n + 1/2)/(M ω). This quantum average has to be further thermally-averaged.
The thermal probability to find a system at temperature T with energy En is the
Boltzmann factor, pn = exp[−βEn ]/Z, where β ≡ 1/(kB T ), and where Z =
n exp[−βEn ] is the normalization, ensuring that the sum of all probabilities is 1
(Z is called the partition function). At zero temperature the probability to find
the system at its ground states is 1, while at infinite temperature there is an equal
probability to find the system at any of its quantum states.
By symmetry, xn = 0; therefore the average standard deviation of x is obtained
by averaging (x − x)2 n = x2 n with the weights pn = exp[−βEn ]/Z,
  1  1 1  ∂
x2  = pn x2 n = (n + )e−βω(n+ 2 ) = − ln Z . (3.58)
n
Mω Z n 2 M ω ∂(βω)

The last step is a mathematical trick that facilitates the calculation. One finds Z
as a sum of an infinite geometrical series,

Z= exp[−βω(n + 1/2)] = 1/[2sinh(βω/2)] .
n
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 163

Radiation scattering off crystals 163

The differentiation yields x2  = [/(2M ω)]coth(ω/[2kB T ]), hence

G2  ω 
2W = coth . (3.59)
2M ω 2kB T
One observes that W indeed increases monotonously between its value at zero tem-
perature and its linear temperature dependence at high temperatures [part (a) of
problem 3.30]; thus, the intensity of the Bragg peaks decays as the temperature is
raised.
The Lindemann criterion for melting. When the average vibration ampli-
tude of the atom in the lattice is large (larger than the lattice constant) the solid
becomes unstable and eventually crosses over to the liquid phase. In other words,
the material melts. The melting temperature can be related to the size of the aver-
age vibration’s amplitude for which melting takes place. Examining various melting
temperatures and lattice vibrations led Lindemann in 1910 to conjecture that cubic
lattices melt once the amplitude of the vibration exceeds a certain percentage of the
lattice constant a, i.e., when (uR )2 1/2 > cL a. It is customary to use cL = 0.1 for
simple lattices; the experimental data is scattered around this value but at times
this estimate yields correctly just the order of magnitude. In more complicated
lattices the lattice constant is replaced by the distance between nearest neighbors.
When there are several lattice constants a is replaced by a typical length scale, for
example, the cubic root of the unit cell volume. As mentioned, at very high tem-
peratures (uR )2  = 3kB T /(M ω 2 ), and melting is expected to occur at the melting
temperature TM , where kB TM ≈ c2L M ω 2 a2 /3. This criterion is called after Lin-
demann. Note that the melting temperature increases with the spring constant
(M ω 2 ), that characterizes the potential V , i.e., when the material becomes more
rigid (as the spring constant is enhanced, it is harder to stretch it, i.e., it is harder
to deform the crystal).
Crystal with a base. For a crystal with a base, Eq. (3.56) is generalized to
be

F (G) = Fi (G)e−iG·ri e−Wi (G) . (3.60)
i

Thus, the relative weights of the contributions to the structure factor from the
various atoms in the unit cell may vary among the Bragg peaks, and that variation
is temperature-dependent.

Problem 3.30.
a. Prove that Eq. (3.59) reproduces the limits of W , for zero temperature and for
very high temperatures.
b. Show that W = B sin2 θ/λ2 , where 2θ is the scattering angle and λ is the wave
length, and find the coefficient B. Which properties of the crystal determine it?
c. Use the Lindemann criterion to show that at high temperatures, kB T  ω, the
intensity of the Bragg peaks is a function of T /TM alone, and that the Debye-Waller
factor includes just the ratio of the lattice constants and not their specific values.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 164

164 Introduction to Solid-State Physics

Problem 3.31.
a. How is the Debye-Waller factor modified when all atoms of a certain type are
replaced by another isotope of them?
b. How should the Debye-Waller factor be modified when the surrounding of each
atom is not spherically-symmetric?

3.11 Scattering off quasicrystals

As mentioned, there exist diffraction patterns that show rotational symmetry of

order 10, which is forbidden for periodic crystals. Such a diffraction pattern is dis-
played in Fig. 3.23; it was obtained by Shechtman in 1982 by scattering X-rays
off Mg-Al alloys. The figure shows diffraction patterns taken at various directions
of the crystal relative to the impinging beam. The different angles represent scat-
tering from different planes; the figures reflect the symmetries that pertain to each
family of planes. In all cases, the Bragg peaks are narrow, but along certain di-
rections, for instance, at 63.43◦ (upper figure) the pattern possesses a rotation axis
of order 10, i.e, rotations by 36◦ . This symmetry is forbidden for a peri-
odic Bravais lattice. Already in 1982 there were theoretical calculations of the
diffraction patterns of a three-dimensional arrangement that consists of two types of
three-dimensional structures which generalize the Penrose tiling; the result resem-
bles very much the patterns in Fig. 3.23. This comparison confirms the conjecture
that three-dimensional quasicrystals can be described by combinations of more than
three lattice vectors, with irrational relations in-between them.
Scattering off one-dimensional quasicrystals. The Fibonacci lattice can
serve to exemplify the diffraction pattern off a quasi-periodic one-dimensional crys-
tal. The scattering amplitude of point-like scatterers arranged on a one-dimensional
lattice is given in Eqs. (3.7) and (3.8), A ∝ Z = n exp[iqxn ], where q is the com-
ponent of q = k − k along the lattice, and {xn } denotes the locations of the lattice
sites. In a Fibonacci lattice the locations are given by xn = S[n+||n/τ ||/τ ]; τ is the
golden ratio and the integer m = ||n/τ || is in the range n/τ − 1/2 < m < n/τ + 1/2
[see Eq. (2.8); S is the length of the shorter segment in the Fibonacci sequence]. It
follows that
 
A∝Z= eiqxn = eiqS(n+m/τ ) Θ(m − n/τ + 1/2)Θ(n/τ + 1/2 − m) ,
n n,m
(3.61)

where Θ(x) is the step function: it is equal to 1 for x ≥ 0 and to 0 otherwise.

The result of the summations is given in Eq. (3.64). It is derived by converting
the sums into integrals
 
Z(qx , qy ) = dx dyei(qx x+qy y) f1 (x, y)f2 (x, y) , (3.62)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 165

Radiation scattering off crystals 165

Fig. 3.23: Diffraction patterns of a quasicrystal. [D. Shechtman, I. Blech, D. Gratis,

and J. W. Cahn, Metallic Phase with Long-Range Orientational Order and No
Translational Symmetry, Phys. Rev. Lett. 53, 1951 (1984).]

with

f1 (x, y) = δ(x − n)δ(y − m) , f2 (x, y) = Θ(y − x/τ + 1/2)Θ(x/τ + 1/2 − y) ,
n,m

and qx = qS, qy = qS/τ . The integral in Eq. (3.62) is the Fourier transform of
a product of two functions; the theory of Fourier transforms [see in particular Eq.
(B.11)] gives for this transform a convolution, i.e.,
 
Z(qx , qy ) = dqx dqy f1 (qx , qy )f2 (qx − qx , qy − qy ) . (3.63)

The Fourier transform of f1 is the two-dimensional version of Eq. (3.14),

 
f1 (qx , qy ) = ei(qx n+qy m) = δ(qx − 2πn)δ(qy − 2πm) .
n,m n,m

The Fourier transform of f2 is a simple integral,

 ∞  y+  ∞
f2 (qx , qy ) = dx dyei(qx x+qy y) = dxei(qx +qy /τ )x 2 sin(qy /2)/qy ,
−∞ y− −∞

y± = x/τ ± 1/2 .
Inserting the two transforms into Eq. (3.63), and using the identity
 ∞
dx exp[ikx] = 2πδ(x) ,
−∞
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 166

166 Introduction to Solid-State Physics

results in
  qy − 2πm  sin(qy /2 − πm)
Z(qx , qy ) = 2(2π)3 δ qx − 2πn +
n,m
τ qy − 2πm
  1  sin(qS/(2τ ) − πm)
= 2(2π)3 δ qS[1 + 2 − 2π[n + m/τ ] . (3.64)
n,m
τ qS/τ − 2πm
The delta function on the right hand-side indicates a Bragg peak when
q = 2π(n + m/τ )/[S(1 + 1/τ 2 )] . (3.65)
As both n and m can attain any integral value, Eq. (3.65) represents a very dense
collection of points on the q−axis. Nonetheless, the intensity of each Bragg peak is
proportional to [sin(qS/(2τ ) − mπ)/(qS/τ − 2πm)]2 , and as such is significant only
when q ≈ 2πmτ /S. Inserting this into Eq. (3.65) implies that n/m ≈ τ . This can
be the case only when both integers are quite large. An inspection of Eq. (3.65)
reveals that the q values that give rise to significant Bragg peaks are themselves a
Fibonacci sequence! (see problem 3.32).
The main conclusion from this analysis is that the diffraction pattern of a
quasi-periodic lattice contains narrow peaks, very much like the usual
Bragg peaks, but these peaks do not correspond to a simple reciprocal lattice.
They have the symmetries of quasicrystals in momentum space, that are decisively
not those of a periodic lattice. Similar considerations for two- and three-dimensional
quasicrystals reach the same conclusion. The diffraction pattern of a quasicrystal
reflects the symmetry of the original crystal, much like the diffraction patterns of
periodic crystals, which display their symmetries.

Problem 3.32.
a. Show that the condition n/m ≈ τ is equivalent to the result of problem 2.27 part
(b), that gives the ratio of the numbers of the long and short segments in a Fibonacci
sequence. What conclusions can be deduced about the sequence of the high peaks in
the diffraction pattern?
b. Compose a code that prepares a table of pairs of values, one for q = π(n +
m/τ )/(τ + 1/τ ) and the other for A = [sin(q − πm)/(q − πm)]2 , for a range of
integral values of n and m (e.g., n, m = 1, 2, . . . , 100) and plot these values in the
q − A plane. Insert into the code the condition that the allowed A values are only
those within the highest percentage of all values obtained for these points, and mark
down the corresponding values of q and the intervals in-between them. What are the
relations between these intervals? What is the conclusion drawn about the sequence
of the q values?

3.12 Scattering of electrons off surfaces

Diffraction of low-energy electrons. As mentioned, for instance in Sec. 2.9,

the surface of a crystal may be considered as one of the planes forming that crystal.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 167

Radiation scattering off crystals 167

Therefore the surface can be denoted by the Miller indices of the family of planes
that contains the surface plane. At times one wishes to explore the surface sep-
arately from the three-dimensional bulk below. In the example discussed in Sec.
2.9 the focus is on the thin layer adsorbed on the surface. Usually X-rays are not
useful for this purpose since they penetrate deep into the material (and thus convey
information on its three-dimensional structure). A possible tool to investigate the
structure of surfaces, already mentioned in Sec. 3.1, is the scanning tunneling mi-
croscope. Another option is to exploit diffraction of electrons at low energies;
this method is termed low-energy electron diffraction, or LEED. The typ-
ical energies involved are 10-200 eV, i.e., wave lengths of the order of λ ∼ 0.8 − 3.8Å
(see Sec. 3.1). Due to the large cross section for inelastic scattering of such electrons
off the crystalline electrons, they do not penetrate far beneath the surface, and are
mainly scattered from the outer atomic layers.
Constructive interference of waves scattered from planes. Several ex-
amples for scattering off two-dimensional crystals are presented in Sec. 3.9. How-
ever, the expressions derived there are based on the assertion that the wave vec-
tors, the impinging and the scattered ones, are also in the plane. In order to
include the possibility that the wave vectors are in space, one may use Eq. (3.9),
under the condition that the lattice vectors R are restricted to atoms belonging to
the outer layer of the crystal. When the surface is represented by the reciprocal-
lattice vector G0 , then the surface itself obeys the equation G0 · R = 0 (see Sec.
3.6). Decomposing the vector q = k − k, that gives the momentum difference be-
tween the incoming and the outgoing waves, into the component q⊥ , perpendicular
to the surface (i.e., parallel to G0 ) and the component q , that is parallel to the
surface, q = q⊥ + q , and using q⊥ · R = 0, Eq. (3.9) implies
q · R = q · R = 2πm . (3.66)
It follows that the vector q is given by a linear combination with integer coef-
ficients of the reciprocal-lattice vectors of the plane, q = Gsurface = hb1 + kb2
(which contains the information on the structure of the surface plane), but there is
no restriction on the component normal to the surface, q⊥ , that thus may attain
continuously any value. Radiation with wave vector k gives rise to constructive in-
terference on lines in momentum space, normal to the surface plane at discrete
points that belong to the reciprocal lattice of the plane,
k − k = q = Gsurface + q⊥ . (3.67)
Assuming that the impinging beam is normal to the planar sample, constructive
interference is expected whenever the scattered beam hits one of the lines perpen-
dicular to the sample at one of its reciprocal-lattice sites. The points on the right
hand-side in Fig. 3.24 represent these sites, and the circle there (whose radius
equals the length of the impinging wave vector and its center is at the origin of that
vector) is a manifestation of energy conservation, |k | = |k|. This circle is part of
the three-dimensional Ewald sphere. As opposed to Fig. 3.12, where constructive
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 168

168 Introduction to Solid-State Physics

interference is realized only for the point C on the sphere, in the present case such
interference emerges whenever one of the parallel lines normal to the plane of the
sample at its reciprocal- lattice sites cuts the sphere. Thus, there are many more
points on that sphere, as compared to the three-dimensional case depicted in Fig.
3.12, for which the interference is constructive. When the point B is connected
with each of these points (marked by ∗), and the lines are continued to a spherical
screen around the point B, the scattering points formed on that screen reflect the
structure of the reciprocal lattice.

Fig. 3.24: The Ewald sphere for scattering off a two-dimensional lattice. The points
on the right represent the plane of the measured sample.

Problem 3.33.
Referring to the geometry of Fig. 3.24, express the direction of the vector k in
terms of the indices h and k that characterize the planar reciprocal-lattice vector
which scatters the radiation, the angle θ, and the wave length of the impinging
wave, and use these data to determine the locations of the radiation points on a
planar screen parallel to the measured sample.
The diffraction picture reflects the sites of the planar reciprocal-lattice of the
measured sample. When it is a planar layer adsorbed on a planar cut of another
crystal, the intensities of the various Bragg peaks indicate the composite structure
of both planes. An example is the structure of krypton placed on the (001) plane
of graphite, displayed in the upper part of Fig. 2.38. As explained in problem 2.32,
the krypton lattice is a triangular one, whose lattice constant is a = 3a, where a
is the distance between nearest-neighbor carbon atoms in the upmost layer of the
graphite. The reciprocal lattice of this triangular lattice is illustrated in Fig. 3.11.
Ignoring the height differences between the krypton atoms and the carbon ones in
the adsorbing layer, the unit cell contains a krypton atom at the origin and six
carbon ones at the sites ±a1 /3, ±a2 /3, ±(a2 − a1 )/3. The planar structure factor
is then F (hk) = FKr + 2FC [cos(2πh/3) + cos(2πk/3) + cos(2π(k − h)/3]. It attains
three values, F (hk) = FKr + 6FC , FKr , FKr − 3FC . The magnitudes of the points in
Fig. 3.25 display these three values, according to the appropriate Miller indices. It
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 169

Radiation scattering off crystals 169

is illuminating to compare this figure with Fig. 3.19, that portrays the intensities of
the Bragg peaks in the diffraction pattern of graphene. In the latter, there appears
a hexagonal lattice of identical intensities (the small points in Fig. 3.19); at the
center of each hexagon there is a peak with a different intensity. The hexagonal
lattice is still there in Fig. 3.25, but inside the hexagons there are now points of
two different intensities, arranged alternately.
When the actual height, u, of the krypton atom above the substrate is accounted
for, the structure factor FKr is multiplied by exp[iq⊥ u] = exp[ik0 [1 − cos(2θ)]u] (see
problem 3.33 and Fig. 3.24). It follows that the structure factor at each site
is a complex number, and the intensity attains an additional dependency on the
scattering angle.

Fig. 3.25: The intensities of the Bragg peaks on the reciprocal lattice of the trian-
gular lattice that represent the periodic arrangement of krypton on graphite.

Problem 3.34.
Describe the diffraction pattern of electrons scattered off a one-dimensional lattice.

3.13 Neutron scattering off magnetic crystals

Advantages of neutron scattering. Scattering by neutrons differs from that by

X-rays since the latter are scattered by the electrons while the neutrons are scattered
by the nuclei or by the atomic magnetic moments. Since the nuclei are much smaller
than the atoms, they can be considered as point-like scatterers; for this reason the
atomic structure factor for neutron scattering hardly decays as the reciprocal-lattice
vector G increases [see the discussion of scattering off a single atom following Eq.
(3.50)]. As opposed to scattering of X-rays, which gives rise to structure factors that
increase with the atomic number of the scatterer (this parameter determines the
number of the scattering electrons), the cross section for neutron scattering is not
sensitive to the atomic number. One of the main advantages of neutrons is that the
scattering intensities, even from light atoms like hydrogen (which is hardly accessed
by X-rays), are large. Another important factor is the penetration depth: the mean
free-path (the distance along which the intensity is reduced by the factor 1/e, where
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 170

170 Introduction to Solid-State Physics

e ≈ 2.73 is the Euler number) of X-rays is of the order of a few millimeters, while
that of neutrons can exceed one centimeter.
Identification of magnetic structures. As mentioned in Sec. 2.10, magnetic
materials undergo phase transitions: above the magnetic transition temperature
the magnetic moment of each atom (or ion) is free to align along an arbitrary
direction, such that the average moment of the entire sample vanishes. Below
that temperature, however, these moments arrange themselves in various periodic
patterns. Neutrons possess a magnetic moment; for this reason they are sensitive
to the magnetic structure of the measured crystal; the scattering amplitude
depends on the angle in-between the magnetic moment of the neutron and that of
the atom (or ion). The magnetic moments of atoms are on the nuclei and on the
electrons, and only the neutrons are sensitive to both types of moments; neutron
scattering is thus the paramount tool for studying magnetic structures. Since above
the transition temperature the atoms (on the average) do not possess a magnetic
moment, the neutrons are scattered solely from the periodic structure of the nuclei
(this structure is termed “chemical” or “nuclear”), very much like X-rays. On
the other hand, below the transition temperature the magnetic moments of the
nuclei and/or the electrons are arranged periodically, and then there is a magnetic
component to the scattering intensity, reflecting the magnetic structure.
Sources of neutrons. Neutrons are largely obtained from nuclear reactors.
Inside these reactors there is a multitude of thermal neutrons, with kinetic energy
of the order of kB T . At room temperature, i.e., T ≈ 300 − 400 degrees K [that is,
kB T = (0.025−0.033)eV] these neutrons have wave lengths of the order of 1-2Å [Eq.
(3.3)]. The disadvantage of this source is the low intensity, that necessitates the use
of relatively large samples (whose typical size is a few centimeters). Larger-intensity
neutrons are obtained from pulsed spallation sources: accelerated protons collide
with heavy metals and release in the process quite a number of neutrons. The
neutron flux from such sources is about a hundred times more than that coming
from reactors.
The discovery of Shull: antiferromagnetism in manganese oxide. A
neutron scattering experiment carried out in 1949 (with wave length λ = 1.057Å)
on a powder of manganese oxide, MnO, had been the first to prove the existence of
the antiferromagnetic structure. The Bragg peaks had been monitored at two
temperatures [displayed in Fig. 3.26(a)]. The mere fact that the low-temperature
data contains peaks that are not there at the higher temperature indicates that
this material has two different periodic structures. The experimenters, Shull and
collaborators, concluded that at the higher temperature the magnetic moments are
not yet periodically-arranged, and therefore the Bragg peaks represent only the
scattering off the “chemical” periodic structure of the atoms in the lattice. These
peaks imply an FCC lattice, similar to that of table salt, in which the manganese
and the oxygen ions replace the sodium and the chlorine ones, respectively (Fig.
2.17), with high intensities for odd indices, i.e., (hk ) = (111), (311), and weak
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 171

Radiation scattering off crystals 171

ones for even indices, (hk ) = (200), (220), and a lattice constant a = 4.426Å [Eq.
(3.53)]. The figure differs from Fig. 3.21 (portraying the data of X-ray scattering off
table salt, which is also an FCC lattice with a base comprising two ions) because, as
opposed to the electronic structure factors of the X-ray data, the nuclear structure
factors FO and FMn are opposite in sign. This explains the disparity between the
intensities of the odd-indices peaks and the even-indices ones: for X-ray the intensity
is higher at the even indices, while in the case of neutron scattering it is higher at
the odd ones. The high-temperature data of the neutron scattering is also in accord
with X-ray scattering off the same powder (allowing for the different intensities).

(a) (b)

Fig. 3.26: (a) Bragg peaks of neutron scattering off a powder of manganese oxide at
two different temperatures. (b) The magnetic structure of the manganese ions at low
temperatures, as deduced by Shull and coworkers. [C. G. Shull, W. A. Strauser,
and E. O. Wollan, Neutron Diffraction by Paramagnetic and Antiferromagnetic
Substances, Phys. Rev. 83, 333 (1951).]

The “chemical” peaks remain in their places in the upper plot of Fig. 3.26(a),
taken at 80◦ K. However, this plot contains more peaks, which can be related to an
FCC lattice with a doubly-larger lattice constant, a = 8.85Å. These peaks belong
to indices which are all even or all odd (in the new basis of the larger FCC lattice),
(hk ) = (111), (311), (331), (511) + (333). These were interpreted as due to the
appearance, at a low temperature, of a periodic pattern of the magnetic moments
which is not there at the higher temperature. As the magnetic moment of the
neutron interacts differently with “upward” directed moments and “downward”-
directed ones, the two moments behave as two different scattering agents on the
lattice, and increase the size of the unit cell (see Sec. 2.10): the scattering angles
identify the double unit cell. It remains to find out the magnetic structure within
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 172

172 Introduction to Solid-State Physics

each unit cell. Testing several options, Shull and coworkers have concluded that
the intensities of the peaks agree with the structure drawn in Fig. 3.26(b). In
this structure, the magnetic moments of the manganese ions are arranged antifer-
romagnetically: half of them points “upwards” and the other half “downwards”.
The figure contains dashed triangles, corresponding to planes of the (111) type in
the original cubic lattice (compare with Fig. 2.24). Within each plane the mo-
ments are parallel to each other, but are antiparallel to those in the neighboring
plane. Along any axis of the cube (as well as upon moving from one plane to the
other) the neighboring moments are anti parallel to each other, as in part C of Fig.
2.42. This picture is re-visited at the end of the section, once simpler structures are
explained. It is worthwhile to note that the total moment of an antiferromagnet
vanishes, since the opposite moments cancel each other; it is therefore not easy
to identify this structure by magnetic measurements, e.g., the measurement of the
total magnetic moment. Neutron scattering is still the best tool to explore complex
magnetic structures.
The magnetic structure factor. In the simplest case, the interaction energy
of the magnetic moment of the neutron, μn (smaller than that of the electron by a
factor of 960, due to the mass ratio) with the magnetic moment of the ion located at
point r in the crystal (that is due to the magnetic moments of electrons in the outer
shells of this ion) is proportional (approximately) to the scalar product μn · μ(r).
The scattering amplitude is thus different when these moments are parallel or are
antiparallel. A detailed calculation yields that the density, n(r), in the previous
expressions, [for instance in Eq. (3.46)] is replaced by the magnetization (defined
as the density of the magnetic moment) at the same point, μ(r). The magnetic
scattering amplitude is then determined by the magnetic structure factor,

FM (G) = d3 rμ(r)eiG·r , (3.68)

which replaces Eq. (3.46). Note that, as opposed to the scalar nuclear structure
factor, the magnetic structure factor is a vector. When the unit cell comprises
several discrete moments μi , whose locations are ri , the magnetic structure factor
is given by a generalization of Eq. (3.51),

nB
FM (G) = μi eiG·ri . (3.69)
i=1

Simple examples. Consider first the one-dimensional arrangements depicted

in Fig. 2.42. All the moments in the ferromagnetic structure B are identical to each
other; the magnetic unit cell thus coincides with the “chemical” or the “nuclear”
one. Neutron scattering then produces the same diffraction pattern as in the absence
of a magnetic order (e.g., when the moments are randomly arranged as in panel A:
there, each moment attains various random directions during the measurement time
and is averaged to zero). On the other hand, the antiferromagnetic order in panel
C is characterized by a double-sized unit cell, with a lattice constant a = 2a, and
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 173

Radiation scattering off crystals 173

with a base comprising two magnetic ions of opposite moments (see Fig. 2.1, second
line). The reciprocal lattice of this lattice is twice as small as its predecessor, and
its lattice constant is b = 2π/a = π/a. When the “positive ” moment, μ0 , resides
at the origin and the “negative” one, −μ0 , is located at r1 = ax̂, the magnetic
structure factor is
FM (G) = μ0 (1 − e−iGa ) = μ0 (1 − e−i(πh/a)a ) = μ0 [1 − (−1)h ] . (3.70)
(It is a scalar since the moments are parallel or antiparallel to one another, and have
a single nonzero component.) The diffraction pattern of the magnetic structure
contains only odd values of h, i.e., G = (π/a)[1, 3, 5, . . .]. On the other hand, the
scattering off the nuclei gives rise to the same structure factor identical for all the
ions, whose contribution is Fn (G) = n0 (1 + exp[−iGa]) = n0 [1 + (−1)h ]. This
interference is constructive for even values of h, G = (π/a)[2, 4, 6, . . .]; these are
exactly the reciprocal-lattice vectors of the original nuclear lattice whose lattice
constant is b = 2π/a and therefore G = (2π/a)[1, 2, 3, . . .]. The magnetic order
produces additional Bragg peaks at the odd multiples of π/a. The appearance of
these additional peaks in the diffraction picture implies that the unit cell is doubled:
the smallest reciprocal-lattice vector is twice as small. An analysis of the scattering
intensities indicates that the neighboring moments are of identical magnitudes but
opposite in signs, and consequently this is the magnetic order of panel C in Fig.
2.42. As opposed, the scattering off the structure D in Fig. 2.42, with two different
moments μ1 and μ2 , yields FM (G) = μ1 − μ2 exp[−iGa] = μ1 − μ2 (−1)h , with the
intensities (μ1 − μ2 )2 or (μ1 + μ2 )2 , for an even or odd h, respectively. In addition
to the appearance of new peaks, the scattering intensities at the nuclear peaks is
modified as well (due to the magnetic intensity added to them).
Likewise, the lattice constant of the pattern marked by E in Fig. 2.42 is a = 3a,
and its base comprises three magnetic moments, for instance, positive, negative,
negative (cf. the lower panel in Fig. 2.1). Thus,
FM (G) = μ0 (1 − e2πih/3 − e4πih/3 ) = μ0 [1 − 2 cos(2πh/3)] .
The possible values are FM (h) = −μ0 when h is an integer product of 3, or FM (h) =
2μ0 for any other value. As opposed, the nuclear structure factor is Fn = n0 [1 +
2 cos(2πh/3)], vanishing unless h is an integer multiple of 3. This result agrees with
the reciprocal-lattice vectors of the original lattice. The magnetic order gives rise
to new Bragg peaks, which indicate that the unit cell is tripled. The intensities
enable one to identify the magnetic order within this new unit cell.
Figure 3.27 displays three types of simple antiferromagnetic structures on the
simple cubic lattice. In the arrangement of the G−type, all the neighbors of a given
moment are opposite to it. In the A−type one, the moments within each horizontal
plane are parallel to each other, but are opposite to those in the nearest-neighbor
planes. In the C−type order, the moments along each axis in the vertical direction
are parallel to each other and opposite to those on the neighboring axes. In all
three types, half of the moments point along one direction, and the other half point
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 174

174 Introduction to Solid-State Physics

along the opposite one, so that the total moment of the sample vanishes. Nonethe-
less, the primitive unit cells, their bases and their diffraction patterns are different
from the “chemical” ones.

Fig. 3.27: Antiferromagnetic arrangements of the types C, A, and G.

The magnetic moment at lattice point R, in each of the structures in Fig. 3.27,
can be written as
μ(R) = μ0 eiK·R . (3.71)
The vectors K in each of the patterns are
KA = (π/a)ẑ , KC = (π/a)(x̂ + ŷ) , KG = (π/a)(x̂ + ŷ + ẑ) . (3.72)
For example, for R = (n1 x̂ + n2 ŷ + n3 ẑ)a, the factor exp[iK · R] equals (−1)n3
for the type-A order, (−1)n1 +n2 for the type C, and (−1)n1 +n2 +n3 for the type G.
In the one-dimensional pattern of Fig. 2.42, the ordering C is given by Eq. (3.71)
with K · R = (π/a)(na) = πn.
None of the vectors in Eq. (3.72) is a reciprocal-lattice vector of the SC lattice,
which represents the “nuclear” crystal [these vectors obey Eq. (3.17), exp[iG ·
R] = 1, and no sign change can occur]. This crystal is built of the vectors G =
(2π/a)(hx̂ + kŷ + ẑ). Nevertheless, the vectors in Eq. (3.72) are parallel to the
reciprocal-lattice vectors, and therefore each of them is normal to a family of planes
in the “nuclear” lattice. The factor exp[iK · R] in Eq. (3.71) has a certain sign
on each plane in the family, and the opposite sign on the neighboring planes. For
instance, inspecting the structure G (the left panel in Fig. 3.27) one observes that
KG is normal to all (111) planes of the original cubic lattice, but the signs of the
magnetic moments alternate from one plane to the next. In a similar fashion, the
signs change alternately as one passes among the (001) planes in structure A, and
among the (110) planes in structure C. Since the moments repeat themselves at
each second plane, the magnetic unit cell is twice that of the nuclear one, along the
corresponding vector K.
This conclusion is based on the notion that in the lattices discussed here, the
magnetic crystal is built of two sublattices, one comprising the positive moments,
the other the negative ones. Therefore it can be viewed as a lattice with a base
of two opposite moments. For example, the structure G resembles the lattice of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 175

Radiation scattering off crystals 175

table salt shown in Fig. 2.17, and hence it is an FCC lattice where each lattice site
represents a base of two opposite moments. In the cubic representation of the FCC
lattice, the lattice constant is twice that of the original cubic lattice. Similar to the
situation with the table salt crystal [Eq. (3.52)], the magnetic structure factor is
FM (hk ) = μ0 [1 + eiπ(h+k) + eiπ(h+) + eiπ(k+) ](1 − eiπ ) . (3.73)

It yields Bragg peaks only when all Miller indices are odd; the “nuclear” peaks
necessitate even Miller indices.
Problem 3.35.
Determine the magnetic structure factors of the antiferromagnetic A and C ar-
rangements in Fig. 3.27. What are the conditions for the appearance of the nuclear
and magnetic Bragg peaks?
Manganese oxide. Returning to the cubic unit cell of manganese oxide, Fig.
3.26(b), one recalls that the lattice constant of this cell, denoted a , is twice that of
the original lattice. It follows that this cell contains eight unit cells of the nuclear
FCC original lattice. Since that cell comprises four manganese ions, the magnetic
unit cell contains 32 ions of Mn. The coordinates of these ions are given by ri =
xi1 a1 + xi2 a2 + xi3 a3 = a (xi1 x̂ + xi2 ŷ + xi3 ẑ), where 4(xi1 + xi2 + xi3 ) is an even
number, and where 4xim = 0, 1, 2, 3, with m = 1, 2, 3 (check!). As noted, the
planes containing parallel moments are parallel to the (111) plane. It is then found
that the antiferromagnetic order proposed by Shull et al. obeys Eq. (3.71), with
K = (2π/a )(x̂ + ŷ + ẑ). Therefore μ(ri ) = μ0 (−1)2(x1 +x2 +x3 ) , as can indeed be
i i i

seen in the figure.  Hence

FM (hk ) = μ0 ei(K−G)·ri
ri

= μ0 [1 − (−1) ][1 − (−1)k ][1 − (−1) ][1 − (−1)(h+k)/2 − (−1)(k+)/2 − (−1)(h+)/2 ]

h

= μ0 [1 + (−1)h+k + (−1)h+ + (−1)k+ ]

× [1 − (−1)h+k+ ][1 − (−1)(h+k)/2 − (−1)(k+)/2 − (−1)(h+)/2 ] .
All 32 terms are listed in the second step. The first square brackets in the third step
represent the FCC lattice, as e.g., in Eq. (3.53), and the factors multiplying them
represent the eight manganese ions of the base of this lattice. Hence all Miller indices
must be odd, but in addition, also (h+ )/2, (k+ )/2, and (h+k)/2 have to be odd as
well. The result is FM (hk ) = 32μ0 , while for any other combination of indices the
result is zero. Indeed, the indices identified by Shull and his collaborators comply
with these conditions. Note also that the magnetic symmetry implies different
results for (311) and for (311): the structure factor is zero for the first, while the
one for the second yields the maximal amplitude.
Problem 3.36.
Show that the scattering angles in Fig. 3.26 indeed correspond to the lattice con-
stants and the “nuclear” and magnetic structures of manganese oxide, as deduced
by Shull and coworkers. Recall that λ = 1.057Å.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 176

176 Introduction to Solid-State Physics

The work of Shull and his collaborators opened a novel era in the research of
magnetism. Nonetheless, later studies that exploited stronger neutron sources have
revealed that some of the lines are split, for example, (115) and (333), and therefore
the actual structure of manganese oxide is rhombohedral, with α = 90.62◦ . It has
also been concluded, in contrast with the previous understanding, that the magnetic
moment vectors lie within the (111) planes.
More complex magnetic structures. The antiferromagnetic structures dis-
cussed above are all based on two sublattices built of opposite-sign moments, but
there exist other configurations. The base of the structure E in Fig. 2.42 contains
three magnetic ions; hence a crystal of this type contains three sublattices. The
structure portrayed in Fig. 2.43 and discussed in problem 2.36 also contains more
than two sublattices. When the different sites in the triangular lattice indeed con-
tain magnetic moments with an angle 120◦ in-between each pair of them, then the
magnetic form factor is a vector,
FM (hk) = μ1 + μ2 e2πi(h+k)/3 + μ3 e4πi(h+k)/3 .
Due to the 120◦ −angle, μi · μj = −μ20 /2 for i = j, which implies that |FM |2 =
3μ20 {1 − cos[2π(h + k)/3]}. Thus, there are no Bragg peaks when h + k is an integer
multiple of 3.
Another, rather extreme, situation is described in problem 2.35. The nuclear
lattice is a SC crystal with a lattice constant a, but the magnetic moments are given
by μ(R) = μ0 cos(Q · R), where Q = (x̂ + ŷ + ẑ)τ /a. Since τ is irrational, the unit
cell is infinite. The magnetic scattering amplitude is thence R μ(R) exp[−iq·R] =
μ0 R (exp[i(Q − q) · R] + exp[−i(Q + q) · R])/2. Each of these two sums resembles
those appearing in Eq. (3.14), and thus
 (2π)3 
μ(R)e−iq·R = μ0 [δ(q − Q − G) + δ(q + Q − G)]/2 . (3.74)
V
R G

The Bragg peaks are “incommensurate”: they appear in pairs around each
point in the reciprocal lattice, q = G ± Q. The Bragg peaks that belong to the
“nuclear” lattice still appear at the reciprocal-lattice sites.

3.14 Answers for the problems in the text

Answer 3.1.
a. Equation (3.1) yields 1.545Å, 1.541Å, and 1.39Å, respectively.
b. The atomic numbers of copper, chromium, cobalt, and molybdenum are 29, 24,
27, and 42, respectively. It follows that the three wave lengths given in part (a) have
to be multiplied by (29/z)2 to obtain the wave lengths of the atom with the atomic
number z. For chromium, one finds 2.256Å, 2.250Å, and 2.03Å, respectively; for
cobalt 1.782Å, 1.778Å, and 1.60Å, respectively; for molybdenum 0.737Å, 0.735Å,
and 0.66Å, respectively.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 177

Radiation scattering off crystals 177

Answer 3.2.
a. As the three points are on the plane, the vectors connecting each two of them
are also on that plane. Therefore the two vectors, a1 = R2 − R1 = a(x̂ − ŷ) and
a2 = R3 − R2 = a(ŷ − ẑ) can be chosen as √ the lattice vectors on the plane. The
length of each such vector is |a1 | = |a2 | = a 2, and the angle in-between them is
given by cos γ = a1 · a2 /(|a1 ||a2 |) = −a2 /(2a2 ) = −1/2, i.e., γ = 120◦ . It follows
that this is a triangular lattice. Its unit cell √is, e.g., a rhombus like in Fig. 2.4, whose
area is S = |a1 × a2 | = |a1 ||a2 | sin γ = a2 3. It is possible to reach an arbitrary
point on the plane upon beginning, for instance, at the point R1 and adding to it
arbitrary vectors oriented along the planar lattice vectors,

a
r = R1 + y1 a1 + y2 a2 = [(2y1 − 1)x̂ + (2y2 − 2y1 + 1)ŷ + (1 − 2y2 )ẑ] ,
2

where y1 and y2 are numbers (not necessarily integers). This is the equation of
the plane. This plane cuts the x̂−axis at the point whose coefficients of ŷ and ẑ
vanish, i.e., for y1 = 1 and y2 = 1/2, and hence at r = (a/2)x̂. Clearly this is
not a lattice point, but it is located at the mid point of the cubic-lattice vector
along this direction. Similarly, the plane cuts the two other axes at r = (a/2)ŷ and
r = (a/2)ẑ. A plane that cuts the axes at these points is indeed shown in Fig. 3.6,
as one of the planes marked there by (222), in the BCC panel. As explained, the
notation (222) indicates that the plane closest to the origin cuts each of the cubic
axes at the mid points of the cubic lattice vectors.
b. As the plane in question should pass through the origin, its equation is r =
y1 a1 + y2 a2 . Its distance from the previous plane equals the height of the pyramid
whose base is the triangle connecting the three points chosen at the initial stage in
(a), and its apex is√at the origin. The length of each edge on the base of the pyramid
is |a1 | =√|a2 | = a 2, and the length of each of the faces’ edge is |R1 | = |R2 | =
|R3 | = a 3/2. Each such edge forms a right triangle, together with the height of
the pyramid and√2/3√of the height  of the equilateral triangle in the base, whose
length is (2/3)(a 2) 3/2 = a 2/3. By the Pythagorean √ theorem, the distance
between the two planes is d = 3a2 /4 − 2a2 /3 = a/(2 3). In order to find the
equation of the next plane, one has to locate a lattice point whose distance from it
is the smallest. From Fig. 3.6 it is seen that such a point is R4 = ax̂, and therefore
the equation of the plane is r = R4 + y1 a1 + y2 a2 . As the first plane cuts the x̂−axis
at (a/2)x̂, and the second plane at ax̂, it follows that the equation of the m−th
plane is r = m(a/2)x̂ + y1 a1 + y2 a2 . Several of these planes are depicted in Fig. 3.6
where they are denoted by (222). √
2
c. The planar density of lattice points √ is 1/S = 1/(a 3). The distance between
two neighboring planes is d = a/(2 3), and therefore the density of lattice points
in space is 1/(Sd) = 2/a3 . This is precisely the density of the BCC lattice, whose
cubic unit cell (of volume a3 ) contains two points. Hence, the family of planes
indeed includes all lattice points.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 178

178 Introduction to Solid-State Physics

Answer 3.3.
The rays, as depicted in Fig. 3.28, hit the points A and B, whose distance from one
another is L. The angle between AB and the distance between the planes, d =AC,
is α. The segments AF and AD are normal to the incoming ray and to the scattered
one, respectively. The angles between AD and AC and between AF and AC are
both θ. As seen in the figure, the difference between the two paths that the rays
traverse is

FB + BD = L sin(θ − α) + L sin(θ + α) = 2L sin θ cos α = 2d sin θ .

Fig. 3.28

Answer 3.4.
a. Since the unit vector ẑ is normal to the plane of the lattice, it can be written as
ẑ = a1 × a2 /S. Exploiting the identity

u × [v × w] = (u · w)v − (u · v)w , (3.75)

(prove!) yields

b1 = 2πa2 × ẑ/S = 2πa2 × [a1 × a2 ]/S 2 = 2π[|a2 |2 a1 − (a1 · a2 )a2 ]/S 2 .

The second equality is proven in the same way.

b. Using part (a) gives

b1 × b2 = (2π)2 [|a1 |2 |a2 |2 − (a1 · a2 )2 ]a1 × a2 /S 4 .

Exploiting in addition the identity S 2 = |a1 × a2 |2 = |a1 |2 |a2 |2 − (a1 · a2 )2 yields

the area of the unit cell of the reciprocal lattice, Srec = |b1 × b2 | = (2π)2 /S.

Answer 3.5.
In addition to the vectors displayed in Fig. 3.11, the lattice vectors of the simple
hexagonal lattice include a3 = cẑ, with a3 = c. Using Eq. (3.23) yields the same
results for the two vectors in the plane, and in addition b3 = 2πẑ/c.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 179

Radiation scattering off crystals 179

Answer 3.6.
The volume of the unit cell of the reciprocal lattice is (up to a sign) Vrec = [b1 ×
b2 ] · b3 . The vectorial product is, from Eq. (3.75),

b1 × b2 = (2π/V )2 [a2 × a3 ] × [a3 × a1 ] = [(2π)2 /V ]a3 .

Consequently, since a3 · b3 = 2π, one finds Vrec = (2π)3 /V .

Another method to solve this problem is the following. The volume V = a1 ·
[a2 × a3 ] is equivalent to the determinant of a 3×3 matrix, whose rows contain the
cartesian components of the lattice vectors a1 , a2 , and a3 . Similarly, the volume
Vrec = [b1 ×b2 ]·b3 is the determinant of a matrix whose columns contain the vectors
b1 , b2 , and b3 . (Note that interchanging the rows with the columns in a square
matrix leaves the determinant unaltered.) The product of these two determinants
is the determinant of the product of the matrices. Equation (3.20) states that the
product matrix is diagonal, with the value 2π for each diagonal entry. Therefore
the determinant of the product is (2π)3 , qed. In one dimension ab = 2π. One may
repeat the proof presented above for two dimensions, to obtain SSrec = (2π)2 , as
has been found in problem 3.4. Hence, in d dimensions, the result is V Vrec = (2π)d .

Answer 3.7.
A straightforward calculation of the lattice vectors of the reciprocal to the reciprocal
lattice is to use Eq. (3.23). One of these vectors is (2π/Vrec )b1 × b2 , where Vrec is
the volume of the unit cell in the reciprocal lattice. The solution of problem 3.6 gives
b1 × b2 = [(2π)2 /V ]a3 . Hence, (2π/Vrec )b1 × b2 = (2π/Vrec )[(2π)3 /V ]a3 = a3 . The
last step exploits the relation V Vrec = (2π)3 , see problem 3.6. The other two lattice
vectors, a1 and a2 , are obtained in a similar way. It follows that the reciprocal of
the reciprocal lattice is the original lattice. An alternative solution is based
on the symmetry of Eqs. (3.20) for the three lattice vectors a1 , a2 , and a3 , and the
three vectors b1 , b2 , and b3 . This symmetry implies that the first triplet contains
the vectors of the reciprocal to the lattice whose vectors form the second triplet.

Answer 3.8.
The lattice vectors of the BCC lattice are given in Eq. (2.5) and the volume
of the unit cell is V = a3 /2. Exploiting Eq. (3.23) gives b1 = (2π/a)(ŷ + ẑ),
b2 = (2π/a)(ẑ + x̂), and b3 = (2π/a)(x̂ + ŷ). A comparison with Eq. (2.6)
shows that this is an FCC lattice, where the length of the cube’s edge is 4π/a.
Alternative method: the reciprocal of the BCC lattice is the FCC one. According
to the solution of problem 3.7, the reciprocal of the BCC lattice is the FCC one.
One has still to determine the lattice constant. Denoting the lattice constant of
the reciprocal lattice of the BCC lattice (which is an FCC lattice) by b, then
Vrec = b3 /4. On the other hand, V = a3 /2, and so, according to the solution of
problem 3.6, b3 /4 = Vrec = (2π)3 /V = 16π 3 /a3 , which yields b = 4π/a.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 180

180 Introduction to Solid-State Physics

Answer 3.9.
As seen in Fig. 3.11, the reciprocal of the triangular lattice is a triangular lattice as
well. Figure 2.5 demonstrates that the Wigner-Seitz cell of the triangular lattice is a
hexagon. It therefore follows that the first Brillouin zone of the triangular reciprocal
lattice is a hexagon. The first three Brillouin zones of this lattice are depicted in
Fig. 3.29. Each of the lines is perpendicular to a reciprocal-lattice vector that
connects the origin with one of the lattice sites (the dark circles), cutting it at its
middle. The nth Brillouin zone is reached after crossing n − 1 border lines.

Fig. 3.29

Answer 3.10.
As discussed in problem 3.2, an arbitrary plane in this family of planes is given by
the equation r = m(a/2)x̂ + x1 â1 + x2 a2 , where a1 = a(x̂ − ŷ) and a2 = a(x̂ − ẑ).
The plane closest to the origin cuts the lattice vectors at a1 /2 = ax̂/2, a2 /2 = aŷ/2,
and a3 /2 = aẑ/2. It follows from Eq. (3.29) that x1 = x2 = x3 = 1/2, and the
Miller indices are h = k = = 2. Indeed, these planes are marked in Fig. 3.6 by
(222), in the line belonging to the BCC lattice. The reciprocal-lattice vector that
corresponds to these indices is G0 = hb1 + kb2 + b3 = (4π/a)(x̂ + ŷ + ẑ). A
scalar product of this vector with the equation of the mth plane, r = m(a/2)x̂ +
x1 a(x̂ − ŷ) + x2 a(x̂ − ẑ), indeed yields G0 · r = 2πm, like in Eq. (3.18). It is
√ found
in problem 3.2 that the distance between neighboring planes √ is d = a/(2 3), as
indeed obtained from Eq. (3.27), d = 2π/G0 , with G0 = 4π 3/a.

Answer 3.11.
a. Inverting the sign of any nonzero index produces a factor of 2 in the multiplicity.
Another factor enumerates all possible permutations of the triplet. Hence, the
answers are 6, 48, 24, 8, 24, and 12, respectively.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 181

Radiation scattering off crystals 181

b. There are eight such planes. For example, the (111) plane joins the tips of the
lattice vectors along the positive directions of the cartesian axes (as in the lower
right part of Fig. 3.6, for the SC lattice). All other planes in the family join such
tips along all other possible directions. The structure thus created is the octahedron
shown on the left in Fig. 3.30. The origin is located at the center of the octahedron,
and the vertices are at the points ±ax̂, ±aŷ, and ±aẑ.
c. The (110) plane is parallel to the ẑ−axis, and passes through two opposite edges
of the cubic unit cell (as in the lower central part of Fig. 3.6, for the SC lattice).
There are four such planes that are parallel to each axis, and hence altogether 12
planes that encompass the origin. They form a symmetric body with 12 square
faces, as displayed in the right panel of Fig. 3.30. This body is the rhombic
dodecahedron; it is identical to the Wigner-Seitz cell of the FCC lattice, Fig.
2.19(a). See also Appendix A.
d. When all indices h, k, and differ from each other, there are six permutations
interchanging them. Each permutation with nonzero indices possesses eight planes
obtained by inverting the signs of each of the indices. Hence the maximal number
of planes in the family is 48.

Fig. 3.30

Answer 3.12.
Since the plane passes through the points a1 /h, a2 /k, and a3 / , it follows that the
independent vectors (a1 /h − a2 /k) and (a1 /h − a3 / ), which connect pairs of such
points, are located on the plane. Then the scalar product of each of these two vectors
with G vanishes, and therefore G is normal to the plane. The distance between the
origin and the plane is thus equal to the projection of each of the original vectors,
e.g., a1 /h, on the direction of G. This projection is (a1 /h) · G/G = 2π/G, using
Eqs. (3.20) and (3.21).

Answer 3.13.
It is found in problem 3.8 that b1 = 2π(0, 1, 1)/a, b2 = 2π(1, 0, 1)/a, and b3 =
2π(1, 1, 0)/a, and hence G = (2π/a)(k + , h + , h + k). It follows that h = k + ,
k  = h + , and  = h + k. Adding together these three relations results in
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 182

182 Introduction to Solid-State Physics

h + k  +  = 2(h + k + ). Thus, the sum of the “new” indices is even: either they
are all even or two of them are odd and the third is even.

Answer 3.14.
The table below lists all possible triplets {hk } with h ≤ k ≤ (so that each triplet
appears only once), the values of S = h2 + k 2 + 2 , the multiplicity√ √p, and the
angle θ [the latter is derived from Eq. (3.30), sin θ = [λ/(2a)] S = S/6]. The
multiplicity for {00 } is 6 since there are three positions for the nonzero index , and
it can have two signs, ±1; p(0kk) = 12 since there are three places to put the zero,
and each of the k’s can have two signs; p(kkk) = 8, as each of the indices can have
two signs. For k = p(0k )=24, as there are 6 permutations of the three indices,
and the two nonzero ones can have two signs; p(kk ) = 24 for k = because there
are three places for the index , and each of the three indices can have two signs.
When all three indices are nonzero and differ from each other the multiplicity is
48 (six permutations and two signs for each of the indices). For each of the values
S = 9, 17, 18 there are two sets of indices that yield the same scattering angle. The
values S = 7, 15 do not appear since there are no three integers that give them.
Since the scattering angle is 2θ, it cannot exceed 90◦ in forward scattering and
therefore there are no scattering points for S > 18.

Answer 3.15.
a. In the orthorhombic lattice G0 = 2π(hx̂/a1 +kŷ/a2 + ẑ/a3 ) [Eq. (3.22)]. Hence,
λ 2  h2 k2 2
sin2 θ = 2 + 2 + 2 .
4 a1 a2 a3
In the tetragonal lattice a1 = a2 , and thus
λ 2  h2 k2 2
sin2 θ = 2 + 2 + 2 .
4 a1 a1 a3
b. For a given wave length, the smallest angles are due to the smallest values of the
expression in the brackets. In the present example there are two wave lengths, 1.39Å
and 1.54Å, and consequently the scattering angles come in pairs. The smallest ones
result from the smallest values of h2 + k 2 + 2 (a1 /a3 )2 = h2 + k 2 + 4 2 /9, i.e.,
(hk ) = (001), (100), (101), and (002), for which h2 + k 2 + 4 2 /9 = 4/9, 1, 15/9,
and 16/9. Hence, the smallest eight angles obey sin2 θ = 0.54, .066, .121, .148,
.174, .214, .215, and .242. Note that the fourth angle of the longer wave length is
smaller than the third one of the shorter wave length. The angles themselves are
θ = 13.4◦ , 14.9◦ , 20.3◦ , 22.6◦ , 24.7◦ , 27.56◦ , 27.6◦ , and 29.4◦ .

Answer 3.16.
a. In the geometry of Fig. 3.3(b), the scattered beam does not hit the screen
when the scattering angle is greater than θmax = 45◦ . Using θmin = 20◦ yields
sin2 θmax / sin2 θmin = 4.27, and hence h2 +k 2 + 2 ≤ 4. For the SC lattice, this condi-
tion allows solely the families {100}, {110}, {111}, and {200}. The scattering angles
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 183

Radiation scattering off crystals 183

h k S p θ
0 0 1 1 6 9.6
0 1 1 2 12 13.6
1 1 1 3 8 16.8
0 0 2 4 6 19.5
0 1 2 5 24 21.9
1 1 2 6 24 24.1
0 2 2 8 12 28.1
0 0 3 9 6 30
1 2 2 9 24 30
0 1 3 10 24 31.8
1 1 3 11 24 33.6
2 2 2 12 8 35.3
0 2 3 13 24 36.9
1 2 3 14 48 38.6
0 0 4 16 6 41.8
0 1 4 17 24 43.4
2 2 3 17 24 43.4
0 3 3 18 12 45
1 1 4 18 24 45
1 3 3 19 24 -
0 2 4 20 24 -
1 2 4 21 48 -

are then 2θ = 40◦ , 58◦ , 73◦ , and 86◦ , respectively. Taking into account the multi-
plicities (see problem 3.14) yields for the intensities I = 6I0 , 12I0 , 8I0 , and 6I0 ,
where I0 is a multiplicative factor that expresses the scattering intensity off each
lattice point.
b. As found in problem 3.15, sin2 θ/ sin2 θ(100) = h2 + k 2 + 2 (a1 /a3 )2 . There-
fore, the scattering angles for (hk ) =(100), (010), (110), (200), and (020) are
not changed; there appear scattering rings for three of the previous angles, 2θ =
40◦ , 58◦ , and 86◦ . However, the scattering intensity of each is 4I0 , since the multi-
plicities are lower. In addition, the angles for which
sin2 θ/ sin2 θ(100) = (a1 /a3 )2 , 1 + (a1 /a3 )2 , 2 + (a1 /a3 )2 , and 4(a1 /a3 )2 ,
gives rise to constructive interference. These angles are 2θ = 36◦ , 55◦ , 70◦ , and
77◦ , with the scattering intensities I = 2I0 , 8I0 , 8I0 , and 2I0 . Altogether, there
are seven scattering circles in place of four.
c. In this case sin2 θ/ sin2 θ(100) = h2 + k 2 (a1 /a2 )2 + 2 (a1 /a3 )2 . Some of the
previous angles attain new values, while others are left unchanged. In addition
the peaks resulting originally from (hk ) = (010), (011), and (020) are split, and
altogether there are ten circles. Assuming that a1 < a2 < a3 , the new rings are due
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 184

184 Introduction to Solid-State Physics

to the indices (hk ) =(001), (010), (100), (011), (101), (110), (111), (002), (020), and
(200). The only remaining multiplicity is that of sign changes of each of the indices,
and therefore the intensities are I = 2I0 , 2I0 , 2I0 , 4I0 , 4I0 , 4I0 , 8I0 , 2I0 , 2I0 ,
and 2I0 , respectively.

Answer 3.17.
From Eq. (3.37),
  b 
1 1 1 b
f(G) = d2 rf (r)e−iG·r = 2 dx 2 dye−i(Gx x+Gy y) ,
S S a 0 b 0
where S = a2 is the area of the unit cell. The integral is a product of two integrals,
each of the
 form
b
dxe−iGx x = (e−iGx b − 1)/(−iGx ) = 2e−iGx b/2 sin(Gx b/2)/Gx ,
0
and therefore
sin(Gx b/2) sin(Gy b/2)
f(G) = 4e−i(Gx +Gy )b/2 .
abGx abGy
In the b → 0 limit one finds sin(Gx b/2)/(Gx b) → 1/2 and therefore f(G) → 1/a2 .
In this limit Eq. (3.34) yields f (r) = (1/a2 ) G exp[iG·r], i.e., the “intensity” is the
same for all plane waves whose wave-vector differences are reciprocal-lattice vectors.
Indeed, in this limit the function f (r) in each cell approaches a delta function (it
diverges to infinity at the origin and vanishes elsewhere, and its integral equals
unity). Adding together the contributions from all cells gives f (r) → R δ(r − R),
and therefore the function is a delta comb, as encountered in Sec. 3.3: the result
obtained here is similar to Eq. (3.14).

Answer 3.18.
X
a. Consider first the one-dimensional integral I = −X
dxf (x). Changing the
−X X
integration variable from x to −x yields I = d(−x)f (−x) = −X dxf (−x).
X
The signs of integration bounds are inverted in the first step, and in the second
they are interchanged, which results in the opposite sign of the integral. Inverting
the sign of the integration
 variable in the expression for the structure factor gives
F (G) = d3 rn(r)eiG·r
 V

−iG·r
= 3
d rn(−r)e = d3 rn(r)e−iG·r = F (−G) = F ∗ (G) .
V V
The last step uses the fact that the scatterers’ density, n(r), is symmetric, n(r) =
n(−r). The integral in the third step is over the unit cell that is the reflection of
the original one, but the two cells are identical.
b. Using spherical coordinates in Eq. (3.46), with the ẑ−axis chosen to be along
G and with the notation μ = cos  θ, yields  ∞ 1
F (G) = 2π drr2 n(r) dμ exp[iGrμ] .
0 −1
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 185

Radiation scattering off crystals 185

The result depends solely on G = |G|. The angular integration is

 1
dμ exp[iGrμ] = 2 sin(Gr)/(Gr) ,
−1

as required.
c. In this case n(r) = |ψ100 (r)|2 = exp[−2r/aB ]/(πa3B ) is spherically symmetric.
Exploiting the result of the previous part yields
 ∞
4 1
F = drre−2r/aB sin(Gr) = .
Ga3B 0 [1 + (aB G/2)2 ]2
Indeed, the structure factor decays for long reciprocal-lattice vectors, but the de-
pendence on G = |G| is weak as G is smaller than 1/aB . The helium atom contains
two electrons in the ground state, but the “radius” of their orbits is aB /2 (because
of the atomic number z = 2). Therefore the result is F = 2/[1 + (aB G/4)2 ]2 . The
electrons are closer to the nucleus, and therefore the decay of the structure factor
with the length of the reciprocal-lattice vector is slower.

Answer 3.19.
The structure factor is

F (G) = FA + FB eiG·r2 = FA + FB e2iπ(h+k)/3 ,

and therefore the scattering intensity is

|F (G)|2 = FA2 + FB2 + 2FA FB cos[2π(h + k)/3] .

There are two possible values for this intensity, |F (G)|2 = (FA + FB )2 when h + k
is an integer multiple of 3, and |F (G)|2 = (FA + FB )2 − 3FA FB when it is not. This
expression reproduces the result for graphene when FA = FB , and yields a vanishing
intensity at the points where h + k is an integer multiple of 3 and FA = −FB . In
the latter case, the reciprocal lattice is the hexagonal one.

Answer 3.20.
Figure 2.9 displays the hexagonal lattice built of a hexagonal lattice in the XY plane
and an additional lattice vector, a3 = cẑ. The reciprocal of the simple hexagonal
lattice is given in problem 3.5. The present case is a lattice with a base: the unit
cell contains four carbon atoms, at the origin, and at the points r2 = (a1 + a2 )/3,
r3 = a3 /2, and r4 = r3 + 2r2 . The structure factor is hence
 
F (G) = F0 1 + e2iπ(h+k)/3 + eiπ (1 + e4iπ(h+k)/3 ) .

It follows that F (G) = 2F0 (1 + cos[2π(h + k)/3]) when is even. On the other
hand, when is odd, then F (G) = 2iF0 sin[2π(h + k)/3], and therefore the intensity
vanishes when h + k is an integer multiple of 3, similar to the situation described
in the previous problem.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 186

186 Introduction to Solid-State Physics

Answer 3.21.
The lattice vectors of the reciprocal to the FCC lattice are given in Eq. (3.24).
Using them, with the coefficients h, k, and , yields
G = hb1 + kb2 + b3 = (2π/a)(h x̂ + k  ŷ +  ẑ) ,
where the right-side equality is written in terms of the reciprocal-lattice vectors of
the simple cubic lattice, with h = h − k + , k  = h + k − ,  = −h + k + .
Substituting the latter into the right hand-side of the equation for G shows that
h + k  = 2h, h +  = 2 , and k  +  = 2k. The sum of two integers is even only
when they both are even or they are both odd. Hence, constructive interference
necessitates the cubic coefficients h , k  , and  to be all even or all odd, as obtained
also from the calculation for the cubic unit cell.

Answer 3.22.
a. The inter-planar distance in a cubic √ lattice, for the family given by the indices
h, k, and is d(hk ) = 2π/G = a/ h2 + k 2 + 2 . As mentioned in Sec. 3.6, the
site density in each plane is d/V , and therefore the largest density is attained at the
smallest value of h2 + k 2 + 2 . In the BCC lattice, the sum h + k + should be even
and hence the smallest value of h2 + k 2 + 2 is realized when two of the indices √ are
1 and the third is 0, e.g., (110). Then, √ the inter-plane distance is d(110) = a/ 2,
and the site density in each plane is 2/a2 (recall that V = a3 /2).
b. The plane (110) in the BCC lattice is parallel to the ẑ−axis, and passes through
two opposite edges of√the cube (see Fig. 3.6). This cut creates a rectangle, whose
edges are a and b = a 2; it contains an additional lattice site at its√center. There-
fore, it is a rectangular-centered
√ lattice. The site density on it is 2/a2 , and the
inter-planar distance is a/ 2.
c. Within the plane, the lines can be marked by the reciprocal-lattice vectors normal
to them. As this is√a rectangular-centered lattice, these vectors are G = 2π(hx̂/a +
kŷ/b), with b = a 2. The inter-line distance is d = 2π/G √ = a/ h2 + k 2 /2, with
the maximum distance attained for (hk) = (01), dmax = a 2. The vector G is then
parallel to the ŷ−axis; consequently the densest line is parallel to the x̂−axis, i.e.,
to the ẑ−axis of the original cubic lattice. The inter-site distance on the densest
line is a.
d. In the FCC lattice, the indices must be either all even or all odd. The lowest √
value of h2 + k 2 + 2 is obtained for h = k√= = 1. Then d(111) = a/ 3
and consequently the planar density is 4/(a2 3) (the volume of the unit cell is
V = a3 /4). The (111) planes in the FCC lattice are those containing the red or the
green spheres in the lowest row of Fig. 2.24(a). As explained there, these planes
comprise triangular
√ lattices, which indeed are the densest planar ones, with the
density 4/(a√2 3). The reciprocal-lattice vectors of this lattice are G = 2π[hx̂ +
(2k − h)ŷ / 3]/a, where x̂ and ŷ define√the coordinate
√ system in that plane √ (see
Fig. 3.11). It follows that d = 2π/G = a 3/[2 h2 + k 2 − hk], and dmax = a 3/2.
This result is obtained for (hk) = (0 ± 1), (hk) = (±10), and for (hk) = ±(11). The
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 187

Radiation scattering off crystals 187

corresponding vectors, G(10) = ±b1 , G(01) = ±b2 , and G(±1 ± 1) = ±(b1 + b2 ),

are normal to the vectors a2 , a1 , and (a1 − a2 ). It follows that the densest lines on
the triangular plane are along the directions connecting nearest-neighbors on this
lattice, as indeed can be expected without carrying out the calculation.

Answer 3.23.
a. Figure 3.21 shows that the interference is constructive for 2θ ≈ 27◦ , 31◦ , 46◦ , 54◦ ,
and 56◦ , which correspond to h2 +k 2 + 2 = 3, 4, 8, 11, and 12. This sequence indeed
fits the FCC lattice (Fig. 3.20); the values confirm the Miller indices that appear
in the figure. They yield approximately sin2 θ/(h2 + k 2 + 2 ) ≈ 0.018. Using this
result in Eq. (3.31) yields [λ/(2a)]2 ≈ 0.018; hence a ≈ 1.545Å/2/0.134 ≈ 5.76Å.
This is only an approximation, because it is hard to read the precise numbers in
the figure. The value cited in the literature is 5.65Å.
b. The intensities decay with |G|2 , which in turn is proportional to h2 + k 2 + 2 .
Ignoring this decay, the first scattering angle corresponds to the family of planes
{111}, and thus is obtained from 8 planes. The second angle comes from the
family {200}, comprising 6 planes. Hence, the intensities ratio is I(111)/I(200) =
8|F (111)|2 /(6|F (200)|2 ). On the other hand, F (111)/F (200) = (FNa − FCl )/(FNa +
FCl ). Subtracting the background (about 30 in the arbitrary relative units of the
figure), one finds I(111) ≈ 40 and I(200) ≈ 270. Assuming the two ions to be
point-like scatterers, their structure factors are real and independent of G, i.e.,
x = FNa /FCl is real. Then the equation (8/6)(x − 1)2 /(x + 1)2 ≈ 40/270 yields
two solutions, FNa /FCl ≈ 2 and FNa /FCl ≈ 0.5. As chlorine has more electrons, it
seems that the lower value should be chosen. Sodium has 10 electrons while chlorine
possesses 18. Therefore a rough estimate would have yielded FNa /FCl ≈ 10/18 =
0.55. Thus, the estimate found here is quite plausible.
c. The assumption that FNa /FCl is real is plausible, as the electrons’ distri-
bution around each ion is spherically-symmetric (see problem 3.18). Assuming
nonetheless that this ratio is complex, one obtains |F (111)/F (200)|2 = (|x|2 + 1 −
2|x| cos φ)/(|x|2 + 1 + 2|x| cos φ), where FNa /FCl = |x| exp[iφ]. There are now two
variables to be determined, and more information is required; this issue is known
in the literature as “the phase problem”.
d. From the given densities it follows that F (hk ) = zi /[1 + αi (h2 + k 2 + 2 )]2 ,
where αi = (πai /a)2 (see problem 3.18). Exploiting the value a = 5.65Å (from the
literature), one obtains FNa (hk ) = zNa /[1 + .279(h2 + k 2 + 2 )]2 and FCl (hk ) =
zCl /[1 + 1.01(h2 + k 2 + 2 )]2 . Adding to this the ratio I(111)/I(200) = 8[FNa (111) −
FCl (111)]2 /{6[FNa (200) + FCl (200)]2 } yields zNa /zCl ≈ .1, and .47. The solution
.47 is closer to the value .55 mentioned above, but does not improve the previous
estimate. As mentioned in the main text, there are other factors which affect the
measured intensities, that are not included in this analysis.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 188

188 Introduction to Solid-State Physics

Answer 3.24.
In this case, the expression in last brackets of Eq. (3.53) is proportional to (1 +
exp[iπk]). This expression is nonzero for even k, and then h and are even as
well. Substituting = 2  , k = 2k  , and h = 2h in the expression for G yields
G = (4π/a)(h x̂ + k  ŷ +  ẑ), i.e., the reciprocal-lattice vectors of a simple cubic
lattice with a lattice constant a/2. Indeed, replacing the two types of atoms in the
right panel of Fig. 2.17 by identical atoms produces such a SC lattice.

Answer 3.25.
In CsCl, the atoms at the origin and at the center of the cube of the cubic lattice have
different structure factors, and therefore the structure factor is (FCs +FCl exp[iπ(h+
k + )]). In contrast to the BCC lattice, here the interference is constructive for
all values of h, k, and , but the scattering amplitude is smaller for odd values of
h + k + (assuming that the ionic structure factors are real and have the same sign).
In diamond, the structure factor is proportional to (1 + exp[iπ(h + k + )/2]), times
the structure factor of the FCC lattice (explain!), which attains the value 2 when
h + k + is an integer product of 4, the value 1 ± i when this sum is odd, and the
value 0 otherwise. Keeping in mind that the three indices have to be all even or all
odd (since this is an FCC lattice), the allowed values of h2 + k 2 + 2 (up to 24)
are 3, 8, 11, 16, 19 , and 24. Ignoring the dependence of the structure factors on
the multiplicities and on the Miller indices, the ratio of the intensities in the three
types of sites are 4, 2, and 0. In zinc-blende, the ions at the origin and at the point
a(1, 1, 1)/4 are different, and consequently the structure factor is proportional to
(FZn + FS exp[iπ(h + k + )/2]). The result is thus FZn + FS , FZn − FS , or FZn ± iFS
when h+k+ is an integer multiple of 4, an odd multiple of 2, or is odd, respectively.

Answer 3.26.
a. Denote a1 = a and √ a3 = c. The lattice vectors of the HCP lattice are a1 =
(a, 0, 0), a2 = (a/2, a 3/2, 0) and a3 = (0, 0, c), see the discussion following Fig.
2.24. The
√ unit cell contains a base comprising two atoms, √ r2 =
at the origin and at √
(a/2, a 3/6, c/2). Using Eq. (3.23) yields b1 = 2π(x̂ − ŷ/ 3)/a, b2 = 4πŷ/(a 3),
and b3 = 2πẑ/c, as also found in problem 3.5. Hence, r2 · G = π + 2π(h + k)/3,
and the structure factor is F (G) = 1 + exp[ir2 · G]. It vanishes when + 2(h + k)/3
is odd, e.g., h + k = 3m, where m is an arbitrary integer, and is odd. Interestingly
enough, the same condition for a null structure factor is derived in problem 3.20 for
graphite.
b. exp[iπ{ + 2(h + k)/3}] can be 1, exp[±iπ/3], exp[±2iπ/3], and −1, for which
|F (G)|2 = 4, 3, 1, and 0. √
c. Since G(hk ) = 2π(h, (2k − h)/ 3, a/c)/a, Eq. (3.30) gives
sin2 θ = λ2 |G/(4π)|2 = λ2 [(h2 − hk + k 2 )/(3a2 ) + 2
/(4c2 )] .
For the ideal HCP lattice this expression is proportional to (h2 − hk + k 2 + 9 2 /32).
The smallest values are for (hk ) = (001), (100), (002), (101), (102). The values
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 189

Radiation scattering off crystals 189

(100) and (110) yield the same angle. Similarly, the pairs (101) and (111), and also
(102) and (112), give rise to identical angles. According to part (a), the structure
factor vanishes for the triplet (001). Hence the smallest angle corresponds to (hk ) =
(100) [or (110)], i.e., sin2 θmin = λ2 /(3a2 ). The next three angles, corresponding to
(hk ) =(002), (101), (102), are given by sin2 θ/ sin2 θmin = 9/8, 41/32, 17/8.
d. The hexagonal unit cell of wurtzite (see Fig. 2.25 and problem 2.17) con-
tains, in addition to the atoms at the origin and at r2 which represent, e.g.,
the sulphur √ions, two additional zinc ions at r3 = (0, 0, 5c/8) = (5/8)a3 and at
r4 = (a/2, a 3/6, c/8) = (a1 + a2 )/3 + a3 /8. The structure factor is thus
F (G) =FS (1 + exp[ir2 · G]) + FZn (exp[ir3 · G] + exp[ir4 · G]) ,
where r3 · G = 5π /4 , and r4 · G = π /4 + 2π(h + k)/3 .

Answer 3.27.
The structure factor of the body-centered orthorhombic crystal is proportional to
(1 + exp[iπ(h + k + )]), precisely as that of the BCC lattice. The structure factor
of the face-centered orthorhombic lattice is proportional to (1 + exp[iπ(h + k)] +
exp[iπ(h + )] + exp[iπ(k + )]), the same as that of the FCC lattice. As opposed,
the structure factor of the base-centered orthorhombic orthorhombic lattice is pro-
portional to (1 + exp[iπ(h + k)]), and yields constructive interference only when
(h + k) is even.

Answer 3.28.
The conditions for constructive interference in the tetragonal lattices are identical to
those of the rhombohedral ones, and hence are also identical to those of the cubic
lattices. The body-centered tetragonal lattice possesses constructive interference
only when (h + k + ) is even. In problem 3.15 this condition is fulfilled for the four
largest angles of the eight ones.

Answer 3.29.
d
Denote the vector by u = i=1 ui x̂i , where x̂i is a unit vector along the i−th axis.
The corresponding coefficient is ui = u · x̂i = u cos φi , and the angle between the
d d
vector and this axis is φi . Since u2 = u · u = i=1 u2i = u2 i=1 cos2 φi , it follows
d d
that i=1 cos2 φii = 1, leading to  i=1 cos2 φi  = 1. As the probabilities of all
directions are identical, all terms in the sum are identical as well, and hence each of
them equals 1/d. One may also calculate this average explicitly, exploiting spherical
coordinates. For instance, in three dimensions
 π  π  1  1
cos θ =
2 2
dθ sin θcos θ/ dθ sin θ = 2
dμμ / dμ = 1/3 .
0 0 −1 −1
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 190

190 Introduction to Solid-State Physics

Answer 3.30.
a. At zero temperature, the argument of the coth diverges, coth(∞) = 1, and
consequently 2W = [/(2M ω)]G2 . At high temperatures, kB T  ω, one may use
coth(x) ≈ 1/x, to obtain 2W = [kB T /(M ω 2 )]G2 .
b. As sin2 θ = λ2 G2 /(4π)2 , [see Eq. (3.30)] one finds the relation given in the
problem, with B = 8π 2 (uR )2 /3. Equation (3.59) implies that this coefficient is
independent of the lattice constants; it depends on the temperature, the atoms’
mass, and the frequency.
c. At high temperatures 2W = [kB T /(M ω 2 )]G2 . Substituting the Lindemann
criterion yields 2W = [c2L T /(3TM )](aG)2 , and the temperature appears solely as
T /TM . According to Eq. (3.21), G = hb1 + kb2 + b3 . For a cubic lattice,
bi = (2π/a)x̂i and consequently abi = 2πx̂i , as well as (aG)2 = 4π 2 (h2 + k 2 + 2 ),
are independent of the lattice constant a. In the general case, the reciprocal-lattice
vectors are given by Eq. (3.23), and hence abi is a dimensionless vector, determined
only by the ratio between the lattice constants.

Answer 3.31.
a. The decay factor W is inversely proportional to the mass of the atom in the
crystal, see Eq. (3.59). It follows that upon replacing the atom by a heavier isotope
(e.g., replacing hydrogen by deuterium) the decay caused by the lattice vibrations
is reduced.
b. Generally,
 
2W = (G · uR )2  = Gα Gβ uRα uRβ  = Bαβ Gα Gβ ,
α,β α,β

where the indices denote the vectors’ components in a certain basis. For instance,
for orthorhombic symmetry, the only nonzero coefficients are the diagonal ones,
and 2W = (2π)2 (B11 h2 /a21 + B22 k 2 /a22 + B33 2 /a23 ). Thus, the decay is different for
different lattice directions.

Answer 3.32.
a. The ratio between the numbers of the short and long segments in a Fibonacci
sequence approaches the golden ratio, x(n) = NL (n)/NS (n) → τ , see problem 2.27.
In the large n limit the distance from the origin to the n−th site contains NS (n)
segments of length 1 and NL (n) segments of length τ . In a similar fashion, the
distance from the origin to a point q in momentum space is proportional to m + nτ ,
and hence the ratio between the number of long segments and the number of the
short ones tends to τ . It follows that the points in momentum space which give rise
to high peaks are arranged as in a Fibonacci sequence.
b. The MATHEMATICA code is shown below. The vector vec in the code, whose
values are depicted in Figs. 3.32, keeps track of the pairs (q, A) for which A ≥ A1.
The right panel displays the case A1 = 0, and it contains all the points in the
diffraction picture. As seen, peaks of various heights are indeed quite dense. The
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 191

Radiation scattering off crystals 191

Fig. 3.31

left panel portrays only the points for which A ≥ A1 = 0.99, i.e., very close to the
maximal value A = 1. For these points the code computes also the vector DQ, that
contains the intervals between successive q’s, divided by the first difference. The
result is
DQ = {1., 1.61803, 1.00000, 1.61803, 1.61803, 0.99999, 1.61803,
1.61803, 1.00000, 1.61803, 1.00000, 1.61803, 1.61803, }
These intervals attain solely the values 1 and τ , and their sequence fits the Fibonacci
sequence SLSLLSLLSLSLL. Thus, the high peaks in the diffraction picture have
the same symmetry as the original lattice.

Fig. 3.32

Answer 3.33.
Place the measured sample (the collection of points on the right side of the figure)
on the XY plane, and the impinging beam (BO) along the ẑ−direction. The figure
shows that the projection of the vector k on the XY plane equals the projection of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 192

192 Introduction to Solid-State Physics

q on the same plane, i.e., it is equal to Gsurface = hb1 +kb2 , while its z−component
(its projection on BO) equals kz = k0 cos(2θ), with k0 = |k | = 2π/λ. On the other
hand, k02 = (hb1 + kb2 )2 + (kz )2 , and therefore the angles at which the scattering
is constructive are given by
(hb1 + kb2 )2 = k02 − (kz )2 = (2π/λ)2 sin2 (2θ) ,
where h and k are integers. The direction of the scattered beam is parallel to the
direction of the unit vector k̂ = k /k0 = [λ/(2π)](hb1 + kb2 ) + cos(2θ)ẑ. When the
planar screen is parallel to the XY plane and is located at a distance D from that
plane, then the triangle comprising the beam hitting the screen and its projections
on the plane of the screen and on the ẑ−axis is similar to the triangle built of the
vector k̂ and its projections along the same directions. This similarity implies that
the ratio between the projection of the scattered beam on the screen, V, and the
vector [λ/(2π)](hb2 + kb2 ) equals the ratio between D and cos(2θ). It follows that
the beam along k̂ hits the screen at a point whose coordinates on the screen plane
are given by the vector V = Dλ(hb1 + kb2 )/[2π cos(2θ)].

Answer 3.34.
There is only one component of q which is parallel to the crystal in this case:
Gsurface = hb (b is the reciprocal-lattice vector of the one-dimensional crystal).
Concomitantly, the vector q⊥ = q−hb represents an entire plane normal to the one-
dimensional reciprocal lattice. It follows that constructive interference is realized
whenever one of these planes cuts the Ewald’s sphere. Each such cut is a circle;
hence there are full lines in the diffraction pattern.

Answer 3.35.
In structure A the unit cell is doubled along the ẑ−direction, and therefore the
magnetic unit cell is tetragonal, with edges a, a, and 2a. The reciprocal-lattice
vector is G = (2π/a)(hx̂ + kŷ + ẑ/2), and the magnetic structure factor is
FM (hk ) = μ0 (1 − exp[iGz a]) = μ0 [1 − (−1) ]. The magnetic Bragg peaks ap-
pear only for odd ’s, while for the nuclear Bragg peaks is even, like in the original
SC lattice. In structure C, the base plane is similar to the planar lattice of table
salt, Fig. 2.7: the two marks there represent in the present case the two directions
of the magnetic moments.
√ Thus, the magnetic unit cell in the plane is a square,
with an edge a = a 2. As the planes parallel to the base are identical, the lattice
constant along the ẑ−axis remains√as in√the original lattice. The result is again a
tetragonal lattice, with the edges a 2, a 2, a (note that the volume remains as for
structure A. Why is this?) The vectors of this lattice in the plane are rotated by
45◦ relative to the original ones, and the reciprocal-lattice vectors are
2π  x̂ − ŷ x̂ + ŷ 
G= h +k + ẑ .
a 2 2
It follows that FM (hk ) = μ0 (1 − exp[iGx a]) = μ0 [1 − (−1)h+k ]. The magnetic
Bragg peaks appear solely for odd (h + k)’s, while for the nuclear ones this sum has
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 193

Radiation scattering off crystals 193

to be even. This condition reproduces the original cubic peaks, G = (2π/a)(h x̂ +

k  ŷ + ẑ), with h = (h + k)/2 and k  = (k − h)/2.

Answer 3.36.
At high temperatures there appear two angles, 2θ ≈ 23.8◦ and 46.5◦ . Assuming that
the structure is an FCC lattice, and hence the indices are all odd, these angles corre-
spond to (hk ) = (111) and (311), √ as indeed marked in the
√ original figure. The peak
at (hk ) = (111) yields a = λ 3/(2 sin θmin ) = 1.057 3/(2 sin 11.9◦ )Å = 4.44Å,
and the second peak gives rise to sin2 θ(311)/ sin2 θ(111) ≈ 11/3, as expected
from an FCC lattice. At low temperatures there are additional peaks at 2θ ≈
11.9◦ , 22.7◦ , 30◦ , 36◦ , which imply (hk√) = (111), (311), (331),
√ (511). The first
peak yields the lattice constant, a = λ 3/(2 sin θmin ) = 1.057 3/(2 sin 5.9◦ )Å =
8.85Å, and the other peaks yield sin2 (hk )/ sin2 (111) ≈ 11/3, 19/3, 27/3, in ac-
cordance with the FCC structure. In other words, the calculation verifies the
identification of Shull and coworkers.

3.15 Problems for self-evaluation

s.3.1.
Prove that the reciprocal of the trigonal lattice is also trigonal, with a lattice constant
b = 2π/[a sin α sin α ], where α is given by the relation cos(α /2) = 1/[2 cos(α/2)].

s.3.2.
Present explicit expressions for the distance between neighboring planes described by
the Miller indices h, k, for the following lattices: triclinic, orthorhombic, tetrago-
nal, hexagonal, monoclinic and trigonal. Determine the multiplicities of each family
of planes.

s.3.3.
a. Find the structure factor of a sphere of radius r0 carrying a constant volume
charge density, n0 = z/(4πr03 /3). What is the limit of F (G) when |G| → 0?
b. Find the structure factor which corresponds to the first five Bragg peaks of
scattering off densely-packed spheres on a square lattice (Fig. 2.6) and off densely-
packed spheres on a simple-cubic lattice (Fig. 2.15). Assume a constant charge
density within each sphere.
c. The structure factor of the molecule C60 [the buckminsterfullerene, or buckyball,
see Fig. 1.9(b)] can be approximated by a constant charge density, with a total
charge z, spread on the surface of a hollow sphere of radius r0 = 3.5Å. Find the
structure factor of each molecule.
d. These molecules create an FCC lattice, of lattice constant a = 14.1Å. Find the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 194

194 Introduction to Solid-State Physics

structure factor for scattering off the planes (111) and (200).

s.3.4.
Consider the orthorhombic lattice, whose edges are 5a, 4a, and 3a. Atoms with
a structure factor F1 are located at the vertices; at the center of the smallest face
there is another atom, with a structure factor F2 . The wave length of the X-rays is
λ = a.
a. Find the structure factor of the crystal.
b. Find the first six scattering angles for this lattice.
c. Which of these angles give constructive interference also when F1 = F2 ?
d. Which of those angles give constructive interference also when F1 = −F2 ?

s.3.5.
Consider a two-dimensional material of a square-lattice structure, with a lattice
constant a. X-rays of wave length λ = 1.5Å are scattered in two dimensions off a
two-dimensional powder of this material. The smallest scattering angle is 2θmin =
60◦ .
a. Find the lattice constant a.
b. Find the second scattering angle.
c. Assuming point-like atoms, find the ratio between the intensity at the second
angle and the intensity in the first one.
d. It is now assumed that each atom is described by a square of edge L with a
uniform charge distribution,
 −ez/L2 , |x| ≤ L/2 and |y| ≤ L/2 ,
n(x, y) =
0, otherwise .

Find an expression for the intensities’ ratio between the scattering at the second
angle and at the first angle, as a function of L/a.

s.3.6.
Figure 3.33 shows a diffraction pattern obtained by a theoretical calculation of X-
rays scattering off a Bravais lattice with a single atom in the unit cell. The radiation
source is copper (wave length 1.54Å).
a. Determine the lattice structure and the lattice constant.
b. Find the scattering angle and the height of the next peak in the sequence shown
in the figure (assuming that the experiment is carried out on a uniform powder, at
zero temperature.)

s.3.7.
The molecule carbon tetrachloride, CCl4 , is built of a tetrahedron with the chlorine
ions at its vertices and the carbon one at its center. The distance between each Cl
and the carbon is 1.78Å. It is assumed that FCl /FC = 3.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 195

Radiation scattering off crystals 195

Fig. 3.33

a. Find the structure factor of this molecule, assuming that the ions are point-like,
and that the molecule is placed on a cubic lattice of lattice constant a, such that the
carbons are located at the lattice sites and the chlorines on the opposite diagonals
of the cube.
b. Upon cooling, this material becomes a solid of an FCC structure, with a lattice
constant a = 8.34Å (at an even lower temperature it crosses over into a monoclinic
structure with 32 molecules in each unit cell). For X-rays of wave length λ =
1.5Å scattered off the FCC phase, what are the two smallest angles for which the
interference is constructive and what is the ratio of the intensities?

s.3.8.
Here is a description of an experiment that measured the intensity of X-rays scat-
tered off four crystalline materials. First it was found that when the materials were
in the form of single crystals, the symmetry of three of them was cubic, and the one
of the fourth crystal was hexagonal. When they were in the form of powders, the
values of the scattering angles (2θ) of the four first diffraction rings were:

A : 42.2◦ , 49.2◦ , 72.0◦ , 87.3◦ , B : 30.0◦ , 31.9◦ , 34.1◦ , 44.3◦ ,

C : 42.8◦ , 73.2◦ , 89.0◦ , 115.0◦ , D : 28.8◦ , 41.0◦ , 50.8◦ , 59.6◦ .

a. Identify the four crystals. Is the identification unique? and if it is not, what else
is needed to make the identification unique?
b. Assuming that the wave length is λ = 1.5Å, what are the lattice constants of
these crystals?

s.3.9.
Problem s.2.6 mentions phase transitions among various structures of sodium and
iron. Find the first four scattering angles obtained from scattering off powders of
each of the structures, for λ = 1.5Å.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 196

196 Introduction to Solid-State Physics

s.3.10.
a. Certain metallic alloys undergo phase transitions from a state (phase) called
“ordered” to another state, called “disordered”. An example is the alloy of copper
and zinc, CuZn: at high temperatures it is a BCC lattice, in which each lattice site
is randomly occupied by a copper or by a zinc atom. This is the disordered phase.
At low temperatures the atoms arrange themselves in the CsCl structure, such that
in each unit cell there is a copper atom at the origin and a zinc atom at the center
of the cube. This one is the ordered phase. Find the structure factors of each of the
phases.
b. Similarly, AuCu3 is transformed from an FCC structure in which the lattice
sites are occupied randomly by a gold atom (with probability 1/4) or by a copper
ion (with probability 3/4), to an ordered phase, in which the gold atoms are at the
corners of the cubic unit cell, and the copper atoms reside at the centers of the faces.
What are the structure factors of the two phases?

s.3.11.
Figure 2.29 displays the ferroelectric phase transition of barium titanate. At this
transition the cubic perovskite structure is transformed into a tetragonal one where
the lattice-constants ratio is c/a = 1.0045 (recall that a1 = a2 = a, and a3 = c).
The titanium and oxygen ions are displaced along the ẑ−axis, by uTi c and uO1 a
(for the oxygens in the base plane, which contains also the barium ions) and uO2 a
in the mid-plane (which contains the Ti ions). Compare the scattering angles that
give rise to Bragg peaks and the intensities in the two phases.

s.3.12.
a. Which values of the Miller indices give constructive interference in a face-
centered tetragonal lattice? Write down the expression for sin2 θ of this lattice.
b. Which lattice is obtained in the limit a3 = a1 ? Write down the seven lowest
values of the ratio sin2 θ/ sin2 θmin for this case. √
c. Write down the seven lowest values of the same ratio when a1 = a3 2. Deter-
mine the lattice structure deduced from the values you obtain. What is its lattice
constant? Explain the result by geometrical arguments.

s.3.13.
a. The distance between two neighboring planes of CuO2 and LaO in Fig. 2.21(b)
is xc. Experiment yields a1 = a2 = a ≈ 3.8Å and a3 = c ≈ 13Å. Calculate the
scattering amplitudes from this lattice and find the first five angles that yield Bragg
peaks for λ = 1.5Å.
b. As described in problem s.2.17, at low temperatures this lattice becomes a face-
centered orthorhombic. For small values of the angle φ defined there, what modifi-
cations are caused in the scattering angles?
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 197

Radiation scattering off crystals 197

s.3.14.
X-ray scattering off a powder of a single-atom FCC lattice yielded that the inten-
sity of the (800) Bragg peak decreases with the temperature as follows: at T [K] =
4, 200, 400, 600 the respective intensities were I(800) = 13.14, 8.06, 3.60, 1.57
(in arbitrary units). It is assumed that the atomic vibrations are described by the
Einstein model, that is, by a single frequency ω.
a. Exploit the Lindemann approximation to find the melting temperature of this
material.
b. Assuming that 4K is low enough to use the zero-temperature expressions, what
is the frequency of the atomic vibrations in the crystal? Is the assumption about the
temperature valid?
c. What is the intensity of the Bragg peak resulting from the (440) planes at T =
200K? Note:  = 6.581 × 10−16 eV sec, and kB = 8.617 × 10−5 eV/K.

3.16 Answers for the self-evaluation problems

Answer s.3.1.
In the trigonal lattice all edges and all angles are equal to each other. Hence
|b1 | = (2π/V )|a2 × a3 | = (2π/V )a2 sin α. The same answer is obtained for the
other two reciprocal-lattice vectors, so their lengths are all equal. As shown in
problem 3.7, the reciprocal of the reciprocal lattice is the original one, and therefore
a = |a1 | = (2π/Vrec )|b2 × b3 | = (2π/Vrec )b2 sin α . Multiplying these expressions
by one another, and using the identity V Vrec = (2π)3 from problem 3.6, results in
b = 2π/[a sin α sin α ]. The angle α is found from the scalar product of the lattice
vectors
b2 cos α = b3 · b1 = (2π/V )2 [a1 × a2 ] · [a2 × a3 ] = (2π/V )2 a4 (cos2 α − cos α) ,
[the right-hand side is derived from the identity
[u × v] · [w × y] = (u · w)(v · y) − (u · y)(v · w) (3.76)
check!] Upon inserting b = |b1 | one finds cos α = cos α(cos α − 1)/(1 − cos2 α) =
− cos α/(1 + cos α). Substituting cos α = 2 cos2 (α/2) − 1 in both sides yields
cos(α /2) = 1/[2 cos(α/2)] (assuming that both angles are acute).

Answer s.3.2.
Consider first the triclinic lattice; all other lattices are special limits of this case.
One begins with the result of Eq. (3.27),
(2π/d)2 = G2 = h2 b21 + k 2 b22 + 2 2
b3 + 2hkb1 · b2 + 2h b1 · b3 + 2k b2 · b3 .
In an arbitrary triclinic lattice (e.g., Fig. 2.13 or the last row in Fig. 2.28) |a1 ×
a2 |2 = a21 a22 sin2 γ and a1 · a2 = a1 a2 cos γ. Using the identity Eq. (3.76) one finds
[a1 × a2 ] · [a2 × a3 ] = a1 a22 a3 (cos γ cos α − cos β) ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 198

198 Introduction to Solid-State Physics

and similar expressions for the other permutations of the vectors. Problem s.2.11
yields that the volume of the unit cell is

V = a1 a2 a3 1 − cos2 α − cos2 β − cos2 γ + 2 cos α cos β cos γ .
Substituting all these expressions into the expression for G2 gives
1 1  2 2 2 2
= h a2 a3 sin α + k 2 a21 a23 sin2 β + 2 a21 a22 sin2 γ
d2 V2
+ 2hka1 a2 a23 (cos α cos β − cos γ) + 2k a21 a2 a3 (cos β cos γ − cos α)

+ 2h a1 a22 a3 (cos α cos γ − cos β) .
For the triclinic lattice, this expression is unchanged only when the signs of all
three indices are inverted together, (h k ) ↔ (h k ). The multiplicity is thus
p{hk } = 2. In the orthorhombic lattice, α = β = γ = 90◦ ; hence 1/d2 = (h2 /a21 ) +
(k 2 /a22 ) + ( 2 /a23 ), with p({hk }) = 8 when all indices are nonzero (the sign of each
one of them can be inverted separately), p({0k }) = p({h0 }) = p({hk0}) = 4,
p({h00}) = p({0k0}) = p({00 }) = 2. In the tetragonal lattice a1 = a2 , and as
a result there are more permutations among h and k, p({hk }) = 16, p({hh }) =
p({0k }) = p({hk0}) = 8, p({hh0}) = p({0k0}) = 4, p({00 }) = 2. In the hexagonal
lattice a1 = a2 = a3 , α = β = 90◦ and γ = 60◦ . It follows that (1/d2 ) =
(4/3)[(h2 −hk+k 2 )/a21 ]+( 2 /a23 ) (sometimes γ is chosen to be 120◦ , and then the sign
of the hk term is reversed). This lattice is invariant with respect to rotations in the
plane by 60◦ , and then p({00 }) = 2 (sign inversion), p({hh0}) = p({0h0}) = 6 [the
cases (±h, 0, 0), (0, ±h, 0), and ±(h, h, 0) are degenerate], p({hh }) = p({h0 }) = 12
(including the sign inversion of ), p({hk0}) = 4, p({hk }) = 8. For the monoclinic
lattice α = γ = 90◦ and then,
1 1  h2 k 2 sin2 β 2
2h cos β 
= 2 2 + 2 + 2 − .
d2 sin β a1 a2 a3 a1 a3
Here p({hk }) = 4, p({h0 }) = 2, p({0k0}) = 2. In the trigonal (or rhombohedral)
lattice, α = β = γ = 90◦ and a1 = a2 = a3 = a; then
1 (h2 + k 2 + 2
) sin2 α + 2(hk + h + kl) cos α(1 − cos α)
= .
d 2 a2 (1 − 3 cos2 α + 2 cos3 α)
Here, p({hk }) = 12 (6 permutations and sign inversion of all of them together)
p({hh0}) = p({h00}) = p({hh }) = p({0k }) = p({hk0}) = 6, p({00 }) = 2. Note
that in certain cases, for example, for the tetragonal lattice, the triplets (34 ) and
(50 ) are degenerate.

Answer s.3.3.
a. Introduce into the result of problem 3.18 the constant value of the density,
n(r) = n0 , and integrate over 0 ≤ r ≤ r0 . The result is

4πn0 r0 3z  
F (G) = drr sin(Gr) = sin(Gr 0 ) − Gr 0 cos(Gr 0 ) .
G 0 (Gr0 )3
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 199

Radiation scattering off crystals 199

In the G → 0 limit, a Taylor expansion gives sin(Gr0 ) − Gr0 cos(Gr0 ) ≈ (Gr0 )3 /3,
and hence F (G) → z, as could have been obtained also by substituting G = 0 in
the original integral.
b. Denote r0 = a/2. In a cubic lattice (Gr0 )2 = π 2 (h2 + k 2 + 2 ), and for the
first five Bragg peaks the sum h2 + k 2 + 2 attains integer values between 1 and 5.
The respective values of F (G) are F (G)/z = 0.304, 0.0075, −0.081, −0.076, and
−0.039, respectively. In the square lattice = 0 and therefore the lowest values of
the sum are 1, 2, 4, 5, and 8, and hence F (G)/z = 0.304, 0.0075, −0.076, −0.039,
and 0.0348. Interestingly enough, the intensities are not monotonic: the intensity of
the second peak is in particular low, because of the interference of the waves within
the sphere.
c. As the charge is spread over the surface, one uses n(r) = δ(r − r0 ). The
integration as obtained in problem 3.18 is only over the angles, which yields F (G) =
z sin(Gr0 )/(Gr0 ), where z is the total number of electrons on the sphere. As each
√ is z = 60 × 6 = 360.
carbon atom contains 6 electrons, the result
d. In the cubic lattice G(hk ) = √ (2π/a) h2 + k 2 + 2 . Using r0 = 3.5Å and a =
14.1Å, results in G(111)r0 = 2π 3r0 /a ≈ 0.86π, and G(200)r0 = 4πr0 /a ≈ 0.99π.
Hence, F (200) ≈ 0.01z and F (111) ≈ 0.16z. The low intensity for the (200) planes
results from the proximity of G(200)r0 to π; it has no special relation to any special
symmetry within the unit cell.

Answer s.3.4.
a. The unit cell contains an atom at the origin and another one at r2 = (a1 + a2 )/2;
the structure factor is thus (F1 + F2 exp[iπ(h + k)]).
b. In the solution of problem 3.15 it is found that sin2 θ = (λ2 /4)[(h2 /a21 )+(k 2 /a22 )+
( 2 /a23 )]; it follows that sin2 θ = (1/4)[h2 /9 + k 2 /16 + 2 /25]. The smallest angles
are obtained for (hk ) = (001), (010), (011), (100), (101), and (002), and their
values are θ = 5.7◦ , 7.2◦ , 9.2◦ , 9.6◦ , 11.2◦ , and 11.5◦ .
c. In this case F (G) = F1 (1 + exp[iπ(h + k)]), and therefore (h + k) must be even
(see also problem 3.25). This condition leaves just the first and the last angles.
d. In this case F (G) = F1 (1 − exp[iπ(h + k)]) and therefore (h + k) must be odd.
Consequently, the scattering angles are the second, third, fourth, and fifth.

Answer s.3.5.
The scattering angles are given by
sin2 θ = [λ2 /(4a2 )](h2 + k 2 ) .
a. The smallest scattering angle corresponds to (hk) = (01), (10), and hence
sin θmin = λ/(2a). As θmin = 30◦ , it follows that a = λ.
b. √ √ scattering angle ◦corresponds to (hk) = (11), and hence sin θ2 =
The second
λ/( 2a) = 1/ 2, yielding θ = 45 .
c. The scattering intensity is independent of G for point-like scatterers, and there-
fore the intensities are all identical.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 200

200 Introduction to Solid-State Physics

d. With the given density, the structure factor is

  L/2
ez L/2 sinux sin uy
F (G) = − 2 dxeiGx x dyeiGy y = −ez ,
L −L/2 −L/2 ux uy
with ux = Gx L/2 = πhL/a and similarly uy = Gy L/2 = πkL/a. The first-peak in-
tensity is thus I1 = |F (10)|2 = (ez)2 [sin(πL/a)/(πL/a)]2 , and the second-peak one
is I2 = |F (11)|2 = (ez)2 [sin(πL/a)/(πL/a)]4 . Thus, I2 /I1 = [sin(πL/a)/(πL/a)]2 ,
tending to 1 when L → 0, and to zero when the material covers the entire plane,
L → a.

Answer s.3.6.
The values of the scattering angles, as can be read from Fig. 3.33, are 2θ 
40◦ , 57◦ , 72◦ , 86◦ , 98◦ , 113◦ , 128◦ , . . . . To identify the type of the lattice one pre-
pares a list of the values sin2 θn / sin2 θmin = 1, 1.95, 2.95, 3.98, 4.87, 5.94, 6.91, . . ..
In a cubic lattice, these values are of the form (h2 + k 2 + 2 )n /(h2 + k 2 + 2 )min . The
ratios obtained from the measurements fit those of the BCC sequence, for which it
is known that the sum h + k + is even, and (h2 + k 2 + 2 )n /(h2 + k 2 + 2 )min =
1, 2, 3, 4, 5, 6, 7, 8, . . .. Note that the difference between the BCC and the SC
lattices is the seventh peak in the sequence, which does not exist for the SC lattice.
To find the lattice constant, one exploits Eq. (3.31), sin2 θin = [λ2 /(4a2 )] × 2; the
lattice constant is then a ≈ 3.2Å.
b. The next peak corresponds to the value 8 of the ratio sin2 θn / sin2 θmin , and
therefore 2θ8 = 151◦ . To estimate the relative height of this peak it is useful to
study the multiplicities of each peak. As they correspond to (hk ) = (110), (200),
(211), (220), (310), (222), (321), (400), the respective multiplicities are 12, 6, 24,
12, 24, 8, 48, 6. The figure shows that the height of the first peak is approximately
120, and thus, according to the multiplicities ratio, the eighth one should be twice
shorter, i.e., 60.

Answer s.3.7.
a. Locating the carbon ion at the origin, the four unit vectors
√ √ √ √
(1, 1, 1)/ 3 , (1, −1, −1)/ 3 , (−1, 1, −1)/ 3 , (−1, −1, 1)/ 3) ,
create a tetrahedron
√ around the origin. Denoting G(hk ) = (2π/a)(h, k, ) and
u = 2πr0 /(a 3) (where r0 is the distance between a chloride and the carbon), one
finds
F (G) = FC + FCl (eiu(h+k+) + eiu(h−k−) + eiu(−h+k−) + eiu(−h−k+) ) .
b. The four vectors listed above point along the cube’s diagonals, as assumed in the
problem. For an FCC lattice, the indices h, k, and are all even or are all odd, and
hence the smallest scattering angles are obtained from (hk ) = (111), (200), i.e.,
sin2 θ/[λ2 /(4a2 )] = 3, 4. The angles are 2θ ≈ 18◦ , 21◦ . The scattering intensity
in the first one is 8|F (111)|2 = 367|FC |2 , in the second is 6|F (200)|2 = 9.65|FCl |2 ;
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 201

Radiation scattering off crystals 201

the ratio is 38. (Note that the measured intensities include an arbitrary multi-
plicative factor and hence an individual intensity cannot be computed; the ratios
of intensities, however, are independent of that factor.)

Answer s.3.8.
a. For the cubic crystals, as seen from Eq. (3.31), sin2 θ = [λ2 /(4a2 )](h2 + k 2 + 2 ),
leading to sin2 θ/ sin2 θmin = (h2 + k 2 + 2 )/(h2 + k 2 + 2 )min . This ratio is 1, 1.34,
2.67, 3.67 for crystal A, 1, 1.13, 1.28. 2.12 for crystal B, 1, 2.67, 3.69, 5.34 for crystal
C, and 1, 1.98, 2.97, 3.99 for crystal D. The values that belong to A, C, and D are
very close to rational fractions, as required for cubic crystals. The values of crystal
A seem to obey sin2 θ/ sin2 θmin = (h2 +k 2 + 2 )/(h2 +k 2 + 2 )min = 1, 4/3, 8/3, 11/3,
fitting the sequence (h2 + k 2 + 2 ) = 3, 4, 8, 11. From the discussion following
problem 3.20 it follows that this is an FCC lattice (Fig. 3.20). Similarly, the
values for crystal C are close to those following from (h2 + k 2 + 2 ) = 3, 8, 11, 16,
which, from the solution of problem 3.25, correspond to diamond. For crystal D,
the values fit (h2 + k 2 + 2 ) = 1, 2, 3, 4, that agree with a SC lattice. However,
they also fit (h2 + k 2 + 2 ) = 2, 4, 6, 8, that agree with a BCC lattice (see the
discussion after problem 3.20 and Fig. 3.20). The data points are not sufficient
to distinguish between these two possibilities. Nonetheless, for the SC lattice the
value (h2 + k 2 + 2 ) = 7 is forbidden, while the BCC lattice allows for 14, e.g.,
for (hk ) = (123). One could have then identify the crystal unambiguously had
one continued to the seventh scattering angle. As three of the crystals are cubic,
one may deduce that crystal B is hexagonal (according to the conditions in the
problem). It remains then to examine whether it is a simple hexagonal lattice, or
may be it has a base. The solution to problem 3.26 shows that for an arbitrary
hexagonal lattice sin2 θ = [4λ2 /(3a21 )][h2 − hk + k 2 + 3 2 a21 /(4a23 )]; consequently
sin2 θ h2 − hk + k 2 + 3 2 a21 /(4a23 )
= 2 .
2
sin θmin [h − hk + k 2 + 3 2 a21 /(4a23 )]min
A simple hexagonal lattice allows for all values of (hk ). Hence, the smallest scat-
tering angle corresponds to (hk ) = (001), where h2 − hk + k 2 + 3 2 a21 /(4a23 ) =
3a21 /(4a23 ). The next angle is for (hk ) = (100), yielding sin2 θ/ sin2 θmin =
4a23 /(3a21 ). A comparison of this expression with the experimental value 1.13 implies
that a3 /a1 = 0.92. However, the next two experimental values cannot be fitted to
any of the triplets (hk ). On the other hand, the HCP structure forbids the triplet
(hk ) = (001), and hence [h2 − hk + k 2 + 3 2 a21 /(4a23 )]min = 1. The next  angles
all agree with those expected for the ideal HCP lattice, with a3 /a1 = 8/3, as
detailed in problem 3.26, sin2 θ/ sin2 θmin = 9/8, 41/32 17/8. This result is further
confirmed by the difference between the second and the fourth ratios, which is about
1.
b. For the cubic lattices sin2 θmin = [λ2 /(4a2 )](h2 + k 2 + )min , and the expression
in the brackets equals 3 for the A and C crystals, and 1 or 2 for the D one, when
it is considered as a SC lattice or as a BCC lattice, respectively. In both cases
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 202

202 Introduction to Solid-State Physics

√
a = λ 3/(2 sin θmin ) ≈ 3.6Å. For crystal D, a(SC)≈ 3Å or a(BCC)≈
 4.3Å. For
crystal B sin2 θmin = λ2 /(3a21 ); hence a1 = 3.35Å and a3 = a1 8/3 ≈ 5.46Å.

Answer s.3.9.
The lattice constants of sodium are a(BCC)=4.23Å and a(HCP)≈ 3.77Å. The
BCC lattice obeys
sin2 θ = [λ(2a2 )]{2, 4, 6, 8} ,
and hence 2θ = 30◦ , 43◦ , 54◦ , 62◦ . Problem 3.26 cites
sin2 θ = [4λ2 /(3a2 )]{1, 9/8, 41/32, 17/8} ,
for the dense HCP lattice and therefore 2θ = 27◦ , 29◦ , 31◦√
, 40◦ . The
√ lattice con-
stant of iron is a(BCC)=2.87Å and therefore a(FCC)=(2 2r/(4r/ 3) = 3.51Å.
For the BCC lattice
sin2 θ = [λ2 /(4a2 )]{2, 4, 6, 8} ,
2θ = 45◦ , 65◦ , 82◦ , 99◦ . For the FCC lattice,
sin2 θ = [λ/(4a2 )]{3, 4, 8, 11} ,
and hence 2θ = 45◦ , 52◦ , 72◦ , 93◦ . In both cases the fourth angle does not produce
a ring on the screen as it is larger than 90◦ .

Answer s.3.10.
a. In the disordered phase, the structure factor at each site of the lattice is an av-
erage of the atomic structure factors, i.e., Fav = (FCu + FZn )/2, as follows directly
from Eq. (3.51) for the scattering amplitude at a reciprocal-lattice site. The struc-
ture factor measured in experiment is then identical to that of a usual BCC lattice,
F (hk ) = Fav [1 + (−1)h+k+ ]. There are no Bragg peaks for odd values of h + k + .
On the other hand, the ordered structure is represented by a simple cubic lattice
with a base, like CsCl, and the structure factor is F (hk ) = FCu + (−1)h+k+ FZn .
The scattering amplitude for even values of h + k + is identical to that obtained for
the disordered phase; the one of odd values is nonzero as well, F (hk ) = FCu − FZn .
b. In the disordered phase, F (hk ) = Fav [1 + (−1)h+k + (−1)h+ + (−1)k+ ],
with Fav = (FAu + 3FCu )/4. Bragg peaks appear only when the three indices are
all odd or are all even. In contrast, in the ordered phase the structure factor is
F (hk ) = FAu + [(−1)h+k + (−1)h+ + (−1)k+ ]FCu . Again, there are more peaks:
indices that are not all odd or all even contribute as well, as in the case of table
salt.

Answer s.3.11.
The first part of the problem is similar to problem 3.15: the scattering angles
for a tetragonal lattice obey sin2 θ/ sin2 θ(100) = h2 + k 2 + 2 (a1 /a3 )2 , and there-
fore all Bragg peaks of the cubic lattice that contain a contribution from the in-
dex are split or displaced relative to their locations in the cubic lattice. For
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 203

Radiation scattering off crystals 203

instance, the scattering angles for (hk ) = (100), (010), (110), (200), (020),
are not changed. Constructive interference also appears for the angles obeying
sin2 θ/ sin2 θ(100) = (a1 /a3 )2 , 1 + (a1 /a3 )2 , 2 + (a1 /a3 )2 , 4(a1 /a3 )2 , due to the
splitting of the Bragg peaks corresponding to (hk ) = (001), (101), (111), (002).
The scattering amplitude in the ferroelectric phases is given by

F (hk ) = FBa + (−1)h+k+ e2iπuTi  FTi

+ [(−1)h+k e2iπuO1  + {(−1)h + (−1)k }(−1) e2iπuO2  ]FO .

It follows that the intensities at the points with = 0 are not changed, while the
other amplitudes acquire a small imaginary part. For instance, the peak at (100)
is divided: F (100) = F (010) = FBa − FTi − FO , but F (001) = FBa − e2iπuTi FTi +
[e2πiuO1 − 2e2πiuO2 ]FO . The changes in the scattering angles and in the intensities
become more pronounced as the ferroelectric deformation is enhanced. They also
increase with .

Answer s.3.12.
a. The reciprocal-lattice vectors are G = (2π/a1 )(hx̂ + ŷ) + (2π/a3 )kẑ, and the
atoms in the base are at the origin and at the points (a1 + a2 )/2, (a1 + a3 )/2, and
(a3 + a2 )/2. The structure factor is thence like in the FCC lattice, {1 + exp[iπ(h +
k)] + exp[iπ(h + )] + exp[iπ(k + )]}, and the Miller indices should be all even or
all odd. From the solution of problem 3.15, sin2 θ = (λ/[4a21 ])[h2 + k 2 + ( a1 /a3 )2 ].
b. For a1 = a3 the lattice becomes an FCC lattice, and the smallest angles
are obtained from h2 + k 2 + 2 = 3, 4, 8, 11, 12, 16, 19. It follows that
sin2 θmin = 3λ2 /(4a1 )2 , and the smallest angles are given by sin2 θ/ sin2 θmin =
1, 4/3, 8/3, 11/3, 4, 16/3, 19/3.
c. In this case sin2 θ = [λ/(4a21 )][h2 + k 2 + 2 2 ], leading to h2 + k 2 + 2 2 =
4, 8, 12, 16, 20, 24, 28. Substituting sin2 θmin = λ2 /a21 yields sin2 θ/ sin2 θmin =
1, 2, 3, 4, 5, 6, 7. Figure 3.20 shows that this is a BCC lattice (the ratio 7 does
not appear in a SC lattice). In a BCC lattice with a lattice √ constant a the result is
sin2 θmin = λ2 /(2a2 ) (check!). It thus follows that a1 = a 2. Figure 3.34 displays
the projection on the base plane of the four unit cells of a BCC lattice. The lattice
constant is a, and the atoms at the center of each cube are at a/2 above the cen-
ter of the
√ square. Consider the square marked by dashed lines: the edge length is
a1 = a 2. Using this square as a base for a prism of height a shows that the prism
is the unit cell of the face-centered tetragonal lattice. The atoms that previously
resided at the centers of the cells of the BCC lattice are now at the centers of the
side faces of the tetragonal lattice.

Answer s.3.13.
a. The lattice is a body-centered tetragonal. As seen in the solution of problem
2.14, the unit cell contains a base with two atoms represented, e.g., by the locations
of the copper ions: the one at the origin and the other at the center of the cell,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 204

204 Introduction to Solid-State Physics

Fig. 3.34

r2 = (x̂ + ŷ)a/2 + ẑc/2. Since the reciprocal-lattice vectors are given by G =

(2π/a)(hx̂+kŷ)+(2π/c) ẑ, the structure factor is F (hk ) = F0 (hk )[1+(−1)h+k+ ],
where F0 (hk ) contains the information on the other ions in the base. Bragg peaks
appear when (h + k + ) is even, like in a BCC lattice. As described in the solution
of problem 2.14, the base connected with each copper ion comprises a copper ion
at the origin, two lanthanum ones at (0, 0, (1/2 − x)c) and (a/2, a/2, xc), and four
oxygens at (a/2, a/2, (1/2 − x)c), (0, 0, xc), (0, a/2, 0) and a/2, 0, 0). It follows that
F0 (hk ) = FCu
+ FLa [eiπ(1−2x) + eiπ(h+k+2x) ] + FO [eiπh + eiπk + e2iπ + eiπ[h+k+(1−2x)] ]
= FCu + 2FLa cos[π(1 − 2x) ] + FO [(−1)h + (−1)k + 2 cos(2πx )] ,
where the second step uses the fact that h + k + is even. As seen in the solution
of problem 3.15, the Bragg peaks appear for the scattering angles
λ 2  h2 k2 2
sin2 θ = 2 + 2 + 2 .
4 a1 a2 a3
As h + k + is even, and given that (c/a)2 = (13/3.8)2 ≈ 11.7, the lowest values
of the indices are (002). The minimal angle is such that sin θmin = λ/c = .115, i.e.,
2θmin = 13.2◦ . The next angles are for (hk ) = (101), (004), (103), (110); thus,
2θ = 13.2◦ , 23.7◦ , 26.6◦ , 30.3◦ , 32.3◦ .
b. At low temperatures the material assumes the face-centered  orthorhombic struc- √
ture; the unit cell is thus larger, with lattice constants a1 = a 2 cos(2φ), a2 = a 2,
a3 = c, see problem s.2.17. This unit cell contains four copper ions, at the origin and
at r2 = (a1 /2)x̂ + (a3 /2)ẑ , r3 = (a1 /2)x̂ + (a2 /2)ŷ , and r4 = (a2 /2)ŷ + (a3 /2)ẑ .
The reciprocal-lattice vectors are
2π 2π 2π
G =  h x̂ +  k  ŷ +   ẑ ,
a1 a2 a3
where x̂ and ŷ are unit vectors, oriented along the dashed lines in the base of the
   
cell in Fig. 2.60. It follows that F (h k   ) = F0 (h k   )[1 + (−1)h +k + (−1)h + +
 
(−1)k + ] (as in an FCC lattice); Bragg peaks appear when the indices are all odd
or they are all even. The scattering angles are given by
λ2  h2 k 2 2 
sin2 θ = + + ,
4 a2 1 a2
2 a23
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 205

Radiation scattering off crystals 205

and the smallest one is for (h k   ) = (002), i.e., sin θmin = λ/c = .115. By assump-
tion, a3 = a3 = c, and therefore the minimal angle is identical to that obtained at
high temperatures. Since
sin2 θ 1  2 k 2  c2
= h + + 2 2 ,
sin2 θmin 8 cos(2φ) a2
the next angles correspond to (h k   ) = (111), (004), (113), (200), (020). Peaks
for which k  = 0 correspond to the same angles as at high temperatures, while
all the others are displaced or even split. Figure 3.35 shows the first six angles as
functions of φ. In particular, the last three triplets, which give rise to a peak at
the same angle in the tetragonal structure, yield two peaks at low temperatures, at
two different angles. The second and fourth angles increase with the orthorhombic
deformation, and the fifth angle is split into two angles.

Fig. 3.35

The figure shows that the angles approach their √ values in the tetragonal
√ phase
when φ tends to zero. In this limit x̂ = (x̂ − ŷ)/ 2 and ŷ = (x̂ + ŷ)/ 2, and thus
G = (π/a)[(h + k  )x̂ + (k  − h )ŷ] + (2π/c)  ẑ. These vectors coincide with those of
the tetragonal phase, once h = (h + k  )/2, k = (k  − h )/2 and =  are used. Note
that then h + k + =  + k  . As the new indices h , k  , and  are all even or are all
odd, this sum is always even, as it should be in the tetragonal phase. Substituting
this in the expression for the scattering angles shows that indeed the two answers
derived above are identical. In general, the structure factor of the base, F0 (hk ), is
modified due to the various displacements of the ions in the unit cell.

Answer s.3.14.
a. The intensity of the Bragg peaks is given by I(hk ) = I0 exp[−2W ]. According
to Eq. (3.59),
 ω
2W = coth( )G2 .
2M ω 2kB T
At high enough temperatures W = (kB T /[M ω 2 ])G2 . In the present case G2 =
(2π/a)2 (h2 + k 2 + 2 ) = 256π 2 /a2 , and therefore the high-temperature value is
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 206

206 Introduction to Solid-State Physics

2W = 256π 2 kB T /(M ω 2 a2 ) = 64π 2 T /(75TM ), where the Lindemann criterion with

cL = 0.1, and kB TM = c2L M ω 2 a2 /3 = M ω 2 a2 /300 √
have been used. (For an FCC √
lattice, the distance between nearest neighbors is a/ 2, and therefore cL → cL / 2,
which halves the melting temperature, but does not affect the results below.) A
plot of ln(I/I0 ) vs. the temperature T (with the given numbers) indeed reveals a
linear dependence for the last three points. The slope can be deduced from the last
two of them, s = 64π 2 /(75TM ) = ln(3.60/1.57)/200[K], and hence TM ≈ 2030K.
From the relation ln I = ln I0 − sT , and from the intensity of the last point follows
ln I0 = ln 1.57 + s600 ≈ 2.94, yielding I0 ≈ 18.9.
b. At sufficiently low temperatures
 ω ω
2W = G2 = (aG)2 = 256π 2 .
2M ω 2M ω 2 a2 600kB TM
Using the value of the intensity at 4 K gives
ω
2W = 256π 2 = ln(I0 /I) = ln(18.9/13.14) ,
600kB TM
and hence ω ≈ 0.086kB TM ≈ 0.015eV. It follows that ω ≈ 2.28 × 1013 sec−1 .
Therefore 4 K kB T /(ω) = T /(0.086TM ) = 0.023  1, justifying the use of the
low-temperature expression.
c. Inserting the information listed above into Eq. (3.59) gives
ω  ω T   T 
2W = coth M
G2 ≈ 0.00566 coth 0.043 M (h2 + k 2 + 2 ) .
600kB TM 2kB TM T T
For (440) and 200 K we find that 2W ≈ 0.44, and therefore I(440) ≈
18.9 exp[−0.44] ≈ 12.2.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 207

Appendix B

Fourier series

Periodic functions in one dimension. This Appendix presents periodic func-

tions as sums of trigonometric functions. In one dimension, a periodic function
f (x) is defined by the relation f (x) = f (x + a). Such a periodicity pertains, e.g.,
to the density of the material within each unit cell of a periodic lattice whose lat-
tice constant is a, or to the potential exerted on electrons moving in a periodic
lattice of ions. Examples for these functions are the trigonometric ones, sin(kn x)
or cos(kn x), where kn = 2πn/a and {n} are arbitrary integers, which describe in-
stantaneous shapes of waves, with wave numbers kn . The Fourier theorem
states that the infinite ensemble of the trigonometric functions forms a basis for
the description of any periodic function of this type. In other words, any function
f (x) that obeys f (x + a) = f (x) can be expressed by the infinite series
∞
a0 
f (x) = + [an cos(kn x) + bn sin(kn x)] , (B.1)
2 n=1

with constant coefficients that do not depend on x. This is a Fourier series.

It is expedient to re-write the Fourier series in terms of exponential functions,
by using cos(kn x) = (exp[ikn x] + exp[−ikn x])/2 and sin(kn x) = (exp[ikn x] −
exp[−ikn x])/(2i), and then
∞

f (x) = fn eikn x , (B.2)
n=−∞

with f0 = a0 /2, fn = (an − ibn )/2, f−|n| = (f|n| )∗ . To calculate the coefficients,
one exploits the orthogonality of the exponential functions,
 a/2
ei(km −kn )a/2 − e−i(km −kn )a/2
dx(eikn x )∗ eikm x = ≡ aδmn , (B.3)
−a/2 i(km − kn )

where δmn is the Kronecker delta function. For n = m the integrand in Eq. (B.3)
equals 1, leading to the result a. (The limit in the second equality should be derived
carefully.) For n = m the denominator is finite while the numerator vanishes [since
(km − kn )a = 2π(m − n)]. Multiplying both sides of Eq. (B.2) by exp[−ikm x], then

207
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 208

208 Introduction to Solid-State Physics

integrating from −a/2 to a/2, and using Eq. (B.3), one obtains the coefficients of
the series

 1 a/2
fm = dxf (x)e−ikm x . (B.4)
a −a/2
√
Since the functions Φn (x) = exp[ikn x]/ a can serve as basis vectors in a linear
space (in analogy with the unit vectors along the axes in a cartesian coordinate
space), Eq. (B.2) implies that any other vector in this space can be described by
a linear combination of the basis vectors, with complex coefficients. To complete
the description of this space, one needs to define a scalar product, i.e., f |g ≡
a/2
−a/2
dxf ∗ (x)g(x). The orthonormality of the basis vectors, Eq. (B.3), is
expressed by the relation Φn |Φm  = δnm .

Problem B.1
a. Prove that if f (x) = f ∗ (x) is a real function, then (fn )∗ = f−n , and if f (x) =
f (−x) is an even function then fn = f−n .
b. Use Eq. (B.4) to compute the coefficients in Eq. (B.1). Derive these coefficients
in the cases mentioned in part a.
c. Derive the coefficients for f (x) = A sin3 (2πx/a).
d. Find the coefficients in Eq. (B.2) for the function
 −1, na < x < (n + 1/2)a ,
f (x) =
1, (n − 1/2)a < x < na ,
where n is an arbitrary integer. Use the result to sum over the first five terms of
the Fourier series, and compare the outcome with the original function.

Dirac delta function. Inserting the expression for the Fourier coefficients, Eq.
(B.4), into the original expansion in Eq. (B.2) yields
∞ 
1 a/2
f (x) = dye−ikn y f (y) eikn x , (B.5)
n=−∞
a −a/2

where the expression in the square brackets is fn (the integration variable is denoted
y to avoid confusion). Interchanging the order of the summation and the integration,
Eq. (B.5) becomes
 a/2 ∞
1  ikn (x−y)
f (x) = dy e f (y) . (B.6)
−a/2 a n=−∞

Comparison with the definition of the Dirac delta function, Eq. (3.12), shows that
the expression in the square brackets is that function itself,
∞
1  ikn (x−y)
δ(x − y) = e . (B.7)
a n=−∞
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 209

Fourier series 209

This identification is valid only for x within the integration range, i.e., −a/2 < x <
a/2. As f (x) = f (x + a), the same result is obtained upon integrating over any
other segment shifted by na, where n is an arbitrary integer. Hence,
∞
 ∞

eikn (x−y) = a δ(x − y + ma) . (B.8)
n=−∞ m=−∞

The sum in the right-hand side comprises narrow peaks at each lattice site; it is
called at times the “Dirac comb”.
Fourier transform. The formalism outlined above can be extended to treat
arbitrary functions that are not necessarily periodic. In other words, it is valid
also when the periodicity (in the present context, the periodicity of the lattice)
is infinitely large, a → ∞. In that case the difference between successive wave
numbers approaches zero, Δk = kn − kn−1 = 2π/a → 0. The Fourier transform of
an arbitrary function f (x) is then defined by
 ∞

f(k) = lim [afm ] = f (x)e−ikx . (B.9)
a→∞ −∞

In this limit Eq. (B.4) takes the form

∞ ∞ 
1 2π   ikn x 1  1 ∞

f (x) = af e = Δk(afn )eikn x → dk f(k)eikx .
2π a n=−∞ n 2π n=−∞
2π −∞

(B.10)

In the last step the sum is converted into an integral, since when a → ∞, the
integration increment approaches zero. Equation (B.10) is the inverse Fourier
transform.

Problem B.2

a. Show that the area under the function f (x) is f(0). What is the area under the

function f(k)?
 ∞
b. Use f(k) to compute the integral Im = −∞ dxxm f (x).
c. Find the Fourier transform of f (x) = exp[−|x|/w]/(2w).
d. Express the Fourier transforms of the functions h1 (x) = f (x−x0 ), h2 (x) = f (ax)

(a > 0), and h3 (x) = exp[iqx]f (x), in terms of the Fourier transform f(k).
e. Prove that the Dirac delta function is the inverse Fourier transform of the con-

stant δ(x) ≡ 1.
f. The generalization of the scalar product for functions defined on the infinite axis

dk f∗ (k)
∞ ∞
is f |g ≡ dxf ∗ (x)g(x). Show that f |g = 1
−∞ 2π g(k).
−∞

Convolution. In certain cases, the Fourier transform of a function is a product

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 210

210 Introduction to Solid-State Physics

 
of two transforms, i.e., F (k) = f(k)g(k). Then
 ∞  ∞
1  1 
F (x) = dk F (k)eikx = dk f(k)
g(k)eikx
2π −∞ 2π −∞
 ∞  ∞
1
= dk dyf (y)e−iky  g(k)eikx
2π −∞ −∞
 ∞  ∞  ∞
1
= dyf (y) dkg(k)eik(x−y) = dyf (y)g(x − y) . (B.11)
−∞ 2π −∞ −∞

The values of the function F (x) are therefore determined by those of f (x) and g(x)
at all points on the axis. The expression on the right hand-side of Eq. (B.11) is
called the convolution of these two functions and is usually denoted by
 ∞
f (x) ⊗ g(x) = dyf (y)g(x − y) . (B.12)
−∞
Periodic functions in three dimensions. Consider the three-dimensional
vector r = (x, y, z), expressed in terms of its cartesian coordinates; a function
periodic along all three axes obeys
f (x + a1 , y, z) = f (x, y + a2 , z) = f (x, y, z + a3 ) = f (x, y, z) . (B.13)
The Fourier theorem, Eqs. (B.2) and (B.4), for the x−dependence gives
∞

fnx (y, z)eiknx x ,
x
f (x, y, z) = (B.14)
nx =−∞

where knxx = 2πnx /a1 and fnx (y, z) = (1/a1 )

a1 /2
−a1 /2
dxf (x, y, z) exp[−iknxx x] .
Using again the theorem for the y−dependence of fnx (y, z), and then again for the
z−dependence, results in
∞ ∞ ∞
f (x, y, z) = f(nx , ny , nz )eik(nx ,ny ,nz )·r , (B.15)
nx =−∞ ny =−∞ nz =−∞

where
k(nx , ny , nz ) = 2π(nx /a1 , ny /a2 , nz /a3 ) , (B.16)
and
 a1 /2  a2 /2  a3 /2
1
f(nx , ny , nz ) = dx dy dzf (x, y, z)e−ik(nx ,ny ,nz )·r .
a1 a2 a3 −a1 /2 −a2 /2 −a3 /2
(B.17)
Note that the vectors k(nx , ny , nz ) in Eq. (B.16) are vectors in the reciprocal space
of the orthorhombic lattice whose lattice constants are a1 , a2 , and a3 .
The three-dimensional Fourier transforms are obtained upon taking the
limit am → ∞,


f (r) = d3 k f(k)eik·r , (B.18)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 211

Fourier series 211

with

 1
f(k) = d3 rf (r)e−ik·r , (B.19)
(2π)3
where in both expressions the integrals are carried out over the entire space (the
reciprocal and the direct one, respectively).
Periodic functions on an arbitrary Bravais lattice. The Bravais lattice
can be defined by the property f (r + a1 ) = f (r + a2 ) = f (r + a3 ) = f (r). Presenting
the vector r in terms of the unit vectors âm ≡ am /|am |, as r = y1 â1 + y2 â2 + y3 â3 ,
one obtains
f (y1 + a1 , y2 , y3 ) = f (y1 , y2 + a2 , y3 ) = f (y1 , y2 , y3 + a3 ) = f (y1 , y2 , y3 ) . (B.20)
This equation resembles Eq. (B.13), and therefore one may exploit the one-
dimensional Fourier theorem for the three coordinates y1 , y2 , and y3 . For instance,
∞

f (y1 , y2 , y3 ) = fn1 (y2 , y3 )ei(2π/a1 )n1 y1 . (B.21)
n1 =−∞

The final result is

∞
 ∞
 ∞
 n y n y n y
f (y1 , y2 , y3 ) = f(n1 , n2 , n3 ) exp[i2π( 1 1 + 2 2 + 3 3 )] ,
n1 =−∞ n2 =−∞ n3 =−∞
a1 a2 a3
(B.22)
with
 a1  a2  a3
n 1 y1 n 2 y2 n3 y3
2 dy1 2 dy2 2 dy3 −i2π( a
f(n1 , n2 , n3 ) =
+ a + a )
f (y1 , y2 , y3 )e 1 2 3 .
−
a1
2
a1 −
a2
2
a2 −
a3
2
a3
(B.23)
The integration in Eq. (B.23) is carried out over a single unit cell. Using Eqs.
(3.20) yields bm ·r = 2πym /am , where {bm } are the vectors of the reciprocal lattice.
Hence, the expression in the exponent is 2π(n1 y1 /a1 + n2 y2 /a2 + n3 y3 /a3 ) = G · r,
where G = n1 b1 + n2 b2 + n3 b3 is an arbitrary vector in reciprocal lattice. The
function ΦG (r) = exp[iG·r] is indeed periodic over the lattice, since from Eq. (3.17)
ΦG (r + R) = exp[iG · (r + R)] = exp[iG · r] = ΦG (r). Thus, Eq. (B.22) implies
that any periodic function on the lattice can be presented as a series of
these functions, with all reciprocal-lattice vectors,

f (r) = f(n , n , n )eiG·r . 1 2(B.24)
3
n1 ,n2 ,n3

It remains to consider the coefficients, Eq. (B.23). In order to carry out the
integration over the cartesian coordinates x, y, and z (in place of the coordinates
defined by r = y1 â1 + y2 â2 + y3 â3 ) one uses the identity dy1 dy2 dy3 = Jdxdydz,
where the Jacobian J is
   
 ∂y1 /∂x ∂y1 /∂y ∂y1 /∂z   
  a1 a2 a3  b1x b1y b1z  a1 a2 a3 a a a
 
J =  ∂y2 /∂x ∂y2 /∂y ∂y2 /∂z  =  b2x b2y b2z  = Vrec = 1 2 3 .
3
(2π)  (2π) 3 V
 ∂y /∂x ∂y /∂y ∂y /∂z  b3x b3y b3z 
3 3 3
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 212

212 Introduction to Solid-State Physics

The first step uses bm · r = 2πym /am . The second step uses the expression for the
volume of a unit cell in reciprocal lattice, Vrec = b1 · [b2 × b3 ], and the third makes
use of problem 3.6, Vrec V = (2π)3 . Hence,

1
f(n1 , n2 , n3 ) = d3 rf (r)e−iG·r , (B.25)
V

where the integration is carried out over the unit cell.

The relation between Fourier series and the reciprocal lattice. As
mentioned, the periodic boundary conditions on the sample’s edges are identical to
its extension over the entire space. The function f (r) can then be written as in Eq.
(B.18), with the Fourier transform given in Eq. (B.19). Exploiting the periodicity,
Eq. (3.33), enables one to displace all points in space by a lattice vector, and thus
Eq. (B.19) yields

 1 
f (k) = d3 rf (r + R)e−ik·(r+R) = f(k)e−ik·R .
(2π)3


It follows that f(k) = 0 only when exp[ik · R] = 1. From Eq. (3.9) one finds that
this condition is fulfilled only when k equals one of the reciprocal-lattice vectors,
i.e., when Bragg’s condition is obeyed. Hence, f (r) is a linear combination of the
periodic functions {exp[iG · r]}, that is, a Fourier series, [see Eqs. (B.24) or (3.34)].
Two alternative calculations of the coefficients are given in the paragraphs leading
to Eqs. (3.37) or (B.25).
The delta function and Dirac’s comb in three dimensions. As in Eq.
(B.5), inserting Eq. (B.25) in Eq. (B.24) yields
 1  

1  iG·(r−r )
f (r) = d3 r f (r )e−iG·r eiG·r = d3 r  e f (r ) .
V V V V
G G

Hence (1/V ) G exp[iG · r] = δ(r), where r is within the unit cell that contains
the origin. Since the function f (r) is periodic, a similar integration over any other
unit cell yields the function at the origin of that cell; consequently

1  iG·r 
e = δ(r − R) . (B.26)
V
G R

This equation generalizes Eq. (B.8), and the right hand-side extends the Dirac comb
to three dimensions. There is a complete symmetry between the “usual” (direct)
space and the reciprocal one; one therefore may interchange G → R, r → q, and
V → Vrec = (2π)3 /V (problem 3.6). It follows that

  (2π)3 
eiq·R = Vrec δ(q − G) = δ(q − G) . (B.27)
V
R G G
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 213

Fourier series 213

B.1 Answers

Answer B.1.
a. Using Eq. (B.4), one finds that when f (x) = f ∗ (x) then
 a/2  a/2
1 1
(fn )∗ = dxf ∗ (x) exp[−ikn x] = dxf (x) exp[ik−n x] = f−n .
a −a/2 a −a/2

Similarly, when f (x) = f (−x) then

 a/2  a/2
1 1
fn = d(−x)f (−x) exp[−ikn x] = dxf (x) exp[ik−n x] = f−n ,
a −a/2 a −a/2

where in the first step the integration variable x is replaced by −x, and the inte-
gration bounds are interchanged in the second, using the fact that the function is
even.
b. From the expressions listed after Eq. (B.2). it follows that a0 = 2f0 = 2/a. For
n > 0 one finds

2 a/2
an = fn + f−n = dx cos(kn x)f (x) ,
a −a/2

  2 a/2
bn = i(fn − f−n ) = dx sin(kn x)f (x) .
a −a/2

For a real function one finds an = 2Re[fn ] and bn = −2Im[fn ]. When the function
is even then bn = 0, and there are only the even functions, i.e., the cosines, in the
Fourier expansion.
c. The identity sin(3α) = 3 sin(α) − 4 sin3 (α) implies that f (x) = A[3 sin(2πx/a) −
sin(6πx/a)]4. Thus, this function represents a special case: all coefficients are zero
save two, b1 = 3A/4 and b3 = −A/4. Alternatively, the coefficients are given by
the integral
 
2A a/2 A a/2
bn = dx sin3 (k1 x) sin(kn x) = dx[3 sin(k1 x) − sin(k3 x)] sin(kn x) .
a −a/2 2a −a/2
a/2
Since −a/2 dx sin(km x) sin(kn x) = (a/2)δnm by Eq. (B.3), the integral is identical
to the aforementioned result.
d. By Eq. (B.4),

1 (1 − e−iπn )2
fn = − (2e−ikn a/2 − 1 − e−ikn a + e−ikn a/2 ) = ,
ikn a iπn

and therefore f|n| = −f−|n| = 2i/(πn) for an odd n, and 0 else. Substituting in Eq.
∞
(B.2) yields f (x) = −4 n=1 sin[π(2n − 1)x]/[(2n − 1)π]. Figure B.1 compares the
original function with the sum of the first five terms.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 214

214 Introduction to Solid-State Physics

Fig. B.1

Answer B.2.
 ∞
a. From Eq. (B.9), f(0) = dxf (x), and the right-hand side is the desired area.
−∞
∞ 
Similarly, Eq. (B.10) yields 2πf (0) = −∞ dk f(k).
b. Differentiating Eq. (B.9) yields  ∞
∂m 
im m f(k) = dxf (x)xm e−ikx ,
∂k −∞

and hence Im = im ∂ m f(k)/∂k m |k=0 .
c. Equation (B.9) gives
 0  ∞
 1
f(k) = dxe(x/w)−ikx + e−(x/w)−ikx
2w −∞ 0
1 1 −1 1
= + = .
2w (1/w) − ik −(1/w) − ik 1 + w2 k2
The function on the right hand-side is called “Lorentzian”.
d. From Eq. (B.9),
 ∞  ∞
 
h1 (k) = dxf (x − x0 )e−ikx = dyf (y)e−ik(y+x0 ) = e−ikx0 f(k) ,
−∞ −∞
 ∞ 
 1 ∞ 
h2 (k) = dxf (ax)e−ikx = dyf (y)e−i(k/a)y = f(k/a)/a ,
−∞ a −∞
 ∞  ∞
 
h3 (k) = dxeiqx f (x)e−ikx = dxf (x)e−k(k−q)x = f(k − q) .
−∞ −∞
e. One begins with Eq. (B.7). In the a → ∞ limit it yields
∞  ∞
1 2π  ikn x 1
δ(x) = e → dkeikx ,
2π a n=−∞ 2π −∞

which is indeed the inverse Fourier transform of δ(k) = 1, as can be seen by com-
paring with Eq. (B.10).
f. One begins with the right hand-side and inserts there the expressions for the
 ∞
Fourier transforms of the
∞
two functions:
∞  ∞

 
∗
dk f (k)g(k) = dk dxf ∗ (x)eikx dyg(y)e−iky
−∞ −∞ −∞ −∞
 ∞  ∞  ∞  ∞  ∞
= dxf ∗ (x) dyg(y) dkeik(x−y) = dxf ∗ (x) dyg(y) 2πδ(x−y)
−∞ −∞ −∞ −∞ −∞
 ∞
= 2π dxf ∗ (x)g(x) = 2πf (x)|g(x) .
−∞
The intermediate step makes use of the result in part e.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 215

Chapter 4

Crystal bonding

This chapter focuses on the various forces that cause the atoms or the molecules
to prefer, at low enough temperatures, the periodic crystalline state. The binding
energy of this state and the manner by which it determines the type of the preferred
lattice are explained. The main considerations are based on electromagnetic forces
that operate in-between the electrons and the nuclei, and on quantum mechanics,
for instance on the uncertainty principle and on the Pauli exclusion principle. Anal-
yses of the various bonds among the ions in the crystal (ionic, covalent, molecular,
hydrogen, and metallic), that follow from these considerations are detailed; like-
wise, the energies which dominate the spatial arrangements of magnetic moments
in magnetic solids are explained.

4.1 Preface

The binding energy of a crystal is the difference between two energies,

the one of its constituents when they are far away from each other, and
the energy of the arrangement in which they are close together in a
periodic array. The crystal is formed when this difference is positive, i.e., when
energy is “gained” upon the crystal creation. In other words, when the potential
energy of the crystal is lower than the one of its separated constituents, then energy
must be invested in order to disassemble it. The crystal is hence stable. The
binding energy of a molecule is defined in a similar fashion: atoms join together to
form a molecule when their total energy is reduced by so doing. In fact, the whole
crystal may be viewed as a very big molecule. For this reason, the present chapter
is based on quite a number of concepts borrowed from the realm of chemistry, that
are needed for understanding the nature of the crystalline bond.
There exist several types of bonds among atoms, depending on the specific elec-
tronic structure of each of them. The potential energy of a pair of atoms (or ions) is
qualitatively illustrated in Fig. 1.1: it corresponds to a repulsion at short distances,
and to an attraction at long distances. Given this form, the total binding energy
of each of the crystalline structures listed in Chapter 2 can be derived. Because

215
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 216

216 Introduction to Solid-State Physics

their number is finite, it is feasible to compare the various binding energies and to
determine which of the structures has the highest binding energy. This is the array
the material forms in its ground state. The results are often in agreement with the
structures derived from scattering experiments (Chapter 3). Certain materials (as
mentioned in Chapter 1) can exist in several different crystalline structures, called
polymorphs. The discussion in this chapter is based only on energetic considera-
tions, which are supposed to identify the structure at sufficiently low temperatures.
Raising the temperature leads to a competition between the energy and the en-
tropy, which in turn induces transitions to other crystalline patterns, as mentioned
in Sec. 2.7. A qualitative survey of the various crystalline bonds is given below; the
next sections offer quantitative analyses of each of them. Several of these analyses
require some knowledge of quantum mechanics.
The ionic bond. The simplest “crystal” is a molecule comprising two atoms.
The nature of the bond between the atoms stems from their respective positions
in the periodic table. The atoms at the far left side of the table (e.g., the alkali
atoms in the first column of the table) have a small number of electrons in their
outer shell. These electrons, termed “valence electrons”, are only weakly coupled
to the atom, and are therefore easily detached from it. Once this happens, the atom
becomes a positive ion with the same completely-full shell as the noble gas atom
situated at the end of the previous row of the periodic table. As opposed, the outer
shell of the atoms on the right side of the table (e.g., the halogens in the column
second to the last one) is almost full, and they tend to absorb additional electrons
in order to completely fill it and to attain the stable configuration of the noble gas
situated after them in the periodic table. When two atoms from the two ends of the
periodic table are brought close to one another, they “gain” energy when the atom
from the left side (for instance, sodium) “donates” an electron to the other atom
(for example, chlorine) such that two electrically-charged ions are created (Na+
and Cl− ). The two ions electrically attract one another via the Coulomb force.
Quantum mechanics prevents the ions from collapsing one of top of the other by
the Pauli principle that forbids two electrons with identical quantum numbers to
occupy the same spatial position. As each of the nuclei is surrounded by a “cloud”
of electrons, this principle prevents the interpenetration of the “clouds”. Thus, once
each ion is assigned an ionic radius dictated by its electronic “cloud”, the optimal
structure is the one in which the corresponding spheres touch one another (as for
example in Fig. 2.7 or in the first line of Fig. 4.3); the distance between their
centers is the sum of the two ionic radii. At this distance, the gain in potential
energy exceeds the energetic loss that results from the ionization and a molecule
is formed (NaCl in the example above). This description can also be viewed as if
there is an effective repulsion force, that prevents the ions to approach one another.
The bond thus created between the two ions is called the “ionic bond”. This
bond exists also in crystals built of such ions, for instance the table salt mentioned
above, and other crystals comprising pairs of ions as described in the previous
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 217

Crystal bonding 217

chapter: at each lattice site there is a point charge (positive or negative), and the
competition between the overall Coulomb energy and the effective repulsion energy
between neighboring ions (that results from the Pauli principle) determines the
crystal structure with its lattice constants. Since the ions’ clouds touch each other
in the optimal configuration, the quintessential ionic lattice is described by close-
packing of two types of spheres. In practice it is not always possible to achieve
such packing, but the ratio between the radii of the two ions does determine the
lattice structure. Indeed, explicit calculations along these lines do reproduce the
structures described in Chapters 2 and 3.
The covalent bond. This bond is formed when the atoms that build the
molecule reside in neighboring columns of the periodic table; in the extreme case
the molecule consists of two identical atoms, e.g., oxygen, nitrogen, or hydrogen,
which form diatomic molecules (O2 , N2 , or H2 ). Each of the atoms “donates” its
valence electrons (for instance, a single electron in the case of the hydrogen atom,
or two electrons in the case of oxygen), and these electrons move around both ions.
The quantum probability for the electrons to be between the two ions is larger than
the probability to find them elsewhere (see below). This is because in the first case
the electrons attract the two ions (the protons in the case of the hydrogen molecule)
and cause the system to gain energy. The gain depends on the distance between the
ions and consequently determines the molecule size. A bond of this type is termed
“covalent bond”. For two identical ions the electrons’ “cloud” is symmetric with
respect to the center of the molecule, and then the covalent bond is unpolarized
(the molecule lacks a dipole moment). When the two atoms are not identical (e.g.,
the CO or the H2 O molecule, see in the following) the electrons’ “cloud” can be
closer to one of the ions, and the molecule becomes electrically polarized. These
two configurations, as well as the ionic molecule, are depicted in Fig. 4.1.
The covalent bond prevails also in numerous crystals. Then, the crystalline
structure is dictated by the number of valence electrons of the atoms. Each of
these electrons can form a covalent bond with an electron of another atom, and
therefore the number of covalent bonds in which a certain atom participates, i.e.,
the coordination number of that specific atom, is just the number of its valence
electrons. For example, in the diamond crystal (left panel in Fig. 2.20) each carbon
“donates” four valence electrons such that each electron is bound to one of the four
carbon neighbors; the coordination number is hence four. The four bonds around
a carbon in diamond are also illustrated in the center of the upper line of Fig. 4.3.
The diamond is in fact a huge molecule inside which any pair of carbons is coupled
together by a covalent bond. In the graphene lattice (Fig. 2.8) each carbon has only
three neighbors in the plane, with which it forms three covalent bonds. The fourth
electron of each carbon is free to move around among the atoms in the crystal.
Hybridization and electronegativity. The ionic bond and the covalent one
constitute two extremes of the molecular bond. In the first case, the atomic wave
function of the electron is centered on the negative ion; in the second, the electron
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 218

218 Introduction to Solid-State Physics

Fig. 4.1: The probability distribution “clouds” of the electrons’ locations in a di-
atomic molecule: right panel–a polarized covalent bond; mid panel–an unpolarized
covalent bond; left panel–an ionic bond. In the right panel there is a positive
charge δ + and a negative charge δ − on the two sides of the molecule, and therefore
it possesses a dipole moment.

is in a quantum molecular state, that belongs to both ions. In general, it may

well be that the quantum wave function of each electron is that of an intermediate
configuration, comprising an atomic wave function on one of the ions and a molecu-
lar function shared by both of them. This is the “hybridization” of the two types
of bonds. The chemists define the “electronegativity” of an atom, which is a
relative measure for its ability to attract valence electrons. The electronegativity
of an atom is a dimensionless number with values between 0.8 for the alkali metals
and about 3 for the halogens. When the difference between the electronegativity
values of two ions is small (for instance, smaller than 0.5), the bond between them
is mainly a covalent one. When that difference is large (e.g., larger than 1.8), the
bond is predominantly ionic. Intermediate values of electronegativity imply that
the two types of bonds are hybridized.
The metallic bond. Extending the physical picture of the covalent bond,
electrons can be found also in quantum states shared by more than two atoms. For
example, the benzene molecule, C6 H6 , has the form of a hexagonal ring built of six
carbon atoms (with a hydrogen atom attached to each of them), see Fig. 4.2. The
carbon atom possesses four valence electrons; three of them are used for covalent
bonds with its two carbon neighbors and one hydrogen atom (see the left panel in
the figure). There remain six “free” electrons (one from each carbon) that are shared
by all six carbons. The right panel exhibits the “clouds” of these electrons. In fact,
the wave function of such an electron is a linear combination of these atomic states,
with equal probabilities to find the electron at any of the carbons. These electrons
move freely around the ring. A further extension of this situation is a metal: each
ion in the metal “donates” its valence electrons, placing them in states where (see
the upper right panel of Fig. 4.3) the probability to find the electron around any
of the ions is the same. Under certain circumstances the electrons are thus free
to move around, hence the high electrical conductivity of metals. In a sense, the
metal is a huge molecule as well. The attraction between the electrons and the
nuclei reduces the potential energy, and consequently causes the binding energy to
increase. This is the basis of the metallic bond.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 219

Crystal bonding 219

Fig. 4.2: Two projections of the benzene molecule. The carbon ions are located at
the six vertices of the hexagon. The “clouds” represent the electrons participating
in the covalent bonds (viewed from above in the left panel) and the free electrons
common to all carbons (a side view). The different color tones indicate opposite
signs of the wave functions.

The van der Waals bond. As opposed to the configurations surveyed above,
where the atoms are relatively close to each other and so can share their elec-
trons, the atoms of the noble gases, which have fully-occupied shells and lack
valence electrons, cannot share electrons with other atoms. The same is true for
neutral molecules, where the electrons remain within each molecule and are not
transferred to any other. Nonetheless, the noble gases and numerous molecules do
arrange themselves in a periodic array, since they do attract one another. The at-
traction is due to the van der Waals force, and the bond so created is termed the
“van der Waals bond”. When two noble-gas atoms approach one another the
electrons’ cloud of each of them is shifted relative to the nucleus, creating a dipole
moment. Figure 4.3 displays (in the lower panel on the right) two possibilities to
form this dipole moment: in both cases the dipole moments attract one another.
The van der Waals energy is an average over the energies of such configurations.
This energy decays (see Sec. 4.4) with the distance R between the atoms as −R−6 .
The Pauli principle again gives rise to an effective repulsion at short distances,
which is customarily described by a steep potential increasing as R−12 (but any
other steeply-rising potential will do). The sum of the two potential energies is the
Lennard-Jones potential, illustrated in Fig. 1.1 [see also Eq. (4.32)]. When
the potential energies of all pairs in the crystal are summed over, it turns out that
the minimal total energy is negative (and consequently the bonding energy is posi-
tive) when the atoms are organized in the periodic crystalline structure. The noble
gases arrange themselves on a FCC lattice, which has the highest bonding energy
in their case. This is confirmed in experiment.
The hydrogen bond. Another way to bond atoms or molecules is by the
hydrogen bond. The conspicuous example of this bond is the attraction among
the molecules of water. These molecules possess a dipole electrical moment, since
the two hydrogen atoms are situated on the same side of the oxygen atom (this
follows from the nature of the covalent bonds within the molecule, see below).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 220

220 Introduction to Solid-State Physics

Hence, the hydrogen atoms of one water molecule are attracted, via the Coulomb
force, to the oxygen of a neighboring water molecule and thus form a bond between
the two. As shown in the following, this type of bonds gives rise to various phases of
ice, and to the beautiful patterns of snow flakes. One such pattern of crystalline
ice, where the hydrogen bonds among neighboring oxygen and hydrogen atoms are
perspicuous, is illustrated in the lower left panel of Fig. 4.3 (the small spheres
that represent the hydrogen atoms are attracted to the large ones marking the
oxygen atoms). The same bond can be formed whenever hydrogen ions in one
molecule approach the negatively-charged face of another molecule (or even the
same molecule, as happens in huge organic molecules). Since biological materials
are rich in hydrogen, the hydrogen bonds play a very important role in biology. For
instance, these bonds couple together the two strands of the double helix of DNA.

Fig. 4.3: Summary of the various crystalline bonds. The upper panel shows the ionic
bond (e.g., table salt), the covalent bond in diamond (the dark regions represent
the electrons’ “clouds” which create the bonds), and the metallic bond (in which
the electrons move freely around). The lower panel exhibits the van der Waals bond
(the median of the attraction energy of the dipoles in the two lines, represented by
the opposite charges δ + and δ − ), and the hydrogen bond, in the cubic crystal of
water.

Numerical values of the bonding energy and the melting tempera-

ture. Figure 4.3 summarizes the types of bonds mentioned so far. Quantitative
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 221

Crystal bonding 221

values of the binding energy are derived in the next sections; these determine the
specific crystalline structures. The calculations detailed in this chapter are often
in agreement with most of the crystal types discussed in Chapter 2 and identified
experimentally, as explained in Chapter 3. The binding energy per unit cell is of
the order 1-15 eV for the ionic and the covalent bonds, 1-5 eV for the metallic
one, 0.5 eV for the hydrogen bond, and 0.01-0.1 eV for the van der Waals bond.
The strength of the crystalline bond, i.e., the actual value of the binding energy, is
reflected in the melting temperature of the crystal. The larger is the bonding
energy, the higher is the melting temperature; the crystal then exists at higher tem-
peratures. Thus the melting temperature of diamond, copper, table salt, nitrogen,
and neon are about 4000◦ K, 1400◦ K, 800◦ K, 63◦ K, and 25◦ K, respectively. Indeed
(see below) the atoms in the first three materials are coupled by covalent bonds,
metallic bonds and ionic bonds, respectively, while those of the last two are van der
Waals ones (between the nitrogen molecules or the noble gas atoms).

4.2 The ionic bond

The bonding energy of a salt comprising two ions. A simple example of

an ionic crystal is a salt, which is built of an alkali metal from the first column
of the periodic table (for instance, sodium or cesium) and an halogenic atom from
the second to the last column (e.g., fluorine or chlorine). The sodium “donates”
its external electron and remains in the configuration of the noble gas neon (the
cesium ion resembles the xenon) with fully-occupied shells of electrons. The chlorine
“absorbs” the electron and attains the configuration of argon, also with full shells
of electrons. For both ions the electrons’ probability density around the nucleus is
approximately spherically-symmetric and the ion can be viewed as a point charge
located at a lattice site. The calculation of the bonding energy therefore requires
just the Coulomb energy of the charges representing the ions – it is the simplest
calculation of a crystalline bond. Consider an infinite lattice that contains a pair
of ions of opposite charges, ±q, in each unit cell (e.g., table salt). The Coulomb
energy of the lattice is given by an infinite double sum, UC = i=j [±q 2 /Rij ]/2,
where i and j denote two ions in the lattice, and Rij is the distance in-between
them. The factor 1/2 arises since each pair of ions appears twice in the double sum
over i and j. The sign of the ij term is determined by product of the signs of the
charges at sites i and j. It is convenient to replace the distance between the ions
by the dimensionless quantity pij = Rij /R01 , where R01 is the distance between
the two ions in the unit cell, i.e., the shortest distance between two ions of opposite
charges. In the simple ionic lattice that comprises only two types of ions and has
a single lattice constant, the ratio pij depends solely on the lattice structure and
not on the lattice constant (convince yourselves that this is true). For instance,
the distance between nearest neighbors sodium and chlorine ions in table salt is
R01 = a/2, where a is the cubic lattice constant of the FCC lattice, Fig. 2.17. The
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 222

222 Introduction to Solid-State Physics

distance between nearest-neighbor√ sodium ions (e.g., at the corner of the cube and√at
the center of a face) is R02 = a/ 2 and consequently the ratio p02 = R02 /R01 = 2
is independent of the lattice constant and has the same value for all crystals of the
table-salt type. One then finds UC = [q 2 /(2R01 )] i=j [±1/pij ] for all crystals of
the table-salt type. When there are more than a single lattice constant, e.g., in the
orthorhombic lattice, then pij is determined also by the ratio among the various
lattice constants, but not by their specific values.
In an infinite lattice one may shift the origin in the sum over j and measure all
distances in this sum relative to the origin at a point i (now marked by 0),
 
[±1/pij ] = ( [±1/p0j ]) .
i=j i j =0

For a crystal comprising two ions of opposite signs, the sum within the round brack-
ets is independent of i (and attains the same value, irrespective whether the origin
hosts a negative or a positive charge); therefore the result is just this sum multiplied
by 2N , where N is the number of unit cells in the lattice. The displacement of the
origin alluded to above is justified only for the infinite crystal and hence N → ∞.
The result, divided by N , is the specific bonding energy per unit cell,
UC 1  q2  1 q2    1 
= ± = ±
N N Rij 2N R01 i pij
ij j =i

q2  1  q2
= ± ≡− α. (4.1)
R01 p0j R01
j =0

The last step introduces the Madelung constant,

 1 
α=− ± . (4.2)
p0j
j =0

Dividing Eq. (4.1) by the volume of the unit cell yields the bonding energy per
unit volume. As mentioned, in a crystal that contains only two types of ions and
has a single lattice constant, the Madelung constant depends on the lattice struc-
ture alone. The numerical value of the Madelung constant is between 1 and 2, as
demonstrated below.
A one-dimensional crystal. Figure 4.4 displays a one-dimensional ionic lat-
tice, similar to the one in the second panel of Fig. 2.1: the unit cell contains two
ions, and the lattice constant is twice the distance between nearest neighbors. At
any distance R0j = jR01 there are precisely two ions (one on each side of the axis)
and hence
 1 1 1  ∞
(−1)n−1
α = 2 1 − + − + ... = 2 = 2 ln 2 ≈ 1.386 . (4.3)
2 3 4 n=1
n
∞
The last step makes use of the expansion ln(1 + x) = n=0 (−x)n /n.
The series in Eq. (4.3) converges only slowly: the outcome of sums over a finite
number of terms oscillates between the result of an even number and that of an
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 223

Crystal bonding 223

Fig. 4.4: One-dimensional ionic lattice built of negative ions (large circles) and
positive ions (small ones).

odd number of summands; the convergence is slow because of the slow decay of
the Coulomb interaction with the distance. At higher dimensions a sum like the
one in Eq. (4.2) is even more problematic, since it depends on the surface up to
which it is carried out. A surplus of positive or negative charges on the surface may
well change the result of the summation. Such a situation is termed “conditional
convergence”. The customary convention is to sum over finite samples for which
the total charge on the surface vanishes. For instance, one may first sum over
the contributions of the charges within the unit cell (the sum of these charges
vanishes) and then add the contributions from all unit cells up to the surface. e.g.,
in the sum of Eq. (4.3), one first combines pairs of successive members, to obtain
∞
α = 2 n=1 {1/[2n(2n − 1)]}. In this sum the terms decay as 1/n2 , which is faster
than the 1/n decay in Eq. (4.3).
Arbitrary dimensions. For a lattice that comprises two ions of opposite
charges, one first sums over the unit cells. Each cell contains two ions, e.g., a
positive one at the origin and a negative one at r1 in the same unit cell. The
distance in-between the two is R01 . Thus, a unit cell situated at R has a positive
charge at R, whose distance from the origin is R = p(R)R01 , and a negative one
located at R + r1 , whose distance from the origin is p(|R + r1 |)R01 . (In the example
given in Chapter 2, that of the table salt, r1 = R01 ŷ.) It therefore follows that
1  1 1
α= − − , (4.4)
p(r1 ) p(R) p(|R + r1 |)
R=0

where according to the definition above p(r1 ) = 1. There exist several ways to
handle these sums. In the Ewald method the sum in the right hand-side is trans-
formed into an integral in momentum space, which in turn becomes a sum over
the reciprocal lattice; thus the method combines together a summation in space
with a summation in momentum space. In other methods one sums over all ions
within a sphere, adjusting the charges on the surface such that the total charge
in the sphere vanishes. In the Evjen method, as applied to a cubic lattice, the
summation is carried out on cubical shells encompassing a central ion, dividing
the charges as to cancel the total charge within each shell. Figure 4.5 illustrates the
Evjen method for the square lattice in the plane, which represents, e.g. the planar
cross section of the table salt (Fig. 2.7). In cartesian coordinates, the lattice point
(n, m) accommodates the charge (−1)n+m (in units of the electron charge, assuming
that there is a single elementary charge in each lattice site). Consider the squares
whose center is at the origin (the midpoint of the figure) and their edge length is
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 224

224 Introduction to Solid-State Physics

2n (in units of the distance between two neighboring ions of opposite charges, R01 ).
Divide the charges on each square such that a charge at the edge contributes half of
itself to the inner side, and half to its outer side, and a charge at the corner gives a
quarter of itself to the inner side and the rest, 3/4, to the outer side. The sum of all
charges within the square vanishes; the sum over the contributions to the Madelung
constant from square shells (between two successive squares) converges relatively
fast to the precise result (see problem 4.1).

Fig. 4.5: The planar cross section of table salt. In the Evjen method one sums
over the contributions of square “shells” to the Madelung constant, for example,
the shell in-between the two dashed squares in the figure. The contributions of each
charge are split between the inner part and the outer part of the dashed line, as
marked in the figure.

Problem 4.1.
a. Show that the total charge within each dashed square in Fig. 4.5 vanishes.
b. Compute the contribution of the charges in the inner side of the nth square to
the Madelung constant.
c. Derive an expression for the contribution of the charges located in-between the
nth and the (n − 1)th squares to the Madelung sum, prove that all these are posi-
tive, and compute explicitly the contributions of the first six squares. How do these
contributions decay with n? [Suggestion: run your formula on your laptop, and find
how many terms are needed for a precision of three digits after the decimal point.]

Three-dimensional crystals with two ions. In the simple cubic lattice each
site has six nearest neighbors (nn) at a distance 1 (in units of R01 ), √
12 neighbors in
√ shell (next-nearest neighbors, nnn), at a distance 2, 8 neighbors
the next spherical
at a distance 3, 6 neighbors at a distance 2, and so on. In the structure of table
salt (right panel in Fig. 2.17) the shells contain successively positive and negative
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 225

Crystal bonding 225

charges, and hence

 (−1)i+j+k−1 √ √ √
α=  = 6/1 − 12/ 2 + 8/ 3 − 6/2 + 24/ 5 + . . . , (4.5)
ijk
i2 + j 2 + k 2

where the sum runs over all (positive and negative) integers, and the notation
indicates that the term i = j = k = 0 does not appear. This sum converges
only modestly, and its result does depend on the point where it is terminated.
Nonetheless, careful computations for the ubiquitous structures of ionic crystals in
three dimensions give
NaCl : α = 1.747565 . . .
CsCl : α = 1.762675 . . .
ZnS (wurtzite) : α = 1.64132 . . .
ZnS (zinc blende) : α = 1.63806 . . . . (4.6)
The results for the two polymorphs of ZnS are quite close to one another. This
is expected, since the corresponding structures are the Bravais lattices HCP and
FCC, with a base comprising two ions. Recall (e.g., Fig. 2.24) that the coordination
numbers of these lattices are identical (12), and the structural differences between
them are minute. In fact, this is the reason for the natural appearance of this
material in two forms; the specific structure it chooses is dictated by the growth
conditions, etc. The same competition exists also in the (similar) materials ZnSe,
ZnO, and ZnTe, for which the charges are ±q = ±2 (in units of the electron charge).

Problem 4.2.
Write down the first five terms in the sum for the Madelung constant of the CsCl
lattice.

Crystals with several ions in the unit cell. The salts considered above are
of the AB type, with ±q charges on the A and B ions, respectively. In the more
general case, the material is built of several ions which have different charges; then
the Coulomb energies of the various ions in the unit cell are not equal to each other;
the Coulomb potential energy is
  qi qj  N    q j  N 
UC = = qi = qi φ i , (4.7)
Rij 2 Rij 2
ij i∈cell j =i i∈cell

where the sums over i in the last two expressions run over the ions within a single
unit cell; φi = j =i (qj /Rij ) is the electric potential which all other ions exert on
the i−th ion. It is straightforward to verify that this expression produces the one
in Eq. (4.1) when q1 = −q2 = q.

Problem 4.3.
A one-dimensional crystal is based on a unit cell comprising three ions, ABBAB-
BABB . . . (like in the third panel of Fig. 2.1, with identical B and C ions). The
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 226

226 Introduction to Solid-State Physics

charges of the ions are qA = 2e and qB = −e, the lattice constant (the distance be-
tween two successive A ions) is a, and the distances within the unit cell between two
B ions and between the ions A and B are RBB = a(1 − 2x) and RAB = RBA = ax.
Find the Coulomb energy per unit cell (once all cells in the chain are summed upon).
Hint: use the digamma function, defined by
∞

Ψ(x) = −γE + [1/(k + 1) − 1/(k + x)] ,
k=0

where γE ≈ 0.577216 is Euler’s constant. The values of this function are docu-
mented in various softwares, and can also be obtained by carrying out the summation
on a laptop.

The short-distance repulsion among the ions. The total negative Coulomb
potential, Eq. (4.1) or Eq. (4.7), increases as the ions with opposite charges ap-
proach each other. Therefore, these ions “try” to to be close to each other as much
as possible. The Pauli principle, as mentioned in Sec. 4.1, forbids the electron
clouds of different ions to interpenetrate. This fact can be accounted for when each
ion is assigned an ionic radius, such that outside this range the probability density
to find electrons is low. The charged spheres of the positive and negative ions are
then brought closer to one another in the unit cell (in order to reduce the distance
in-between them). Measurements of the structure factors of ionic crystals indeed
confirm this qualitative picture. (Recall that the structure factor is the Fourier
transform of the electron density in the unit cell and its measurement conveys
information about the volume around each ion where the electrons’ cloud is con-
centrated.) Figure 4.6 displays typical values of the atomic radii and the different
ionic radii in ionic crystals (the radii of the ions are also affected by their crystalline
environment; the values listed in the table are approximate). As can be expected,
the radii increase as one descends from row to row in the periodic table (and the
number of electronic shells increases). Usually the radii of the elements become
shorter along the row while the charge of the nucleus increases. The differences
between the ionic radii (e.g., the cesium ion is much larger than the sodium one,
and they are both smaller than the chlorine ion) explain the variety of crystalline
structures (e.g., the difference between table salt and cesium chloride).
Comparison of various lattice structures. Assuming that the charged
spheres that represent the positive and the negative ions touch one another en-
ables the comparison of the resulting lattice structures. Figure 4.7 illustrates the
charged spheres that represent table salt and cesium chloride. As seen, the size
of the chlorine ions is almost twice that of the sodium ones; therefore the latter
fit in the cavities of the FCC lattice of the chlorine ions. The cesium ions are
bigger, and consequently the simple-cubic arrangement of the chlorine ions, with
the cesium ions placed at the center of each cube, is preferred. This geometrical
consideration can be given a quantitative basis. As mentioned, the Coulomb energy
is largest when the positive and negative spheres are contiguous. Then the distance
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 227

Crystal bonding 227

Fig. 4.6: Typical values (in nanometers) of atomic radii and ionic radii in
ionic crystals. Taken from http://www.boomeria.org/chemlectures/textass2/
table10-9.jpg.

between their centers is R01 = r< + r> , where r> is the larger ion radius (chlorine),
and r< is the radius of the smaller one (sodium or cesium). It follows that the
length
√ of the diagonal of the cubic unit cell in the cesium chloride structure is
a 3 = 2(r< + r> ). This structure is feasible only when the lattice constant (which
is the distance between two√neighboring chlorine ions) is such that a > 2r> , and
therefore 1 < r> /r< < 1/( 3 − 1) ≈ 1.366. On the other hand, in the structure
of table salt, the sodium ion is located at the midpoint of the edge of the cubical
unit cell, and therefore (since it is contiguous to the chlorine) the lattice constant
must be a =√2(r< + r> ). The length of the diagonal √ of the face of the cubic cell is
such that a 2 > 4r> ; consequently r> /r< < 1/( 2 − 1) ≈ 2.414. Apparently, the
FCC lattice fits better the table salt crystal for 1.366 < r> /r< < 2.414, because
for smaller ratios the cubic lattice with the higher Coulomb energy is preferred [see
Eq. (4.6)], and this is the lattice of cesium chloride. A further increase of this ratio
leads to the zinc-blende structure (ZnS), (Fig. 2.20, right panel), which may be the
most suitable for 2.414 < r> /r< < 4.55 (see problem 4.4). Indeed it is known, from
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 228

228 Introduction to Solid-State Physics

various sources (see Fig. 4.6), that the radii ratio for sodium chloride is approxi-
mately 1.9, and that of cesium chloride is about 1.1, in accordance with the lattice
structures associated with these crystals. Interestingly enough, high pressures may
change the ratio of the radii, and cause at times phase transitions among various
structures.

(a) (b)

Fig. 4.7: The packing of (a) the table-salt lattice and (b) the CsCl one. The larger
spheres are the chlorines, and the smaller ones are the sodium ions or the cesium
ones.

So far only geometrical arguments have been exploited to deduce the crystal
structure. One can also compare the Coulomb energies in various lattice structures.
Consider a certain salt, in which the radii of the ions are r> and r< . In the cesium-
chloride structure of this salt, and when the ions of opposite charges in the unit cell
are contiguous to one another, the distance between√them is R01 = r< + r> . This is
the preferred structure as long as 1 < r> /r< < 1/( 3 − 1) ≈ 1.366. When 1.366 <
r> /r< these ions cannot touch one another any more (since now the negative ions
are contiguous). In this case, the distance between√the neighboring
√ positive and
negative ions is half of the cube diagonal, R01 = a 3/2 = 3r> . Therefore the
√ is −N αCsCl q /(r> + r< ) in the first range of radii’s length, and
2
Coulomb energy
−N αCsCl q /( 3r> ) in the second. Had the same ions been
2
√ arranged in the structure
of the table salt, then R01 = r> + r< for r> /r< < 1/( 2 − 1) ≈ 2.414. Since the
Madelung constant of the cesium-chloride structure is slightly larger than that of
the table salt, the preferred pattern is the one of the cesium-chloride in the first
range. In the range 1.366 < r> /r< < 2.414 the Coulomb energy of the table-salt
structure is −N αNaCl q 2 /(r> + r< ), and this energy is lower than the one pertaining
to the cesium-chloride structure only when
√
r> /r< > 1/[(αNaCl /αCsCl ) 3 − 1] ≈ 1.3943 , check!
According to this argument, the cesium-chloride structure is still the “winner” in
the range 1.366 < r> /r< < 1.3943, though the ions with opposite charges are not
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 229

Crystal bonding 229

contiguous there, and the table-salt structure exists only when 1.3943 < r> /r< .
However, when the energies of the two structures differ only slightly, other factors,
like the repulsion between the ions at short ranges (though it has been assumed
tacitly that ions of opposite charges are touching one another, the repulsion at
short distances may take them apart), the van der Waals interaction (considered
below), the quantum zero-point vibrational energy (discussed in the next chapter),
and more, have to be accounted for. In most of the salts the bond is not purely
ionic but is a combination (hybridization) of an ionic bond with a covalent one,
and this also affects the result. Therefore, one cannot always predict the structure
from simple geometrical considerations. Nonetheless, the outcome of these simple
considerations is often confirmed in experiment.
Problem 4.4.
a. Find the upper bound on r> /r< up to which the ions can touch each other in the
zinc-blende structure.
b. Determine the radii ratio at which the transition between the table-salt and the
zinc-blende structures takes place, by comparing the Coulomb energies of the two
structures.
Problem 4.5.
The lattice constant of a zinc-blende crystal is 5.42Å.
a. What is the distance R01 between the two ions in the unit cell?
b. Assuming a close-packed pattern, i.e., all nearest-neighbor ions are contiguous,
find the radii of the two ions.
c. The same ions, with the same radii, create also the wurtzite lattice of the
quintessential HCP structure. Assuming that the ions are still contiguous, de-
termine the lattice constants of this lattice.
d. What is the specific volume of a single ZnS unit in each of the aforementioned
cases?
The bonding energy and the effective repulsive potential. A quan-
titative way to account for the effect of the Pauli principle, that prevents the
electronic “clouds” of neighboring ions from “interpenetrating” each other, is to
build an effective potential energy which represents a repulsive force between the
neighboring ions in the unit cell, that keeps them apart. As mentioned, as long as
this potential is steep enough, the results it yields are not sensitive to its details.
It is customary to choose a convenient functional form for this repulsive potential,
and use it to compute the distance between nearest neighbors that produces the
minimal energy. A comparison of the outcome of this calculation with several mea-
sured quantities allows for the identification of the parameters that characterize the
repulsive potential. Two ubiquitous forms for such a potential are
m
UP 1 = C/R01 , UP 2 = λ exp[−R01 /ρ] . (4.8)
The potential UP 1 lacks any theoretical base, but is very convenient for computa-
tions. The potential UP 2 , at times named after Born and Mayer, mimics the decay
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 230

230 Introduction to Solid-State Physics

of the electronic wave functions around the atomic nucleus, and consequently also
that of the probability density for an electron to be at a radial distance R01 away
from the nucleus (for instance, in the ground state of the hydrogen atom). The
decay length ρ is of the order of r< + r> . The total potential energy of the crystal
takes the form
Utot (R01 ) = N [zUP (R01 ) − αq 2 /R01 ] , (4.9)
where z is the number of the negatively charged nearest-neighbors of each positive
ion (and vice versa). The lattice structure is reflected in the coefficients α and
z, and also in the parameters that characterize the effective potential UP . The
minimal value of Utot as a function of R01 yields the optimal distance R; inserting
the latter into Eq. (4.9) gives the binding energy per a unit cell of the lattice,
u = −Utot (R)/N .
The minimal value of Utot for the potential UP 1 is at R = [mzC/(αq 2 )]1/(m−1) ;
the binding energy per unit cell is then u = (1−1/m)αq 2 /R. Experiment shows that
the binding energy of ionic lattices is approximately the Coulomb energy of nearest-
neighbor ions, u ≈ αq 2 /R, and hence the parameter m is chosen to be much larger
than 1. This verifies the assertion of a steep repulsive potential. Measurements
of the binding energy and the distance R between nearest neighbors, allow for the
identification of the parameters m and C when the crystal structure is known.
Assuming that all salts share the same value for m, the binding energy is deter-
mined by the distance R among nearest-neighbor ions in the unit cell, at equilibrium.
As seen, this distance is mainly dictated by the radii of the two oppositely-charged
m−1
ions. From the expression R = [mzC/(αq 2 )]1/(m−1) follows that zC ∝ αR .
That is, the strength of the repulsive potential increases with the radii of the ions.
The relation u ≈ αq 2 /R allows to estimate the magnitude of the binding energy
in ionic crystals. It is expedient to measure R in units of the Bohr radius, aB =
2 /(me2 ) ≈ 0.53Å, and the binding energy in units of the Rydberg constant which
determines the ground-state energy of the hydrogen atom, Ry = m0 e4 /(22 ) =
e2 /(2aB ) ≈ 13.6eV (m0 is the electron mass, e is the electron charge). In these
units u/Ry = α(q/e)2 (2aB /R). Experiment shows that R is of the order of a few
Bohr radii, so when q = e, u is of the order of the Rydberg constant, and is in the
range 5-10 eV per unit cell. This energy increases with the amount of charge on
the ion.
It is worth emphasizing that the binding energy is derived from a comparison of
the energy of the ions when far apart from each other and the energy they acquire
when arranged on a lattice. At times it is preferable to compare the crystal energy
with the energy of the isolated atoms and not that of the ions. Then one has
to account for the ionization energy. In the case of the table salt, for instance,
the ionization of the sodium atom requires 5.1 eV, but 3.6 eV are gained since
the chlorine atom is ionized (i.e., by “adding” an electron). The energy u is then
about 7.9 eV. Therefore, the crystalline arrangement still leads to an energy gain,
(7.9 − 5.1 + 3.6)eV=6.4 eV.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 231

Crystal bonding 231

The bulk modulus. An independent identification of the parameters that

characterize the effective repulsive potential between neighboring ions (i.e., C and
m) is accomplished by a comparison of the measured and calculated values of the
bulk modulus, defined as B = −V (∂P/∂V )T , where P is the pressure and V = N v
is the volume (v is the volume of a unit cell). The inverse of the bulk modulus, i.e.,
1/B, is the compressibility coefficient, that gives the relative change of the volume
resulting from the applied pressure. As the pressure is given by P = −dUtot /dV ,
with the total energy Utot (R) = −N u, one finds

∂  ∂u  ∂2u
B = −v = −v 2 . (4.10)
∂v ∂v ∂v
3
By a dimensional argument, the volume of the unit cell is v = cR , where c is a
dimensionless factor (of order one). For instance, in the table-salt crystal, the ions
are arranged on an FCC lattice with a lattice constant a = 2R and a primitive unit
3
cell of volume v = a3 /4 = 2R , that is, c = 2. Upon changing the variables from v
to R,
∂u 1 ∂u 
= 2  ,
∂v 3cR ∂R01 R01 =R
and a further differentiation leads to
∂2u  ∂2u  2 ∂u  
1 
= 2 2  −  . (4.11)
∂v 2 (3cR )2 ∂R01 R01 =R R01 ∂R01 R01 =R

At equilibrium, R corresponds to the minimal energy, and therefore

(∂u/∂R01 )R01 =R = 0 and the second term in the brackets of Eq. (4.11) vanishes.
Inserting into Eq. (4.9) the potential UP 1 yields that for the table-salt structure
4
B = (m − 1)αq 2 /(18R ) .

Measurements of the bulk modulus and the lattice constant, in conjunction with
the identification of the lattice structure, allows to derive the decay factor m and
to verify that the two parameters m and C of the potential UP 1 , Eq. (4.8), indeed
yield plausible values for R, u, and B. For the alkali-halide crystals like table salt,
this calculation gives for m values between 6 (for the smaller ions) and 10 (for the
larger ones). Small errors in this parameter do not affect significantly the estimate
of the binding energy, which attains values between 6 eV to 10 eV (per unit cell). As
mentioned, the binding energy of ionic crystals is larger than the binding energies
of the other types of bonds.

Problem 4.6.
Repeat the above calculation for the table-salt crystal, with the potential UP 2 , Eq.
(4.8): express the binding energy, the bulk modulus, and the parameter zλ, in terms
of the distance in-between neighboring ions in the unit cell and the distance ρ.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 232

232 Introduction to Solid-State Physics

4.3 Thecovalent bond

The Born-Oppenheimer approximation. A crystal comprises atoms (or ions)

built of nuclei and electrons. As the nuclei are much heavier than the electrons, they
move much slower. The Born-Oppenheimer approximation is based on this fact:
one assumes that the nuclei are inert and calculates the quantum wave-functions
of the electrons in the presence of the electric potential of these static nuclei. This
approximation is used in computations of bond energies of molecules and solids.
For a certain set of the nuclei locations, the electrons are in the state with the
lowest energy (excitations to higher-energy levels have smaller Boltzmann proba-
bilities, since the energy differences are large compared with the thermal energy).
This ground-state energy, which depends on the aforementioned set, can be further
decreased provided that those locations are chosen suitably: minimizing the energy
determines the equilibrium locations in the molecule or in the solid. The next stage
of the calculation (discussed in Chapter 5) accounts for the small vibrations around
these equilibrium locations.
The hydrogen molecule. The simplest molecule is that of hydrogen, H2 . It
contains two protons and two electrons. Instead of treating these two electrons
together, it is advantageous to begin with a system of two protons and a single
electron, as shown in Fig. 4.8. This figure displays the ion H+ 2 . In the Born-
Oppenheimer approximation the two nuclei, at distance R from one another, are
placed on the ẑ− axis, (the points A and B in the figure) and the Schrödinger
equation is solved for the wave function of the electron (marked in the figure by e).
The ground-state energy resulting from this procedure is inevitably a function of
R. An approximate calculation of this function is shown by the lower curve in Fig.
4.10: it has a single minimum as a function of R, and the value of R there, denoted
R, is the equilibrium distance between the two nuclei of the H+ 2 ion.

Fig. 4.8: The ion H+

2 . The two protons, each of a charge +e, are nailed at the points
A and B. The third particle (marked by e) is the electron. The vector R connects
nucleus A with nucleus B.

Placing the origin at the mid-point O between the two nuclei, and the electron
at the point r relative to this origin (Fig. 4.8), the Hamiltonian of the electron in
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 233

Crystal bonding 233

the H+
2 ion is

2 2
Ĥ = − ∇ + U (r) , (4.12)
2m
where the first term represents the kinetic energy (the gradient comprises derivatives
with respect to the components of r), and the electric potential (including also that
between the nuclei) is
e2 e2 e2
U (r) = − − + . (4.13)
rA rB R
[Equation (4.13) exploits CGS units; to convert to the SI units, one has to divide
the Coulomb energies by 4π0 , where 0 is the dielectric constant of the vacuum.]
The dependence on r on the right-hand side of Eq. (4.13) is via rA = (R/2) + r,
rB = −(R/2) + r.
The wave function for well-separated nuclei. When the distance in-
between the nuclei exceeds considerably the Bohr radius, R  aB = 0.529Å, the
electron can be in the ground state of the hydrogen atom near any of the nu-
clei. For instance, when the electron is near
 the A nucleus, then its wave func-
tion is ψA = ψ100 (rA ) = exp[−rA /aB ]/ πa3B , and the ground state energy is
E1 = −Ry = −13.6eV, where Ry is the Rydberg constant. The indices 100 are
the quantum numbers, n m, of this state (energy, angular momentum, and the
z−component of the angular momentum). In such a configuration rB ≈ R, the last
two terms in Eq. (4.13) nearly cancel one another and the Hamiltonian becomes
that of the hydrogen atom around the nucleus A. Alternatively, the electron may
be found near nucleus B, and then its wave function is ψB = ψ100 (rB ), with the
same energy. Each of these configurations describes an “ionic” state, in which the
electron is “located” near one of the nuclei. These two states have the same en-
ergy and hence are degenerate. Therefore, any linear combination of them (with
arbitrary complex coefficients), Ψ(r) = αA ψ100 (rA ) + αB ψ100 (rB ), is a solution of
the same Schrödinger equation. The normalization of the wave function implies
|αA |2 + |αB |2 = 1 (as is shown below, this expression for the normalization is valid
only when the nuclei are far apart). |αA |2 or |αB |2 are the probabilities to find
the electron around nucleus A or nucleus B, respectively. When the two nuclei are
identical then |αA |2 = |αB |2 = 1/2 by symmetry. The state Ψ(r), in which the
electron is “shared” by the two nuclei, is called a covalent state. When the nuclei
are not identical, the probabilities are different but still the energy of the covalent
state can be lower than that of the ionic states, in which the electron is near one
of the nuclei.
The symmetry of the wave functions. Denote the full solution of the
molecular Schrödinger equation [including the full potential, Eq. (4.13)] by Ψ(r) =
Ψ(rA , rB ). Often, the symmetries of the wave functions can be deduced from the
symmetries of the Hamiltonian, prior to undertaking the technical task of solving
the corresponding differential equation. The Hamiltonian (4.12) possesses several
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 234

234 Introduction to Solid-State Physics

symmetries: it is invariant under a rotation of the electron location around the

molecular axis AB, upon interchanging the nuclei (i.e., replacing R by −R), and
also under a reflection through a mirror perpendicular to the molecular axis at
point O (check!). The Hamiltonian is also unchanged when the sign of r is inverted
(this latter symmetry is also obtained by a combination of the interchange of the
nuclei locations and a rotation around the axis). This inversion is identical to the
interchange rA ↔ rB . Hence, the Hamiltonian obeys Ĥ(r) = Ĥ(−r).
Each eigenfunction of such a Hamiltonian is either symmetric or antisymmetric
with respect to this interchange. The Schrödinger equation, Ĥ(r)Ψ(r) = EΨ(r),
can be also written as Ĥ(r)Ψ(−r) = Ĥ(−r)Ψ(−r) = EΨ(−r), and therefore the
two functions, Ψ(r) and Ψ(−r), solve the same Schrödinger equation, with the
same energy E. It has been already found that in the limit of an infinite distance
in-between the nuclei the ground state is doubly-degenerate. When the distance
between the nuclei is finite, the corrections to the single-atom potential in Eq.
(4.13) remove the degeneracy, and split the energy of the ground state into two
different levels (see below). In the non-degenerate situation the two functions, Ψ(r)
and Ψ(−r), are identical up to a multiplicative constant, Ψ(−r) = CΨ(r) (since
the Schrödinger equation may be multiplied by a constant). Therefore, Ψ(r) =
CΨ(−r) = C 2 Ψ(r); it follows that C = ±1, i.e., Ψ(−r) = ±Ψ(r). The function
is even or is odd under the interchange rA ↔ rB .
When the two nuclei are far apart, the even and the odd functions are given by
Ψ± (r) = A± (ψA ± ψB ) = A± [ψ100 (rA ) ± ψ100 (rB )] , (4.14)
where the plus (minus) sign denotes the limit of the symmetric (anti-symmetric)
solution.
√ The normalization of the wave function implies that αB = ±αA = ±A± =
±1/ 2, and therefore |αA |2 = |αB |2 = 1/2, as is expected from symmetry consid-
erations; there is an equal probability to find the electron in the vicinity of each of
the nuclei. These two functions and the corresponding probabilities are displayed
in Fig. 4.9.
An approximate solution for the electron in the H+ 2 ion. Though the
Schrödinger equation of this system can be solved exactly (by transforming it to
elliptic coordinates), the solution presented here is obtained in the variational
approximation which is further exploited below. This approximation is explained
in Appendix C. Within the variational-method scheme one uses a trial wave function
for the ground state, which is in fact a guess, and calculates the average of the
quantum energy for this function. The trial wave functions used here, even when
the nuclei are not located far away from one another, are those appearing in Eq.
(4.14). The calculated energy is an upper bound on the exact energy of the desired
ground state. The results of this calculation are presented in Fig. 4.10. As seen
there, the energy E+ is minimal for a certain distance between the nuclei R, and
therefore the energy of the “true” ground state also possesses such a minimum.
The conclusion is that energy can be “gained” by pushing the nuclei towards each
other, thus allowing the electron to be simultaneously on both nuclei. This energetic
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 235

Crystal bonding 235

Fig. 4.9: The wave functions (top panel) and the corresponding probability density
(mid panel), for θ = 0 and X = R/aB = 4. The dashed curves in the upper
panel are the atomic wave functions on each of the nuclei. The lower panel displays
equal-probability curves on a plane normal to the molecular axis (brighter regions
correspond to higher probabilities).

“gain” (as compared with the configuration where the electron is around only one
of the nuclei) is the binding energy of the molecule.
The variational approximation. As explained in Appendix C, the variational
approximation is based on a calculation of the expectation value of the energy
for a broad collection of trial wave-functions, and then choosing out of them the
one that yields the lowest energy. Here it is applied to the two combinations of
atomic wave functions [Eq. (4.14)], assuming an arbitrary distance in-between
the nuclei (though these wave functions are derived for two nuclei which are far
apart from one another). Then, the (real) atomic wave functions ψA and ψB are
not orthogonal because they overlap, and consequently the normalization constants
A± are determined by the requirement Ψ± |Ψ±  = 1, i.e.,

|A± |−2 = dr(|ψA |2 + |ψB |2 ± 2ψA ψB ) = 2(1 ± S) , (4.15)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 236

236 Introduction to Solid-State Physics

Fig. 4.10: The averaged energies (in units of Ry , and relative to the energies of the
configuration in which the two nuclei are well separated) for the symmetric wave
function Ψ+ (solid curve) and the anti-symmetric function Ψ− (dashed line), as
functions of X = R/aB .

where S = drψA ψB is the overlap integral. The first two integrals in the
mid equality of Eq. (4.15) are equal to 1, since the atomic wave functions are
normalized. The overlap integral contains the product of the two wave functions;
as each of them decays exponentially with the distance, this integral diminishes as
R becomes longer. Using the identity rB 2
= R 2 + rA 2
− 2RrA cos θ, and integrating
over rA in spherical coordinates (with the notations X = R/a √ B , x = rA /aB , and
transforming the θ−integration into an integration over y = X 2 + x2 − 2Xx cos θ,
dy = Xx sin θdθ/y) gives

1
S= drA e−rA /aB e−rB /aB
πa3B
 ∞  π √
2 2
=2 dx dθe−[x+ X +x −2xX cos θ] x2 sin θ = e−X (1 + X + X 2 /3) . (4.16)
0 0
Within the variational approximation, each of the functions, the symmetric Ψ+ and
the antisymmetric Ψ− , can be chosen as a trial ground state. The energy of the
first is then the upper bound on the lowest energy of a symmetric state, while the
energy of the second is the upper bound on the lowest energy of an anti-symmetric
state (the symmetric and the anti-symmetric states do not mix with one another as
the eigen states have a definite symmetry).
 The upper panel in Fig. 4.9 displays the
functions Ψ± = (e−rA /aB ± e−rB /aB )/ 2πa3B (1 ± S) for the special case θ = 0, with
rA,B = |R/2±r|. The dashed lines exhibit each of the summands. The mid panel of
the figure shows the probability density to find the electron on the molecule axis for
each of the wave functions, and the lower panel portrays lines of equal probability
(of the electron location) in the plane normal to that axis. The results are not
modified when this plane is rotated around the axis. These probability “clouds”
are described below by such curves.
The variational method can be used to estimate the ground-state energy of the
electron in the H+2 ion. Inserting the function Eq. (4.14) into Eq. (C.1) yields

E±  = (2 ± 2S)−1 [ψA |Ĥ|ψA  + ψB |Ĥ|ψB  ± 2ψA |Ĥ|ψB ]

= (1 ± S)−1 [ψA |Ĥ|ψA  ± ψA |Ĥ|ψB ] . (4.17)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 237

Crystal bonding 237

The symmetry of the Hamiltonian under the interchange of the locations of the
two nuclei is exploited in the second step. Using the Hamiltonian, Eqs. (4.12) and
(4.13), gives
e2 e2 e2  R  −2R/aB
ψA |Ĥ|ψA  = E1 + − ψA | |ψA  = E1 + 1+ e . (4.18)
R rB R aB
The first term on the right hand-side of Eq. (4.18) arises since ψA is the ground-state
wave function of the hydrogen atom near ion A,
 2 2 e 2 
− ∇ − ψ = E1 ψ A .
2m rA A
The second term results from a straightforward integration,

e2
ψA | |ψA  = e 2
d3 rA |Ψ100 (rA )|2 /rB .
rB
Similarly,
 e2  e2  e2  e2  R  −R/aB
ψA |Ĥ|ψB  = E1 + S − ψA | |ψB  = E1 + S− 1+ e .
R rB R aB aB
(4.19)
The ground-state energy of the electron when the nuclei are infinitely far away
from one another is E1 , and the additional energy due to the bond between the
two members of the molecular ion is E±  − E1 . This difference, whose explicit
expression is obtained upon inserting the results derived above into Eq. (4.17), is
displayed in Fig. 4.10 for the two wave functions Ψ+ and Ψ− . As seen, the anti-
symmetric wave function Ψ− does not yield an energetic gain (at least, not in the
approximation used here). For this reason, this function is called “anti-bonding”,
and is not a suitable candidate for being the ground state of the molecule. On the
other hand, the symmetric function Ψ+ yields negative energies for a large range of
X = R/aB values; it is called a “bonding” function. The quantitative reason for
this disparity can be seen from Fig. 4.9: in the symmetric state the electron has a
larger probability to be in-between the two nuclei, it is then close to both of them,
and “gains” energy from the Coulomb attraction of the two nuclei. As the energy
calculated above is an upper bound on the true energy of the ground state, the latter
is clearly more negative, and thus holds together the two parts of the molecule. This
calculation yields an optimal distance of 1.32Å and a binding energy of 1.76 eV,
while the exact values are 1.06Å and 2.79 eV. The approximation can be improved
by exploiting better trial wave functions. For example, one may add atomic wave
functions of higher levels to the functions ψA and ψB , as described in Appendix C,
see Eq. (C.3). The main point is that the variational method proves that a bound
state with a positive binding energy does exist. (Recall that the binding energy is
the difference between the energy of the nuclei (with the electron) when far apart
and the energy when they are in the bound state. Therefore a minimal negative
energy in the bound state is identical to a positive binding energy.)
Problem 4.7.
Prove the results (4.16), (4.18), and (4.19).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 238

238 Introduction to Solid-State Physics

Covalent bonds between different atoms. When the two atoms that par-
ticipate in the covalent bond are not identical, the symmetry for interchanging them
is lost and Eq. (4.14) is replaced by a linear combination of the two real atomic
wave functions, Ψ± = A± [ψA (rA ) ± β± ψB (rB )]. Assuming that β± are real and
positive, the normalization yields

|A± |−2 = dr(|ψA |2 + β± 2
|ψB |2 ± 2β± ψA ψB ) = 1 + β±
2
± 2β± S , (4.20)

and the condition for orthogonality gives


dr[|ψA |2 − β+ β− |ψB |2 + (β+ − β− )ψA ψB ] = 1 − β+ β− + (β+ − β− )S = 0 .
(4.21)
It follows that the choice β+ = β− = 1 made in Eq. (4.14) is still possible, though is
not enforced by symmetry, and does not necessarily yield the lowest energy. Instead,
one may consider the coefficient β+ as a variational parameter, and choose it such
as to minimize the energy (see problem 4.8). The other coefficient is determined by
β− = (1 + Sβ+ )/(β+ + S). The resulting energy turns out to be negative at large
R, and hence corresponds to a bonding state. The ratio between the probability to
find the electron near ion B to that of finding it near A in that bonding state is
|β+ |2 .

Problem 4.8.
Extend Eq. (4.17) to allow for a bond between different states, and determine the
coefficients β+ and β− which maximize the binding energy. Verify that the values
obtained comply with the orthonormalization of the functions Ψ± .

Molecular orbitals and the H2 molecule. Since the bonding wave function
of H+2 as described above (whether in the variational approximation or else) cor-
responds to symmetric states in which the electronic orbits are shared by both
atoms (i.e., the electrons “move” around both nuclei simultaneously), it is called a
“molecular wave function”, or a wave function of a molecular orbital. Con-
sider the H2 molecule, that has two nuclei and two electrons. This molecule is
created when a second electron is added to the ion H+ 2 . By the Pauli principle,
each molecular state can be occupied by two electrons provided that their spin states
are opposite to one another. This is similar to the ground state of the helium atom,
in which there are two electrons in the lowest electronic orbital ψ100 . Ignoring the
Coulomb repulsion between the two electrons, each of them is in a bonding ground
state and the wave function of the two electrons is Ψ2 (r1 , r2 ) = Ψ+ (r1 )Ψ+ (r2 ),
where r1 and r2 are the coordinates of the two electrons. In this case the bind-
ing energy is roughly twice that of the ion H+ 2 . The Coulomb repulsion can be
accounted for by using Ψ2 (r1 , r2 ) as a variational wave function, and then the cor-
rection to the ground-state energy is the expectation value Ψ2 |e2 /r12 |Ψ2  of the
repulsion energy. The same result is obtained to first order in perturbation theory.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 239

Crystal bonding 239

This positive correction reduces the binding energy, but the total remains positive,
higher than that of the H+2 ion.
The Heitler-London approximation for a covalent bond with two elec-
trons. Inserting Eq. (4.14) in the expression Ψ2 (r1 , r2 ) = Ψ+ (r1 )Ψ+ (r2 ) gives
Ψ2 (r1 , r2 ) = Ψ+ (r1 )Ψ+ (r2 ) = A2+ [ψA (r1A )ψA (r2A ) + ψB (r1B )ψB (r2B )
+ ψA (r1A )ψB (r2B ) + ψA (r2A )ψB (r1B )] , (4.22)
where the radii r1A , r1B , r2A , and r2B are the radius vectors of each of the electrons
to each of the nuclei. The first two terms on the right hand-side of Eq. (4.22)
represent states in which the two electrons are on the same nucleus, and therefore
each of them is in an “ionic” state (one of the electrons moved from its “parent”
atom to the other one). The last two terms describe states in which each electron
resides on a different nucleus. In the next step of the calculation the energy is
augmented by the average of the repulsion between the two electrons. As quite
generally this one is higher when the two electrons are located on the same ion,
it is customary to carry out the variational computation with different coefficients
for the first two, “ionic” terms, and for the last two, “molecular” terms. The more
similar the two atoms are, the higher is the weight of the molecular terms. In the
Heitler-London approximation the two “ionic” functions are completely ignored,
and the symmetric variational wave function is
ΨHL
2+ (r1 , r2 ) = A+ [ψA (r1A )ψB (r2B ) + ψA (r2A )ψB (r1B )] . (4.23)
This function yields expressions like those in Eq. (4.17) (where each integral is
carried out over the coordinates of both electrons), and energies whose qualitative
dependence on R is similar to the ones shown in Fig. 4.10 (with different numeri-
cal values). The Heitler-London approximation gives an improved estimate of the
ground-state energy as compared to the result obtained from the wave function
based on the molecular orbitals, Ψ2 (r1 , r2 ) = Ψ+ (r1 )Ψ+ (r2 ). Extending Eq. (4.23)
to include the ionic terms (in the variational approximation),
 
Ψion
2 (r1 , r2 ) = B ψA (r1A )ψA (r2A ) + ψB (r1B )ψB (r2B ) ,

with various coefficients, improves the approximation. Such a combination, already

mentioned in Sec. 4.1, is termed hybridization of an ionic and a covalent bond.
Exchange energy, spin and exchange symmetry. The electronic Hamil-
tonian of the hydrogen molecule H2 contains the single-electron Hamiltonian (for
each of the two electrons), Eq. (4.12), and the interaction between the two elec-
trons, e2 /r12 . This Hamiltonian is symmetric with respect to the interchange of
the two electrons, and therefore it can be proved that the wave function of the two
electrons is either symmetric or is antisymmetric under such an interchange. The
proof resembles the one presented for the symmetry of interchanging the two nuclei
in the H+2 molecule. Indeed, in the examples of two-electrons’ Hamiltonian treated
above, the wave function of the two electrons is symmetric with respect to their
interchange, Ψ2 (r1 , r2 ) = Ψ2 (r2 , r1 ). For instance, the variational energy obtained
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 240

240 Introduction to Solid-State Physics

with the antisymmetric state, Eq. (4.23) with the negative sign between the two
terms,

2− (r1 , r2 ) = A− [ψA (r1A )ψB (r2B ) − ψA (r2A )ψB (r1B )] ,

ΨHL
is higher. This is since this wave function vanishes when r1 = r2 , i.e., when the
electrons are on the plane that bisects the line connecting the two nuclei. In the
symmetric case the probability to find the two electrons in-between the two nuclei
is higher; as a result there is an energy gain due to the Coulomb attraction between
each of the nuclei and the electronic “cloud” located in-between them. As in Eq.
(4.17), the difference between the energy of the symmetric state and that of the
antisymmetric one contains the energy
ψA (r1A )ψB (r2B )|Ĥ|ψA (r2A )ψB (r1B ) ,
and the overlap integral, ψA (r1A )ψB (r2B )|ψA (r2A )ψB (r1B ). In both expressions
the electrons are interchanged between the wave functions on the left and on the
right sides, and therefore this energy difference is termed “exchange energy”.
The considerations given so far pertain to the spatial part of the quantum wave
function, which gives the probability density of the electrons in space. Electrons
possess also the spin degree of freedom; the spin is described by the vector operator
Ŝ. The square of this vector is Ŝ2 = s(s + 1), with s = 1/2, and its ẑ− component
attains two values, Ŝz = ±s. The Hilbert space of the spin states is thus spanned
by two base vectors, denoted | ± s. The Pauli principle dictates that when the
spatial state of two electrons is symmetric with respect to their interchange (as in
the examples discussed above), their spin state must be anti symmetric with respect
to the spins’ interchange, such that the total wave function is antisymmetric upon
interchanging all degrees of freedom of the two electrons. The total spin state is
then
1  1   1   1   1  
   
χS = √    −  −  −  .
2 2 1 2 2 2 2 2 1
The operator of the total spin of the two electrons is Ŝ12 = Ŝ1 + Ŝ2 . Operating with
Ŝ212 on the state χS yields zero. The ẑ−component of the total spin vanishes as
well. There is only a single state with these properties, called singlet. Therefore,
the ground state of the hydrogen molecule is a singlet.
The hydrogen molecule possesses also the antisymmetric spatial state, e.g.,
HL
Ψ2− (r1 , r2 ), whose energy is higher than that of the symmetric state. Since the
spatial state is antisymmetric with respect to the exchange of the two electrons,
then according to the Pauli principle the spin state is symmetric. There are three
symmetric spin states of two electrons,
1 1  1  1
   
χT 1 =  , χT 1 =  − − ,
2 1 2 2 2 1 2 2
  
1 1   1       
1   1 
χT 0 = √    −  +  −  .
2 2 1 2 2 2 2 2 1
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 241

Crystal bonding 241

These three states are termed “triplet”. They are eigenfunctions of (Ŝ12 )2 , with
the eigenvalue 2, and of (Ŝ12 )z with the eigenvalues 1, −1, and 0 (as the subscripts
indicate). The triplet states correspond to states of unit (in the relevant units) spin
angular-momentum. In certain situations, the ground state can be a triplet.
Heavier atoms. The calculation above shows that the covalent bond between
two atoms is formed when two electrons “occupy” the same molecular orbital, which
is built of a linear combination of atomic wave functions on each of the ions. This cal-
culation is relatively simple since the atomic wave function of the hydrogen ground-
state, ψ100 , is spherically symmetric, and because the orbital ground-state is not
degenerate. It becomes more complex when the valence electrons participating
in the covalent bond are in higher electronic shells, as such atoms can contribute
to the bond more than one electron. The electronic states in the atom are usually
identified by the quantum number of the angular momentum. The state with
= 0 is denoted by the letter s, those with = 1 by p, with = 2 by d, etc. The
electronic configurations of hydrogen, helium, lithium, beryllium, boron, and
carbon atoms are then 1s, 1s2 , 1s2 2s, 1s2 2s2 , 1s2 2s2 2p, 1s2 2s2 2p2 , respectively,
where the “large” numbers on the left denote the quantum number n, and the small
numbers (superscripts) indicate the number of electrons in the state with the cor-
responding angular momentum. For instance, hydrogen has a single electron in the
shell with n = 1, = 0, while carbon has two electrons in this first shell, and also
four electrons in the next one (n = 2), two with = 0 and two with = 1. Each of
the valence electrons in the n = 2 shell can form a covalent bond with an electron
of another atom.
Examples for such configurations include, e.g., helium. In helium, the shell
n = 1, = 0 is occupied with two electrons which are then prohibited, by the
Pauli principle, from forming covalent bonds. The next atom in the periodic table
is lithium, with a single valence electron with n = 2, = 0. Lithium often “con-
tributes” this electron, forming an ionic bond. The atoms in the second row of the
periodic table are considered in the following.
Covalent bonds of beryllium. In the lowest atomic state of beryllium, the
two valence electrons (those in the outer shell, which are available for chemical
bonding) are in the 2s state to which another electron cannot be added (because of
the Pauli principle). Nonetheless, as the energy difference between this state and
the 2p one is relatively small, it is sometimes “profitable” for one of the electrons to
cross over to the 2p state, such that the electronic configuration becomes 1s2 2s2p.
Once this happens, the two valence electrons of the n = 2 shell can combine with
electrons from other atoms and form two covalent bonds. The functions with = 0
are spherically symmetric, e.g.,
√
ψ200 = 2[z/(2aB )]3/2 [1 − zr/(2aB )] exp[−zr/(2aB )]/ 4π ,

where z is the effective charge, due to the nucleus and the electrons in the first shell,
as “seen” by the electrons in the second shell. On the other hand, the functions
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 242

242 Introduction to Solid-State Physics

with = 1 are not symmetric, as they depend on the spherical coordinates θ and
ϕ, via the spherical harmonic functions Ym (θ, ϕ):
 z 3/2  zr  exp[−zr/(2a )]
B
ψ21m (r) = R21 (r)Y1m (θ, ϕ) , R21 (r) = √ ,
2aB aB 3
where m = −1, 0, 1 and R21 is the radial function. In place of the three functions
Y1m one may choose the real linear combinations,

3
ψz = ψ210 = R21 (r) cos θ ,
4π

1 3
ψx = √ (ψ21−1 − ψ211 ) = R21 (r) sin θ cos ϕ ,
2 4π

i 3
ψy = √ (ψ21−1 + ψ211 ) = R21 (r) sin θ sin ϕ . (4.24)
2 4π
The angular parts of the three functions are identical to those that appear in the
spherical coordinates, x = r sin θ cos ϕ, y = r sin θ sin ϕ, z = r cos θ. The regions
of high probability of the electron’s density are thus centered around each of the
axes (for a given radial distance, the absolute value squared of each wave function
is maximal along one of the axes). Figure 4.11 illustrates the probability “cloud” of
the wave function ψx (denser points indicate higher probability to find the electron).
The two valence electrons of beryllium can be in the state ψ200 , or in any of the
three states (4.24).

Fig. 4.11: The density of points reflects the probability density to find the electron
in the ψx state. Only probabilities higher than a certain threshold are depicted.

Hybridization of the sp type. The arguments given above demonstrate that

the larger is the probability to find the two electrons which participate in a covalent
bond in-between the two nuclei, the higher is the binding energy. In the ψx state
the probability to find the electron on the x̂−axis is the same on both sides of
the nucleus; hence this is not the quintessential state. The probability √ density is
concentrated on one side of the nucleus in the state Ψ± = (ψ200 ± ψx )/ 2. Such
functions are “hybridized states of the sp type” (i.e., linear combinations of
= 0 states with = 1 ones). Since ψx has opposite signs on the two sides of
the x̂−axis, Ψ+ gives a larger probability on the positive side, while Ψ− gives the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 243

Crystal bonding 243

larger probability on the negative side. The two states are described on the right
side of Fig. 4.12. Their probability “clouds” are concentrated on the positive or
on the negative side of the x̂−axis, as dictated by the relative sign. When each of
the hybrid states Ψ± is occupied by a single electron, the average repulsion energy
is smaller than in any of the original states, because in the hybridized state the
electrons“reside” on the two sides of the nucleus, and hence are relatively far apart.
In the hybridized configuration, the beryllium atom can form covalent bonds
with two other atoms. For instance, it can build the linear molecule BeH2 , in which
the two hydrogens are on the x̂−axis on both sides of the beryllium. The electron of
the right hydrogen hybridizes its hydrogen state with the state Ψ+ of the beryllium,
and the one of the left hydrogen hybridizes with the state Ψ− (the coefficients of
the wave functions of the beryllium and the hydrogen in each molecular state are
not identical, see problem 4.9). Pictorially, this molecule is referred to as H−Be−H,
where each hyphen represents a covalent bond comprising two electrons, one from
each atom. The atoms in this molecule are located on a straight line (the angle
between the two bonds is 180◦ ), since bending the molecule brings closer the two
electron clouds (as well as the two protons) on the two sides of the beryllium, and
thus costs Coulomb energy.
In a similar fashion, a beryllium atom can form covalent bonds with two
beryllium atoms on its two sides, thus creating the infinitely-long linear molecule
. . .Be−Be−Be−Be−Be− . . . . This is a simple realization of a one-dimensional
periodic lattice, built of atoms connected by covalent bonds. The lattice con-
stant and the bonding energy can be estimated from a calculation of the average
variational energy for the (symmetric) molecular wave functions of the two electrons
[based on Ψ+ (rA ) in the left atom and Ψ− (rB ) in the right one]. The distance be-
tween the two atoms at which this energy is minimal corresponds to the lattice
constant.

Fig. 4.12: From left to right: ψ200 state, ψx state, and the hybrid states Ψ± .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 244

244 Introduction to Solid-State Physics

Covalent bonds of boron, hybridization of the sp2 type. The next atom
in the periodic table is boron (z = 5). Again it is “better” for the three valence elec-
trons to change the configuration 2s2 2p into 2s2p2 , to enable all three to participate
in covalent bonds. Assuming that the two p−electrons are in linear combinations of
the states ψx and ψy , a preferable situation is that of hybrid states, which tend to
separate the electrons’ clouds and thus to decrease the average energy of the electric
repulsion. These states, termed hybridization states of the sp2 type, are
√ √
Ψ1 = (ψ200 + 2ψx )/ 3 ,
√ √ √
Ψ2 = (ψ200 − 2[ψx − 3ψy ]/2)/ 3 ,
√ √ √
Ψ3 = (ψ200 − 2[ψx + 3ψy ]/2)/ 3 . (4.25)
These three states are occupied by the three electrons of the boron. The choice of
the coefficients is dictated by two considerations. First, the three functions should
be orthogonal to each other, i.e., Ψi |Ψj  ≡ drΨ∗i Ψj = δij as can be proven by the
orthonormality of the atomic functions (check, see also problem 4.9); second, the
probability clouds formed by them should be identical (up to rotations), as there is
no reason for one electron to be different than the others. It is straightforward to
show that the probability clouds of these three states, depicted in Fig. 4.13, attain
their largest values along three directions, all in the XY −plane, and are oriented
at an angle of 120◦ relative to each other. As the clouds are rather far apart, the
repulsion energy, in a quantum state in which there is a single electron in each state,
is relatively low.
Each such electron can be also bound covalently with an electron of a hydrogen
atom, to form the molecule BH3 . This is a planar molecule, symmetric under
rotations of order 3 around the nucleus of the boron (that is, the nuclei are all
in-plane; the electrons’ clouds can be also outside the plane). Each boron atom
has three bonds of this type, and therefore boron can create an infinite hexagonal
lattice, like that of graphene (Fig. 2.8); each atom in this lattice has three covalent
bonds with three neighbors, making an angle of 120◦ in-between each pair. This
type of covalent bonds characterizes numerous hexagonal lattices in Nature.

Fig. 4.13: From left to right: an s−state, p−states, and the hybrid states. Each
“cloud” in the right panel illustrates one of the states listed in Eq. (4.25).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 245

Crystal bonding 245

Covalent bonds of carbon, hybridization of the sp3 type. The next atom
in the periodic table is carbon; it participates in organic molecules, indispensable
for many biological functions. In order for all four valence electrons of carbon to
form covalent bonds, they must be in the configuration 2s2p3 . Then hybridization
among these states, where the probability clouds are distanced as far as possible
from each other, can be established. These states are called hybridization states
of the sp3 type,
Ψ111 = (ψ200 + ψx + ψy + ψz )/2 ,
Ψ111 = (ψ200 + ψx − ψy − ψz )/2 ,
Ψ111 = (ψ200 − ψx − ψy + ψz )/2 ,
Ψ111 = (ψ200 − ψx + ψy − ψz )/2 . (4.26)
The subscripts of each function represent the signs of the three p−states on the right
hand-side, with 1 denoting −1. The four probability clouds are directed towards
the four corners of a tetrahedron, shown by thick lines in Fig. 4.14 (see problem
4.9). The angle between the axes of each pair of clouds is 109.5◦ (check that this
is indeed the angle in the tetrahedron). Indeed, in the simplest organic molecule,
CH4 (methane), Fig. 4.15(a), the hydrogen atoms are at the vertices of such a
tetrahedron, with the electron of each one of them forming a covalent bond with
one of the carbon electrons that occupies one of the states of Eqs. (4.26). The ion
(NH4 )+ has the same structure, as the nitrogen ion possesses four valence electrons
like the carbon atom. As opposed, in the molecule C2 H6 (ethane) [shown in Fig.
4.15(b)] each carbon is connected with three hydrogens and with another carbon,
and thus the four bonds around each C are not identical. As a result, the angles
between the bonds, as well as their lengths, are different. One may compute the
angles by utilizing the variational approximation, with different coefficients for the
terms in Eqs. (4.26) (see also problem 4.10).
Problem 4.9.
Two hybridizations of the 2s state with the 2p ones are Ψ1,2 = A[ψ200 + β n̂1,2 · Ψ],
where n̂1,2 are two unit vectors in space, and Ψ = {ψx , ψy , ψz }.
a. Show that in the state Ψ1 , the probability to find the electron is centered around
the direction of the vector n̂1 .
b. Use the orthogonality of the original basis functions Eq. (4.24) to demonstrate
that ψi |ψj  = |A|2 (1 + |β|2 n̂i · n̂j ).
c. Utilize the result in part (b) to derive A and β when the hybrid states are
orthonormal, and when the angle between the two unit vectors is α.
d. The hybridization of the sp2 type corresponds to three states, with three unit
vectors in the plane, n̂1 , n̂2 , and n̂3 . Explain why the angles in-between these unit
vectors have to be identical, and derive Eq. (4.25).
e. Derive in a similar way Eq. (4.26).
f. In a planar molecule like BH2 F the bonds are not entirely symmetric: the angle
between the bonds of the boron with each of the hydrogens is γ, while the other two
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 246

246 Introduction to Solid-State Physics

Fig. 4.14: The probability clouds of the four valence electrons of carbon (at the
center of the cube), for the wave functions in Eq. (4.26). The probability clouds
point at the corners of a tetrahedron (shown by thick lines). Alternatively, the
probability clouds are directed at the opposite corners of the cube: two vertices are
at the ends of one of the face diagonals, and the other two are at the ends of the
diagonal on the opposite face, which is normal to the first one.

angles (in-between each BH bond and BF one) are α = π − γ/2. Determine the
hybrid functions that comply with this geometry.
g. Given the angle α = 111.17◦ between the CC bond and the CH one in ethane
[Fig. 4.15(b)], what is the angle γ between the two CH bonds? What can be deduced
about the coefficients in the corresponding hybrid functions?

(a) (b)

Fig. 4.15: The methane molecule (a) and the ethane molecule (b). The bonds’
lengths are measured in picometers (10−2 Å).

π−bonds, double and triple bonds. The ethene (C2 H4 ) molecule is illus-
trated in Fig. 4.16(b). Around each carbon there are two hydrogens; this carbon
is also coupled to the other carbon. These three bonds are described by the planar
wave functions of Eq. (4.25), and are located on the same plane. Nonetheless,
due to the difference between carbon and hydrogen, the three angles between the
bonds are not precisely the same, and the coefficients in Eqs. (4.25) are slightly
modified [see part (g) of problem 4.9]. Since carbon has four valence electrons,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 247

Crystal bonding 247

there remains in each carbon one electron, in the state ψz , whose probability cloud
is normal to the plane. The hybridization of the functions ψz on the two carbons,
Ψ = A[ψzA (rA ) + ψzB (rB )], creates an additional bond in-between them, as de-
scribed in Fig. 4.16(a). This bond is weaker than the in-plane ones, as the overlap
between the wave functions oriented normal to the plane is smaller than the one of
two wave functions concentrated on the same axis. This bond is termed π−bond;
the one in which the electron clouds are concentrated on the line connecting the
nuclei is termed σ−bond. In the C2 H4 molecule there is therefore a double bond
between the carbons, comprising the two bonds. The double bond, marked by =,
is stronger than a single bond, as it is due to four electrons, two from each atom.
This is illustrated in Fig. 4.16(b). Note that without the π−bond it is possible to
rotate one of the planes containing one carbon with the two hydrogens connected
to it, such that the two planes will be perpendicular to one another. This will
shift the hydrogens away from those of the second carbon; however, the overlap
between the ψz functions will diminish, weakening the π−bond between the car-
bons. This is the reason for the hydrogens in the ethene molecule to be located
on the same plane. Similarly, in acetylene (C2 H2 ) there is a triple bond be-
tween the two carbons,
√ comprising one σ− bond (between the two linear functions
Ψ± = (ψ200 ± ψx )/ 2) and two π−bonds, consisting of the functions normal to this
line (ψy and ψz ). This molecule is denoted H−C≡C−H.

(a) (b)

Fig. 4.16: (a) The π−bond between two p states for which the probability clouds are
normal to the bond direction. The clouds on the right result from a combination of
two atomic functions of the π type, shown separately on the left of that panel. (b)
The planar molecule ethene, C2 H4 , with the double bond in-between the carbons.

Benzene, the metallic bond. In the benzene molecule, C6 H6 , mentioned in

Sec. 4.1, each carbon is located on a corner of a hexagon, and is connected to a
hydrogen atom (see Fig. 4.2). Three of the carbon electrons are in the plane of
the hexagon, in wave functions similar to those in Eq. (4.25), with different angles
for the different types of bonds: one electron participates in a covalent bond with
the hydrogen, and the other two form σ−bonds with the neighboring carbons. The
fourth electron is in the ψz state, with a probability cloud normal to the plane. By
the Heisenberg uncertainty principle, the kinetic energy is decreased as the clouds’
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 248

248 Introduction to Solid-State Physics

spatial extension is widened. (Increasing the uncertainty in the location reduces

that of the momentum. When the average momentum vanishes, then the mean-
square deviation equals the square of the momentum, and hence it is proportional
to the kinetic energy. As the latter is reduced, the bonding energy is enhanced.)
The uncertainty in the locations of the electrons is largest when their wave functions
are spanned over the entire hexagon; extending Eq. (4.14),
 6
Ψ= a ψ (r ) , (4.27)
=1
where the coefficients are chosen so that the wave function is normalized; r is the
distance of the electron from the carbon, and ψ is the ψz function of that carbon.
The six ψz functions accommodate each a single electron and are orthonormal. As in
the case of the hydrogen ion H+2 , symmetry considerations imply that the probability
to find the electron on any of the carbons is the same, and hence |a |2 ≡ A2 and
is independent of . The complex coefficients differ from each other only in their
phases (see Chapter 6). The covalent bonds between the perpendicular states, like
in ethene (Fig. 4.16), are more effective when the probability clouds of these states
are parallel, and therefore the other bonds around each carbon are in the same
plane, as seen in Fig. 4.2. Since the probability to find the electron on any of the
carbons is the same, these electrons are free to move around the hexagonal ring
once subjected to external fields. For example, a uniform magnetic field causes
the electrons to move around the ring, and to create a magnetic moment normal
to the plane. The wave functions (4.27) contribute as well to the binding energy.
This configuration resembles metals: in Chapter 6 it is found that in a metal each
electron has the same probability to be near any of the ions in the crystal, and wave
functions similar to those of Eq. (4.27) are the basis for the metallic bond. Thus,
the benzene molecule is a simple realization of a metallic crystal.
Covalent bonds of nitrogen, oxygen, and fluorine. The n = 2 electronic
shell contains four orbital states, one s and three p. Various hybridizations of
these four states can be designed in order to create covalent bonds. The atoms
in the second row of the periodic table possess 1-8 electrons. Up to carbon, each
of the four states can be singly-occupied (or be empty), and each such electron
can participate in a covalent bond. The next atom is nitrogen, which has five
electrons in this shell. Consequently, two of these share the same orbital state, and
therefore, by the Pauli principle that forbids the addition of another electron to this
state, cannot participate in covalent bonds. Nitrogen has thus three free electrons,
that may be in any of the hybridizations listed above. Similarly, oxygen has two
doubly-occupied states and only two electrons that are free to form bonds. Fluorine
possesses three states occupied by two electrons, and hence only a single electron
that can be bonded. The configurations of the corresponding electronic states are
shown in Fig. 4.17.
The water molecule. The water molecule is obviously of paramount impor-
tance. It is built of an oxygen atom and two hydrogens, H2 O. In the second shell
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 249

Crystal bonding 249

Fig. 4.17: From left to right: probability clouds to find the electrons in the second
shell of carbon, nitrogen, oxygen, and fluorine. The bigger clouds contain two
electrons each. The smaller ones accommodate a single electron, “free” to form a
covalent bond.

of oxygen there are two orbital states which are doubly occupied, so only two of
the electrons are available for forming bonds. In order to build the water molecule,
it is expedient to begin with the four electronic states of Eq. (4.26), which create
separate clouds (as far as possible from each other) as to reduce the Coulomb in-
teraction among the electrons. Two of these states contain two electrons each (see
Fig. 4.17), and the other two are coupled with the hydrogens as depicted in Fig.
4.18. The angle between the bonds with the hydrogens is 104.5◦ , slightly smaller
than that of the tetrahedron of the symmetric case in Fig. 4.14 (see problem 4.10).
This difference stems from the repulsion between the electrons taking part in the
bonds with the hydrogens, and the other electrons of the oxygen, concentrated
on the other side of this atom. Technically, the difference manifests itself in slight
modifications of the coefficients of the base functions in Eq. (4.26). A similar differ-
ence is encountered in the ethane molecule, Fig. 4.15. As a result of this deviation
from complete symmetry, the water molecule has an electric dipole moment.

Problem 4.10.
Describe the geometry of the electrons’ clouds in the molecules N2 , O2 , F2 , NH3 ,
and FH. Which of those possesses an electric dipole moment?

Fig. 4.18: The water molecule. The darker clouds represent the two doubly-occupied
states of oxygen (indicated by the two black points). The angle between the OH
bonds is 104.5◦ .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 250

250 Introduction to Solid-State Physics

Giant carbon molecules. The three-dimensional crystals of carbon are di-

amond and graphite. Carbon, as mentioned in Sec. 1.2, forms also very large
molecules, which have become an important topic of research in recent years. Ex-
amples are C60 [the Bucky ball, Fig. 1.9(b)] and carbon nanotubes [Fig. 1.9(a)].
Though quite large, these molecules are nonetheless finite; they are considered as
being of zero dimension. The nanotubes can be extremely long, and then they are
viewed as one-dimensional lattices (of finite thicknesses in the directions normal
to the lattice). Each carbon in these molecules has three neighbors, and therefore
the bonds are approximately of the sp2 −type. They are slightly distorted because
of the deviation of the sphere or the tube from perfect planarity. Because of this
deviation, the functions in Eq. (4.25) are augmented by a contribution from the
wave function ψz , normal to the tube (or to the sphere), and then the electrons’
clouds are not exactly in the same plane (see also problem 4.10). The wave func-
tions of the surplus electrons may be spanned over the whole molecule, similar to
Eq. (4.27).
Carbyne. The four valence electrons of carbon can be in any of the hybrid
states described above, according to the geometric configuration of the covalent
bonds at hand. Several allotropes of carbon, in various spatial dimensions, are
mentioned in Chapter 1. Each such lattice requires different electronic states on
the carbon. The simplest example of a one-dimensional carbon lattice is a
linear chain of carbons, called carbyne. The chain may take one of two forms,
cumulene, . . .C=C=C=C=C. . . , with a double bond between two nearest neighbors,
and polyyne, . . .C−C≡C−C≡C. . . , in which each carbon has a single bond on one
side and a triple one on the other. As explained in the context of beryllium, the
σ−bonds between nearest-neighbor carbons cause the bonds to be on the same
straight line.
Graphene. In two dimensions carbon creates the hexagonal planar lattice
of graphene, Fig. 1.6 or Fig. 2.8. Similar to the bonds of boron, three of the carbon
electrons are in the hybridized state of the type sp2 , Eq. (4.25). Each of these
electrons forms a σ−bond with the neighboring carbon, and all these bonds are in
the same plane. The remaining electron, as in benzene, is in a ψz state, normal to
this plane. In order to reduce the kinetic energy, then according to the uncertainty
principle of Heisenberg, these extra electrons occupy linear combinations of the ψz
states spread on all carbons in the lattice, e.g., Eq. (4.27), with |a |2 = A2 . These
π−states contribute to the binding energy and are also responsible for the high
electrical conductance and the metallic behavior of this material. The graphene
lattice is planar; this is because the charge clouds of the π−states are parallel to
each other. Such electronic states are discussed in Chapter 6.
Boron nitride, silicene and hexagonal germanium. The hexagonal lattice
of graphene appears in other materials as well, e.g., boron nitride (BN). In this
compound the boron and the nitrogen atoms occupy alternating sites of a hexagonal
lattice, like the large and small points in Fig. 2.8. The electronic configurations of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 251

Crystal bonding 251

the borons and the nitrogens are 2s2 2p1 and 2s2 2p3 , respectively; however, when
in a bulk form, one electron is transferred from the nitrogen to the boron and the
two ions attain the 2s1 2p3 configuration, as in carbon. In this way bonds of the
sp2 −type are formed, like in graphene, with the ψz −electrons common to all ions in
the lattice. Because of the transferred electron, there are excess charges on the ions,
and the resulting bonds comprise both covalent and ionic ones. To compute the
binding energy of BN one utilizes hybridizations of the two types of wave functions,
cf. the discussion following Eq. (4.23). Other examples of hexagonal lattices
are silicene and germanene, Fig. 2.11. There, similar to BN, there are two
sublattices which in this case are located on different planes. In the structure of
Fig. 2.11 the electronic wave functions around each ion include also a component
of ψz , in addition to the functions of Eq. (4.25). The deformation of the planar
lattice stabilizes it mechanically (see also Chapter 5).
Graphite. Graphite is built of planar layers of graphene (see Fig. 2.11). The
attraction between the layers is mostly due to the van der Waals interaction, which
is related to the fluctuating dipole moments of the electrons in each carbon (see
below). The inter-plane attraction is thus relatively weak and allows for the peeling
of planar layers off the bulk graphite (this is the reason why graphite is used in
pencils). Such a peeling (accomplished with a piece of celo-tape) enabled Geim and
Novoselov to obtain single layers of graphene. The bond between atoms in each
layer in graphite (and in graphene) is stronger even than that in diamond. This is
the reason for graphite to be the preferred three-dimensional phase of carbon under
ambient conditions. Diamonds are formed only under high pressures. Graphite
is created upon cooling liquid carbon under atmospheric pressure below 4000C,
while it is only when the liquid is cooled at pressures of about half a millions atms
that diamond is created. The reason is related to the dependence of the binding
energy on the distance between the carbons’ nuclei in the crystal. Under pressure,
the nuclei approach each other, and at a certain distance the bonding energy of
diamond becomes smaller than that of graphite. Indeed, natural diamonds are
found in places which were subjected to high pressures (due to certain geological
phenomena); they are thus rather rare (and hence expensive).
The diamond crystal. The four directions of the probability clouds shown in
Fig. 4.14 are identical to those between nearest-neighbor carbons in the diamond
lattice. The cube in that figure is 1/8 of the unit cell of diamond (see Fig. 2.20). In
this structure, each carbon contributes four electrons that form four covalent bonds
of the sp3 −type with the four neighbors; each carbon is located at the center of a
tetrahedron built of its four neighbors. This structure is very strong and not easily
distorted, as any deformation brings together the electrons’ clouds on the various
bonds and requires considerable Coulomb energy.
Diamond and zinc-blende structures. Silicon and germanium, located at
the same column of the periodic table as carbon, have the crystal structure of
diamond. Interestingly enough, AlAs, GaAs, and InAs have the very similar zinc-
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 252

252 Introduction to Solid-State Physics

blende structure (Fig. 2.20). The valence-electron configuration in aluminium,

gallium, and indium is ns2 np1 (n = 3, 4, and 5 for Al, Ga, and In, respectively)
and that of arsenic is 4s2 4p3 . In the bulk one electron of As is transferred to the
other atom of the compound, and then the two ions attain the ns1 np3 configuration,
which forms bonds of the sp3 -type. The resulting bond again comprises ionic and
covalent bonds. GaAs is vastly exploited in electronic devices.
In summary, the simple geometrical picture of the various hybridizations of the
p electrons explains the experimentally-determined structures (Chapters 2 and 3)
of materials based on the elements in the fourth column in the periodic table, and
of materials comprising elements from the third and fifth columns.
Covalent bonds with d electrons. Downwards from the third row of the
periodic table, electrons can fill also the 3d shell, where their angular momentum is
= 2. The angular parts of the wave functions of these electrons are
ψ322 − ψ32−2 ψ +ψ
ψz2 = ψ320 ∝ 3z 2 − r2 , ψxy = √ ∝ xy , ψyz = 321 √ 32−1 ∝ yz ,
2 2
ψ321 − ψ32−1 ψ322 + ψ32−2
ψzx = √ ∝ zx , ψx2 −y2 = √ ∝ x2 − y 2 . (4.28)
2 2
The probability clouds of these states attain their maximal density at angles 90◦
or 180◦ with each other. [For instance, the probability density of the function
proportional to xy is (xy)2 ∝ (sin2 θ cos φ sin φ)2 ; it is thus maximal for θ = π/2
and φ = ±π/4, ±3π/4, i.e., when the vector r = r(sin θ cos φ, sin θ sin φ, cos θ)
is along the diagonals in the XY −plane. The probability cloud proportional to
3z 2 − r2 is concentrated on the ẑ−axis.] Hence, bonds based on these functions
build up molecules with right angles between the bonds, or lattices with orthogonal
base vectors.
Ions taking part in five covalent bonds often form bonds of the sp3 d-type, in
which a single s−function combines with three p ones and a single d state. This
combination leads to five functions whose peaks point at the corners of a trigonal bi-
pyramid, along directions forming an angle of 120◦ in the plane (like in graphene) or
along the normal to the plane, Fig. 4.19(a). In the PF5 molecule, the configuration
3s2 3p3 of phosphor changes into 3s1 3p3 3d1 , such that each of the five states can
contain two electrons. These five wave functions have trigonal symmetry, where the
angles between their probability clouds are 120◦ , 90◦ , or 180◦ . These clouds can
therefore build e.g., hexagonal lattices.
Likewise, ions with six covalent bonds form them in the sp3 d2 -type hybridiza-
tion, in which two d states are involved. The six functions so created point at
the corners of an octahedron, as in Fig. 4.19(b). This type of bond exists, e.g., in
the SF6 molecule, where the 3s2 3p4 configuration of the sulphur atom changes into
3s1 3p3 3d2 , such that each of the states can accommodate two electrons and thus
can participate in a covalent bond. These states have octahedral symmetry, and
form lattices of this symmetry.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 253

Crystal bonding 253

(a) (b)

Fig. 4.19: (a) The five fluorines in the molecule PF5 , with bonds of the sp3 d−type;
the phosphor is at the center of the planar triangle. (b) The six fluorines in SF6 ,
with bonds of the sp3 d2 −type; the sulphur is at the center of the square.

Problem 4.11.
a. Propose two linear combinations of the functions ψz and ψz2 that can serve as
the wave functions of electrons which take part in the covalent bonds normal to the
plane, as those illustrated in Figs. 4.19(a) and (b).
b. Propose five functions, appropriate for the electrons in the sp3 d bonds as shown
in Fig. 4.19(a).
c. In addition to the functions found in part (a), show that the wave functions of
the four electrons participating in the planar bonds of Fig. 4.19(b) can be built from
the linear combinations
Ψ = A(ψ200 + βx ψx + βy ψy + βxy ψxy ) .
Propose numerical values for the coefficients of this linear combination.
Crystals with transition metals. There are many crystals in which the
electrons that take part in the bonds are only in the d−shell, e.g., crystals which
contain transition metals like chromium, manganese, iron, cobalt, nickel, copper,
and zinc; the d−shell is partially occupied in these elements. For example, in the
lanthanum-cuprate crystal (see Fig. 2.21) the copper ions have nine electrons in the
3d−shell. One of those participates in the bonds with the nearest neighbors. Figure
4.20 illustrates the electrons’ charge clouds in the Ti and O ions in the perovskite
LaTiO3 . In this crystal, the Ti ion has a single valence electron in the 3d−shell,
and the figure displays the configuration in which the wave function of this electron
is ψxy [Eq. (4.28)]. The electrons of the oxygens are in the states ψx , ψy , or ψz .
In this configuration, nearest-neighbor Ti ions in the XY −plane and the oxygen
located in-between them are hybridized. In this way the square structure of the
plane can be stabilized. The bonds between the planes along the normal direction
are formed by hybridization of the wave functions ψxz or ψyz .
Hybridization between ionic and covalent bonds. The ionic bond and the
covalent one are the two extremal possibilities for the bonding of two ions in the
Heitler-London approximation. The more ubiquitous bond is a combination of the
two. For example, the zinc-blende and wurtzite crystals can be considered as ionic
crystals. The sulphur ion “acquires” two electrons from the zinc; the occupations
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 254

254 Introduction to Solid-State Physics

Fig. 4.20: The wave functions of the Ti ions (ψxy ) and the oxygens in the
XY −planes of the perovskite crystal LaTiO3 . The darker and lighter areas of
the charge clouds indicate positive and negative functions, respectively. [A. B. Har-
ris, T. Yildirim, A. Aharony, O. Entin-Wohlman, and I. Ya., Korenblit, Unusual
symmetries in the Kugel-Khomskii Hamiltonian, Phys. Rev. Lett. 91, 087206
(2003).]

of the lattice sites then alternate between charges of +2 or −2 each. On the other
hand, since the electronic configuration of sulphur resembles that of carbon, the
former can participate in four covalent bonds. Indeed, the coordination of the
sulphur ion in both lattices, where each sulphur is surrounded by four zinc ions at
the corners of tetrahedra, indicates that these bonds should be of the sp3 −type. As
each zinc ion is surrounded by four sulphur ions, it is plausible that the zinc has
also four covalent bonds of the same type. Glancing at the periodic table shows
that the basic configuration of zinc includes 3d10 4s2 . Exciting three electrons into
the configuration 3d8 4s1 4p3 enables this ion to form sp3 bonds. The result is a
lattice similar to that of diamond, where all valence electrons are distributed over
the covalent bonds between the zinc ions and the sulphur ones. This picture can
be backed by, e.g., a variational calculation which includes both the ionic and the
covalent descriptions, or by a numerical solution of the Schrödinger equation for the
electrons. Such a mixture of ionic and covalent bonds appears also in more complex
crystals, like the perovskite or the cuprate structures (Fig. 2.21). In many cases
the bonds include also d−states, like in Eq. (4.28), which reinforces the square or
tetragonal symmetry of these crystals.

4.4 Molecular crystals–van der Waals bond

Induced fluctuating dipoles. The bonds described so far have all their origin in
the electric attraction between pairs of opposite charges. “Secondary” bonds exist
in crystals built of neutral units, e.g., noble gases or neutral molecules. The
first part of this section presents calculations that lead to the attraction force of van
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 255

Crystal bonding 255

der Waals. As mentioned in Sec. 4.1, the coupling between two noble atoms results
from their mutual polarization; the electrons’ clouds in each atom are displaced
relatively to the nucleus, such that each atom acquires a fluctuating electric dipole.
When the moment on one atom is p1 = Qr1 (Q is the charge of the nucleus and
r1 is the distance in-between the centers of the positive and negative charges) then
at a distance R away there is an electric field, E ∝ [(p1 · R̂)R̂ − p1 ]/R3 (where
R̂ ≡ R/R is a unit vector along R). This field exerts force on the electrons of
the other atom, that causes them to move a little and create a dipole moment,
p2 = αE, where α is the polarization coefficient. The magnitude of this moment is
p2 ∼ p1 /R3 . Between the two dipoles operates a force, whose electrostatic energy
is ∝ −p2 p1 /R3 ; inserting here p2 gives an energy proportional to −p21 /R6 . As
the average of p21 does not vanish (though the average of p1 does), there arises an
energetic gain, proportional to R−6 . This gain is the origin of the van de Waals
attraction between the two atoms.

Problem 4.12.
Classically (i.e., at temperatures higher than all energies of the problem), the van der
Waals interaction can be derived from a thermodynamic average over the interaction
between two classical dipoles, that may point along any direction in space. The
energy of this dipole-dipole interaction is U (R) = (p1 ·p2 )/R3 −3(p1 ·R)(p2 ·R)/R5 ,
where R is the distance between the dipoles, and the average is carried out with the
Boltzmann weight exp[−βU ], where β = 1/(kB T ). This means that U (R) =
dΩ1 dΩ2 U (R) exp[−βU ]/ dΩ1 dΩ2 exp[−βU ], where the integrals are performed
over all possible directions of the dipole moments. Show that at large distances
U (R) ≈ −4βp21 p22 /(3R6 ).

Quantum calculation of van der Waals force. The phenomenon of induced

fluctuating dipole moments may be exemplified by considering two hydrogen atoms.
Though hydrogen is not a noble gas, it has a single electron that can create a dipole
moment. Figure 4.21 illustrates the coordinates of the two protons and the two
electrons. When the nuclei are close enough to one another they create the molecule
H2 , discussed in Sec. 4.3. For the present discussion it is assumed that the distance
R between the nuclei is far longer than the Bohr radius (which is roughly the size
of the atoms) so that there is almost no overlap of the electrons’ clouds, and each
electron is near its “own” nucleus. Similarly to the discussion in Sec. 4.3, the
energy of the two atoms as a function of R is found within the Born-Oppenheimer
approximation (one may then view the result as an effective potential between the
two nuclei).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 256

256 Introduction to Solid-State Physics

Fig. 4.21: Two hydrogen atoms whose nuclei are located at A and B, and their
electrons at a and b.

Given the locations of the two nuclei, the Hamiltonian of the electrons is Ĥ + Ĥ ,
where
2 2 e2 2 2 e2
Ĥ = − ∇A − − ∇B − ,
2m rA 2m rB
1 1 1 1 
Ĥ = e2 + − − , (4.29)
R |R + rB − rA | |R + rB | |R − rA |
(in CGS units; in the SI ones the e2 terms in the Coulomb interaction are divided by
4π0 ). The Hamiltonian Ĥ describes two isolated hydrogens, whereas Ĥ pertains
to the Coulomb interactions between the charges in these atoms. This interaction
diminishes as the distance R becomes longer, and hence at large distances it is a
small perturbation. When this perturbation is ignored, the electronic wave function
in the ground state of Ĥ is Ψ0 (rA , rB ) = ψ100 (rA )ψ100 (rB ), with the energy 2E1 =
−2Ry = −27.2eV, where ψnm (r) is the eigenfunction of the hydrogen atom with
the quantum numbers n, , and m, and energy En = −[13.6/n2 ]eV.
To account for the effect of Ĥ , one takes ẑ to be the direction of R, and expands
the electrostatic interaction in rA and rB . To leading order, the result is
e2  
Ĥ = 3 rA · rB − 3zA zB + . . . . (4.30)
R
The notation . . . indicates higher-order terms, with higher powers of R in the
denominator. As erA and erB are the dipole moments of the two atoms, Eq.
(4.30) is the electric interaction between two dipoles. Within the variational
approximation, the ground-state energy is augmented by the expectation value
of Ĥ in the trial function. With the trial wave function of the ground state,
i.e., Ψ0 (rA , rB ) = ψ100 (rA )ψ100 (rB ), the second term in Eq. (4.30) contains
ψ100 |zA |ψ100 . Since the wave function in the ground state of the hydrogen atom
is symmetric with respect to reflection, this integral vanishes. For the same rea-
son the first term does not contribute as well. To modify the ground-state energy
(away from the one that pertains to the wave function Ψ0 ) one adds to the trial
wave function excited states of each atom, e.g., Ψ(rA , rB ) = A(Ψ0 + cΨ1 ), where
Ψ1 = ψ210 (rA )ψ210 (rB ), with a small complex coefficient c. Expanding the expec-
tation value of the energy in this state in c gives
E = 2E1 + cΨ0 |Ĥ |Ψ1  + c∗ Ψ1 |Ĥ |Ψ0  + 2c∗ c(E2 − E1 ) + . . . , (4.31)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 257

Crystal bonding 257

where the normalization condition, Ψ|Ψ = |A|2 (1 + |c|2 ), implies |A|2 ≈ 1 − |c|2 .
The notation . . . again indicates higher-order terms; the other terms vanish (see
problem 4.14). Treating c and c∗ as two independent variables, the expectation
value is minimized with respect to both. Differentiating Eq. (4.31) then yields
2c∗ (E2 − E1 ) + Ψ0 |Ĥ |Ψ1  = 0 , 2c(E2 − E1 ) + Ψ1 |Ĥ |Ψ0  = 0 .
It follows that c is proportional to R−3 , and the correction to the energy is propor-
tional to R−6 . The minimal variational energy is
|Ψ1 |Ĥ |Ψ0 |2 4e4 |ψ100 (r)|z|ψ210 (r)|4 e 2  a B 6
E ≈ 2E1 − = 2E1 − ≡ 2E − ζ ,
2(E2 − E1 ) 2R6 (E2 − E1 ) 1
aB R
(4.32)
where ζ = O(1) is a dimensionless constant, determined by the wave functions and
the energies of the hydrogen atom. This constant is calculated in problem 4.13, for
both the trial wave function proposed above, and for more general trial functions.
The correction in Eq. (4.32) is negative, showing that the trial wave function yields
a lower energy. This correction is the van der Waals attractive potential energy
between the two atoms.

Problem 4.13.
a. Prove Eq. (4.31).
b. Calculate the coefficient ζ of Eq. (4.32).
c. Extend the calculation above by adding more excited states to the variational trial
wave function of the ground state. Find the value of ζ in this case.

Problem 4.14.
Prove the expansion in Eq. (4.30). Determine the next correction to the Hamilto-
nian when this expansion is extended to higher-order terms. What is the order of
magnitude of this correction to the energy in Eq. (4.32)?

It is worthwhile to note that the discussion above (for both the leading order
and its correction as analyzed in problem 4.13) is valid as long as the ground-state
wave function is spherically symmetric. This is true for the noble gases, but not so
for many molecules, see Sec. 4.3 and the end of this section.
Pauli’s repulsion and the Lennard-Jones potential. The calculation car-
ried out above is valid for R longer than the atomic radius (i.e., aB ). When the
distance R is shorter, the expansion in Eq. (4.30) is not valid anymore; a solution of
the complete Hamiltonian, Eq. (4.29), is required. The resulting energy is positive
at short distances; this can be interpreted as a repulsive potential. Such a repulsion
follows from the Pauli principle, which prevents two atoms to approach one another
too closely. Instead of carrying out this calculation, and similar to the discussion
of the ionic bond in Sec. 4.2, the effective energy which represents the interaction
between the two atoms is augmented by an effective repulsion term, that increases
sharply as R decreases. Experiment shows that the detailed form of this term is
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 258

258 Introduction to Solid-State Physics

not essential (see the examples in Sec. 4.2). It is customary to utilize the potential
UP 1 in Eq. (4.8), with m = 12. The combination of this repulsion with the van der
Waals potential is the effective Lennard-Jones potential,
 σ 12  σ 6
U (R) = 4 − . (4.33)
R R
This potential is depicted in Fig. 1.1. The parameters  and σ [equivalent to the
original coefficients in Eqs. (4.8) and (4.32)] set the energy and the length scales.
They are chosen such that at R = σ the effective potential vanishes. These param-
eters can be estimated from measurements of the thermodynamic properties of the
gaseous and the liquid phases of the material (e.g., from the van der Waals equa-
tion of state). For the noble gases xenon, krypton, argon and neon, the values are
 = 0.0031, 0.0104, 0.0140, 0.0200 eV and σ = 2.74, 3.40, 3.65, 3.98 Å, respectively.
These numbers are slightly modified in the solid phase.
The bonding energy of the crystal. Similarly to Eq. (4.1), the potential of
the crystal consists of
1  σ 12  σ 6
Utot = U (Rij ) = 2N  A12 − A6 , (4.34)
2 R01 R01
i=j

where R01 is the distance between nearest neighbors, and the coefficients An are
given by the lattice sums
 1
An = , (4.35)
pn0j
j =0

with R0j = p0j R01 . For the main cubic crystals

AFCC
n = 12 + 6/2n/2 + 12/2n + . . . ,
ABCC
n = 8 + 6(3/4)n/2 + 12(3/8)n/2 + . . . ,
ASC
n = 6 + 12/2
n/2
+ 8/3n/2 + 6/2n + 24/5n/2 + . . . . (4.36)
These sums converge relatively fast (check!). For large enough n they are almost
entirely dominated by the first term (which results from the interactions of nearest
neighbors, for which p0j = 1), with a small correction due to the interactions with
the next neighbors. In this limit An → znn + znnn /pn02 , where znn and znnn are the
coordination numbers (for nearest and next-nearest neighbors), and where p02 R01
is the distance to the next-nearest neighbors. Extended summations over numerous
terms yield
AFCC
12 = 12.13188 , AFCC
6 = 14.45392 ,
AHCP
12 = 12.13229 , AHCP
6 = 14.45489 ,
ABCC
12 = 9.11418 , ABCC
6 = 12.25330 ,
ASC
12 = 6.20 , ASC
6 = 8.40 . (4.37)
The minimal value of Utot of Eq. (4.34) is reached for R/σ = (2A12 /A6 )1/6 . For the
four lattices mentioned above, R/σ ≈ 1.090, 1, 090, 1.068, 1.067; the corresponding
binding energies are u/ = −Utot (R)/(N ) = A26 /(2A12 ) ≈ 8.610, 8.611, 8.237, 5.69.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 259

Crystal bonding 259

The lattice structure of noble gases. Since the sums in Eq. (4.35) are
dominated by the number of nearest neighbors, it is not surprising that the maximal
binding energy is the one of the FCC and HCP lattices. These two lattices are
the densest ones, with 12 nearest neighbors. The contributions from the third
neighbor onwards differ in the two lattices, and give rise to a minute, 0.01%, energy
gain, in favor of HCP. Nonetheless, the noble gases are usually arranged in the
FCC structure. Apparently, the final structure is sensitive to small corrections in
the binding energy, that are ignored in the considerations above. For instance, at
finite temperatures the entropy has to be accounted for, in addition to the energy
(i.e, contributions due to the number of degenerate states at the same energy, and
to the Boltzmann weight-factors of excited states). It seems that the free energy
(comprising the energy and the entropy) of the FCC structure is the lower one.
Another energy ignored in the considerations above is the kinetic energy; this is
due to the use of the Born-Oppenheimer approximation, which is based on the
assumption that the nuclei are immobile at the lattice sites (where the the potential
energy is minimal). The oscillations of the nuclei around these minima are discussed
in Chapter 5. These oscillations exist even in the quantum ground state, similar
to the zero-point motion of harmonic oscillators. The zero-point energy of the four
noble gases alluded to above is relatively small; it is however significant enough to
favor the FCC structure over the HCP one. (The full explanation necessitates an
exact calculation.)
This argument breaks down in the case of helium: the small mass of the helium
atom causes the energy of the zero-point motion to be relatively large, and leads to
a quantum instability of the helium lattice. The Lindemann criterion (Sec. 3.10)
examines the stability of the solid phase by comparing the average amplitude of the
zero-point vibrations with the lattice constant. At low temperatures (and within the
approximations used in Sec. 3.10) this criterion predicts that the crystal is unstable
when u2R  = 3/(2M ω) > (cL a)2 , with cL = 0.1. Helium obeys this inequality.
Under ambient pressure conditions, helium remains in the liquid phase down to zero
temperature. Quantum effects, related to the fact that atoms of helium 4 are bosons
(that is, they obey the Bose-Einstein statistics) turn this liquid at low temperatures
into a superfluid, i.e., a liquid with no viscosity. High pressures (more than 25
atms) stabilize a solid phase of helium at temperatures of about 1◦ K, and then it
is arranged in an HCP lattice. At much higher pressures (more than 1000 atms)
there is a narrow temperature window (below the melting temperature) in which
helium is an FCC lattice.
The disparity between the HCP structure, formed of ABABA. . . layers of tri-
angular lattice, and the FCC one, where the layers take the ABCABC. . . structure
(Fig. 2.24) is still being explored. The paper cited in the caption of Fig. 4.22
points out that the preference of FCC structure stems from the growth procedure
of close-packed spheres. The authors followed the absorption dynamics of addi-
tional spheres on a dense planar layer of (other) spheres. As expected, the spheres
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 260

260 Introduction to Solid-State Physics

sink into the dips in-between the ones in the previous layer. Figure 4.22 displays
the pyramids formed by the spheres in the third and fourth layers, for each of the
structures. The researchers found that the right pyramid is less stable than the
left one, and the absorption of the upmost sphere frequently causes the runaway of
spheres below (along the direction of the thick arrow in the figure). This does not
happen in the left pyramid (there the arrow indicates the direction where runaway
is the hardest). This leads to the preference of the FCC structure. It is somewhat
surprising that after more than a hundred years of research, there are still open
questions of this type. As mentioned, most of the noble gases are arranged on an
FCC lattice, for which R = 1.09σ and u = 8.6. These values result in a rather
small binding energy, of the order of 0.03−0.17 eV per unit cell.
Problem 4.15.
Find the bulk modulus of an FCC structure dominated by van der Waals interac-
tions.

Fig. 4.22: Two pyramids formed during a dynamical growth of close-packed spheres.
[S. Heitkam, W. Drenckhan, and J. Fröhlich, Stability on Close-Packed Sphere Struc-
tures, Phys. Rev. Lett. 108, 148302 (2012).]

Van der Waals lattices of molecules. The discussion above pertains also
for spherically-symmetric molecules. However, most of the molecules in Nature
lack this symmetry. For example, the H2 molecule, Sec. 4.3, is linear, with a charge
distribution as displayed in Fig. 4.9 (|ψ+ |2 there). Such linear molecules have a
rotation symmetry around the molecule axis, and can be viewed as ellipsoids, as in
the mid panel of Fig. 4.1. The ellipsoidal geometry introduces two modifications
into the previous discussion. First, the close-packed arrangement of ellipsoids is not
necessarily identical to that of spheres; recall that for a spherical symmetry, the van
der Waals interaction prefers close-packing of spheres like in the FCC lattice. Sec-
ond, each ellipsoidal molecule has a quadruple moment, proportional to the spatial
average of 3(r· n̂)2 −r2 , where n̂ is a unit vector along the molecule axis (see problem
4.16). Consequently, there exists a quadrupole-quadrupole interaction between any
pair of molecules. As found in that problem, the energy of this interaction decays
like R−5 , and depends on the angles between the two axes of the molecules and the
line joining them.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 261

Crystal bonding 261

Problem 4.16.
a. Prove that the quadrupole-moment tensor of a molecule symmetric with respect
to rotations around ẑ is diagonal, with the components Qxx = Qyy = −Qzz /2 =
i qi (ri − 3zi )/2, where the sum runs over all charges qi located at ri in the
2 2

molecule, and where the angular brackets denote quantum averaging.

b. Show that the average electrical dipole moment of the hydrogen molecule is zero.

The arrangements of linear molecules, viewed as very narrow ellipsoids, may be

quite complicated. Figure 4.23 displays four structures of ellipsoids on the plane.
In the first row the ellipsoids’ axes are normal to the plane, and in the second they
are in the plane or are inclined with respect to it. These structures are all realized
in Nature; in three-dimensional patterns the successive planes can be identical and
can also be different.

Fig. 4.23: Various planar lattices of ellipsoids. [B. Donio, S. Buathong, I. Bury, and
D. Guillon, Liquid crystalline dendrimers, Chem. Soc. Rev. 36, 1495 (2007).]

As for spherical molecules, the van der Waals interaction leads generally to an
FCC lattice also for ellipsoidal molecules. These molecules, though, are not parallel
to each other, and have different inclinations at different lattice points. Figure 4.24
illustrates one of the lattice structures of CO2 molecules. Were these molecules just
points, then this would have been an FCC lattice. The packing of the ellipsoids
and the interaction between the quadruple moments cause the molecules in each
XY −plane to arrange themselves parallel to each other, with their axes inclined
relative to the plane. On the other hand, the molecules in the neighboring planes
arrange themselves almost normal to each other; the emerging structure is then
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 262

262 Introduction to Solid-State Physics

tetragonal, or even orthorhombic. The F2 molecules have similar structures, and so

do long organic molecules like those that form liquid crystals, or biological molecules,
e.g., DNA.

Fig. 4.24: One of the crystal structures of CO2 . Taken from http://www.
webelements.com/compounds/carbon/carbon_dioxide.html

The crystal structure of hydrogen molecules, H2 , is still a topic of active

research. Figure 4.25 portrays four possibilities to arrange H2 in the plane. Under
strong enough pressures the ions can approach closely each other, and the molecules
then dismantle, to yield a lattice of ions where the electrons are shared by all ions,
i.e., a metal.
The situation becomes even more subtle when the molecules possess a dipole
moment, see the right panel in Fig. 4.1 (and also Sec. 4.3). The interaction
between the dipoles decays as R−3 with the distance between them [Eq. (4.30)].
The arrangement of the molecules involves their locations, the directions of their
axes, and also the vectorial directions of their dipole moments. Since dipoles sums,
i.e.,

[pi · pj − 3(pi · R̂ij )(pj · R̂ij )]/Rij
3
, (4.38)
<ij>

are only conditionally convergent (much like the sums over interactions between
point charge, Sec. 4.2), the result depends on the shape of the surfaces of the
specimen. The dipolar energy varies with the dipoles’ directions. When they are
both parallel to the vector joining them, Rij , the energy is −2pi pj /Rij
3
, and hence
energy is gained when the dipoles are parallel. As opposed, when they are both
3
normal to Rij , the energy is pi pj /Rij and thus energy is gained when the dipoles
are antiparallel. As an example, the results of a numerical minimization of the
dipolar energy for an arbitrary planar lattice are shown in Fig. 4.26. The arrows
there represent the projections of the dipole moments on the plane; the unit cell
contains eight dipoles.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 263

Crystal bonding 263

Fig. 4.25: Patterns of the hydrogen molecules in the plane. In the one labeled
C2/c the triangular lattices in the figure arrange themselves in space as ABCDA.
In all other cases, the spatial ordering is ABA. [V. Labet, P. Gonzales-Morelos,
R. Hoffmann, and N. W. Ashcroft, A fresh look at dense hydrogen under pressure.
I. An introduction to the problem, and an index probing equalization of H-H dis-
tances, J. Chem. Phys. 136, 074501 (2012).]

4.5 The hydrogen bond

The hydrogen bond. Numerous molecules adopt a polar structure in which the
positive and the negative charges are concentrated on the opposite sides of the
molecule (see, e.g., the right panel in Fig. 4.1 and Sec. 4.3). Certain effects of
the interactions among molecules that possess dipole moments are reviewed in Sec.
4.4. A more intricate situation arises when the molecules contain hydrogens. Since
hydrogen has just a single valence electron, that is located on a covalent bond
within the molecule, its nucleus (i.e., the proton) remains exposed on the outer side
of the molecule. This proton can then be at the vicinity of the electrons’ clouds on
the negatively-charged sides of neighboring molecules, and then the system gains
electrostatic energy. The binding energy involved is higher relative to that caused
by the van der Waals interaction. This bond is termed the “hydrogen bond”; it
is the dominant one in many natural materials, e.g., organic or biological molecules
that contain many hydrogen atoms.
Hydrogen bonds among hydrogen-fluoride molecules. A simple polar
molecule is that of hydrogen fluoride, HF. As seen in Fig. 4.17 and in problem 4.11,
the electrons’ clouds around the fluorine nucleus are directed towards the vertices of
a slightly-distorted tetrahedron (a triangular pyramid built of isosceles triangles);
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 264

264 Introduction to Solid-State Physics

Fig. 4.26: The dipole pattern obtained by minimizing the energy on a planar
lattice. Taken from http://loto.sourceforge.net/loto/, courtesy of Takeshi
Nishimatsu.

three of the base vertices are occupied by identical doubly-occupied states of the
fluorine, and the fourth accommodates the hydrogen ion, coupled to the fluorine
by a covalent bond. The outer part of the positively-charged hydrogen nucleus is
exposed and can then come closer to the negative charges on a neighboring molecule.
When two such molecules approach one another, the hydrogen ion of one of them
prefers to approach one of the doubly-occupied electrons’ clouds of the other; a
hydrogen bond between the molecules is then formed. The bond is along the axis
that connects the two fluorine nuclei, such that on one side there is the hydrogen ion
of one molecule, and the other side accommodates one of the doubly-occupied states,
as shown in Fig. 4.27(a). Additional molecules can now approach the two molecules
and form more hydrogen bonds with them. In the case of solid HF, the structure
created is that of a zig-zag one-dimensional chain, as portrayed schematically in
Fig. 4.27(b). On the line that joins the two fluorines, there is a single hydrogen ion.
This hydrogen has a covalent bond with one of its nearest-neighbor fluorines, and
a hydrogen bond with the negative charge clouds of the other. The angle between
the covalent bond and the hydrogen one on the two sides of a fluorine ion in the
chain is 116◦ . Each fluorine is surrounded by hybrid functions of three types, one
that participates in the covalent bond, one, with a double occupation, is directed
towards the hydrogen bond, and the other two are the remaining doubly-occupied
states. A full calculation of the configuration requires an optimization of all angles
among these bonds; however, the measured values are quite similar to the angles
between other sp3 −bonds, e.g., those discussed in Sec. 4.3. Note that the chain
has two possible states: the covalent and the hydrogen bonds between all pairs of
fluorines can be interchanged while leaving the energy intact. In other cases the
degeneracy is even larger.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 265

Crystal bonding 265

(a) (b)

Fig. 4.27: (a) A pair of HF molecules; the larger spheres are the fluorine ions, the
smaller are the hydrogen ions, and the intermediate-size ones represent the doubly-
occupied electronic states of the fluorine. The straight thick line that connects the
two fluorines is the hydrogen bond between the hydrogen ion which couples the
right fluorine and one of the doubly-occupied states of the left fluorine. (b) A chain
of such molecules. The short, solid lines describe covalent bonds, and the long
dashed ones represent hydrogen bonds (the distances between the nuclei are given
in picometers).

Lattice of hydrogen fluoride. At room temperature, hydrogen fluoride is a

gas. When cooled below 190◦ K, it becomes a solid built of chains, see Fig. 4.27(b).
The periodicity along the chain is of two molecules: each second fluorine has the
same surrounding. In one of the three-dimensional structures of this material the
chains are arranged parallel to each other along one of the axes of an orthorhombic
lattice. Nevertheless, the dipolar interaction between the molecules, together with
the van der Waals interaction, and the possibility to form more than a single hydro-
gen bond around each fluorine, give rise to a multitude of phases of this material,
which are still being explored.
The hydrogen bond as a covalent bond. The hydrogen bond is particularly
effective in binding hydrogen and ions of the upper right hand-side of the periodic
table, e.g., nitrogen, oxygen, or fluorine. These ions contain only the first two
electronic shells, and hence the hydrogen ion can be rather close to their nuclei.
The hydrogen bond is presented above as resulting from the “classical” Coulomb
attraction between the charge on the hydrogen and the charge on the negative ion
in the neighboring molecule, very similar to the ionic bond. However, as in the case
of HF, the hydrogen bond stems from the proximity of the hydrogen to a doubly-
occupied state on the negative ion (e.g., F). Without this negative ion, the ψ100
state of H would have taken part in a molecular covalent bond with one of the
F1
states of the F1 ion, for example, ψ111 ; this joint state, termed “bonding state”, is
occupied by two electrons. The other state of the two ions (the “anti-bonding” one)
is empty. Adding together the doubly-occupied state of the other negative ion, say
F2
ψ111 , enables the formation of linear combinations of three states: the two that take
part in the original covalent bond, and the state of the other negative ion. In this
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 266

266 Introduction to Solid-State Physics

way three new states, common to the hydrogen and its two neighbors, are created:
F1 F2
Ψi = A(ψ100 + βi ψ111 + γi ψ111 ). Each of these states is a sum of the three original
states, with different coefficients. These states are occupied by the four electrons,
one from the hydrogen, one from the negative ion to which it has been originally
coupled, and two from the doubly-occupied states of the other negative ion. The
four electrons can occupy two states out of the three, those of the lower energies; the
third state remains empty, like the anti-bonding state of the previous case. In many
cases the total energy of the new states is lower than that of the separate states in
the original covalent bond and the negative ion. Now the four electrons are all in
molecular states shared by the three ions. For the bond F−−H−F, where on both
sides of H there is the same ion, these states are symmetric and the hydrogen is thus
expected to be in the midpoint between the two F’s. Figure 4.27, which is based on
experiment, locates the hydrogen closer to one of the fluorines; this indicates that
the actual bond joins together an ionic bond and a molecular one.
Ice. The paramount examples of hydrogen bonds are crystals of water, i.e.,
the various phases of ice. The angle between the two covalent bonds of the water
molecule (between the oxygen and the two hydrogens) is 104.5◦ (Fig. 4.18). Each
of the oxygens has in addition two pairs of electrons, which occupy hybrid states
that point at the other two corners of the slightly-distorted tetrahedron. As for
the molecule of HF, when two water molecules approach one another, the Coulomb
attraction between a hydrogen ion (that possesses an extra positive charge, exposed
towards the outer side of the molecule) of one of them and the doubly-occupied
state of the oxygen of its neighboring molecule, leads to a hydrogen bond between
those, as depicted in Fig. 4.28(a). Since each oxygen has two clouds of negative
charges, it attracts a hydrogen ion from each of two other water molecules, thus
surrounding itself by four other O’s, in a tetrahedron-type structure. Two of the
tetrahedral directions accommodate covalent bonds with the H’s within the same
molecule, and the other two support hydrogen bonds with neighboring molecules.
Repeating these arrangements in all directions, it is not surprising that the oxygen
ions create a diamond-like lattice. This is indeed the cubic phase of ice, shown in
Fig. 4.28(b). This phase is rather rare, and appears at low enough temperatures.
The ubiquitous phase is the hexagonal phase of ice (also built of tetrahedrons,
like wurtzite) displayed in Fig. 4.28(c) which is obtained when water freezes under
normal conditions. Viewed from above, it looks like graphite; however, each of the
planes here is distorted, to accommodate the angles between the bonds in the water
molecules.
Snowflakes. The hexagonal structure of ice leads to the beautiful patterns of
snowflakes, like those displayed in Fig. 4.29. Typically, a snowflake grows out of a
tiny ice crystal, that is frozen under ambient conditions. This tiny crystal has the
shape of a hexagonal prism, with six faces normal to the base and oriented at 120◦
relative to each other (see Fig. 3.15). The little crystal is growing on each face,
but water molecules that attempt to reach the space in-between the branches which
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 267

Crystal bonding 267

(a) (b)

(c)

Fig. 4.28: (a) The bonds of a water molecule with its four neighbors. Each oxygen
ion is surrounded by four hydrogen ones; it is coupled to two of them by covalent
bonds and to the other two by hydrogen bonds. (b) The cubic (diamond-like)
structure of ice. (c) The hexagonal structure of ice.

stem form each face collide with the latter, and are unable to enter inside. The
result are bifurcations of the branches while the hexagonal symmetry is preserved.
Ice and entropy. The main reason for the numerous phases of ice (sixteen
different phases have been identified so far) is related to the large entropy of the
crystalline structures described above: each oxygen is surrounded by two covalent
bonds and two hydrogen ones, and thus the same structure can be achieved by many
different configurations in which these bonds are interchanged. The entropy of the
ground state, as calculated by Pauling, is determined by the “ice rules”, according
to which the ice crystallines are formed under two constraints. First, around each
oxygen ion there are precisely two covalent and two hydrogen bonds. There are six
possibilities to choose two bonds out of four, and for N water molecules there are all
in all 6N such possibilities. Second, on each bond between two oxygens there resides
a single hydrogen ion. The lattice contains 2N bonds (around each oxygen there are
four bonds but each of them is shared by two oxygens). The 6N possibilities include
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 268

268 Introduction to Solid-State Physics

Fig. 4.29: Snowflakes. [K. G. Libbrecht, The physics of snow crystals, Rep. Prog.
Phys. 68, 855 (2005).]

also those in which there are two covalent bonds that take part in the coupling (one
bond at each of the oxygens), or two doubly-occupied charge clouds (one near each
O). In addition, there are two possibilities in which there is a covalent bond on one
side and a charge cloud on the other. Only the last two options can be realized,
and therefore the total number of possibilities is multiplied by 1/22N . All in all the
number of configurations which obey the ice rules is W = 6N /22N = (3/2)N . By
Boltzmann’s definition, the entropy is extensive, S = kB ln W = kB N ln(3/2), i.e.,
it is proportional to the size of the specimen. Experiment indeed yields numbers
close to this estimate.

Problem 4.17.
Consider the formaldehyde molecule, CH2 O.
a. What is the shape of this molecule? Determine the type of the hybridization of
the wave functions that form the covalent bonds.
b. Show that these molecules can construct a rectangular-centered lattice, in which
each oxygen has two hydrogen bonds with neighboring molecules. Is this the only
possible structure?

4.6 Metallic crystals

Insulators and metals. In the crystals studied so far, the electrons are localized
near their “parent” atoms, either in the atom itself, like in the noble gases, or in the
ions (in ionic bonds) or on the covalent bonds shared by two atoms. (Exceptions
are the π−electrons in graphene, whose wave functions are spread over the entire
plane.) As the electrons are not free to move around, they can hardly respond to an
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 269

Crystal bonding 269

external electric field; those materials are hence all insulators. As opposed (and
as mentioned in Sec. 4.1) a metal can be considered as a huge molecule, in which
each atom “contributes” its valence electrons; these electrons are described by wave
functions spread on all ions in the lattice, as in Eq. (4.27) that corresponds to
electrons shared by all carbons in the benzene molecule. The kinetic energy of an
electron whose wave function is spread over the entire lattice is reduced considerably.
This effect enables the electrons to move around under the influence of an electric
field, and the material attains good electrical conductivity.
Localized and delocalized states. Figure 4.30 illustrates localized states
of the electrons on the ions (on the left) and delocalized states where the wave
function of each electron is spread over the entire lattice (on the right). When each
atom contributes a single valence electron (e.g., sodium) then a delocalized state
is a linear combination of the N atomic states, and thus there are N independent
such states. Pauli principle dictates that each electronic state can accommodate at
most two electrons. The energies of these states are gradually increasing, and at
low temperatures the states are filled up to an energy called “the Fermi energy”
(see chapter 6).

Fig. 4.30: The passage from localized electronic states (each electron is at the vicin-
ity of its parent atom, on the left) to delocalized states (each electron is “spread”
over the entire lattice, on the right).

Qualitative discussion: the bonding energy of metals. The detailed

calculation of the energies of metals necessitates the knowledge of the quantum
properties of electrons in the periodic potential of the ions, see Chapter 6. Here
are mentioned general considerations that determine the crystal structure and the
binding energy. In crystals of alkali-metals (e.g., sodium or lithium), each atom
contributes a single electron. A useful length scale in this context is the radius of
a sphere representing the average volume occupied by an electron in the lattice,
in this case the volume of the unit cell, V = 4πrs3 /3. Calculations yield that the
momentum of an electron at the Fermi level (i.e., the momentum of an electron
whose energy is the Fermi energy) is of the order of /rs and therefore the average
kinetic energy of an electron in a metal is C1 /rs2 = C 1 (/rs )2 /(2m); m is the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 270

270 Introduction to Solid-State Physics

electron’s mass and C 1 is a dimensionless constant. (Apart from this coefficient,

the general form is dictated by dimensional considerations: when rs is the only
length scale in the problem, the momentum must be proportional to /rs .) The
detailed calculation [Eq. (6.34)] leads to C 1 = (3/5)(9π/4)2/3 . This positive energy
is augmented by the Coulomb potential energy, of the electrons with the nuclei, and
of the electrons with themselves. The calculation of the potential energy resembles
that of the Madelung sum, Sec. 4.2, with continuous integrals over the electrons’
locations. The answer can be written as −αe2 /(2rs ), where α is a constant that
characterizes the lattice. With the assumption that the electrons are distributed
uniformly in space, one finds α ≈ 1.792 for the three dense lattices FCC, HCP,
and BCC, and smaller values for SC (1.760) and diamond (1.671). The higher
values cause most of the metals to be arranged in a dense structure (see Fig. 2.22).
The total energy includes also contributions due to the exchange energy, which
is related to the overlap of the electronic wave function near a certain ion with
those of the neighboring ones [see, e.g., Eq. (4.19)]. The exchange terms, which
also stem from the Coulomb potentials, contribute the energy −α e2 /(2rs ), where
α ≈ α/2 (this can be expected, again from dimensional arguments: recall that the
only length scale in the problem is rs and the only charge scale is e). All in all, the
energy is
C1 e2
E= 2
− (α + α ) . (4.39)
rs 2rs
The minimal value is reached at rs /aB = 4C1 /[(α + α )e2 aB ] ≈ 1.6. This is smaller
than the typical values of alkali metals, which are between 2 and 6. The error is
due to the rather simplistic considerations.
Transition metals. The heuristic considerations given above account solely for
the valence electrons in the outer shell of each atom. In reality, there are interactions
with the electrons in the inner shells. For example, transition metals’ ions like
copper, iron, manganese, or titanium, have d−electrons available for bonding, and
these can indeed create covalent bonds with the neighbors (similar to Fig. 4.19),
in addition to the metallic bond between the s−electrons. The ions may also form
van der Waals bonds with other ions in the lattice, and so further strengthen the
total binding energy.

4.7 Magnetism

The atomic dipole moment. Magnetic systems deserve a separate treatise;

nonetheless, it is illuminating to present several of the physical principles of mag-
netism. As mentioned in Sec. 2.10, many materials undergo magnetic phase tran-
sitions. This phase transition can occur when each atom or ion possesses an aver-
age magnetic dipole moment, that vanishes above the transition temperature and
increases as the temperature is reduced towards zero below the transition tem-
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 271

Crystal bonding 271

perature; the value of this moment at low temperatures is μi , where i labels the
lattice site. Several examples of magnetic patterns formed by the moments are
presented in Secs. 2.10 and 3.13; these arrangements usually reduce the symmetry
(as compared to that of the disordered phase above the transition temperature)
and increase the unit cell. Neutron-scattering experiments that identify the mag-
netic structures are described in Sec. 3.13. The magnetic moment of an electron
is related to the angular momentum L and the spin S (measured in units of ),
μ = −μB (L + ge S) where ge ≈ 2.00232 is the gyromagnetic coefficient of an
electron and μB = e/(2mc) = 5.7884 × 10−9 eV/Gauss is the Bohr magneton.
The magnetostatic interaction between two magnetic dipole moments, [μi · μj −
3(μi · R̂i )(μj · R̂j )]/Rij
3
is of the order of μ2B /R3 , approximately 10−4 eV for R = 2Å
(check!) Except in special situations, this energy is much lower than kB Tc , where Tc
is the typical transition temperature of a magnetic phase transition; this interaction
therefore cannot explain the existence of such transitions. Usually, the transition
temperature of a phase transition is of the order of the energy of the interaction that
leads to this transition (divided by the Boltzmann constant kB ). The conclusion
is that the main source of magnetism in solids is not the dipole-dipole interaction,
but quantum effects.
Singlet and triplet energies. The spin state of two electrons is a singlet
or a triplet, in correspondence with their symmetric or the antisymmetric orbital
state (with respect to the interchange of the locations of the two electrons, see
Sec. 4.3). The energy difference between these two states is the exchange energy J,
J = ET −ES (J < 0 for the hydrogen molecule). The square of the total spin of two
electrons can attain only two values, 0 for the singlet and 2 for the triplet. The total
energy of the two states can therefore be formally expressed as E = ES − JS212 /2,
where S12 is the total spin of the two electrons. This energy is ES for the singlet
state and ET for the triplet one. Recall that the energy difference stems from the
Coulomb energies of the electrons in space; the expression in terms of the spins is
just formal. Using
3
S212 = S21 + S22 + 2S1 · S2 = + 2S1 · S2 ,
2
gives
E = E0 − JS1 · S2 , (4.40)
where E0 = ES − 3J/4 = (ES + 3ET )/4 is the “center of mass” of the two energies,
weighed by their ratio of degeneracies, 1:3.
When J < 0 (as for the hydrogen molecule) the singlet state is the preferred
one. Then the scalar product of the two spins is negative,
S1 · S2 = (S212 − S21 − S22 )/2 = −3/4 ,
and they are antiparallel to each other in the ground state. On the other hand,
sometimes the energy of the triplet state is the lower one (this is determined by the
various integrals on the wave functions and the Coulomb energies, that are more
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 272

272 Introduction to Solid-State Physics

complicated than in the hydrogen molecule). When J > 0, the scalar product of
the two spins is positive,
S1 · S2 = (S212 − S21 − S22 )/2 = 1/4 ,
and the two spins are parallel to one another in the ground state. (In fact, the angle
between the two spins, though being acute, is not strictly zero due to quantum
effects, and the spins are not precisely parallel. Inspection shows that in two of the
three triplet states the two electrons’ spin projections on the ẑ−axis are identical
but in the third one the projections have opposite signs.)
Super-exchange. Another process that can contribute to the effective energy
between two neighboring spins, of the form (4.40), was proposed by Kramers and
significantly developed later by Anderson. Consider two neighboring ions in the
lattice, each having a singly-occupied orbital state. When the electron is confined
to the vicinity of its parent ion, the kinetic energy is relatively large (due to the
uncertainty principle, the momentum is of the order of /R, where R is the size of
the confinement region). On the other hand, were this electron to hop over to the
neighboring ion, and thereby reduce the kinetic energy, it would have created there
an excited state, with a higher potential energy because of the Coulomb repulsion
between that electron and the other one, already present there. The Pauli principle
allows this configuration only when the spins of the two electrons are opposite to
one another. Exploiting Eq. (C.3), with the excited state as one of the n−states in
the sum, yields an energy similar to Eq. (4.40), with a negative J. The same result
can be derived by the variation method: one begins with the state in which there is
an electron on each ion, and adds an excited state in which the two electrons reside
on the same ion. The latter, due to the Pauli principle, is a singlet. The correction
it causes to the ground-state energy is negative, as for example in Eq. (4.32). It
thus follows that it “pays” for the two electrons to be in the singlet state. Equation
(4.40), with a negative J, is then valid.
This process is frequently encountered in transition-metal oxides, where the
electron tunnels from a metal ion to the oxygen, and then to the neighboring metal
ion. In fact, Kramers proposed this mechanism to explain antiferromagnetism in
manganese oxide (see Sec. 3.13). This process also contributes to the magnetism
of the lanthanum cuprate, as discussed in problem s.2.17. The relative geometry
of the various ions’ locations can cause the respective interaction constant J to be
positive or negative.
Arbitrary angular momentum. When there are numerous electrons in the
atom (or in the ion), their magnetic moments are summed over according to the
addition rules of quantum angular momenta. In the context of magnetism in solids,
the total angular momentum of atom i is frequently presented as a (dimensionless)
vector Si [the dipole moment is μi = gμB Si , where g is the (dimensionless) gy-
romagnetic coefficient (of order unity)]. The magnetic energy may include various
combinations of these angular momenta. The treatment of the general case is be-
yond the scope of this book; note though, that the main features of the magnetic
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 273

Crystal bonding 273

structures presented in Secs. 2.10 or 3.13 may be explained by exploiting the super-
exchange energy. The remaining part of this section is devoted to certain simple
models, and discusses solely the “classical” ground states of magnetic moments on
atoms in the crystal, ignoring quantum effects.
The Heisenberg model. In the Heisenberg model, the energy of the spin
degrees of freedom is [analogous to Eq. (4.40)]

ĤM = − Jij Si · Sj . (4.41)
<ij>

The quantum number that corresponds to the vector S can be an integer or a half-
integer, with |S|2 = S(S + 1). In the ground state the quantum averages of the
vectors Si , and consequently of the magnetic moments μi = gμB Si , attain definite
values which determine the magnetic arrangement of the system.
A complete calculation of the energy spectrum of the Hamiltonian (4.41) is
rather complicated. However, a qualitative estimate of the ground-state energy
can be derived by treating the vectors Si as classical vectors, i.e., as vectors of a
definite length and a certain direction in space. The problem is then to find the
configurations of these vectors that minimize the energy (4.41). If all coefficients
Jij are positive, the energy is minimal when all vectors Si are parallel to each other,
like in Fig. 2.42B; this arrangement constitutes the ferromagnetic phase: the
total magnetic moment of the crystal is then the sum over the average magnetic
moments on all atoms (ions).
The exchange energy originates from integrals over wave functions in which the
locations of the two electrons are interchanged between the two atoms; it thus
decays exponentially with the distance in-between the atoms. Hence, it suffices to
include in the sum only pairs of close-by neighbors. For example, when only nearest
neighbors are included, and when the coefficients Jij are all negative, the energy is
minimal when Si and Sj on nearest-neighbor atoms are antiparallel to one another.
All neighboring moments of a certain moment are antiparallel to it, e.g, as in Fig.
2.42C, or the configuration G in Fig. 3.27. This is the simplest antiferromagnetic
state. Antiferromagnetic interactions between both nearest neighbors and next-
nearest neighbors (i.e., the corresponding J’s are negative) give rise to “frustration”
(Sec. 2.10), which in turn implies complicated magnetic structures. This point is
demonstrated on a simple-cubic lattice, see Fig. 3.27. Let the Hamiltonian be of
the form (4.41), with an exchange energy J1 between nearest neighbors (nn), and
an exchange energy J2 between next-nearest neighbors (nnn). The minimal energy
of this classical Hamiltonian is derived as follows. According to Eq. (3.71), the
three structures are described by Si = S0 exp[iK · Ri ]. This is a plane wave of
wave vector K (various magnetic structures can be described by such waves). The
wave vectors of Eq. (3.72) are such that exp[2iK · Ri ] = 1, and therefore the scalar
product Si · Sj = |S0 |2 exp[iK · (Ri − Rj )] depends only on the distance in-between
the pair of moments. It follows that all contributions of the neighbors to site i are
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 274

274 Introduction to Solid-State Physics

identical to those of the neighbors of the origin; hence

N  
EM = − (J1 S 0 · S j + J2 S0 · Sj ) , (4.42)
2 j=nn j=nnn

where the first sum runs over the six nearest neighbors of the origin, Rj =
±ax̂, ±aŷ, ±aẑ, and the second over the twelve next-nearest neighbors, Rj =
±ax̂ ± aŷ, ±aŷ ± aẑ, and ±aẑ ± ax̂. The factor of 1/2 in front of the sums appears
since the sum over pairs, <ij> is transformed into two sums, over i and over j;
thus each pair appears twice. Using the explicit expressions for the scalar product,
and the normalization |S0 |2 = 1 of the spins’ lengths, the energy per unit cell is
E(K)
e(K) = = −J1 [cos(Kx a) + cos(Ky a) + cos(Kz a)]
N
− J2 [cos(Kx a + Ky a) + cos(Ky a + Kz a) + cos(Kz a + Kx a)
+ cos(Kx a − Ky a) + cos(Ky a − Kz a) + cos(Kz a − Kx a)]
= −J1 [cos(Kx a) + cos(Ky a) + cos(Kz a)]
− 2J2 [cos(Kx a) cos(Ky a) + cos(Ky a) cos(Kz a) + cos(Kz a) cos(Kx a)] .
(4.43)
At zero temperature the entropy contribution is absent, and stable state is reached
when the energy is minimal. Indeed, the derivative of the energy with respect to
any of the components of K yields
sin(Kx a){J1 + 2J2 [cos(Ky a) + cos(Kz a)]} = 0 , (4.44)
and two more equations like (4.44) in which the roles of the axes are interchanged.
The energy is extremal when sin(Kx a) = 0 or when cos(Ky a) + cos(Kz a) =
−J1 /(2J2 ), and two more similar conditions obtained by interchanging the axes.
At the end of the day, Eqs. (4.44) give the following possibilities:
K(πππ)a = (π, π, π), K(0ππ)a = (0, π, π), K(00π)a = (0, 0, π), K(000)a = (0, 0, 0),
K(φφφ)a = (φ, φ, φ), K(πθ θ )a = (π, θ θ ), K(0θθ)a = (0, θ, θ) , (4.45)
where cos θ = −1−J1 /(2J2 ), cos θ = 1−J1 /(2J2 ), and cos φ = −J1 /(4J2 ). The last
three solutions do not comply with exp[2iK · Ri ] = 1, and therefore are not allowed
in the present case. Upon inserting these three solutions into the expression for the
energy, Eq. (4.43), they do not yield a minimum of the energy. As for the other
options, one notes that together with each solution for which one of the axes differs
from the others, there are also solutions with interchanged axes. The corresponding
energies are
e(000) = −3J1 − 6J2 , e(00π) = −J1 + 2J2 , e(0ππ) = J1 + 2J2 , e(πππ) = 3J1 − 6J2 ,
e(πθ θ ) = −J1 + 2J2 + J12 /(2J2 ), e(0, θ, θ) = J1 + 2J2 + J12 /(2J2 ),
e(φφφ) = −3J12 /(8J2 ) . (4.46)
Figure 4.31(a) shows these six energies (in units of J2 ), each in the range where
the corresponding state exists, as functions of J1 /J2 . When both J1 and J2 are
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 275

Crystal bonding 275

positive, the minimal energy is that of the ferromagnetic state, with K(000) = 0;
the two types of bonds prefer the spins to be parallel. When J1 < 0 but J2 > 0,
the minimal energy is that of a simple antiferromagnet (G in Fig. 3.27), with
K(πππ)a = (π, π, π) (the spins on nearest neighbors are antiparallel, and the spins
on next-nearest ones are parallel; there is no competition between the magnetic
interaction of nearest neighbors and that of next nearest-neighbors). When J2 < 0,
the signs of the energies in Fig. 4.31 have to be reversed, and then the ferromagnetic
state is realized for J1 > −4J2 > 0, and the simple antiferromagnetic one for J1 <
4J2 < 0 (in the two configurations the energy of the nearest-neighbor interaction
overcomes that of the next nearest-neighbor ones). On the other hand, in the range
4J2 < J1 < 0 the preferred state is the antiferromagnetic one of type C (right
side of Fig. 3.27), with K(0ππ)a = (0, π, π), and in the range 0 < J1 < −4J2 ,
the ground state is the antiferromagnetic state of the A type (mid panel in Fig.
3.27), with K(00π)a = (0, 0, π). (The axes can be interchanged in both cases.).
Figure 4.31(b) depicts the four regions in the J1 − J2 plane, where each of the
aforementioned structures can exist. Such a figure is called a “phase diagram”,
and each region in it represents a different phase of the magnetic material. Similar
considerations enable the identification of more complex structures. for instance,
that of the manganese oxide.

(a) (b)

Fig. 4.31: (a) The energies of Eq. (4.46), in units of J2 , as functions of the ratio
J1 /J2 , when J2 < 0. (b) The phase diagram of the minima of Eq. (4.43). The
letters F, G, C, and A represent the four options for magnetic structures in the
ground state of the Hamiltonian (4.42).

The Ising model. The Heisenberg model does not suffice to determine the
spatial directions of the spin on each ion. Those are dictated by anisotropic
terms in the energy, due e.g., to the coupling between the orbital angular momentum
and the spin, that is, the spin-orbit interaction. This interaction relates the
orientations of the magnetic moments to the spatial lattice vectors and determines
special directions called “easy axes”, along which the spins point. The dipole-
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 276

276 Introduction to Solid-State Physics

dipole interaction is also capable of relating the directions of the magnetic moments
and the lattice vectors. This interaction is given by Eq. (4.38), when the electric
dipoles there are replaced by magnetic dipoles or spins, that is, pi → μi → Si . As
described in Sec. 4.4, the spin components parallel to the spatial vector connecting
them tend to be parallel to one another, while the normal components prefer to be
antiparallel to one another). In the extreme situation, the anisotropic interaction
that determines the easy axes is very strong, such that the magnetic moments are
confined to point along a single axis, e.g., the ẑ−axis. In that case, Eq. (4.41)
becomes

ĤM = − Jij Siz Sjz . (4.47)
<ij>

In the simplest situation, each spin may attain one of two values, Siz = ±1. This
expression for the energy is called the Ising model. The spin patterns implied by
this energy differ from those that emerge from the Heisenberg model. For instance,
all solutions listed in the second line of Eq. (4.45) cannot be realized in the Ising
model.
The triangular lattice and frustration. Another example is related to mag-
netic moments arranged on a triangular lattice. When Eq. (4.41) is applied to such
a lattice, with identical exchange couplings, J < 0 for each pair of nearest neighbors
(and zero otherwise), it is found that there is no configuration for which the mo-
ments of all nearest-neighbor pairs are antiparallel. The energy of a single triangle
is E = −J(S1 · S2 + S2 · S3 + S3 · S1 ). When spins 1 and 2 are antiparallel to one
another the energy is −|J|, as the third spin is “frustrated”: its interactions with
the other two cancel one another, and the energy of the tiny system is independent
of its direction. On the other hand, once the spins are allowed to assume arbitrary
angles in-between them, then for an angle 120◦ between any two spins the energy
is minimal, −3|J|/2 (problem 4.18). The triangular lattice is built of triangles; its
ground state is as shown in Fig. 2.12, in which the sites marked by A, B, and C are
occupied by spins with angles of 120◦ with each other (see also Fig. 2.43). Choosing
the spins in the sites A and B to have an angle of 120◦ with respect to one another,
the spin in the sites C should point out along the third direction. Each of the edges
of the triangle ABC determines the spin at the point that closes the triangle. In this
way all spins’ directions in the ground state of the triangular lattice are determined.

Problem 4.18.
Prove that the energy of a triangle with antiferromagnetic couplings is optimal when
the angle between each pair of spins is 120◦ .

Frustration and degeneracy. The solution of the triangle derived in problem

4.18 is valid for the Heisenberg model, but not for the Ising one. In the latter,
all spins point out along the same direction, and therefore the frustration is not
removed. The Ising model gives rise to many patterns with the same energy; con-
sequently the ground state possesses a finite entropy (like ice). For instance, in
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 277

Crystal bonding 277

Fig. 2.43 all spins on one of the sublattices (e.g., the large points) can be chosen
to point “upwards”, and all those on the other (e.g., the intermediate points) to
point “downwards”. The couplings between these two sublattices are hence not
frustrated, and each of them contribute an energy −|J| per site. Each spin on the
third sublattice is coupled antiferromagnetically to three spins that belong to each
of the previous sublattices. These triplets are opposite to one another, and thus
that spin is frustrated; it can be directed along any of the two possible directions
without affecting the energy. This causes a degeneracy of at least 2N/3 of the ground
state energy (where N is the total number of the spins in the system). In fact the
degeneracy is much higher, as this energy is shared by many more arrangements.
The degeneracy is even higher for other lattices, like the Kagomé one (problem
s.2.2) in which each triangle can assume several magnetic patterns.

Problem 4.19.
A one-dimensional magnet is described by the Hamiltonian (4.41), with exchange
energy J1 between nearest neighbors (nn) and J2 between next-nearest neighbors
(nnn). These couplings may be positive or negative. The spins are normalized, i.e.,
|Si | = 1 (one may absorb |Si |2 in the exchange energies). Determine the magnetic
order in the classical ground state, as a function of J1 /J2 . How is this answer
modified for the Hamiltonian (4.47)?

4.8 Answers for the problems in the text

Answer 4.1.
a. Within the first square, there is a +1 charge at the √ origin (the center of the
square) four +1 charges at the four corners (at a distance 2 from the center; each
one contributes 1/4) and four −1 charges at the midpoints of the edges (at a distance
1 from the center, each one contributes 1/2). The total sum is 1+4(1/4)−4(1/2) = 0.
The perimeter
√ of the nth square includes four +1 charges at the corners (at a
distance n 2 from the center, each contributes 1/4 to the inner side of this square),
four (−1)n charges at the midpoints of the edges, on the axes (at a distance n
from the center, each gives 1/2), and eight (−1)n+m charges
√ at the eight points
(±n, ±m), which are located on the edges at a distance n2 + m2 from the center,
with 1 ≤ |m| ≤ n − 1 (each one contributes1/2). Therefore, the total contribution
from the inner side of this square is

n−1
Qin = (1/4)4 + (1/2)(−1)n + (1/2)8 (−1)n+m = −1 .
m=1

In a similar fashion, the outer part of the (n − 1)th square contributes Qout =
Qin + 2 = 1 (the additional factor of 2 is due to the different weights of the four
charges at the corners, each giving 3/4 to the outer side instead of 1/4 to the inner
side). Adding up the two contributions yields that the total charge in between the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 278

278 Introduction to Solid-State Physics

(n − 1)th and the nth squares vanishes. The sum of all charges within the nth
square, as obtained from a sum over all the regions, vanishes.
b. The contribution of the√inner part of the first square to the Madelung constant
α is S1 = (1/2)4 − (1/4)4/ 2 = 1.29289. The contribution to α from the inner side
of the nth square (n > 1) is
√ 
n−1 
Sn = −[(1/4)4/ 2n2 + (1/2)4(−1)n /n + (1/2)8 (−1)n+m / n2 + m2 ] .
m=1
c. The net contribution
 of the charges in between the two squares to the coefficient
α is Sn + Sn−1 + 2/ 2(n − 1)2 (the last term is due to the corners of the inner
square; the contribution of each of them to the outer side is 3/4). Hence, α =
1.29289 + 0.313981 + 0.003648 + 0.002988 + 0.000981 + . . . = 1.61544. The sum of
the first two terms yields 1.61, and the sum of the first six gives 1.615. The value
1.61554 is obtained from the summation over 26 terms. Plotting the terms as a
function of n shows that they decay as n−4 .

Answer 4.2.
CsCl consists of two simple cubic sublattices.  For a positive ion at the origin,
the positive charges are located
√ at distances √i2 + j 2 + k 2 (there are 6 ions at a
distance 1, 12 at a distance 2, 8 at a distance
 3, 6 at a distance 2, etc.) and the
negative ones are situated at distances (i + .5) 2 + (j + .5)2 + (k + .5)2 (there are
√ √ √
8 at a distance 3/2, 24 at a distance 11/2, 24 at a distance 19/2, etc.) from
the origin.
√ The distance between the ions with opposite charges within the unit cell
is R01 = 3/2. (All distances are in units of the lattice constant.) Hence,
√  
3 6 12 8 8 24
αCsCl = − + √ + √ + ··· − √ −√ − ...
2 1 2 3 3/2 11/2
 √ 
= 8 + 24 3/11 + . . . − 3 3 − 6 3/2 − 4 − . . . .

Answer 4.3.
Inspection of the various distances in the unit cell yields
2e   2  2e
∞
1 1
φA = − − = [2γE + Ψ(x) + Ψ(1 − x)] ,
a k k−x k−1+x a
k=1

(the factor of 2 multiplying the sum accounts for the charges on both sides of A);
similarly
e  2 1
∞
1 1 2 2
φB = − − + − −
a k − x k k − 2x k − 1 + x k − 1 + 2x k
k=1
e
= [−2γE − 2Ψ(1 − x) − 2Ψ(x) + Ψ(2x) + Ψ(1 − 2x)] .
a
The Coulomb energy per unit cell is (UC /N ) = (qA φA + 2qB φB )/2 = e(φA − φB ).
The binding energy, in units of e2 /a, u = −(UC /N )/(e2 /a) is given as a function of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 279

Crystal bonding 279

x in Fig. 4.32. This energy diverges near x = 0, because of the attraction between
the ions A and B; it changes sign around x = 1/2, where the two negative ions are
close to one another. The crystal is stable for u > 0, i.e., for x < 0.45. In reality,
the lattice constants a, RAB = RBA = ax, and RBB = a(1 − 2x) are determined
by the repulsion among the ions at short distances.

Fig. 4.32

Answer 4.4.
a. Placing the bigger ion, of radius r> , at the origin, then one of the nearest
neighbors (with the opposite charge and of radius r< ) is located at a(1/4, 1/4, 1/4).
The distance between the centers when the two spheres √ touch one another is a
quarter of the cube diagonal, R10 = r< + r> = a 3/4. Each face of the unit
cell contains one of the bigger ions, again at its center √ (this is an FCC lattice)
and therefore√the diagonal on the face is such that a 2 > 4r> . It follows that
r> /r< < 2 + 6 ≈ 4.45.
b. When 2.414 < r> /r< and the structure is that of the table-salt crystal, the two
ions of opposite charges cannot touch one another any more. Instead, the bigger
negative ions are contiguous, and √ then the distance between them is half of the
lattice constant, R10 = a/2 √ = 2r> . Consequently, the Coulomb energy of this
structure is −N αNaCl q 2 /( 2r> ). The Coulomb energy of the zinc-blende structure
is −N αZB q 2 /(r> + r< ). This latter energy “wins” when
r> 1
>√ ≈ 3.047 .
r< 2(αZB /αNaCl ) − 1
The table-salt structure is realized in the range 2.414 < r> /r< < 3.047, unless
other considerations are involved. Indeed, the MgS crystal has the same structure
as table salt, albeit the ionic radii there obey r> /r< ≈ 2.83 > 2.414.

Answer 4.5. √
a. The distance between the ions is a quarter of the diagonal, R10 = a 3/4 ≈ 2.35Å.
b. In the close-packing arrangement,
√ the diagonal of the face equals four radii of
the larger ion, hence r> = a 2/4 ≈ 1.92Å. The sum of the two radii is the distance
between nearest neighbors, hence r< = R10 − r> ≈ 0.43Å.√The ratio of the radii is
4.47, slightly higher than the limiting value r> /r< = 2 + 6 ≈ 4.45.
c. As in the zinc-blende structure, in the one of wurtzite the zinc ions occupy the
cavity at the center of the tetrahedron formed by the sulphur ions. The edge of the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 280

280 Introduction to Solid-State Physics


a = R10 8/3 ≈
tetrahedron is the lattice constant a of the HCP lattice; therefore,
1.633R10 ≈ 3.84Å. The lattice is a perfect HCP one; thus c = a 8/3 = 8R10 /3 ≈
6.27Å.
d. The cubic unit√cell contains four base units, and therefore the specific volume
per unit cell is a2 c 3/4 ≈ 40(Å)3 . Both structures are close packed; it is hence not
surprising that their densities are identical (see also problem 2.15).

Answer 4.6.
The distance between the ions (for the table-salt crystal) is given by the solution
2
of the equation R exp[−R/ρ] = ραq 2 /(zλ). The binding energy per unit cell is
4
u = αq 2 (1 − ρ/R)/R, and the bulk modulus is B = αq 2 /(18R )[(R/ρ) − 2]. As
mentioned, the measured binding energy is about u ≈ αq 2 /R, implying R  ρ.

Answer 4.7.
The overlap integral is
 
1 1
S = ψA |ψB  = drA e−rA /aB e−rB /aB = d3 xe−x−y .
πa3B π
In spherical coordinates, where the ẑ−direction is along molecule axis, d3 x =
x2 sin θdxdθdϕ. The integrand is independent of the rotation angle ϕ around the
axis, and hence its integration yields 2π. The substitution y 2 = X 2 + x2 − 2Xx cos θ
yields 2ydy = 2Xx sin θdθ, and consequently
  X+x
2 ∞ −x
S= dx xe dy ye−y
X 0 |X−x|
  X+x  ∞  x+X
2 X −x −y −x

= dx xe dy ye + dx xe dy ye−y ,
X 0 X−x X x−X

where the limits on the y−integrals are derived from those on θ, 0 ≤ θ ≤ π. Making
use of the (indefinite) integrals
 
−2x 1 + 2x −2x 1 + 2x + 2x2 −2x
dx xe =− e , dx x2 e−2x = − e ,
4 4
yields Eq. (4.16). In a similar fashion,
  ∞  x+X
e2 e2 −2rA /aB 2e2 −2x
ψA | |ψA  = dr A e = dx xe dy
rB πa3B rB XaB 0 |x−X|
 ∞
e2 −2x e2
= dx xe (x + X − |X − x|) = [1 − (1 + X)e−2X ] ,
πa3B rB 0 R
which yields Eq. (4.18). Finally,
  ∞  x+X
e2 e2 −rA /aB −rB /aB 2e2 −x
ψA | |ψB  = dr A e e = dx xe dy e−y ,
rB πa3B rB XaB 0 |x−X|

and a similar algebra leads to Eq. (4.19).

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 281

Crystal bonding 281

Answer 4.8.
The variational average of the energy is
E±  = (1 + β±
2
± 2β± S)−1 (ψA |Ĥ|ψA  + β±
2
ψB |Ĥ|ψB  ± 2β± ψA |Ĥ|ψB ) .
2
The derivatives with respect to β± vanish when aβ± ± bβ± + c = 0, with
a = SψB |Ĥ|ψB  − ψA |Ĥ|ψB  , b = ψB |Ĥ|ψB  − ψA |Ĥ|ψA  ,
c = ψA |Ĥ|ψB  − SψA |Ĥ|ψA  .
It is straightforward to verify that in the symmetric case, where ψB |Ĥ|ψB  =
ψA |Ĥ|ψA , one finds β+ = β− = ±1, reproducing the bonding and the anti bonding
states discussed
√ in the main text. In the general case one has to choose the solution
β± = (∓b + b2 − 4ac)/(2a) (otherwise the roots are such that β+ = −β− ). The
chosen roots do fulfil the orthonormality condition, 1 − β+ β− + S(β+ − β− ) = 0.

Answer 4.9.
a. Using Eq. (4.24) gives

3 r
ψ(r) = R21 (r) = f (r)r̂ ,
4π r
where r̂ is a unit vector along the r−direction. It follows that
Ψ1 (r) = A[ψ200 (r) + βf (r)n̂1 · r̂] .
The second term is proportional to the cosine of the angle between r and n̂1 , and
hence (assuming that β is positive) is maximal for r  n̂1 . The two terms are
positive for r < 2aB /z, and thus their sum is maximal along the vector n̂1 . When
r > 2aB /z, the two terms have opposite signs; in that case the square of the sum is
maximal along −n̂1 . In both configurations the sum is maximal on the axis through
which passes the vector n̂1 .
b. The quantum scalar product of the two states is

Ψ1 |Ψ2  = |A|2 ψ200 + β n̂1 · Ψ|ψ200 + β n̂2 · Ψ = |A|2 (1 + |β|2 n1i n2j ψi |ψj ) ,
ij

where i, j = x, y, z indicate the cartesian components of the vectors, and ψ200 |ψi  =
0. Using ψi |ψj  = δij yields the desired result.
c. For the hybrid states to be orthonormal, the scalar product should be Ψi |Ψj  =
δij ; consequently, for i = j |β|2 = −1/n̂i · n̂j = −1/ cos α, and for i = j |A|2 =
1/(1 + |β|2 ) = cos α/(cos α − 1).
d. As there is no reason to distinguish between the three bonds, symmetry implies
that the angles between pairs of unit vectors are identical. This is also the config-
uration in which the clouds of negative charge are as far away from each other as
possible. Since the sum of the three angles is 360◦ , each of√them is 120◦ . Such a
triplet of unit vectors is, for instance, n̂1 = x̂, n̂2,3 = −x̂/2± 3ŷ/2. Inserting these
into part (c) gives |β|2 = −1/ cos 120◦ = 2 and |A|2 = cos 120◦ /(cos 120◦ −1) = 1/3,
from which follows Eq. (4.25).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 282

282 Introduction to Solid-State Physics

e. The sp3 hybridization allows for four bonds; this necessitates four unit vectors,
with an identical angle in-between each pair of them. For example,
 
1 1
n̂111 = (x̂ + ŷ + ẑ) , n̂111 = (x̂ − ŷ − ẑ) ,
3 3
 
1 1
n̂111 = (−x̂ + ŷ − ẑ) , n̂111 = (−x̂ − ŷ + ẑ) .
3 3
The angle between any two unit vectors is given by cos α = n̂i · n̂j = −1/3, leading
to |A|2 = 1/4 and |β|2 = 3. The √ coefficients in Eq. (4.26) are chosen to be real; it
follows that A = 1/2 and β = 3.
f. In the general case Ψi = Ai [ψ200 + βi n̂i · ψ], with a unit vector n̂1 along the bond
with the fluorine, and unit vectors n̂2 and n̂3 along the bonds with the hydrogens.
Hence Ψi |Ψj  = A∗i Aj (1 + βi∗ βj n̂i · n̂j ). The bonds with the two hydrogens are
identical; hence the angles between them and the bond with the fluorine are equal as
well, β1∗ β2 = β1∗ β3 = −1/ cos α, and β2∗ β3 = −1/ cos γ. It follows (from the  first rela-
tion) that β2 = β3 . Choosing the coefficients
 to be real, then β 2 = β 3 = −1/ cos γ
and β1 = − − cos γ/ cos α = − − cos γ/(cos (γ/2). The other coefficients are
2 2

obtained from |Ai |2 = 1/(1 + |βi |2 ).

g. Figure 4.33 illustrates the triangular pyramid whose base is formed by the three
hydrogens (H1 , H2 , H3 ), and the carbon C1 is at its apex. The height of the
pyramid is the line Y C1 , the continuation of the straight line connecting the two
carbons, C1 C2 . The angle between Y C1 and C1 H3 is 180◦ − α. Denoting the length
of C1 H3 by d, then the length of Y H3 is d sin α. Denoting √ the distance between two
hydrogens by a, the height of the base triangle, XH3 , is a 3/2. The point Y is the
locus of the heights of the isosceles triangle of the base, √ and therefore√it is located
at the third of the height. It follows that d sin α = a 3/3, i.e., a = d √ 3 sin α. The
angle can be derived from the triangle XC1 H1 : sin(γ/2) = a/(2d) = ( 3/2) sin α.
Using the value given in the problem, γ ≈ 107.7◦ . In general, the coefficients
of the hybrid functions are such that β1∗ β2 = β1∗ β3 = β1∗ β4 = −1/ cos α, where
the index 1 refers to the CC bond, and the other three indices refer to the CH
bonds. In addition, β2∗ β3 =β2∗ β4 = β3∗ β4 = −1/ cos γ. Hence, |β2 |2 = |β3 |2 =
|β4 |2 = −1/ cos γ, and β1 = − cos γ/ cos2 α. In the case of ethane, β1 ≈ 1.53 and
β2 ≈ 1.81.

Answer 4.10.
Nitrogen has five electrons in the second shell, and therefore two of them have
to be in the same orbital state. The other three are available for forming covalent
bonds (Fig. 4.17). As the nitrogen molecule comprises two atoms, three electrons
of each atom form a triple bond like in acetylene. One of those is a σ−bond,
of the sp type, like in Fig. 4.12, and the other two are π−bonds, with the two
p−states perpendicular to this bond (Fig. 4.16). The remaining two electrons in
each nitrogen are in the hybrid state concentrated on the line joining the two nuclei,
on the two far sides of the molecule. Oxygen has six electrons in the second shell,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 283

Crystal bonding 283

Fig. 4.33

four of which are in doubly-occupied states and the other two bond the two atoms
(Fig. 4.17). Therefore the oxygen molecule is planar, with hybridization of the sp
type, in each atom. The double connection between the two oxygens consists of
a σ−bond and a π−bonds, like in ethene (Fig. 4.16), but each hydrogen there is
replaced here by an oxygen state accommodating two electrons. Six of the fluorine
electrons are in three orbital states, which have to be as far as possible from each
other, and the seventh electron occupies the state forming the bond (Fig. 4.17). It
follows that four different directions (three for the doubly-occupied states and one
for the bond) are required. The hybridization on each fluorine atom must be of the
sp3 -type, and the geometry of the F2 molecule is similar to that of methane (Fig.
4.15), where each of the six hydrogens there is replaced here by a doubly-occupied
state.
The diatomic molecules analyzed above are all symmetric under the interchange
of the two nuclei, and therefore lack an electric dipole moment. The next two
molecules do have such a moment.
In the FH molecule there are four directions around the fluorine, one for the
covalent bond and three for the doubly-occupied states (the right panel of Fig.
4.17). As shown there, the three clouds of the doubly-occupied states terminate on
an equilateral triangle which is a base of a triangular pyramid of equilateral edges,
with the hydrogen at its apex. As these clouds are oriented towards the right in the
figure, i.e., in the opposite direction to that of the bond with the hydrogen (which
is oriented to the left) there is a considerable surplus of negative charges on the
right side, which causes this molecule to have an electric dipole moment parallel to
its axis (the line connecting the nuclei). In the ammonia molecule, NH3 , there are
three different directions of the bonds with the hydrogens, and another direction
is needed for the doubly-occupied state of the nitrogen. Therefore, the suitable
hybridization is that of the sp3 −type, like in ethane (Fig. 4.15). The difference is
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 284

284 Introduction to Solid-State Physics

that one of the hydrogens there is replaced here by a doubly-occupied state, and
therefore the angles between the bonds are not equal to the angle between each of
them and the direction of the double state. Here again there is a dipole moment
along the molecule axis. Figure 4.34 displays the electrons’ probability clouds of
this molecule.

Fig. 4.34

Answer 4.11. √
a. The probability clouds of the two functions (ψz ± ψz2 )/ 2 point along the two
directions of the ẑ−axis.
b. Equation (4.24) contains three wave functions that are maximal at the vertices
of an equilateral triangle, as required in Fig. 4.19(a).
c. First, by symmetry considerations |βx | = |βy |. Second, the solution to problem
4.9 shows that the probability clouds of the √ functions ψ200 ± ψx ± ψy point along
the diagonals of the planar square, [±1, ±1]/ 2. These four functions, though, are
not orthonormal to each other (in the quantum-mechanical sense). To make them
orthonormal, one begins with a base of four functions, and adds to it the function
ψxy , which yields maximal probability at the directions of all these diagonals. The
four combinations

Ψ110 = (ψ200 + ψx + ψy + ψxy )/2 ,

Ψ110 = (ψ200 + ψx − ψy − ψxy )/2 ,
Ψ110 = (ψ200 − ψx − ψy + ψxy )/2 ,
Ψ110 = (ψ200 − ψx + ψy − ψxy )/2 ,

are orthonormal to each other and yield maximal probabilities along the diagonals.
They are also normal to the functions of part (a); thus these six functions together
can form a base for bonds of the sp3 d2 −type.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 285

Crystal bonding 285

Answer 4.12.
The average of U is given by a logarithmic
 derivative,
U (R) = −∂{ln dΩ1 dΩ2 exp[−βU ]}/∂β .
Taking R to lie along the ẑ−axis leads to U (R) = (p1 · p2 − 3p1z p2z )/R3 . In
a Taylor expansion, exp[−βU ] ≈ 1 − βU + β 2 U 2 /2 + . . . . As dΩ1 p1n = 0
for any cartesian component n (the positive values cancel the negative ones),
dΩ1 p1n p1m = p21 δnm /3 (since p21x + p21y + p21z = p2 , and the averages of all terms
are identical)
 one finds that
U =0 ,
 
U = [(p1 · p2 )2 − 6(p1 · p2 )p1z p2z + 9p21z p22z ]/R6 = 4p21 p22 /(3R6 ) ,
2

(the integrals are carried out over the spatial angles). It follows from the logarithmic
derivative that U (R) = −4βp21 p22 /(3R6 ).

Answer 4.13.
a. A complete calculation of Eq. (4.31) yields
E = Ψ|Ĥ + Ĥ |Ψ = |A|2 Ψ0 + cΨ1 |Ĥ + Ĥ ||Ψ0 + cΨ1 
 
= |A|2 Ψ0 |Ĥ+ Ĥ |Ψ0 +c∗ Ψ1 |Ĥ+ Ĥ |Ψ0 +cΨ0 |Ĥ+ Ĥ |Ψ1  + |c|2 Ψ1 |Ĥ + Ĥ |Ψ1  .
Symmetry considerations imply that
Ψ0 |Ĥ |Ψ0  = Ψ1 |Ĥ |Ψ1  = Ψ0 |Ĥ|Ψ1  = Ψ1 |Ĥ|Ψ0  = 0 ,
since the integrands of all these expressions are antisymmetric with respect to re-
flection. In addition,
Ψ0 |Ĥ|Ψ0  = 2E1 , Ψ1 |Ĥ|Ψ1  = 2E2 ,
since Ψ0 and Ψ1 are eigenfunctions of Ĥ. Taylor expanding the coefficient, |A|2 ≈
1 − |c|2 , and discarding terms of order three or four in c leads to Eq. (4.31).
b. Equation 4.30) yields
e2
Ĥ = 3 (xA xB + yA yB − 2zA zB ) .
R
It follows that  
e2
Ψ0 |Ĥ |Ψ1  = 3 ψ100 |x|ψ210 2 + ψ100 |y|ψ210 2 − 2ψ100 |z|ψ210 2 .
R
The integrals are products of integrals; they can be carried out within each atom
separately. For the chosen function, ψ100 |x|ψ210  = ψ100 |y|ψ210  = 0 (the wave
functions are independent of the angle φ, which causes the integrals over sin φ and
cos φ to vanish). It remains to consider the term
 ∞  π  2π
e−r/aB e−r/(2aB ) r
ψ100 |z|ψ210  = r2 dr sin θdθ dφ 3/2 √ r cos θ 3/2 √ cos θ
0 0 0 aB π aB 4 2π aB
 ∞
a 256
= √B duu4 e−3u/2 = √ aB ≡ ζaB ,
3 2 0 243 2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 286

286 Introduction to Solid-State Physics

where the following relations are used

 2π  π  ∞
256
dφ = 2π , dθ sin θ cos2 θ = 2/3 , duu4 e−3u/2 = ,
0 0 0 81
(with the variable u = r/aB ). The second term in Eq. (4.32) is
4e4 |ψ100 (r)|z|ψ210 (r)|4 2e4 ξ 4 a4B 16ξ 4 e2 a5B e 2  a B 6
= = = ζ ,
2R6 (E2 − E1 ) R6 Ry (1 − 1/4) 3R6 aB R
with ζ = 16ξ 4 /3 = 0.115.
c. In the general case,
   
Ψ = A Ψ0 + cnm,n  m ψnm (rA )ψn  m (rB ) .
n,n >1 ,m  ,m

The coefficients are obtained by the variational method. The result is

e 2  a B 6
E = 2E1 − ζ ,
aB R
e2   |ψ100 (rA )ψ100 (rB )|3zA zB − rA · rB |ψnm (rA )ψn  m (rB )|2
ζ= 5 .
aB  
En + En − 2E1
n,,m n ,,m

This result is clearly larger than the one obtained before, as all terms in the sum are
positive. The sum can be estimated once En + En in the denominator is ignored.
Then, ζ ≈ 6.

Answer 4.14.
The expansion is based on
1 x 3x2 5x3 35x4
√ =1− + − + + ... .
1+x 2 8 16 128
Upon expanding each the terms in Eq. (4.29), e.g., |R + r|−1 = R−1 [1 + 2z/R +
(r/R)2 ]−1/2 and collecting terms, one finds that the next term in the expansion
(4.30) is
3e2  
[2r A · r B − 5z A z B ](z A − z B ) + r 2
A z B − r 2
B z A + O(rAB
4
/R6 ) .
2R4
Each of the terms here contains a product of the dipole moment on one of the
atoms with a quadrupole moment on the other. These terms are antisymmet-
ric with respect to reflection, and therefore their quantum averages vanish in the
symmetric ground state. As in the calculation leading to Eq. (4.32), one has
to include in the variational wave function excited states; the correction is then
quadratic in the coefficient of the additional wave function. The contribution to
Eq. (4.32) is of the order of (aB /R)8 . As is always the case in variational calcula-
tions, the contribution to the ground-state energy is negative. It is a small correction
to the leading order and therefore is frequently ignored. Note that higher orders in
the multipole expansion yield corrections of the order of (aB /R)10 , (aB /R)12 , etc.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 287

Crystal bonding 287

Answer 4.15. √ 3 √
For the FCC lattice, a = R/ 2; hence v = a3 /4 = R / 2. Inserting these into
Eqs. (4.10) and (4.11) yields
B = 4(/σ 3 )A12 (A6 /A12 )5/2 = 75/σ 3 .

Answer 4.16.
a. The potential created at a point R due to charges qi located at ri can be
expanded as
 qi 1  qi

=  
i
|ri − R | R i 1 + (ri /R ) − 2ri · R /R2
 2

1   r · R 3(ri · R )2 − ri2 R2 

=  qi 1 + i 2 + + . . .
R i R 2R4
  2
Q P · R αβ Qαβ (3Rα Rβ − δαβ R )
= + + + ... .
R R3 6R5
In the first stage the potential is expanded in the charges’ coordinates; the total
charge, Q = i qi , the dipole moment, P = i qi ri , and the components of the
quadrupole-moment tensor, Qαβ = i qi (3riα riβ − δαβ ri2 ), are introduced in the
second stage. Next, the quantum-average over the coordinates of the electrons in
the molecule is carried out, with weights determined by the absolute-value squared
of their wave functions. Due to the cylindrical symmetry assumed in the problem,
this average yields xi  = yi  = 0, and xi yi  = zi xi  = yi zi  = 0. In addition,
x2i  = yi2  = ri2 − zi2 /2. Therefore, the dipole moment has at most a component
along the molecule axis, and the quadrupole moment obeys the relations given in
the problem.
b. The hydrogen molecule is also invariant to the inversion of r (the coordinate of
the electron, see Fig. 4.8 and the discussion following it); therefore the probability
to find the electron near any of the nuclei is the same, and the dipole moment
vanishes.

Answer 4.17.
a. Oxygen has two valence electrons; both participate in a double bond with the
carbon, built of a σ−bond and a π−one. Therefore both the oxygen and the carbon
have two additional hybrid states per each, that are farther away from the bond
and from one another, available for a planar hybridization of the sp2 −type, as in
Fig. 4.13 or in Eq. (4.25). The remaining two states in the oxygen are occupied by
two electrons each, while in the carbon they take part in covalent bonds with the
hydrogens. The result resembles the planar ethene molecule, Fig. 4.16, with one of
the carbons replaced by the oxygen, and the two adjacent hydrogens replaced by
clouds containing two electrons each.
b. The answer is in Fig. 4.35. The solid lines represent covalent bonds, the double
ones are double bonds, and the dashed lines are hydrogen bonds. The carbons build
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 288

288 Introduction to Solid-State Physics

a rectangular-centered lattice. It is not the only possible structure: for instance,

one may build a planar lattice in which the molecules in neighboring columns are
normal to each other (and not parallel, as in Fig. 4.35).

Fig. 4.35

Answer 4.18. The three spins are written in spherical coordinates,

Si = {sin θi cos φi , sin θi sin φi , cos θi }.

Without loss of generality, one of the spins, say 1, may be chosen to be S1 = {0, 0, 1}.
The energy of the triangle is

E = |J|[cos(θ2 ) + cos(θ3 ) + sin(θ2 ) sin(θ3 ) cos(φ2 − φ3 ) + cos(θ2 ) cos(θ3 )] .

As 0 ≤ θi ≤ π, it follows that sin θ2 sin θ3 is positive, and thus the minimum is

reached for cos(φ2 − φ3 ) = −1, that is, φ2 = π + φ3 . Hence, the spin components
normal to the ẑ−axis are antiparallel to each other. The energy is then

E = |J|[cos(θ2 ) + cos(θ3 ) + cos(θ2 + θ3 )] .

Differentiating with respect to θ2 and to θ3 gives

sin(θ2 ) = − sin(θ2 + θ3 ) = sin(θ3 ) .

The equality sin(θ2 ) = sin(θ3 ) has two solutions: θ2 = θ3 and θ2 = π − θ3 . The

energy corresponding to the second one is E = −|J|; the energy of the first is
E = |J|[2 cos(θ2 ) + 2 cos(2θ2 )], which is extremal for

2 sin(θ2 ) + 2 sin(2θ2 ) = 2 sin(θ2 )[1 + 2 cos(θ2 )] = 0 ,

with the solutions θ2 = θ3 = 0 or π, or cos θ2 = −1/2, implying θ2 = θ3 = 2π/3.

The energy of the first two solutions is E = 3|J| (a ferromagnetic state) or E = −|J|
(a frustrated state with two parallel spins and the third antiparallel to them). The
energy of the last solution is E = −3|J|/2, the smallest possible energy of the
problem. In this solution S1 · S3 = S1 · S2 = −1/2, and the corresponding angles
are indeed θ2 = θ3 = 2π/3. Similarly, S2 · S3 = cos(θ2 + θ3 ) = −1/2 and hence the
angle between these two spins is also 120◦ .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 289

Crystal bonding 289

Answer 4.19.
The simplest cases are those for which J2 = 0. The material is then a ferromagnet
(all spins are parallel) for J1 > 0, and a simple antiferromagnet where all neigh-
boring spins are antiparallel, when J1 < 0. When J2 > 0, both arrangements are
benefitted, as in both the next-nearest neighbors are parallel to each other; the
total energy is E = −N (|J1 | + J2 ). The case J2 < 0 is more subtle, since the
interaction between next-nearest neighbors (that requires them to be antiparallel)
competes with the nearest-neighbor interaction that prefers them to be parallel. In
any event, it cannot be that the two types of bonds, between nearest neighbors and
between next-nearest neighbors, will have their optimal energy. This is the situation
called frustration in Sec. 2.10. To treat it, one may begin by presuming that the
spins laying in the XY −plane, and that the angle between nearest neighbor spins is
ϑ, i.e., Sn = (cos[nϑ], sin[nϑ]). The total energy is E = −N [J1 cos(ϑ) + J cos(2ϑ)],
where N is the number of spins in the system. Differentiating with respect to ϑ
gives
J1 sin(ϑ) + 2J2 sin(2ϑ) = sin(ϑ)[J1 + 4J2 cos(ϑ)] = 0 .
Hence, either sin(ϑ) = 0 or cos(ϑ) = −J1 /(4J2 ). The second solution exists provided
that |J1 /(4J2 )| < 1, and then the corresponding energy is E2 = N J2 [1+(J1 /J2 )2 /8].
The energy of the first solution is E0 = −N [J1 + J2 ] for a ferromagnetic struc-
ture (ϑ = 0), or Eπ = N [J1 − J2 ] for an antiferromagnetic one (ϑ = π). One
easily convinces oneself that in its range of existence, the energy of the solu-
tion cos(ϑ) = −J1 /(4J2 ) is lower than that of the other two solutions: E2 =
−N |J2 |[1 + (J1 /J2 )2 /8] < −N |J2 |(|J1 /J2 | − 1), when |J1 /J2 | < 4. The two en-
ergies are equal for |J1 /J2 | = 4; when |J1 /J2 | > 4, the second solution does not
exist at all, and therefore either the ferromagnetic state (for J1 > 0), or the antifer-
romagnetic one (for J1 < 0) prevails. The dependence of ϑ on the ratio j = J1 /|J2 |
in the optimal situation (of the ground state) is depicted in Fig. 4.36. In the inter-
mediate state, described by the second solution, the spins along the lattice rotate
around the ẑ−axis normal to the XY − plane of the spins. When that plane is per-
pendicular to the one-dimensional lattice, then the edge-points of the spins create a
screw that advances along the lattice. In most of the lattice points the ratio ϑ/π is
not rational, and therefore this is an incommensurate pattern: the magnetic unit
cell is infinite (see Sec. 2.9). For j < −4 the magnetic structure is antiferromagnetic
and the magnetic unit cell coincides with the atomic one. At j = ±4 there occur
continuous phase transitions, in which the angle ϑ loses its dependence on j.
For the Ising model, Eq. (4.47), the incommensurate solution is not valid. Again,
the subtle configuration occurs for J2 < 0. When J1 > 0, the nearest neighbors
“wish” to be parallel, and the next-nearest ones “wish” to be antiparallel. For
strong enough J1 , the arrangement is ferromagnetic with the total energy −N J1 .
For large enough J2 , the next-nearest neighbors arrange themselves successively in
opposite directions, creating the pattern + + − − + + − − + + − − + + . . . , in
which the next-nearest neighbors are antiparallel (the first spin in the line can be
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 290

290 Introduction to Solid-State Physics

Fig. 4.36

“up” or “down” without affecting the energy). The total energy is −N |J2 | as half
of the nearest-neighbor pairs yields J1 , and the other half gives −J1 . The magnetic
unit cell is four times larger than that of the original one. The transition between
these two solutions occurs when J1 = |J2 |. Alternatively, when J1 < 0 then for
|J2 | < |J1 | the simple antiferromagnetic state is realized, − + − + − + − + . . . , but
for |J2 | > |J1 | again appears the state − − + + − − + + − − + + . . . . Therefore,
for J2 < 0 the simple antiferromagnetic state prevails over the entire range j < −1,
and the ferromagnetic state is the ground state in the range j > 1, and in-between
there is the new state − − + + − − + + − − + + . . . .

4.9 Problems for self-evaluation

s.4.1.
The following values have been measured for the binding energy, the distance between
nearest neighbors, and the bulk modulus of table salt: u = 7.9eV, R = a/2 = 2.82Å,
B = 2.4 × 1010 N/m2 .
a. Use these values to identify the parameters m and C in Eq. (4.8) and check
whether they are not self-contradictory.
b. What is the pressure one should apply in order to squeeze table salt as to reduce
its volume by 1% ?

s.4.2.
a. Assuming that the ions A, B, and O of the perovskite structure shown in
Fig. 2.21 are spheres of radii rA , rB , and rO , respectively, and denoting the lattice
constant by a, obtain five inequalities that these radii have to obey.
b. With the assumption that the spheres of A and O are touching each other, and
so are also the spheres of B and O (why is this advantageous?) find the range of
allowed values for the ratios xA = rA /rO , xB = rB /rO .
c. What happens when the radii are not in that range? Is the perovskite structure
still possible?
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 291

Crystal bonding 291

d. In the barium titanate compound, a ≈4Å, and rA ≈ rO ≈ 1.4Å. Which of the

conditions in part (a) becomes (approximately) an equality? What can be deduced
about the radius of the titanium ion?
e. Explain how to derive the Coulomb energy of this crystal.

s.4.3.
Assuming that the lattice constant of a certain alkali halide is determined by the radii
of the ions, and that the binding energy is given by Eq. (4.9) with the repulsion UP 2
of Eq. (4.8) with a constant ρ, find the value of λ. How does the repulsive potential
vary as a function of the ions’ radii?

s.4.4.
Sulphur appears in nature in several forms (allotropes); several of those are lattices
(orthorhombic, rhombohedral, monoclinic, and more), with a base that comprises
the molecule S8 or the molecule S12 , depicted in Fig. 4.37.
a. Determine the type of the bond between two neighboring atoms in these molecules.
Find approximately the value of the angle between neighboring bonds in the molecule.
b. Do the molecules possess electrons that are free to move around them, like in
benzene?
c. Occasionally (though rarely) sulphur appears in the molecular form S2 . What is
the structure of this molecule?
d. There are six possible forms of the molecule S4 ; in one of the them the atoms
build a planar square. What is the electronic configuration of this molecule?
e. The left panel of Fig. 4.37 displays two versions of the structure of the polymer
S∞ , which creates a spatial helix. What is the type of the bond along the molecule?
Does the molecule conduct electricity? What is the lattice constant of the one-
dimensional lattice it creates?
f. What are the types of the bonds between the molecules in crystals built of the
molecules mentioned in the previous parts?
g. In rare cases it is possible to grow a lattice of sulphur atoms. Which types of
bonds can be formed between the atoms in the lattice? Propose lattice structures
based on these bonds.

Fig. 4.37
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 292

292 Introduction to Solid-State Physics

s.4.5.
a. Propose a general expression for the van der Waals binding energy of the
rectangular lattice, and show that it depends solely on the ratio of the two lattice
constants, x = b/a. For which value of x is it maximal? Is it possible to extend the
expression to the family of simple three-dimensional orthorhombic lattices? What
can be deduced about the symmetry of simple molecular crystals?
b. Find the expression for the binding energy due to van der Waals interactions
in the oblique planar lattice, which has equal lattice vectors with an arbitrary angle
between them.

s.4.6.
The potential between pairs of molecules on a fictitious BCC lattice is described by
the Morse function

Utot (R) = N U0 (exp[−2(R − σ)/r0 ] − 2 exp[−(R − σ)/r0 ]) ,

where R is the distance between nearest neighbors. Find the distance between nearest
neighbors at equilibrium, R, and calculate the binding energy and the bulk modulus.

s.4.7.
The structure of the ammonia molecule, NH3 , is discussed in problem 4.11. Based
on this structure, determine the type of the bonds among these molecules. Propose
a crystal structure for a solid built of such molecules.

s.4.8.
The acetylene molecule, C2 H2 is linear, with a triple bond between the carbons (see
the discussion before Fig. 4.16). Can a hydrogen bond be formed between the hy-
drogen in one molecule and the π−states perpendicular to a neighboring molecule?
Assuming that there are such bonds, describe the three-dimensional crystal that can
be created. Discuss the hydrogen bonds based on a classical Coulomb interaction and
on a quantum hybridization.

s.4.9.
Like in problem 2.35, the magnetic moments of spins on a square lattice are ex-
pressed in the form μ(R) = μ0 cos(Q · R). The magnetic Hamiltonian includes the
exchange interaction J1 between nearest neighbors (nn), and the exchange interac-
tion J2 between next-nearest neighbors (nnn). Determine the magnetic energy as
a function of Q, and find the magnetic order, i.e., the spin configuration, in the
ground state, as a function of the exchange energies.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 293

Crystal bonding 293

4.10 Answers for the self-evaluation problems

Answer s.4.1.
4
a. Changing the units, B = 0.15eV/(Å)3 . Hence, m = 1 + 18R B/(αe2 ) ≈ 7.78
[using the relations 1 Rydberg=e2 /(2aB ) ≈ 13.6eV, aB ≈ 0.53Å, and αNaCl =
3
1.747565]. It follows that u = (1−1/m)αe2 /R = 18R /m ≈7.8 eV, in fair agreement
m−1
with the measured binding energy. In addition, zC = R αe2 /m ≈ 3660eV(Å)m ,
with z = 6.
b. By Eq. (4.10) and the discussion preceding it, the pressure is given by p =
v
− v0 dvB(v)/v, when the specific volume changes from v0 to v = xv0 . Since B =
4
(m − 1)αe2 /(18R ) = B0 (v0 /v)4/3 , one finds
 v/v0
p = −B0 d(v/v0 )(v0 /v)7/3 = 3B0 (x−4/3 − 1)/4 .
1

Substituting x = 0.99 and the given bulk modulus yields p = 2.4 × 108 N/m2 .

Answer s.4.2. √
a. The distance
√ between two centers of neighboring
√ O ions is a/ 2, and hence
2r√O < a/ 2. Similarly, 2rA < a, 2(rA + rO ) < a 2, 2(rB + rO ) < a, 2(rB + rA ) <
a 3.
b. The Coulomb energy increases as the positive ions approach their negative
neighbors. Thus, this energy is maximal when the positive ions touch the oxygen √
ions, and consequently two of the inequalities become equalities: 2(rA + rO ) = a 2
and 2(rB + rO ) = a. Substituting the expressions √for a in the remaining inequalities
√
yields six inequalities. The inequality
√ 2r O < a/ 2 gives xA > 1 and xB > 2 − 1.
Similarly, 2rA√ < a gives x√ A < 2 + 1 and √ xA < xB + 1,√and the third condition
gives xA < ( 3 − 1)xB + √ 3 and xB < ( 6/2 − 1)xA + 6/2. √It follows that xA
√ range 1 < xA√< 2 + 1, and xB is in the range max{ 2 − 1, xA − 1} <
is in the
xB < ( 6/2 − 1)xA + 6/2.
c. In this case one of the conditions for the spheres to touch each other is abandoned,
and a structure where the opposite-sign ions are as close as possible is to be sought
for. The five conditions of part (a) leave room for many possibilities.
d. The numbers given in the problem show that (approximately) the barium ions
are touching the oxygen ones, and the oxygen ions are touching each other. The
remaining three conditions require that rTi < a/2 − rO ≈ 0.6Å. Figure 4.6 indeed
shows that the titanium ion is smaller then the barium and the oxygen ions.
e. The Coulomb energy is UC = (N/2) i∈cell qi φi , where φi = j =i (qj /Rij ) is
the electric potential at i due to all other ions [see Eq. (4.7)]. Thus one has to find
the three potentials for the ions A, B, and O, to obtain UC = (N/2)(qA φA + qB φB +
3qO φO ). In the case of the barium titanate compound, the required potentials are
for the barium, for which qA = +2, for the titanium (qB = +4) and for the oxygen
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 294

294 Introduction to Solid-State Physics

(qO = −2). Denote by R = a/2 the distance between the Ti and the O (this is the
shortest distance between opposite-sign charges), and measure distances
√ in units of
R. Each barium  ion is surrounded by 12 oxygens
√ at a distance 2/2, 24 oxygens
at a distance 3/2, 24 oxygens at a distance 10/2, etc. √ It is also surrounded
√ by
6 barium ions at a distance 1, 12
√ bariums at a distance
√ 2, eight at 3, etc., and
also by 8 Ti ions at a distance 3/2, 24 Ti’s at 11/2, etc.. The electric potential
on the barium is
 qj  12 24 24 
φBa = = −2 √ + + √ + ...
j
R0j ( 2/2) 3/2 ( 10/2)
6 12 8   8 24 
+2 + √ + √ + ... + 4 √ + √ + ... .
1 2 3 ( 3/2) ( 11/2)

Answer s.4.3.
The maximal binding energy is achieved when zλ = (αq 2 /ρ) exp[R/ρ](ρ/R)2 (see
the solution to problem 4.6). The function exp[x]/x2 has a minimum at x = 2. One
generally expects that R/ρ > 2, and in this range zλ increases exponentially with
R. As mentioned, usually R = r< + r> , and thus zλ increases exponentially with
the sum of the radii.

Answer s.4.4.
a. The ground configuration of the sulphur atom is similar to that of oxygen, 3s2 3p4 ;
as in oxygen, a possible hybrid state is 3s1 3p5 , which contains four states of the type
given in Eq. (4.26) and displayed in Fig. 4.14. Two of these states are each doubly-
occupied, and the other two have a single electron and can form covalent bonds like
in Fig. 4.17 and similarly to the bonds of oxygen with the hydrogens in water –
Fig. 4.18. In the symmetric case, Fig. 4.14, the angles between neighboring bonds
is 109.5◦ . In water, the presence of extra negative charges on the doubly-occupied
states reduces the angle to 104.5◦ . Since sulphur resembles oxygen, its atom too
has four bonds of the type given in Eq. (4.26), two are doubly-occupied and two
create covalent bonds, e.g., with the neighboring sulphur atoms in the molecules
mentioned in the problem. One therefore expects the angle to be of the same order
of magnitude as given above, about 105◦ . In fact, this angle is 106◦ . In order to
close the loop in the S8 molecule, the atoms have to move up and down successively,
forming the zig-zag structure displayed in the figure, with four sulphur atoms in each
plane. The same structure could have been realized for the S12 molecule. However,
the Coulomb repulsion between the charges in the doubly-occupied states, which
are located opposite to the covalent bonds, causes this molecule to adopt another
structure in which there are six sulphur ions in the intermediate plane, and three
atoms in the planes below and above that plane.
b. As the six electrons of each sulphur atom are in states localized near that atom
(whether in doubly-occupied states or in a covalent bonds) there are no electrons
free to move around the ring.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 295

Crystal bonding 295

c. There are only two sulphur atoms in S2 and thus its configuration is similar
to that of the O2 molecule. Each sulphur (or oxygen) has six electrons in the
outer shell; all in all there are four orbital states (one s and three p). Two of
the orbital states are doubly-occupied, and the other two (each accommodating a
single electron) can form covalent bonds. In the previous example, similar to the
water molecule, each atom has two covalent bonds along different directions, and
therefore it is expedient to exploit the states given in Eq. (4.26). In the linear
diatomic molecule both bonds have to be along the same direction, and therefore
the bond is a double one. In this situation it is better to use the basis given in
Eq. (4.25), depicted in Fig. 4.13. Two of the planar states are occupied by two
electrons each and form an angle close to 120◦ with one another, and the third state
forms a σ−bond with the other atom. The additional bond between the atoms is
a π one, coupling the states perpendicular to the plane. The result resembles the
ethene molecule, Fig. 4.16, with the hydrogens there replaced by doubly-occupied
sulphur states.
d. As explained in part (d) of problem 4.10, one may assume arbitrary angles
between the four hybrid states given by Ψi = Ai [ψ200 + βi n̂i · ψ i ]. If the sulphur
atoms are arranged on a square, then one of the angles, (say between the states 1
and 2) is 90◦ , and therefore β1 = β2 = ∞ and |A1 |2 = |A2 |2 = 1/(1 + |β1 |2 ) = 0.
These states are then ψx or ψy . From the scalar product of each of the other
functions with the first two it √ follows that n̂3 and n̂4 are normal to n̂1 and n̂2 ,
and hence Ψ3,4 = [ψ200 ± ψz ]/ 2. The first two states are exploited for covalent
bonds, and therefore the remaining two – which point along the positive and the
negative directions of the ẑ−axis – contain each two electrons. The binding energy
of this configuration is not as high as in the previous cases, because of the larger
proximity of the electrons’ clouds. Alternatively, it is possible to excite the sulphur
to the configuration sp3 d2 , and then the emerging bonds are with right angles, as
displayed in Fig. 4.19. In this configuration there remain un-paired electrons, free
to move around the square.
e. In the polymer molecule of sulphur each atom may retain the configuration sp3 ,
as described in part (a). Therefore, the angle between two neighboring bonds is
approximately 106◦ . Given two successive bonds like this, the third one has to form
an identical angle with its predecessor; however, the atom at the edge can rotate in
a plane perpendicular to the bond. Its final location is determined by the repulsion
between the electrons’ clouds in the two states that do not take part in the bond and
the other electrons. This repulsion turns out to be minimal for the structures shown
in Fig. 4.37, which differ from one another in their “screw” direction. According
to various measurements and calculations, the third bond forms an angle of about
85.3◦ with the plane spanned by its two predecessors; in this way the screw structure
that characterizes the helix is created. As seen in Fig. 4.37, the unit cell contains
ten atoms. Like in part (a), the electrons are localized close to the sulphur atoms,
and hence there is no electric conductivity along the molecule.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 296

296 Introduction to Solid-State Physics

g. As all molecules are electrically neutral and have no electric dipole moment, the
interactions among them are of the van der Waals type.
f. To build a lattice it is needed that each atom will be bound to at least three
other atoms (like in graphene). For a three-dimensional bond even more neighbor-
ing atoms are needed. As the configurations 3s1 3p5 or 3s2 3p4 allow for no more
than two bonds per atom, it is clear that a lattice structure requires excitations to
higher hybridization states, e.g., 3s1 3p4 3d1 or 3s1 3p3 3d2 . The latter configuration
allows for six bonds normal to one another, see Fig. 4.19. Locating sulphur atoms
at the end point of such a bond yields a simple tetragonal lattice. The first config-
uration requires that one of the five states [three in the plane, Eq. (4.25), and two
perpendicular to the plane, e.g., ψz2 ± ψz , see the solution to problem 4.11] will
be doubly occupied, and therefore there remain only four states capable of forming
bonds. When the double occupancy fills the states perpendicular to the plane, the
resulting lattice is a hexagonal one, as in the case of graphene.

Answer s.4.5.
a. The binding energy is given by extending Eq. (4.33),
Utot = 2N [F12 (x)(σ/a)12 − F6 (x)(σ/a)6 ] ,
2 2 2 −n/2
with Fn = n1 ,n2 (n1 + n2 x ) . The minimum of Utot yields u =
−Utot (min)/N = F6 /(2F12 ), and indeed the result depends only on the ratio
2

x. When the two directions in the lattice are interchanged, that is, the dummy
indices in the sum of Fn are interchanged, then Utot = 2N [F12 (1/x)(σ/b)12 −
F6 (1/x)(σ/b)6 ], and hence u(x) = u(1/x). The derivative with respect to x gives
u (x) = −u (1/x)/x2 , and thus u (1) = 0: the function has an extremum at x = 1.
Numerical test shows that this is indeed a maximum: the sum over |n1 |, |n2 | ≤ 3
gives
47449 47449 4 4 4 4
F6 (x) = + + + + +
23328 23328x6 (1 + x2 )3 (4 + x2 )3 (9 + x2 )3 (1 + 4x2 )3
4 4 4 4 4
+ 2 3
+ 2 3
+ 2 3
+ 2 3
+ + ... ,
(4 + 4x ) (9 + 4x ) (1 + 9x ) (4 + 9x ) (9 + 9x2 )3
2177317873 2177317873
F12 (x) = +
1088391168 1088391168x12
4 4 4 4
+ + + +
(1 + x2 )6 (4 + x2 )6 (9 + x2 )6 (1 + 4x2 )6
4 4 4 4 4
+ + + + + + ... .
(4 + 4x2 )6 (9 + 4x2 )6 (1 + 9x2 )6 (4 + 9x2 )6 (9 + 9x2 )6
Additional terms almost do not change the result. The right panel in Fig. 4.38
displays the dependence of u on x (in units of /2). It is seen that the square lattice
with x = 1 has the maximal binding energy. A heuristic way to understand this
result is to note that the sums are dominated by their first term. The atom at the
origin has two nearest neighbors at a distance a, and two at a distance b, and hence
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 297

Crystal bonding 297

Fn ∝ 2(1/an + 1/bn ). In this approximation u = F62 /F12 = (1 + y)2 /(1 + y 2 ),

where y = 1/x6 . This function is maximal for y = 1. For the orthorhombic lattice,
denote y = b/a and z = c/a, and then
u = −Utot (min)/N = F62 (y, z)/[2F12 (y, z)] ,
where Fn (y, z) = n1 ,n2 ,n3 (n21 + n22 y 2 + n23 z 2 )−n/2 . Then u(y, z) = u(1/y, z/y) =
u(y/z, 1/z), and hence
uy (y, z) = −[uy (1/y, z/y) + zuz (1/y, z/y)]/y 2 ,
uz (y, z) = −[yuy (y/z, 1/z) + uz (y/z, 1/z)]/z 2 ,
yielding uy (1, 1) = −[uy (1, 1) + uz (1, 1)], and uz (1, 1) = −[uy (1, 1) + uz (1, 1)], i.e.,
uy (1, 1) = uz (1, 1) = 0. Again, a numerical check shows that y = z = 1 is a
maximum. The number of nearest neighbors at the same distance is maximal for
the cubic lattice (as compared with the orthorhombic one), and thus the binding
energy is higher for this more symmetric lattice.

Fig. 4.38: Left panel: the binding energy of the oblique lattice; right panel: the
binding energy of the rectangular lattice.

b. For the oblique lattice, the base vectors are a1 = ax̂, a2 = a cos γ x̂ + a sin γ ŷ,
and therefore Fn (γ) = n1 ,n2 (n21 + n22 + 2n1 n2 cos γ)−n/2 . The left panel in Fig.
4.38 shows that the binding energy is maximal for cos γ = 1/2, i.e., for the trian-
gular lattice. Indeed, this is the densest planar lattice. In three dimensions, most
materials with van der Waals interactions are arranged on an FCC or a HCP
lattice, which are the densest lattices in three dimensions. Indeed, when only the
neighbors next to the origin are taken into account, the atom at the origin has
four neighbors at a distance a, and two at a distance 2a sin(γ/2). In the crudest
approximation, Fn = 4/an + 2/[2a sin(γ/2)]n , and hence u ∝ (4 + 2y)2 /(4 + 2y 2 ),
where y = 1/[2 sin(γ/2)]6 . The maximal binding energy is now obtained for y = 1,
i.e., γ = 60◦ .

Answer s.4.6.
The potential is extremal for
∂Utot (R) N U0  −2(R−σ)/r0 
= −2 e − e−(R−σ)/r0 = 0 ,
∂R r0
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 298

298 Introduction to Solid-State Physics

and therefore R = σ. At equilibrium, the binding energy is u = −Utot (R)/N = U0 .

√ 3 3 √
In a BCC lattice, R = a 3/2, and thus v = cR = 8R /(3 3). Making use of
Eqs. (4.10) and (4.11) yields

∂ 2 u  U0
B = −v  = √ .
∂v 2 R=σ 4 3σr0

Answer s.4.7.
Each nitrogen ion is surrounded by three covalent bonds with the hydrogens, and a
single cloud of a doubly-occupied state. The four states are described by hybridiza-
tions of the sp3 −type. Along each line that joins the nitrogen with a hydrogen
it is possible to create a hydrogen bond with the doubly-occupied state near a
neighboring nitrogen. In a similar fashion, along the line that accommodates the
doubly-occupied state it is possible to form a hydrogen bond with a hydrogen from
a neighboring molecule. Thus, each nitrogen ion creates four hydrogen bonds with
four neighboring nitrogens, with the angles between the bonds that are similar to
those of the diamond lattice. Like in the solution of problem 4.17, it can be as-
sumed that when the crystal begins to grow from a single molecule, there is only
one pattern which will be eventually formed. Such tetrahedral patterns exist in
the diamond lattice and in the wurtzite one. Experimentally, it is found that the
ammonia lattice has a cubic symmetry, like the diamond lattice.

Answer s.4.8.
Consider first a single molecule, H−C≡C−H. The triple bond comprises a σ−bond
on the line that joins the carbons, e.g., along the x̂−axis, and two π−bonds per-
pendicular to the molecule, one along ŷ and another along ẑ. The XZ plane then
contains a cloud of electrons’ probability around the ẑ−axis, normal to the mid-
point of the molecule (see Fig. 4.16). This negatively-charged cloud can attract
the positive hydrogen ion located at the edge of another molecule, and therefore
this other molecule lies along the line normal to the molecule in the XZ plane.
Figure 4.39 displays this plane. Each thick line corresponds to a certain molecule;
another molecule lies on the normal line, such that its hydrogen is close to the cloud
of π−electrons along the same direction. A similar cloud is located on the other
side of the molecule, and hence another molecule places itself there. The result is
a planar lattice: the thin lines describe the square unit cell, which contains four
molecules (at the lower left corner, at the mid-point of the edges close to it, and
at the center of the square). The primitive unit cell is a square rotated by 45◦ ,
with a base that contains two neighboring molecules, one horizontal and the other
vertical. The crystal can be extended along the ŷ−direction. Each of the molecules
in the plane possesses another π−bond, normal to the plane at the mid-point of
the molecule. Placing there another molecule yields a plane of molecules along the
ŷ−axis. Above this plane it is possible to locate the original plane, and so forth.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 299

Crystal bonding 299

The result is a simple cubic lattice, with a base of eight molecules in the unit cell
(four in the lower plane and four in the plane above).

Fig. 4.39

So far the molecules are arranged such that energy is gained from the Coulomb
attraction between the hydrogen ion and the negatively-charge electrons’ clouds.
However, recall that each π−bond in the plane is built of two functions of the
ψz −type, from each of the carbons. Thus, in the molecule placed on each side of the
ẑ− axis there are two quantum states yielding an sp−bond between the hydrogen
and the carbon. As a result there are six wave functions along this direction, two
in each molecule. Six hybrid states can be formed out of those. Each molecule
can also contribute two electrons: two from the sp−bond on the perpendicular
molecules, and two from the π−bond. These six electrons can occupy the three new
hybrid states of the lower energies, and thus form a covalent bond between the three
molecules. The emerging bond is a joint one of an ionic bond in which the electrons
remain within the molecule, and a covalent bond shared by all three molecules.

Answer s.4.9.
The nearest neighbors are located at R = ±ax̂ ± aŷ and the next-nearest neighbors
at R = ±a(x̂ + ŷ), ±a(x̂ − ŷ). Hence the magnetic energy is

E = −N {J1 cos(Qx a) + cos(Q)y a)] + J2 [cos(Qx a + Qy a) + cos(Qx a − Qy a)]}

= −N {J1 [cos(Qx a) + cos(Qy a)] + 2J2 cos(Qx a) cos(Qy a)} ,

where N is the number of bonds (twice the number of sites) in the lattice. Differ-
entiating with respect to the components of the wave vector yields

sin(Qx a)[J1 + 2J2 cos(Qy a)] = 0 , sin(Qy a)[J1 + 2J2 cos(Qx a)] = 0 .

These equations have five solutions:

a. a ferromagnet, Qx a = Qy a = 0, with energy E(00) = −2N (J1 + J2 );
b. an antiferromagnet along the x̂−direction, Qx a = π, Qy a = 0, with energy
E(π0) = 2N J2 ;
c. an antiferromagnet along the ŷ−direction, Qy a = π, Qx a = 0, with energy
E(0π) = 2N J2 ;
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 300

300 Introduction to Solid-State Physics

d. an antiferromagnet along the diagonal (11), Qx a = Qy a = π, with energy

E(ππ) = 2N (J1 − J2 );
e. an incommensurate antiferromagnet along the diagonal, cos(Qx a) = cos(Qy a) =
cos ϑ = −J1 /(2J2 ), which can exist only for |J1 /(2J2 )| ≤ 1, with energy E(ϑϑ) =
N J12 /(2J2 ).
A plot of E/(2N J1 ) as a function of J2 /J1 , for J1 > 0, reveals that the ground
state is ferromagnetic for J2 > −J1 /2, or antiferromagnetic along one of the axes
when J2 < −J1 /2. Similar plots of E/(2N |J1 |) as a function of J2 /|J1 | for J1 <
0 show that the ground state is a simple antiferromagnet along the diagonal for
J2 > −|J1 |/2, or an antiferromagnet along one of the axes for J2 < −|J1 |/2. Figure
4.40 displays the regions in parameter space, where in each region there appears
the vector (Qx a, Qy a), that indicates the magnetic order.

Fig. 4.40
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 301

Appendix C

Topics in quantum mechanics

Wave functions and Hilbert space. A quantum state is described by a complex

wave function, Ψ, which is a function of the spatial coordinates of the particles in
the system, and depends on time. The focus of this book is on stationary situations,
and therefore the time dependence is omitted. Thus, the quantum state of a single
particle in a three-dimensional space is described by the function Ψ(r), where r is
the particle’s location in space; |Ψ(r)|2 d3 r is the probability to find the particle in
the volume element d3 r around r. When there are many particles in the system,
Ψ depends on the locations of all particles and |Ψ|2 is the probability density to
find the particles at these locations. Borrowing the terminology of linear algebra,
all (allowed) wave functions form a “linear space”: if Ψ1 and Ψ2 are two wave
functions that describe the system, then their linear combination, β1 Ψ1 + β2 Ψ2 ,
with complex coefficients β1 and β2 , is a possible wave function as well, and is also
a member of that linear space, which is called the Hilbert space. This addition
of the wave functions reflects the wavy nature of the quantum system: the linear
combination represents the interference of the two waves described by the two
wave functions. The scalar product of two wave functions is Ψ1 |Ψ2  ≡ Ψ∗1 Ψ2 ,
where the integration is carried out over all coordinates on which the wave functions
depend. When they refer to a single particle, the integration is over the entire three-
dimensional space in which the particle exists. Since the sum of the probabilities
over this space is 1, it follows that the wave functions are normalized, i.e., Ψ|Ψ ≡
|Ψ|2 = 1. (For a single particle, the integration is over the volume which contains
the particle.)
The Schrödinger equation and the eigenenergies. A wave function is
a solution of the Schrödinger equation. The time-independent functions solve
ĤΨn = En Ψn , where Ĥ is the Hamiltonian operator of the system. (The “hat”
symbol is used for a quantum operator.) The index n, i.e., the energy quantum
number, denotes the different solutions. The Hamiltonian operator of a single par-
ticle, for instance, is given in Eq. (4.12). Each of the solutions Ψn is character-
ized by an eigenvalue of the equation, En . A measurement of the system energy
yields one of these eigenvalues. The corresponding wave function Ψn is also termed
“eigenstate” (or eigenfunction) of the energy. The set of wave functions is or-

301
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 302

302 Introduction to Solid-State Physics

thonormal, i.e., Ψn |Ψm  = δnm , where δnm is the Kronecker delta-function. In
many cases the energy levels are degenerate, i.e., a certain energy is an eigenenergy
of several eigenstates. In such a situation, the degenerate eigenstates are distin-
guished by additional quantum numbers. The ensemble of eigenstates in a basis
for the Hilbert space, which includes all possible wave functions of the system.
As is the case of a cartesian coordinate system, the basis vectors are normalized
(i.e., they are unit vectors) and are orthogonal to each other. This means that
∞
an arbitrary state of the system can be written as Ψ = n=0 an Ψn , with complex
coefficients an . As the wave function is normalized, Ψ|Ψ ≡ |Ψ|2 = 1, and the set
∞
of eigenfunctions is orthonormal, Ψn |Ψm  = δnm , it follows that n=0 |an |2 = 1
(check!).
Free particles. The √eigenstates of a free particle confined to a space of volume
V are Ψk = exp[ik · r]/ V , where V is the system volume and the wave functions
obey periodic boundary conditions. The corresponding eigenenergies are E(k) =
2 k 2 /(2m). By the Fourier theorem (see Appendix B), these functions are indeed
a basis for all functions in space. Thus, the Fourier theorem proves the assertion
stating that the solutions of the Schrödinger equation are a basis of the Hilbert
space. In this case the energies are degenerate, because E(k) = E(−k). The various
degenerate functions are distinguished from each other by the different values of k.
Averages (expectation values). Applying the Hamiltonian operator to the
∞
function Ψ = n=0 an Ψn , and carrying out the scalar product of the result with
this very function, yields
 
Ψ|ĤΨ = a∗n am Em Ψn |Ψm  = |an |2 En .
n,m n

n |an | = 1, the right-hand side can be interpreted as a weighted av-

2
Since
erage of the values picked up in measurements of the energy, where |an |2 is the
probability to detect a certain eigenvalue En . It is therefore customary to denote
E = Ψ|Ĥ|Ψ. (Here, the notation Ψ1 |Ĥ|Ψ2  ≡ Ψ1 |ĤΨ2  is introduced.) This
average is termed the expectation value of the energy in the Ψ state.
The variational approximation. As all |an |2 are non negative and their sum
is 1, there exists the inequality
 
E = Ψ|Ĥ|Ψ = Ψ∗ ĤΨ = |an |2 En ≥ E0 . (C.1)
n
(If all En in the sum are replaced by E0 the result is reduced.) Therefore, the
average of the energy in an arbitrary quantum state is always larger or
equal to the ground-state energy, E0 . It follows that one may derive approxi-
mately the wave function of the ground state by calculating the average energy for
a selection of wave functions, and then choosing the wave function with the lowest
value of those averages. This value may well be higher than the true one; however,
given an ample selection of wave functions, it might be a good approximation for
the true ground-state energy. This is the variational method in quantum me-
chanics: the expectation value of the energy in an arbitrary “guess” of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 303

Topics in quantum mechanics 303

the wave function is an upper bound on the ground-state energy of the

system at hand. Improving the trial wave function, e.g., by trying several more
functions, tends to diminish the computed energy and to push the result towards
the true value of the ground-state energy.

Problem C.1.
a. Given the Hamiltonian of a one-dimensional harmonic oscillator,
2 d 2 mω 2 2
Ĥ = − 2
+ x ,
2m dx 2
and a trial function of the form ψ(x) = (α/π)1/4 exp[−αx2 /2], find the expectation
value of the energy, and its minimal value with respect to the parameter α. Com-
pare the result with the exact ground-state energy of the one-dimensional harmonic
oscillator.
b. Add to the Hamiltonian a “perturbation” term, Ĥ = λx6 . Find the expectation
value of the energy with the same function as above. For which α is it minimal?
Calculate the variational ground-state energy to leading order in λ.
Perturbation theory for non degenerate states. A ubiquitous method to
find approximately the eigenenergies and the eigenstates of a quantum system is
termed perturbation theory. In many cases, the eigenstates and the eigenvalues
(0) (0) (0)
of a Hamiltonian Ĥ0 , i.e., Ĥ0 Ψn = En Ψn , are known; but this Hamiltonian

is augmented by a new term, λĤ , called ‘perturbation’, where λ is a small di-
mensionless real number. One wishes to find the eigenenergies and eigenstates of
the new total Hamiltonian, Ĥ = Ĥ0 + λĤ . Consider first the case in which the
(0)
eigenvalues of the Hamiltonian Ĥ0 are not degenerate, i.e., a single eigenstate, Ψn ,
(0)
corresponds to each eigenenergy, En . Solving the Schrödinger equation with the
Hamiltonian Ĥ = Ĥ0 + λĤ might be a difficult task, and therefore one seeks ap-
proximate solutions. It is assumed that the dimensionless coefficient λ is smaller
than 1; the perturbation theory yields the solutions as power series in λ. Denote the
eigenfunctions of the new Hamiltonian by Ψn , and the corresponding eigenenergies
by En , i.e., ĤΨn = En Ψn . Expanding the wave function and the energy in the
small parameter λ,
En = En(0) + λEn(1) + λ2 En(2) + O(λ3 ) , n + λΨn + O(λ ) ,
Ψn = Ψ(0) (1) 2

and inserting the expansions into the Schrödinger equation ĤΨn = En Ψn , one
compares terms of the same order in λ. The ones to first order in λ are
Ĥ Ψ(0)
n + Ĥ0 Ψn = En Ψn + En Ψn .
(1) (0) (1) (1) (0)

(0)
Taking the scalar product with Ψm yields

Ψ(0)
m |Ĥ |Ψn  = (En − Em )Ψm |Ψn  + En δnm ,
(0) (0) (0) (0) (1) (1)
(C.2)
where the identity
 
∗ ∗ (1)
[Ψ(0)
m ] Ĥ Ψ
0 n
(1)
= [Ĥ0 Ψ(0)
m ] Ψn = Em Ψm |Ψn  ,
(0) (0) (1)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 304

304 Introduction to Solid-State Physics

which can be proven by integrating by parts, is used (the left equality expresses the
fact that the Hamiltonian is a hermitian operator). The terms for which n = m
give

En(1) = Ψ(0)
n |Ĥ |Ψn  .
(0)

For n = m one obtains


Ψ(0)
m |Ψn  = Ψm |Ĥ |Ψn /(En − Em ) ,
(1) (0) (0 (0) (0)

and hence
  Ψ(0)  (0)
m |Ĥ |Ψn 
Ψ(1)
n = Ψ(0)
m |Ψn Ψm =
(1) (0)
(0) (0)
Ψ(0)
m .
m m=n En − Em
(0)
The first equality can be checked by taking the scalar product with Ψm , and
ignoring the term with n = m. Indeed, the normalization of the wave function
yields
1 = Ψn |Ψn  = Ψ(0)
n + λΨn + . . . |Ψn + λΨn + . . .  = 1 + 2λRe[Ψn |Ψn ] + . . .
(1) (0) (1) (0) (1)

(0) (1)
(. . . represent higher orders in the expansion). It follows that Ψn |Ψn  = iy is
purely imaginary. Therefore in this order in λ one finds

n + iλyΨn + · · · = exp[iλy]Ψn + . . . .
Ψn = Ψ(0) (0) (0)

The phase factor exp[iλy] is cancelled in any calculation of a physical quantity and
therefore may be ignored.
In a similar fashion, the terms of order λ2 yield
Ĥ Ψ(1)
n + Ĥ0 Ψn = En Ψn + En Ψn + En Ψn .
(2) (0) (2) (1) (1) (2) (0)

(0)
Again, the scalar product with Ψn and the use of the previous results yield
 |Ψ(0)  (0) 2
m |Ĥ |Ψn |

n |Ĥ |Ψn  − λ
En = En(0) + λΨ(0) (0) 2
(0) (0)
+ ... . (C.3)
m=n Em − En
One may obtain the same results by the variational method. This can be demon-
strated by considering the ground state of the full Hamiltonian. As the functions
(0)
Ψn are a basis for the space of all functions, the required solution can be expanded
in terms of this basis, i.e., Ψ0 = n an 0 Ψn . The first coefficient, of n = 0, is
(0)

a00 = 1 + O(λ). All other coefficients are of order λ. Since the wave function is
∞
normalized, Ψ0 |Ψ0  = n =0 |an 0 |2 = 1. The average energy in this ground state
is

E = Ψ0 |Ĥ0 + λĤ |Ψ0  = a∗n0 am0 Ψ(0) 
n |Ĥ0 + λĤ |Ψm 
(0)

n,m
 
a∗n0 am0 Ψ(0) 
(0) (0)
= E0 + |an0 |2 (En(0) − E0 ) + λ n |Ĥ |Ψm  .
(0)

n=0 n,m
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 305

Topics in quantum mechanics 305

The last sum contains a single term of order λ, contributed by n = m = 0 (one

may write a00 = 1, since the corrections to this coefficient are of higher orders in
λ because of the normalization). The terms in the sum in which one of the two
indices is zero yield contributions of order λ2 . The result is hence

E = Ψ0 |Ĥ0 + λĤ |Ψ0  + O(λ2 ) .

(0) (0)

Namely, to linear order the correction to the ground-state energy is the average of
the additional Hamiltonian λĤ , calculated with the original ground-state function.
This is a special case of Eq. (C.1), where the “trial” function is that of the unper-
turbed ground state. Equation (C.3) shows that this result is exact to linear order
in λ.
The second-order (in λ) correction of the energy, within the variational method,
is found from the values of an0 which yield a minimal average energy. Since these
are complex numbers, an0 and a∗n0 are considered as two independent variables,
and E is differentiated with respect to each of them. Up to linear order in λ, this
procedure produces
∂E
= a∗n0 (En(0) − E0 ) + λΨ0 |Ĥ |Ψ(0)
(0) (0)
n  + O(λ ) = 0 ,
2
∂an0
where a00 = 1 is used in the second term. The derivative with respect to a∗n0 gives
the complex conjugate of this equation. Using the solutions for an0 and a∗n0 of these
equations in the expression for E, yields
 |Ψ(0) |Ĥ |Ψ(0)
n |
2
E = E0 + λΨ0 Ĥ |Ψ0  − λ2
(0) (0) 0
(0) (0)
,
n=0 En − E0

which is identical to Eq. (C.3) (for the ground state), derived by perturbation
theory. The variational-method treatment of excited states necessitates also the
requirement that the state with the minimal energy is orthogonal to the ground
state found above. The results do reproduce Eq. (C.3) for the excited states.
Perturbation theory for degenerate states. The derivation above is re-
stricted to the case where the eigenstates of Ĥ0 are not degenerate. When they are,
(0)
i.e., there are more than a single state with the same energy En , Eq. (C.3) cannot
hold; the denominator in the last term there vanishes for pairs of degenerate states,
and the sum diverges. That divergence is eliminated as follows. Consider Eq. (C.2)
(0)
for the set of degenerate states alone. These are denoted Ψn , where the index
distinguishes among the degenerate states. Then

Ψn |Ĥ |Ψn  = En δ .

(0) (0) (1)

The index is added to the energy on the right-hand side, as the perturbation-theory
correction need not be the same for all (original) degenerate states. This equation is
valid only when the original states are eigenstates of the perturbation Hamiltonian,
i.e., Ĥ Ψn = En Ψn , such that the matrix Ψn |Ĥ |Ψn  is diagonal. Had not
(0) (1) (0) (0) (0)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 306

306 Introduction to Solid-State Physics

(0)
Ψn possessed this property, one would have to construct linear combinations of
(0) (0)
them, Ψn = α cα Ψnα , with coefficients cα such that
 
Ĥ Ψn = En Ψn , or cα Ĥ Ψ(0)
(0) (1) (0) (1)
nα = En cα Ψ(0)
nα .
α α
(0)
Performing the scalar product (from left) with givesΨnα
 (0)
Ψn |Ĥ |Ψ(0)
(1)
nα cα = En cnα . (C.4)
α

The coefficients’ vector, {cα } has thus to be an eigenvector of the matrix

Ψnα |Ĥ |Ψnα , with an eigenvalue Enα , the first-order correction to the energy.
(0) (0) (1)

In many cases these eigenvalues are not identical to each other, and thus the de-
generacy of the Hamiltonian Ĥ0 + λĤ is smaller than that of the original one
Ĥ0 .
Other operators. Any arbitrary physical quantity, A, is associated with an
hermitian operator Â, with eigenfunctions Ψα and real eigenvalues α, ÂΨα = αΨα .
The ensemble of functions {Ψα } is a possible basis for the Hilbert space of the
system’s states. For instance, the momentum operator of a single particle is p̂ =
−i∇. In a box of volume V , on which periodic boundary √ conditions are applied,
the eigenstates of the momentum are Ψk = exp[ik · r]/ V , and the eigenvalues
are k (with discrete values dictated by the boundary conditions). As mentioned,
the state Ψk is also an eigenstate of the Hamiltonian Ĥ = p̂2 /(2m), with the
eigenvalue E(k) = (k)2 /(2m). This eigenstate of the energy is degenerate, since
E(k) = E(−k). The corresponding eigenstates are distinguished from each other
according to the eigenvalues of the momentum (that are not degenerate). Another
example is that of a single particle in a central potential, e.g., in the hydrogen
atom. The eigenstates of the Hamiltonian with energies En are characterized also
by the quantum numbers and m, which represent the eigenstates of L̂2 and L̂z ,
respectively, where L̂ = r̂ × p̂ is the operator of the angular momentum.
Commutative operators. In the two examples mentioned above, the basis
functions of the Hilbert space are simultaneously eigenfunctions of more than one
operator; in the example of the free particle in a box these are the Hamiltonian
and the momentum; and for the particle in the central potential these are the
Hamiltonian, the square of the angular momentum, and its component along one
of the axes. The reason for the appearance of such combinations stems from a
theorem which states that the basis of the Hilbert space can comprise states
which are simultaneously eigenstates of several operators provided that
those operators commute with each other. Two operators, Â and B̂, are
commutative when ÂB̂f = B̂ Âf for any arbitrary function f in space.
proof: When all basis states Ψαβ are eigenstates of the two operators, with eigen-
values α and β, ÂΨαβ = αΨαβ and B̂Ψαβ = βΨαβ , then
ÂB̂Ψαβ = αβΨαβ = B̂ ÂΨαβ , i.e., ÂB̂Ψαβ = B̂ ÂΨαβ .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 307

Topics in quantum mechanics 307

As this identity pertains to all basis states, it is valid for any other function in the
Hilbert space.
This is a sufficient condition as well. Suppose all eigenfunctions of  are
known, i.e., ÂΨα = αΨα . Operating on this equation with the operator B̂ gives
B̂ ÂΨα = αB̂Ψα . The left hand-side, due to the commutativity, can be also written
as B̂ ÂΨα = ÂB̂Ψα , and therefore Â[B̂Ψα ] = α[B̂Ψα ]. It follows that the function
B̂Ψα is an eigenstate of Â, with the same eigenvalue α. Provided that the eigen-
states of  are non-degenerate, this implies that B̂Ψα is proportional to Ψα , i.e.,
B̂Ψα = βΨα , where β is the proportionality coefficient. Hence, Ψα is an eigenstate
of B̂, with the eigenvalue β. This completes the proof for non-degenerate states
of Â.
We now demonstrate the proof for the simplest degenerate case, when the degen-
eracy of the eigenvalue α is 2. In this case there exists a two-dimensional subspace
of the Hilbert space, that includes all eigenstates of  with the eigenvalue α. It
is always possible to choose a orthonormal basis for this subspace, such that the
basis functions are also eigenstates of B̂. Had one begun with a different basis,
with basis functions Ψα1 and Ψα2 , then any other linear combination of these two
basis functions is another eigenstate, with the same eigenvalue α. In particular, the
functions B̂Ψα1 and B̂Ψα2 are eigenfunctions of Â, with the eigenvalue α. That is,
B̂Ψα1 = β11 Ψα1 + β12 Ψα2 and B̂Ψα2 = β21 Ψα1 + β22 Ψα2 . An arbitrary state in
the subspace, χ = c1 Ψα1 + c2 Ψα2 , under the operation of B̂, becomes
B̂χ = c1 B̂Ψα1 + c2 B̂Ψα2 = c1 (β11 Ψα1 + β12 Ψα2 ) + c2 (β21 Ψα1 + β22 Ψα2 ) .
χ is an eigenstate of B̂, with the eigenvalue β, for B̂χ = βχ = β(c1 Ψα1 + c2 Ψα2 );
carrying out a scalar product of this equation with the functions Ψα1 and Ψα2 yields
    
Ψα1 |B̂|Ψα2  Ψα1 |B̂|Ψα2  c1 c
=β 1 . (C.5)
Ψα2 |B̂|Ψα1  Ψα2 |B̂|Ψα2  c2 c2
This equation resembles Eq. (C.4): two linear, homogeneous equations for two
unknowns. These have a solution provided that β is an eigenvalue of the matrix
on the left hand-side. The matrix in fact possesses two eigenvalues, β1 and β2 ,
(that need not be identical), and two eigenstates χ1 and χ2 . These functions are
simultaneously eigenfunctions of the two operators. The extension of this proof to
degeneracies higher than two enlarges the matrix one has to diagonalize, but leads
to the same result.

C.1 Answer

Answer C.1.
a. The average value is E = (2 α + m2 ω 2 /α)/(4m); the minimum is obtained for
α = α0 ,
α0 = mω/ ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 308

308 Introduction to Solid-State Physics

and is ω/2. It is equal to the exact result.

b. The expectation value of the perturbation Hamiltonian is
  ∞
 α 2 15λ
Ĥ  = 2λ dxx6 e−αx = .
π 0 8α3
It follows that
2 α2 15λ ω  1 β α 15λ
E = (α + 0 ) + 3 = u+ + 3 , u= , β= . (C.6)
4m α 8α 4 u u α0 2m3 ω 4
The extrema are obtained from the equation 1 − 1/u2 − 3β/u4 = 0, whose solutions
√
are u2 = [1 ± 1 + 12β]/2. As the negative sign  yields an imaginary u, the only
√
solution is the one with the positive sign, u = [1 + 1 + 12β]/2. Inserted into
Eq. (C.6), it yields the variational approximation for the ground-state energy.
Expanding in powers of β, one finds
3β
u=1+ + ... ,
2
ω ω 15λ
E = (2 + β + . . .) = + 3 + ... .
4 2 8α0
Hence, in this order, the correction to the energy is just the expectation value
of the perturbation Hamiltonian, calculated with the ground-state function of the
Hamiltonian without the perturbation.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 309

Chapter 5

Lattice vibrations

This chapter discusses on the small oscillations of atoms (or ions) around their equi-
librium locations. This picture is based on the harmonic approximation, in which
the potential energy of each atom is replaced by a parabola centered around the
minimum of that energy. The eigenmodes of the lattice are expressed in terms of
Fourier transforms of the vibrations’ amplitudes of individual atoms. The eigen-
modes and the related frequencies are derived from both the classical Newtonian
equations of motion, which give rise to waves similar to sound waves that propa-
gate in the material, as well as by the quantization of the vibrational modes, i.e.,
the phonons (the particles dual to these waves). The lattice vibrations exist at
finite temperatures, and even at zero temperature (“zero-point motion”). They
contribute to the specific heat of the material, and affect its stability. Anharmonic
effects, that give rise to the thermal expansion and to the heat conductance of
crystals, are presented. The experimental methods by which eigenfrequencies of
lattice vibrations are detected, e.g, using inelastic scattering of neutrons or of light,
are described. Finally, spin waves, the excitations of magnetic materials that are
analogous to lattice vibrations, are considered.

5.1 Classical equations of motion in one dimension

Preface. The Born-Oppenheimer approximation, based on the assumption

that the nuclei are static, is used in Chapter 4 to calculate the quantum energies of
atoms or molecules. Alternatively, Chapter 4 presents calculations of the Coulomb
energy of the ions located right at the sites of the periodic lattice. In reality, the
nuclei (or the centers of mass of the atoms, molecules, or ions of which the crystal
is built) oscillate around the lattice sites; the latter are in fact just their average
locations. The term “atoms” used below refers to the centers of mass. These
vibrations contribute significantly to the specific heat, to the thermal expansion,
to the heat conductance, to the sound propagation, and to the temperature depen-
dences of many properties of the crystal (for instance, the electrical conductance,
see Chapter 6). The lattice vibrations cause the decay of the intensity of radiation

309
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 310

310 Introduction to Solid-State Physics

scattered off the crystal as the temperature is raised (the Debye-Waller factor) and
lead to the melting of the crystal (the Lindemann criterion, Sec. 3.10).
Longitudinal vibrations in one dimension. The simplest case is that of a
one-dimensional lattice of lattice constant a, built of identical atoms. At equilib-
rium, the atoms are located on the periodic lattice, at R0n = nax̂, where n is an
arbitrary integer. Consider small longitudinal oscillations around these equilibrium
sites, such that the instantaneous position of the atom is Rn = R0n +un = (na+un )x̂
(see Fig. 5.1). The longitudinal vibrations are the displacements of the atoms
that are parallel to the direction of the lattice. Transverse vibrations along the
normal directions are discussed in the following. It is also assumed, for simplic-
ity, that forces are active only between nearest neighbors, and that the force be-
tween two neighbors depends solely upon the distance in-between them. That
is, the force is derived from a potential energy of the form U (Rn,n+1 ), where
Rn,n+1 = Rn+1 − Rn = a + un+1 − un . (The vector notations are superflu-
ous when discussing a one-dimensional lattice.) The total potential energy is
Utot = n U (Rn,n+1 ), where the sum runs over the N atoms in the lattice. To
avoid complications at the end points, it is assumed for the time being that the
lattice is infinite. The boundary conditions are accounted for afterwards.
In Chapter 4 it is found that the equilibrium configuration corresponds to the
pattern for which the potential energy is minimal. In the present context this is
achieved when each of the summands is minimal, i.e., when the atoms reside in
the periodic-lattice sites, Rn,n+1 = Rn+1 − Rn = a or un = 0. At the minimum,
the derivative vanishes, U  (R)|R=a = 0. According to the Lindemann criterion, the
deviation from the equilibrium location is smaller than the distance between nearest
neighbors, lest the crystal melts into the liquid phase. Thus we assume that the
deviation is much smaller than the lattice constant, |un |  a, and consequently the
potential energy can be expanded as a power series in un ,

Utot = [U (a) + U  (a)(un+1 − un ) + U  (a)(un+1 − un )2 /2 + . . . ] ,
n

(. . . indicate higher orders in the un ’s). In an infinite lattice the linear terms
disappear, as un attains opposite signs in the (n − 1)th term and in the nth one.
(In the finite lattice, the boundary conditions ensure that these terms vanish.) Also,
when the forces act only between nearest neighbors then U  (a) = 0 as no force is
exerted at equilibrium. (Recall that −∂Utot /∂un is the force that acts on the nth
atom, and this force vanishes at equilibrium when un = 0.) Keeping only the
second-order terms,
DL 
Utot = U0 + (u − un+1 )2 , (5.1)
2 n n

where DL = d2 U (x)/dx2 |x=0 (L indicates the longitudinal vibrations). Adding

the kinetic energy of the vibrations, the Hamiltonian (that corresponds to the total
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 311

Lattice vibrations 311

energy) is
 p2
H= n
+ Utot , (5.2)
n
2M
where pn is the momentum of the nth atom, and M is the mass of each atom in
the crystal.
The classical equations of motion. Newton’s second law for the potential
energy (5.1) yields
∂Utot
M ün = − = −DL (un − un−1 ) − DL (un − un+1 ) . (5.3)
∂un
(A dot above the letter indicates a derivative with respect to time.) The right hand-
side contains the forces exerted on the nth atom by its left and right neighbors. Had
it been assumed that each pair of atoms is connected by a spring, with a spring
constant DL and an equilibrium length a, the result would have been the same:
each of the forces is given by the spring constant times the elongation (or the short-
ening) of the spring as compared with its equilibrium length, as in Hooke’s law.
Hooke’s law applies also to the oscillations of a harmonic oscillator; hence the
term “harmonic approximation” for this description. The higher-order terms
in the expansion of the energy are thus named anharmonic terms (they are dis-
cussed below). Figure 5.1 displays the equilibrium configuration of the lattice, as
well as the instantaneous one.

Fig. 5.1: Longitudinal oscillations of a one-dimensional, single-atom, lattice. Upper

row: equilibrium configuration, the length of each spring (helical lines) is the lattice
constant a. Lower row: an instantaneous configuration of the vibrating lattice, the
nth atom is located at xn = na + un . The deviations from the equilibrium sites are
marked by the small arrows.

Dispersion laws. Techniques for solving equations of the form (5.3) are de-
tailed in the following. Here it suffices to present the solution as a sum over
“monochromatic” waves which oscillate in time with definite frequencies ω. Con-
sider one of these monochromatic solutions, which can be written as un (t) =
An exp[−iωt], where the complex amplitude An depends on the coordinate (i.e,
on n), but not on time. It is convenient to represent the deviation un (t) as a com-
plex function; at the end of the calculation only the real part (or alternatively, the
imaginary one) is to be taken into account. Both real and imaginary parts solve the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 312

312 Introduction to Solid-State Physics

same equation; the choice between them, or the proper linear combination, is made
according to the boundary and initial conditions. With this form of the deviation,
the equation of motion (5.3) becomes
M ω 2 An = DL (2An − An−1 − An+1 ) . (5.4)
In an infinite lattice, one may shift all lattice points by the lattice constant, to
obtain the equation
M ω 2 An+1 = DL (2An+1 − An − An+2 ) .
Denoting Bn = An+1 shows that Bn solves precisely the same equation as An , i.e.,
Eq. (5.4): M ω 2 Bn = DL (2Bn − Bn−1 − Bn+1 ). As a linear homogeneous equation
has a single solution up to a multiplicative factor, one concludes that Bn = CAn ,
where C is yet to be determined. Hence, the solution is such that An+1 = CAn .
Repeating this procedure over and over again, assigning the origin to the point
n = 0, yields An = C n A0 . Had the absolute value of C been different from 1, then
the solution |An | would have attained very large values for positive (or negative)
n’s. This is not acceptable, since the deviation must be small. Hence C has to be a
complex number, of unit absolute value, i.e., C = exp[ika], where k is real, leading
to An = exp[ikna]A0 . This argumentation is typical to periodic systems, and the
result is a particular case of the Bloch theorem presented in Chapter 6. Inserting
this result into Eq. (5.4) yields a relation between the vibration frequency ω and k,
M ω 2 = DL (2 − e−ika − eika ) = 2DL [1 − cos(ka)] = 4DL sin2 (ka/2) , (5.5)
namely,

DL
ω(k) = 2 | sin(ka/2)| = ω0 | sin(ka/2)| . (5.6)
M
The solution of Eq. (5.4) is un = An exp[−iω(k)t] = A0 exp[i(kna − ω(k)t)], well-
known from wave mechanics. It is a wave, with a wave number k and a wave
length λ = 2π/|k|, propagating along the positive x̂−axis when k > 0 with a
phase velocity vph = ω/k [e.g., the maxima of the real part of un are located
at the points where kna − ωt = 0, and therefore they are moving to the right
according to na = (ω/k)t = vph t. The wave length is the distance in-between
neighboring maxima.] Similarly, the function un = A0 exp[i(kna − ω(k)t)], with
k < 0, that describes a wave moving along the opposite direction, also solves Eq.
(5.4), with the same frequency. The function ω(k) is the dispersion relation
relating the frequency to the wave number; it is displayed in Fig. 5.2. The frequency
is symmetric with respect to the sign of the wave number, i.e., there are two solutions
for each frequency.
Periodicity and Brillouin zones. An important property of the solution
derived in the previous paragraph is the periodicity in reciprocal space: Eq. (5.6)
obeys
ω(k) = ω(k + 2πh/a) = ω(k + G) , (5.7)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 313

Lattice vibrations 313

Fig. 5.2: Dispersion relation of a one-dimensional, single-atom lattice. The thick

line marks the dispersion within the first Brillouin zone (see below). The frequency
is measured in units of ω0 = 2 DL /M , and the wave number in units of the basis
vector of the reciprocal lattice, G0 = 2π/a.

where G = h(2π/a) is an arbitrary reciprocal-lattice vector of the one-

dimensional lattice (h is an arbitrary integer). Recall (Sec. 3.5) that such a vector
displaces a point from one Brillouin zone to another zone. Equation (5.7) expresses
the fact that the frequency is a periodic function of the wave number, with a period
2π/a (as seen in Fig. 5.2). Another aspect of this periodicity refers to the wave
functions un : the solution with the wave number k is also periodic between different
Brillouin zones, i.e.,
un (k) = A0 ei(kna−ω(k)t) = A0 ei(kna+2πhn−ω(k)t) = un (k + 2πh/a) = un (k + G) .
It follows that it suffices to obtain the solutions for wave numbers within the first
Brillouin zone of the reciprocal lattice, −π/a ≤ k ≤ π/a. All other solutions
are identical to those, and do not convey any additional information. As found
below, this conclusion is quite general. In all cases it suffices to calculate the
dispersion laws and the wave functions within the first Brillouin zone.
This is the main reason for the introduction of Brillouin zones in Sec. 3.5
The continuum limit. The relation (5.6) has other remarkable properties.
When the wave length is much longer than the lattice constant, a  λ or ka  1,
the right hand-side may be expanded in a Taylor series in ka; this leads to a linear
dispersion relation, ω = c|k|, where c = ω0 a/2 = a DL /M is the velocity
of the longitudinal wave. This is the sound velocity in the material. (The
sound velocity coincides with the phase velocity for long wave lengths.) Indeed,
when the lattice constant is very small, the discrete nature of the material becomes
approximately continuous along the x̂−axis, with a mass density per unit length
ρ = M/a. The location of the atom can then be viewed as a continuous variable,
na → x; consequently, 2un − un−1 − un+1 ≈ −u (x)a2 (the double prime denotes
the second derivative with respect to x). With these assumptions Eq. (5.3) becomes
the continuum wave equation,
∂2u ∂2u
ρ 2
=Y , (5.8)
∂t ∂x2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 314

314 Introduction to Solid-State Physics

where Y = DL a is the Young modulus, i.e., the compressibility of the mate-

rial, in units appropriate for one dimension. This is the equationfor longitudinal

sound waves in the continuum, with the sound velocity c = a DL /M = Y /ρ
(see Appendix D). Hence, in the small wave-number limit, the wavy solutions of
the equations of motion are the well-known sound waves in the continuum. For
this reason it is customary to term those as acoustic solutions. As for acoustic
waves ka  1, it follows that the ratio An+1 /An = exp[ika] ≈ 1, and therefore the
displacements of nearest neighbors are quite close to one another, justifying the use
of a continuous variable, na → x. More details of the continuum limit are given in
Appendix D.
“Forbidden” frequencies. As opposed to sound waves in the continuum, Eq.
(5.6) yields
 solutions with a real wave number only for frequencies smaller than
ω0 = 2 DL /M = 2c/a. The sound velocity in solids is typically of the order
of 105 cm/sec, the lattice constant is of the order of 1Å = 10−8 cm, and hence
ω0 ≈ 2 × 1013 sec−1 , like the frequencies of light waves in the infra red. When
one tries to force upon a crystal a vibration of higher frequency, the solutions of
Eq. (5.6) become complex, with a complex “wave number” k = π/a + iκ, where
ω/ω0 = cosh(κa/2) (check!). However, inserting this expression into the wave
function yields |un | = |A0 | exp[−nκa], a solution which diverges to infinity when
n is a very large positive or negative integer (depending on the sign of κ). This
solution, being far away from the harmonic approximation (that assumes small
displacements), is not acceptable in the infinite system. (Recall the Lindemann
criterion which states that large displacements cause the lattice to disintegrate.)
In a semi-infinite system these frequencies correspond to solutions that decay as
the wave propagates into the lattice (the diverging solution is rejected on physical
grounds). In other words, vibrations with such frequencies cannot propagate over
large distances within the material. The maximal frequency that still allows for
a wavy solution, ω0 , is reached at ka = π, i.e., when the wave vector is at the
edge of the first Brillouin zone, or when λ = 2a. This is the shortest wave that
can propagate in the system. (The physical origin is clear: shorter waves describe
vibrations of the material in-between neighboring atoms, but in the lattice there
are no atoms there; i.e., these vibrations are meaningless.) For this wave vector,
un (k = π/a) = A0 (−1)n exp[iωt]; this is a standing wave in which neighboring
atoms vibrate opposite to one another (contrary to the acoustic wave, for which
neighboring atoms vibrate in unison along the same direction).
Wave packets. The appearance of a standing wave at the edge of the Bril-
louin zone can be understood by examining the group velocity of the waves. As
mentioned, the solutions of the type un = A0 exp[i(kna − ωt)] describe propagating
monochromatic waves (of a single frequency). As all these waves, with various
frequencies, solve the original equations of motion (5.3), the general solution is a
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 315

Lattice vibrations 315

linear combination of them,


un (t) = A(k)ei[ka−ω(k)t] , (5.9)
k

where ω(k) is given in Eq. (5.6), and the coefficients {A(k)} are determined by
the appropriate initial and boundary conditions. [For the finite-length system, the
allowed values of k are discrete, and then Eq. (5.9) is a Fourier series of the
vibration. A continuum of k values is allowed for the infinite system, and then
the sum in Eq. (5.9) becomes an integral, which is the Fourier transform of the
vibration, see Appendix B.] The solution (5.9) is a wave packet. The maximal
intensity of this packet is reached at k values around k = k0 , where the amplitude
A(k) is maximal. As the amplitude is multiplied by the phase factor exp[ikna −
iω(k)t], the intensity is larger when this phase factor is slowly varying around k0 .
(For other values of the wave number the oscillating term changes its sign upon
small variations of k, making the contribution to the sum much smaller.) This slow
variation of the phase, and consequently this peak in the sum, are obtained for
(∂/∂k)[kna − ω(k)t]k=k0 = 0, that is, for na = vg t, where the group velocity vg is
vg = [∂ω/∂k]|k=k0 . (5.10)
The group velocity is the speed with which the peak of the wave packet evolves
in time. In the first Brillouin zone, (−π/a ≤ k0 ≤ /π/a),
 this velocity is vg =
±c cos(k0 a/2), with the sound velocity c = ω0 a/2 = a DL /M , and with a sign
given by that of k0 [see Eq. (5.6)]. Thus, the group velocity is maximal (in its
absolute value) and equals the sound velocity (of the acoustic limit) at the center
of the Brillouin zone, and vanishes at the zone edges. Zero group velocity indeed
implies standing waves, as found above.

Problem 5.1.
a. Find the dispersion relation for the longitudinal vibrations of a single-atom,
one-dimensional lattice with “spring constants” DL,m between neighbors located at
a distance ma from one another. Find the group velocity, the condition for the
dispersion law to be linear at small wave-numbers, and the sound velocity in that
case.
b. Find the dispersion relations for the longitudinal vibrations of a single-atom,
one-dimensional lattice with the Lenard-Jones interaction, Eq. (4.33), among the
atoms. What form is taken by these relations at long wave-lengths?
c. Plot the dispersion relation for DL,2 = DL,1 /2, and DL,m = 0 for m ≥ 2. How
many wavy solutions are there for each value of the frequency?

Transverse vibrations. Suppose that the atoms can move out of the one-
dimensional lattice, i.e., along the ẑ and ŷ directions. The location of the nth atom
is then a three-dimensional vector, Rn = nax̂ + un . Assuming central forces
between pairs of atoms, the potential energy of the interaction between two atoms
located at a distance Rn,n+m = max̂ + un+m − un from one another depends solely
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 316

316 Introduction to Solid-State Physics

on the length of this vector. For reasons to be clarified later on, the present analysis
includes also pairs of atoms that are far apart. Taylor-expanding the length of the
vector Rn,n+m ≡ |Rn,n+m | yields
Rn,n+m = ma + δux + (δu⊥ )2 /(2ma) + . . . ,
with δu = un+m −un , the square of the transverse displacement is (δu⊥ )2 = (δuy )2 +
(δuz )2 , and . . . represents higher-order powers of the deviations from equilibrium
(check!). Inserting this expansion into the Taylor expansion of the corresponding
potential yields
 (δu⊥ )2  (δux )2
U (Rn,n+m ) = U (ma) + U  (ma) δux + + U  (ma) + . . . . (5.11)
2ma 2
The term linear in δux = un+m,x − un,x vanishes as before, since at equilibrium the
resultant force on each atom vanishes. The equation of motion for the transverse
vibrations, e.g., along ŷ, is M ün,y = −∂Utot /∂un,y . (The one for the longitudinal
oscillations is identical to that derived in problem 5.1.) In the harmonic approxi-
mation,

M ün,y = − DT,m (2un,y − un−m,y − un+m,y ) , with DT,m = U  (ma)/(ma) .
m

When the force is active only between nearest neighbors, U  (a) = 0, leading to
M ün,y = 0. That is, in this approximation any transverse displacement remains
constant or increases with time until it becomes too long to be in the regime of
small oscillations, and thus negates the harmonic approximation. This pitfall can be
overcome by the inclusion of anharmonic terms, that give rise to nonlinear equations
of motion. These are beyond the scope of this book (but do note Secs. 5.5 and 5.8).
It can be also overcome by adding the effect of farther-away neighbors. Conjecturing
that the solution is still wavy, of the form un,y = B0 exp[i(kna − ωT t)], results in
the dispersion relation
  sin2 (mka/2)
M ωT2 = 4 DT,m sin2 (mka/2) = 4 U  (ma)
m m
ma
 k4
⇒ U  (ma)(ma) k 2 − (ma)2 + ... ,
m
12
where the second expression pertains to the large wave-length limit. As the equi-
∞
librium energy per site, utot = 2 m=1 U (ma), is minimal at the lattice constant,
the derivative there vanishes, i.e., m>0 mU  (ma) = 0. This implies that the first
term on the right hand-side (of the large wave-length limit result) disappears, and
there is no acoustic behavior (for which the frequency is proportional to the wave
vector). The leading term yields a quadratic dependence of the frequency on the
wave number, ωT ∝ k 2 , assuming that m U  (ma)(ma)3 < 0. For instance, when
there are only two terms in the sum, then U  (2a) = −U  (a)/2, and the sum equals
U  (a)+8U  (2a) = −3U  (a). Generally, the sum is indeed negative. Had it been pos-
itive, ωT would have been imaginary; the crystal is then unstable, and anharmonic
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 317

Lattice vibrations 317

terms must be added to stabilize it. In such cases there appears a new minimum in
the energy, which describes a new crystalline structure.
Bending transverse vibrations. Another possibility to stabilize transverse
vibrations is to add non-central forces, e.g., forces that depend on the angles
between neighboring bonds. Consider for instance the covalent bond. Section
4.3 lists several hybridizations of wave functions on an atom in the lattice, e.g.,
of the sp type, or the sp2 one, which are hybridized with wave functions on the
neighboring atoms. In a linear chain like Be–Be–Be–Be–Be–. . . there are two elec-
trons on each beryllium√atom occupying the hybridized atomic states of the type
sp, Ψ±x = [ψ200 ± ψx ]/ 2, and each of them creates a covalent bond with an elec-
tron on a neighboring atom. These bonds are located on the molecule axis, and
the angle between nearest neighbors is 180◦ . The energy of the bond depends upon
the distance between the neighboring atoms, and its minimal value determines the
lattice constant of this one-dimensional crystal. When the transverse displacements
are ignored, an expansion of this potential, as done in Eq. (5.1) gives rise to longi-
tudinal vibrations like the ones discussed at the beginning of this section. On the
other hand, a traverse motion of an atom modifies the angle between the bonds
that this atom has with its neighbors, say from π to γ; the hybridized states then
assume the form Ψi = A[ψ200 + β n̂i · ψ] (see problem 4.9) where n̂1,2 are unit
vectors along the new directions of the two bonds. Part (f) of problem 4.9 details
the dependence of the coefficients A and β on the angle γ. Because it is an angle
between two bonds, the effective interaction under consideration depends on the
coordinates of three atoms, extending the pair potential given in Eq. (5.1). The
new terms in the interaction refer to bending, flexural, or buckling deviations
of the lattice. As the angle γ decreases, the probability clouds of the two electrons
in the atom approach one another, thus increasing the repulsive Coulomb energy.
When the deviations are small one may express the energy as a power series of
the small parameter (1 + cos γ); problem 5.2 demonstrates that the dispersion law
obtained from this analysis is proportional to k 2 in the long-wave limit.
The transverse vibrations are in particular complex in a one-dimensional lattice,
because there are no atoms surrounding the chain that can exert transverse forces
and hinder the “runaway” of the atoms along the perpendicular directions. Similar
difficulties arise for vibrations normal to the plane of a two-dimensional crystal,
e.g., graphene. In that case again the lattice is stabilized by additional interactions,
like those described above (see also Sec. 5.5). Coupling the lattice to the substrate
upon which it lies may also stabilize the transverse vibrations. These difficulties
disappear in three dimensions, where each atom is surrounded all around by other
atoms. Below it is simply assumed that the spring constants of the transverse
oscillations are finite.

Problem 5.2.
Write down an expression for the dependence of (1 + cos γn ) on the displacements of
the atoms, where γn is the angle between two successive bonds in a one-dimensional
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 318

318 Introduction to Solid-State Physics

lattice, coupling the atoms at n − 1, n, and n + 1; find the leading terms of the
result for small deviations (longitudinal and transverse). Write down the equations
of motion for the transverse vibrations and obtain the dispersion relation.

Crystal with a basis. Next consider a one-dimensional lattice with two atoms
(or ions) in the unit cell, e.g., the one-dimensional version of table salt, Fig. 4.4.
The two atoms have different masses, M and m, but all longitudinal springs are
identical, with a spring constant D. The longitudinal displacements in the nth unit
cell, un and vn , are defined in Fig. 5.3. The equations of motion for those are
M ün = −D(un − vn−1 ) − D(un − vn ) ,
mv̈n = −D(vn − un ) − D(vn − un+1 ) . (5.12)
These equations are invariant under the translation n → n+1; one may expect then
wavy solutions of the form un = A exp[i(kna − ωt)] and vn = B exp[i(kna − ωt)].
Inserting those into Eqs. (5.12) shows that these are indeed the solutions, provided
that the coefficients obey the two equations
M ω 2 A = D[2A − (e−ika + 1)B] ,
mω 2 B = D[2B − (1 + eika )A] . (5.13)
These two homogeneous linear equations, for the two unknowns A and B, have a
unique solution provided that the determinant of the coefficients vanishes, i.e.,
(M ω 2 − 2D)(mω 2 − 2D) − 4D2 cos2 (ka/2) = 0 . (5.14)
2
This is a quadratic equation for ω , whose solutions are

D 1 4 sin2 (ka/2)
2
ω± = ±D − , (5.15)
μ μ2 Mm
where μ = mM/(M + m) is the reduced mass of the atoms in the unit cell. The
two solutions for the frequency are plotted in Fig. 5.4. They are both periodic
with the periodicity given in Eq. (5.7), and hence it suffices to examine them in the
first Brillouin zone, −π/a ≤ k ≤ π/a. As opposed to the example of a single-atom
lattice, each wave vector of the two-atom chain corresponds to two frequencies.
In the limit of small wave numbers the lower one is
Da2
ω−2
≈ k 2 = c2 k 2 . (5.16)
2(M + m)
This is precisely the acoustic frequency of a sound wave in a continuum of mass
density ρ = (M + m)/a [see the discussion before Eq. (5.8)]. The higher frequency
is approximately
2D Da2
2
ω+ ≈ − k2 . (5.17)
μ 2(M + m)

2D/μ is a rather high frequency, of the order of the frequency of light in the
infra red. Because of this, the lower branch of the dispersion relation is called the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 319

Lattice vibrations 319

Fig. 5.3: Oscillations of a one-dimensional two-atom lattice. Top: the equilibrium

configuration. The dashed frame encloses the unit cell at n = 0. The nth cell
contains two atoms (or ions); the bright circle represents an atom of mass M and
the dark one an atom with mass m. The equilibrium distance between the two atoms
in the unit cell is b, and the lattice constant (the length of the dashed rectangle)
is a = 2b. Bottom: an instantaneous glimpse at the vibrating lattice; the atoms in
the nth cell are located at na + un and na + vn + b.

“acoustic branch” and the upper one is termed the “optical branch”. One has
to keep in mind, though, that these are not electromagnetic waves.
As the wave number increases towards the edge of the Brillouin zone, the acoustic
frequency increases while the optical one decreases. At k = π/a the two frequencies
are (assuming that M > m) [ω− (π/a)]2 = 2D/M and [ω+ (π/a)]2 = 2D/m. Thus,
there is a frequency gap in-between the two: in the range 2D/M ≤ ω 2 ≤ 2D/m
there are no wavy solutions. All solutions in this range decay (or increase) exponen-
tially with the distance. In addition, as in the single-atom one-dimensional lattice,
solutions with frequencies above the maximum of the dispersion relation, i.e., above
the optical branch ω 2 ≥ ω02 = 2D/μ, do not propagate.

Fig. 5.4: Dispersion relation of a one-dimensional, two-atom lattice, with M =

3m/2. The solid line shows the dispersion in the first Brillouin zone of this crystal,
whose unit cell contains two atoms.
 The frequency is measured in units of the
maximal optical frequency, ω0 = 2D/μ.

It is illuminating to consider the amplitudes themselves. For each of the fre-

quencies ω = ω± , Eqs. (5.13) yield only the ratio between the two amplitudes,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 320

320 Introduction to Solid-State Physics

(explain!)
B 2 − M ω 2 /D 1 + eika
= = . (5.18)
A 1 + e−ika 2 − mω 2 /D
In the k ≈ 0 limit, one finds that A ≈ B for the acoustic branch and B ≈ −(M/m)A
for the optical one. In the acoustic branch the atoms move in unison, with almost
equal amplitudes of displacements. This indeed can be expected for very long waves,
for which the wave function changes only slightly over short distances. A similar
situation occurs for the sound waves of the single-atom lattice. For this reason the
sound velocity in Eq. (5.16) contains the total mass of the two atoms, (M + m):
the two atoms oscillate together as a single entity. On the other hand, in the
optical branch the signs of the two amplitudes are opposite to one another. The
two atoms in the unit cell vibrate one against the other; the instantaneous distance
between them oscillates periodically while their center of mass remains immovable.
[The displacement of the center of mass is proportional to (M A + mB).] This is
the reason for the optical frequency in Eq. (5.17) to depend on the reduced mass
of the two atoms, μ = M m/(M + m). A reduced mass always corresponds to a
relative motion of two particles in the rest reference frame of the center of mass.
The picture is significantly modified for wave numbers at the edge of the Brillouin
zone, k = π/a. There [ω− (π/a)]2 = 2D/M and [ω+ (π/a)]2 = 2D/m, and therefore
Eq. (5.18) yields A = 0 for the optical branch and B = 0 for the acoustic one.
(One should be cautious with the order of limits, though, as both numerator and
denominator vanish.) The other atom vibrates with an amplitude (−1)n ; this is
again a standing wave, for which one of the atoms remains at rest and the other
oscillates around its equilibrium location.
The limit of identical atoms. Equation (5.15) yields two frequencies for each
wave number even for identical masses,
2
ω± = 2D[1 ± cos(ka/2)]/M , (5.19)
as shown on the right hand-side of Fig. 5.5. Since the masses are equal, it could
have been expected that the solution coincides with that obtained for a single-
atom chain, with a lattice constant b = a/2, and with a double-size Brillouin zone,
−2π/a = −π/b ≤ k ≤ π/b = 2π/a, as portrayed in Fig. 5.2 and reproduced
by the thick line in the left panel of Fig. 5.5. As implied from Eq. (5.5), ω 2 =
2D[1 − cos(kb)]/M = 2D[1 − cos(ka/2)]/M for the single-atom chain. This result
coincides with the acoustic branch of the “new” frequency for |k| ≤ π/a, and with
the optical one for π/a ≤ |k| ≤ 2π/a. Due to the periodicities of the two-atom
chain solutions, the wave number may be shifted by a wave vector of the reciprocal
lattice, k → k − 2π/a, which transforms the left panel of Fig. 5.5 into the right one
(or vice versa). When the two masses are the same, both descriptions are identical:
the frequency is given either by the thick line in the left panel of Fig. 5.5 with
the wider Brillouin zone of the single-atom chain, or by the two thick lines in the
right panel there, with the narrower Brillouin zone of a lattice with a basis of two
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 321

Lattice vibrations 321

atoms. Recall that at times it is convenient to describe a crystal by a larger unit

cell with several atoms rather than by the primitive one. In the present context, the
result is a smaller Brillouin zone, which contains several branches. The conclusion
is that the number of longitudinal frequencies which pertain to each wave
vector coincides with the number of atoms in the cell. When the transverse
frequencies are included, the total number of frequencies is the number of the atoms
in the unit cell times the dimensions (i.e., times 3 in three dimensions). In three
dimensions, three frequencies are acoustic (i.e., approach zero in the limit of long
waves) and the rest are optical (i.e., tend to a finite, relatively large, constant
frequency).

Fig. 5.5: Dispersion relations of two equal masses in the unit cell. The thick lines
in the right panel display the optical and the acoustic branches within the first
Brillouin zone of the lattice with the double-size unit cell, −π/a ≤ k ≤ π/a. The
thick lines in the left panel are the dispersion relations of the single-atom lattice,
whose Brillouin zone is −2π/a = −π/b ≤ k ≤ π/b = 2π/a.

Problem 5.3.
Show that Eq. (5.14) is the eigenvalue equation of a 2×2 matrix, and that Eq.
(5.18) yields the eigenvectors. Determine the matrix.

Problem 5.4.
The two-atom crystal depicted in Fig. 5.3 may describe an ionic crystal, in which
the ions of mass m are positively charged, with charge +e, while those of mass M are
negative, with charge −e. An electromagnetic wave, whose electric field oscillates
along the longitudinal direction with frequency ω, is applied on this system. The
wave vector is k = ω/clight , where clight is the speed of light. Thus, at a point
x = na the electric field is E(n, t) = E0 exp[i(kna − ωt)]. Find the induced dipole
moment. Find the frequency dependence of the dielectric function.

Problem 5.5.
Suppose that there are additional “springs”, of spring constant D2 , between equal-
mass atoms in neighboring cells of the two-atom one-dimensional lattice of Fig. 5.3.
Find the dispersion relations of the longitudinal oscillations. Discuss the D → 0
limit.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 322

322 Introduction to Solid-State Physics

Problem 5.6.
As mentioned, the polyyne crystal contains two types of bonds . . . C − C ≡ C − C ≡
C . . .. [Fig. 1.7(b)] The spring constants of these two types differ from each other.
Write down the equations of motion for the longitudinal oscillations of this crystal.
Find the dispersion relations.

5.2 Classical equations of motion (arbitrary dimensions)

The harmonic approximation. Applying the harmonic approximation to dimen-

sions higher than 1, for a potential energy that is a sum over the energies of pairs of
atoms, is straightforward: the explicit expression is Utot = i=j U (Rij )/2, where
Rij = Ri − Rj is the vector connecting two atoms. (The sum is divided by 2 since
it contains the contribution of each pair twice.) Qualitatively, the potential U (R)
is repulsive at short distances and attractive at longer ones, as illustrated in Fig.
1.1. Similar to the analysis carried out in Sec. 5.1, only small deviations from equi-
librium are considered; the instantaneous locations of the atoms are Ri = R0i + ui ,
where ui is the displacement of the i−th atom away its equilibrium location on the
lattice, R0i (where the potential is minimal). In the harmonic approximation the
potential energy is expanded around its minimum in the small deviations ui ’s (and
therefore all first derivatives vanish),
1 1 
Utot = U (R0ij + ui − uj ) = U0 + (uiμ − ujμ )Dμν (R0ij )(uiν − ujν ) + ..
2 4 μ,ν
i=j i=j
1 
= U0 + uiμ Kμν (R0ij )ujν + . . . . (5.20)
2 μ,ν i=j

Here, μ and ν denote the cartesian components of the vectors, i.e., μ, ν = 1, 2 (or
x and y) in two dimensions, μ, ν = 1, 2, 3 (or x ,y, and z) in three dimensions;
R0ij = R0i − R0j , and the definitions

Kμν (R0ij ) = δij Dμν (R0ij  ) − Dμν (R0ij ) ,
j  =i


Dμν (R0ij ) = ∂ 2 U/(∂Rij,μ ∂Rij,ν ) , (5.21)
Rij =R0ij

have been introduced. The notation δij stands for the Kronecker delta, δij = 1 when
i = j and 0 else. The constant term U0 is the equilibrium-stateenergy (denoted in
Chapter 4 as U0 = −N u, where u is the crystal binding energy per unit cell, and
there are N unit cells. The boundaries are ignored in this discussion; they are
accounted for in Sec. 5.3.) The . . . in Eq. (5.20) indicate the anharmonic terms,
of higher orders in the u’s; these are discarded in the harmonic approximation.
This approximation allows one to obtain the temporal dependence of the small
displacements ui (in the classical approximation). In the particular situation in
which the potential depends only on the distance between the two atoms, i.e., it
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 323

Lattice vibrations 323

is a central potential, the harmonic approximation amounts to joining each pair of

atoms by a spring obeying Hooke’s law.
The harmonic coefficients Dμν (R0ij ) are determined by the potentials, see Eqs.
(5.21). In general, these coefficients are interrelated, and therefore only a small
number of them is needed. For instance, Bravais lattices are invariant under in-
version, R0ij ⇔ −R0ij (Chapter 2). As a result, the terms in Eq. (5.20) appear in
pairs, with the same coefficients for R0ij and −R0ij . Another example is the simple
cubic lattice, whose basis vectors are along the cartesian axes, ax̂, aŷ, and aẑ. The
lattice is invariant under reflections through the planes (100), (010), or (001), that
are normal to each of the basis vectors. The lattice is also invariant under rotations
by 90◦ around each of the axes. It therefore follows that, for bonds between near-
est neighbors, Dxx (±ax̂) = Dyy (±aŷ) = Dzz (±aẑ) = DL (nn) (nn denotes nearest
neighbors, and L stands for a longitudinal displacement, parallel to the bond un-
der consideration). Similarly, Dyy (±ax̂) = Dzz (±ax̂) = Dxx (±aŷ) = Dzz (±aŷ) =
Dxx (±aẑ) = Dyy (±aẑ) = DT (nn), where T stands for transverse motion, along the
direction normal to the bond. (The complexities related to transverse modes are
ignored here, and arbitrary spring constants are assumed also along the transverse
directions. Those, as shown, may vanish, or may be finite due to additional inter-
actions.) For nearest neighbors in a SC lattice, all non-diagonal elements in the
matrices Dμν (nn) vanish, and the matrix comprises only two independent coeffi-
cients, DL (nn) and DT (nn). The bonds between next-nearest neighbors (denoted
nnn) on this lattice, e.g., at ±ax̂ ± aŷ, obey similar symmetries; the D matrix for
them comprises three independent coefficients, one for the vibrations parallel to the
bond, and two for those normal to it (check!).
Central potentials. The explicit values of the matrices Dμν or Kμν are found
once the binding energy is known. As seen in Chapter 4, the potentials of pairs of
ions (or atoms, or molecules) are often central, that is, they depend only on the
length of the vector that joins the two members of the pair (this is not the case for
covalent bonds). The harmonic coefficients Dμν which result from a central potential
can be derived from an extension of Eq. (5.11). Using the notation uij = ui − uj ,
 u2ij
Rij = |R0ij + uij | = (R0ij )2 + u2ij + 2R0ij · uij ≈ Rij
0
+ 0
2Rij
R0ij · uij (R0ij · uij )2
+ 0 − 0 )3 + ... ,
Rij 2(Rij
0
with Rij = |R0ij |. Adding to this the expansion
1
U (Rij ) ≈ U (Rij
0
) + U  (Rij
0
)(Rij − Rij
0
) + U  (Rij
0
)(Rij − Rij
0 2
) + ... ,
2
leads, up to second order in the displacements, to
u2ij − (R̂0ij · uij )2 1
U (Rij ) ≈ U (Rij
0
) + U  (Rij
0
) R̂0ij · uij + 0 + U  (Rij
0
)(R̂0ij · uij )2 ,
2Rij 2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 324

324 Introduction to Solid-State Physics

where R̂0ij = R0ij /Rij

0
is a unit vector along the direction of the vector R0ij .
The total potential energy is Utot = i=j U (R0ij + uij )/2. In general, the sum
runs over all lattice sites, but in most cases it is well converging, since the potential
decays with the distance. The expansion of the total energy includes the linear

i=j U (Rij )(R̂ij · uij )/2. All terms in this sum have to vanish, that is,
0 0
term
the condition for equilibrium reads i=j U  (Rij 0
)R̂0ij = 0. As a Bravais lattice is
invariant with respect to inversion, Rij ⇔ −Rij , this condition is always fulfilled
0 0

by central potentials. The next term in the expansion yields

U  (Rij
0
) 
U  (Rij
0
) 0
Dμν (R0ij ) = δμν 0 + U (R 0
ij ) − 0 R̂ijμ R̂0ijν . (5.22)
Rij Rij
Problem 5.7.
a. Show that in one dimension Eq. (5.22) reproduces Eq. (5.11). What are the
longitudinal and transverse spring constants in this case?
b. Find the coefficients Dμν (ma) for the Lennard-Jones potential, Eq. (4.33).

The equations of motion and the dispersion relations. The classical

equations of motion of a single-atom crystal are [see Eq. (5.20)]

M üiμ = − Kμν (R0ij )ujν . (5.23)
j ν

Examination of the expression for the “spring constants”, Eq. (5.21), reveals that
0
j Kμν (Rij ) = 0, and therefore all forces vanish when the entire crystal is displaced
by a constant displacement, ui = u0 (check!). Indeed, forces appear only when the
atoms move each relative to the others. One may now generalize the argument of
Bloch: upon shifting all atoms by a basis vector of the lattice, e.g., R0i → R0i =
R0i +ai , the difference R0ij = R0i −R0j remains as is, and the new variables ui μ obey
the same equations. Barring exceptional cases, Eq. (5.23) has a unique solution up
to a multiplicative factor, and the new variables are such that ui μ = Cuiμ . For
the above-mentioned displacement, the constant C is of the form C = exp[ik · ai ]
(see problem 5.8). Repeating the argument for each of the basis vectors results in
ui = A exp[i(k · R0i − ωt)], which describes a planar wave in space (see problem
5.8). Inserting this expression for ui into the equations of motion (5.23) shows that
this indeed is a solution, provided that three equations (in three dimensions) are
obeyed,
  
M ω 2 Aμ =
0
Kμν (R0ij )e−ik·Rij Aν =  μν (k) Aν ,
K (5.24)
ν j ν

where
 
 (k) =
K
0
Kμν (R0ij )e−ik·Rij = Dμν (R0ij )(1 − e−ik·Rij ) ,
0
(5.25)
μν
j j

is the Fourier transform of Kμν (R0ij ) of Eq. (5.21) (check!). Equation (5.24 rep-
resents three linear equations for the three components of the vector A; the latter
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 325

Lattice vibrations 325

are nonzero provided that the determinant of the coefficients, a polynomial of order
three in ω 2 , vanishes. As a result, there are three frequencies for each wave vector
k. In other words, the vectors A which solve these equations are eigenvectors of the
matrix K  μν (k), with eigenvalues M ω 2 (see also problem 5.3). One may convince
oneself that Eq. (5.5) is a particular case of Eq. (5.25), for the one-dimensional
configuration in which A is a one-dimensional vector, i.e., a number.
Problem 5.8.
Prove that the coefficient C in the equation ui μ = Cuiμ is given by C = exp[ik · a1 ];
hence the wave function is of the form ui = A exp[i(k · R0i − ωt)].
When the unit cell contains several atoms, the wave function is described by
several different amplitudes A(s) for each atom in the unit cell [as in Eq. (5.14),
which contains two such amplitudes]. For nB atoms in the unit cell, there are nB d
equations in d dimensions for the components of those vectors. It follows that there
are nB d different frequencies for each wave vector k. Out of all dispersion relations,
d are acoustic (i.e., yield zero frequency in the limit of zero wave vector), and the
remaining (nB − 1)d ones are optical.
Square lattice. A simple example is that of a single-atom square lattice (in
the XY plane) with interactions just between nearest neighbors. The “spring con-
stants” are then characterized by DL for the vibrations along the bond in-between
two neighbors, DT 1 for the vibrations in the plane perpendicular to this bond, and
DT 2 for the vibrations along the normal to that plane. As mentioned, the transverse
spring constants may vanish, unless the interactions are augmented by couplings
with farther-away neighbors, or by interactions resulting from non-central poten-
tials. There are four neighbors to each atom in the lattice, at ±ax̂ and ±aŷ. Ex-
ploiting the above-mentioned symmetries, one finds Dxx (±ax̂) = Dyy (±aŷ) = DL ,
Dyy (±ax̂) = Dxx (±aŷ) = DT 1 , and Dxx (±ẑ) = Dyy (±ẑ) = DT 2 . Hence [see Eq.
(5.21)] Kxx (0) = Kyy (0) = 2(DL + DT 1 ), Kzz (0) = 4DT 2 , and Kμν = 0 for μ = ν;
the matrix in Eq. (5.25) is diagonal. Consequently one finds three equations,
 xx (k) = 2DL [1 − cos(kx a)] + 2DT 1 [1 − cos(ky a)]
K
kx a ky a
= 4DL sin2 ( ) + 4DT 1 sin2 ( ),
2 2
 yy (k) = 2DL [1 − cos(ky a)] + 2DT 1 [1 − cos(kx a)]
K
ky a k a
= 4DL sin2 ( ) + 4DT 1 sin2 ( x ) ,
2 2
k a
K (k) = 2D [2 − cos(k a) − cos(k a)] = 4D [sin2 ( kx a ) + sin2 ( y )] . (5.26)
zz T2 x y T2
2 2
The μ−component of the amplitude A oscillates with a frequency given by M ω 2 =
K μμ (k). Figure 5.6 illustrates the dispersion relations as functions of the com-
ponents of the wave vector, for transverse vibrations normal to the lattice plane
(μ = z) and for the ones along one of the plane axes (μ = x). The results are dis-
played in the first Brillouin zone of the square lattice, as they are periodic among
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 326

326 Introduction to Solid-State Physics

the zones. Methods for measuring the dispersion relations are discussed in the fol-
lowing. Instead of depicting the dispersion relations in a complex three-dimensional
space as in Fig. 5.6, it is convenient to present them only along special routes in
the Brillouin zone. Figure (5.7(b) displays the dispersion for the three vibrational
modes along the path Γ → M → K → Γ, as defined in Fig. 5.7(a).
The dispersion relations of a simple cubic lattice in higher dimensions are derived
in a similar fashion. Though the derivation is rather straightforward, the plot of the
dispersion relations as functions of the three components of the wave vector requires
a four-dimensional space. It is therefore customary to exhibit the dispersion as
a function of the wave vector along special routes in the first Brillouin zone [an
experimental plot of this kind is presented in Fig. 5.17(a) and discussed in problem
(s.5.6)]

(a) (b)

Fig. 5.6: Dispersion relations of the vibrations of a square lattice with nearest-
neighbor interactions. The wave number is measured in units of 2π/a. (a) Dis-
persion
 of oscillations normal to the lattice plane; the frequency is in units of
2 2DT 2 /M . (b) Dispersion ofthe vibrations along the x̂−axis, with DT 1 = DL /4;
the frequency is in units of 2 (DL + DT 1 )/M (note that while the vibrations are
along x̂, k might have a component along ŷ).

When the vibration is described by the wave function ui = A exp[i(k · R0i −

ωt)], atoms residing on the plane normal to the wave vector k move in unison,
without being displaced relative to each other within the plane. Therefore, such
a wave corresponds to a motion of the planes one against the other, like in the
one-dimensional examples given in Sec. 5.1. In particular, when the wave vector k
lies along one of the special lines in Fig. 5.7(a) (e.g., Γ → M ), the wave describes a
motion of planes normal to the reciprocal-lattice vector corresponding to that line
[e.g., G = (2π/a)x̂].
In one dimension, the longitudinal and the transverse vibrational modes are well
separated, with distinct frequencies and sound velocities for each of them. In the
present situation, however, such a separation pertains only to waves moving along
one of the axes, e.g., when k = (k, 0). In that case the frequency of the longitudinal
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 327

Lattice vibrations 327

(a) (b)

Fig. 5.7: (a) The first Brillouin zone of the square lattice. (b) Dispersion relations
for the three vibrational modes along a path from the center of the zone, Γ ≡ (0, 0),
to the center of the face, M = (π/a, 0), then to the corner, K = (π/a, π/a), and
then back to the center. The spring constants are DT 2 = DL /8, DT 1 = DL /4.

wave with the amplitude Ax is given by M ω 2 = K  xx (k) (on the path Γ → M ), and
that of the transverse planar wave, of amplitude Ay , is given by M ω 2 = K  yy (k).
These results resemble the ones of the one-dimensional system, with one longitudi-
nal wave and two transverse ones, and with three acoustic branches of the dispersion
relation. This similarity is not accidental: in such a wave, all atoms residing on a
lattice plane (a line in two dimensions) normal to the direction of the motion (i.e.,
to the x̂−axis) move in unison with the same displacement, and the vibrations be-
come effectively
√ one-dimensional. For a wave along (1, 1), i.e., for a wave vector
k = (k, k)/ 2 (the path K → Γ in Fig. 5.7), one can clearly identify a longitudi-
√
nal and a transverse wave. In this case K  yy = 4(DL + DT 1 ) sin2 (ka 2/4),
 xx = K
and consequently the equations
√ for Ax and Ay are identical. The longitudinal-wave√
amplitude is (Ax + Ay ) 2 and the amplitude of the transverse one is (Ax − Ay ) 2,
both with the same frequency. However, the identification of three dispersion re-
lations with one longitudinal wave and two transverse ones is generally meaningful
only along special symmetry directions of the wave vector. In other directions there
still exist three branches of dispersion relations (for the single-atom lattice), but
there is no relation between a certain branch and the longitudinal wave (along the
direction of the wave vector) or the transverse one (normal to the wave vector). It
is also interesting to note that the degeneracy between the two transverse modes is
lifted along the other paths in Fig. 5.7, where there are three different frequencies
of vibrations along each of the lattice directions.
The dispersion relation is linear in the wave vector length at the vicinity of the
center of the Brillouin zone, as expected for sound waves. The sound velocity of
 vibrations normal to the plane is independent of the direction, ω ≈ c|k|, with c =
the
a DT 2 /M . On the other hand, the sound velocity of vibrations within the plane
depends on the direction of the wave vector. The frequency of a wave oscillating
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 328

328 Introduction to Solid-State Physics


along the x̂−axis, as seen in Fig. 5.6(b), is ω ≈ a (DL /M )[kx2 + (DT 1 /DL )ky2 ],
and cannot be declared as a transverse or a longitudinal wave. Arbitrary sound
waves in the continuum limit are discussed in Appendix D.
The explicit values of the spring constants, DL , DT 1 , and DT 2 , are specific for
each lattice. For instance, the constants in a square lattice with central-potential
forces among nearest neighbors are [see Eq. (5.22)]

DT 1 = Dxx (±aŷ) = Dyy (±ax̂) = DT 2 = Dzz (±ax̂) = U  (a)/a ,

DL = Dxx (±ax̂) = Dyy (±aŷ) = U  (a) , (5.27)

since the nearest neighbors lie along the axes. The equilibrium energy is Utot =
2N U (a); U  (a) = 0, and consequently DT 1 = DT 2 = 0: the frequencies of the
transverse motions vanish for a central potential between nearest neighbors. Like
in one dimension, it is customary to assume that the forces are not between nearest
neighbors alone, and/or are not central, and to adjust the spring constants DT 1
and DT 2 so that the dispersion relations comply with the experimental values. As
seen in problem 5.9 and also in the discussion of the one-dimensional configuration,
additional interactions between farther-away neighbors give rise to transverse modes
with non-vanishing sound velocities.

Problem 5.9.
a. How are the solutions for the vibrations of the square lattice modified upon
adding central-potential interactions in-between second neighbors, at distances R0ij =
a(±x̂ ± ŷ)?
b. Find the elastic constants in the continuum limit of this system. Use the defini-
tions of the elastic constants in Appendix D.

Vibrations of a triangular lattice. The nearest neighbors of an atom on the

triangular lattice
√ are located at R0nn = ±a1 , ±a2 , ±(a2 − a1 ), where a1 = ax̂, and
a2 = (x̂ + 3ŷ)a/2 (see Fig. 3.11). It is assumed that the interactions are solely
with the six nearest neighbors of each atom, and that the spring constants are given
by

Dμν (R0nn ) = DT δμν + (DT − DL )R̂0nn,μ R̂0nn,ν .

Treating DL and DT as given quantities, this specific form for the spring constants
implies that the constant for the longitudinal vibrations is DL , and that for the
transverse ones is DT . For instance, for R0nn = a1 = ax̂ one finds Dxx = DL ,
and Dyy = Dzz = DT . (The spring constants for the remaining neighbors are
obtained by applying 60◦ −rotations on these coefficients.) The vibrations normal
to the plane are described by the spring constant Dzz (R0nn ) = DT (the same for all
neighbors). These vibrations are decoupled from those pertaining to the vibrations
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 329

Lattice vibrations 329

within the plane; their dispersion relation is


M ω2 = K  zz (k) = 0
DT (1 − e−ik·Rnn )
nn

= 2DT [3 − cos(k · a1 ) − cos(k · a2 ) − cos(k · (a2 − a1 ))]

= 4DT [sin2 (k · a1 /2) + sin2 (k · a2 /2) + sin2 (k · (a2 − a1 )/2] . (5.28)

The frequency is periodic in the reciprocal lattice,

ω(k) = ω(k + hb1 + b2 ) , (5.29)

for arbitrary integers h and (the basis vectors in the reciprocal lattice are defined in
Fig. 3.11). The dispersion relation obeys the symmetry of the reciprocal triangular
lattice: a rotation of k by 60◦ leaves ω as is, since it just interchanges the vectors
a1 , a2 , and (a2 −a1 ) in Eq. (5.28). Figure 5.8(a) depicts the frequency as a function
of the components of the wave vector k; the periodicity and the symmetry are quite
clear.

(a) (b)

Fig. 5.8: The dispersion law of vibrations in a triangular lattice with interactions
between nearest neighbors, with DL = 4DT . (a) Vibrations normal to the plane,
(b) vibrations
 of one of the modes within the plane. The frequency is measured in
units of 2 3DT /M .

The vibrations within the plane are described by two coupled equations,
 xx Ax + K
M ω 2 Ax = K  xy Ay ,  xy Ax + K
M ω 2 Ay = K  yy Ay , (5.30)

where
     
 (k) = 4D sin2 k · a1 + (D + 3D ) sin2 k · a2 + sin2 k · (a2 − a1 )
K ,
xx L L T
2 2 2
     
 (k) = 4D sin2 k · a1 + (3D + D ) sin2 k · a2 + sin2 k · (a2 − a1 )
K ,
yy T L T
2 √ 2 2
√    3k a 
 xy (k) = 3(DL − DT ) sin kx a sin
K
y
(5.31)
2 2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 330

330 Introduction to Solid-State Physics

(check!). The allowed frequencies result then from the equation (M ω 2 −K  xx )(M ω 2 −
K yy ) = (K
 xy ) . The solutions of the latter are qualitatively similar to those per-
2

taining to the vibrations normal to the plane. One of them is portrayed in Fig.
5.8(b); it is periodic and symmetric as discussed above. Figure 5.9(b) displays the
three dispersion relations on the path shown in Fig. 5.9(a).

(a) (b)

Fig. 5.9: (a) The path Γ → M → K → Γ in the first Brillouin zone of the triangular
lattice. (b) The frequencies of the three vibrational modes along this path (from
Fig. 5.8) .

The spring constants derived from a central potential are DL = U  (a) and
DT = U  (a)/a. The condition of equilibrium leads to DT = 0, and hence to zero
frequencies for the vibrations normal to the plane. Adding interactions with farther
away neighbors gives rise to finite frequencies for those modes; the sound velocity,
however, still vanishes (i.e., the frequency is not linear in the wave vector).
Vibrations of graphene. As illustrated in Fig. 2.8(a) and in the central panel
of Fig. 2.9, graphene is represented by a triangular lattice with two atoms per unit
cell. As a result, six dispersion laws (nB = 2 times the three components of each
displacement) are expected. The unit cells are indexed by (n, m), as defined by
the lattice vectors R0n,m = na1 + ma2 [these lattice vectors are depicted in Fig.
2.8(a) ]. Consider the vibrations normal to the plane. The normal-to-the-plane
displacement of the point at the origin of the unit cell (i.e., at R0n,m ) is denoted
un,m , and that of the second point in the cell, located at R0n,m + 2(a1 + a2 )/3, is
labeled vn,m , see Fig. 5.10. Ignoring all interactions except those between nearest
neighbors (on the hexagonal lattice), then each point on one of the two sublattices
is coupled by a spring to three neighboring points on the other sublattice. All spring
constants for the vibrations normal to the plane are identical, and are denoted by D
(at equilibrium, the length of this spring is the distance between nearest neighbors
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 331

Lattice vibrations 331

√
in the hexagonal lattice, a/ 3). The equations of motion are
M ün,m = −3Dun,m + D(vn−1,m + vn,m−1 + vn−1,m−1 ) ,
M v̈n,m = −3Dvn,m + D(un+1,m + un,m+1 + un+1,m+1 ) . (5.32)
Substituting the wavy solutions un,m = A exp[i(k · R0n,m − ωt)] and
vn,m = B exp[i(k · R0n,m − ωt)], yields

M ω 2 A = 3DA − D(e−ik·a1 + e−ik·a2 + e−ik·(a1 +a2 ) )B ,

M ω 2 B = 3DB − D(eik·a1 + eik·a2 + eik·(a1 +a2 ) )A . (5.33)

Fig. 5.10: The notations of the atoms’ displacements in the hexagonal lattice. The
indices (n, m) represent the rhombic unit cell marked by thick lines. Each such cell
contains two atoms, whose displacements are unm and vnm .

The dispersion relations are the solutions of the quadratic equation (M ω 2 −

3D)2 = D2 |eik·a1 + eik·a2 + eik·(a1 +a2 ) |2 (check!). These solutions are plotted in Fig.
5.11. The explicit values of the frequencies along the Brillouin-zone path defined
in Fig. 5.9 are shown in Fig. 5.12. Note in particular the degeneracy of the two
eigenfrequencies at the K point of the Brillouin zone. This degeneracy, as well as
the linear in the wave-number dependence of the frequency around this point (see
problem 5.10) are typical to graphene. Around the Γ point one may use the acoustic
approximation to expand the dispersion relations in the wave vector. This yields
M ω 2 ≈ D(|k|a) 2
/4, and hence the sound velocity is the same along all directions,
c = ω/|k| = a D/M /2. The thin lines in Fig. 5.12 display this approximation.
The vibrational modes within the plane are given by four coupled equations (two
atoms in the unit cell, and two directions of the displacement of each of them), and
therefore the frequencies are derived from a quartic equation.

Problem 5.10.
a. Prove that the two eigenfrequencies in graphene are degenerate at the K point.
b. Show that around this point the eigenfrequencies
√ are linear in the distance from
that point, ω(kK + q) ≈ (3D/M )[1 ± ( 3/6)a|q|].
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 332

332 Introduction to Solid-State Physics

(a) (b)

Fig. 5.11: Dispersion relations of the two eigenmodes normal to the plane of a
graphene lattice with nearest-neighbor interactions.
 (a) The acoustic branch; (b)
the optical branch. Frequencies are in units of D/M .

Fig. 5.12: The frequencies of the two eigenmodes normal to the plane in graphene,
along the path displayed in Fig. 5.9. The thin lines pertain to the acoustic approx-
imation.

Problem 5.11.
a. Find the dispersion relations of the vibrations normal to the plane in graphene
when there are also interactions with next-nearest neighbors, with spring constants
D2 .
b. Prove that the dispersion relations of these vibrations are periodic according to
Eq. (5.29), and are symmetric with respect to rotations of the wave vector by 60◦ .
c. Derive explicit expressions for the sound velocities around the center of the
Brillouin zone.
d. Find the frequency of the optical branch at the center of the zone. What is the
ratio of the amplitudes of the vibrations in the unit cell there?
e. Find the frequencies of the two branches at the M and K points. What are the
ratios between the two amplitudes there?

Problem 5.12.
Boron nitride, BN, has the same hexagonal structure as graphene: the boron and
nitrogen ions occupy the two sites in the unit cell of the triangular lattice. Write
down the equations of motion for the vibrations normal to the plane of this crystal,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 333

Lattice vibrations 333

when solely nearest-neighbor interactions are accounted for. Find the dispersion
relations. Point out the main difference between them and those of graphene.

5.3 The eigenmodes of a finite lattice

Boundary conditions in one dimension. The monochromatic-wave functions,

un = A0 exp[i(kna − ωt)], and the wave packets in Eq. (5.9), are obtained by using
the Bloch argument, which is based on certain shifts of the coordinates. While these
shifts are harmless in an infinite system, they require special boundary conditions
in a finite system. The ubiquitous choice are the periodic boundary conditions
of Born and von Karmann. In one dimension they imply that any function un
defined on the lattice sites is periodic

un+N = un , (5.34)

where N is the number of unit cells in the finite system. The periodic bound-
ary conditions validate Bloch’s argument, and lead to wavy solutions even for a
finite lattice. They correspond to atoms placed on a circle, that vibrate along the
tangential to that circle (see problem 5.14).
Inserting un = A0 exp[i(kna − ωt)] into Eq. (5.34) gives rise to the condition
exp[ikN a] = 1, that is, the wave numbers are restricted to the N discrete values
k = k = 2π /(N a), with = 0, 1, 2, . . . , N − 1. Other integer values of , derived
by displacing any of the former values → + N , yield precisely the same wave
functions. This numerical range covers the section 0 ≤ k < 2π/a at equal-length
intervals, Δk = 2π/(N a) = 2π/L, where L = N a is the length of the sample. The
length of this range coincides with that of the first Brillouin zone. As mentioned,
the wave functions are invariant under displacements by vectors in the reciprocal
lattice, exp[i(k + G)na] = exp[ikna], where G = h(2π/a) is an arbitrary reciprocal-
lattice vector. This implies that the range −π/a ≤ k < π/a might be used as
well, i.e., the wave vectors are confined to the first Brillouin zone. When N is even,
then = −N/2, −N/2 + 1, −N/2 + 2, . . . , N/2 − 1, while for odd N the enumeration
begins with −(N −1)/2 and spans a similar range. Below, N is chosen for simplicity
to be even; the disparity between even and odd N ’s is negligible for large N ’s. The
sums that appear below encompass this range of ’s.
The eigenmodes. According to the Fourier theorem (see Appendix B) the
collection of functions {exp[ik na], = −N/2, . . . , N/2−1} is a basis for the Fourier
series of functions defined on the lattice sites. One may therefore define the Fourier
transform of the function un ,


N 
N
(k ) = N −1/2
u un e−ik na = N −1/2 un e−2πin/N , (5.35)
n=1 n=1
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 334

334 Introduction to Solid-State Physics

and its inverse (i.e., the original function)


N/2−1

N/2−1
un = N −1/2 (k )eik na = N −1/2
u (k )e2πin/N .
u (5.36)
=−N/2 =−N/2

Equation (5.36) is similar to Eq. (5.9); here, however, both un and u (k ) are
time-dependent. Inserting Eq. (5.36) into the equation of motion (5.3) yields the
(k ),
equation of motion of u
Mu(k ) = −D(2 − e−ik a − eik a )
¨ u(k ) = −M ω 2 (k )
u(k ) (5.37)
(check!). This is a separate equation for each of the variables { u(k )}. Equation
(5.37) describes the motion of a harmonic oscillator, i.e., Hooke’s law, with the
frequency ω(k ) derived in Eq. (5.6). The transformation (5.35) thus maps the
functions {un } into {
u(k )}, which are the vibrational eigenmodes of the system.
An alternative way to obtain the vibrational eigenmodes is found by inserting
into the Hamiltonian (5.2) the coordinates’ transformation Eq. (5.36), and the
parallel transformation for the momenta,
1  1 
pn = 1/2 p(k )eik na = 1/2 p(k )e2πin/N . (5.38)
N 
N 
(Recall that the sum over is confined to the first Brillouin zone.) The sum over
the momenta in the kinetic-energy term becomes

N
1 
N  
p2n = p(k )
p(k )ei(k +k )na = |
p(k )|2 . (5.39)
n=1
N 1/2 n=1   
N
[This manipulation uses the identity n=1 exp[i(k + k )na] = N δk ,−k which is
derived from a summation of the geometrical series, with exp[i(k + k )N a] = 1
(check!). In fact, this sum is N when k + k = G for any reciprocal-lattice vector
N
G; therefore n=1 exp[i(k + k )na] = N

G δk ,G−k . However, because the
analysis is confined to the first Brillouin zone, it suffices to use the identity with
G = 0, as done above.] Note that the relation p(k) = [ p(−k)]∗ [which follows
directly from Eq. (5.38) and the fact that pn is real] is used as well. Similarly,
un+1 − un = N −1/2  u (k )(exp[ik a] − 1) exp[ik na], and therefore

N
1 
N 
(un+1 − un )2 = (k )
u u(k )(eik a − 1)(eik a − 1)ei(k +k )na
n=1
N 1/2 n=1  

=4 |
u(k )|2 sin2 (k a/2) . (5.40)

Collecting these results, the Hamiltonian becomes a sum of N independent Hamil-
tonians. Each of those is the Hamiltonian of a harmonic oscillator whose coordinate
(k ), its conjugate momentum is p(k ), and the frequency is ω(k ). That is,
is u

N/2−1
1 M ω 2 (k )
H= |
p(k )|2 + |
u(k )|2 . (5.41)
2M 2
=−N/2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 335

Lattice vibrations 335

Indeed, in the classical limit, each of these Hamiltonians gives rise to the equations
of motion as in Eq. (5.37). Note that the N variables { u(k )} of Eq. (5.41) are
each complex. However, with the notations u (k) = x(k) + i
y (k), where x(k) and
y(k) are both real, it follows from the relation u u(k)]∗ , that x
(−k) = [ (−k) = x(k)
and y(−k) = − y (k). Hence, the N variables u (k) may be replaced by N/2 pairs
of independent variables, x  and y, whose argument k is positive (the limits on the
values of k have to be chosen carefully, but the result remains the same: there are
N independent harmonic oscillators).

Problem 5.13.
Derive the eigenmodes of the longitudinal vibrations in a finite one-dimensional lat-
tice comprising N atoms, coupled together by nearest-neighbor interactions; assume
rigid boundary conditions, for which uN +1 = u0 = 0. How many eigenmodes are
there in the first Brillouin zone?

Problem 5.14.
a. Show that periodic boundary conditions in one dimension, Eq. (5.34), correspond
to N atoms on a circular ring, like the carbon ions in the benzene molecule (Fig.
4.2), when the atoms move along the tangent to the ring.
b. Derive the equations of motion of the angles ϑn , which describe the angular
deviations of the atoms away from their equilibrium sites. Use the harmonic ap-
proximation, and assume central-force interactions between nearest neighbors. Show
that the vibrational modes are derived from the diagonalization of an N × N matrix.
Find the eigenvalues and the eigenvectors of this matrix.

Higher dimensions. The extension of the derivation above to dimensions

higher than 1 is straightforward. Consider a single-atom lattice comprising Nm
unit cells along the basis vector am . The Born-von Karmann boundary condition
along this direction is

u(R0i + Nm am ) = u(R0i ) . (5.42)

This implies that exp[ik · Nm am ] = 1. As for the one-dimensional lattice, the peri-
odic boundary conditions turn the crystal into a three-dimensional torus. According
to Eq. (3.17) the vector Nm k coincides with one of the reciprocal-lattice vectors and
therefore in three dimensions it is a combination of the vectors {b , = 1, 2, 3} with
integer coefficients. Since it is only the vector bm that survives the scalar product
with am [from Eq. (3.20)], one chooses km = m bm /Nm . As a displacement of
this vector by a reciprocal-lattice vector does not change the wave function, i.e.,
exp[ikm · Nm am ] = 1, one may choose the independent values of the wave numbers
within the first Brillouin zone, i.e., −Nm /2 ≤ m < Nm /2 − 1. The outcome is the
basis functions {exp[ik1 ,2 ,3 · R0i ]}, with the discrete wave vectors

k1 ,2 ,3 = ( 1 /N1 )b1 + ( 2 /N2 )b2 + ( 3 /N3 )b3 . (5.43)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 336

336 Introduction to Solid-State Physics

These wave vectors are all within the first Brillouin zone where they are distributed
homogeneously, i.e., with a constant density that increases as the lattice is enlarged.
The Fourier transform of u(R0i ), with this basis, is
 0
u(R0i ) = (N1 N2 N3 )−1/2  (k1 ,2 ,3 )eik1 ,2 ,3 ·Ri .
u (5.44)
1 ,2 ,3

Inserting it into the Hamiltonian yields

 p2iμ 1 
H= + uiμ Kμν (R0ij )ujν + . . . , (5.45)
i,μ
2M 2 μ,νi=j

(where piμ is the momentum conjugate to the displacement uiμ ). The Hamiltonian
becomes a sum of N1 N2 N3 Hamiltonians,
 1  1 
H= |
pμ (k)|2 + u  μν u
(k)K ν (−k) . (5.46)
2M μ 2 μ,ν
k

(The indices of k are omitted for brevity.) The three coupled equations of motion
in Eq. (5.24) emerge from this Hamiltonian in the classical limit. Three vibrational
eigenmodes for each wave vector are obtained from a diagonalization of the 3 × 3
matrix K  μν . In symmetric configurations, those comprise a longitudinal wave and
two transverse ones. When the unit cell, at d dimensions, contains nB atoms, each
of the discrete wave vectors in the first Brillouin zone corresponds to nB d harmonic
oscillators, with the same frequencies as calculated in the previous paragraphs.
Though the derivations of the Hamiltonians (5.41) and (5.46) are based on clas-
sical mechanics, the transformations between the variables on the lattice, e.g., (un
and pn ) and the variables in momentum space [ u(k) and p(k) or x
(k) and y(k),
with the appropriate wave vectors] can be applied also on the quantum operators
which correspond to these variables. It follows that the quantum description of
lattice vibrations is derived from the Hamiltonians (5.41) or (5.46), upon replacing
the coordinates and the momenta by the their quantum operators.

5.4 Density of states, specific heat of phonons

Energy levels, average energy, specific heat. Various thermal properties of

crystals, e.g., the specific heat, follow from the thermodynamics of lattice vibrations.
The thermodynamics of a single harmonic oscillator is discussed in Sec. 3.10. The
extension to a system described by many vibrational eigenmodes is quite straightfor-
ward. As derived in the previous section, the Hamiltonian of the lattice vibrations in
the harmonic approximation is a sum of Hamiltonians of one-dimensional harmonic
oscillators,
 1 M 
H= p2α + α ωα2 u2α , (5.47)
α
2Mα 2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 337

Lattice vibrations 337

where uα and pα denote the displacement and the momentum of the α−th eigen-
mode [these replace u μ (k) and pμ (k) of Sec. 5.3]. The index α replaces the index
μ (that has d values), the wave vector k (that has N1 N2 . . . Nd values in the first
Brillouin zone) and the index of the atom in the unit cell (that has nB values).
As explained, each of these variables (that represent the eigenmodes) is a linear
combination of the variables of a single atom, [see e.g., Eqs. (5.35) and (5.38),
leading to Eq. (5.41)]. Each summand in Eq. (5.47) corresponds to a harmonic
oscillator, of mass Mα , spatial coordinate uα , and momentum pα . Within a classical
approach, it describes a periodic vibration whose frequency is ωα . In a quantum-
mechanical calculation, the energy levels of each oscillator are ωα (nα +1/2), where
nα = 0, 1, 2, . . . , ∞ are integers that classify the eigenstates. In addition to the
ground-state energy, ωα /2, the energy of a single vibrational mode contains an in-
teger number of basic units (or quanta) of energy, ωα . Very much like photons in
quantum electrodynamics (e.g., black-body radiation) and as implied for quantum
particles by the wave-particle duality principle of de Broglie, such a basic unit can
be considered as a “particle” carrying this energy unit. In the present context these
“particles” are termed phonons. In the quantum terminology, the classical wavy
motion, given by the vibrational mode of frequency ωα , is replaced by the motion
of phonons, each carrying a discrete amount of energy ωα and momentum kα ,
related to that frequency by the respective dispersion relation. Substituting these
eigenvalues in Eq. (5.47) yields the energy levels of the entire system,

E({nα }) = ωα (nα + 1/2) . (5.48)
α
The thermal average of any physical quantity is carried out with the Boltzmann
weights, exp[−βE]/Z, where β = 1/(kB T ), and the normalization factor of the dis-
tribution (that ensures that the sum over all weights is 1) is the “partition function”
Z, Z = E exp[−βE], where the sum comprises all quantum states of the system.
In the present context,
  
Z≡ e−βE = e−β α ωα (nα +1/2)
E {n1 ,n2 ,n3 ,... }
  
= e−βωα (nα +1/2) ≡ Zα . (5.49)
α nα α

(The sum over the states is replaced by a sum over all quantum numbers {nα } in the
second step.) As these sums are independent of each other, one may calculate each
of them separately; the partition function is a product of the partition functions of
each oscillator, Zα . The latter is given by an infinite geometrical series,
 e−βωα /2 1
Zα = e−βωα (nα +1/2) = −βωα
= . (5.50)
n
1 − e 2sinh(βω α /2)
α

The free energy of the entire system is given by

  
F = −kB T ln Z = −kB T ln Zα = ωα /2 + kB T ln[1 − e−βωα ] . (5.51)
α α α
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 338

338 Introduction to Solid-State Physics

The average energy of the system is obtained using the Boltzmann weights,
Ee−βE ∂ ln Z  ∂ ln Z
E = E −βE = − =− α

E e ∂β α
∂β
  ω ω  
= α
+ βω α = Eα  . (5.52)
α
2 e α −1 α

The last term is a sum over the average energy of each oscillator separately,
1  1 1 
Eα  = ωα + nα = ωα + βω . (5.53)
2 2 e α −1
The right hand-side of Eq. (5.53) consists of two terms: the first, ωα /2, is the
energy of the ground state of the oscillator. This is the “zero-point energy” of
the oscillator, which does not affect the thermodynamic properties, being a number
independent of the temperature (recall that energies are defined up to an additive
constant). One does have to include this term in the total energy, as it might com-
pensate for the “classical” binding energy and thus destroy the crystalline structure.
As mentioned in Chapter 4, even in the ground state the location of the oscillator
(i.e., its coordinate) is determined up to a certain uncertainty; when the amount
of this uncertainty exceeds the lattice constant the crystal disintegrates. For this
reason helium remains liquid down to ultra-low temperatures. The second term in
Eq. (5.53) stems from the average over the quantum numbers nα of the phonons
that belong to an oscillator with frequency ωα ,
1
nα  = βω . (5.54)
e α −1
This number decays to zero at zero temperature (β → ∞). In this limit there are no
phonons, and the system is in its ground state. The number of phonons increases
as the temperature is raised. Equation (5.54) also describes the number of photons
in the black-body radiation, and in fact is the expression for the number of bosons
(particles that obey the Bose-Einstein statistics) of a given energy.
Specific heat (at constant volume). The expression for the specific heat of
a system is obtained by differentiating the energy with respect to the temperature,
 ∂E   eβωα
CV = = kB (βωα )2 βω . (5.55)
∂T V α
(e α − 1)2
Figure 5.13 illustrates the temperature dependence of the specific heat due to a
single vibrational mode. It is exponentially small at low temperatures, and ap-
proaches the Boltzmann constant, kB , at high ones (check!). The latter result is
a manifestation of the equipartition law, which holds at high temperatures, i.e.,
in the classical limit: to each quadratic term in the Hamiltonian (for instance, the
potential energy which is quadratic in the coordinate, or the kinetic energy, that
is quadratic in the momentum) there corresponds an average energy kB T /2. Thus,
the total average energy of each oscillator (in the classical limit) is kB T , and its
contribution to the specific heat is kB . This result also follows from the analysis
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 339

Lattice vibrations 339

above: when kB T  ωα , Eq. (5.54) yields nα  ≈ kB T /(ωα ), and Eq. (5.53)
gives Eα  ≈ kB T . Inserting those into Eq. (5.55) multiples kB by the number of
terms in the sum, that is, by the number of vibrational modes. Each mole in a
single-electron material contains 3N vibrational modes, where N is the Avogadro
number. The specific heat per mole is CV = 3N kB = 3R, where R is the gases’
constant. This is the Dulong and Petit law, that is indeed observed in numerous
high-temperature experiments. It is worthwhile to note that the function in Fig.
5.13 depends upon a single “scaled” variable kB T /(ωα ). The plot is “universal”:
when the temperature is scaled by the frequency of the vibration (in energy units),
and the specific heat is scaled by the Boltzmann constant, the plot describes the
specific heat contributed by any vibrational mode.

Fig. 5.13: The specific heat of a single vibrational mode of frequency ωα , in units
of kB , as a function of the temperature measured in units of ωα /kB .

Problem 5.15.
In classical statistical mechanics, the partition function is an integral over
the entire phase space. For a one-dimensional harmonic oscillator Z =

α (1/h) dp α du α exp[−βH], where H is given in Eq. (5.47), and the Planck
constant appears in order to make the partition function dimensionless. Derive the
free energy, the energy, and the specific heat, and prove that the latter obeys the
Dulong-Petit law.

Problem 5.16.
The area enclosed by the horizontal line and the curve of the specific heat in Fig.
5.13 represents the difference between the energy in the classical calculation and the
one resulting from the quantum-mechanical one. Find the value of that difference.

Density of states. As explained in Sec. 5.4, the index α in Eq. (5.47) – and
hence also in Eq. (5.55) – represents the N1 N2 . . . Nd wave vectors k in the first
Brillouin zone; each of those has nB d vibrational modes that correspond to the d
components of the displacement vector, and to the nB atoms in the unit cell. The
frequency of each of the nB d vibrational modes is related to the wave vector by the
relevant dispersion law, ωm (k), where m = 1, 2, . . . , nB d. Because the wave vectors
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 340

340 Introduction to Solid-State Physics

cover densely the Brillouin zone, it is expedient at times to turn the sum over k into
an integral. For instance, the k−dependence of the contribution to the specific heat
is solely via the dependence on the frequency ωα (k); by a variables’ redefinition,
the sum over k becomes an integral over ω.
The density of states in one dimension. The simplest configuration is that
of the longitudinal oscillations of a single-atom, one-dimensional chain of length
L = N a (a is the lattice constant), built of N atoms with nearest-neighbor couplings.
Here d = nB = 1, and the sum runs over the wave vectors in the first Brillouin
zone, k = k = 2π /(N a), with = −N/2, −N/2 + 1, −N/2 + 2, . . . , N/2 − 1 [see
the discussion following Eq. (5.34)]. The sum over all these wave vectors of an
arbitrary function f [ω(k)] is
 
N/2−1  N/2
f [ω(k)] = f [ω(k )] = d f [ω(k )]
k =−N/2 −N/2
 π/a  π/a
d L
= dk f [ω(k)] = dkf [ω(k)] ,
−π/a dk 2π −π/a

where the discrete sum is turned into an integral, of “integration step” Δ = 1,

with the integration variable k (exploiting the relation dk/d = 2π/L). In the next
step one uses the dispersion relation Eq. (5.6). Because two vibrational modes with
±|k| correspond to each frequency, the integral over ω is multiplied by 2,
    ω0
L π/a L ω0 dk
f [ω(k)] = dkf [ω(k)] = dω f (ω) = dωg(ω)f (ω) ,
2π −π/a a 0 dω 0
k
(5.56)
where g(ω) = (L/π)(dk/dω) is the density of states. It is defined such that g(ω)dω
is the number of vibrational modes with frequencies in the range {ω, ω + dω}. From
Eqs. (5.6) and (5.10) one finds that the group velocity is

dω/dk = vg = c cos(ka/2) = c 1 − (ω/ω0 )2 ,
(the sound velocity c is c = ω0 a/2) and thus the density of states can be written in
the form

g(ω) = (L/π)/vg (ω) = L/[πc 1 − (ω/ω0 )2 ] . (5.57)
Figure 5.2 shows that as ω approaches ω0 , the group velocity decreases, thence the
density of states, displayed in Fig. 5.14(a), increases. In particular, the density of
states diverges at the upper edge of the Brillouin zone where ω = ω0 and the group
velocity vanishes. Such divergences are termed after van Hove.
The specific heat of a one-dimensional chain. The specific heat of the
one-dimensional lattice discussed above is
 ω0 
(βω)2 eβω 2
kB T X x2 ex
C V = kB dωg(ω) βω = L dx  .
0 (e − 1)2 πc 0 1 − (x/X)2 (ex − 1)2
(5.58)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 341

Lattice vibrations 341

Here X = βω0 , and the integration variable is changed to x = βω. As expected,

the specific heat, which is an extensive property, is proportional to the length of the
lattice, L. At temperatures high enough compared to ω0 /kB the frequencies are
such that βω < βω0  1, and consequently (βω)2 exp[βω]/(exp[βω]−1)2 ≈ 1.
In this limit, the integral is conveniently calculated using the second expression in
ω
Eq. (5.58). The result, 0 0 dωg(ω) = N , is the total number of vibrational modes
in the range 0 < ω < ω0 . [The equality stems from both the definition of the
density of states, that counts the number of states per frequency unit, and from the
explicit integration, see problem 5.17(a).] The specific heat at these temperatures
is CV ≈ kB N , and is a particular case of the Dulong-Petit law. In contrast, at low
temperatures one has X  1. Inserting into the integral on the right-hand side of
Eq. (5.58) X = ∞, this integral converges, and its magnitude is
 ∞
ex π2
dxx2 x = 2ζ(2) = , (5.59)
0 (e − 1)2 3
∞
where ζ(2) = m=1 m−2 = π 2 /6 is a particular case of the Riemann zeta function,
∞ −s
ζ(s) = m=1 m , see problem 5.17(b). It can be proven that the corrections
to this result that arise from large but finite values of X are small (of the order
of X 2 exp[−X], see also problem 5.19). It follows that at low temperatures the
leading-order behavior of the specific heat of these vibrational modes (which are in
fact acoustic phonons) in one dimension is linear in the temperature. This approx-
imation fits quite well the low-temperature specific heat data, see Fig. 5.14(b) (the
thin line there). This behavior differs significantly from the one depicted in Fig.
5.13: the sum over the small frequencies compensates for the exponential behav-
ior, and the result is a much higher specific heat compared to the one of a single
frequency, which increases much slower with the temperature.
The Debye approximation. At intermediate temperatures the integral on
the right hand-side of Eq. (5.58) cannot be calculated analytically; a numerical
computation validates the linear behavior of the specific heat at low temperatures,
and the saturation at high temperatures [see Fig. 5.14(b)]. In order to simplify the
calculations, and to obtain expressions that are insensitive to details of the density
of states, Debye had proposed an approximation, named after him, for the density
of states. The exact form, Eq. (5.57), represented by the thick line in Fig. 5.14(a),
is replaced by a constant density of states, whose value is equal to that of the true
density of states in the acoustic limit of low frequencies, i.e., g(ω) = g0 = L/(πc).
This is depicted by the horizontal thin line in Fig 5.14(a). As seen, the two lines are
close to one another in the acoustic (low-frequency) range. Since the approximate
density of states is smaller than the true one (in particular at frequencies close
to ω0 ), it yields less states. To compensate for this fault, Debye had proposed
to extend the approximate density of states up to a new upper limit, termed the
Debye frequency and marked by ωD = kB ΘD / (ΘD is the Debye temperature;
it expresses the same frequency in units of temperature). This frequency is chosen
such that the total areas under the two curves in Fig. 5.14(a) are equal to one
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 342

342 Introduction to Solid-State Physics

(a) (b)

Fig. 5.14: (a) The thick upper line is the exact density of states of a single-atom,
one-dimensional lattice with nearest-neighbor interactions, in units of g0 = L/(πc),
as a function of ω/ω0 . The darker area below the horizontal line g = g0 , that extends
up to ωD /ω0 = π/2, is identical to the light area below the (exact) density of states
curve. (b) The specific heat of the same system in units of N kB , as a function of
kB T /(ω0 ). The thick solid line: an exact computation of the integral. The dashed
line: the Debye approximation. The thin solid line is obtained by approximating
the integral on the right-hand side of Eq. (5.58) by π 2 /3.

another, and the area is equal to the total number of vibrational modes. In the
units used
√ in the figure, the area under the curve of the exact density of states
1
is 0 dx/ 1 − x2 = π/2, while the new area [the darker one in Fig. 5.14(a)] is
xD ω
0
dx = xD , hence one chooses ωD /ω0 = xD = π/2. As 0 D dωg(ω) = N , one
finds N = LωD /(πc), i.e., g0 = N/ωD . It follows that

Debye kB N X ex
CV = dxx2 x , X = βωD . (5.60)
X 0 (e − 1)2
Figure 5.14(b) compares the exact result for the specific heat [Eq. (5.58)] with the
Debye approximation [Eq. (5.60)]. The two results are close to one another at the
two edges: at low temperatures the specific heat is dominated by the low frequen-
cies, where the density of states is indeed almost constant. At high temperatures
the specific heat is proportional to the total number of vibrational modes and the
Debye approximation is designed to comply with this number. At intermediate
temperatures there is a difference between the Debye approximation and the exact
result.
Typical values of the Debye temperatures are of the order of a few hundreds of
degrees. As in the discussion of Fig. 5.13, the Debye approximation is universal:
though the frequencies ω0 , and consequently the Debye temperatures, differ among
materials, the coefficient in front of the integral on the right hand-side of Eq. (5.60)
is C0 = kB N/(βωD ) = kB N T /ΘD , and the integral itself depends solely on X =
βωD ∝ ΘD /T . It follows that the specific heat depends on the temperature only
via the ratio T /ΘD ∝ kB T /(ω0 ). Hence, when the temperature and the specific
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 343

Lattice vibrations 343

heat of the acoustic vibrational modes of various one-dimensional materials are

scaled as in Fig. 5.14(b), the scaled experimental results lie on the same curve.
The contribution of the optical branch to the specific heat. The optical
branch appears when there are two atoms in the unit cell (nB = 2). From the
derivatives of the two dispersion relations in Eq. (5.15) (see Fig. 5.4), in conjunction
with the expression g(ω) = (L/π)/(dω/dk) for the density of states, it is possible
to deduce the density-of-states curves depicted in Fig. 5.15. As expected from Fig.
5.4, the density of states of the acoustic branch resembles that of the single-atom
system [Fig. 5.14(a)]. In fact, the density of states is “flatter”, implying that the
Debye approximation is more reliable in this case. The optical branch possesses van
Hove’s singularities as well, since the group velocity vanishes at its both ends. It
comprises relatively high frequencies within the range ω+ (π/a) < ω < ω+ (0). Its
contribution to the specific heat is
 ω+ (0)
eβω
CVoptical = kB dωg(ω)(βω)2 βω .
ω+ (π/a) (e − 1)2
At low temperatures this contribution is bounded by a constant proportional to
exp[−xmin ], where xmin = βω+ (π/a)  1, and hence it is rather tiny [see problem
5.17(c)], similarly to the illustration in Fig. 5.13. At high temperatures the integral
approaches N (the number of optical vibrational modes), and then the specific heat
is N kB , as expected from the Dulong-Petit law.

Fig. 5.15: The density of states of a one-dimensional crystal with two atoms in the
unit cell, in units of g0 = 2L/(ω0 a), as a function of ω/ω0 . The mass ratio is 3/2,
like in Fig. 5.4.

The Einstein approximation. Due to the exponential decay of the specific

heat of the optical modes at low temperatures, and the approach to a constant
value at high ones, and also since the integrals over complicated densities of states
(like the one that displays the optical branch in Fig. 5.15) are quite cumbersome,
it is customary to compute the contribution of the optical branch by the Einstein
approximation. In this approximation, one chooses a typical optical frequency
(for instance, the average frequency of the optical branch), finds the corresponding
specific heat [a single term in Eq. (5.55)], and then multiplies it by the number of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 344

344 Introduction to Solid-State Physics

vibrational modes of this branch (N in the case of a one-dimensional lattice). The

result resembles the one depicted in Fig. 5.13, multiplied by the number of vibra-
tional modes of this branch. This approximation is in particular successful when
the frequencies’ range of the optical branch is narrow, a rather ubiquitous situation.
Adding together the contributions of the two branches, yields a specific heat that
increases linearly with the temperature at low temperatures (much lower than the
maximal acoustic frequency or than the Debye temperature). At high temperatures
the acoustic-branch specific heat reaches the value N kB , in accordance with the
Dulong-Petit law. The contribution of the optical branch is negligible at tempera-
tures smaller than the minimal optical frequency (in units of temperature), but it
increases precipitately towards N kB at higher temperatures, so that asymptotically
the specific heat tends to the value 2N kB .
Problem 5.17.
a. Prove, by a direct calculation of the integral, that the density of states given in
ω
Eq. (5.57) has the normalization 0 0 dωg(ω) = N .
b. Prove Eq. (5.59). [Hint: to prove the first step, expand the integrand in powers
of exp[−x], and integrate by parts each term in the sum. To prove the second
step compare the Taylor series of the function f (x) = sin x/x with the mathematical

identity f (x) = m (1−x/xm ), where the product comprises all zeros of the function
on the real axis.]
c. Prove that at low temperatures the contribution of the optical branch to the
specific heat obeys the inequality CVoptical < kB N x2min exp[−xmin ], where xmin =
βω+ (π/a)  1, and therefore it decays exponentially when β increases.
Arbitrary dimensions. To derive the density of states at dimensions higher
than 1, recall that for each wave vector there are nB d vibrational modes; the cor-
responding frequencies are indexed by m = 1, 2, . . . , nB d. The sum over the wave
vectors is replaced by an integration for each value of m,
 
nB d

f (ωα ) = f [ωm (k)] = n(k)dd kf [ωm (k)] , (5.61)
α m=1 k m
where f (ω) is an arbitrary function of the frequency, e.g., as in Eq. (5.56). In
the last step, the discrete sum over k is transformed into an integral over the first
Brillouin zone; n(k)dd k is the number of (discrete) wave vectors in the volume
element dd k. Recall that, quite generally, the discrete wave vectors are given by
k1 ,2 ,3 = ( 1 /N1 )b1 + ( 2 /N2 )b2 + ( 3 /N3 )b3 , with −Ni /2 ≤ i < Ni /2 − 1. The
volume of the first Brillouin zone, which is the volume of a unit cell in reciprocal
lattice, is Vrec = (2π)d /V , where V is the volume of the unit cell in the original
lattice (see problem 3.6). The Brillouin zone comprises N = N1 N2 . . . Nd discrete,
uniformly distributed, wave vectors. Therefore, the density of points in mo-
mentum space (i.e., the number of points in the Brillouin zone) is
N NV
n(k) = = , (5.62)
Vrec (2π)d
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 345

Lattice vibrations 345

where N V is the total volume of the crystal (the number of unit cells times the
volume of one of them).
In the next step, the integration variables are changed, from k to ω,
  NV  
n(k)dd kf [ωm (k)] = d d
kf [ω m (k)] = dωgm (ω)f (ω) ,
m m
(2π)d m
(5.63)
where

NV
gm (ω) = dd kδ[ω − ωm (k)] (5.64)
(2π)d
is the density of states of the m−th vibrational mode [substituting Eq. (5.64)
in the integral on the right hand-side of Eq. (5.63) gives exactly the terms that
appear in the second equality there]. Phonons, in addition to their contribution
to the specific heat, participate in many physical processes, e.g., heat conductance,
scattering of electrons (that lead to a reduction of the electrical conductance), and
more. These all involve the phonon density of states, given by Eq. (5.64).
In Eq. (5.64), gm (ω)dω is the number of vibrational modes (of the m−th type),
located in-between two equal-frequency surfaces in the Brillouin zone, of frequency
ω and ω + dω, respectively, see Fig. 5.16. As the density of the wave vectors in the
Brillouin zone is uniform, it suffices to find the volume enclosed in-between the sur-
faces, and to multiply it by that density. Consider a point k located on the “lower”
surface; the vector ∇k ω(k) is perpendicular to that surface, and points along the
direction of increasing ω. The distance between the two surfaces in k−space, at
this point, Δk⊥ , is given by the relation Δω = |∇k ω(k)|Δk⊥ . Defining an area
element on the surface by dSω , the volume in-between it and the area above on the
neighboring surface is dSω Δk⊥ = ΔωdSω /|∇k ω(k)|, leading to

NV dSω
gm (ω) = , (5.65)
(2π) S |∇k ωm |
d

where the integration is carried out over the entire surface (which is a closed one,
encompassing the origin) on which the frequency is constant and equals ω. This
result is a generalization of Eq. (5.57): the group velocity in arbitrary dimension
is given by vg = ∇k ω|k=k0 . In one dimension, the coefficient in Eq. (5.65) is
N a/(2π) = L/(2π), and the “area” S encompassing the first Brillouin zone includes
solely two points. Equation (5.65) can be also derived directly from Eq. (5.64): the
identity δf [(x)] = δ[f  (x0 )(x − x0 )] = δ(x − x0 )/f  (x0 ), valid when f (x0 ) = 0 [see
part d of problem (B.2)] enables one to write
   
d kδ[ω − ωm (k)] =
d
dSω dk⊥ δ[k⊥ − k⊥ (ωm )]/(dω/dk⊥ ) = dSω /|∇k ωm | ,
S
(5.66)
where k⊥ is the component of k normal to the surface.
As seen in Sec. 5.2, the dispersion relations in higher dimensions are quite com-
plicated, and consequently the corresponding densities of states are complicated as
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 346

346 Introduction to Solid-State Physics

Fig. 5.16: Two equal-frequency surfaces in two dimensions. The dark area represents
dSω dk⊥ (see text).

(a) (b)
Fig. 5.17: Equal-frequency curves (in kx − ky plane) for the dispersion relation of
vibrations normal to the plane due to nearest-neighbor interactions. (a) The square
lattice of Fig. 5.6(a). (b) The triangular lattice of Fig. 5.8(a).

well. Figure 5.17 displays the equal-frequency curves (corresponding to the “area”
S) for the vibrations normal to the plane in the square and the triangular lattices,
with nearest-neighbor interactions, according to Figs. 5.6 and 5.8. Around the
center of the Brillouin zone, in the acoustic region, these lines are circles, the dis-
tances in-between them are identical along any direction, and the density of states
takes a relatively simple form. However, farther away from the center the curves
follow the symmetry of the lattice, and the distances between them depend on the
direction of the wave vector. Moreover, for the vibrations within the plane of the
square lattice the equal-frequency
 curves are not circular even in the acoustic range:
the relation ω ∝ a (DL /M )[kx + (DT 1 /DL )ky2 ] (Sec. 5.2) gives rise to elliptical
2

equal-frequency lines. Nonetheless, given the explicit expression of the dispersion

relation, the density of states can be derived from Eq. (5.66). Though the group
velocity vanishes at special points in the Brillouin zone, the integration over the
equal-frequency surfaces weakens the divergencies of the density of states, and thus
the van Hove singularities of dimensions higher than one are weakened. As seen in
problem 5.18, the divergence is logarithmic in two dimensions, and is not as strong
as the one of one dimension, Eq. (5.57). In three dimensions, the density of states
does not diverge at all, but its derivative does, see Fig. 5.19(b).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 347

Lattice vibrations 347

Problem 5.18.
a. Find the group velocity of the vibrations normal to the plane in the single-atom
square lattice with interactions between nearest-neighbors, whose dispersion relation,
M ω2 = K  zz (k), is displayed in Eq. (5.26) and in Fig. 5.6(a). At which points in
the first Brillouin zone does this velocity vanishes?
b. Give an expression for the dispersion relation at the corners of the Brillouin
zone. Find the density of states at the vicinity of such a corner.
c. As seen in Fig. 5.17(a), the straight lines that join the points of the type M [i.e.,
(π/a, 0) → (0, π/a) → (−π/a, 0) → (0, −π/a) → (π/a, 0] form an equal-frequency
“surface” for this lattice. Prove this, and show that the density of states is infinite at
the frequency corresponding to this curve, and diverges logarithmically at its vicinity.
d. Describe schematically the dependence of the density of states on the frequency.

Fig. 5.18: The measured dispersion relations at various directions in the first Bril-
louin zone of copper. The frequency is ν = ω/(2π). [E. C. Svensson, B. N. Brock-
house and J. M. Rowe, Crystal dynamics of Copper, Phys. Rev. 155, 619 (1967)].

An example of a three-dimensional lattice. The complexity of the disper-

sion relations in three dimensions, in particular when there are many branches, is
illustrated in Fig. 5.18, which displays those relations for copper (an FCC lattice).
The curves in the figure are experimental data measured by inelastic neutron
scattering (discussed below) for the wave-vector dependence of the frequency at
various directions [see Fig. 5.19(a)] in the first Brillouin zone. Figure 5.19(b)
illustrates the density of states as found from model calculations designed to fit
the experimental data. Such a model can be based on educated “guesses” for the
spring constants of the different interactions, which are adjusted until a plausible
agreement with the data is reached [see also problem (s.5.6)]. At low frequencies,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 348

348 Introduction to Solid-State Physics

(a) (b)

Fig. 5.19: (a) The first Brillouin zone of an FCC lattice. The coordinates of the
special points Γ, X, W , K, and L are marked below the horizontal line in Fig. 5.18;
the wave vectors are measured in units of 2π/a. For instance, the wave vector in the
left part of Fig. 5.18 refers to the wave vectors (0 0 ζ)(2π/a), where 0 < ζ < 1, and
that in second segment refers to (0 ζ 1)(2π/a). (b) The density of states derived
under certain assumptions about the “spring” constants (the thin line) and the
Debye approximation for the same system (the thick line). Part (b) is based on the
reference cited in Fig. 5.18.

i.e., in the acoustic branch, the density of states is proportional to ω 2 , as would have
been obtained from a linear dispersion relation, like the one derived in the Debye
approximation. The three-dimensional density of states contains quite a number
of singular points, at which the derivative dg/dω is discontinuous. These points
appear whenever an equal-frequency surface cuts the face of the first Brillouin zone.
This behavior is related to the van Hove singularities: the group velocity vanishes
at certain points near the corresponding surface in the Brillouin zone, but the two-
dimensional integration [Eq. (5.65)] “mitigates” the divergence. Figure 5.19(b)
shows the density of states found by the Debye approximation for this system.
The Debye and the Einstein approximations; arbitrary dimensions.
The approximations of Debye and Einstein are also useful at dimensions higher
than one. That of Einstein is the same in any dimension: each of the optical
branches contributes to the specific heat a function similar to the one in Fig. 5.13,
whose shape is dictated by the frequency typical to that branch. As opposed, the
Debye approximation requires a re-examination. Recall that in one dimension, this
approximation replaces the density of states of the material by a simple form, that
captures the acoustic behavior at low frequencies, and extends that form up to
the Debye frequency, which in turn is determined such that the total number of
vibrational modes is conserved. This rather crude approximation produces results
quite reliable for the specific heat. In the simplest version of this approximation
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 349

Lattice vibrations 349

the dispersion relation is assumed to be linear in the wave vector magnitude, i.e.,
it is independent of the direction of the wave vector, ωm = cm |k|. Consequently,
the equal-frequency surfaces are spheres (in three dimensions) or circles (in two
dimensions, as the ones near the origin in Fig. 5.17) in momentum space. The
density of states of a single acoustic vibrational mode in Eq. (5.65) is
NV
gm (ω) = S ω d−1 , (5.67)
(2πcm )d d
where Sd is the area of the sphere (or the perimeter of the circle) whose radius
is 1 in d−dimensions [S3 = 4π, S2 = 2π, S1 = 2] and V is the volume of the
d−dimensional crystal. The integrations over the frequency are carried out up to
the Debye frequency, ωDm , determined by the conservation of the total number of
modes,
 ωDm
N V Sd  ωDm d
N= dωgm (ω) = , (5.68)
0 d 2πcm
and hence ωDm = 2πcm (V Sd /d)−1/d , and gm (ω) = N dω d−1 /ωDm d
. Inserting Eqs.
(5.67) and (5.68) into Eqs. (5.55) and (5.61) yields the specific heat in the Debye
approximation, for each of the acoustic branches,
 βωDm
CV = kB dN (βωDm )−d dxxd+1 ex /(ex − 1)2 . (5.69)
0
At low temperatures, kB T  ωDm , the integration may be carried out up to
infinity [like in Eq. (5.59)]; the integral then converges to a constant value,
 ∞
dxxd+1 ex /(ex − 1)2 = Id = (d + 1)!ζ(d + 1) ,
0

where I1 = 2ζ(2) = π 2 /3 ≈ 3.29, I2 = 6ζ(3) ≈ 7.212, and I3 = 24ζ(4) = 4π 4 /15 ≈

∞
25.98. [ζ(s) = n=1 n−s is the Riemann zeta function; its numerical values are de-
rived in a method similar to that presented in connection with the one-dimensional
case, problem 5.17(b).] The corrections to this approximation are exponentially
small (see problem 5.19). The specific heat at low temperatures of the acoustic
modes of the m−th branch is thus proportional to (kB T /ωDm )d . As the den-
sity of states varies smoothly at low frequencies, this result pertains also to the
full (exact) solution of the specific heat. It is worthwhile to emphasize again that
the low-temperature specific heat is dominated by the acoustic modes, as the con-
tributions of the optical ones decay there exponentially. At high temperatures,
kB T  ωDm , the integral encompasses all states, and gives the Dulong-Petit law,
CV → kB N , for each of the nB d modes. The Debye temperatures in three dimen-
sions, ΘDm = ωDm /kB , is of the order of a few hundreds degrees K.
Three dimensions. Equation (5.68) yields a different Debye temperature for
each of the acoustic branches, i.e., the density of states of each branch is parabolic
(in three dimensions) up to the corresponding Debye frequency, and then drops to
zero. Accordingly, the specific heat is a sum of three different integrals, extending
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 350

350 Introduction to Solid-State Physics

up to three different bounds. At times it is convenient to return to Eq. (5.67),

and to write down the total density of states of the three vibrational modes (the
longitudinal and the two transverse ones) in the small-frequency limit,
 NV  1 2 
3
3N V 2
g(ω) = gm (ω) = 3 3 + 3 S3 ω 2 ≡ 2 c3
ω , (5.70)
m=1
(2π) c L c T 2π
where the average sound velocity c is defined by 3/c3 = [1/c3L ] + [2/c3T ]. The Debye
temperature is chosen so that the total number of vibrational modes is 3N ,
 ωD
3N V  ωD 3
3N = dωg(ω) = , (5.71)
0 2π 2 c
i.e., g(ω) = 9N ω 2 /ωD
3
, and ωD = (6π 2 /V )1/3 c. This expression is rather useful for
explicit calculations, and gives the correct behavior of the specific heat at both low
and high temperatures.

Problem 5.19.
The Debye function in d dimensions [appearing in Eq. (5.69)] is defined by
 X
d xd+1 ex
Dd (X) = d dx x .
X 0 (e − 1)2
a. For X  1 (which corresponds to high temperatures), show that Dd (X) =
1 − dX 2 /[12(d + 2)] + . . .. This is the leading correction to the Dulong-Petit law.
b. For X  1 (corresponding to low temperatures), prove that Dd (X) = [d/X d ]Id −
dX exp[−X] + O(exp[−X]). This is the leading correction to the T d behavior of the
vibrational modes’ specific heat at low temperatures.

Problem 5.20.
When the matrix Kμν is diagonal, the dispersion relations are M ωμ2 = K  μμ , where
the right hand-side is given by the Fourier transform of Kμμ (R) [Eq. (5.25)]. For
long-range interactions, for which Kμμ (R) = D1 /Rη+d , with η > 0, the leading
term in the Fourier transform is K μμ ∝ |k|η (under appropriate conditions). In the
extended Debye approximation it is assumed that the dispersion relation ω 2 ∝ |k|η
is valid at all frequencies, up to a generalized Debye frequency; the latter is chosen
so that the total number of the vibrational modes is conserved. Give expressions for
the density of states and for the Debye frequency in this model. Derive the low and
high-temperature limits of the specific heat.

Problem 5.21.
Use the Debye approximation to obtain an expression for the acoustic specific heat
of a system of the shape of box, of size Lx × Ly × L, with Lx = Ly  L. When
the thickness L of the system is significantly shorter that the other dimensions,
a two-dimensional behavior is expected. Use tetragonal symmetry with the “thin”
direction normal to the plane of the system, and show that at low temperatures the
temperature dependence of the specific heat crosses over from the two-dimensional
behavior to the three-dimensional one; explain how this transition depends on L.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 351

Lattice vibrations 351

5.5 Crystal stability and the Debye-Waller factor

Mermin-Wagner theorem. The Lindemann criterion for the melting temper-

atures of solids is discussed in Sec. 3.10. There, it is assumed that each of the atoms
in the lattice oscillates independently of the motions of all other; the estimates for
the melting temperature are then based in turn on estimates of the average of the
square of the oscillation’s amplitude. In view of the discussions in the present chap-
ter, that analysis can be improved. As seen from Eq. (5.44), The amplitude of the
vibration of each atom, u(R0i ), is affected by the motions of all others. In fact, it
is a sum over all dnB N vibrational modes (recall that d is the dimension, nB is the
number of atoms in the unit cell, and N = N1 N2 . . . Nd is the number of unit cells).
The average of the square of this amplitude is u2m (R0i ), where m = 1, 2, . . . , nB d
counts the various vibrational modes. Due to the symmetry with respect to dis-
placements of the lattice, the mean-square deviations of all atoms of the same type
are identical. This allows for an average over all unit cells,
1  2 1 
u2m (R0i ) = um (R0i ) = 
um (k)
um (−k) , (5.72)
N i N
k

where the expansion in the eigenmodes, Eq. (5.44), is used; the rest of the derivation
follows the same route as that leading to Eq. (5.46) (check!). The k−sum runs over
the discrete values of the wave vectors, as defined in Eq. (5.43).
The right hand-side of Eq. (5.72) is the average of the amplitude squared of a
harmonic oscillator, whose frequency is ωm (k). As seen in Sec. 3.10, one has first to
calculate the quantum average, and then the thermal (statistical-mechanical) one,
with the Boltzmann weights. The result is [see Eq. (3.59)]
  βω (k) 

um (k)um (−k) = coth m
, (5.73)
2M ωm (k) 2
where the angular brackets represent the double average, over the quantum proba-
bility and over the thermal one. When this expression is inserted into Eq. (5.72),
and the sum there is replaced by an integration over the frequency, one finds
  βω 
1   dω
u2m (R0i ) = 
um (k)
um (−k) = gm (ω)coth . (5.74)
N 2M N ω 2
k

The integral on the right side is over a limited range of frequencies for the optical
branches, and therefore it converges. This is in particular true when the Einstein
approximation, which accounts for a single frequency, is used. In this approxima-
tion, the results derived in Sec. 3.10 are valid. As opposed, the same integral for
the acoustic modes is not well-behaved: as seen from Eq. (5.67), at low frequencies
g(ω) ∝ ω d−1 . Consider first the limit of zero temperature. There, Eq. (5.74) yields

um (k)um (−k) = /(2M ωm ), and consequently
 
 dω
u2m (R0i ) = gm (ω) ∼ dωω d−2 .
2M N ω
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 352

352 Introduction to Solid-State Physics

As this expression is inversely proportional to the mass of the atoms, crystals com-
prising lighter atoms are less stable. The integral converges for d > 1, but diverges
(due to the lower bound) at one dimension. This divergence obviously invalidates
the Lindemann criterion: the average amplitude of each atom is by far longer than
the lattice constant, and thus the crystal cannot remain stable, that is, in one
dimension the harmonic vibrations destroy the crystalline structure even at zero
temperature. At higher dimensions, however, the harmonic vibrations are stable
down to zero temperature.
What happens at finite temperatures? At low frequencies, ω  kB T /, one
finds um (k)u(−k) ≈ kB T /(M ωm 2
), and therefore the integrand in Eq. (5.74) is
proportional to ω d−3
. It thus diverges (at the lower bound) for d ≤ 2, and the Lin-
demann criterion implies that a two-dimensional crystal disintegrates at any finite
temperature. This has been already noted by Landau and Peierls and precisely
phrased later by Mermin and Wagner. The Mermin-Wagner theorem states
that within the harmonic approximation, there are no stable one- and
two-dimensional crystals at finite temperatures. At zero temperature
there are no one-dimensional stable crystals.
Does that mean that one should abandon the conclusions drawn from the pre-
vious discussions regarding one- and two-dimensional solids? Not necessarily. Con-
trary to the Mermin-Wagner theorem, experimental observations indicate that effec-
tive one-and two-dimensional materials do exist in Nature (recall, e.g., that graphene
is a two-dimensional lattice). One has to understand why the Mermin-Wagner the-
orem does not pertain to them; several possible explanations are listed below.
Anharmonic effects. As mentioned, the Mermin-Wagner theorem is valid for
harmonic vibrations alone, and does not pertain to anharmonic terms. Indeed,
when the potential energy of a certain vibrational mode is of the form U (u) =
M ω 2 u2 /2 + Bu4 /4, with B > 0, then even though the frequency found within
the harmonic approximation tends to zero, the quartic term confines the motion.
Classically, in the ω → 0 limit, the energy of the particle is E = p2 /(2M ) + Bu4 /4;
the particle oscillates between the two points u = ±[4E/B]1/4 , and the amplitude
squared is of the order of u2 ∝ 4E/B. The classical Boltzmann average (discussed
in problem 5.22) is

duu2 exp[−βBu4 /4] 4kB T dxx2 exp[−x4 ]
u  =
2
=
du exp[−βBu4 /4] B dx exp[−x4 ]
(the integration variable is changed in the second step). The integrals on the right
hand-side converge [their values are tabulated, and yield Γ(3/4)/Γ(5/4) ≈ 1.35].√ It
follows that the average of the amplitude squared is finite and proportional to T .
The quantum-mechanical calculation is subtle, as there is no analytical solution for
the anharmonic oscillator; however, the motion is bounded, and the anharmonic
terms stabilize the crystal.
Growth on a substrate and meta-stability. Often the low-dimensional sys-
tem is attached to a three-dimensional substrate (examples are the absorbed layers
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 353

Lattice vibrations 353

described in Sec. 2.9 and Sec. 3.12). Even when the coupling to the substrate is
weak, it suffices to stabilize the system. The two-dimensional layers fabricated in
the lab are usually grown on a substrate, and then are peeled off. This is in par-
ticular true for graphene. Attempts to grow free crystals of graphene from isolated
carbon atoms end up with small non-planar crystals, e.g., the Buckminster-Fuller
spheres or the nano tubes (Fig. 1.9). Apparently during the peeling operation,
the atoms keep the hexagonal structure they have had in the graphite, as this is a
metastable state of the system (similar to the one of a glass): to reach a more stable
state the system has to overcome numerous potential barriers, a process which is
usually rather slow.
Finite size. As mentioned, samples of graphene are not in particular large;
they are of the order of 100 microns. A possible explanation for their stability
is based on their finite, rather small, size. The integral on the right hand-side
of Eq. (5.74) is not continuous for a finite crystal; it is an approximation of a
sum over discrete frequencies that correspond to discrete wave vectors (i.e., the
central part of the equation). The latter, for a two-dimensional system, are k1 ,2 =
( 1 /N1 )b1 + ( 2 /N2 )b2 . The length of the shortest wave vector is of the order of
|bi |/Ni . Assuming that the lengths of the crystal in both directions are similar,
L1 = N1 a1 and L2 = N2 a2 , one finds that kmin = 2π/max{L1 , L2 } = 2π/Lmax .
At the small frequencies that belong to this wave number the dispersion relation is
acoustic, and therefore ωmin = ckmin = 2πcm /Lmax . Even if the sum in Eq. (5.72)
is approximated by an integral, the lower bound of the latter is (up to a constant)
ωmin . Indeed, the Debye approximation (5.67) in conjunction with Eq. (5.74) for
d = 2, gm (ω) = N Sω/(2πc2m ), yields

  βω   sinh[βω /2] 

 ωD
dω SkB T
u2m (R0i ) = gm (ω)coth = ln D
,
2M N ωmin ω 2 M 2πc2m sinh[βωmin /2]
(5.75)

(check!) where S is the area of the two-dimensional unit cell. This expression di-
verges when applied to an infinite crystal, for which Lmax → ∞. However, for a
finite crystal for which Lmax  βπcm , the logarithmic term on the right hand-side
is dominated by − ln[sinh(βπcm /Lmax )] ≈ ln[Lmax /(βπcm )]. As the logarithmic
function varies slowly with its argument, the divergence of the amplitude squared
as the crystal is increased is slow as well, and it is possible to achieve small dis-
placements in relatively large samples. Moreover, the argument of the logarithmic
function depends on the ratio Lmax /β, which decreases as the temperature is re-
duced. At sufficiently low temperatures, i.e., when kB T  ωmin = πcm /Lmax , the
logarithm on the right hand-side of Eq. (5.75) is approximately β(ωD − ωmin )/2,
and consequently the amplitude squared tends to a finite limit which is independent
of the temperature. Thus, relative large stable crystals can exist within a range of
non zero temperatures.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 354

354 Introduction to Solid-State Physics

Problem 5.22.
Find the leading dependence of the average of the amplitude squared in a one-
dimensional lattice (with a lattice constant a) on the sample’s length, L. Is this
lattice stable according to the Lindemann criterion?
Long-range forces. The denominator in Eq. (5.75) contains the sound velocity
of the material. As shown in Sec. 5.1, the sound velocity increases when interactions
with fartheraway neighbors are included. For instance, in problem 5.1 it is found
that c = a 2
m m DL,m /M , which obviously becomes larger as more terms are
added to the sum. It follows that when more couplings are taken into account, the
average of the amplitude squared decreases, and larger crystals can be stabilized.
This is even more so when the interaction decays slowly with the distance, e.g., like
in problem 5.20: a spring constant that decays with the spring’s length, D(R) =

D1 /|R|d+η yields K(k) ∝ k η , and in turn gives rise to g(ω) ∝ ω d/η−1 . Inserting this
ω
form into Eq. (5.74) gives, in the zero-temperature limit, u2 (Ri0 ) ∼ 0 D dωω d/η−2 ;
this integral converges for η < d. As opposed, at finite temperatures, the behavior
ω
near the lower bound of the integral is u2 (Ri0 ) ∼ 0 D dωω d/η−3 , and this integral
converges for η < d/2.
Topological phases. The heating of a solid crystalline causes the creation
of crystalline imperfections whose number increases till the melting temperature is
reached, and the crystal melts to become a liquid. As an example, consider a type
of defects that is relevant to the issue of stability in two dimensions: dislocations.
The dislocation displayed in Fig. 5.20(a) is a linear default, which is formed in a
three-dimensional system when one of the lattice planes terminates along a line,
while the distances between the other planes decrease, such that far away from the
linear defect the ordered periodic structure is restored. In two dimensions [Fig.
5.20(b)] the dislocation is a point in the plane where one of the lattice lines ter-
minates. The dislocation is a topological phenomenon: in an infinite lattice it
cannot be eliminated even when the lattice is deformed continuously. (For instance,
one can deform a cup with a handle into a ring, but one cannot eliminate the single
hole that exists in the ring and in the handle of the cup; in fact, the number of holes,
as the number of dislocations, is not changed by the deformation.) The dislocation
in a finite lattice can be removed only by shifting it to the boundary, or when two
“inverse” dislocations are joined so that they cancel one another. The energy “cost”
of the creation of a dislocation in two dimensions increases logarithmically with the
distance from the dislocation; as a result, a finite concentration of dislocations de-
stroys the solid phase. On the other hand, when two dislocations are joined together
(without cancelling one another), their joint effect at long distances is smaller. In
fact, as discovered by Kosterlitz and Thouless and in parallel by Berezinskii, the pe-
riodic crystalline phase in two dimensions is replaced by another phase, in which all
dislocations are paired. In this phase, called hexatic, the Bragg peaks are widened,
which indicates the destruction of the periodic order. The relationships between the
directions of the bonds, however, are maintained. For instance, the angles between
the bonds are 0◦ or 60◦ , as in the triangular lattice. The hexatic phase appears also
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 355

Lattice vibrations 355

in liquid crystals, see Fig. 7.2. At the melting temperature of the hexatic phase
the pairs of dislocations disintegrate, and the material becomes a liquid. This is
the Berezinskii-Kosterlitz-Thouless phase transition, that led to the discoveries of
other topological phase transitions.

(a) (b)

Fig. 5.20: (a) A dislocation in three dimensions. (b) A dislocation in two dimensions.

The Debye-Waller factor. The vibrational modes affect the form of the
Debye-Waller factor, introduced in Sec. 3.10. The right hand-side of Eq. (3.57) is
replaced by sums over the various modes that participate in the oscillations of the
atoms. Using the Debye approximation for arbitrary dimension, Eqs. (5.67) and
(5.74), one finds
  ωD  βω 
 d
u2m (R0i ) = d
dωω d−2
coth (5.76)
D 2M ωD ωmin 2
for each acoustic branch. Each optical mode, in the Einstein approximation, gives
   βω 

u2m (R0i ) = d
coth E
, (5.77)
E 2M ωE 2
where ωE is the frequency which represents the optical mode.

Problem 5.23.
Assuming that the melting temperature is higher than the Debye one, use the Linde-
mann criterion to express the melting temperature of a single-atom, d−dimensional
lattice, in terms of its Debye temperature.

5.6 Impurities, localized modes, scattering of sound waves

In reality, crystals are not perfectly periodic; they contain a finite (though at times
minute) concentration of “foreign” atoms, (“impurities”), or other flaws. The full
description of disordered crystals, with randomly-distributed defects, is beyond the
scope of this book. Instead, the effect of a small amount of impurities on the lattice
vibrations of a simple system is studied. Consider for simplicity a single foreign
atom that occupies one of the lattice sites of a one-dimensional lattice compris-
ing 2N + 1 atoms. Let this atom be located at n = 0, and all other atoms at
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 356

356 Introduction to Solid-State Physics

n = ±1, ±2, . . . , ±N . Though the following solution is obtained for rigid boundary
conditions, u±(N +1) = 0, similar solutions can be worked out for other boundaries
(see problem 5.24), and for an infinite lattice as well. The atom at n = 0 is specified
by its mass, m, which differs from the mass M of the host atoms. The constants
of the springs that couple it to its nearest neighbors, D0 , also differ from all other
spring constants, D. The equations of motion are
mü0 = − D0 (2u0 − u1 − u−1 ) ,
M ü±1 = −D0 (u±1 − u0 ) − D(u±1 − u±2 ) ,
M ü±n = −D(2u±n − u±(n−1) − u±(n+1) ) , 1<n<N −1 ,
M ü±N = −D(2u±N − u±(N −1) ) . (5.78)
These equations are invariant with respect to reflection through the origin, n ↔ −n.
Therefore the solutions are symmetric or anti-symmetric with respect to such a
reflection. It is thus helpful to define symmetric and anti-symmetric functions,
u±
n = un ± u−n , for 1 ≤ n ≤ N ; the equations of motion then separate,

mü0 = −D0 (2u0 − u+

1) , M ü− − − −
1 = −D0 u1 − D(u1 − u2 ) ,

1 = −D0 (u1 − 2u0 ) − D(u1 − u2 ) ,

M ü+ + + +

± ±
M ü± ±
n = −D(2un − un−1 − un+1 ) , 2 ≤ n ≤ N − 1 ,
M ü± ± ±
N = −D(2uN − uN −1 ) . (5.79)
Thus there are N equations for u− − − + + +
1 , u2 , . . . uN , and N +1 ones for u0 , u1 , u2 , . . . uN .
For 2 ≤ n ≤ N − 1, the equations are the “usual” ones for an homogeneous
chain; the solutions are un ∝ C ±n exp[−iωt] (this is implied by the Bloch theorem
mentioned in Sec. 5.1). For C = exp[ika], where ka is real, these are propagating
wavy solutions. Otherwise, these solutions grow or decay exponentially. Hence
u± ± n ± −n −iωt
n = (A C + B C )e , (5.80)
provided that the “usual” dispersion relation is obeyed,
M ω 2 = D(2 − C − 1/C) . (5.81)
Anti-symmetric solutions. These solutions obey two more equations, for
n = 1 and for n = N . The symmetric case is discussed separately below. Indeed,
inserting Eqs. (5.80) into the equations of motion yields
(M ω 2 − D − D0 )(A− C + B − /C) + D(A− C 2 + B − /C 2 ) = 0 ,
(M ω 2 − 2D)(A− C N + B − /C N ) + D(A− C n−1 + B − /C N −1 ) = 0 . (5.82)
Using the dispersion relation (5.81), one obtains
B− (D − D0 )C − D
− −
= C 2N +2 = . (5.83)
A (D − D0 )/C − D
When D = D0 , then C 2N +2 = 1, which entails C = exp[ika], with k = k =
π /[(N +1)a], = 1, 2, . . . , N . In this case B − = A− , and thus u− −
n = 2iA sin(nka).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 357

Lattice vibrations 357

These are just the wavy solutions that would have been obtained were the foreign
atom identical to all others (see e.g., problem 5.13), since the symmetric vibrations’
amplitudes, including u0 , vanish. As the oscillating atoms are not connected to the
foreign atom, their amplitudes do not depend on its mass.
One may still look for wavy solutions when D = D0 . Indeed, using C = exp[ika]
in Eq. (5.83), assuming that ka is real, turns that equation into exp[2i(N + 1)ka] =
exp[2iχ], with
(D − D0 ) sin(ka)
tanχ = .
(D − D0 ) cos(ka) − D
It follows that k = k + χ(k)/[(N + 1)a]; this transcendental equation is solved
graphically or numerically for each . Thus, there are N wavy solutions, with k
displaced from its previous value π /[(N + 1)a]. The corresponding frequencies,
however, are still within the acoustic range found before.
Symmetric solutions. To derive these solutions, one first substitutes u0 =
−D0 u+ 1 /(mω − 2D0 ) in the equation for u1 , and then finds that the coefficients
2 +
+ +
A and B obey
 2D02  B+   B+ 
M ω 2 − D − D0 − A+
C + + D A + 2
C + =0,
mω 2 − 2D0 C C2
  B+   B+ 
M ω 2 − 2D A+ C N + N + D A+ C N −1 + N −1 = 0 . (5.84)
C C
The second equation is identical to that obtained in the solution for the anti-
symmetric functions. All in all, one finds
B+ (D − D0 − X)C − D
− = C 2N +2 = , (5.85)
A+ (D − D0 − X)/C − D
where X = 2D02 /(mω 2 − 2D0 ) and M ω 2 = D(2 − C − 1/C). As before, one may
insert C = exp[ika], to obtain N wavy solutions, with k = k + χ1 (k)/[(N + 1)a],
with
D − D0 − X) sin(ka)
tanχ1 (k) = .
(D − D0 − X) cos(ka) − D
Localized vibrational mode. As all in all there are 2N + 1 solutions, out
of which 2N are wavy ones as found above, there is yet one more solution to be
derived. One tries a solution for |C| < 1 (check that a trial solution with |C| > 1
gives rise to a similar result). For large enough N , the central part in Eq. (5.85) is
very small. When it is completely ignored, one finds (D − D0 − X)C − D = 0. With
the dispersion relation (5.81), there emerges a cubic equation for C. The trivial
value C = 1 is always a solution, and thus there remains a quadratic equation.
Usually only one of its solutions complies with all the requirements. Consider for
simplicity the case D = D0 . Then the afore-mentioned equation becomes
 2M  2  M
1− C −2 1− C +1=0 ,
m m
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 358

358 Introduction to Solid-State Physics

with the trivial solution C = 1 and the non trivial one C = m/(m − 2M ). The
requirement |C| < 1 is equivalent to m < M , and then −1 < C < 0, implying that
C n = (−|C|)n and the wave function alternates its sign between nearest neighbors.
In case M < m, then |C| > 1, and the whole procedure has to be repeated for that
value.
Inserting the result into Eq. (5.81) yields
4DM 2
M ω2 = . (5.86)
m(2M − m)
One easily verifies that ω 2 > 4D/M = ω02 , where ω0 is the maximal frequency of
acoustic waves in one dimension (check!). When m = M one finds that ω = ω0 , as
expected. As seen in Sec. 5.1, frequencies higher than ω0 correspond to exponential
functions, which are not waves. Such solutions lead to unphysical consequences in
the homogeneous system (as they diverge to infinity at one of the edges). However,
here they decay on both sides of the impurity atom, and therefore they are accept-
able. It is worth noting that for 2M < m the frequency becomes imaginary, and
consequently the solutions decay with time.
Returning to Eq. (5.85), one finds that for sufficiently large N , |B + /A+ | =
|C| 2(N +1)
 1, and thus u+n ≈ A C = A (− exp[−κa]) ; that is, the amplitudes
+ n + n

of the displacements decay with the distance from the foreign atom, with the decay
coefficient κ = − ln |C|. The coefficient is larger as the two masses become more
disparate. This special mode is confined to the vicinity of the foreign atom; it is
called a localized vibrational mode.
In summary, in the presence of a foreign atom, there are 2N + 1 vibrational
modes, as found also for a homogeneous system; the difference is that in the present
case only 2N of them possess frequencies in the acoustic range. One vibrational
mode is distinct, and has a discrete frequency. When there is a single foreign
atom on the lattice, the contribution of its vibrations to the specific heat is smaller
by a factor of 2N compared to the contributions of all other vibrations, making
it difficult to be detected in experiments. In reality, however, there is a finite
concentration, p, of foreign atoms. When this concentration is small, it is plausible
to assume that the foreign atoms are far away from each other; consequently, the
localized modes that are centered around each of the foreign atoms are similar
to each other, and thus have the same frequency. In this case the contribution
of the acoustic branch to the specific heat is multiplied by 1 − p; the impurities’
contribution is of the Einstein approximation type, with the frequency found above,
and is proportional to the number of foreign atoms, i.e., 2N p. Such a term in the
specific heat indicates the existence of impurities in the crystal.

Problem 5.24.
Find the solutions of Eqs. (5.78) for periodic boundary conditions, i.e., uN = u−N .

The generalization of this discussion to more complex one-dimensional systems,

e.g., with larger unit cells, yields similar results. In higher dimensions the equations
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 359

Lattice vibrations 359

turn out to be rather complicated. For instance, the localized states are not spher-
ically symmetric (i.e., the decay coefficient depends on the direction); the analysis
of those is beyond the scope of this book.
Scattering of phonons off an impurity. Impurities, beside causing localized
vibrational modes, can also scatter the waves (i.e., scatter the phonons), and thus
hinder the propagation of these waves in the crystal. Consider for concreteness a
heavy foreign atom; the scattering off such an atom can be viewed as elastic – there is
no energy exchange between the phonons and the scattering atom. To demonstrate
this important phenomenon it is useful to consider a wave that impinges from the left
on the one-dimensional crystal discussed above. The original wave propagates from
left to right, and therefore its wave function is of the form un = A exp[i(kna − ωt)].
Once this wave hits the foreign atom, a reflected wave is created that moves to
the left, with the same wave number and with the same frequency. Denoting the
amplitude of the reflected wave by Ar (r is the reflection amplitude of the wave),
its function is un = Ar exp[i(−kna − ωt)]. One therefore conjectures a solution of
the form
un = A(eikna + re−ikna )e−iωt , for n ≤ 0 . (5.87)
On the other hand, the wave on the right side of the impurity moves entirely to
the right; its origin is that part of the impinging wave that succeeded to pass the
impurity. One therefore conjectures
un = Ateikna e−iωt , for n ≥ 0 , (5.88)
where t is the transmission amplitude (beware not to confuse it with the
time!). The two functions, Eqs. (5.87) and (5.88), solve Eq. (5.79) for any
|n| > 0, provided that the dispersion relation, ω(k) = ω0 | sin(ka/2)|, inferred
from Eq. (5.6) is fulfilled. That same equation, however, requires that mω 2 t =
K{2t − t exp[ika] − (exp[−ika] + r exp[ika])}. Using also the relation u0 = 1 + r = t
(that follows from the equations for n = ±1) one obtains
i sin(ka)
t = ika . (5.89)
e − 1 + 2(m/M ) sin2 (ka/2)
For equal masses, m = M , the transmission amplitude is 1 (check!) since then the
lattice is homogeneous and the wave is transmitted perfectly. In contrast, when
the mass of the foreign atom differs from that of all other atoms, the transmission
amplitude
√ is a complex number, whose absolute value is smaller than 1. Denoting
t=√ T exp[iα], the function of the wave on the right side of the impurity is un =
A T exp[i(kna + α)] exp[−iωt]. The√phase of the wave is shifted by α, and its
amplitude is decreased by the factor T . The dependence of T , which is termed
the transmission coefficient, on the masses’ ratio is displayed in Fig. 5.21, for
several values of ka. All curves possess a maximum, T = 1, reached when the masses
are equal; the transmission coefficient decreases as the wave number increases.
Generalizations. Three comments are in order. First, at higher dimensions,
a plane wave, exp[ik · Ri − iωt], impinging on a foreign atom, is scattered at all
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 360

360 Introduction to Solid-State Physics

Fig. 5.21: The transmission coefficient, T , as a function of the masses’ ratio m/M .
It decreases as the original wave-number increases from ka = 0.1π (upper curve) to
ka = 0.9π (lowest curve), in steps of 0.2π.

directions, with an amplitude that depends on the direction. Second, the classical
treatment described above can be carried out quantum mechanically, e.g., see the
arguments presented in Sec. 5.3, that lead to Eq. (5.41). Third, foreign atoms
scatter other particles as well, predominantly electrons (see Chapter 6).

5.7 Lattice momentum and inelastic scattering of phonons

The lattice momentum of phonons. As mentioned in the context of Eq. (5.48),

the description of vibrational modes in terms of waves can be translated into the
“language” of particles: a vibrational mode of frequency ω, that possesses an average
energy ω(1/2 + n), is equivalent to a motion of n phonons, each with an energy
ω. In analogy with the de Broglie duality of particles and waves in quantum
mechanics, each phonon may be also assigned a momentum, k. In the previous
section it is found that when a wave is scattered off an impurity elastically in one
dimension, the frequency does not change but the momentum of the reflected wave
is anti parallel to that of the impinging wave, and is −k. The momentum lost
by the scattered phonon is absorbed by the lattice. As the lattice is very large,
the momentum absorbed from a single phonon is negligible (it is not measurable).
In higher dimensions, the phonon can be scattered into various directions, and its
momentum is changed accordingly. In the harmonic approximation, though, the
energy is conserved.
In contrast with electrons or photons, whose momentum is uniquely defined by
the de Broglie principle, p = k, the momentum of the phonon is different. As
seen, it is enough to specify the wave vector of a vibration in the first Brillouin
zone, as the properties of the wave are not changed upon displacing that vector by
a reciprocal-lattice vector, k ⇒ k + G. It follows that phonons whose momenta
differ by G describe exactly the same physical phenomenon. For this reason, the
phonon’s momentum is termed lattice momentum, and is defined within the
first Brillouin zone. In case this momentum, following a certain physical process
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 361

Lattice vibrations 361

(examples are given below) is augmented to lie beyond the first zone, one uses the
translation k ⇒ k + G to bring it back into the first Brillouin zone. The physical
outcome is not modified.
Scattering of radiation by phonons. The concept of lattice momentum is
elucidated in the description of scattering of radiation off solids. This scattering
may well be inelastic, i.e., the energy of the scattered particles differs from that of
the impinging ones. To account for the inelasticity the real wave function is written
in the form cos(k · r − Ωt) = {exp[i(k · r − Ωt)] + exp[−i(k · r − Ωt)]}/2, where
Ω = Ω(k) is the frequency derived from the dispersion relation of the radiation, see
Sec. 3.1. Assuming for simplicity point-like scatterers, then the generalization of
Eq. (3.7) yields that the function of the scattered wave is
  0 0
A∝ eiq·R−iΩt + cc = eiq·[R +u(R ,t)] e−iΩt + cc , (5.90)
R R0

where cc denotes the complex conjugate of the other term and q = k − k is

the “momentum” transfer between the impinging particles (photons, neutrons, or
electrons) and the scattered ones (up to the multiplicative factor ).
Momentum and energy conservation. On the right hand-side of Eq. (5.90)
appears the location of the nucleus, R, expressed in terms of its equilibrium site
on the lattice, R0 , and its displacement, u(R0 ) (the index i is omitted here for
brevity). As the displacement is small, the right hand-side of the equation may be
expanded as a power series in u,
 0
A∝ eiq·R e−iΩt [1 + iq · u(R0 ) + . . .] + cc . (5.91)
R0

When the lattice vibrations are ignored, this expression reproduces the result of
Chapter 3, that gives rise to Bragg peaks whenever q = G. Then the frequency of
the outgoing wave is also Ω, and the scattering is elastic, i.e., it conserves the energy
of the impinging particle. As in Eq. (5.44), the oscillation’ amplitude of each atom
in the crystal is a sum over the contributions of all eigenmodes. Classically, the
amplitude can be written in the form
 0
u(R0 , t) = N −1/2  (K)eiK·R −iω(K)t ,
u (5.92)
K

where the sum contains only those discrete wave vectors which are allowed by the
boundary conditions, K = ( 1 /N1 )b1 + ( 2 /N2 )b2 + ( 3 /N3 )b3 , and where the
explicit time dependence is kept, with the frequency that obeys the dispersion law,
ω = ω(K). Inserting Eq. (5.92) in each of the summands in Eq. (5.91) yields
 0  0
e±i(q·R −Ωt) [1 ± iN −1/2 q·u (K)eiK·R −iω(K)t + . . .] . (5.93)
R0 K

The second factor in the square brackets contains the sum R0 exp[i(K ± q) · R0 ].
As seen in the discussion following Eq. (3.11), this sum equals N when K ± q = G,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 362

362 Introduction to Solid-State Physics

for any reciprocal-lattice vector G. Using the original wave numbers of the radiation
thus results in an appreciable scattering amplitude only when
k − k = q = ±(K − G) . (5.94)
The left hand-side is the momentum transfer of the scattered particles, while the
right hand-side comprises the lattice momentum of the relevant phonon. It follows
that the radiation “emits” or “absorbs” a phonon (according to the sign on the right
hand-side), and changes its momentum such that the new momentum is conserved,
according to Eq. (5.94). The time dependence of the outgoing wave is exp[−i(ω ±
Ω)t]; one may thus identify the frequency of the outgoing wave as Ω = Ω ± ω; hence
Ω − Ω = ±ω(K) . (5.95)
This relation expresses the energy conservation of the process. The upper sign
corresponds to a “creation” of a new phonon in the system, while the radiation loses
energy and momentum; the lower sign represents the “annihilation” of an existing
phonon, with the radiation absorbing its energy and momentum. Recall that the
dispersion relation of the phonons is periodic in momentum space, ω(K − G) =
ω(K), and thus the right hand-sides of Eqs. (5.94) and (5.95) represent the same
vibrational mode. A quantum-mechanical treatment of this scattering event gives
rise to Eqs. (5.94) and (5.95) as well; in such an analysis the Hamiltonian contains
the interaction between the radiation and the lattice vibrations.
Bragg peaks of inelastic scattering. Once Eqs. (5.94) and (5.95) are used
in Eq. (5.93), it is found that each Bragg peak, that results from elastic scattering,
is accompanied by additional peaks, for which the momentum and the energy of
the scattered particles obey the conservation laws of the energy and the lattice
momentum. These new Bragg peaks represent inelastic scattering events. To
detect the energy transfers in experiments, the energies of the scattered particles
have to be of the same order of magnitude as those of the phonons. As the lattice
constants are of the order of a few Å’s, the relevant wave numbers are of the order
of Å−1 . Such wave numbers correspond to photons’ energies of the order of a few
thousands electron-volts, or neutrons’ energies of the order of a few tens of milli-
electron volts. The energies of the phonons are comparable to those of the neutrons,
and therefore neutron radiation is the most suitable one for the research of phonons
by inelastic scattering. Indeed, this radiation has been exploited in the experiments
discussed in connection with Figs. 5.18 and 5.19.

Problem 5.25.
Neutrons of mass mn and momentum k impinge on a single-atom one-dimensional
lattice (along the direction parallel to the lattice). The lattice consists of atoms of
mass M coupled together by nearest-neighbor springs of a constant D. Find the
possible values of the momentum and frequency of the scattered neutrons.

Inelastic scattering of light. As opposed to neutrons, the typical photon

energy of the visible light is of the order of a few electron volts; the corresponding
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 363

Lattice vibrations 363

wave length is of the order of a few thousands Å. It follows that inelastic scat-
tering of light off lattice vibrations might be realized only for wave vectors much
shorter than the lengths of the reciprocal-lattice vectors, i.e., vectors at the vicinity
of the center of the Brillouin zone. Inelastic scattering of light off phonons is called
Brillouin scattering; when the scattering phonons are optical, it is referred to as
Raman scattering. Equations (5.94) and (5.95) prevail for this type of scattering
as well. The emission of a phonon is termed Stokes scattering, and the absorption
process is the anti-Stokes scattering. As c|k| = Ω  ω(K), then approximately
|k| = |k| , and consequently |K| ≈ 2k sin θ (see Fig. 3.7).

5.8 Phenomena related to anharmonic effects

The importance of anharmonic terms in the potential energy. The discus-

sion above is focused on the outcome of the harmonic approximation, Eqs. (5.1)
or (5.20), in which the potential between two atoms in the crystal is replaced by
a paraboloid centered around around its minimal value. Figure 5.22 displays the
Lennard-Johns potential, already discussed in previous chapters (see also Fig. 1.1),
together with the parabola that passes through its minimum, such that the second
derivative there is the same for both curves. Obviously the parabola is a faithful
description of the potential close to the minimum, but it deviates from it farther
away. At low temperatures the harmonic approximation is quite reliable, as only
small deviations from the minimum are thence important. In the quantum lan-
guage, the properties calculated within this approximation are determined mainly
by the lower energy levels of each vibrational mode, i.e., by a small number of
phonons. However, the harmonic potential misses several important aspects. First,
the actual potential is steeper than the parabola to the left of the minimum, and
smoother than it to the right. As a result, it is harder to compress the material
than to stretch it. In contrary, the harmonic potential is symmetric and does not
reproduce this feature.
Second, in the harmonic approximation the average distances between the atoms
remain unchanged at all temperatures; the phenomenon of thermal expansion
does not exist, whereas experiments show that crystals expand when heated. The
volume thermal expansion coefficient of a material in d dimensions is defined
by
1  ∂V  d  ∂L 
αV = = dαL = , (5.96)
V ∂T p L ∂T p
where αL is the longitudinal thermal expansion coefficient (the factor d results from
the relation V ∝ Ld , from which it follows that ∂V /V = d∂L/L). It is found below
that this coefficient vanishes in the harmonic approximation; it becomes finite only
when anharmonic terms are included.
Third, in the harmonic approximation the vibrational waves extend over the
entire lattice. In other words, in the absence of foreign atoms and other defects the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 364

364 Introduction to Solid-State Physics

Fig. 5.22: A comparison between the Lennard-Johns potential, Eq. (4.33), and the
harmonic approximation at the vicinity of its minimum. Distances are measured in
units of σ and the energies in units of .

phonons move freely over the entire system. Phonons carry heat and contribute
significantly to the heat conduction in solids. In experiments, the heat conduc-
tance, κ, is defined by the Fourier law, which relates the heat current, jU , to a
temperature gradient imposed on the material, ∇T ,

jU = −κ∇T . (5.97)

The heat current is the amount of thermal energy that passes through a unit area
per unit time; it is induced by the temperature gradient. (Note the analogy with
the electric current in a metal, which is induced by an electric field, that is given
by the gradient of the electrical potential.) When the phonons passing through
the material are not hindered by any perturbation, there is no resistance to the
heat conduction, that is, the heat conductance is infinite. However, as discussed,
phonons are scattered by foreign atoms. The main cause for the finite value of the
heat conductance in solids, though, are collisions in-between phonons. In the
harmonic approximation, the Hamiltonian of the lattice vibrations is a sum over
the separate Hamiltonians of each vibrational mode, and there are no interactions
among the various modes; the modes do not evolve in time. The anharmonic terms,
which exist beyond the harmonic approximation, give rise to interactions among
the phonons; these cause collisions that lead in turn to a reduction in the heat
conductance. As mentioned, other factors, e.g., the collisions of the phonons with
impurities or with the sample’s walls, or with other defects in the lattice, give rise
to a further decrease in the heat conductance.
Thermal expansion. To understand the notion of thermal expansion it is
useful to consider the simple anharmonic potential U (x) = Cx2 /2−g3 x3 +g4 x4 +. . . .
The sign of the third-order term is chosen to make the leading-order correction to
the parabola negative on the right of the maximum, and positive on its left, in
accordance with Fig. 5.22. In the classical approximation, the thermal average of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 365

Lattice vibrations 365

the particle’s location (relative to its equilibrium site) is

2
dxxe−βU (x) dxxe−βCx /2 (1 + βg3 x3 − βg4 x4 + . . .) 3g
x = ≈ 2 /2 ≈ 23 kB T .
dxe−βU (x) dxe −βCx (1 + βg3 x − βg4 x + . . .)
3 4 C
(5.98)
In the right-hand parts of the equation, both the numerator and the denomina-
tor are expanded in the anharmonic terms; the end result consists of the leading
contribution (check!). As seen, the average position vanishes in the harmonic ap-
proximation, but increases linearly with the temperature once the first anharmonic
correction is accounted for.

Problem 5.26.
At zero temperature the energy of the crystal can be written in the form F0 =
Utot (V ) ≈ Utot (V ) + [B/(2V )](V − V )2 , where V is the volume attained at the
minimal value of the energy, and where B = V (∂ 2 U/∂V 2 ) is the coefficient of the
volume expansion, see the discussion around Eq. (4.10). At finite temperatures the
energy is augmented by the free energy of the phonons.
a. Show that the total free energy is
B 
F0 (T, V ) = Utot (V ) + (V − V )2 + kB T log[2sinh(βωα /2)] ,
2V α

where the sum encompasses all vibrational modes.

b. The anharmonic effects can be accounted for by allowing the harmonic frequen-
cies to depend on the volume. Assuming that the crystal is not under pressure, i.e.,
p = −(∂F0 /∂V )T = 0, show that the change of the volume of the crystal caused by
the phonons is
1  ωα βωα  ∂ log ωα 
V =V + coth − .
B α 2 2 ∂ log V

c. To a certain extent, ∂ ln ωα /∂lnV = −γ is the same for all modes (i.e., it

is independent of α). This quantity is termed the Grüneisen constant. Show
that then V = V + γUph /B, where Uph = Eph is the total average energy of the
phonons.
d. Show that in this case αV = [γ/(BV )]CV , where CV = (∂Eph /∂T )V is the
phonon specific heat.

Heat conductance. Heat conductance is mainly determined by the collisions

among the phonons. These collisions can be treated in much the same way as done
for those between the phonons and the photons, culminating in Eq. (5.94),
k1 − k2 = ±(k3 − G) , (5.99)
where k1 is the momentum of the phonon scattered into a state of momentum k2 ,
once another phonon of momentum k3 is emitted or absorbed. Another way to
present this relation is k1 ± k3 = k2 − G. The left hand-side is the total lattice
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 366

366 Introduction to Solid-State Physics

momentum of the incoming phonons, and the right one contains the lattice momen-
tum of outgoing phonon, up to a reciprocal-lattice vector. When the three momenta
are small, then G = 0, the momentum of the outgoing phonon equals the sum of
the momenta of the two incoming ones; i.e., the total momentum of all phonons is
not changed. On the other hand, when the sum of the two incoming momenta is
a vector ending outside of the first Brillouin zone, the outgoing momentum is that
sum minus a reciprocal-lattice vector. The outgoing phonon may thus propagate
in a direction opposite to the original direction, and hence diminish the original
phonons’ stream. Such a process, that gives rise to an inverse momentum (relative
to the incoming one) is termed Umklapp(the German word for inverse). Another
example for such a process appears in Sec. 5.7, in connection with the scattering of
a phonon off an impurity.
The kinetic theory. The simplest treatment of heat conductance is carried
out in the context of the kinetic theory of a phonon gas, in which the phonons
are viewed as particles moving in the material. Each phonon moves with a constant
velocity v during a typical time interval τ , traversing a distance termed the “mean
free-path”, until it collides with a defect (or the vessel wall, or an electron, or
another phonon). Imagine that the two opposite faces of the container are held
at two different temperatures, such that the temperature decreases towards the
positive x̂−axis. Assume further that the thermal-equilibrium energy of the phonon
is determined by the local temperature, T (x), i.e., E(x) = Eeq [T (x)]. In the absence
of a net current, half of the phonons moves to the right and the other half moves
towards the left. The net number of phonons passing from left to right through a
unit area normal to the x̂−axis per unit time is nvx /2. (The number of phonons
per unit volume is n, and it is assumed that vx > 0). The phonons that impinge on
this unit area from the left traveled a distance vx τ and therefore carry the energy
Eeq [T (x − vx τ )]. Similarly, the phonons that come from the right carry the energy
Eeq [T (x + vx τ )]. Thus, the amount of energy passing through the unit area per
unit time is
n n ∂Eeq ∂T
jU,x = vx {Eeq [T (x − vx τ )] − Eeq [T (x + vx τ )]} ≈ − vx 2v τ , (5.100)
2 2 ∂T ∂x x
where both the energy and the temperature are expanded to leading order in the
respective variable’s difference. Recalling that the specific heat per unit volume is
CV = n(∂Eeq /∂T ), one retrieves Eq. (5.97), with the heat conductance
1 2
v CV τ .
κ = CV vx2 τ = (5.101)
3
The average is carried out over all velocity’s directions (as there is no preference
ˆ
to the x−direction). When the phonons are acoustic, their velocity is the sound
velocity, which is independent of the temperature. At temperatures higher than
the Debye temperature the specific heat is close to the its value derived from the
Dulong-Petit law, and thus it is also independent of the temperature. It remains to
estimate the time in-between collisions, τ . It turns out that the most detrimental
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 367

Lattice vibrations 367

processes for the heat conduction are the Umklapp ones, in which two phonons are
colliding with one another. The probability of a certain phonon to hit another one
is proportional to the density of phonons in the crystal. From Eq. (5.54), one finds
that at high temperatures nα  = 1/(exp[βωα ] − 1) ≈ kB T /(ωα ); the probability
is proportional to the temperature and consequently the mean free-time in-between
collisions is inversely proportional to the temperature, τ ∝ 1/T . In this limit
κ ∝ 1/T . Experiments find a somewhat higher power of 1/T , probably because of
collisions among more than two phonons. At low temperatures most of the phonons
have small momenta; collisions among them leave the outgoing momentum within
the first Brillouin zone, and consequently do not affect the heat conductance. To
obtain Umklapp processes the two colliding phonons should have momenta larger
than half of the reciprocal-lattice vector (in suitable units), and the probability to
find such phonons at low temperatures is exponentially small, nα  ≈ exp[−βωα ],
where ωα is of the order of the Debye frequency; the mean free-path may then
be even longer than the sample itself. It follows that the heat conductance at
low temperatures is dominated by other processes, as collisions with the sample’s
walls or collisions with impurities. The dependence of the heat conductance on the
temperature is then determined by the temperature dependence of the specific heat,
CV ∝ T 3 in three dimensions. The outcome of these observations is that, roughly
speaking, the heat conductance increases like T 3 at low temperatures, reaches a
maximum, and then decays as T −1 at high temperatures.

5.9 Magnons

Magnetic excitations. In addition to phonons, there are other excitations in the

crystal, some of them resembling phonons, for instance, magnetic excitations
termed magnons. Recall that the magnetic behavior of quite a number of materials
is represented by the Heisenberg model Eq. (4.41): ĤM = − nm Jnm Sn · Sm ,
where μn = −gμB Sn is the magnetic moment located at the lattice site n, Jnm
is the exchange coefficient (in units of energy), and the sum encompasses all pairs
of sites nm. At any given time, the spin (i.e., Sn ) at site n is subjected to a
magnetic field, induced by the instantaneous magnetic moments of all its neighbors,
Bn = m Jnm Sm (as the “spins” are dimensionless, the magnetic field is in units
of energy). Consider the classical equations of motions that emerge from this
description. A magnetic field B, applied on a classical magnetic moment μ, gives
rise to a torque, such that the equation of motion is μ̇ = γμ × B, where γ =
ge/(2mc) = μB /. In the present case,

μ̇n = μn × J nm μm , (5.102)
m

where J nm = γJnm /(gμB )2 = Jnm /(gμB ) (see problem 5.27). Equation (5.102)
is a non-linear equation for the variables μn . However, at low temperatures the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 368

368 Introduction to Solid-State Physics

magnetic moments are ordered, that is, they mainly point along a certain direction,
and it is plausible to consider only small deviations from this ordered state [in
analogy with the lattice vibrations, where only small deviations (off the equilibrium
sites) in the locations of the atoms are considered]. This situation is demonstrated
by two examples.
Ferromagnets. In the fully-ordered ferromagnetic state, all magnetic moments
are equal in magnitude and are parallel to each other, for instance along the ẑ−axis.
Thus, in the ground state μn = μẑ. Deviations from the ground state can be
considered by assuming that the instantaneous components μnx and μny are small.
Up to first order in the x and y components, the length of the z−component is
unchanged (its variation is quadratic in the lengths of the other components), μ2nz =
μ2 − μ2nx − μ2ny ≈ μ2 . With these assumptions, the equation of motion (5.102) yields
dμnx  
= J nm (μny μmz − μnz μmy ) = μ J nm (μny − μmy ) ,
dt m m
dμny  
= J nm (μnz μmx − μnx μmz ) = μ J nm (μmx − μnx ) . (5.103)
dt m m
These are two linear equations in two variables, whose treatment is similar to the
one carried out above, e.g., Eqs. (5.12), (5.23), or (5.32). Inserting a wavy solution,
(0)
μnx,y = μx,y exp[i(k · Rn − ωt)], leads to
−iωμy = −μ[J(0) − J(k)]μ
 x , −iωμx = μ[J(0) − J(k)]μ
 y , (5.104)
where 

J(k) = J nm e−ik·Rnm
0
(5.105)
m
is the Fourier transform of the exchange coefficient, and R0nm = R0n − R0m . For
instance, for a nearest-neighbor interaction in one dimension J(k)  = J(exp[−ika] +
exp[ika]) = 2J cos(ka). Inserting the second of Eqs. (5.104) into the first one yields
 − J(k)]
ω 2 μy = μ2 [J(0)  2
μy ; hence this is a valid solution provided that the disper-
2  
sion relation ω = μ [J(0) − J(k)]
2 2
is obeyed. The similarity with the dispersion
relation of lattice vibrations is quite clear, for example, Eq. (5.5). Nonetheless, the
dispersion relation is different,
 − J(k)]
ω = μ[J(0)  , (5.106)
where the positive value of the frequency is chosen. For instance, in one dimension
with solely nearest-neighbor interactions
 − J(k)]
ω = μ[J(0)  = μJ[2 − 2 cos(ka)] = 4μJ sin2 (ka/2) . (5.107)
The right hand-side may be expanded in k at small frequencies, yielding a quadratic
dispersion relation, ω ≈ μJ(ka)2 . Note the disparity with the linear dispersion in
the acoustic branch of lattice vibrations.
Problem 5.27.
a. Show that the solutions of the equation μ̇ = γμ × B correspond to a precession
of the vector μ around the vector B, with the Larmor frequency ωL = γB.
b. Derive Eq. (5.102) from the equation in part (a).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 369

Lattice vibrations 369

Spin waves. The waves derived in the calculation above are termed “spin
waves”. Indeed, upon inserting Eq. (5.106) in Eqs. (5.104), one finds that μy =
−iμx . This implies that there is a phase difference between the oscillations of the
two spin components: the component of the magnetic moment in the XY plane
rotates around the ẑ−axis together with the propagation along the direction of the
wave vector k. Figure 5.23 depicts the spins (by the arrows) of a ferromagnet whose
moment is perpendicular to the direction of the (one-dimensional) lattice, and hence
is also normal to the direction of k itself. Has the magnetic moment been parallel to
the lattice direction, it would have made a screw-like rotation around that direction.

Fig. 5.23: Spin waves of a one-dimensional ferromagnet with a ferromagnetic mo-

ment (the vertical arrow) normal to the crystal, for a wave that propagates along
the lattice (the horizontal direction in the figure). The component of the magnetic
moment perpendicular to the ferromagnetic one rotates around the ferromagnetic
moment along with the propagation of the wave.

Problem 5.28.
An alternative method to solve the equations of motion for spin waves of a ferro-
magnet is to construct equations for the functions μn± = μnx ± iμny , and to solve
them. Show that this method produces the same solutions as those obtained above.

Fourier transforms. An alternative way to obtain the spin-wave solutions is

to Fourier transform the magnetic moments,
 0
 (k) =
μ eik·Rn μn , (5.108)
n

and then to insert the transforms into the equations of motion (5.103), which become
d
μx  − J(k)]

= μ[J(0) μy ,
dt
d
μy
 − J(k)]
= −μ[J(0)  μx . (5.109)
dt
Differentiating the first equation with respect to the time t and then inserting the
result into the second gives an equation for μx ,
x
d2 μ  − J(k)]

= −μ2 [J(0) 2
x .
μ (5.110)
dt2
The same equation is obtained for μ y . Equation (5.110) is identical to the one
 − J(k)];
that describes the harmonic oscillator, with the frequency ω = μ[J(0) 
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 370

370 Introduction to Solid-State Physics

thus the previous solutions are reproduced. This mapping upon the harmonic os-
cillator also allows for the construction of a quantum Hamiltonian for the vari-
able μ x . The energy levels of each such “oscillator” are of the form ω(n + 1/2),
and their statistical mechanics is the same as that of phonons. In the present
context the “particles” that represent spin waves are termed “magnons”. The
only difference between (ferromagnetic) magnons and phonons is in the disper-
sion relation. Choosing in addition boundary conditions identical to the ones ap-
plied for phonons, i.e., periodic boundary conditions, yields discrete wave numbers,
k1 ,2 ,3 = ( 1 /N1 )b1 + ( 2 /N2 )b2 + ( 3 /N3 )b3 , where −Nm /2 ≤ m < Nm /2.
The calculation of the specific heat of the magnons is very similar to that of the
phonons, but the different dispersion relation leads to a different dependence on the
temperature, see problem 5.29.

Problem 5.29.
As for phonons, one may also describe a Debye model for magnons, by extending
the dispersion relation of long waves up to an appropriate Debye frequency.
a. Find the dispersion relation of a cubic ferromagnet with nearest-neighbor in-
teractions, and show that in the long-wave limit it can be written as ω ≈ Δ|k|2 .
Determine the value of the coefficient Δ.
b. Find the leading-order dependence of the magnons’ specific heat on the temper-
ature, at low and at high temperatures.

Antiferromagnets. For simplicity, the equations of motion for the spin waves
of an antiferromagnet are derived for a simple configuration: a one-dimensional
lattice, in which the exchange interactions are solely between nearest neighbors.
The ground state of this system is depicted in part C of Fig. 2.42. The unit cell
(1) (2)
contains two magnetic moments, e.g., μ1 on the left and μ1 on the right; at
(1) (2)
equilibrium, μn = μẑ and μn = −μẑ. The equations of motion are
(1)
dμn (2)
= μ(1)n × J(μn + μn−1 ) ,
(2)
dt
(2)
dμn (1)
= μ(2)n × J(μn + μn+1 )
(1)
(5.111)
dt
(check!). As again only linear-order terms in the deviations from the equilibrium
configuration are accounted for, it is assumed that the z−component of the magnetic
moment does not vary in time. Then, using the explicit forms of the vectors
μ(1) (1) (1)
n = (μnx , μny , μ) ,

μ(2) (2) (2)

n = (μnx , μny , μ) ,

one obtains four linear equations for the components of the magnetic moments in
the XY plane (see problem 5.30). The solution of these equations yields spin waves
with the dispersion relation
ω = 2μJ| sin(ka/2)| , (5.112)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 371

Lattice vibrations 371

where a is the lattice constant.

Problem 5.30.
Write down the explicit equations of motion for spin waves that propagate in a
one-dimensional antiferromagnet with nearest-neighbor exchange interactions, and
prove Eq. (5.112).

Other excitations. In addition to phonons and magnons, there are several

other types of excitations in condensed-matter systems. These are described clas-
sically by waves connected with the oscillations of a certain physical property, and
quantum-mechanically are characterized by the appearance of “particles” moving
in the material. Several such excitations are related to the motion of electrons in
the crystal, discussed in Chapter 6. Examples are plasmons, polarons, and excitons.
The plasmon refers to a collective excitation of the electron gas in a metal, relative
to the lattice of ions. The origin of the polaron is the coupling between the electron
and the vibrations of the crystal in which it resides. It describes a motion of an
electron accompanied by a distortion of the surrounding ions’ locations. The exci-
ton results when an electron from the valence band is excited into the conduction
band, leaving behind a “hole”; the electron and the hole remain connected, like the
positive and negative charges in the hydrogen atom, and therefore move together.
Detailed descriptions of these excitations are beyond the scope of the book.

5.10 Answers for the problems in the text

Answer 5.1.
a. Extending Eq. (5.3), the equations of motion for harmonic interactions between
neighbors located at a distance ma from one another, are

M ün = − DL,m (2un − un−m − un+m ) .
m

Considerations similar to those presented at the beginning of Sec. 5.1 lead to a

solution of the form un = A0 exp[i(kna − ωt)]. This is indeed a solution provided
that
 
M ω2 = DL,m (2 − e−ikma − eikma ) = 4 DL,m sin2 (mka/2) . (5.113)
m m

This dispersion relation has the periodicity given in Eq. (5.7) and hence it suffices
to investigate it in the first Brillouin zone. Due to the symmetry of k with respect
to inversion, it is enough to consider the range 0 ≤ k ≤ π/a. The group velocity is
given by the derivative of the dispersion law, vg = [a/(M ω)] m DL,m m sin(mka).
It vanishes at the edges of the Brillouinzone, ka = π. The expansion around k = 0
yields the sound velocity, c = vg = a 2
m m DL,m /M . This is valid as long as
mka  1 for each m in the sum.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 372

372 Introduction to Solid-State Physics

b. The interaction is [see Eq. (4.33)]

 σ 12  σ 6
U (R) = 4 − ,
R R
and hence, using DL,m = ∂ 2 U (R)/∂R2 |R=ma , one finds
96   σ 14  σ 8
M ω2 = 2 26 −7 sin2 (mka/2) .
σ m ma ma

Once the summation is carried out, the value of the lattice constant, (a/σ)6 =
2A12 /A6 , is introduced [see discussion and notations after Eq. (4.34)]. In one
∞
dimension An = m=1 1/mn = ζ(n), and the values of the zeta function, ζ(n),
can be found in mathematical codes or treatises of mathematical functions. In
particular, ζ(6) = π 6 /945 ≈ 1.0173 and ζ(12) = 691π 12 /638512875 ≈ 1.000246.
In the long wave-length limit sin x can be approximated by x (though, strictly
speaking, this approximation is valid only for kam  1 and therefore fails at long
distances). The result is
  σ 12  σ 6  A2 
M ω 2 = 24 26A12 − 7A6 k 2 = 72 6
k 2 ≈ 74.5k 2 .
a a A12

This leads to an acoustic dependence, with a sound velocity c ≈ 8.63 /M . Inter-
estingly enough, a numerical computation of the sum (discarding the approximation
of sin x) gives rise to a similar behavior.
c. In this case
ω 2 = ω02 sin2 (ka/2)[1 + (4DL,2 /DL,1 ) cos2 (ka/2)] ,

where ω0 = 2 DL,1 /M . Figure 5.24 displays the dispersion law. One observes
that at the edge of the Brillouin zone the second term in the brackets vanishes
and there appears a local minimum, ω = ω0 . Between the origin and this point
there is a maximum, ωmax = ω0 (DL,1 + 2DL,2 )/(4 DL,1 DL,2 ), at sin2 (kmax a/2) =
(DL,1 + 4DL,2 )/(8DL,2 ). The group velocity vanishes at this maximum (like at
any extremal point), leading to the appearance of standing waves there. For the
parameters considered here, the maximum appears at kmax a = 2π/3. As seen, there
is a single wavy solution (for each sign of the wave number) for ω < ω0 . In the
range ω0 ≤ ω ≤ ωmax there are two wavy solutions, and for ω ≥ ωmax there are
no wavy solutions at all. The number of solutions is doubled when one takes into
account the negative k’s.

Answer 5.2.
The vectors between the nth atom and its two neighbors are Rn,n±1 ; the angle in-
between the bonds is given by cos γn = Rn,n−1 ·Rn,n+1 /(Rn,n−1 Rn,n+1 ). Expanding
Rn,n±1 = ±ax̂ + un±1 − un in a Taylor series in the displacements, results in
cos γn +1 = (2vn −vn−1 −vn+1 )2 /(2a2 )+. . ., where vn is the transverse component
of the vector Rn . Assuming that the energy is proportional to this expression, i.e.,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 373

Lattice vibrations 373

Fig. 5.24

U =C n (1+cos γn ), the equations of motion obeyed by the transverse components

are
C
M v̈n = − [2(2vn − vn+1 − vn−1 )
a2
+ (vn + vn−2 − 2vn−1 ) + (vn + vn+2 − 2vn+1 )] .
The dispersion relation for vn = v0 exp[i(kan − ωT t)] is
C
M ωT2 = 2 [6 + 2 cos(2ka) − 8 cos(ka)] ≈ Ca2 k 4 + . . . .
a

Answer 5.3.
One may write Eq. (5.13) in the form
    
A 2/M −(e−ika + 1)/M A
ω2 =D .
B −(1 + e )/m
ika
2/m B
The vector whose components are A and B is an eigenvector of the matrix on the
right hand-side, with the eigenvalue ω 2 . Denoting this matrix by Ω, the eigenvalues
are obtained by equating the determinant of (Ω−Iω)ˆ to zero, where Iˆ is the 2×2 unit
matrix. This requirement is identical to Eq. (5.14). Inserting one of the eigenvalues
into the equation yields the ratio between the components of the corresponding
eigenvector, like in Eq. (5.18).

Answer 5.4.
The equations of motion are
M ün = −D(un − vn−1 ) − D(un − vn ) − eE0 ei(kna−ωt) ,
mv̈n = −D(vn − un ) − D(vn − un+1 ) + eE0 ei(kna−ωt) .
These describe a driven harmonic motion. One seeks for a particular solution of the
equations: assuming that the atoms “follow” the oscillations of the electric field,
one uses the forms un = A exp[ikna − iωt] and vn = B exp[ikna − iωt], with the
same wave vector and the same frequency as those of the electric field. Then
M ω 2 A = D[2A − (1 + e−ika )B] + eE0 ,
mω 2 B = D[2B − (1 + eika )A] − eE0 ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 374

374 Introduction to Solid-State Physics

and consequently
1
A= [mω 2 − 2D + D(1 + e−ika )]eE0 ,
Det
1
B=− [M ω 2 − 2D + D(1 + eika )]eE0 ,
Det
Det = (M ω 2 − 2D)(mω 2 − 2D) − 4D2 cos2 (ka/2) .

Fig. 5.25

The induced dipole moment in a unit cell is the charge times the distance between
the negative and the positive ions, i.e.,
p(n, t) = e(vn − un ) = −e2 E(n, t)[(m + M )ω 2 − 4D sin2 (ka/2)]/Det.
Because the speed of light is about 105 larger than the velocities of the acoustic and
optical waves in the crystal, one may assume that ka  1. Using strictly zero ka
yields an electric field and an induced dipole moment which do not depend on the
coordinate. The induced dipole moment per unit volume is then P = αE, with the
polarizability α = ρe2 /[μ(ω02 − ω 2 )], where ρ is the density
 (the number of unit cells
per unit volume), μ is the reduced mass, and ω0 = 2D/μ is the optical frequency
in the k = 0 limit, Eq. (5.17). The polarizability is related to the dielectric function
ε by ε = 1 + 4πα and hence ε = 1 + 4πρe2 /[μ(ω02 − ω 2 )].
The dielectric function is depicted in Fig. 5.25. The speed of light in the
√ √
material is c = clight / ε, and the wave number is k = ω/c = ω ε/clight . The
dielectric function is negative over a range of frequencies and there the wave number
is imaginary, k = iκ. Then the light waves decay with the distance as exp[−κna],
and the material is opaque for radiation of frequencies in this range.

Answer 5.5.
The “new” equations of motion are
M ün = −D(un − vn−1 ) − D(un − vn ) − D2 (un − un−1 ) − D2 (un − un+1 ) ,
mv̈n = −D(vn − un ) − D(vn − un+1 ) − D2 (vn − vn−1 ) − D2 (vn − vn+1 ) .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 375

Lattice vibrations 375

Inserting the wave functions un = A exp[i(kna − ωt)] and vn = B exp[i(kna − ωt)]

yields
M ω 2 A = D[2A − (e−ika + 1)B] − 2D2 [1 − cos(ka)]A ,
mω 2 B = D[2B − (eika + 1)A] − 2D2 [1 − cos(ka)]B ,
with the determinant
Det = [M ω 2 − 2D − 4D2 sin2 (ka/2)][mω 2 − 2D − 4D2 sin2 (ka/2)] − 4D2 cos2 (ka/2) .

Fig. 5.26

Equating the determinant to zero gives two solutions,

 1 1  4D2 
ω2 = D + 1+ sin2 (ka/2)
M m D
 1 1 2  4D2 2
±D + 1+ sin2 (ka/2)
M m D
2
4 sin (ka/2)  4D2 4D2  1/2
− 1+ + 22 sin2 (ka/2) .
Mm D D
These are shown in Fig. 5.26 for two values of D2 , with the same notations as in Fig.
5.4. Note in particular the qualitative modifications in the optical branch. When
D2 = 0, one retrieves Fig. 5.4. As the forces in-between next-nearest neighbors
increase, this branch “flattens”, and the maximum at k = 0 turns into a minimum
which decreases as D2 is increased. In the D = 0 limit, the only forces are those
in-between identical atoms. The equations of motion then coincide with Eqs. (5.5),
and give rise to the dispersion relations shown in Fig. 5.2 – but with different
masses for each sublattice, and hence with different sound velocities and different
maximal frequencies.

Answer 5.6.
The unit cell still contains two atoms, but their masses are identical. Denoting the
spring constant between the atoms in the cell (i.e., the single bond) by D1 , and
the one in-between nearest neighbors belonging to neighboring cells (i.e., the triple
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 376

376 Introduction to Solid-State Physics

bond) by D2 (these are the springs to the left and to the right of the solid circle in
Fig. 5.3), the equations of motion are
M v̈n = −D1 (vn − un ) − D2 (vn − un+1 ) ,
M ün = −D2 (un − vn−1 ) − D1 (un − vn ) .
One then finds the two equations
M ω 2 B = (D1 + D2 )B − (D1 + D2 eika )A ,
M ω 2 A = (D1 + D2 )A − (D1 + D2 e−ika )B ,
which have a non trivial solution when

M ω 2 = D1 + D2 ± D12 + D22 + 2D1 D2 cos(ka) .
This dispersion relation is qualitatively similar to the one illustrated in Fig. 5.4.

Answer 5.7.
a. The lattice vectors of a one-dimensional lattice lying along the x̂−axis are
R0m = max̂, and hence
U  (ma) U  (ma)
DL (R00m ) = Dxx (R0m ) = + U  (ma) − = U  (ma) .
ma ma
On the other hand, the second term in Eq. (5.22) does not contribute to the
transverse vibrations, and hence
U  (ma)
DT (R00m ) = Dyy (R0m ) = Dzz (R0m ) = .
ma
b. The Lenard-Jones potential, when differentiated with respect to the distance
between two atoms, gives
24  σ 7  σ 13 24   σ 14  σ 8
U  (ma) = −2 , and U  (ma) = 2 26 −7 .
σ ma ma σ ma ma
The remaining steps in the solution are rather straightforward.

Answer 5.8.
The function uiμ depends on the coordinate R0i . Denoting this function by uiμ =
Uμ (R0i ), then ui μ = Uμ (R0i +a1 ), and hence Uμ (R0i +a1 ) = C(a1 )U (R0i ). Repeated
displacements by the same lattice vector yield Uμ (R0i + 2a1 ) = C(2a1 )U (R0i ) =
[C(a1 )]2 U (R0i ), and consequently C(na1 ) = [C(a1 )]n , or alternatively ln[C(na1 )] =
n ln[C(a1 )]. The only function of na1 that increases linearly with n is the function
proportional to this vector, i.e., ln[C(na1 )] = na1 · K, where K is an arbitrary
vector. Since the absolute value of C is 1, the vector K has to be imaginary, i.e.,
K = ik, with k being a real vector. This completes the proof: C(a1 ) = exp[ik · a1 ].
A similar argumentation with the other basis vectors yields C(n1 a1 +n2 a2 +n3 a3 ) =
exp[ik · (n1 a1 + n2 a2 + n3 a3 )]. Choosing R0i = 0, i.e., ui = A exp[−iωt], and R0i =
n1 a1 +n2 a2 +n3 a3 , results in ui = C(n1 a1 +n2 a2 +n3 a3 )ui = A exp[i(k·R0i −ωt)].
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 377

Lattice vibrations 377

Answer 5.9.
a. Following Eq. (5.22), the spring constants between second neighbors are denoted
by Dμν (R0 ) = D3 δμν + (D4 − D3 )R̂0μ R̂0ν . This entails the following additions to
the diagonal harmonic coefficients: ΔDzz √= D3 ,√ΔDxx = ΔDyy = D√3 + (D4 −
D3 )/2 = (D3 + D4 )/2, where D3 = U  (a 2)/(a 2) and D4 = U  (a 2). Since
R̂0ijx R̂0ijy = ±1/2, two additional non diagonal terms are added to the matrix of
coefficients, ΔDxy = ±(D4 − D3 )/2, where the positive sign is for R0ij = ±a(x̂ + ŷ),
and the negative one is for R0ij = ±a(x̂ − ŷ), and ΔDxz = ΔDyz = 0. Equation
(5.25) then yields
ΔK  yy (k) = (D + D )[4 − 2 cos(k a + k a) − 2 cos(k a − k a)]/2
 xx (k) = ΔK 3 4 x y x y
= 2(D3 + D4 )[1 − cos(kx a) cos(ky a)] ,
 zz (k) = 4D3 [1 − cos(kx a) cos(ky a)] ,
ΔK
 xy (k) = (D3 − D4 )[cos(kx a + ky a) − cos(kx a − ky a)]
ΔK
= −2(D3 − D4 ) sin(kx a) sin(ky a) .
The equations of motion are [see Eq. (5.24)]
 xx (k) + ΔK
M ω 2 Ax = [ K  xx (k)]Ax + ΔK
 xy Ay ,
 yy (k) + ΔK
M ω 2 Ay = [ K  yy (k)]Ay + ΔK
 xy Ax ,
 zz (k) + ΔK
M ω 2 Az = [ K  zz (k)]Az .

Therefore, the frequency of the modes which vibrate along the normal to the plane
is
 zz (k) = 4DT 2 [sin2 (kx a/2) + sin2 (ky a/2)] + 4D3 [1 − cos(kx a) cos(ky a)] .
M ωT2 2 = K
The in-plane vibrations’ frequencies are derived by solving two coupled linear
equations, leading to a quadratic equation for the allowed frequencies, (M ω 2 −
K  yy ) = (K
 xx )(M ω 2 − K  xy )2 . Ignoring the relations between the spring constants
implied by the equilibrium-statecondition of the central forces, leads to results not
much different than those that pertain to the situation where only forces between
nearest neighbors are involved. The three vibrational modes are presented in Fig.
5.27(b), for the three main directions in the Brillouin  zone (with D4 = 0.5DL ,
D3 = 1.5DL , and DT 1 = 2DL , and in units of a DL /M ). For a longitudinal wave
propagating along one of the axes, say the x̂− axis, the off diagonal element K  xy van-
ishes, and the dispersion relation is M ωL 2
=K  xx (k) = 4(DL +D3 +D4 ) sin2 (kx a/2).
This dispersion law is similar to the one obtained for a one-dimensional chain, and
for the same wave in the case discussed above: atoms on a line normal to the
direction
 of motion vibrate in unison. The sound velocity, though, is modified,
c = a (DL + D3 + D4 )/M . In a similar way, the frequency of the transverse wave
vibrating along the ŷ−axis and propagating along x̂ is given by M ωT2 1 = K  yy (k) =
2
4(DT 1 + D3 + D4 ) sin (kx a/2). The one of the transverse wave oscillating along
ẑ and moving along x̂ is M ωT2 2 = K  zz (k) = 4(DT 2 + 2D3 ) sin2 (kx a/2), with the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 378

378 Introduction to Solid-State Physics


sound velocity c = a DT 2 + 2D3 )/M . All in all, one obtains three distinct disper-
sion relations, with three different sound velocities. All degeneracies are removed,
and there are three branches in each Brillouin zone.

(a) (b)

Fig. 5.27

As in one dimension, the constraint imposed by the condition √ of equilibrium

modifies
√ the picture.√ At equilibrium U tot = 2N [U (a) + U ( 2a)], leading to
U  ( 2a) = −U  (a)/ 2. It follows that DT 2 + 2D3 = 0, and then the sound
velocity of vibrations normal to the plane vanishes. The dispersion law of these
vibrations is M ωT2 2 = K zz (k) = 2DT 2 [1 − cos(kx a)][1 − cos(ky a)] (check!), it van-
ishes on the axes in the Brillouin zone, and is approximately M ωT2 2 ≈ DT 2 a4 kx2 ky2
at the vicinity of the Γ point. The behavior of the in-plane modes is qualitatively
the same. Dispersion relations along chosen directions are depicted in Fig. 5.27(a).
Note in particular the parabolic dependence of the frequency on the wave vector of
the transverse oscillation normal to the plane around the Γ point in Fig. 5.27(b).
b. In the acoustic limit, the equations are

M ω 2 Ax ≈ a2 {[(DL + D3 + D4 )kx2 + (DT 1 + D3 + D4 )ky2 ]Ax + 2(D3 − D4 )kx ky Ay } ,

M ω 2 Ay ≈ a2 {[(DL + D3 + D4 )ky2 + (DT 1 + D3 + D4 )kx2 ]Ay + 2(D3 − D4 )kx ky Ax } ,
M ω 2 Az ≈ a2 (DT 2 + 2D3 )(kx2 + ky2 )Az .

The system has a square symmetry in the plane. A comparison with the equations
in Appendix D enables the identifications Cxx,xx = Cyy,yy = a(DL + D3 + D4 ),
Cxx,xy + Cxy,xy = 2a(D3 − D4 ), and Czx,zx = Czy,zy = a(DT 2 + 2D3 ). When the
forces are central the last two coefficients are zero.

Answer 5.10.
a. From the expression following Eq. (5.33) one finds ω 2 = (3D/M )[1 ± |U|/3],
where U = exp[ik · a1 ] + exp[ik · a2 ] + exp[ik · (a1 + a2 )]. √
The wave vector at the
K point is kK = 4πx̂/(3a). Using a1 = ax̂ and a2 = (x̂ + 3ŷ)a/2, shows that U
vanishes, and hence the two solutions are degenerate.
b. Substituting k = kK +√ q and expanding up to linear order in the components of
q, results in U = −ia(3i+ 3)(qx +iqy )/4+. . .. It follows that |U|2 = 3a2 |q|2 /4+. . .,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 379

Lattice vibrations 379

 √
and hence ω(kK + q) ≈ (3D/M )[1 ± ( 3/6)a|q|]. The frequency is linear in the
deviation of the vector from its value at the K point.

Answer 5.11.
a. The next-nearest neighbors of the atom whose displacements have been denoted
by un,m are precisely the atoms in the nearest cells that are of the same type;
the contributions of the interactions with them are thus those listed in Eq. (5.28)
(which pertains to the triangular lattice). The same argument holds for the atoms
of the other type. Consequently, the equations of motion (5.33) become
 − D(e−ik·a1 + e−ik·a2 + e−ik·(a1 +a2 ) )B ,
M ω 2 A = [3D + ΔK]A
 − D(eik·a1 + eik·a2 + eik·(a1 +a2 ) )A ,
M ω 2 B = [3D + ΔK]B
where
k · a  k · a 
 k · (a − a ) 
 = 4D2 sin2
ΔK 1
+ sin2 2 2
+ sin2 1
.
2 2 2
The frequencies (again derived by solving a quadratic equation) are M ω 2 = 3D +
ΔK  ± D| exp[ik · a ] + exp[ik · a2 ] + exp[ik · (a + a )]|. Figure 5.28(a) shows
1 1 2
the acoustic branch, and Fig. 5.28(b) the optical one, for D2 = 0.4D. As seen,
the competition between the two interactions creates a complex structure of the
dispersion relations.

(a) (b)

Fig. 5.28

b. The dependence of the expressions above on the wave vector k is via the terms
exp[ik · a1 ] and exp[ik · a2 ]. Since exp[iG · a1 ] = exp[iG · a2 ] = 1 for any reciprocal-
lattice vector, the equations are invariant under the translation k → k + G, and
the frequencies are not changed. In addition,
| exp[ik · a1 ]+ exp[ik · a2 ] + exp[ik · (a1 + a2 )]|2 = 3 + 2 cos(k · a1 ) + 2 cos(k · a2 )
+ 2 cos[k · (a2 − a1 )] .
 are invariant under the transformation
Both this expression and the one for ΔK
a1 → a2 → a2 − a1 → −a1 , which occurs when k is rotated by 60◦ . The dispersion
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 380

380 Introduction to Solid-State Physics

relations are thus invariant with respect to this rotation, as indeed can be seen in
Figs. 5.28.

(a) (b)

Fig. 5.29

c. The limit |k|a  1 allows one to expand the expressions obtained above. The
frequency of the acoustic branch is given by M ω 2 = (D/4 + 3D2 /2)(|k|a)2 . It is
independent of the
 direction of the wave vector; the corresponding sound velocity
is c = ω/|k| = a (D/4 + 3D2 /2)/M .
d. At k = 0 one finds M ω 2 (k = 0) = 3D ± 3D, yielding M ω 2 (k) = 6D for the
optical frequency. The equations of motion give A = −B, that is, the two sublattices
are vibrating one against the other (as in the case of two different atoms in the unit
cell of a one-dimensional lattice).
e. At the M point k = b1 /2 (where b1 is a reciprocal-lattice vector, Fig. 3.11
or Fig. 5.9), and therefore k · a1 = π and k · a2 = 0. It follows that M ω 2 (M ) =
(3 ± 1)D + 8D2 , and A = ±B (the upper sign pertains to the acoustic branch). At
the K point k = 4πx̂/(3a), and then k · a1 = 4π/3 and k · a2 = 2π/3, leading to
M ω 2 (K) = 3D + 9D2 , i.e., the two branches are degenerate. The amplitudes ratio
in the optical branch is

A eik·a1 + eik·a2 + eik·(a1 +a2 )

= − −ik·a ,
B |e 1 + e−ik·a2 + e−ik·(a1 +a2 ) |

as shown in Fig. 5.29(b) (the thick and thin lines correspond to the real and
imaginary parts of the expression, respectively). The same expression but with
the inverse sign describes the acoustic branch. Both numerator and denominator
vanish at the K point√leading to a discontinuity there: the real part is ±0.5, and
the imaginary one is 3/2. As all coefficients in Eqs. (5.33) vanish at that point,
one may choose any values for A and B. In any event, the absolute values of the
two amplitudes are equal; only their phases vary with the wave vector.

Answer 5.12.
The only difference between the two crystals is that in the present case the two ions
in the unit cell are not identical, i.e., their masses differ. Hence, Eqs. (5.32) are
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 381

Lattice vibrations 381

modified to be
M ün,m = −3Dun,m + D(vn−1,m + vn,m−1 + vn−1,m−1 ) ,
mv̈n,m = −3Dvn,m + D(un+1,m + un,m+1 + un+1,m+1 ) .
Equations (5.33) are modified to be
M ω 2 A = 3DA − D(e−ik·a1 + e−ik·a2 + e−ik·(a1 +a2 ) )B ,
mω 2 B = 3DB − D(eik·a1 + eik·a2 + eik·(a1 +a2 ) )A ,
and the dispersion relations are the solutions of the quadratic equation
(M ω 2 − 3D)(mω 2 − 3D) = D2 | exp[ik · a1 ] + exp[ik · a2 ] + exp[ik · (a1 + a2 )]|2 .
Figure 5.30 displays the acoustic and optical branches for M/m = 1.1. The prime
difference between it and the corresponding one for graphene is at the K point.
Whereas in graphene the two branches merge together there, this degeneracy is
lifted for the boron nitride. This is clear from the quadratic equation: at this point
2 2
the right hand-side vanishes, and then the two roots are M ω− = 3D and mω+ = 3D
(note in particular the similarity with the one-dimensional lattice whose unit cell
contains two different atoms).

Fig. 5.30

Answer 5.13.
The equations of motion for the “inner” points n = 2, 3, . . . , N − 1 are intact, and
are given by Eq. (5.3). These are solved by the monochromatic wave functions
exp[i(kna − ωt)] or exp[i(−kna − ωt)], and hence by any linear combination of
them. The equations of motion for the edges, n = 1 and n = N , are again given by
Eq. (5.3), but these should be augmented by boundary conditions. The condition
u0 = 0 is fulfilled by choosing the difference between the two wave functions, i.e.,
un = A sin(kna) exp[−iωt]. These functions, which describe a standing wave, indeed
solve Eq. (5.3) (this can be shown by a direct substitution), with the same dispersion
relations derived before, Eq. (5.5). It remains to consider the second boundary
conditions, uN +1 = sin[k(N + 1)a] = 0, which implies k = k = π /[(N + 1)a].
Though it seems that the density of wave numbers is doubled (as compared to the
solution based on periodic boundary conditions), the difference between subsequent
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 382

382 Introduction to Solid-State Physics

wave numbers is just π/[(N +1)a] as compared to 2π/(N a). (The disparity between
the two denominators is negligible for large N ’s.) Moreover, the solutions for k and
−k are identical up to a sign, and thus describe the same eigenmode. It follows
that one has to include only the positive solutions within the (half) Brillouin zone,
= 1, 2, . . . , N . All in all, there are N independent solutions, as before.

Answer 5.14.
a. At equilibrium, the locations of the atoms are R0n = R(cos[2nπ/N ], sin[2nπ/N ]);
note that R0n+N = R0n , which implies that the ring obeys Eq. (5.34). The same
relation between the atoms’ locations pertains also to the atoms vibrating around
the equilibrium positions.
b. Denote the instantaneous locations of the atoms by Rn = R(cos[2nπ/N +
ϑn ], sin[2nπ/N + ϑn ]). The distance between two nearest neighbors is
π δϑn+1,n  π π 
Rn,n+1 = 2R sin + ≈ R 2 sin + cos δϑn+1,n + . . . ,
N 2 N N
where δϑn+1,n = ϑn+1 −ϑn . Thus, the central potential between these two neighbors
is
U (Rn,n+1 ) = U (X0 ) + U  (X0 )Y0 δϑn+1,n + U  (X0 )(Y0 δϑn+1,n )2 /2 + . . . ,
where
X0 = 2R sin(π/N ) , Y0 = R cos(π/N ) , δϑN +1,N = ϑ1 − ϑN .
N
Using n=1 δϑn+1,n = 0, it follows that

N 
N
U (Rn,n+1 ) = N U (X0 ) + U  (X0 )(Y0 )2 (δϑn+1,n )2 /2 + . . . .
n=1 n=1
The equations of motion are
M ϑ̈n = −D(2ϑn − ϑn−1 − ϑn+1 ) , n = 2, 3, . . . , N − 1 ,
M ϑ̈N = −D(2ϑN − ϑN −1 − ϑ1 ) M ϑ̈1 = −D(2ϑ1 − ϑN − ϑ2 ) ,

where D = U (X0 )Y02 .
For N = 6, the equations can be written as
⎡ ⎤ ⎡ ⎤⎡ ⎤
ϑ̈1 −2 1 0 0 0 1 ϑ1
⎢ ϑ̈ ⎥ ⎢ 1 −2 1 0 0 0 ⎥ ⎢ ϑ ⎥
⎢ 2⎥ ⎢ ⎥⎢ 2⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎢ ϑ̈ ⎥ ⎢ 0 1 −2 1 0 0 ⎥ ⎢ ϑ3 ⎥
M ⎢ 3⎥ = D⎢ ⎥⎢ ⎥ .
⎢ ϑ̈4 ⎥ ⎢ 0 0 1 −2 1 0 ⎥ ⎢ ϑ4 ⎥
⎢ ⎥ ⎢ ⎥⎢ ⎥
⎣ ϑ̈5 ⎦ ⎣ 0 0 0 1 −2 1 ⎦ ⎣ ϑ5 ⎦
ϑ̈6 1 0 0 0 1 −2 ϑ6
The extension to other values of N is obvious. The column vector
(exp[ik ], exp[2ik ], exp[3ik ], exp[4ik ], exp[5ik ], exp[6ik ]) is an eigenvector of the
matrix on the right hand-side, with the eigenvalue 2[cos k − 1], provided that
exp[iN k ] = 1, which follows from the first and the last raws in the matrix. There-
fore, there are just N allowed values of the “wave number”, k = 2π/N , with
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 383

Lattice vibrations 383

= 1, 2, 3, . . . , N . The solution of the equations of motion can be presented as a

sum of the eigenvectors, ϑn = =1 ϑ exp[ik n], and then
N


N
¨ 
N
M ϑ exp[ik n] = 2D [cos k − 1]ϑ exp[ik n] ,
=1 =1

i.e.,
¨
M ϑ = −2D[1 − cos k ]ϑ .
This is the equation of an harmonic oscillator, that gives the time dependence of
the amplitudes ϑ , with the frequencies given in Eq. (5.6).

Answer 5.15.
The integrals, for instance for one
 of the vibrational modes, are computed using the
∞
identity −∞ dx exp[−ax2 /2] = 2π/a. One finds
 
Zα = ( 2πM/β 2π/(βM ωα2 )/h) = (βωα )−1 ,
and hence
 
ln Z = ln Zα = −[N ln(β) + ln(ωα )] .
α α

It follows that
E = −∂ ln Z/∂β = N kB T ,
and
CV = (∂E/∂T )V = N kB .

Answer 5.16.
The area below the dashed line in Fig. 5.13 is the total classical energy. The area
up to a temperature T is kB T . The dark area in the figure is
 ∞  ∞
dT CV = (∂E/∂T )V = E(T = ∞) − E(T = 0) ,
0 0
where the expressions on the right-hand side are the average quantum energies,
given by Eq. (5.52). This area is then
 ∞  T
A= dT [kB − CV (T )] = lim [kB T − dT CV (T )
0 T →∞ 0
= lim [kB T − E(T ) + E(T = 0)] .
T →∞

Equation (5.52) implies that E(T = 0) = ωα /2. Expanding this equation at high
temperatures yields
E(T ) = ωα /2 + ωα /[βωα + (βωα )2 /2 + O((βωα )3 )] ≈ kB T + O(βωα ) .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 384

384 Introduction to Solid-State Physics

All in all,

A = lim [kB T − (kB T − ωα /2)] = ωα /2 ,

T →∞

precisely the zero-point motion energy, which distinguishes the quantum-mechanical

result from the classical one.

Answer 5.17.
a. Equation (5.6) implies that ω = ω0 | sin(ka/2)|, which in the long-wave-length
limit becomes ω = ω0 |ka/2|, and thus the sound velocity is c = ω0 a/2. The total
number of states in the range 0 < ω < ω0 , from Eq. (5.57), is
 ω0 
L ω0 dω
dωg(ω) = 
0 πc 0 1 − (ω/ω0 )
 1 1
Lω0 dx 2L  L
= √ = arcsin(x) = = N .
πc 0 1−x 2 πa 0 a
b. Expanding the integrand in Eq. (5.59) in exp[−x] yields
∞ ∞
ex −x −x −2 d 1 d  −mx 
= e (1 − e ) = − = − e = me−mx .
(ex − 1)2 dx 1 − e−x dx m=0 m=1
∞ ∞
Inserting this expression into the integral results in 0 dxx2 m=1 m exp[−mx] =
∞ 2
m=1 2/m = 2ζ(2) (integrating twice by parts). It remains to find ζ(2). This can
be accomplished by expanding the function sin x/x up to second order in x. The
function vanishes at the points x = ±mπ, and therefore
sin x  x2  x2  x2  x2
= 1 − 2 1 − 2 1 − 2 . . . = 1 − 2 ζ(2) + O(x4 ) .
x π 4π 9π π
On the other hand, a straightforward Taylor expansion gives sin x/x = 1 − x2 /6 +
O(x4 ), thus ζ(2) = π 2 /6.
c. At low temperatures, where xmin = βω+ (π/a)  1, the function
x2 exp[x]/(exp[x] − 1)2 is bounded from above by x2min exp[xmin ]/(exp[xmin ] − 1)2 ≈
x2min exp[−xmin ]. Replacing this function in the integrand by its bound (which can
be taken out of the integral), it remains to calculate the integral that appears in
part (a); in this way one derives an upper bound on the integral,

CVoptical < kB N x2min exmin /(exmin − 1)2 ≈ kB N x2min e−xmin .

As the upper bound decays exponentially when β increases, the specific heat follows
this behavior (or even decays faster).

Answer 5.18.
a. The dispersion relation is

M ω 2 = 2DT 2 [1 − cos(kx a) + 1 − cos(ky a)] ,

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 385

Lattice vibrations 385

and therefore the group velocity is

vg = ∇k (ω) = [DT 2 a/(M ω)][sin(kx a)x̂ + sin(ky a)ŷ] .
This vanishes only when both components are zero, i.e., when kx a = 0, π and
ky a = 0, π; these are the points of the type M (the centers of the faces) and the
points of the type K (the corners) in the Brillouin zone.
b. At the vicinity of the corner K = (π/a, π/a) one may write kx a = π − kx a and
ky a = π − ky a, and obtain
M ω 2 = 2DT 2 [2 + cos(kx a) + cos(ky a)] ≈ DT 2 [8 − {(kx )2 + (ky )2 }a2 ] .
The equal-frequency surfaces are circles around the point K. The group velocity on
a circle of radius k  is radial, and its magnitude is |vg | = DT 2 a2 k  /(M ω) (check!).
The circumference is 2πk  , and therefore Eq. (5.65) yields

NMω N 2M
g(ω) = ≈
2πDT 2 π DT 2
(note that V = a2 ). The density of states is independent of the frequency.
c. It suffices to calculate the contribution of k’s from the upper right quarter of the
Brillouin zone, as each quarter contributes the same amount. The line (π/a, 0) →
(0, π/a) is ky a = π − kx a. Inserting it into the expression for the frequency confirms
that it is indeed an equal-frequency curve, with M ω 2 = 4DT 2 . Substituting into
the expression for the group velocity,
vg = [DT 2 a/(M ω)][sin(kx a)x̂ + sin(ky a)ŷ] ,
√
gives |vg | = [DT 2 a/(M ω)] 2 sin(kx a). Hence, the contribution of this segment to
π
the integral in Eq. (5.66) is proportional to 0 dx/(sin x) = ln[tan(x/2)]|x=πx=0 = ∞,
where x = kx a. This logarithmic dependence of the integral on its bounds implies
that the same dependence still prevails upon shifting the frequency a bit below or
above this value.
d. The line described in part (c) is the only equal-frequency line in the Brillouin
zone that includes points where the group velocity vanishes. It follows that the
density of states is finite at any other point in the Brillouin zone.  It is a linear
line near ω = 0 [see Eq. (5.67)], diverges logarithmically at ω = 4DT 2 /M , and
returns
 to the constant value calculated in part (b) around the maximal frequency
ω = 8DT 2 /M . The density of states vanishes for frequencies above this value.
Figure 5.31 illustrates this schematic behavior.

Answer 5.19.
a. For X  1 the integrand may be expanded in powers of x, to yield
 X  X
xd+1 ex x2 x2 x3
dx x = dxxd+1 (1 + x + + . . .)/(x + + + . . .)2
0 (e − 1) 2
0 2 2 6
 X
x2 Xd X d+2
= dxxd−1 (1 − + . . .) = − + ... ,
0 12 d 12(d + 2)
leading to the required answer.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 386

386 Introduction to Solid-State Physics

Fig. 5.31

b. It is possible to express the integral in the form

 X  ∞  ∞
xd+1 ex xd+1 ex xd+1 ex
dx x = dx − dx .
0 (e − 1)2 0 (ex − 1)2 X (ex − 1)2
As mentioned after Eq. (5.69), the first integral is Id , which attains finite values.
Since X  1, the integrand in the second term can be expanded in powers of
exp[−x],
 ∞  ∞
xd+1 e−x
dx = dxxd+1 e−x [1 + O(e−2x )] . (5.114)
X (1 − e−x )2 X

Integrating by parts,
 ∞ ∞

Jd = dxxd+1 exp[−x] = −xd+1 e−x  + (d + 1)Jd−1 = X d+1 e−X + O(X d e−X ) .
X X

This completes the proof.

Answer 5.20.
In spherical symmetry, Eq. (5.65) yields g(ω) = [N V Sd /(2π)d ][k d−1 /vg ]. The
group velocity, from the dispersion relation, is vg = dω/dk ∝ k η/2−1 , and hence
g(ω) ∝ ω 2d/η−1 . As η = 2 for the acoustic dispersion, Eq. (5.67) is reproduced.
The Debye frequency is derived from a generalized version of Eq. (5.68): N =
ωD 2d/η
0
dωg(ω) ∝ ωD . Equation (5.69) yields
 ωD
eβω
CV = dωg(ω)(βω)2 βω
0 (e − 1)2

2k dN  kB T 2d/η βωD 2d ex
= B dxx η +1 x .
η ωD 0 (e − 1)2
When kB T  ωD , the upper bound may be taken as infinity, to obtain CV ∝
T 2d/η . In the opposite limit one obtains the Dulong-Petit law, CV ≈ kB T , which is
independent of the details of the dispersion relation.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 387

Lattice vibrations 387

Answer 5.21.
The specific heat due to a single acoustic branch is given by Eq. (5.55),
 eβω(k)
CV = kB [βω(k)]2 ,
k
(eβω(k) − 1)2
where the sum runs over the discrete values of the wave vector, k1 ,2 ,3 =
( 1 /N1 )b1 + ( 2 /N2 )b2 + ( 3 /N3 )b3 , with −Nm /2 ≤ m < Nm /2 − 1. The three
sums are turned into integrals in Eq. (5.63). However, the thickness along a3 is fi-
nite: N3 = L/a3  N1 , N2 , and may be quite small; the corresponding sum cannot
be turned into an integral. (The other two sums are still replaced by integrals.) It
follows that Eq. (5.55) becomes
 N N a a 
1 2 1 2
f [ω(k)] = d2 kf [ω(k)] .
(2π)2
k 3

The Debye approximation, when applied to a system of tetragonal symmetry, im-

plies that the dispersion relation is
ω 2 = c2 (kx2 + ky2 ) + c2⊥ kz2 .
As the system is symmetric in the XY plane, the integrals are conveniently carried
out using cylindrical coordinates, and thus d2 k = 2πkdk = 2πωdω/c2 , using the
dispersion relation. Summing over all waves with frequency smaller than the Debye
frequency, i.e., ω 2 < ωD
2
, results in
  N N a a  ωD
1 2 1 2
f [ω(k)] = 2 dω ωf (ω) , (5.115)
(2π)c2 c ⊥ kz
k kz

where the factor 2 is due to the restriction of the sum to positive values, 0 ≤ kz =
2π 3 /L ≤ πN3 /L. A positive contribution is obtained only when c⊥ kz < ωD . The
total number of vibrational modes is N = N1 N2 N3 = k 1. One therefore inserts
f (ω) = 1 in Eq. (5.115) and finds
N   ωD 1 2
= ωdω = (ωD − c2⊥ kz2 )
Z c ⊥ kz 2
kz ≥0 kz
2
ωD N3 1  πc⊥ 2
= − N3 (N3 + 1)(N3 + 2) ,
4 12 L
N/2
where Z = N1 N2 a1 a2 /(πc2 ), and =0 2 = N (N +1)(N +2)/24. When N3 is very
large, the sum over kz can be turned into an integral; in this way the expressions for
the three-dimensional system, Eq. (5.68) are reproduced, with N3 = ωD L/(πc⊥ ).
In contrast, when the sum comprises only a small number of terms, the second term
in small, and the expression is reduced to the formula for a two-dimensional crystal.
For the calculation of the specific heat one inserts
eβω
f (ω) = kB (βω) ,
eβω − 1)2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 388

388 Introduction to Solid-State Physics

into Eq. (5.115). Changing there the integration variable leads to

3 /2  βωD
 k T 2 N ex 2π
CV = kB Z B
dxx3 x , α = βc⊥ .
 α3 (e − 1) 2 L
3 =0

The two-dimensional expression for the specific heat is reproduced upon keeping
only the first term, 3 = 0, in the sum. For α > 1, i.e., c⊥ 2π/L > kB T , each of the
other terms in the sum is in fact the specific heat of an optical mode: it is exponen-
tially small at low temperatures and tends to kB (the Dulong-Petit law) at higher
ones. At low temperatures the sum is dominated by its first term, and the resulting
specific heat is that of a two-dimensional system, CV ∝ T 2 . In the intermediate
range c⊥ 2π/L < kB T < ωD the lower bounds of the integrals for α 3 < 1 are
relatively small, and their contributions are comparable to that of the first term.
When this range is wide enough (i.e., L is large enough) the transformation to a
(three-dimensional) integration results in a three-dimensional behavior, CV ∝ T 3 .
Figure 5.32 displays the specific heat, CV /T 2 (scaled by the temperature) in arbi-
trary units, as a function of the temperature. The curves, in descending order, are
for N3 /2 = 1 , 3 , 10 , 1000. The horizontal range at low temperatures corresponds
to the two-dimensional dependence, while the region to its right (N3 > 2) portrays
an effective three-dimensional behavior.

Fig. 5.32

Inter-dimensional crossovers like the one described here appear quite often in
condensed matter. They occur also in one-dimensional systems, for instance, nano
tubes covered with cylindrical layers, and quasi-two-dimensional systems (or quasi-
one-dimensional ones), built of planar lattices (or wires) with very weak couplings
in-between them.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 389

Lattice vibrations 389

Answer 5.22. The one-dimensional density of states in the Debye approximation,

from Eq. (5.57), is g = L/(πc). It follows that
 x2
a dx βωmin βωD
u2i (Ri0 ) = cothx , x1 = , x2 = .
2πM c x1 x 2 2
The integral does not possess an analytical from, but when x is small, the lower
limit implies that
 x2
a dx a LakB T
u2i (Ri0 ) ≈ 2
≈ = .
2πM c x1 x 2πM cx1 2π 2 M c2
According to the Lindemann criterion, the crystal is stable for u2 (Ri0 ) < a2 /100,
and therefore kB T < [π 2 M c2 /(50N )], where N = L/a.

Answer 5.23. At high temperatures, Eq. (5.76) yields

 ωD  −1
d d−2 βω dkB T
u2 (R0i ) ≈ dωω = 2 .
d
2M ωD ωmin 2 (d − 2)M ωD
According to the Lindemann criterion, u2 (R0i ) < a2 /100, and therefore the melt-
ing temperature is kB Tm = a2 M (d − 2)ωD
2
/(100d).

Answer 5.24. Periodic boundary conditions imply that uN = u−N . The last two
equations of motion in (5.78) are thus modified to be
M ü±n = −D(2u±n − u±(n−1) − u±(n+1) ) , 1 < n < N − 2 ,
M ü±(N −1) = −D(2u±(N −1) − u±(N −2) − uN ) ,
M üN = −D(2uN − uN −1 − u−(N −1) ) .
The corresponding equations for the symmetric and anti-symmetric functions are
modified as well,
± ±
M ü±
n = −D(2u±
n − un−1 − un+1 ) , 1 < n < N − 2 ,

N −1 = −D(2uN −1 − uN −2 − 2uN ) ,
M ü+ + +

M ü− − −
N −1 = −D(2uN −1 − uN −2 ) , M üN = −D(2uN − u+
N −1 ) .

The equations of motion for the anti-symmetric functions are similar to those ob-
tained previously, upon interchanging N by N − 1. It follows that there are N − 1
anti-symmetric wavy solutions. The equations of motion for the symmetric func-
tions are similar to the previously-derived ones as well; the difference is that one
has to insert in the equation for u+N −1 the expression uN = DuN −1 /(D − M ω ).
+ 2

With solutions of the type (5.80), one finds

 2D02  B+  B+ 
M ω 2 − D − D0 − A +
C + + D(A + 2
C + =0,
mω 2 − 2D0 C C2
 2D2  B+   B+ 
+ N −1 + N −2
M ω 2 − 2D − A C + + D A C + =0,
M ω 2 − 2D C N −1 C N −2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 390

390 Introduction to Solid-State Physics

modifying Eq. (5.85) to be

B+ C 2 + 3 2n (D − D0 − X)C − D
− = C = .
A+ 3 + C2 (D − D0 − X)/C − D
For C = exp[ika] one obtains the same wavy solutions as found in the main text.
For |C| < 1 and N  1 the left hand-side may be again discarded, leading to
precisely the same equation as found in the main text for the localized mode, (D −
D0 − X)C − D = 0. This could have been anticipated, as the localized state decays
exponentially and thus is insensitive to conditions on boundaries that are situated
far away from the foreign atom.

Answer 5.25. Denoting the momentum of the outgoing neutrons by k  , momen-

tum conservation yields k  = k ± (G − K), where G = 2πh/a, and h is an arbitrary
integer. Energy conservation implies
2 2 2
ω0 | sin(Ka/2)| = ω(K) = ± [k − (k  )2 ] = ± k 2 − [k ± (G − K)]2 ,
2mn 2mn
(5.116)

where ω0 = 2 D/M . Figure 5.33 displays a graphical solution of this equation: the
thin line is the left hand-side, and the thick solid line (or the thick dashed one) is
the right hand-side for phonon emission (upper sign) or absorption (lower sign), as
functions of K − G in units of G0 = 2π/a. As seen, the right hand-side is a parabola
cutting the horizontal axis at the origin and at ±2k, with a maximum (minimum)
which equals (up to a sign) the energy of the impinging particle, ±2 k 2 /(2mn ), at
K − G = ±k. Each crossing of the two curves gives a possible value of the energy
and momentum of the phonon. The number of solutions varies with the value of k.

Fig. 5.33: The two sides of Eq. (5.116).

Answer 5.26.
a. Equation (5.51) is equivalent to the expression Fph = kB T α ln[2sinh(βωα /2)].
b. Equating the derivative of F0 (T, V ) from part (a) with respect to V to zero
yields the required result.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 391

Lattice vibrations 391

c. Equation (5.52) is equivalent to Eph = α (ωα /2)coth(βωα /2). Assuming

that ∂ ln ωα /∂ ln V is independent of α gives the required result.
d. Differentiation of the result of the previous part with respect to the temperature
leads to the required result.

Answer 5.27.
a. Postulating that B is along ẑ, B = Bẑ, then μ̇z = 0, that is, the z−component of
the magnetic moment does not vary in time. The equations for the two components
normal to the field are μ̇x = γBμy and μ̇y = −γBμx . An equation for μx is
derived by differentiating the first and inserting the result into the second, μ̈x =
−(γB)2 μx . The solution is μx = μ(0) cos(ωL t), with the Larmor frequency ωL = γB.
Substituting in the first equation gives μy = μ̇x /ωL = −μ(0) sin(ωL t). Therefore,
the component of the magnetic moment which is perpendicular to the field rotates
around it with the Larmor frequency. The magnetic moment vector propagates at
a fixed rate on a cone whose axis is parallel to the magnetic field.
b. The energy of the system, in terms of the magnetic moments, is

ĤM = − J nm μn · μm ,
nm

where J nm = Jnm /(gμB )2 . It follows that Ĥ = − n μn · Bn , where Bn =

m J nm μm . Hence,

μ̇n = γμn × Bn = μn × J nm μm ,
m

J nm = γJ nm = γJnm /(gμB )2 = Jnm /(gμB ) .

Answer 5.28.
Inserting μn± in Eqs. (5.103) yields
dμn± 
= ±iμ J nm (μn± − μm± ) .
dt m

With wavy solutions, μn± = μ± exp[i(k · R0n − ωt)], this equation yields −iωμ± =
 − J(k)]μ
±iμ[J(0)   
± , i.e., ω = ∓[J(0) − J(k)]. It follows that the solution μn− =
μ− exp[i(k · Rn − |ω|t)] propagates along the vector k, while the other solution,
0

μn+ = μ+ exp[i(k · R0n + |ω|t)], propagates in the opposite direction. Assuming that
μ+ = 0 results in μx = μ− /2 and μy = iμ− /2, and thus there is a phase difference
of π/2 between the spin components; the spins rotate in the XY plane.

Answer 5.29.
a. The equations of motion in three dimensions are identical to those presented in
Eq. (5.102), but the sums run over nearest neighbors alone. For a cubic crystal
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 392

392 Introduction to Solid-State Physics

of lattice constant a, whose nearest-neighbor exchange coefficient is J, Eq. (5.105)

yields


J(k) = J nm eik·Rnm = 2J[cos(kx a) + cos(ky a) + cos(kz a)] ,
m

(Rnm is a vector connecting two nearest neighbors) and Eq. (5.106) gives

ω = 4μJ[sin2 (kx a/2) + sin2 (ky a/2) + sin2 (kz a/2)] ≈ Δ|k|2 ,

with Δ = μJa2 .
b. The dispersion relation of magnons in a ferromagnet is that found in problem
5.20, with η = 4. It follows from the results of that problem that CV (FM) ∝
T d/2 at low temperatures. At high temperatures one obtains the Dulong-Petit law,
CV (FM) = kB N , where N is the number of magnetic moments in the system.

Answer 5.30.
The equations of motion are
(1)
dμnx (2)
= −μJ(2μ(1) (2)
ny + μny + μn−1y ) ,
dt
(1)
dμny (2)
= μJ(2μ(1) (2)
nx + μnx + μn−1x ) ,
dt
(2)
dμnx (1)
= μJ(2μ(2) (1)
ny + μny + μn+1y ) ,
dt
(2)
dμny (1)
= −μJ(2μ(2) (1)
nx + μnx + μn+1x ) .
dt
(1,2) (1,2) (1,2)
Inserting the variables μn± = μnx ± iμny yields

(1)
dμn+ (1) 2) (2)
=iμJ(2μn+ + μn+ + μn−1+ ) ,
dt
(2)
dμn+ (2) 1) (1)
= − iμJ(2μn+ + μn+ + μn+1+ ) .
dt
Finally, exploiting the wavy form for the spin waves, one finds

−ωμ+ = μJ[2μ+ + (1 + e−ika )μ+ ] ,

(1) (1) (2)

(2) (2) (1)

ωμ+ = μJ[2μ+ + (1 + eika )μ+ ] .

The determinant of the coefficients vanishes when ω = 2μJ| sin(ka/2)|, and there-
fore the dispersion relation is linear in the wave number for long waves, exactly like
(1,2)
the one of phonons. The equations for μn− yield the same dispersion relation.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 393

Lattice vibrations 393

5.11 Problems for self-evaluation

s.5.1.
As explained in Sec. 4.4, the van der Waals interaction stems from displacements
of the electrons’ cloud (whose total mass is m) relative to the position of the nu-
cleus (whose mass is M ). Strictly speaking, there is no reason to assume identical
displacements for both the electrons’ center of mass and the nucleus’ center of mass
(these are located at the same place at equilibrium). Consider longitudinal vibra-
tions in a single-atom, one-dimensional lattice, and denote the displacements of the
nucleus and the electrons in the n−th atom by un and ue,n , respectively. Assume
further that the potential energy of the relative motion is described by the form
U = D0 (ue,n − un )2 /2, and that the harmonic interactions between nearest neigh-
bors in the lattice result predominantly from the interaction between the electrons’
clouds on these atoms; these harmonic interactions are characterized by a spring of
constant D.
a. Find the frequencies of the vibrational modes, and write down the expression for
the dispersion relation of the acoustic branch in the m → 0 limit.
b. Discuss the effects of the coupling between the nuclei and the electrons on the
sound velocity, the maximal acoustic frequency, and the group velocity.

s.5.2.
The lattice constant of a single-atom chain is 4.85Å, the mass of each atom is
6.81 × 10−26 Kg, and the maximal frequency of the lattice vibrations is 4.46 × 1013
sec−1 .
a. Find the values of the spring constant and of the sound velocity.
b. One of the atoms is forced to oscillate at the frequency 5.75 × 1013 sec−1 , with
amplitude U0 . Find the dependence of the displacement amplitudes of the other
atoms on the distance from that atom. When is that amplitude equal to U0 /10?

Fig. 5.34

s.5.3.
A chain that consists of two different masses m and M is coupled to the table on
which it lies by springs (see Fig. 5.34). The constant of the spring joining two
neighboring (different) masses is D, and that of the spring coupling each mass to
the table is Ds . It is assumed that the masses oscillate only parallel to the table.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 394

394 Introduction to Solid-State Physics

a. Solve the classical equations of motion and find the dispersion law.
b. Find the frequencies for which the system is transparent. What can be concluded
about the stability of this lattice against melting?
c. Plot schematically the density of states of the lattice vibrations as a function of
the frequency, and the specific heat as a function of the temperature.

s.5.4.
The unit cell of a one-dimensional lattice contains three atoms, two of mass m and
one of mass M . The spring constants are all equal to D.
a. Write down the equations of motion and derive the expression that determines
the dispersion relations. How many branches are there?
b. Show that one of the branches is an acoustic one, and find the sound velocity.
c. Show that the solution approaches that of the single-atom lattice when the masses
are equal, M = m.
d. Draw schematically the dispersion relations for all branches in the first Brillouin
zone.

s.5.5.
Consider a rectangular lattice with lattice constants a and b. The area of the lattice
is A. The lattice sites contain two types of atoms, of masses M1 and M2 , arranged
such that the nearest neighbors of an atom of one type are all of the other type. The
atoms oscillate only perpendicular to the plane, and the spring constants are Da
and Db for the two lattice directions.
a. Determine the lattice structure and find the primitive unit cell. How many atoms
of each kind are there in the crystal?
b. Write down the classical equations of motion and derive the frequency spectrum
of the lattice. Describe the frequency spectrum when Da = 0 or Db = 0. The
crystal is perfectly symmetric when Da = Db and a = b. Repeat part (a) for this
case. Discuss in particular the spectrum of a wave along the (111) direction in the
symmetric configuration.
c. Find the equal-frequency lines in momentum space in the low-frequency limit.
Use them to derive the density of states.
d. Use the Debye approximation to obtain an expression for the contribution of the
the acoustic modes to the specific heat, and find the specific heat at low temperatures.
e. Find the contribution of the high-frequencies modes to the specific heat. Which
approximation is most suitable for its calculation? What is the relation between this
contribution and the one of low temperatures, calculated in part (c)?
f. Give an expression for the high-temperature specific heat.

s.5.6. The nearest-neighbor interaction in an FCC lattice is described by “springs”

whose constants are
Dμν (R0 ) = D1 δμν + (D2 − D1 )R̂0μ R̂0ν ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 395

Lattice vibrations 395

where D1 = DT and D2 = DL are related to the transverse and longitudinal vibra-

tions, respectively.
a. Derive the equations that relate the eigenfrequencies to the wave vector.
b. Obtain explicit expressions for these frequencies for waves corresponding to the
following lines in the Brillouin zone: k = (k, 0, 0), k = (2π/a, k, 0), k = (k, k, 0),
and k = (k, k, k). Determine the range of k−values for each of those, needed to
reproduce the path Γ → X → W → X → Γ → L in the Brillouin zone (Fig.
5.18). In particular, explain why the points (2π/a, 0, 0) and (2π/a, 2π/a, 0) are both
marked by X.
c. Use the results of part (b) to reproduce Fig. 5.18 (choose for D2 /D1 a value that
yields approximately the ratio of the frequencies at the X point).
d. Deduce from Fig. 5.18 the sound velocities for a wave along the cube diagonal,
and use them to estimate the Debye temperature of copper. The lattice constant of
copper is a = 3.61Å.
e. Explain qualitatively (without an explicit calculation) the way the figure is mod-
ified when the crystal under consideration has two ions in the unit cell, e.g., table
salt.

s.5.7.
a. The density of states depicted in Fig. 5.19(b) is singular at certain values of the
frequency. Can these points be identified from Fig. 5.19(a)?
b. Some of these van Hove singularities pertain to points where the dispersion
relation is maximal. Assuming that at the vicinity of such a maximum ω(k) ≈
ω0 − ω1 (k − k0 )2 , find the singularity of the density of states near the maximum.
c. What is the shape taken by the density of states of the vibrations in graphene near
the frequency corresponding to the K point in the Brillouin zone (problem 5.10)?

s.5.8.
Consider the two-atom, one-dimensional system displayed in Fig. 5.3. Assume that
the Einstein frequency ωE coincides with the maximal optical frequency.
a. Find the Debye frequency ωD , and the ratio ωD /ωE .
b. Copy Fig. 5.15, and add to it the density of states calculated within the Debye
approximation and the Einstein one.

s.5.9.
Express the wave length that corresponds to the Debye frequency in terms of the
volume of the unit cell. What is the ratio between this wave length and the distance
between nearest neighbors in crystals of copper (FCC) and iron (BCC)?

s.5.10. The crystalline structures of NaCl and KCl are identical, though the lattice
constants and the magnitudes of the “spring constants” differ. The sound velocities
in the three acoustic branches of each crystal are rather comparable. The Debye
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 396

396 Introduction to Solid-State Physics

temperatures are 310 K and 230 K, respectively. The specific heat of KCl at 5 K is
0.038J/mol/K. What is the specific heat of NaCl at 2 K?

s.5.11. The dispersion relation of a two-dimensional lattice is ω(k) = c(|kx | + |ky |).
Derive the density of states of this lattice. Compare it with the result of the Debye
approximation. Determine the Debye frequency. How does the specific heat vary
with the temperature?

s.5.12. A beam of neutrons of wave length λ = 3.5Å impinges on a single-atom SC

lattice, of lattice constant a = 4.25Å, along the normal to one of its faces. Part of
the neutrons leave the lattice along its principal diagonal, [111], with wave length
λ = 2.33Å. What are the frequency and the wave length of the phonon involved in
this inelastic scattering? Is the phonon absorbed, or is it emitted?

s.5.13. Find the frequencies of the spin waves in a one-dimensional ferri-magnet

(whose unit cell contains two antiparallel moments of unequal magnitudes), in which
the exchange interactions are only between nearest-neighbors.

5.12 Answers for the self-evaluation problems

Answer s.5.1.
a. The equations of motion are
M ω 2 un = D0 (un − ue,n ) ,
mω 2 ue,n = D0 (ue,n − un ) + D(2ue,n − ue,n−1 − ue,n+1 ) ,
where ω is the frequency of the motion. Assuming un = A exp[ikna] and un,e =
B exp[ikna], one finds
M ω 2 A = D0 (A − B) ,
mω 2 B = D0 (B − A) + 4D sin2 (ka/2)B .
The frequencies are the roots of a quadratic equation, as is also obtained for the
two-atom chain. Hence there are two branches, acoustic and optical. However, in
the limit where the electronic mass is extremely small compared to the mass of the
nucleus, the optical branch comprises very high frequencies, approaching infinity.
Only the acoustic branch survives, with the dispersion relation
M ω 2 = 4D sin2 (ka/2)/[1 + (4D/D0 ) sin2 (ka/2)] .
This dispersion is depicted in Fig. 5.35, for 4D/D0 = 0, .5, 1, 5, 10 (the smaller is
the ratio, the higher is the frequency).
b. In the acoustic limit one expands the dispersion relation in powers of the wave
number k; to leading order, M ω 2 ≈ 4D(ka/2)2 . This expression is identical to
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 397

Lattice vibrations 397

the dispersion relation of a chain in which the electrons and the nuclei move in
unison. The sound velocity is thus unaffected by the internal motion within each ion.
Likewise, the specific heat in the Debye approximation is not affected. In contrast,
2
the maximal frequency, attained for ka = π, is M ωmax = 4D/[1 + (4D/D0 )], and it
decreases as D0 is reduced. When the latter is small, it dominates the dispersion,
and the dispersion relation becomes M ω 2 ≈ D0 . It follows that the dependence
of the frequency on the wave number is gradually weakened, and so is the group
velocity. In this limit most of the motion takes place within each atom, and the
relative displacements of neighboring atoms have almost no effect.

Fig. 5.35

Answer s.5.2.
a. Using Eq. (5.6), one finds D = M ω02 /4 = 3.39×104 dyn/cm. The sound velocity
is c = aω0 /2 = 1.08 × 106 cm/sec.
b. The forced frequency exceeds the maximal frequency of the dispersion rela-
tion, thence the amplitudes of the vibrations decay with the distance away from the
special atom, un = U0 exp[−κan]. The coefficient that characterizes the decay is de-
rived from the relation ω/ω0 = cosh(κa/2). It follows that κa = 2arccosh(ω/ω0 ) ≈
1.5. A decay by a factor of 10 requires 10= exp[κna] ≈ exp[1.5n], i.e., n ≈ 1.55 or
na ≈ 7.5Å. Since n turned out to be non integer, the amplitude of the vibration is
smaller than U0 /10 already at the third atom in the chain.

Answer s.5.3.
a. A generalization of Eqs. (5.12) gives that the equations of motion are

M ün = −D(2un − vn−1 − vn ) − Ds un ,

mv̈n = −D(2vn − un+1 − un ) − Ds vn ,

where the notations are defined in Fig. 5.3. Assuming the forms un = A exp[i(kna−
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 398

398 Introduction to Solid-State Physics

ωt)] and vn = B exp[i(kna − ωt)] extends Eqs. (5.13), to yield

M ω 2 A = D[2A − (1 + e−ika )B] + Ds A ,
mω 2 B = D[2B − (1 + eika )A] + Ds B .
The determinant of the coefficients vanishes for [cf. Eq. (5.14)]
(M ω 2 − 2D − Ds )(mω 2 − 2D − Ds ) − 4D2 cos2 (ka/2) = 0 ,
and thus
 #
2D + Ds 4μ cos2 (ka/2)
ω2 = 1± 1− 1− ,
2μ M +m [1 + Ds /(2D)]2

where μ = M m/(M + m) is the reduced mass. The acoustic and optical branches
shown in Fig. 5.4 are reproduced when Ds vanishes. They are presented in Fig.
5.36 by the dashed curves (for M = 2m). The solid curves in Fig. 5.36 display
the dispersion relations for Ds = D, with the frequency measured in units of ω0 =

2D/μ.
b. When Ds = 0, the lowest possible frequency (for k = 0) is finite; the acoustic
regime in which the frequency is linear in the wave number disappears. The two
branches are both “optical”; this means that the average of the displacement squared
does not diverge, and the instability problem discussed in Sec. 5.5 does not arise.
As mentioned, the coupling to the substrate stabilizes the crystal. At the
edge of the Brillouin zone, where ka = π, the frequencies are ω22 = (2D + Ds )/M
or ω32 = (2D + Ds )/m. In the middle of the Brillouin zone one finds
 #
2D + Ds 4μ 1
2
ω± = 1± 1− 1− .
2μ M +m [1 + Ds /(2D)]2
The material is transparent to radiation in the ranges ω < ω− , ω2 < ω < ω3 , and
ω+ < ω.

Fig. 5.36
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 399

Lattice vibrations 399

c. In one dimension, the density of states is g(ω) = (L/π)/(dω/dk). Figure 5.36

shows that the density of states in both branches diverges at the edges of the
Brillouin zone. The plot portrayed in Fig. 5.37(a) is obtained by differentiating
the dispersion law, with the same parameters as used in Fig. 5.36 [the frequency
is in units of ω0 , and the density of states is in units of g0 = 2L/(ω0 a)]. The
contribution to the specific heat of each of the frequencies has the same form as
the one resulting from the Einstein model, Fig. 5.13. Figure 5.37(b) shows the
contributions of the two branches, for a mass ratio of M/m = 10; such a ratio
allows for two distinct branches. The two dotted curves display the contribution to

(a) (b)

Fig. 5.37

the specific heat of each branch separately (those are computed from a summation
over 300 frequency values, that correspond to 300 wave numbers in the first Brillouin
zone). The specific heat of each branch is exponentially small at temperatures lower
than its threshold frequency; as the temperature is raised the specific heat saturates
towards the Dulong-Petit constant (300kB in the plots). The solid line is the total
contribution. As seen, at low temperatures the specific heat is predominantly given
by the lower branch; when the temperature approaches the threshold frequency
of the upper branch, the latter begins to contribute as well, such that at high
temperatures the total specific heat reaches the Dulong-Petit value 600kB .

Answer s.5.4.
a. Denoting the displacements of the three atoms in the cell (from left to right) by
un = A exp[i(kna − ωt)], vn = B exp[i(kna − ωt)], and wn = C exp[i(kna − ωt)],
the equations of motion are

mω 2 A = D(2A − B − C exp[−ika]) ,
mω 2 B = D(2B − A − C) ,
M ω 2 A = D(2C − B − A exp[ika]) .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 400

400 Introduction to Solid-State Physics

When the determinant of the coefficients’ matrix vanishes,

⎡ ⎤
mω 2 − 2D D De−ika
det ⎣ D mω 2 − 2D D ⎦=0,
De ika
D M ω − 2D
2

the solutions are given by

mω 2 M
(x − 2)2 (ξx − 2) − (2 + ξ)x + 6 + 2 cos(ka) = 0 , x ≡ , ξ= .
D m
The three roots of the cubic equation give rise to three branches of frequencies, as
expected for a one-dimensional lattice with three atoms in the unit cell.
b. When ka = 0, one of the solutions of the cubic equation is x = 0, while the
two others follow from the quadratic equation ξx2 − 2(2ξ + 1)x + 3(ξ + 2) = 0.
The vanishing of the first solution at ka = 0 implies that it is an acoustic branch.
Indeed, an expansion of the equation to the lowest possible order in x and in ka
the origin x = mω 2 /D ≈ (ka)2 /[3(ξ +2)], and therefore
yields that at the vicinity of 
the sound velocity is c ≈ a D/[3(2m + M )]. As expected, the sound velocity is
determined by the joint motion of all three atoms in the unit cell, with the total
mass 2m + M .
c. Figure 5.38(a) displays the three roots for the case of equal masses, ξ = 1.
Exploiting the symmetry of the solution with respect to a displacement by a recip-
rocal lattice-vector, and the one with respect to reflections, shows that the three
branches represent three parts of the Brillouin zone of a single-atom lattice, whose
lattice constant is b = a/3, and its Brillouin zone extends up to π/(a/3) = 3π/a.
d. Figure 5.38(b) displays the results for ξ = 2 (obtained by an explicit solution
of the cubic equation for an arbitrary wave number). An expansion around the
origin shows that the successive branches have a sequence of minimum–maximum–
minimum at the origin. One may therefore “guess” this form, and draw it schemat-
ically. One may also derive the result starting from the plots in Fig. 5.38(a), and
then opening gaps in-between the branches at the locations where they coincide in
Fig. 5.38(a).

(a) (b)

Fig. 5.38
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 401

Lattice vibrations 401

Answer s.5.5.
a. The lattice is a rectangular centered one; its primitive unit cell, encompassed
in Fig. 5.39 by the thick-line rhombus, comprises a single atom of each type. The
lattice vectors are a1 = ax̂ − bŷ and a2 = ax̂ + bŷ, and the locations of the atoms in
(1) (2)
this unit cell are Rn,m = na1 +ma2 +unm ẑ and Rn,m = na1 +ma2 +ax̂+vn,m ẑ. The
displacements away from the equilibrium locations are marked in Fig. 5.39. The
area of the unit cell is 2ab, and the number of cells in the given area is N = A/(2ab).

Fig. 5.39

b. The classical equations of motion are

M1 ün,m = −Da (2un,m − vn,m − vn−1,m−1 ) − Db (2un,m − vn,m−1 − vn−1,m ) ,
M2 v̈n,m = −Da (2vn,m − un,m − un+1,m+1 ) − Db (2vn,m − un,m+1 − un+1,m ) .
Substituting un,m = A exp[ik · (na1 + ma2 ) − iωt] and vn,m = B exp[ik · (na1 +
ma2 ) − iωt] yields
M1 ω 2 A = Da (2A − B − Be−ik·(a1 +a2 ) ) + Db (2A − Be−ik·a1 − Be−ik·a2 ) ,
M2 ω 2 B = Da (2B − A − Aeik·(a1 +a2 ) ) + Db (2B − Aeik·a1 − Aeik·a2 ) ,
and thus
  
M1 ω 2 − 2(Da + Db ) 2e−ikx a [Da cos(kx a) + Db cos(ky b)] A
=0.
2eikx a [Da cos(kx a) + Db cos(ky b)] M2 ω 2 − 2(Da + Db ) B
A non trivial solution is obtained when the determinant vanishes,
M1 M2 ω 4 −2(Da + Db )(M1 + M2 )ω 2 + 4(Da + Db )2
− 4[Da cos(kx a) + Db cos(ky b)]2 = 0 .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 402

402 Introduction to Solid-State Physics

There are therefore two dispersion branches,

D + Db (Da + Db )2 16  
2
ω± = a ± 2
− Da sin2 (kx a/2) + Db sin2 (ky b/2)
μ μ M1 M2
  1/2
× Da cos2 (kx a/2) + Db cos2 (ky b/2) ,

where μ = M1 M2 /(M1 + M2 ) is the reduced mass. In the limits Da = 0 or Db = 0

this spectrum coincides with that of a one-dimensional lattice with two atoms in the
unit cell, Eq. (5.15), with lattice constants 2a or 2b, respectively. In the symmetric
configuration, where Da = Db and a = b, the result corresponds to that of the
table salt, Fig. 4.4. Along the diagonal, kx = ky , the expression resembles again
Eq. (5.15) (once the lattice constant and the coupling coefficients are properly
adjusted), since in this case neighboring planes contain atoms of the same type; the
system becomes similar to a one-dimensional lattice.
c. In the long-wave limit, the acoustic branch is
2
ω− ≈ (cx kx )2 + (cy ky )2 ,
c2x = 2μDa a2 /(M1 + M2 ) , c2y = 2μDb b2 /(M1 + M2 ) .
cx and cy are the sound velocities along each of the axes. It follows that the equal-
frequency curves are ellipses,
 k 2  k 2
x y
+ =1.
ω/cx ω/cy
The area of such an ellipse is π(ω/cx )(ω/cy ). The area of the unit cell in the re-
ciprocal lattice (which equals the area of the first Brillouin zone) is (2π)2 /(2ab)
(see problem 3.6), and it contains N points; therefore the area per site is
2π 2 /(N ab) = 4π 2 /A. Consequently, the total number of points within the ellipse is
[πω 2 /(cx cy )]/[4π 2 /A] = Aω 2 /(4πcx cy ), and the density of states is
d[Aω 2 /(4πcx cy )] Aω
g(ω) = = .
dω 2πcx cy
d. The Debye approximation assumes that the acoustic density of states, obtained
in the previous part, is valid up to the Debye frequency. The total number of
acoustic vibrational modes is
 ωD 2
A ωD
N= = g(ω)dω = A ,
2ab 0 4πcx cy
2
and hence ωD = 2πcx cy /(ab). The contribution of the acoustic modes to the specific
heat is
 ωD 
2 eβω T 2 ΘD /T 2 ex
CV = kB dωg(ω)(βω) βω = 2N k dxx ,
0 (e − 1)2 B 2
ΘD 0 (ex − 1)2
and at low temperatures, CV ≈ 2N kB I2 (T /ΘD )2 , where I2 is defined after Eq.
(5.69).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 403

Lattice vibrations 403


e. The frequencies in the optical branch are of the order of ω+ ≈ 2(Da + Db )/μ.
In the Einstein approximation, their contribution to the specific heat is
CV ≈ N kB (βω+ )2 exp[βω+ ]/(exp[βω+ ] − 1)2 ,
yielding at low temperatures CV  N kB (βω+ )2 exp[−βω+ ]. This contribution is
exponentially smaller than that of the acoustic modes.
f. At sufficiently high temperatures the specific heat obeys the Dulong-Petit law,
CV ≈ 2N kB .

Answer s.5.6.
a. Each atom in the FCC lattice has 12 nearest neighbors, at the points
(±1, ±1, 0)a/2, (±1, 0, ±1)a/2, and (0, ±1, ±1)a/2. Inserting these into the ex-
pression for the spring constant yields
Dzz [(±1, ±1, 0)/2] = D1 , Dxx [(±1, ±1, 0)/2] = Dyy [(±1, ±1, 0)/2] = (D1 + D2 )/2 ,
Dyy [(±1, 0, ±1)/2] = D1 , Dxx [(±1, 0, ±1)/2] = Dzz [(±1, 0, ±1)/2] = (D1 + D2 )/2 ,
Dxx [(0, ±1, ±1)/2] = D1 , Dyy [(0, ±1, ±1)/2] = Dzz [(0, ±1, ±1)/2] = (D1 + D2 )/2 ,
Dxy [±(1, 1, 0)/2] = −Dxy [±(1, −1, 0)/2] = (D2 − D1 )/2 ,
Dyz [±(0, 1, 1)/2] = −Dyz [±(0, 1, −1)/2] = (D2 − D1 )/2 ,
Dxz [±(1, 0, 1)/2] = −Dxz [±(1, 0, −1)/2] = (D2 − D1 )/2 .
All other non diagonal spring constants vanish. (The lattice constants are omitted
on the left hand-sides for brevity.) Equation (5.25) then yields
 xx = (D1 + D2 ){4 − cos[(kx + ky )a/2] − cos[(kx − ky )a/2] − cos[(kx + kz )a/2]
K
− cos[(kx − kz )/2]} + 2D1 {2 − cos[(ky + kz )a/2] − cos[(ky − kz )a/2]} ,
 yy = (D1 + D2 ){4 − cos[(kx + ky )a/2] − cos[(kx − ky )a/2] − cos[(ky + kz )a/2]
K
− cos[(ky − kz )/2]} + 2D1 {2 − cos[(kx + kz )a/2] − cos[(kx − kz )a/2]} ,
 zz = (D1 + D2 ){4 − cos[(kx + kz )a/2] − cos[(kx − kz )a/2] − cos[(ky + kz )a/2]
K
− cos[(ky − kz )/2]} + 2D1 {2 − cos[(ky + kx )a/2] − cos[(ky − kx )a/2]} ,
 xy = (D2 − D1 ){− cos[(kx + ky )a/2] + cos[(kx − ky )a/2]} ,
K
 xz = (D2 − D1 ){− cos[(kx + kz )a/2] + cos[(kx − kz )a/2]} ,
K
 yz = (D2 − D1 ){− cos[(kz + ky )a/2] + cos[(kz − ky )a/2]} .
K
The values of M ω 2 for the eigenfrequencies are derived from a diagonalization of
this matrix.
b. For k = (k, 0, 0) the matrix is diagonal,
 xx = 8(D1 + D2 ) sin2 (ka/4) , K
K  yy = K
 zz = 4(3D1 + D2 ) sin2 (ka/4) .

As the wave propagates along the x̂−direction, the longitudinal motion has the
2
frequency ωL =K xx (k00)/M , while the frequency of the two transverse modes is
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 404

404 Introduction to Solid-State Physics

 yy (k00)/M . The path Γ → X in the Brillouin zone is given by 0 < k < 2π/a.
ωT2 = K
The matrix is again diagonal for k = (2π/a, k, 0),
 xx = 4(D1 + D2 )[2 − sin2 (ka/4)] + 8D1 sin2 (ka/4) ,
K
 yy = 4(3D1 + D2 ) ,
K
 zz = 4(D1 + D2 )[1 + sin2 (ka/4)] + 8D1 cos2 (ka/4) .
K

Hence, there are three different eigenfrequencies, ωμ2 = K  μμ /M . The path X →

W → X is given by 0 < k < 2π/a. The end point of this path, (2π/a, 2π/a, 0) is
marked in Fig. 5.18 by X, much as the starting point, (2π/a, 0, 0). Indeed, when the
point (2π/a, 2π/a, 0) is displaced by the reciprocal-lattice vector (2π/a, 2π/a, 2π/a),
it becomes the point (0, 0, −2π/a), which is equivalent to the point (2π/a, 0, 0), but
along the ẑ−direction. Recall that around the origin of the first Brillouin zone
there are six points of the type X, along all axes. In fact, Fig. 5.18 and Fig. 5.40
demonstrate that the dispersion relation on the path X → W → X is symmetric
under a reflection through the point W , since such an operation transforms the
Brillouin zone back to itself. For k = (k, k, 0) one finds

Fig. 5.40

 xy = 2(D2 − D1 ) sin2 (ka/2) ,

K  xz = K
K  yz = 0 ,
 xx = K
K  yy = 4(3D1 + D2 ) sin2 (ka/4) + 2(D1 + D2 ) sin2 (ka/2) ,
 zz = 8(D1 + D2 ) sin2 (ka/4) + 4D1 sin2 (ka/2) ,
K

and therefore the frequencies are M ω± 2

= K xx ± K xy and M ω 2 = K  zz . The
T2
corresponding eigenvectors are (1, ±1, 0) and (0, 0, 1), which can be identified as
belonging to a longitudinal vibration (parallel to the wave vector, with ω+ ), and
two transverse modes, with ωT 1 = ω− and ωT 2 . The path X → Γ pertains to
0 < k < 2π/a. Finally, for k = (k, k, k),

K  yy = K
 xx = K  zz = 4(2D1 + D2 ) sin2 (ka/2) ,
 xy = K
K  yz = K
 xz = 2(D2 − D1 ) sin2 (ka/2) ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 405

Lattice vibrations 405

which constitutes a 3 × 3 matrix,

⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
K  xy K
 xx K  xy 1 1 1
⎢   xy ⎥
 xx K
⎣ Kxy K ⎦ , whose eigenvectors are ⎣ 1 ⎦ , ⎣ −1 ⎦ , ⎣ 1 ⎦ ,
 xy K
K  xy K
 xx 1 0 −2

(the eigenvectors are not normalized). The corresponding eigenvalues are K  xx +

2K xy (for the left vector) and K xx − K yy (for the other two). The first is the
longitudinal frequency, as its eigenvector is parallel to the wave vector; the other
two are transverse, as their vectors are perpendicular to the first one. In the latter
case, the path Γ → L corresponds to 0 < k < π/a.
 At the point X, the expressions found in part (b) yield ωL /ωT =
c.
2(D1√+ D2 )/(3D1 + D2 ). This ratio, as judged from Fig. 5.18, is approximately
1.4 ≈ 2; it implies that D2  D1 . Figure 5.40 displays the results of part (b),
for D2 = 100D1 . The plots are quite similar to those portrayed in Fig. 5.18. This
indicates that accounting only for nearest-neighbor interactions, with forces which
are predominantly longitudinal, is quite a reliable approximation.
d. The left part of Fig. 5.18 shows that the slopes are cL ≈ 4.4 × 105 cm/sec,
and cT ≈ 2.8 × 105 cm/sec (note that the figure refers to frequencies and not to
angular frequencies). It follows that c ≈ 2.9 × 105 cm/sec. Equation (5.71) then
yields ωD = (24π 2 /a3 )1/3 c ≈ 5 × 1013 sec−1 , where V = a3 /4. Consequently,
ΘD = ωD /kB ≈ 360 K.
e. The acoustic branches of table salt are qualitatively similar to those calculated
above, as they pertain to the joint vibrations of the two ions in the unit cell. Be-
cause the unit cell comprises two atoms, there are also three optical branches with
higher frequencies, which describe motions within the unit cell. For the acoustic
branches, and for specific directions in the Brillouin zone, the two transverse modes
are degenerate. The degeneracies are implied by the symmetries with respect to
rotations around the longitudinal direction (and do not result from the approxima-
tions). One therefore expects two optical frequencies for the paths Γ → X (due to
the symmetry with respect to rotations around the edge of the cube) and Γ → L
(due to symmetries with respect to rotations around the cube’s diagonal), and three
optical frequencies along other directions.

Answer s.5.7.
a. The van Hove singularities appear when the group velocity vanishes; this happens
when the frequency is maximal, minimal, or at a saddle point, as a function of the
wave numbers. Indeed, a vanishing slope of the frequency can be detected in Fig.
5.18 at points X (two frequencies), W (a single frequency), and L (two frequencies),
and also at a point in-between X and Γ. The values of these frequencies agree with
those at which the derivatives in Fig. 5.19 are discontinuous.
b. The equal-frequency surfaces are spheres centered around the point k0 in the
Brillouin zone, ω = ω0 − ω1 |k − k0 |2 . The group velocity is ∇k ω = −2ω1 (k − k0 ).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 406

406 Introduction to Solid-State Physics

For a sphere of radius q = |k − k0 | = [(ω0 − ω)/ω1 ]1/2 , Eq. (5.65) yields


NV dS N V 4πq 2 NV
g(ω) = = = (ω0 − ω)1/2 , ω < ω0 ,
(2π) S |∇k ω|
3 3
(2π) 2ω1 q 4π 2 ω
3/2
1
g(ω) = 0 , else .
The density of states of this branch tends continuously to zero as ω approaches ω0
from below, but the derivative diverges.
c. At the vicinity of point K in the Brillouin zone of graphene ω = ω0 ±ω1 |q|, where
q = k − kK . The group velocity is ∇k ω = ±ω1 q/q, and therefore |∇k ω| = ω1 . For
this two-dimensional case, Eq. (5.65) gives

2N V dS 2N V 2πq NV
g(ω) = = = 2 |ω0 − ω| ,
(2π)2 S |∇k ω| (2π)2 ω1 ω1
where the factor of 2 reflects the existence of two points of the type K in each
Brillouin zone. The density of states vanishes at ω = ω0 , and increases linearly as
the frequency moves away from it along the two directions. The derivative flips its
sign at the singular point.

Answer s.5.8. 
 that ωE = 2D/μ. Equation (5.16) yields that
a. Using Eq. (5.17), one finds
the sound velocity is c = a D/[2(M + m)]. The density of states in the Debye
approximation is
g(ω) = L/(πc) = N a/(πc) .
Hence,
 ωD
N= g(ω)dω = N aωD /(πc) ,
0

i.e., ωD = πc/a = π D/[2(M + m)]. It follows that

ωD /ωE = (π/2) m/M /(1 + m/M ) .
The ratio of the frequencies depends only on the masses’ ratio. It increases from 0
to π/4 when this ratio varies between 0 and 1.
b. Figure 5.41 illustrates the approximate density of states: a constant density
between 0 and ωD , and a delta function at ωE .

Answer s.5.9.
The wave length corresponding to the Debye frequency is obtained from the re-
lation ωD = ckD = 2πc/λD . Equation (5.71) implies that V = 6π 2 (c/ωD )3 =
6π 2 [λD /(2π)]3 , and therefore λD = (4πV /3)1/3 . The volume of the unit cell in an
FCC lattice with a lattice constant
√ a is V = a3 /4, and the distance√between nearest-
neighbor atoms is d = a/ 2. Consequently, (λD /d)|FCC = (2π 2/3)1/3 ≈ 1.44.
The volume of the unit cell of a BCC lattice is V = a3 /2, and the distance between
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 407

Lattice vibrations 407

Fig. 5.41
√ √
nearest-neighbor atoms is d = a 3/2. Thus, (λD /d)|BCC = [16π/(9 3)]1/3 ≈ 1.48.
In both cases the wave length is longer than the nearest-neighbor distance, and
therefore the description of the vibrations in terms of waves is plausible. Such a
description is meaningless once the wave length is shorter than the distance between
nearest-neighbor atoms.

Answer s.5.10.
The temperatures at which the specific heat is measured are much lower than the
Debye temperature, and therefore Eq. (5.69), with the integral there replaced by I3 ,
suffices for the calculation of the specific heat. These two crystals possess two atoms
in the unit cell, and consequently they have also optical branches. However, the
specific heat contributed by the latter is exponentially small and may be ignored.
Assuming that the sound velocities are the same along all directions, one finds CV ≈
C0 (βωD )−3 , where the coefficient C0 is the same for both crystals. Using ωD =
kB ΘD yields CV ≈ C0 (ΘD /T )3 , and consequently C0 = CV (KCl,5K)(5/230)3 . It
follows that
 310 5 3
CV (NaCl, 2K) = CV (KCl, 5K) = 1.45 J/mol/K .
2 230

Answer s.5.11. √
The equal-frequency lines, at√frequency ω, are squares of edge 2ω/c, and hence
their circumference is S = 4 2ω/c. The group velocity (which is normal to √ the
equal-frequency curves) is vg = ∇k ω = c(sgn[kx ], sgn[ky ]) and thus |vg | = c 2.
Equation (5.65) yields
√
N a2 4 2ω/c N a2 ω
g(ω) = √ = .
(2π)2 c 2 (2π)2
As the density of states is linear in the frequency, all thermodynamic properties
calculated with it are identical to those obtained in the Debye approximation at
two dimensions, except for a multiplicative factor that determines the scale. A
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 408

408 Introduction to Solid-State Physics

calculation
√ of the total number of vibrational modes yields the Debye frequency,
ωD = πc 2/a. Each acoustic branch then yields
 T 2  ΘD /T ex
CV = 2kB N dxx3 x , kB ΘD = ωD .
ΘD 0 (e − 1)2
This result is identical to the one derived from the Debye model in two dimensions.

Answer s.5.12.
The wave vectors of the√ neutrons are k = (2π/λ)(1, 0, 0) = (1.795(Å)−1 , 0, 0) and
k = (2π/λ )(1, 1, 1)/ 3 = (1.5571, 1.557, 1.557)(Å)−1 . For neutrons, 2 /(2mn ) =
1.965×10−3 eV (Å)2 , which implies that the energy change of the scattered neutron
is Ω − Ω = 2 (k 2 − k 2 )/(2mn ) = −7.96 × 10−3 eV. This is a negative change, i.e.,
the neutron emits a phonon. The momentum the neutron loses is K − G = k − k =
(0.238, −1.557, −1.557)(Å)−1 . The boundaries of the first Brillouin zone along each
of the axes are |Kα | < π/a = 0.739(Å)−1 . One therefore concludes that the y− and
z−components of the momentum exchange vector are outside the first Brillouin
zone. Hence, the reciprocal-lattice vector G = (0, 2π/a, 2π/a) is used, and then
K = k − k = (0.238, −0.079, −0.079)(Å)−1 .

Answer s.5.13.
The magnetic moments in the ordered state (i.e., the ground state) are denoted
μ(1) = μ1 ẑ and μ(2) = −μ2 ẑ. The equations of motion given in problem 5.29 are
then
(1)
dμnx (2)
= −J[2μ2 μ(1) (2)
ny + μ1 (μny + μn−1y )] ,
dt
(1)
dμny (2)
= J[2μ2 μ(1) (2)
nx + μ1 (μnx + μn−1x )] ,
dt
(2)
dμnx (1)
= J[2μ1 μ(2) (1)
ny + μ2 (μny + μn+1y )] ,
dt
(2)
dμny (1)
= −J[2μ1 μ(2) (1)
nx + μ2 (μnx + μn+1x )] .
dt
(1,2) (1,2) (1,2)
Introducing the variables μn± = μnx ± iμny changes these equations to be
(1)
dμn+ (1) (2) (2)
= iJ[2μ2 μn+ + μ1 (μn+ + μn−1+ )] ,
dt
(2)
dμn+ (2) (1) (1)
= −iJ[2μ1 μn+ + μ2 (μn+ + μn+1+ )] .
dt
Finally, with the wavy-solution form of spin waves, one finds
−ωμ+ = J[2μ2 μ+ + μ1 (1 + e−ika )μ+ ] ,
(1) (1) (2)

(2) (2) (1)

ωμ+ = J[2μ1 μ+ + μ2 (1 + eika )μ+ ] .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 409

Lattice vibrations 409

The determinant vanishes for


ω = J[(μ1 − μ2 ) ± (μ1 − μ2 )2 + 4μ1 μ2 sin2 (ka/2)] ,

(a is the lattice constant). The two solutions for the frequency, in units of μ1 J,
are presented in Fig. 5.42 for μ2 = 2μ1 . One of the modes is similar to that
found for a ferromagnet, with a dispersion relation quadratic in the wave number
(at long waves). The other mode is “optical”. The sign does not matter, as it just
determines the direction of the propagation. When the two moments are equal, the
linear dispersion law of long waves, as for an antiferromagnet, is reproduced.

Fig. 5.42
December 12, 2014 15:40 BC: 9334 – Bose, Spin and Fermi Systems 9334-book9x6 page 2

This page intentionally left blank

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 411

Appendix D

The continuum limit

Equation (5.8) shows that the equations of motion approach the continuum limit
when the lattice constant is significantly shorter than the wave length of the acous-
tic wave. The theory of elasticity yields these equations of motion, irrespective of
the lattice structure of the material. Here are listed certain basic concepts of elas-
ticity theory, from which are derived the equations governing acoustic waves in the
continuum.
Consider first the one-dimensional system, lying on the x̂−axis. Denoting the
longitudinal displacement of a point x by ux (x), the strain is defined by εxx =
∂ux /∂x. According to Hooke’s law, the stress, σxx (that is, the force per unit area
normal to the x̂−axis) is linear in the strain, i.e., σxx = Y εxx = Y (∂ux / ∂x), where
Y is a material constant called the Young modulus. For a cylinder lying along
the x̂−axis, whose height is dx and whose base area is A, Newton’s second law takes
the form
∂ 2 ux (x, t) ∂σ ∂ 2 ux (x, t)
(ρAdx) 2
= A[σxx (x + dx) − σxx (x)] = A xx dx = Y A dx ,
∂t ∂x ∂x2
where ρ is the mass density. The result is the wave equation,
∂ 2 ux (x, t)  Y  ∂ 2 ux (x, t)
= .
∂t2 ρ ∂x2
A monochromatic wave, ux ∝ exp[i(kx − ωt)], of wave vector k and frequency ω,
has a linear dispersion law, ω = ck, with the sound velocity c = Y /ρ.
In three dimensions, the displacement vector u has three components, and the
strain tensor is defined by
1  ∂uμ ∂uν 
εμν = + .
2 ∂xν ∂xμ
The stress is a tensor as well, σμν (i.e., the ν−component of the force on a unit
area normal to the μ̂−axis), and Hooke’s law becomes σμν = α,β Cμν,αβ εαβ . The
tensor Cμν,αβ contains the elastic constants (or elastic moduli) of the material. By
definition, the strain tensor is symmetric, εμν = ενμ . When no moment is exerted
on the material so that its angular velocity is zero, the stress tensor is symmetric

411
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 412

412 Introduction to Solid-State Physics

as well, σμν = σνμ . It follows that the tensor comprises six independent quantities,
conventionally denoted by
e1 = εxx , e2 = εyy , e3 = εzz , e4 = 2εyz , e5 = 2εzx , e6 = 2εxy ,
(at times, these are defined without the factor 2 in the last three). Similarly,
the elastic constants’ tensor, Cij ≡ Cμν,αβ is of order 6 × 6 and contains 21
independent quantities (check!). The free energy connected with the elasticity of
the material is given by
  6
1
F = d3 r ei Cij ej . (D.1)
2 i,j=1

The elastic tensor determines the coefficients which relate the response of the mate-
rial to various forces. All 21 components of the triclinic lattice differ from each other.
The number of non-zero coefficients of more symmetric crystals is smaller. For in-
stance, in cubic symmetry there are four rotation axes of order 3 (around the cube’s
diagonals). Rotations by 120◦ around these axes give rise to the transformations
x → y → z → x, x → z → −y → x, −x → z → −y → −x, −x → y → z → −x.
Because of these mappings, the free energy (D.1) is invariant under the transforma-
tions εxy → εyz → εz(−x) → εx(−y) = −εxy . As a result, the free energy contains
only three independent coefficients,
C12 = C23 = C31 , C44 = C55 = C66 , C11 = C22 = C33 , (D.2)
and all other coefficients vanish, e.g., C45 = C56 = C64 = 0 and C14 = C15 = C16 =
0. By similar considerations one finds that the monoclinic lattice has 13 elastic
constants, the orthorhombic one has 9, 6 or 7 coefficients describe the tetragonal
and rhombohedral lattices, and 5 are needed for the hexagonal lattice.
The three-dimensional wave equation in cubic symmetry is obtained in a similar
way. For a cube whose edges are dx, dy, and dz, along the directions of the three
axes, the force acting on this cube along the x̂−direction is
∂ 2 ux (r, t) ∂σxx ∂σyx ∂σzx
(ρdxdydz) 2
= + + dxdydz .
∂t ∂x ∂y ∂z
In cubic symmetry, Hooke’s law yields
∂ux  ∂u ∂uz 
y
σxx = C11 εxx + C12 (εyy + εzz ) = C11 + C12 + ,
∂x ∂y ∂z
 ∂u ∂uy   ∂u ∂uz 
x x
σxy = 2C44 εxy = C44 + , σxz = 2C44 εxz = C44 + .
∂y ∂x ∂z ∂x
Collecting all terms, one obtains the wave equation in the form
∂ 2 ux (r, t) ∂u2x (r, t)  ∂u2 (r, t) ∂u2 (r, t) 
x x
ρ = C 11 + C 12 +
∂t2 ∂x2 ∂y∂x ∂x∂z
 ∂u2 (r, t) ∂u2 (r, t) ∂u2 (r, t) ∂u2 (r, t) 
x y x z
+ C44 + + + ,
∂y 2 ∂y∂x ∂z 2 ∂z∂x
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 413

The continuum limit 413

and two more equations for the forces along the ŷ and ẑ directions. Inserting
the plane-wave function, u = A exp[i(k · r − ωt)], results in three coupled linear
equations,
ρω 2 Ax = [C11 kx2 + C44 (ky2 + kz2 )]Ax + (C12 + C44 )(kx ky Ay + kz kx Az ) ,
ρω 2 Ay = [C11 ky2 + C44 (kx2 + kz2 )]Ay + (C12 + C44 )(kx ky Ax + kz ky Az ) ,
ρω 2 Az = [C11 kz2 + C44 (ky2 + kx2 )]Az + (C12 + C44 )(kx kz Ax + ky kx Ay ) .
For instance, for a wave
 along the x̂−direction, there is a longitudinal wave with
the dispersion
 ω L = C11 /ρkx , and two transverse waves with the dispersion law
ωT = C44 /ρkx . In the more general case ρω 2 Aμ = σ,ν,τ Cμσ,ντ kσ kν Aτ .

Problem D.1.
a. Derive the dispersion relations for acoustic waves in cubic symmetry, for a plane
wave along an arbitrary direction in the XY −plane.
b. Repeat part (a) for a plane wave along (111).

D.1 Answer

Answer D.1.
a. When kz = 0, the wave equations are
ρω 2 Ax = [C11 kx2 + C44 ky ]Ax + (C12 + C44 )kx ky Ay ,
ρω 2 Ay = [C11 ky2 + C44 kx ]Ay + (C12 + C44 )kx ky Ax ,
ρω 2 Az = C44 [kx2 + ky2 ]Az .
Hence, 
there are transverse vibrations normal to the plane, whose dispersion is
ωT 2 = C44 /ρ|k|. To obtain the in-plane vibrations one has to solve the first two
equations; the requirement that the determinant vanishes gives
[C11 kx2 + C44 ky2 − ρω 2 ][C11 ky2 + C44 kx2 − ρω 2 ] − (C12 + C44 )2 kx2 ky2 = 0 .
It follows that

1 1
ρω = (C11 + C44 )k ±
2 2
(C − C44 )2 k4 + [C12 2 − C 2 + 2C (C 2 2
11 + C12 )]kx ky .
2 4 11 11 44

Generally, there is no special relation between the eigenvectors and the direction
of k. Such a relation pertains to symmetry directions, like (100) or (110). For a
2
wave along the latter direction, ρωL = (C11 + C12 + 2C44 )k2 /2, with a longitudinal
vibration along (110), or ρωT2 1 = (C11 − C12 )k2 /2, with a transverse motion normal
to that direction. √
b. For a wave along (111), i.e., kx = ky = kz = |k|/ 3, the wave equations are
3ρω 2 Ax = [(C11 + 2C44 )Ax + (C12 + C44 )(Ay + Az )]k2 ,
3ρω 2 Ay = [(C11 + 2C44 )Ay + (C12 + C44 )(Ax + Az )]k2 ,
3ρω 2 Az = [(C11 + 2C44 )Az + (C12 + C44 )(Ay + Ax )]k2 .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 414

414 Introduction to Solid-State Physics

2
The eigenfrequencies are ρωL = (C11 + 2C12 + 4C44 )k2 /3, corresponding to a lon-
gitudinal motion along (111), and two transverse modes with ρωT2 = (C11 − C12 +
C44 )k2 /3. Problem 5.9 compares these equations with those of the acoustic branch
of the square lattice.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 415

Chapter 6

Electrons in solids

This chapter discusses the thermal and electric properties of electrons in solids.
Certain phenomena can be explained in terms of the classical theory of Drude, while
others are analyzed within the framework of the quantum theory of free electrons,
as formulated by Sommerfeld. The main part of the chapter focuses on phenomena
that result from, or that are related to, the periodic potential of the lattice of ions
in which the electrons move. This potential gives rise to bands of energies accessible
for the electrons, that in turn determine the electronic properties of solids, and the
classification of those into metals, insulators and semiconductors. The fascinating
quantum behavior of electrons subjected to magnetic fields, e.g., the quantum Hall
effect, is discussed towards the end of the chapter.

6.1 Preface

Metals are characterized by their high electrical conductance. The electric cur-
rent is carried by the electrons, that move relatively freely through the solid. Each
atom in the metal “releases” its valence electrons (usually one or two electrons per
atom), generally from the outer “shells” of the free-atom electronic states. These
electrons are termed “conduction electrons”. In the classical picture of Drude
(called below the “Drude theory”) these electrons are treated as particles whose
motion is governed by Newton’s laws, until they collide with a certain defect. (In
the original formulation of the theory it was assumed that the collisions are with the
ions forming the metal, but in fact other types of collisions dominate the conduc-
tance.) The collisions change the direction of the electrons’ propagation, and may
slow down the flow of electrons along the electric field. This is the physical origin of
the electrical resistance (the inverse of the conductance), R, and the explanation of
Ohm’s law, which relates the electrical current, I, to the voltage difference across
the sample, V : I = V /R.
In the Drude theory the electrons are treated as classical particles, and their
quantum nature is ignored. As found in Sec. 6.2, even this naı̈ve picture captures
many of the phenomena observed in experiments. However, there exist phenomena

415
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 416

416 Introduction to Solid-State Physics

whose origin is in the quantum aspect of the electrons and in the Pauli principle.
These are included in the Sommerfeld theory, described in Sec. 6.3. Both the
Sommerfeld and the Drude theories ignore the periodic potential of the crystal
in which the electrons reside, as well as the Coulomb interactions among them.
Though there are properties that cannot be explained without invoking the periodic
potential, surprisingly enough, many other phenomena can be described in terms
of free electrons.
The central theorem that dictates the form of the solutions of the Schrödinger
equation with a periodic potential is the Bloch theorem. It is presented and
proven in Sec. 6.4. The next three sections contain examples of calculations of
the wave functions: the exact solution of the Kronig and Penney potential in
Sec. 6.5, the nearly-free electron approximation in Sec. 6.6, and the tight-
bindingapproximation in Sec. 6.7. These calculations reveal the existence of
energy bands, (comprising energies which the electrons can assume) with energy
gaps in between them, of forbidden energies. The mere existence of such bands
explains the disparities between insulators, conductors, and semiconductors, as ex-
plained in Sec. 6.8. The electronic density of states is analyzed in Sec. 6.9,
together with the occupations of the various bands and the specific heat, and Sec.
6.10 discusses the equations of motion of an electron moving in a periodic potential;
these enable, e.g., the calculation of the charge conductance. Section 6.11 explains
the motion of an electron in a magnetic field and the quantum Hall effect. Section
6.12 surveys the Anderson and the Mott insulators.

6.2 The Drude theory

As mentioned, the Drude model (that preceded the appearance of quantum mechan-
ics) disregards the Coulomb interactions among the electrons. Rather, the electrons
are viewed as a gas of free particles. The electron gas is described by the kinetic
theory of gases, the same as the one exploited for gases of freely-moving molecules.
Time in-between collisions (the relaxation time) and the electrons’
velocity. The kinetic theory presumes that the probability per unit time for the
electron to collide with a defect is given by 1/τ . Problem 6.1 demonstrates that
τ is in fact the mean time in-between collisions. It is also shown below that
the average velocity of the electron decays during a time interval τ to a stationary
value (that is independent of time), rendering τ to be also the “relaxation time”.
Both terms are used below. Drude conjectured that the electrons collide with the
ions situated at the lattice sites. As the ions are far heavier than the electrons, he
assumed that the collisions are instantaneous and elastic, such that the direction of
the electron’s velocity is changed randomly, but not its magnitude. In reality the
collisions that lead to relaxation are not with the ions in the lattice. Nonetheless, the
existence of defects that scatter elastically the electrons suffices for the derivation
of Ohm’s law (without specifying the nature of those defaults).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 417

Electrons in solids 417

Problem 6.1.
Assuming that the probability per unit time for a collision between the electron and
a certain defect is 1/τ , prove that:
a. the probability that the electron will not collide during a time interval t is P (t) =
exp[−t/τ ];
b. the mean time in-between collisions is τ .

In-between collisions, the motion of the electron (whose mass is m and whose
velocity is v) is determined by Newton’s second law, ṗ = mv̇ = ma = F, where
p = mv is the momentum of the electron and F is the force that acts on it. In
case the electron at time t is subjected to an electric field E(t), then F = −eE,
where −e is the electron’s charge. It follows that during a very short time interval
dt p(t + dt) = p(t) − eE(t)dt. As the collisions the electron performs are random,
and since eventually the result is summed over many electrons, it follows that the
momentum is averaged over all events that can occur. The probability for the
electron to undergo a collision during the time interval dt and to lose the velocity
along the direction of propagation is dt/τ . The electron will then attain a new
random velocity v , or a new random momentum p = mv . On the other hand, it
may not collide at all and then its momentum is not changed. The probability for
the latter scenario is 1 − dt/τ . The average over all possible events during dt yields
the new momentum, p(t + dt) = (1 − dt/τ )[p(t) − eEdt] + p dt/τ , and thus

d p(t)
p(t) = − − eE(t) . (6.1)
dt τ
(It is assumed that the random velocity vanishes upon averaging, v  = 0.) Equa-
tion (6.1) resembles the equation of motion of a particle moving in a viscous fluid,
with 1/τ playing the role of the friction coefficient (in appropriate units). The
solution of Eq. (6.1), for an electric field that does not vary in time, is

p(t) = −eEτ + Ae−t/τ , (6.2)

where A = p(0) + eEτ is determined by the initial conditions. At times longer

than the relaxation time, t  τ , the second term on the right hand-side decays and
becomes exponentially small; the solution is then stationary (independent of time),

v(t) = p(t)/m ⇒ −eEτ /m = −μE , μ = eτ /m , (6.3)

where μ is the electronic mobility. The relaxation time is in fact the time interval
during which there still remains a certain memory of the initial conditions. Once
time longer than the relaxation has elapsed, t  τ , the average velocity is stationary,
with its magnitude determined by the electric field.
Another point of view on Eq. (6.3) is the following. In-between collisions the
electron moves under the effect of the electric field with a constant acceleration, and
therefore its velocity is v(t) = v − eEt/m, where v ≡ v(0) is the initial velocity
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 418

418 Introduction to Solid-State Physics

(i.e., the one it had immediately after the previous collision). This expression is valid
on the average, at times shorter than τ ; after a time τ has elapsed, the electron
collides again and attains a new velocity. Upon averaging over all possible collisions,
the average v vanishes, as after each scattering event the velocity is directed at
random. It follows that since the typical time between collisions is τ , the average
velocity is v(t) = v  − eEt/m = −eEτ /m, as in Eq. (6.3).

Problem 6.2.
So far, it is assumed that after each collision the electron loses its original velocity,
and attains a new one, v , at a completely random direction. Find the velocity of
the electron in a stationary state when v is not randomly oriented, but is changed
at most within an angle β.

Ohm’s law. Consider a gas of electrons whose density (the number of electrons
per unit volume) is n, and where each electron is moving with the average velocity
v. The number of electrons that pass through an area A normal to this average
velocity during a unit time is nA|v| (this is the number of electrons located within
a volume whose base area is A and whose height is the velocity). The charge
current-density, j (a vector), is defined as the amount of charge that is transported
through a unit area orthogonal to j, per unit time, i.e.,
ne
j = −nev = − p . (6.4)
m
Using here Eq. (6.3) yields
1
j = ne2 τ E/m = σE = E , (6.5)
ρ
where
σ = 1/ρ = ne2 τ /m = neμ (6.6)
is the specific conductivity of the material, while ρ is the corresponding specific
resistivity. Once a voltage drop of magnitude V is applied across a sample whose
length, along the direction of the current, is L and the cross section normal to it is
A, then the electric field is E = V /L. Denoting the total current by I, the current
per unit area is j = I/A. One then finds Ohm’s law, V = IR, with the resistance
R = Lρ/A = L/(σA) . (6.7)
Equation (6.6) relates the macroscopic specific conductivity of the material to the
microscopic processes of electrons’ scattering in the sample. Equation (6.5) repre-
sents the local Ohms’s law.
Characteristic values of the specific resistivity of metals are of the of the or-
der of micro-Ohm cm, ρ ≈ 1 μΩ cm. Typical electrons’ densities are those of
the ions’ densities, i.e., n ≈ 1023 cm−3 = 1029 m−3 . Using m = 9.1 × 10−31 Kg
and e = 1.6 × 10−19 Coul leads to τ ≈ 10−14 sec. Drude assumed that the av-
erage velocity of a particle in a gas can be estimated from the equipartition law,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 419

Electrons in solids 419

mv2 /2 = 3kB T /2. At room temperature, T = 300K, this relation yields |v| ≈ 107
cm/sec. It follows that the mean free-path of an electron, i.e., the mean dis-
tance it traverses in-between successive collisions, is = |v|τ ≈ 10Å. This is a
rather short distance, which fits Drude’s conjecture that the electrons collide with
the ions, and consequently the length of their mean free-path is of the order of the
lattice constant. However, (see Sec. 6.3), the estimate of the velocity magnitude
given above is not correct, and in fact the mean free-path is considerably longer.
Moreover, using the velocity derived √ from the equipartition law implies that the
mean free-path is proportional to T , whereas the lattice constant is almost inde-
pendent of the temperature. Nonetheless, the wrong estimate of the velocity is the
only fault in the argument above; the assumption that there exists a relaxation time
that determines the electrical resistance is benign. Once its correct value is used in
Eq. (6.6), the result is a faithful description of the resistances of many metals. This
topic is resumed below; for the time-being, the Drude picture is used to address
three more phenomena.
The Hall effect. Adding a constant magnetic field (that does not vary neither
in time nor in space) leads to the Hall effect, discovered by Hall in 1879. Such a
field, when directed normal to the electric field, gives rise to a voltage drop normal
to its direction and to that of the electric current. The measurement of this voltage
enables the determination of the sign of the charge carriers, and also their density.
The experimental setup is displayed in Fig. 6.1, which illustrates the situation
where two fields affect the electron’s motion: an electric field E, and a magnetic
one, B, with B ⊥ E; both fields are constant in time and in space. The classical
equation of motion, obeyed by the electron in-between collisions, is dominated by
the Lorentz force, ṗ = −e(E + v × B/c) where p = mv and c is the speed of light
(this is the Lorentz force in CGS units; in the SI ones the speed of light does not
appear). The constant magnetic field does not affect the velocity component parallel
to it. When one adds the probability for collisions, and averages over all possible
scattering events that occur within a unit time, one obtains a generalization of Eq.
(6.1),
p(t)  p(t) 
ṗ(t) = − −e E + ×B . (6.8)
τ mc
When the magnetic field is along the ẑ−axis and the electric field and the velocity
are in the XY plane, the equations of motion for each component of the momentum
are
ṗx  = −eEx − ωc py  − px /τ ,
ṗy  = −eEy + ωc px  − py /τ .
At a stationary state these equations become [see Eq. (6.2)],
− eEx − ωc py  − px /τ = 0 ,
− eEy + ωc px  − py /τ = 0 , (6.9)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 420

420 Introduction to Solid-State Physics

where ωc = eB/(mc) is the “cyclotron frequency” of the electron. (This is the

angular frequency of the circular motion in the plane normal to the magnetic field
in the absence of an electric field, and when collisions are ignored, see problem 6.3.)
Replacing the momentum by the current density from Eq. (6.4), j = −nev =
−nep/m, leads to
Ey = ρ(−ωc τ jx + jy ) , Ex = ρ(ωc τ jy + jx ) (6.10)
where the specific resistivity ρ is given in Eq. (6.6). Using the latter allows one to
introduce the notation ρωc τ = B/(nec) = −RH B, where the Hall coefficient RH
is
1
RH = − , (6.11)
nec
which brings the equations into the compact form
       
Ex ρ ρ jx ρ −RH B jx
= xx xy = . (6.12)
Ey ρyx ρyy jy −RH B ρ jy
In the presence of the magnetic field the specific resistivity turns out to be a tensor,
that relates the components of the current density with the components of the
electric field, in the general case where these two vectors are not parallel to one
another. The minus sign in Eq. (6.11) is designed such the sign of RH reflects that
of the charge carriers, i.e., RH is negative when the charge carriers are electrons. It
happens at times that the Hall coefficient is positive; the positive sign of the charge
carriers needs then to be explained.
Inverting the matrix in Eq. (6.12) yields
    
jx σxx σxy Ex
= , (6.13)
jy σyx σyy Ey
which is an extension of Eq. (6.5), with the specific conductivity replaced by a 2×2
tensor, whose components are
σ σ
σxx = σyy = = ,
1 + (τ ωc )2 1 + (σRH B)2
στ ωc σ 2 RH B
σxy = −σyx = = . (6.14)
1 + (τ ωc )2 1 + (σRH B)2
Problem 6.3.
The motion of an electron in a magnetic field. Prove that in the absence of an
electric field, and when collisions are ignored, the electron moves with a constant
velocity along the direction of the magnetic field, and performs a circular motion of
period 2π/ωc in the plane perpendicular to the magnetic field.

Measuring the Hall coefficient. Consider a sample of the shape of a box, as

illustrated in Fig. 6.1, and assume that a current flows only along the x̂−direction
(i.e., jy = 0). The electric field along the current direction follows Ohm’s law [Eq.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 421

Electrons in solids 421

(6.5)], Ex = ρjx . The current flowing along the x̂−direction creates an electric field
along the ŷ−axis,
B
Ey = ρyx jx = −ρτ ωc jx = − j ≡ RH Bjx . (6.15)
nec x
As no current flows along the ŷ−direction, negative charges are accumulated on the
lower face, and positive charges are collected on the upper one.
It is interesting to note that several of the coefficients specific to a certain system
(e.g., the relaxation time) disappear from the Hall coefficient, Eq. (6.11). This
coefficient, detected in experiment by monitoring the voltage created between the
opposite faces of the sample (along the ŷ−direction in the example above) depends
only on the charge of the carriers (−e in the present discussion) and their density
n. When it is known that the charge carriers are electrons, as assumed here, the
measurement yields their density. There are quite a number of materials for which
this description is faithful enough. However, as mentioned, there are materials with
positive Hall coefficients; these deserve a special consideration.

Fig. 6.1: The Hall effect geometry: the magnetic field is along the ẑ−axis, the cur-
rent flows along the x̂−axis, creating an electric field along the ŷ−axis (in addition
to the external electric field pointing along the x̂−axis).

Problem 6.4.
Compute the Hall coefficients of lithium (a BCC lattice with a lattice constant
a = 3.49Å) and copper (an FCC lattice with a lattice constant a = 3.61Å). Each
atom donates a single electron to the total conductance.

ac transport. When the electric field depends periodically on time, it gives

rise to an alternating current (ac). To find the corresponding ac conductivity,
it is convenient to present the field in the form E(t) = Re{E(ω) exp[−iωt]}. The
equation of motion (6.1) then requires the use of an analogous expression for the
average momentum, p(t) = Re{p(ω) exp[−iωt]}. Hence
−iωp(ω) = −p(ω)/τ − eE(ω) .
The resulting expression for p(ω) yields the local Ohm’s law for alternating
(ac) current,
j(ω) = −nep(ω)/m = σ(ω)E(ω) , (6.16)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 422

422 Introduction to Solid-State Physics

where
σ
σ(ω) = (6.17)
1 − iωτ
is the complex conductivity for an alternating current.
The Maxwell equations for electromagnetic fields that oscillate with a frequency
ω, E(t) = Re{E(ω) exp[−iωt]}, in the presence of an electric current, are
4π iω iω
∇ × B(ω) = j(ω) − E(ω) , ∇2 E(ω) = ∇ × B(ω) . (6.18)
c c c
Combining the two and using Eq. (6.16) leads to
ω2
−∇2 E(ω) = (ω)E(ω) , (6.19)
c2
with the complex dielectric function
(ω) = 1 + 4πiσ(ω)/ω . (6.20)
In the high-frequency limit, ωτ  1, Eq. (6.17) yields σ(ω) ≈ iσ/(ωτ ); hence
(ω) ≈ 1 − ωp2 /ω 2 , (6.21)
where ωp is the plasma frequency of the electrons, ωp2 = 4πne2 /m. In SI units,
the plasma frequency reads ωp2 = ne2 /(m0 ), where 0 is the dielectric constant
of the vacuum. At frequencies smaller than the plasma frequency (but still higher
than 1/τ ), the dielectric function (ω) is negative, and therefore the solutions
of Eq.
(6.19) decay exponentially with the distance, e.g., like exp[−κx], with κ = ω ||/c
(check!). The fields decay, and waves of those frequencies cannot penetrate into
the material; instead, they are reflected from the surface. This is the reason that
metals reflect light and behave like mirrors. At frequencies higher than the plasma
frequency the solutions are waves that propagate through the material. The metal
is thus transparent to fields at such frequencies.

Problem 6.5.
A metal is subjected to a constant magnetic field, B = Bẑ, and to an oscillat-
ing electric field, E(t) = Re[E 0 exp[−iωt]] oriented at an arbitrary direction. The
oscillating magnetic field connected to the latter is ignored.
a. Find the tensor of the electrical resistivity and that of the electrical conductivity.
b. Find the conductivity tensor in the limit of high frequencies, ωτ  1. What
happens when ω ≈ ωc = eB/(mc)?

Thermal conductivity. The contribution of the electrons to the thermal

conductivity of metals is described in the Drude picture within the kinetic theory,
(as is also argued for the thermal conductivity of phonons, Sec. 5.8). By Eq.
(5.101), the heat current is
jU = −κ∇T , (6.22)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 423

Electrons in solids 423

where
nv2 τ c v 2 τ cV 3
κ= = = nk 2 τ T (6.23)
3 3 2m B
is the thermal conductivity of the electrons. The last expression on the right
hand-side follows from a double use of the equipartition theorem that assigns to each
electron the energy u = mv2 /2 = 3kB T /2, and hence a specific heat, c = 3kB /2
(per electron). The central expression in Eq. (6.23) exploits the specific heat per
unit volume, cV = nc.
Note the similarity between Eq. (6.22) and Eq. (6.5), keeping in mind that the
electric field is (minus) the gradient of the electric potential, E = −∇φ. One may
obtain Eq. (6.5) from arguments similar to those that lead to Eq. (6.22). In both
cases the carriers are the same, carrying charge in the first case, and energy in the
second. This similarity gives rise to the Lorenz ratio,
L = κ/(σT ) . (6.24)
Indeed, using Eqs. (6.6) and (6.23), results in
3  k B 2
L= . (6.25)
2 e
Thus, the Lorenz ratio depends solely upon the physical constants e and kB ; it
is independent of any material coefficient. Equation (6.25) is the Wiedemann-
Franz law. Its numerical value is L = 1.11 × 10−8 Watt Ω/K2 (K represents
Kelvin’s scale). Using experimental values (for the charge conductivity and for the
thermal conductivity of the electrons), one finds that indeed the Lorenz ratio is
approximately the same constant for quite a number of metals. Nonetheless, that
constant is about twice the one derived form the Drude theory. Improved estimates
of both conductivities, as done in the next section, leads to a ratio much closer to
the experimental values.
The thermoelectric (Seebeck) effect. As seen, a temperature gradient
causes the electrons on the warmer side of a sample to be faster than those on
the colder side. Such a gradient creates an average velocity in the opposite direc-
tion of the gradient. This nonzero average velocity gives rise to an electric current,
induced by a temperature gradient (without an electric field). This is the ther-
moelectric effect; the electric current is proportional to the temperature gradient
that causes it. Alternatively, when no current is flowing (e.g., because the circuit is
not closed) the temperature gradient creates a voltage drop due to the accumulation
of electrons, i.e. charge, on the edge of the sample. The voltage drop that forms is
termed the Seebeck effect. This voltage cancels the electric current.
Problem 6.6.
Use the Drude model to obtain an expression for
a. the electric current created by a temperature gradient (the thermoelectric effect),
and
b. the electric field created when there is no current flowing (the Seebeck effect). Can
the magnitude of this field be estimated from the equipartition law for the specific
heat of the electrons?
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 424

424 Introduction to Solid-State Physics

6.3 The Sommerfeld theory

Flaws in the Drude theory. The Drude theory is based on the classical
picture. For instance, one of the assumptions of this theory is that the spe-
cific heat contributed by each electron is given by the equipartition theorem,
c = 3kB /2, and therefore the total contribution of the electrons to the specific
heat is CV,e = Ne c = 3Ne kB /2, where Ne is the number of conduction electrons.
Experiment finds that the specific heat of metals at room temperature is essentially
given by the Dulong-Petit law [see the discussion after Eq. (5.55)] that includes
solely the phonons’ contribution, CV,tot ≈ 3N kB . The actual contribution of the
electrons to the specific heat is hence much smaller than 3N kB . It follows that the
equipartition theorem does not pertain to electrons at room temperature. The use
of the equipartition theorem leads to other discrepancies with experiment, e.g., the
coefficient L of the Wiedemann-Franz law. These caveats of the Drude theory are
removed by using quantum mechanics for the electrons. That is, the electrons
are still considered as a gas of free particles (i.e., the electronic interactions, as well
as the interactions of the electrons with the ions, are ignored), but the particles are
treated quantum-mechanically. The theory that describes electrons as a gas of free
quantum particles is named after Sommerfeld.
Free quantum electrons. Quantum mechanics modifies the calculations in
two ways. First, the free electron is described by the Schrödinger equation,
2 2
− ∇ ψ(r) = Eψ(r) , (6.26)
2m
and its energy levels are given by the eigenvalues of this equation. Assuming that
the electron is confined in a box whose dimensions are Lx × Ly × Lz , then by
a variable separation the eigenfunctions are ψ(r) = ψx (x)ψy (y)ψz (z), where each
multiplier is a solution of the one-dimensional Schrödinger equation. In most of
the following considerations it is assumed that the wave function ψ obeys periodic
boundary conditions, as the ones exploited in Sec. 5.3 for the lattice vibra-
tions. Thus, e.g., for ψx (x) = ψx (x + Lx ), the normalized solution is a wavy one,
−1/2
ψx (x) = Lx exp[ikx x], with discrete values of the wave number, kx = 2πnx /Lx ,
nx = 0, ±1, ±2, . . ..√ Similarly, ky = 2πny /Ly and kz = 2πnz /Lz , such that
ψ(r) = exp[ik · r]/ V , where V = Lx Ly Lz is the volume of the box. Inserting
this solution into the Schrödinger equation (6.26) yields discrete energy levels,
2 2 2π 2 2  nx 2  ny 2  nz 2
Enx ,ny ,nz = k = + + . (6.27)
2m m Lx Ly Lz
Because the wave number is related to the momentum of the electron by p = k,
k is referred to as either a “wave vector” or a “momentum”.
As mentioned, other boundary conditions give rise to different discrete energies;
e.g., the vanishing of the wave functions on the sample’s edges yields ψx (x) =
(2/Lx )1/2 sin(kx x), with kx = πnx /Lx , nx = 1, 2, . . ., and similar expressions for the
other two directions (see problem 5.13). The physical quantities of a macroscopic
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 425

Electrons in solids 425

sample are not affected by these minor changes. The energy levels given in Eq.
(6.27) are discrete, but they become very dense as the size of the sample grows. For
instance, the interval between two levels corresponding to successive values of nx
is proportional to 1/L2x , and as such it tends to zero for an infinite system. Thus,
the infinite system possesses a continuum of energy levels. In most cases the focus
is on large but finite systems; one has at hand a discrete continuum of levels. A
similar discrete continuum appears in Chapter 5, in the context of the dispersion
relations of lattice vibrations.
The Pauli principle and the Fermi energy. The second modification intro-
duced by quantum mechanics is much more significant. As electrons are fermions,
they obey the Pauli principle, a basic concept of quantum mechanics: like
electronic levels in atoms, each of the single-particle energies in Eq. (6.27) can
accommodate solely two electrons, with opposite spins. Thus, not only the thermal
fluctuations prevent the electrons from being in their lowest energy level [out of
all those given in Eq. (6.27)]; it is also the Pauli principle. The most conspicuous
manifestation of the Pauli principle is at zero temperature, where there appears a
degenerate electron gas: the electrons fill in pairs the lowest Ne /2 energy levels,
such that their total energy is minimal. This is the ground state of the free
electron gas.
The topmost occupied energy level is termed the Fermi level, and its energy
(the Fermi energy) is denoted EF . For free particles one may write EF = 2 kF2 /(2m),
where kF is the Fermi momentum (the maximal momentum an electron can
have in the degenerate gas at zero temperature). kF is the Fermi wave num-
ber, corresponding to the Fermi momentum. From the discussion that leads to
Eq. (6.27) it can be deduced that the momentum component (in units of )
along the i−th axis, i = x, y, z, assumes discrete values in steps of Δki = 2π/Li ;
in the three-dimensional space of the momentum there is a single “allowed” value of
the quantum momentum in each basic cube, whose volume (in momentum space)
is Δkx Δky Δkz = (2π)3 /V , where V = Lx Ly Lz is the volume of the sample. The
energy levels below the Fermi energy, for which
2 k 2 2 kF2
Enx ,ny ,nz = < EF = ,
2m 2m
are located within a sphere in momentum space (the “Fermi sphere”), whose
radius is kF [an example of such a “sphere” with the discrete points it contains,
for a (two-dimensional) square system with Lx = Ly , is depicted in Fig. 6.3(a)].
The number of points contained in the Fermi sphere equals its volume divided by
the “basic” volume of each point,
Ne = 2[4πkF3 /3]/[(2π)3 /V ] = kF3 V /(3π 2 ) = nV , (6.28)
where n = Ne /V = kF3 /(3π 2 ) is the electrons’ density, and the factor of 2 counts
for the two spin directions. As explained below, the electrons that can change their
states at zero temperature are those located on the surface of the Fermi sphere, since
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 426

426 Introduction to Solid-State Physics

the other electrons are surrounded by occupied states which are blocked for them
due to the Pauli principle. Consequently, the typical momentum of the electrons
that are free to move in the quantum gas is kF , and the corresponding velocity is
the Fermi velocity, vF = kF /m. At low enough temperatures (as found shortly,
room temperature is very low in this context), this is the velocity that has to be
used in the Drude formula, replacing there the one derived from the equipartition
law. Recall that n ≈ 1023 cm−3 , and thus kF ≈ Å−1 , implying vF ≈ 108 cm/sec
(check!) This value is ten times larger than the one used by Drude, and therefore
the mean free-pathis significantly longer than the lattice constant. This finding
invalidates the conjecture that most of the electron’s collisions are with the ions in
the lattice.

Problem 6.7.
The Fermi temperature is defined by the relation kB TF = EF . Estimate the Fermi
temperature of a three-dimensional gas of free electrons whose density is n ≈ 1023
cm−3 . [ = 1.05 × 10−27 erg sec, m = 9.1 × 10−28 g, kB = 1.38 × 10−16 erg/K].

The energy of the electron gas at zero temperature. Given the electronic
population of the single-particle energy levels, the total kinetic energy of the electron
gas at zero temperature is given by

Etot = 2 Enx ,ny ,nz , (6.29)
E<EF

where the sum runs over all discrete values of the energy within the Fermi sphere.
For a large-volume system, the discrete points in momentum space are dense (they
form a “discrete continuum”), and the sum may be replaced by an integral,
 EF
Etot = V Eg(E)dE , (6.30)
0
of which V g(E)dE is the number of discrete energy levels with energies in-between
E and E + dE, multiplied by 2 due to the two spin states. As E = (k)2 /(2m), this
number is given by the volume of the spherical shell with wave number in-between
k and k + dk, divided by the volume assigned to a point in momentum space,
8πk 2 dk m √
V g(E)dE = = V 3 2 2mEdE (6.31)
3
(2π) /V  π
(check!). The density of states per unit energy and per unit volume, is given by
[using Eq. (6.28)],
3n √
g(E) = 3/2
E. (6.32)
2EF
It follows that the kinetic energy of the degenerate electron gas, at zero temperature,
is
 EF
Etot = V Eg(E)dE = (3/5)V nEF ,
0
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 427

Electrons in solids 427

from which follows the average kinetic energy per an electron,

E = Etot /Ne = Etot /(V n) = 3EF /5 . (6.33)
With kF = (3π 2 n)1/3 [Eq. (6.28)] this becomes
3 2 32 Ne  2 Ne 2/3
Etot = Ne (3π 2 n)2/3 = 3π , (6.34)
5 2m 10m V
and therefore the pressure of the electron gas is p = −(∂Etot /∂V )Ne = 2Etot /(3V ),
and the bulk modulus is B = −V (∂p/∂V )Ne = 5p/3 = 10Etot /(9V ) (check!).
The binding energy of metals. When each atom in the system contributes
a single electron, then V /Ne is the volume of the unit cell, and is of the order of R3 ,
where R is the distance between nearest-neighbor ions. Hence, Eq. (6.34) implies
that Etot /Ne ∝ 1/R2 . This result could have been derived also from arguments
based on dimensional analysis: when the only length in the problem is R, then the
only momentum possible is proportional to /R, and the average kinetic energy is
proportional to (/R)2 /(2m). By similar dimensionality arguments, the Coulomb
potential energy of the interaction between the electrons and the ions, and the
interaction among the electrons themselves, is of the order of −e2 /R. The total
energy of the electron gas takes the form Etot /Ne = A/R2 − B/R, with the positive
constants, A, B > 0. Its minimization with respect to R provides an estimate for
the size of the unit cell and for the binding energy of the metal: R = 2A/B and
(Etot /Ne ) = −B 2 /(2A). Thus, the minimization of the energy of the electron gas
determines the average distance between neighboring ions in the metal, i.e., the
lattice constant.
The Fermi-Dirac distribution. At finite temperatures, there is a finite prob-
ability for the electrons from the degenerate gas to go into energy levels higher than
the Fermi one. According to the Pauli principle, each single-electron quantum state
of energy Ei is occupied by ni electrons (or more generally, fermions), where ni
can be assigned only the values 0 or 1 (ignoring the spin degree of freedom). The
average value of ni is given by the Fermi-Dirac distribution,
1
ni  = f (Ei ) = , (6.35)
1 + eβ(Ei −μ)
where β = 1/(kB T ) and μ, which depends on the temperature, is the chemical
potential. The chemical potential is the free energy that is added to the system
with each extra electron. Ignoring the spin degree of freedom, the total number of
electrons is N = i ni . The chemical potential is determined by the requirement
that the average number of electrons which occupy all possible states equals the
total number of electrons in the system,
 
Ne = 2N  = 2 ni  = 2/[1 + eβ(Ei −μ) ] . (6.36)
i i

Equation (6.35) resembles in certain aspects Eq. (5.54) for the average number of
phonons in a given vibrational mode. As mentioned there, Eq. (5.54) is a particular
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 428

428 Introduction to Solid-State Physics

case of the Bose-Einstein distribution, which pertains to particles of integer-value

spins. In contrast, the Fermi-Dirac distribution describes particles of half-integer
spin values, e.g., electrons. In the case of bosons, ni can have any integer value
between 0 and ∞, while for fermions ni is 0 or 1.
As for fermions 0 ≤ ni ≤ 1, the function f (Ei ), Eq. (6.35), can be regarded as
the average probability to find a particle in a state of energy Ei . A simple way to
derive that expression is by conjecturing that the ratio of this probability to the
probability of that state to be empty, [1 − f (Ei )], is proportional to the Boltzmann
distribution, i.e., f (Ei )/[1 − f (Ei )] = z exp[−βEi ]. Indeed, when z = exp[βμ] Eq.
(6.35) is reproduced.
A more technical proof of Eq. (6.35) is based on statistical physics (see also
problem 6.8). The thermodynamic properties at finite temperatures are derived
from the partition function of the electron gas, i.e., its free energy. It is convenient
to calculate the latter for a fixed value of the chemical potential μ (i.e., for an
undetermined value of the number of electrons), and then to compute the average
number of electrons which corresponds to that chemical potential, i.e., to use Eq.
(6.36). For a certain configuration {ni } of the states’ occupations, the energy of the
electron gas (omitting the spin) is Etot = i ni Ei , and their number is N = i ni .
The Boltzmann’s weight of such a configuration is exp[−β(Etot − μN )], and the
partition function is the sum of all weights over all possible configurations, that is,
over all possible occupations {ni },
  
Z= e−β(Etot −μN ) = e−β i ni (Ei −μ)
Etot ,N {ni }
  
= e−βni (Ei −μ) = [1 + e−β(Ei −μ) ] . (6.37)
i ni =0,1 i

The average number of electrons at energy Ei (excluding spin) is obtained upon

averaging ni with the Boltzmann weights,

{nj } ni e−β j nj (Ej −μ)
∂ ln Z 1
ni  = f (Ei ) = −β

nj (Ej −μ)
=− = , (6.38)
{nj }e
j ∂(βEi ) 1 + eβ(Ei −μ)

which is the result given in Eq. (6.35). The Fermi-Dirac distribution replaces the
Maxwell-Boltzmann one, fMB (Ei ) = exp[−β(Ei − μ)], in the quantum regime. The
latter gives rise to the equipartition theorem used in the classical Drude theory.
The Fermi-Dirac distribution is identical to the Maxwell-Boltzmann one only in
the limit of energies much higher than β −1 = kB T , or at temperatures that largely
exceed the Fermi temperature (see problem 6.7), where the exponential dominates
the denominator in the right-hand side of Eq. (6.35).

Problem 6.8.
Another way to derive Eq. (6.35) is based on combinatorics. One assumes that each
quantum state may contain gi electrons at most (including a possible degeneracy of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 429

Electrons in solids 429

the energy level Ei , but ignoring the spin), and is occupied by ni electrons. The total
energy and the total number of electrons in the gas are given by Etot = i ni Ei and
N = i ni , respectively.

a. Show that the number of occupations’ possibilities is W = i {gi !/[ni !(gi − ni )!]}.
b. Use the Stirling formula, ln n! ≈ n ln n − n, to derive expressions for the entropy,
defined by S = kB ln W , and for the free energy, F = Etot − T S.
c. Assuming that the free energy is not changed when δn electrons are transferred
from the state m to the state , show that ∂F/∂nm = ∂F/∂n , and therefore this
derivative is identical for all states, and can be denoted as the chemical potential,
∂F/∂nm = μ.
d. Use the expression for F derived in part (b) to find μ, calculate the average ni 
and show that the result reproduces Eq. (6.35).

At zero temperature, i.e., when the electron gas is in the ground state, the
function f (E) becomes a step function: it is 1 for E < μ and zero otherwise (Fig.
6.2). Only the levels below μ are occupied, and the chemical potential coincides
with the Fermi level, μ(T = 0) = EF . At finite temperatures, as seen in Fig. 6.2,
f (E) is close to 1 as long as E − μ  kB T , and is close to zero in the opposite limit,
E − μ  kB T . In the latter case f (E) ≈ exp[−β(E − μ)]; the “tail” of f (E) decays
to zero like the Maxwell-Boltzmann distribution [see also part (c) in problem 6.10].

Fig. 6.2: The Fermi-Dirac distribution, Eq. (6.35). The thin line is for T = 0, the
thick one (with the shadowed area below) is for kB T /μ = 0.05.

The specific heat. At any finite temperature, f (μ) = 1/2. The exponential
factor exp[β(E − μ)] differs significantly from 1 or from 0 only when β|E − μ| ≤ 1,
and therefore the gradual decrease of the function from 1 to zero takes place at
energies in the range |E − μ| ≤ kB T . As found in problem 6.7, at room temperature
kB T  EF = kB TF , and so this width of this energy range is much smaller than the
Fermi energy, |E − μ|  EF . The probability of an electron to pass from an energy
within the Fermi sphere to a state of energy above that sphere is the ratio of the
area in-between the two lines (which is not shadowed) below μ and the total area
below the thin line in the figure. The former equals the shadowed area above μ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 430

430 Introduction to Solid-State Physics

which then represents the probability density of the electrons that are moved from
below to above the Fermi sphere. A rough estimate yields that each of these areas
is of the order of kB T (the area of a triangle whose base is kB T and whose height
is 1/2). A more careful calculation yields for the shadowed area above μ
 ∞ ∞

dEe−β(E−μ) /[1 + e−β(E−μ) ] = −kB T ln[1 + e−β(E−μ) ] = kB T ln 2
μ μ

(check that the area in-between the two lines below μ is indeed the same as this
result). The area of the (big) rectangular in Fig. 6.2 is μ, and at low enough
temperatures μ ≈ EF (as the temperature-dependent corrections to the chemical
potential of the degenerate gas are small). It follows that the probability of an
electron to ascend above the Fermi level is of the order of kB T /EF . As the total
number of electrons is Ne , the total number of electrons that move above the Fermi
level is ΔN ≈ Ne (kB T /EF ) (up to a coefficient of order 1). The total energy of
the gas then increases as compared to its zero-temperature value, since some of the
electrons have changed their states from below to above the Fermi sphere. As each
of those changes the energy by |E − μ| ≈ kB T , the total change in the energy is
ΔEtot ≈ ΔN kB T ≈ Ne (kB T )2 /EF , and hence the specific heat is (approximately)
linear in the temperature, CV ≈ 2Ne kB (kB T /EF ). As found in problem 6.7, the
Fermi temperature of a typical electron gas is TF = EF /kB ≈ 90, 000 K. (In fact, it
is TF = 136, 000 K for aluminium, and TF = 18, 400 K for cesium.) Hence, at room
temperature, kB T /EF ≈ 1/300, and the electronic specific heat is much smaller
than that of the phonons, which at room temperature follows the Dulong-Petit law,
CV (phonons) = 3N kB , in agreement with experiment.
The above calculation is an approximate one; in reality the changes in the en-
ergies of the electrons are not identical; a full calculation requires the computation
of the total energy at finite temperature, which averages over all those energies
 ∞  ∞
3Ne E 3/2
Etot  = V Eg(E)f (E)dE = 3/2
dE . (6.39)
0 2E 0 1 + eβ(E−μ)
F
The lower bound is set to zero because the density of states vanishes below zero
energy. The electronic specific heat is
 ∞
∂Etot  ∂f (E)
CV (electrons) = =V Eg(E) dE . (6.40)
∂T 0 ∂T
An expansion to leading order in the temperature (see problem 6.9) yields
π2 T
CV (electrons) ≈ V (3π 2 n)1/3 mkB2
T /(32 ) = N kB ≡ γT . (6.41)
2 TF
The result is indeed linear in the temperature, but the coefficient slightly differs
from the one derived in the crude approximation used above. At temperatures
lower than the Debye temperature of the phonons, the phonons’ contribution to the
specific heat, at three dimensions, is CV (phonons) ≈ AT 3 [see the discussion after
Eq. (5.69)]; the total specific heat is thus given by
CV ≈ γT + AT 3 . (6.42)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 431

Electrons in solids 431

Plotting the experimental values for CV /T as a function of T 2 , one usually obtains

a straight line. The measured value of γ is not the one found from the expression
given in Eq. (6.41). As γ is proportional to the mass of the electron [Eq. (6.41)],
the discrepancy is resolved by using an effective mass m∗ , different from that of free
electrons. This issue is resumed in the following.
Problem 6.9.
a. Plot the function −f  (E) ≡ −∂f /∂E for several values of the temperature. Show
that the integral of this function is 1, and that at low temperatures it has a symmetric
narrow peak around the Fermi level. Estimate the dependence of the peak’s width
on the temperature. Show that in the T → 0 limit, the function tends to a delta
function, −∂f /∂E → δ(E − μ).
b. The Sommerfeld expansion. Use the results of part (a) to show that the
leading correction to average quantities at low temperatures is proportional to T 2 .
[Hint: integrate by parts the integral that yields the average, say dEf (E)A(E),
and zoom in on the contributions around the Fermi level.]
Problem 6.10.
Denote the surface of the unit sphere in d dimensions by Sd [e.g., S2 = 2π and
S3 = 4π].
a. Find the relation between the electrons’ number and the Fermi energy at zero
temperature. Express the zero-temperature Fermi energy and the zero-temperature
Fermi momentum in terms of the density of electrons at arbitrary dimensions.
b. Find the electrons’ density of states at arbitrary dimensions, d. Find the ratio
of the total energy of the gas to the Fermi energy.
c. Given the number of electrons in the system, compare it with the average num-
ber found from the Fermi-Dirac distribution at arbitrary temperatures, and derive
thence an equation which yields the dependence of the chemical potential μ on the
temperature and on the Fermi energy. Use this equation to prove that in the high-
temperature limit, the chemical potential of a system with a finite number of electrons
tends to μ ≈ −(d/2)(kB T ) ln(kB T ) → −∞. Show that in this limit all thermody-
namic averages can be obtained from the Maxwell-Boltzmann distribution, and thus
the equipartition theorem and the Dulong-Petit law are valid at high temperatures.
d. Show that the chemical potential of a quantum electron gas at two dimensions is
given by μ(T ) = kB T ln{exp[EF /(kB T )] − 1}.
e. Find the exponents in Eq. (6.42) at arbitrary dimensions.
The electrical conductivity. The quantum description of a free electron,
according to the de Broglie principle, is by a wave of frequency ω = E/ and of
wave vector k = p/. (E is the energy of the electron and√p is the momentum.)
The plane-wave eigenfunctions ψk (r) = exp[ik · r − iω(k)t]/ V are spread over the
entire space. An electron confined to a finite volume is described by a wave packet.
The (average) velocity of the electron is then the group velocity of the wave packet,
v = ∇k ω(k). For a free particle, with E = 2 k2 /2m, this velocity is v = k/m.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 432

432 Introduction to Solid-State Physics

(a) (b)

Fig. 6.3: The dots represent discrete values of the momentum for the electron’s
eigenstates in a finite square sample, Lx = Ly . The states within the solid circles
are occupied by electrons (at zero temperature). (a) Without an electric field. (b)
In the presence of an electric field pointing along the negative x̂−axis.

At zero temperature, when the electron gas is in its ground state, all states within
the Fermi sphere are occupied. Consequently the total current density vanishes,
e 
J=− v =0, (6.43)
V i i
as for each point ki in the sphere there is also the point ki = −ki , see Fig. 6.3(a).
When the electron gas is subjected to an electric field, E, the picture is changed. At
distances long compared with the wave length of the electron the quantum averages
of the location and of the momentum fulfill the classical equations of motion; that
is, in-between collisions, the momentum, and hence the wave vector, obeys
ki (t) − ki (0) = −eEt/ . (6.44)
The Pauli principle, though, forbids the electron to move into a state already occu-
pied by two other electrons. Assuming that the electric field is along the negative
x̂−axis, the electrons move along the positive x̂−direction. When the field is weak
the change in the momentum of each electron is small, and therefore the electron
moves to a close-by momentum state. The only electrons that are capable of doing
so are those represented by points on the right side of the Fermi sphere. These move
to the right according to Eq. (6.44), vacating their original locations for electrons
on their left, which then also move to the right according to the same equation, i.e.,
the entire Fermi sphere is shifted towards the right, as illustrated in Fig. 6.3(b).
The states in-between the two spheres there, on the left (kx < 0), remain empty,
and the states in-between the two spheres on the right become occupied.
Reproducing the Drude formula. Recall that the electrical resistivity origi-
nates in collisions of the electrons with various flaws in the sample. As room tem-
perature is always significantly lower than the Fermi temperature, the only electrons
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 433

Electrons in solids 433

that can collide and as a result change their states are those located near empty
states, i.e., the electrons close to the surface of the Fermi sphere. The electrons that
reside deep in the sphere have no place to move to after the collision. Therefore, the
typical velocity of the colliding electrons is the Fermi velocity, and not the thermal
one that appears in the Drude theory. Equation (6.6) of the Drude theory does not
contain the velocity; hence, the relaxation time is still of the order of τ ∼ 10−14
sec. With the value vF ∼ 108 cm/sec, the mean free-path is = vF τ ∼ 100Å,
implying that the collisions cannot be with the ions of the lattice, whose distances
from each other are of the order of the lattice constant. The periodic array does
not scatter the electrons for reasons similar to those discussed in Chapter 3. The
quantum electron is described by a wave, and that wave can be scattered only when
its wave vector is changed by a reciprocal-lattice vector, as in the case of the Bragg
peaks of waves scattered off a periodic structure. As discussed below, a change by
a reciprocal-lattice vector does not alter the state of the electron, and hence cannot
contribute to the electrical resistivity.
In a stationary situation, the Fermi sphere is shifted by the amount mΔv/ =
Δk = −eEτ /. Most of the electrons still occupy pairs of states with opposite
momenta. Nonetheless, Fig. 6.3(b) illustrates that the points in-between the solid
and dashed circles do not have a counterpart. The number of these points is of the
order of Ne |Δv|/vF = Ne |eτ E/m|/vF . The velocity of each such electron is vF , and
thus the resultant current density is
ne2 τ
|J| ≈ n(|Δv|/vF )evF = neΔv = |E| . (6.45)
m
This is the Drude formula; the relaxation time, however, is that of electrons that
move with the Fermi velocity.
The heat conductivity and the Seebeck coefficient. The expression found
above for the heat conductivity is κ = nv 2 τ c/3 [Eq. (6.23)]; however, in this
expression one has to use the Fermi velocity and the specific heat from Eq. (6.41).
Substituting in the Lorenz ratio yields
κ π 2  k B 2
L= = . (6.46)
σT 3 e
With the numerical values of the physical constants, this becomes L = 2.45 × 10−8
Watt Ω/K2 . In contrast with the result of the Drude model, this value of L is in
fair agreement with experimental values of metals.
It is found in problem 6.6 that the Seebeck coefficient is Q = −c/(3e). With the
equipartition-law value for the specific heat this gives a result higher by a factor
of 100 as compared with experimental measurements. Equation (6.41) implies that
the ratio of the electronic specific heat calculated within the Sommerfeld theory to
its counterpart resulting from the Drude model is (π 2 /3)(T /TF ). As according to
problem 6.7 T /TF ≈ 1/300, the Seebeck coefficient within the Sommerfeld theory
is smaller by a factor of 100 as compared with its Drude’s analogue, and thus is in
agreement with experiment.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 434

434 Introduction to Solid-State Physics

The collisions that cause the electrical resistivity. As mentioned, the

collisions leading to the electrical resistivity cannot be among the electrons and the
ions forming the periodic lattice. In fact, the electrons collide with various types of
impurities, that might be foreign atoms or other flaws in the lattice structure, or
they are scattered by phonons or other electrons. As 1/τ is the probability for a
collision to occur, it can be presented as a sum over the probabilities pertaining to
the various collision processes,
1 1 1 1
= + + + ... , (6.47)
τ τimp τph τee
where the terms on the right hand-side refer to the three scattering processes men-
tioned above. The first of them is estimated as follows. Denoting the impurities’
concentration (their number per unit volume) by Nimp , and the scattering cross
section of each of them (the effective area normal to the propagation direction from
which the electron is scattered) by σimp , then the mean free-path of the electron
in-between such collisions, imp , is determined by the relation imp σimp Nimp = 1.
Here, imp is the distance the electron travels between collisions with the impuri-
ties, i.e., it is the average distance between the impurities. (The left hand-side is
the product of the impurities’ density by the volume that contains one of them.)
It follows that 1/τimp = v/ imp = vσimp Nimp , proportional to the impurities’
concentration. This density is independent of the temperature, and therefore com-
prises the only contribution to the resistivity at zero temperature. This is called
the residual resistivity, and is given by
m mvσimp
ρres = 2 = Nimp . (6.48)
ne τimp ne2
Next consider the probability of collisions with phonons. At temperatures higher
than the Debye one, the number of phonons in the α−vibrational mode is [from
Eq. (5.54)] nα  ≈ kB T /(ωα ). Some of these phonons have momenta of the order
of the Fermi momentum, and so might upon colliding with an electron cancel its
average momentum along the direction of the current. Their number is proportional
to kB T , and therefore the contribution to the electrical resistivity is proportional to
kB T as well. At lower temperatures the typical phonon momentum is proportional
to the temperature; consequently the change in the momentum of an electron after
colliding with a phonon is of the order of kph 2
/kF ∝ T 2 (this is the momentum
component along the direction of the motion after losing a momentum kph along
the normal direction, see the discussion of scattering into small angles in problem
6.2). In d dimensions, and within the Debye approximation [Eqs. (5.54) and (5.67)],
the number of such phonons, at low temperatures, is
  ωD  βωD
gph (ω) xd−1
Nph = nα  = dω βω ∝ Td dx x ∝ Td ,
α 0 e − 1 0 e − 1
where in the last step the upper limit of the integration is set to infinity. The phonon
contribution to the resistivity of a three-dimensional material is proportional to T 5
at low temperatures, and to T at higher ones.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 435

Electrons in solids 435

The probability of an electron to collide with other electrons is proportional

to the square of the density of the electrons that take part in the transport (this
is the probability for two electrons to be close to one another in space). The two
electrons have to be at an (energy) distance of the order of kB T from the Fermi
energy (to have empty states to move to), and therefore the contribution to the
resistivity is proportional to T 2 . At low temperature, this contribution is greater
than that arising from collisions with phonons. Figure 6.4 illustrates the resistivity
as a function of the temperature: the curve begins from the value ρres close to
T = 0, then ascends and becomes linear with the temperature at T higher than the
Debye temperature, TD . This behavior of ρ(T ) is termed “Matthiessen’s law”,
$ %
ρres + AT 2 , T  TD ,
ρ(T ) = A, B > 0 . (6.49)
BT , T  TD ,
Since ρres depends on the impurities’ concentration, it varies from sample to sample,
as opposed to the temperature-dependent part which is not, because it is determined
by the properties of the material. Indeed, data taken on different samples fall one
on top of the other once the resistivity curves are shifted vertically to account for
the differences in ρres . To illustrate this point, Fig. 6.4 presents two such curves.
It is also customary to define the resistivity ratio, the ratio between the room-
temperature resistivity and the very-low-temperature one. This ratio increases as
the material becomes cleaner.

Fig. 6.4: Schematic dependence of the resistivity on the temperature according to

the Matthiessen law, for two samples with different residual resistivities.

Problem 6.11.
a. The mass density of copper is ρm = 8.95 g/cm3 , and its specific resistivity at
room temperature is ρ = 1.55 μΩ cm (beware of the notation ρ, for the mass density
and for the resistivity). Assuming that each atom donates a single electron, whose
mass is that of a free electron, find the concentration of conduction electrons, the
relaxation time, the Fermi velocity, the Fermi energy, the Fermi temperature, and
the mean free-path of the electrons.
b. The residual resistivity of copper that contains one atomic percent of arsenic
is ρ = 6.8 μΩ cm. Find the cross section for the scattering of an electron by an
arsenic atom in the copper.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 436

436 Introduction to Solid-State Physics

6.4 Electrons in a periodic potential: the Bloch theorem

Preface. The free-electron theories discussed above, that ignore the periodic poten-
tial in which the electrons reside, fail to explain various phenomena. For instance,
certain materials are insulators though the Sommerfeld theory predicts a metallic
behavior for them; another example is the Hall coefficient that can be positive.
Recall that in the presence of a periodic potential the Schrödinger equation for a
single electron [Eq. (6.26)] is replaced by
2 2
Ĥ(r)ψ(r) = − ∇ + U (r) ψ(r) = Eψ(r) , (6.50)
2m
where the potential U has the periodicity of the lattice,

U (r) = U (r + R) , (6.51)

for any displacement vector in the lattice, R = n1 a1 + n2 a2 + n3 a3 : the

potential “seen” by an electron at each point in the unit cell coincides with that
felt at the same point in any other unit cell. The wave functions which solve Eq.
(6.50) in conjunction with Eq. (6.51), possess special properties, that are described
by the Bloch theorem. This theorem is mentioned also in the context of lattice
vibrations, see the discussion following Eq. (5.4).
The Bloch theorem.1 Any wave function ψnk (r) that is a solution for
the Schrödinger equation (6.50) with the periodic potential Eq. (6.51),
is characterized by a discrete quantum number n, and by a vector k of
quantum numbers, and can be written as a product,

ψnk (r) = eik·r unk (r) , (6.52)

where the function unk (r) is periodic on the lattice,

unk (r) = unk (r + R) (6.53)

with an arbitrary displacement vector on the lattice, R. For an infinite

system, the values of k form a continuum, while for a finite system they
are discrete and are given by the dense values listed in Eq. (5.43). The
solutions of the Schrödinger equation for a certain vector k are indexed
by the quantum number n; the eigenenergies are denoted by En (k).
The second version of the Bloch theorem. Upon inserting Eq. (6.53) into
Eq. (6.52), one finds

ψnk (r + R) = exp[ik · (r + R)]unk (r + R) = exp[ik · R]{exp[ik · r]unk (r)} ,

i.e.,

ψnk (r + R) = eik·R ψnk (r) . (6.54)

1 Bloch’s paper appeared in the same year, 1928, when he had been awarded his PhD degree.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 437

Electrons in solids 437

This means that a function which complies with the Bloch theorem obeys Eq. (6.54).
The opposite is true as well: if ψnk (r) obeys Eq. (6.54), one may define another
function,

unk (r) = e−ik·r ψnk (r) . (6.55)

Displacing the coordinate by R and using Eq. (6.54) yield

unk (r + R) = e−ik·(r+R) ψnk (r + R)

= e−ik·r [e−ik·R ψnk (r + R)] = e−ik·r ψnk (r) = unk (r) . (6.56)

It follows that the function unk (r) is periodic with the lattice vectors, as in Eq.
(6.53). Equation (6.54) is identical to the Bloch theorem (this equation is
referred to as “the second version of the Bloch theorem”). Though the quantum
numbers k in Eq. (5.43) coincide with the discrete values of the momentum of a free
particle (with periodic boundary conditions on the surface of an arbitrary triclinic
box), they do not coincide with the momentum of the electron. However, as
shown in Sec. 6.10, they do possess properties similar to those of the momentum,
and therefore k is referred to as the “lattice momentum” of the electron.
A qualitative proof. Suppose that ψ(r) is a solution of the Schrödinger equa-
tion in a certain cell, say the one at the origin, and denote the wave function in a
neighboring cell, displaced from the first one by the lattice vector a1 , by ψ(r + a1 ).
Due to the periodicity, the surroundings of r in the first cell and (r + a1 ) in the
second are identical; it is hence expected that the probability densities to find
the electron at these two points are also identical, i.e., |ψ(r)|2 = |ψ(r + a1 )|2 .
It follows that the ratio A = ψ(r + a1 )/ψ(r) (whose absolute value is 1) must
be of the form of A = exp[iα1 ], where α1 is a real phase. Consider next the
wave function in the next cell, ψ(r + 2a1 ). Again due to the periodicity, it is ex-
pected that this cell does not differ from the first and the second ones, and hence
A = ψ(r+2a1 )/ψ(r+a1 ). Likewise, A = ψ(r+(n1 +1)a1 )/ψ(r+n1 a1 ), and therefore
ψ(r + n1 a1 ) = An1 ψ(r). Periodic boundary conditions on the face of the crystal
along a1 imply that ψ(r + N1 a1 ) = ψ(r); consequently AN1 = exp[iN1 α1 ] = 1,
i.e., α1 = 2πh/N1 , where h is an arbitrary integer. Hence α1 is independent of r.
Nonetheless, A = exp[iα1 ] may depend on a1 . This discussion leads to the con-
clusion that A(n1 a1 ) = [A(a1 )]n1 for any integer n1 , i.e., α(n1 a1 ) = n1 α(a1 ). The
only possible solution for this equation is a linear function of a1 , α(n1 a1 ) = n1 k · a1 ,
where k is a real vector. The boundary condition α(N1 a1 ) = N1 k · a1 = 2πh is
fulfilled only when k = hb1 , where b1 is a reciprocal-lattice vector [Eq. (3.20)].
Hence ψ(r + na1 ) = exp[ik · n1 a1 ]ψ(r). Repeating this consideration for the other
directions of the lattice vectors results in Eq. (6.54), where k attains one of the
values listed in Eq. (5.43). As the change of the function between one of the unit
cells and another depends on k, the function itself has to depend on k as well. The
Schrödinger equation possesses many solutions for Eq. (6.54); the index n is added
to distinguish among them, and hence the notation ψnk (r). This completes the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 438

438 Introduction to Solid-State Physics

proof of the second version of the Bloch theorem, which is equivalent to the original
Bloch theorem.
Translation operators. The proof given above is based on plausibility ar-
guments and hence a more mathematical derivation is called for. Such a proof is
presented here; another one is given in Sec. 6.6. To this end, one needs to introduce
translation operators on the lattice. The translation operator T̂ (R) shifts the
coordinate r of an arbitrary function f (r) by the vector R = na a1 + n2 a2 + n3 a3 ,
T̂ (R)f (r) = f (r + R) . (6.57)
Obviously
T̂ (R)T̂ (R ) = T̂ (R + R ) = T̂ (R )T̂ (R) , (6.58)
and therefore all translation operators commute with each other [it is worthy to
note that according to Eq. (6.58) the collection of all translation operators forms
a group in the algebraic sense. That is, the members of this group are in one-to-
one correspondence with those of the translation group discussed at the beginning
of Sec. 2.7]. According to the theorem proved at the end of Appendix C, this
implies that there is a basis of the Hilbert space in which each of the basis states
ψ is simultaneously an eigenstate of all translation operators, T̂ (R)ψ(r) =
t(R)ψ(r), where the eigenvalue t(R) is a number.
Applying Eq. (6.58) on the eigenfunction ψ(r) yields t(R)t(R ) = t(R + R ),
where the eigenvalue t(R) is a number. The logarithm of this equation is ln t(R) +
ln t(R ) = ln t(R + R ). The only function that obeys this additive relation for
any pair of vectors R and R is linear in R, i.e., ln t(R) = c · R [one may add a
constant, but that constant is zero as t(0) = 1], leading to t(R) = exp[c · R]. The
normalization condition requires the absolute value of t(R) to be 1, and therefore c
is imaginary, c = ik, with a real k. Thus, the result is t(R) = exp[ik · R]. Periodic
boundary conditions on an arbitrary lattice with lattice vectors {am , m = 1, 2, 3},
comprising Nm unit cells along the lattice vector am , imply that f (r + Nm am ) =
f (r), that is, t(Nm am ) = t(am )Nm = exp[ik · Nm am ] = 1. One may repeat the
arguments in Sec. 5.3 (also used in the “qualitative proof” of the Bloch theorem),
to obtain that the independent quantum numbers that characterize the eigenvalues
of the translation operators are those given in Eq. (5.43), k1 ,2 ,3 = ( 1 /N1 )b1 +
( 2 /N2 )b2 + ( 3 /N3 )b3 , with −Nm /2 ≤ m < Nm /2 − 1. In summary, an eigenstate
of any translation operator obeys
T̂ (R)ψk (r) = eik·R ψk (r) = tk (R)ψ|k (r) , k = k1 ,2 ,3 . (6.59)
A simple example is that of a free particle, for which the potential energy vanishes,
U (r) = 0. The eigenfunctions
√ of Eq. (6.59) are then those of a free particle in a
(0)
box, Ψk (r) = exp[ik · r]/ V , for which
(0) (0) 1 (0)
T̂ (R)Ψk (r) = Ψk (r + R) = √ eik·(r+R) = eik·R Ψk (r) .
V
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 439

Electrons in solids 439

Proof of the Bloch theorem. Equation (6.51) implies that the Hamiltonian
Ĥ(r) = −(∇)2 /(2m) + U (r) is such that Ĥ(r + R) = Ĥ(r) [the kinetic energy is
invariant to translations, as, e.g., d(x + a) = dx]. Hence, any function obeys the
relations
T̂ (R)[Ĥ(r)f (r)] = Ĥ(r + R)f (r + R) = Ĥ(r)f (r + R) = Ĥ(r)[T̂ (R)f (r)] ,
i.e.,
Ĥ(r)T̂ (R) = T̂ (R)Ĥ(r) ,

the Hamiltonian Ĥ(r) commutes with all the translation operators of the
lattice, {T̂ (R)}. The theorem proven at the end of Appendix C implies that there
is a basis of the Hilbert space comprising states that are simultaneously eigenstates
of all translation operators and of the Hamiltonian. Each of these states has to
obey Eq. (6.59); such a state is also a solution of the Schrödinger equation,
Ĥ(r)ψnk (r) = En (k)ψnk (r) , (6.60)
where the index n distinguishes among the various solutions of the equation, all
belonging to the subspace of the Hilbert space with the eigenvalue exp[ik · R] of
T̂ (R) (for any R). As all functions within this subspace depend on k, the eigenvalues
of the energy, En (k), depend on it as well. In other words, one may choose the
eigenfunctions of the Schrödinger equation with a periodic potential such that they
obey both Eq. (6.59) and Eq. (6.60). As Eq. (6.59) coincides with Eq. (6.54), this
is a proof of the Bloch theorem in its second version. As shown in Eq. (6.56), it
also proves the first version of the Bloch theorem.
Conclusions drawn from the Bloch theorem. The Bloch theorem has quite
an impact on the possible solutions of the Schrödinger equation with a periodic
potential. Inserting Eq. (6.52) into Eq. (6.60) yields
 2 2    
− ∇ + U (r) eik·r unk (r) = En (k) eik·r unk (r) . (6.61)
2m
By differentiating (in the left hand-side) and then eliminating exp[ik · r], this be-
comes
 2 
(−i∇ + k)2 + U (r) unk (r) = En (k)unk (r) . (6.62)
2m
According to the Bloch theorem [Eq. (6.53)], the function unk (r) is periodic, and
hence it suffices to solve this differential equation in a single unit cell, with
periodic boundary conditions on its faces, and then to exploit Eq. (6.52) to obtain
the solution in the entire space.
The eigenfunctions unk (r) and the eigenenergies En (k) depend on k, as it ap-
pears explicitly in the left hand-side of Eq. (6.62). However, Eq. (6.62) is a
second-order differential equation within the unit cell, and as such it may have
many solutions with different eigenenergies and different eigenfunctions (recall the
examples of an infinite potential well, a harmonic oscillator, or the hydrogen atom,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 440

440 Introduction to Solid-State Physics

whose respective Schrödinger equations have many discrete energy levels). As men-
tioned, the index n distinguishes among these various solutions. It is worthwhile
to note that in three dimensions a single discrete index usually does not suffice to
characterize the eigenfunctions. For instance, in the case of the hydrogen atom
there are the three indices n, , and m, while for a free particle in a box these are
the indices nx , ny , and nz [Eq. (6.27)]. For brevity, a single index n is kept below.
Periodicity in the reciprocal lattice. The Bloch theorem is phrased for
k’s whose values are confined to the first Brillouin zone, Eq. (5.43). This equation
follows from the periodic boundary conditions, exp[ik·Nm am ] = 1; these are obeyed
also by vectors shifted into other Brillouin zones. In other words, the eigenvalues of
Eq. (6.59) obey tk+G (R) = exp[i(k + G) · R] = exp[ik · R] = tk (R), where G is an
arbitrary reciprocal-lattice vector. The central step in the equation results from Eq.
(3.17), exp[iG · R] = 1. Consequently, there is no difference between the solutions
in the first Brillouin zone, and those in any other zone shifted by a reciprocal-lattice
vector G: one may replace in the previous equations k by k + G, to obtain
ψn,k+G (r) = ψn,k (r) , (6.63)
and
En (k + G) = En (k) . (6.64)
Note the similarity between Eq. (6.64) and Eq. (5.7): the spectrum of the lattice
vibrations is also periodic in reciprocal lattice; in both cases it suffices to obtain the
wave functions and the eigenvalues in the first Brillouin zone.

Problem 6.12.
∗
a. Prove that ψn,k (r) = ψn,−k (r), and hence En (−k) = En (k). [Complex conju-
gating the time-dependent Schrödinger equation, Ĥ(r)ψ(r, t) = i(∂/∂t)ψ(r, t), is
equivalent to time inversion, t → −t. Therefore this result proves that the Hamil-
tonian is invariant under time reversal.]
b. Repeat the proof for an orthorhombic lattice in three dimensions, and show that
En (−kx , ky , kz ) = En (kx , ky , kz ).

Energy bands. As mentioned, the values of k are rather dense (they form a
“discrete continuum”); they become a continuum in the infinite-lattice limit. As
opposed, the values of n are discrete. Thus, for each n there is, in the continuum
limit, a separate continuous function of k, that repeats itself in the various Bril-
louin zones (or among the unit cells in reciprocal lattice). Several examples for the
calculation of such functions are presented in the next three sections. They all give
rise to the same qualitative picture as illustrated in Fig. 6.5: each of the sinusoidal
lines in Fig. 6.5 belongs to a different energy function En (k), with a different index
n. For a specific value of n the function En (k) is periodic in k. In part (a) of
the figure, the period is G0 = 2π/a (a is the lattice constant, and the horizontal
coordinate is k/G0 ). The vertical lines in that figure are the boundaries of three
unit cells in reciprocal lattice (that are equivalent to three Brillouin zones).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 441

Electrons in solids 441

(a) (b)
Fig. 6.5: (a) Energy bands of an electron in a one-dimensional periodic potential,
computed in the tight-binding approximation (Sec. 6.7). The thick sinusoidal curves
are the energies as functions of the lattice momentum, and the shadowed areas
indicate the range of energies in each band. The vertical lines are the boundaries
between the Brillouin zones in reciprocal lattice. The central cell is the first Brillouin
zone. (b) An example of energy bands of a two-dimensional periodic potential on a
square lattice, computed within the same approximation.

Each of the functions En (k) comprises energy values confined to a certain range.
This range, shadowed in Fig. 6.5(a), is termed an “energy band”. In simple cases,
the bands are separated by energy gaps, in which the Schödinger equation has no
eigenvalues (see also problem 6.13). Two types of states are frequently encountered
in quantum mechanics: bound states, e.g., those of an electron in the hydrogen
atom, or the ones of a harmonic oscillator; these belong to discrete energy levels
separated from each other by finite energies. Alternatively, the energy levels of a
free particle confined to a large box are very dense, see for instance those given in
Eq. (6.27), and then they form a “discrete continuum”; in the limit of an infinite
system those levels become a continuum of positive energies. The spectrum of the
periodic potential joins together the two types of states: each band comprises many
dense energy levels, but the bands are separated by gaps in which there exist no
levels, similar to the gaps between the energies of the bound states in the atom.
These gaps remain finite also in the limit of an infinite system. It is worthwhile to
note the similarity between the energy gaps of the periodic potential and the gap
separating the acoustic and optical modes of the lattice vibrations, see Chapter 5.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 442

442 Introduction to Solid-State Physics

The energy levels within each band are the available energies for a single electron
in a periodic potential.

Fig. 6.6

Problem 6.13.
Figure 6.6 exhibits bands that “penetrate” into each other. Prove that such a pen-
etration is impossible in one dimension. Hence, in one dimension the successive
bands are always separated by energy gaps, and at most they may touch each other.
Why is a penetration possible at higher dimensions?

Note that the energy levels are derived for a single electron. These levels group
themselves into bands that are filled with electrons according to the Pauli principle.

6.5 The Kronig-Penney potential

A square-shape periodic potential. A simple example that illustrates the band

structure of the energy levels is based on the one-dimensional potential of Kro-
nig and Penney, built of periodic steps of square-shape potential,
 0, na ≤ x ≤ na + b ,
U (x) = (6.65)
U0 , na + b ≤ x < (n + 1)a ,
for any integer n (b < a). As seen in Fig. 6.7, the lattice constant is a. The
segments with the lower potential (0 in the present choice) represent regions to
which the electron is attracted, similarly to the Coulomb potential of the positive
ions arranged periodically in a crystal. Since the potential is periodic, the solutions
of the Schrödinger equation obey the Bloch theorem, and can be represented by
Eq. (6.52), ψk = exp[ikx]uk (x). The periodic function uk (x) is the solution of the
one-dimensional version of Eq. (6.62),
 2 d 
(−i + k)2 + U (x) uk (x) = E(k)uk (x) , (6.66)
2m dx
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 443

Electrons in solids 443

Fig. 6.7: The periodic Kronig-Penney potential. The width of each “well” is b, and
the width of each potential barrier (whose height is U0 ) is a − b.

within a single unit cell, 0 ≤ x ≤ a, with periodic boundary conditions, uk (a) =

uk (0) and uk (a) = uk (0).
The wave function. Consider the solutions for 0 < U0 < E; the treatment
of the case 0 < E < U0 is quite similar, see below. As can be verified by a direct
substitution, the solutions of Eq. (6.66) are
uk (x) = e−ikx [AeiKx + Be−iKx ] , for 0 < x < b ,
uk (x) = e−ikx [CeiQx + De−iQx ] , for b < x < a , (6.67)
where
K 2 = 2mE/2 , Q2 = 2m(E − U0 )/2 . (6.68)
By continuity, uk (b+ ) = uk (b− ) and uk (b+ ) = uk (b− ), and hence
AeiKb + Be−iKb = CeiQb + De−iQb ,
K[AeiKb − Be−iKb ] = Q[CeiQb − De−iQb ] . (6.69)
The periodicity of uk (x) implies that uk (a) = uk (0) and uk (a) = uk (0), and so
A + B = e−ika [CeiQa + De−iQa ] ,
K(A − B) = Qe−ika [CeiQa − De−iQa ] . (6.70)
Adding and subtracting the two equations in (6.70) give
1 −ika
A= e [(1 + Q/K)CeiQa + (1 − Q/K)De−iQa ] ,
2
1
B = e−ika [(1 − Q/K)CeiQa + (1 + Q/K)De−iQa ] . (6.71)
2
In a similar fashion one finds from Eqs. (6.69)
1 −iKb
A= e [(1 + Q/K)CeiQb + (1 − Q/K)De−iQb ] ,
2
1
B = eiKb [(1 − Q/K)CeiQb + (1 + Q/K)De−iQb ] . (6.72)
2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 444

444 Introduction to Solid-State Physics

A comparison of the two pairs of expressions for A and B yields two homogeneous
equations for C and D. A non-trivial solution exists when the determinant vanishes,
i.e., when
K 2 + Q2
cos(ka) = cos[Q(a − b)] cos(Kb) − sin[Q(a − b)] sin(Kb) (6.73)
2KQ
(check!) As seen from Eqs. (6.68), both K and Q depend on the energy (and on the
height of the potential barriers, U0 ). Therefore, Eq. (6.73) gives explicitly k (that
appears only on the left hand-side) as a function of E (contained solely in the right
hand-side). A plot of k as a function of E, together with a rotation of the resulting
figure by 90◦ , produces the energy levels as functions of k, i.e., the energy bands.
As the same value of k pertains for numerous energies, there exist numerous bands.
The eigenenergies. Introducing the dimensionless variables,
E a 2mU0 2
= , z = −1 , W = b , (6.74)
U0 b 2
transforms Eq. (6.73) into the form
 √
cos(ka) = F () = cos[z W ( − 1)] cos( W )
2 − 1  √
−  sin[z W ( − 1)] sin( W ) . (6.75)
2 ( − 1)

When  > 1 then Q = W ( − 1)/b is real, and consequently all quantities on
the right hand-side are real. On the other hand, when  < 1 then Q is purely
√ √
imaginary; using the identities  − 1 = i 1 − , cos(iα) = cosh(α), and sin(iα) =
isinh(α), one obtains again real quantities in the right-hand side. The function
F () is depicted in Fig. 6.8(a), for W = 200 and z = 0.1. (Other values give
similar results.) As seen, there are energy segments for which 1 < |F ()|. In
those, Eq. (6.75) yields complex values for (ka); these are unphysical as the Bloch
theorem requires k to be real (the normalization implies that | exp[ika]| = 1). Hence,
the allowed energy levels are in the segments for which |F ()| < 1. Each such
segment (shadowed in the figure) corresponds to a band of allowed energies. In-
between these segments there appear energy gaps, in which the Schrödinger equation
has no acceptable solutions. For each allowed value of the energy, Eq. (6.75)
produces definite values for cos(ka), and then the solutions repeat themselves with
the period ka → ka+2π , i.e., under a translation of k by a reciprocal-lattice vector,
G = 2π /a = G0 . These solutions, as computed from Eq. (6.75), are displayed
in Fig. 6.8(b). Indeed, the energies that correspond to a certain value of k, in the
various Brillouin zones, coincide with each other. Each such value allows for many
values of the energy: the horizontal line [that represents cos(ka)] has many cuts
with the branches of F () in Fig. 6.8(a).
As seen in Fig. 6.8(a), the function F () is monotonic in each shadowed area;
this implies that the energy is minimal or maximal at the ends of the segment, i.e.,
when cos(ka) = 1 or cos(ka) = −1. In particular, there are minima or maxima for
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 445

Electrons in solids 445

(a) (b)
Fig. 6.8: The right hand-side of Eq. (6.75), for W = 200 and z = .1. The shadowed
areas contain the “allowed values” of the energy, i.e., the energy bands. (b) The
solutions of Eq. (6.75), for the same parameters. The thick lines mark the first
Brillouin zone.

ka = 0, that is, at the center of the Brillouin zone, or for ka = ±π, i.e., at the
ends of the first Brillouin zone, as indeed seen in Fig. 6.8(b). Figure 6.8(a) also
shows that the slopes of F () change sign between successive bands, and therefore
the locations of the minima and the maxima alternate between such bands. This
observation is confirmed in Fig. 6.8(b).
The wave function. Figure 6.9 illustrates the absolute value squared of the
wave function, that is, the probability density to find the electron, for parameters
given in the caption, with the energies (in units of U0 )  = .6 and  = 2. In
both cases the probability is oscillatory in the ranges where the potential barrier
vanishes. In Fig. 6.9(a) the energy is lower than the height of the potential barrier
and consequently Q is purely imaginary; the probability to find the electron in the
range b < x < a is exponentially decaying on the two sides of this segment. Figure
6.9(b) displays the case where the energy is above the potential step, and therefore
the probability to find the electron in that range is considerably higher.
The “nearly-free” electron. Equation √ (6.73) has simple solutions in two
limits. When U0 = 0 then Q = K = 2mE/, and the right hand-side becomes
cos(Ka). In that case ka = Ka + 2 π for any integer , yielding E = 2 (k −
G )2 /(2m), with G = G0 = 2π /a. One might have expected that when the
periodic potential vanishes the solutions of the Schrödinger equation are those of the
free particle, E0 (k) = 2 k 2 /(2m). This parabola is indeed obtained in the present
case for = 0. Nonetheless, even when the periodic potential is infinitesimally small,
almost unnoticed in the energy levels, the system does “identify” its existence, and
then the Bloch theorem yields Eq. (6.64). This equation in turn dictates that the
energy is periodic in reciprocal lattice, and therefore the parabola E0 (k) can be
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 446

446 Introduction to Solid-State Physics

(a) (b)

Fig. 6.9: The probability density to find the electron within two neighboring unit
cells, for W = 100, z = −.3, and (a)  = .6, (b)  = 2.

shifted to the left or to the right by an arbitrary reciprocal-lattice vector, giving

rise to an infinite ensemble of parabolae, E (k) = E0 (k − G ) = 2 (k − G )2 /(2m).
This ensemble, as derived from the limiting solution of Eq. (6.73), is displayed in
Fig. 6.10(a). It is discussed in more detail in the next section.

(a) (b)

Fig. 6.10: The energy levels of the Kronig-Penney potential. (a) The free-electron
limit, U0 = 0. [The energy is in units of 2π 2 2 /(6ma2 ).] (b) The limit of infinite
potential wells, U0 → ∞. [The energy is in units of π 2 2 /(2mb2 ).]

The tight-binding limit. Another simple case is obtained when the potential
barrier is quite high, U0 → ∞. Then Q2 → ∞ and the right-side member of Eq.
(6.73) is considerably larger than the other two. As a result, the equality exists
only when sin(Kb) = 0, i.e., E = 2 K 2 /(2m) = 2 π 2 n2 /(2mb2 ), where n is an
arbitrary integer. These are precisely the discrete energy levels in a single infinite
potential well of width b (check!). The energy levels do not depend on the wave
number k, and therefore each of the energy bands becomes an horizontal line at one
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 447

Electrons in solids 447

of the discrete levels, see Fig. 6.10(b). As opposed to the single-well case, there
are now N identical wells, and therefore each energy level is N -fold degenerate.
Decreasing gradually U0 widens these horizontal lines and narrows the gaps in-
between them, until the free-particle limit is reached, and then neighboring bands
touch one another. Figures 6.8(b) and 6.11 display intermediate situations between
these two limits.

(a) (b)

Fig. 6.11: Energy bands of the Kronig-Penney potential, with z = .1 (a) Weak
potential, W = 10. (b) Strong potential, W = 1000. Energy in units of U0 .

As seen in Fig. 6.8(a), the amplitude of the oscillations of the function F ()
decreases when the energy is increased for finite values of U0 . Therefore, the regions
where |F ()| > 1 (that represent the energy gaps in-between the energy bands) are
narrowed down
√ for high-energy bands. Indeed, when   1, one finds F () ≈
cos[(z + 1) W ] = cos(Ka), which reproduces the free-particle solutions (check!).
In other words, electrons of high energies do not “feel” the potential.

Problem 6.14.
a. Find the limit of Eq. (6.73) for U0 → ∞ and (a − b) → 0 under the condition
that u0 = U0 (a−b) remains constant. This is the limit in which each step becomes a
delta function, and the potential has the shape of a “comb”, U (x) = u0 n δ(x−na).
Show that in this limit the value Ka = nπ is always the border between an energy
band and an energy gap.
b. Calculate the energy bands for large values of the constant φ = u0 am/2 =
−Q2 (a − b)a/2.
c. Find the energy gaps between the bands that correspond to small values of φ.
What is then the energy at the bottom of the lowest band?
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 448

448 Introduction to Solid-State Physics

6.6 Nearly-free electrons

Weak and strong potentials. The Kronig-Penney potential is a special case,

since it allows for an exact solution of the Schrödinger equation. As in most cases
an exact solution does not exist, approximations or numerical methods are called for.
Figure 6.10 illustrates the energy levels in two extreme limits, when the potential
is vanishingly small and when it is very strong. In the first case, the solution of the
Schrödinger equation is that of a free particle; in the second the solution coincides
with that of a single “unit”, described in Sec. 6.5 by an infinite potential well
corresponding to the attractive potential of a single ion in the lattice. It turns out
that one may build on any of these extreme limits, to find approximate solutions
that yield the energy levels slightly away from those limits. The analysis of the
weak-potential situation [see Fig. 6.11(a)] is carried out in the present section; the
next one is devoted to the opposite case [Fig. 6.11(b)]. The approximation based
on a weak potential describes faithfully the alkali metals (the ones in the first two
columns of the periodic table), whose valence electrons are relatively far away from
the nucleus and are only loosely bound to it. The discussion below pertains to a
general three-dimensional periodic potential, denoted U (r).
Bragg conditions for scattering of electrons off a periodic lattice. A
comparison of Figs. 6.10(a) and 6.11(a) shows that the effect of the weak peri-
odic potential on the energy levels is conspicuous close to the boundaries of the
Brillouin zones, that is, when k is near an integral multiple of π/a. A qualitative
understanding of this feature is gained by considering the discussion in Chapter 3.
It is shown there that a plane wave that impinges on a periodic lattice is signifi-
cantly scattered only when the difference between the incoming and outgoing wave
vectors equals one of the reciprocal-lattice vectors, Eq. (3.9). This condition leads
to Eq. (3.26), k · Ĝ = −G/2, that defines the boundaries of the Brillouin zones.
Indeed, in one dimension this equation yields k = G/2 = π/a. In the absence of
the periodic potential√the wave function of the electron is that of a free particle,
(0)
Ψk (r) = exp[ik · r]/ V (V is the total volume of the crystal), and this is a plane
wave of the same type as discussed in Chapter 3. A wave with k far away from
the boundaries of the Brillouin zone traverses the lattice almost unscattered. The
energy corresponding to such a wave vector is hence very close to that of the free
particle, E(k) ≈ E (0) (k) = 2 k2 /(2m). The probability density of the particle is
(0)
constant in space, |Ψk (r)|2 = 1/V . As opposed, when k complies with the con-
dition for the Bragg scattering, Eq. (3.26), the wave (that describes the electron)
is scattered; the wave function contains the original wave together with all the
scattered ones. √
In one dimension the incident
√ wave is exp[ikx]/ L, and the only possible scat-
tered wave is exp[−ikx]/ L. The wave function is a linear combination of these
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 449

Electrons in solids 449

two waves,
1   1  
Ψ → √ αeikx + βe−ikx = √ (α + β) cos(kx) + i(α − β) sin(kx) ,
L L
|α| + |β|2 = 1 .
2

The probability density of the electron’s location is thence a function of the co-
ordinate. Assuming that the positive ions are located at x = (n + 1/2)a, i.e., at
the centers of the unit cells, then the potential there is attractive, tending to bind
the electron to the ion. It is thus plausible that the energy of the electron is low-
est when the probability density to find it at these locations is maximal. This is
particularly prominent at the boundary of the Brillouin zone, k = π/a. The proba-
bility density of the electron
√ at the ions’ locations is maximal when its wave function
is Ψ− (x) = sin(πx/a)/ 2L. As opposed, the energy of the electron is highest √ in
the other extreme case, when the wave function is Ψ+ (x) = cos(πx/a)/ 2L, be-
cause then the probability density at the vicinity of the ions is the lowest. One
therefore may expect that when k = π/a there is an energy gap between the two
states Ψ+ and Ψ− , as can indeed be observed in Fig. 6.11(a). Interestingly enough,
both these states represent standing waves, with equal weights for the incident and
reflected waves, |α|2 = |β|2 . This feature of the wave functions stems from the total
reflection that is realized when the wave vector obeys the Bragg condition (3.26).
Recall that standing waves also characterize lattice vibrations at the vicinity of the
Brillouin-zone boundaries [see the discussion following Eq. (5.8)]. A systematic
derivation of this result, as well as the magnitude of the energy gap, is presented in
the following.
Bragg scattering does not contribute to the electrical resistance. In
one dimension, the wave function of the scattered particle comprises the incident
wave, with wave vector k, and the scattered wave, with the wave vector −k = k +G.
In three dimensions there are many scattered waves, and the wave function of the
electron is a linear combination of the incident wave of wave vector k, and all
scattered waves, each with a wave vector k → k + G. The coefficients of this linear
combination are computed below. Note though that by Eqs. 6.63) and (6.64) a shift
of the lattice momentum by a reciprocal-lattice vector does not change the wave
function and the corresponding energy. This is the reason why Bragg scattering
does not contribute to the electrical resistance of the metal, contrary to
the original conjecture of Drude.
The Schrödinger equation in momentum space. The analysis in this
section provides yet another proof of the Bloch theorem. Consider the Fourier
transforms of the periodic potential and of the electron’s wave function. As seen in
Sec. 3.8, the Fourier transform of a function that is periodic on a lattice contains
only components with wave vectors belonging to the reciprocal lattice [Eq. (3.34)],
and therefore the potential energy can be written as

U (r) =  (G)eiG·r ,
U (6.76)
G
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 450

450 Introduction to Solid-State Physics

where the sum runs over all N reciprocal-lattice vectors (N is the number of unit
cells in the original lattice). The coefficients in the sum are given by the Fourier
transforms,

 1
U (G) = d3 re−iG·r U (r) , (6.77)
v v
where the integration is carried out over a single unit cell, v = V /N . For G = 0,
the Fourier transform is U  (0) = (1/v) d3 rU (r), and is a constant that represents
v
the choice of the coordinate origin. This constant is subtracted from all energies,
so that the sum over G in Eq. (6.76) comprises only nonzero reciprocal-lattice
vectors [i.e., it is practically assumed that U  (0) = 0]. The potential U (r) is real,
  ∗
and hence U (−G) = U (G). When the potential is invariant under reflection (as
is usually the case for Bravais lattices), then U (r) = U (−r), and consequently
U ∗ (G) = U (−G) = U  (G); it follows that U
 (G) is real [see also problem (B.1)].
The Fourier transform of the wave function is

ψ(r) = 
ψ(k)e ik·r
, (6.78)
k

where periodic boundary conditions along each of the basis vectors imply that
exp[ik · Ni ai ] = 1, and hence k = i (mi /N )bi . Here, bi are the reciprocal-lattice
basis vectors, and the sum encompasses the vectors k in the first Brillouin zone.
Equation (6.78) presents the wave function in a periodic potential as a linear com-
bination of the free-particle wave functions, that by themselves constitute a basis
of the Hilbert space of all wave functions (this is the Fourier theorem, see App. B).
The Schrödinger equation. Applying the Hamiltonian on Eq. (6.78) yields
2 2      2 k 2

− ∇ ψ(k)eik·r = ψ(k)e ik·r
, (6.79)
2m 2m
k k

and
   
U (r)ψ(r) =  (G)eiG·r
U   )eik ·r =
ψ(k U  − G) eik·r .
 (G)ψ(k
G k k G
(6.80)
In the second equality of Eq. (6.80) the summation index is changed, k → k =
k +G. Inserting Eqs. (6.79) and (6.80) into the Schrödinger equation (6.50) results
in
  2 k 2  

− E ψ(k) + U  − G) eik·r = 0 .
 (G)ψ(k (6.81)
2m
k G

This sum vanishes only when all the coefficients (in the circular brackets) vanish,
i.e.,
 2 k 2  

− E ψ(k) + U  − G) = 0 .
 (G)ψ(k (6.82)
2m
G
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 451

Electrons in solids 451

Equation (6.82) is a set of linear homogeneous equations, involving N unknowns

 − G)}, where G is one of the reciprocal-lattice vectors (including G = 0).
{ψ(k
Each of the k vectors in the first Brillouin zone possesses such a set. Except for
certain particular values of k (discussed below) these sets do not mix coefficients
related to different k’s, and therefore each set can be treated separately. Equation
(6.82) can be presented in a matrix form, with an N × N matrix multiplying the
vectors comprising the N unknowns. In the one-dimensional example discussed at
the beginning of this section, the matrix equation is
⎡ ⎤
... ... ... ... ... ...
⎢ . . . E (0) (k − G ) − E U  (−G ) U (−2G ) U (−3G ) ...⎥
⎢ 0 0 0 0 ⎥
⎢    ⎥
⎢... U (G0 ) E (k) − E
(0)
U (−G0 ) U (−2G0 ) ...⎥
⎢  (2G )  (G )  (−G ) ⎥
⎢ ... U U E (0 (k + G0 ) − E U ...⎥
⎢ 0 0 0 ⎥
⎣ ...  (3G )
U  (2G )
U U (G ) E (0) (k + 2G0 ) − E . . . ⎦
0 0 0
... ... ... ... ... ...
⎡ ⎤
...
⎢ ψ(k
 −G ) ⎥
⎢ 0 ⎥
⎢  ⎥
⎢ ψ(k) ⎥
×⎢  ⎥=0, (6.83)
⎢ ψ(k + G0 ) ⎥
⎢ ⎥
 + 2G ) ⎦
⎣ ψ(k 0
...
where G0 = 2π/a is the basic reciprocal-lattice vector, and where k is in the first
Brillouin zone. Note that the number of rows and columns equals the number of
reciprocal-lattice vectors, that is, the number of lattice sites, N . Equation (6.83) has
nontrivial solutions when the determinant vanishes; this requirement determines the
values of the energy E for each k. These values are termed En (k); the corresponding
wave functions, ψnk (x), are obtained by inserting the solutions of Eq. (6.83) in Eq.
(6.78).
Another proof of the Bloch theorem. The solutions of the Schrödinger
equation, as obtained from Eq. (6.82) for a given k, are

ψnk (r) = ψn (k − G)ei(k−G)·r . (6.84)
G
They may be written in the form

ψnk (r) = eik·r ψn (k − G)e−iG·r ≡ eik·r unk (r) . (6.85)
G

The function unk (r) defined in the right hand-side obeys Eq. (6.53), unk (r) =
unk (r + R), which validates the Bloch theorem. It can be verified that

ψn (k + G − G )e−i(k+G−G )·r

ψn,k+G (r) =
G

ψn (k − G )ei(k−G )·r = ψnk (r)

= (6.86)
G
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 452

452 Introduction to Solid-State Physics

(the summation index is changed, G → G = G − G); this proves Eq. (6.63).
Equation (6.64) follows from the latter. Thus, it suffices to investigate the solutions
in the first Brillouin zone.
The weak-binding approximation. The complete solution of the
Schrödinger equation can be derived, in principle, from Eqs. (6.82). It is a
formidable task to solve those because of the huge number of equations; hence
approximation methods are called for. When the lattice is “empty”, i.e., U  (G) = 0,
 
Eq. (6.82) yields [ k /(2m) − E]ψ(k) = 0; consequently ψ(k) = 0 only when the
2 2

energy is E = E (0) (k) = 2 k 2 /(2m). Then all other coefficients in Eq. (6.84),
 − G) with G = 0, have to vanish, and so the sum in Eq. (6.84) comprises a
ψ(k
single term, ψk (r) = ψ(0) (k) √ exp[ik · r]. This is the wave function
√ of the free elec-
tron, Ψk (r) = exp[ik · r]/ V , implying that ψ(0) (k) = 1/ V . Suppose that the
(0)

periodic potential is very weak, such that it is a small fraction of the free-particle
energy. Specifically, |U  (G)| ≤ |U
 (G0 )| = λE (0) (G0 ), where λ is a small dimension-
less parameter, λ  1, which characterizes the weakness of the potential, and G0 is
the shortest reciprocal-lattice vector. [The first inequality follows from the integral
in Eq. (6.77): it is smaller when |G| is larger, due to the oscillations of the factor
exp[iG · r] in the integrand.] Barring exceptional cases, to be discussed separately,
the other coefficients in the sum are such that |ψ(k  − G)/ψ(k)|
 = O(λ), and hence
can be considered as small corrections to the wave function: both the wave function
(0)
ψ(r) and the energy E(k) are quite close to their free-particle values, Ψk (r) and
E (0) (k). For one of these other coefficients, ψ(k  − G) with G = 0, Eq. (6.82) reads

 − G) = −
[E (0) (k − G) − E(k)]ψ(k U  − G − G ) .
 (G )ψ(k
G

The right hand-side includes the term U 

 (−G)ψ(k), that comes from G = −G.
This is the largest term in the sum; all other terms are smaller as they contain the
 − G − G ), with G + G = 0. They thus
“other” coefficients of the solution, ψ(k
can be discarded, leading to
 − G) ≈ −U
ψ(k 
 (−G)ψ(k)/[E (0)
(k − G) − E (0) (k)] , (6.87)
where E(k) ≈ E (0) (k) is used. Assuming that the denominator is not too small, then
 − G)/ψ(k)
ψ(k  = O(λ), making the discarded terms of the order of λ2 . Inserting
Eq. (6.87) in Eq. (6.82) yields
  (G)|2
|U
E(k) = E (0) (k) − + ... , (6.88)
G
E (0) (k − G) − E (0) (k)
where . . . represent higher-order terms in λ. Finally, the assumption that none of the
denominators in the sum is too small implies that the correction to the energy due
to the periodic potential is of the order of λ2 ; thus it is indeed small. Higher-order
corrections are even smaller.
The weak-binding approximation near Brillouin-zone boundaries. The
expansion in Eq. (6.88) is valid as long as |U  (G)/[E (0) (k − G) − E (0) (k)]| < λ, but
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 453

Electrons in solids 453

it fails for pairs of wave vectors for which the difference E (0) (k − G) − E (0) (k) is
small [e.g., of the order of λE (0) (G0 )], because then the corrections to E(k) might
be larger than the intervals between the original energies. This energy difference
vanishes for
2
E (0) (k − G) − E (0) (k) = [(k − G)2 − k2 ] = 0 ,
2m
i.e., when 2k · G = G2 . This is, up to a sign, the condition for Bragg scattering,
Eq. (3.26), discussed also at the beginning of this section. The condition implies
that the energies of the states corresponding to k and to (k − G) are degenerate
whenever the tip of k is located on the plane perpendicular to a reciprocal-lattice
vector at its mid point, i.e., it is located on the border between two Brillouin zones.
The energy gaps in one dimension. In a one-dimensional lattice with a con-
stant a Eq. (3.26) is fulfilled at a single point for each reciprocal-lattice vector, i.e.,
when k = G/2 = π/a, where the “free states” with k and (k − G) are degenerate.
For such states, the energy E(k) is close to both E (0) (k) and E (0) (k − G), but is
rather far away from all other energies E (0) (k − G ), with G = G, 0. Considerations
similar to those leading to Eq. (6.87) yield that the coefficients ψ(k  − G ), with

G = G, are still rather small. Nonetheless, in the present situation there are two
coefficients that are not necessarily small, ψ(k)  − G). For these two special
and ψ(k
coefficients, Eq. (6.82) yields

[E (0) (k) − E(k)]ψ(k)  − G) = 0 ,
 (G)ψ(k
+U
 − G) + U
[E (0) (k − G) − E(k)]ψ(k 
 (−G)ψ(k) =0. (6.89)
All other, much smaller, coefficients are discarded. Equations (6.89) may be ob-
tained from Eq. (6.83) once only the first two rows and the first two columns there
are retained. In general, keeping only the rows and the columns related to degener-
ate energies gives a reliable approximation for the energy bands. [A similar result
is derived in perturbation theory for degenerate states; see Eq. (C.4) and problem
6.15.]
The two linear homogeneous equations (6.89) have a nontrivial solution when
the determinant of their coefficients vanishes. On the border of the Brillouin zone
E (0) (k) = E (0) (k − G) = E 0) (G/2); then this requirement yields two possible
energies
(1)  (G)| ,
E± (G/2) = E (0) (G/2) ± |U (6.90)
with a gap in-between them, 2|U  (G)|. In contrast to Eq. (6.88), which gives
2
corrections of the order λ to the energy, the corrections obtained here are far
larger, of order λ.
 (G) = U
As the potential energy is real, Eq. (6.77) implies that U  ∗ (−G). When
it is also even, U (x) = U (−x), then U  (G) = U (−G) [see Problem (B.1)]. Hence,
 
U (G) = U0 (G) is real, and the eigenvectors of Eq. (6.89) are 
√ such that ψ(−G/2) =
 ± (0) (0)
±ψ(G/2). The eigenfunctions are ΨG/2 = [ΨG/2 ± Ψ−G/2 ]/ 2L; these are precisely
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 454

454 Introduction to Solid-State Physics

the two standing-wave functions obtained qualitatively in the discussion of Bragg

reflections. Indeed, the potential energy “responsible” for these wave functions, for
G = G0 = 2π/a, is
 (G)(eiGx + e−iGx ) = 2U
U (x) = U  (G) cos(Gx) = 2U
 (G) cos(2πx/a) .
0 0 0

Assuming that U 0 (G) > 0, the minima of the potential are at the points x =
√
(n + 1/2)a; therefore the function Ψ−G/2 (x) = sin(πx/a)/ 2L has the lower energy.
Energies near Brillouin-zone boundaries. In the general case where k is
close to the border of a Brillouin zone, the eigenvalues of Eq. (6.89) are

(1)  (G)|2 /2 .
E± (k) = E (0) (k) + E (0) (k − G) ± [E (0) (k) − E (0) (k − G)]2 + 4|U
(6.91)
Denoting k = G/2 + δ an expansion in δ up to second order gives

(1) 2  G2   2 2
 (G)|2 +  Gδ
E± = + δ ± |U
2
2m 4 2m
2  
 2 G2  (G)| +   2 2
G
≈ ± |U 1± δ2 + . . . . (6.92)
8m 2m 2m|U  (G)|
(Show that the approximation is valid only when
 (G)|2 /E (0) (G) = O[λ2 E (0) (G)] ,
2 δ 2 /(2m)  U
i.e., when δ  λG, in the proximity of the point k = G/2.) Thus, an energy
gap of width 2|U  (G)| opens at k = G/2. Around each of the “new” energies
(1)
E± (G/2) =  G /(8m) ± U
2 2 0 (G) the energy depends quadratically on δ. Equation
(6.92) gives an approximate expression for the energies near each border of the
Brillouin zones, that is, at the vicinity of the points k = G/2 = π/a. Away from
these points, the degeneracy may be ignored, and a good approximation for the
energies is the form (6.88).

Problem 6.15.
Show that Eqs. (6.88) and (6.89) are derived directly from the perturbation theory,
as summarized in Appendix C.

Problem 6.16.
An alternative way to solve problem 6.14, with the potential U (x) = u0 n δ(x −

na): calculate the Fourier transform of the wave function, ψ(k), using Eq. (6.82).

[Hint: Derive an equation for the function f (k) = G ψ(k − G), and exploit the
∞ −1
mathematical identity cotz = n=−∞ (z − nπ) , documented in mathematical-
formulae books.]

Problem 6.17.
Use the weak-binding potential to derive the energy gaps for the Kronig-Penney
potential, Eq. (6.65). Confining yourself to the delta-function limit, compare the
results with those of problem 6.14.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 455

Electrons in solids 455

Figure 6.12(a) compares the energy of a free particle with the energy
corresponding to a periodic potential described by the shortest reciprocal-lattice
vector G0 = 2π/a. The figure displays the jump between the two energies on the
boundary of the first Brillouin zone, Eq. (6.90). Similar jumps appear at the other
boundaries of the Brillouin zones, as seen in Fig. 6.12(b). In each case, the behavior
near k = G/2 is derived from Eq. (6.92). For the sake of clarity, the parameters are
chosen such that all gaps are equal, 2U  (2G0 ) = 2U
0 (G0 ) = 2U  (3G0 ) = 0.6 [in en-
ergy units of E (G0 /2) =  π /(2ma )]. In general one expects U
(0 2 2 2 0 (G) to decrease
as G increases, due to the oscillations of the factor exp[iGx] in the integrand of the
Fourier transform, Eq. (6.77).
Figure 6.12(b) illustrates the modifications in the parabolic energy of the free
particle stemming from the weak periodic potential, for all k values −∞  k  ∞.
For each value of k (except the boundaries of the Brillouin zones) there is a single
value of the energy, derived from the original free-particle energy with an additional
small correction. This description is termed the “extended spectrum”.

(a) (b)

Fig. 6.12: (a) The dashed line is the parabolic energy of a free particle in one
dimension, as a function of k, E (0) (k) = 2 k 2 /(2m), in units of E (0) (G0 /2) =
2 π 2 /(2ma2 ). The solid lines are the energies of a particle moving in a periodic
potential described by the reciprocal-lattice vector G0 = 2π/a, with U 0 (G0 ) = 0.3
(1)
(in the same units). These lines are computed from Eq. (6.92), E− (k) for k < π/a,
(1)
and E+ (k) for k > π/a. (b) The dashed line illustrates the energy of the free
particle, while the solid lines display the energies close to additional boundaries
of Brillouin zones, due to periodic potentials that correspond to other reciprocal-
0 (2G0 ) = U
lattice vectors, with U 0 (3G0 ) = 0.3. The behavior at k < 0 is obtained
by reflection, E(−k) = E(k).

The reduced spectrum. The Bloch theorem states that it is impossible to

distinguish between the energies in the different Brillouin zones, see Eq. (6.64).
Indeed, once the energies in Fig. 6.12(b) are displaced by a reciprocal-lattice vec-
tor to the right or to the left, the resulting figure resembles Fig. 6.11(a). In the
limit of a very weak periodic potential the figure resembles Fig. 6.10(a), which at
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 456

456 Introduction to Solid-State Physics

times is referred to as the empty-lattice approximation. Figure 6.13(a) repro-

duces Fig. 6.10(a) for a very weak potential. As mentioned in the context of Fig.
6.10(a), this approximation is plausible provided that one keeps in mind that it is
an approximation of Fig. 6.11(a), which displays the spectrum for a weak periodic
potential.
Figures as Fig. 6.10(a) or Fig. 6.11(a) comprise numerous energy levels for each
value of k. Such figures contain many repetitions that convey the same information
numerous times. To avoid this extensive redundancy one may employ the extended
spectrum, Fig. 6.12(b). Another option in which each energy level appears only
once is presented in Fig. 6.13(b) (showing the spectrum in the empty-lattice ap-
proximation). In this figure all energies of the extended spectrum are moved into
the first Brillouin zone. The energy levels of the extended spectrum are in one-
to-one correspondence with the shifted ones in the first Brillouin zone. The latter
description is called the reduced spectrum. For each value of k, the first Bril-
louin zone contains many energy levels, shifted from other Brillouin zones. These
are distinguished by the index n which counts the bands, and are denoted En (k).
The function corresponding to the nth energy band is obtained upon displacing the
energy from the nth Brillouin zone into the first one,
En (k) = E[k − 2(n − 1)π/a] , or En (k) = E[k + 2(n − 1)π/a] . (6.93)
The dashed curve in Fig. 6.13(b) is the parabola E (k) =  k /(2m), that is,
(0) 2 2

the extended spectrum of the free particle. Figure 6.13(a) comprises all parabolas
displaced by reciprocal-lattice vectors (as compared to the original one). When
E (0) (k) < 2 (π/a)2 /(2m), the wave number k is within the first Brillouin zone,
(0)
and the corresponding energy is E1 = E (0) (k) = 2 k 2 /(2m), again within the
first Brillouin zone. When  (π/a) /(2m) < E (0) (k) < 2 (2π/a)2 /(2m), the wave
2 2

number k is in the second Brillouin zone, and consists of two segments, π/a < k <
2π/a and −2π/a < k < −π/a. These segments are moved into the first Brillouin
zone by k ⇒ k − 2π/a and k ⇒ k + 2π/a, respectively. The corresponding shifted
(0) (0)
energies are E2 (k) = 2 (k − 2π/a)2 /(2m) and E2 (k) = 2 (k + 2π/a)2 /(2m),
respectively. In a similar fashion, the k values in the nth Brillouin zone are displaced:
(0)
k ⇒ k − 2(n − 1)π/a and k ⇒ k + 2(n − 1)π/a, giving rise to the energies En (k) =
2 [k ± 2(n − 1)π/a]2 /(2m) in the reduced spectrum. The energies shifted into
the first Brillouin zone are presented by the thick curves in Fig. 6.13(b), and
the corresponding displacements of the wave numbers are marked by the arrows
there. The boundaries of the first Brillouin zone are the vertical lines at k = ±π/a,
and the borders between the various n values are the vertical dashed lines. The
figure illustrates that the spectrum can be presented in two equivalent forms: by
(0)
the original parabola E (0) (k) or by the energy levels En (k) in the first Brillouin
zone, with a one-to-one correspondence between the two. Figure 6.13(a) emphasizes
the periodicity of the energy levels in the reciprocal lattice; the information on
the electronic energies is contained within the first Brillouin zone. One therefore
(0)
confines oneself to the energies En (k) in the first Brillouin zone, and examines the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 457

Electrons in solids 457

corrections caused to them by the periodic potential. These corrections are crucial
(0)
when the original levels are degenerate, i.e., when the curves En (k) cut each other.
In Fig. 6.13(b) such cuts occur at the origin or at the edges of the Brillouin zone.
At each of them the periodic potential causes the appearance of an energy gap,
with the energies nearby varying quadratically as a function of the distance from
the cut.

(0)
Fig. 6.13: (a) The periodic functions En (k) in several Brillouin zones in the empty-
lattice approximation [see also Fig. 6.10(a)]. The vertical lines are the zone bound-
aries [the wave number k is measured in units of the shortest reciprocal-lattice
vector, 2π/a, and the energy is measured in units of E (0) (G0 /2) = 2 π 2 /(2ma2 )].
(b) The dashed parabola [E (0) (k) = 2 k 2 /(2m) in suitable units] is the free-particle
energy, in the absence of the periodic potential. This is the extended spectrum of
the particle in the empty-lattice approximation. The thick lines display the reduced
spectrum: these are the energies displaced by reciprocal-lattice vectors, from the
(0)
various parabolas of part (a). The index n “counts” the energy levels En (k) for a
given k value in the first Brillouin zone.

The energy bands in the empty-lattice approximation in two dimen-

sions. At higher dimensions the description of the spectrum is more involved,
as the free-particle energy, the paraboloid E (0) (k) = 2 k2 /(2m), cuts the bound-
aries of the Brillouin zones at various energies for various direction of k. Consider
for instance the square lattice in the plane. Figure 6.14(a) shows this paraboloid
in the first Brillouin zone. As seen, the paraboloid meets the center of the face,
kM = (π/a, 0) at the energy E (0) (kM ) = 2 (π/a)2 /(2m), but touches the corner,
kK = (π/a, π/a) at twice that energy, E (0) (kK ) = 22 (π/a)2 /(2m). The notations
for these special points of the Brillouin zone are defined in Fig. 5.7(a). Figure
6.14(b) displays the equal-energy curves corresponding to this paraboloid. At the
center of the zone these curves are perfect circles, but as the radius becomes larger
than π/a, part of the circle penetrates into the second Brillouin zone. As noticed in
the one-dimensional case, the energies located within the second zone are part of the
second energy band. Indeed, exploiting the periodicity of the energies in reciprocal
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 458

458 Introduction to Solid-State Physics

(a) (b)

(c) (d)

Fig. 6.14: The energy in the first Brillouin zone [measured in units of E (0) (kM )
and k is measured in units of the shortest reciprocal-lattice vector, 2π/a]. (a) The
free-particle energy as a function of the two components of k. (b) Equal-energy
curves for part (a): the brighter areas represent lower energies. (c) The energy
 (kM ) = 0.2, U
of a particle in a weak periodic potential, U  (kK ) = 0.15. Note in
particular the modifications near the faces of the Brillouin zone: the energy surface
is orthogonal to the faces. (d) Equal-energy curves for part (c). Again notice that
the equal-energy curves are normal to the faces at the faces [see also Fig. 6.18(b)].

lattice enables one to bring them all into the first Brillouin zone. Figure 6.15(a)
reproduces Fig. 3.14(a), with different notations, (2a, 2b, 2c, 2d), for the various
parts of the second Brillouin zone. Displacements by basic reciprocal-lattice vec-
tors along the two axes transfer these parts into the first zone, as illustrated in Fig.
6.15(b). Figure 6.16(a) shows those parts of the original free-particle paraboloid
that belong to the second zone, and Fig. 6.16(b) exhibits them after being dis-
placed into the first zone, according to the scheme of Figs. 6.15. As seen, the
energies of the second band are minimal at the centers of the faces, and are max-
imal at the center of the zone. Figure 6.17 portrays the same construction of the
second band in the first zone, in terms of the equal-energy curves. It is convenient
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 459

Electrons in solids 459

Fig. 6.15: Mapping of the second Brillouin zone [panel (a)] into the first Brillouin
zone [in panel (b)].

(a) (b)

Fig. 6.16: (a) Parts of the original paraboloid in the second Brillouin zone (that
belong to the extended spectrum). (b) The same parts after being shifted into the
first Brillouin zone (these form the second band in the reduced spectrum).

at times to present the energy bands as a function of the wave vector along certain
lines in the Brillouin zone (similarly to the figures showing the frequencies of the
normal modes in Chapter 5). Such a path, pertaining to the square lattice, is dis-
played in Fig. 5.7. The energies of the first two bands along this path are shown in
Fig. 6.18(a).

Problem 6.18.
Find the criteria for mapping the third Brillouin zone into the first one for a free
particle on a square lattice. Display the equal-energy curves that belong to the third
band in the first Brillouin zone.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 460

460 Introduction to Solid-State Physics

(a) (b)

Fig. 6.17: (a) Equal-energy curves for the paraboloid of a free particle. The straight
lines are the boundaries between the Brillouin zones. (b) Equal-energy curves from
the second band after being shifted into the first Brillouin zone, as illustrated in
Fig. 6.15.

(a) (b)

Fig. 6.18: The first two energy bands of a square lattice, on special lines in the first
Brillouin zone, as defined in Fig. 5.7. (a) A free particle. (b) Particle in a weak
periodic potential [as in Fig. 6.14(c)].

Band structure for a weak potential. Once the first two bands of the
free particle are constructed, one can “switch-on” a weak periodic potential, and
explore where are there degenerate states that cause energy gaps between these
bands. A comparison of Fig. 6.14(a) (that illustrates the first energy band in the
first zone) with Fig. 6.16(b) (which displays the second energy band in that zone)
shows that the energies in the two bands are equal only on the perimeter of the
zone. As noticed, this is a general property of neighboring bands (that have the
same energies of the free particle on the zone boundaries). Indeed, the energies
of the bands coincide along the line connecting points M and K in Fig. 6.18(a).
The displacements of the two bands against one another created upon the opening
of the gaps is shown schematically in Fig. 6.18(b). These gaps are calculated in
problem 6.19. When the potential is weak the gaps are quite narrow, and then the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 461

Electrons in solids 461

minima of the second band, located at the centers of the first zone’s faces, are lower
than the maxima of the first band, located at the corners. Hence, as opposed to
the situation in one dimension, in dimensions higher than 1 a band may penetrate
into its neighbors (problem 6.13).
Figure 6.14(c) portrays the energy of the lowest band for a weak periodic poten-
tial (calculated in problem 6.19), and Fig. 6.14(d) displays the corresponding equal-
energy curves. Note the differences between these two figures and Figs. 6.14(a) and
(b) that present the bands for a free particle. These are in particular conspicuous
near the zone boundaries. Due to the opening of the gaps, the energy of the first
band is maximal there, and hence its derivative with respect to the component of
the lattice momentum normal to this curve vanishes. For the same reason, the
equal-energy curves are always normal to this boundary. This property of the spec-
trum is in particular visible in Fig. 6.18(b), that illustrates the energy along certain
lines in the Brillouin zone.

Problem 6.19.
a. Obtain an approximate equation for the energies near the zone boundaries
by exploiting the nearly-free electron model, for an arbitrary lattice in arbitrary
dimensions, when there are solely two degenerate states on each such boundary.
Use the result to compute the energy gap created at a point q0 on one of these
boundaries.
b. Calculate the energies at the two sides of the gap found in part (a) at q = q0 + δ,
for small |δ|. Discuss in particular the dependence of the energies on the normal
and tangential components (with respect to the boundary surface) of the wave vector.

6.7 The tight-binding approximation

c. Use the results to obtain roughly Figs. 6.14(c) and (d), and explain the differences
between your plots and Figs. 6.14(a) and (b).
d. Optional: the results above do not pertain to the corners of the Brillouin zone.
Explain why, and describe how should the bands be calculated there. Apply this
calculation to a square lattice, for which the only nonzero Fourier components are
 (kM ) = 0.2 and U
U  (kK ) = 0.15 (together with all other points obtained from those
by symmetry operations).

States built of atomic wave functions. Section 6.6 zeroes on the limit of a
weak periodic potential, such that the eigenenergies remain close to those of a free
particle, with minor corrections. When the binding potential-energy of the electron
to one of the ions is rather strong, the energy, to a good approximation, is found
by ignoring all other ions to which the electron is only loosely bound. That is, the
calculation is based on the state in which the electron is bound solely to its parent
atom, as in the isolated atom. This description is in particular useful when the ions
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 462

462 Introduction to Solid-State Physics

are relatively far away from each other. It is also suitable for electrons located in
the inner electronic shells, as these are quite close to the nucleus. Generally, the
potential energy is the sum of the potential energies arising from the attraction of
the electron to all ions; the Hamiltonian of a single electron is thus
2 2 
Ĥ = − ∇ + u(r − R) , (6.94)
2m
R

where u(r − R) = −e /|r − R| is the potential energy of the Coulomb attraction

2

between the electron and the ion at the lattice site R. (For simplicity, we ignore
here the screening of these potentials by the other electrons in the system.) This
Hamiltonian is an extension of the one of the molecule H+ 2 , Eqs. (4.12) and (4.13),
that contains two ions. Strictly speaking, the various states of the electron on a
single ion (or on all ions in the unit cell) have to be considered. This point is elab-
orated upon at the end of the section. Here is analyzed the simplest configuration
in which there is a single ion possessing a single state in each unit cell, ϕ(0) (r). (In
a similar fashion to the treatment of the hydrogen molecule, where only the ground
state of each ion is included.) Assuming also that the corresponding atomic energy
E (0) is not degenerate (save for the trivial degeneracy of the spin states), then when
the ions are far away from each other there are N degenerate states on the lattice.
The electron’s wave function is the atomic ϕ(0) (r − R) on each of the ions located at
each of the N lattice sites {R}, and the energy is the discrete energy of the electron
on that ion, E (0) ,
2 2
− ∇ + u(r − R) ϕ(0) (r − R) = E (0) ϕ(0) (r − R) . (6.95)
2m
The treatment of the H2+ molecule exploits the degenerate states of the electron
on each of the hydrogen ions, see Sec. 4.3. As in that case, one aims to find a
variational approximation for the wave function of the electron in the lattice, by
using an arbitrary linear combination of the “atomic” wave functions,

ϕ(r) = a(R)ϕ(0) (r − R) . (6.96)
R

This wave function is termed LCAO, which abbreviates “Linear Combination of

Atomic Orbitals”. A similar combination is encountered in the context of electrons
in the benzene molecule, Eq. (4.27).
The conditions for the Bloch theorem. The coefficients a(R) have to fulfill
certain requirements for the function (6.96) to comply with the Bloch theorem,
Eq. (6.54), ϕ(r + R) = exp[ik · R]ϕ(r). Displacing r by R, and changing the
summation index R → R + R in Eq. (6.96) give
 
ϕ(r + R) = a(R )ϕ(0) (r + R − R ) = a(R + R )ϕ(0) (r − R ) . (6.97)
R R

The change of summation index does not cause any harm, as (upon employing
periodic boundary conditions) the sum comprises N displacement vectors on the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 463

Electrons in solids 463

lattice. On the other hand, the second version of the Bloch theorem requires that
any solution of the Schrödinger equation with a periodic potential is described by
the quantum number k, such that
 
a(R + R )ϕ(0) (r − R ) = ϕ(r + R) = eik·R ϕ(r) = eik·R a(R )ϕ(0) (r − R ) .
R R
(6.98)
Adding k as an index to the function, ϕk (r), and to the coefficients, ak (R), in all
previous equations, and comparing the coefficients on both sides of Eq. (6.98) yield
the relation
ak (R + R ) = eik·R ak (R ) . (6.99)

For R = 0 one finds
ak (R) = eik·R ak (0) ≡ Ak eik·R , (6.100)
where Ak = ak (0) is a constant, and therefore

ϕk (r) = Ak eik·R ϕ(0) (r − R) . (6.101)
R

Normalization. The coefficient Ak is determined by the normalization of the

wave function,
 
1 = ϕk |ϕk  = |Ak |2 eik·(R−R ) ϕ(0) (r − R )|ϕ(0) (r − R) . (6.102)
R,R

The expression on the right-hand side includes the overlap integral between the two
wave functions at the two sites,

ϕ (r − R )|ϕ (r − R) = d3 r[ϕ(0) (r − R )]∗ ϕ(0) (r − R) ≡ α(R − R) ,
(0)  (0)

(6.103)
which depends only on the distance between the two ions, |R − R |, since the origin
within the integral can be shifted to the point R. This integral decays exponentially
with |R − R |, because |ϕ(0) (r)| decays with r. For instance, for the ground-state

wave function of the hydrogen atom, ϕ(0) (r) = ψ100 (r) = exp[−r/aB ]/ πa3B , Eq.
(4.16) yields
α(R) = e−R/aB [1 + R/aB + (R/aB )2 /3] . (6.104)
Replacing the dummy index R in the sum in the right hand-side of Eq. (6.102) by
R − R gives N identical sums, and hence
  
|Ak |−2 = N 1 + eik·R α(R) , (6.105)
R=0

where it is assumed that each of the original functions ϕ(0) (r) is normalized,
α(0) = ϕ(0) |ϕ(0)  = 1. When the ions are far away from each other the over-
lap of wave functions localized around two different ions may be ignored, and
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 464

464 Introduction to Solid-State Physics

|ak (R)|2 = |Ak |2 ≈ 1/N , is then roughly the probability to find the electron around
the ion located at R. Equation (6.100) then implies that these probabilities are all
equal, as can indeed be expected since the lattice is invariant under displacements
by lattice vectors.

Problem 6.20.
Use Eq. (6.101) to prove that the function ϕk (r) obeys the two versions of the Bloch
theorem, Eqs. (6.53) and (6.54). Prove that this function is periodic in reciprocal
space, Eq. (6.63).

In the case of a one-dimensional lattice with periodic boundary conditions, the

quantum numbers that characterize the wave functions are k = km = 2πm/L [Eq.
(6.60)], where L = N a. For N = 2 one finds km = 0, π/a, and hence there exist two
solutions, ϕ0 = A0 [ϕ(0) (x) + ϕ(0) (x − a)] and ϕπ/a = Aπ/a [ϕ(0) (x) − ϕ(0) (x − a)].
These solutions resemble the symmetric (bonding) and the antisymmetric (anti
bonding) ones obtained for the molecule H2+ , Eq. (4.14). Another relevant example
is that of the benzene molecule, also discussed in Sec. 4.3: each of the six carbons
there has four valence states. The electronic states in the molecule’s plane create
local covalent bonds, called sp2 . At low temperatures these are fully occupied and
do not contribute, neither to the electrical conductance around the ring, nor to
the specific heat. Each of the carbons has in addition an electron in the pz state,
whose wave function is “normal” to the molecule’s plane. These can serve as a basis
for single-electron molecular states as the one used in Eq. (6.101) [or Eq. (4.27)].
These six electrons occupy six states similar to those in Eq. (6.101), with equal
probability to occupy each of the carbon atoms. Each such state can accommodate
two electrons; there are, therefore, empty states to which the electrons can move
under the effect of external fields. The electrons in the pz states are capable of
moving around the ring and hence contribute, e.g., to the magnetic susceptibility
of the molecule.
The energy bands in the variational approximation. As argued in the
discussion of the hydrogen molecule, the average energy in the state (6.101), E(k) =
ϕk |Ĥ|ϕk  is an upper bound on the ground-state energy of the system; in the lowest
variational approximation it is the energy of the electron in the ground-state of the
periodic lattice. This energy is
 
E(k) = |Ak |2 eik·(R −R) ϕ(0) (r − R)|Ĥ|ϕ(0) (r − R )
R,R

= N |Ak |2 ϕ(0) (r − R)|Ĥ|ϕ(0) (r) . (6.106)
R

The same manipulations exploited in treating Eq. (6.105) are used: in the second
line, the origin in the integral is shifted,

ϕ(0) (r − R)|Ĥ|ϕ(0) (r − R ) = ϕ(0) (r − R + R )|Ĥ|ϕ(0) (r) ,

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 465

Electrons in solids 465

and the summation index is changed R → (R − R ). The terms in the sum are
independent of R , which then comprises N identical summands.
Next, the Hamiltonian is written in the form Ĥ = Ĥ0 + ΔU , where
2 2
Ĥ0 = − ∇ + u(r) (6.107)
2m
is the Hamiltonian of an electron bound solely to the ion located at the origin, and

ΔU = [u(r − R) + e2 /|R|] (6.108)
R=0

is the potential energy of an electron near R due to all other ions (save the one
at the origin), to which the repulsive energy between each such ion and the ion
at R is added. The wave function ϕ(0) (r) is a solution of the atomic Schrödinger
equation (6.95). Multiplying that equation from the left by [ϕ(0) (r − R )]∗ and then
integrating gives
ϕ(0) (r − R)|Ĥ0 |ϕ(0) (r) = E (0) α(R) , (6.109)
where α is defined in Eq. (6.103); hence,
β+ R=0 eik·R γ(R)
E(k) = E (0) − , (6.110)
1+ R=0 eik·R α(R)
with
γ(R) = −ϕ(0) (r − R)|ΔU (r)|ϕ(0) (r) , (6.111)
and
β = −ϕ(0) (r)|ΔU (r)|ϕ(0) (r) . (6.112)
The ground state of hydrogen is ϕ(0) = ψ100 , and Eqs. (4.18) and (4.19) show that
e2  R  −R/aB
γ(R) = 1+ e , (6.113)
aB aB
and
 e2  R  −2R/aB
β=− 1+ e . (6.114)
R aB
R=0

When R exceeds the Bohr radius, R  aB , then β  γ. This small ratio is found
also for heavier atoms. As mentioned, for sufficient long |R| the integrals α(R)
and γ(R) decay exponentially with |R|, due to the small overlap of wave functions
localized around different ions, in addition to the exponential decay of the atomic
wave functions themselves. The constant β decreases as the lattice constant is
increased, and therefore in the limit of a very large lattice constant the energy
bands are extremely narrow, with E(k) ≈ E (0) . As the distance between nearest
neighbors is diminished, the correction to this value gradually increases and the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 466

466 Introduction to Solid-State Physics

bands broaden. Since the integrals α(R) decay exponentially with the distance |R|,
one may expand the denominator in Eq. (6.110); retaining only the first-order,

E(k) ≈ E (0) − β − eik·R Γ(R) , (6.115)
R=0

where Γ(R) = γ(R) − βα(R) ≈ γ(R). For a Bravais lattice α(R) = α(−R) and
γ(R) = γ(−R) (check!). The summations in Eqs. (6.110) or (6.115) can then be
carried out pairwise, e.g.,
 
eik·R Γ(R) = cos(k · R)Γ(R) (6.116)
R=0 R=0

(the sum includes both R and −R, thus each term appears twice). Note the resem-
blance of Eqs. (6.115) and (5.35). In both cases there appear sums of the type dis-
played in Eq. (6.116). The difference is that Eq. (5.35) yields the frequency-squared
of the lattice vibrations, while Eq. (6.115) gives the energy of the electron. Nonethe-
less, this similarity might be useful: details of the calculations presented in Chapter
5 can be exploited to obtain energy bands in thetight-bindingapproximation.
Many bands. In the calculation presented above, it is assumed that each ion
has a ingle atomic energy level; upon adding the other ions in the crystal this level
creates an energy band. The isolated atom has many discrete energy levels, and the
calculation can be repeated for each of them. Since the wave functions of different
atomic states are orthogonal to each other the variational approximation gives rise
to separate energy bands for each atomic level, with wave functions built of the
atomic wave functions of that same level. Below, the energies within the energy
(0)
band obtained from the atomic level En are denoted En (k). These bands are very
narrow when the lattice constant is large (that is, the overlap of wave functions
of neighboring ions is small, which leads to a small |Γn (R)|), and they broaden
gradually as the ions approach each other. In many cases the sign of the coefficient
Γn (R) alternates for successive states of the ions. As a result, the energies of
neighboring bands have maxima or minima at the center of the first Brillouin zone,
as seen in Fig. 6.5.
Example: one-dimensional lattice. For a single-atom, one-dimensional
chain, with lattice constant a, Eq. (6.115) gives
∞

En (k) = En(0) − βn − 2 Γn (ma) cos(mka) (6.117)
m=1

for the nth band (the factor 2 results from m = ±|m|). As Γn (ma) decays sharply
with the distance between the ions ma, the summands in the sum may all be ignored
save those of the nearest neighbors. Then, denoting Γn (a) = Γn ,
En (k)  En(0) − βn − 2Γn cos(ka) = E n + 2Γn [1 − cos(ka)] . (6.118)
(0)
Except for the constant E n = En − βn − 2Γn , that shifts the band’s location on
the energy axis, the last term in Eq. (6.118) is the same as the one that appears
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 467

Electrons in solids 467

in Eq. (5.5) for the dispersion relation of a one-dimensional, single-atom chain

with nearest-neighbor spring constants. Equation (5.5) pertains to the square of
the frequency and therefore Fig. 5.2 displays its root. In the present case there
is no such root; hence, leaving aside the constant E n , the energy in each band
is proportional to 1 − cos(ka) = 2 sin2 (ka/2). Figure 6.5(a) shows three bands
described by this expression. Γn of the first and the third bands is positive, and
that of the second is negative. It is worth noting that near the bottom of the lower
band the energy is approximately E1 (k) ≈ E 1 + Γ1 k 2 a2 . It is parabolic in k, as is
the energy of a free particle [E = (k)2 /(2m)], but with a coefficient that may be
very different from that of a free particle. Nonetheless, it is customary to denote
Γ1 = 2 /(2m∗ a2 ), where m∗ is the effective mass of the electron. The meaning
of this “mass” is discussed in the following.
Wave functions. It is of interest to examine the wave function that cor-
responds to each of the energies in the band. From Eq. (6.101), ϕk (r) =
Ak m exp[ikma]ϕ(0) (r − ma), where the sum runs over the unit cells, R = ma.
An example is a linear chain of hydrogen ions for which it is assumed that
ϕ(0) (r) = ψ100 . This function appears in the sum multiplied by an oscillating phase
factor, and therefore the sum is larger for smaller wave numbers. Figure 6.19 shows
the real part of the wave function and its absolute value squared on the x−axis (on
which the ions are located), for several values of k = 2π /(N a), for a lattice of 10
ions. For = 0 the wave function is real, and oscillates between its value on the ions
and the one in-between them. For = 0 the real and the imaginary parts of the
wave function alternate in sign, but the absolute value (and hence the probability to
find the electron) oscillates periodically. As seen, the probability to find the electron
at the vicinity of any of the ions is equal (about 0.1 in the figure). The probability
to find it at the mid-point between two neighbors is maximal in the ground state,
= 0. This probability decreases as increases, and vanishes at the middle of the
band ( = 5), where the wave function changes sign at nearest neighbors (compare
with Fig. 4.9, which is similar to the case with two ions, N = 2).
Square lattice. When only the overlaps of the wave functions on nearest
neighbors are included, Eq. (6.115) for a square lattice reads
En (k) ≈ E n + 4Γn (a)[sin2 (kx a/2) + sin2 (ky a/2)] , (6.119)
(0)
where E n = En − βn − 4Γn (a) (check!) Again it is useful to compare Eq. (6.119)
with the last equation in (5.26), which gives the dispersion relation of vibrations
normal to the plane in a square lattice. Figure 6.5(b) displays three energy bands
described by Eq. (6.119), as a function of the two components of the vector k. A
parabolic behavior is observed near the center of the Brillouin zone for Γ1 > 0,
E1 (k) = E 1 + Γ1 k2 a2 ≡ E 1 + (k)2 /(2m∗ ). The equal-energy curves for each
band are similar to those presented in Fig. 5.17(a): circles near the center of the
Brillouin zone and its corners, and a square connecting the mid-points on the faces
of the zone in-between these two ranges, see Fig. 6.20(b). [In the middle of the
band cos(kx a) + cos(ky a) = 0, which implies that ky a = ±π − kx a. These are
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 468

468 Introduction to Solid-State Physics

Fig. 6.19: Wave functions (their real parts and their absolute values squared) on
the lattice axis for a few values of the wave number k = 2π /(N a). The lattice
comprises N = 10 hydrogen ions (with periodic boundary conditions) and the lattice
constant, for the sake of the demonstration, is a = 3aB .

the equations of the straight lines that build the square whose edges connect the
mid-points on the faces of the Brillouin zone.] Like in Fig. 6.18, the energy along
the path Γ → M → K → Γ is presented in Fig. 6.20(c). It is illuminating to
compare Fig. 6.20(b) with Fig. 6.14(b), that displays the lower energy band of a free
particle. The equal-energy curve is deformed away from the center of the zone, so
that it gradually attains the square symmetry of the lattice. A similar deformation
is found in the nearly-free electron model, as calculated in problem 6.19: there,
the sharp maxima at the corners of the Brillouin zone are diminished and become
smoother due to the periodic potential, leading to a paraboloid behavior near the
maxima. The spectrum of the lattice vibrations behaves similarly, see Sec. 5.2. In
particular, the dependence of the energy on the vector k is sensitive to its direction.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 469

Electrons in solids 469

0.4

0.2

0.0

-0.2

-0.4

-0.4 -0.2 0.0 0.2 0.4

(a) (b)

Fig. 6.20: (a) An energy band in the square lattice, according to Eq. (6.119) (with
overlaps between nearest neighbors alone). (b) Equal-energy curves for the same
band. (c) The energies along the path Γ → M → K → Γ in the Brillouin zone.

Figure 6.20 illustrates the following features. (1) The derivative of the energy with
respect to the component of k normal to the face of the Brillouin zone vanishes on
that face. (2) At the center of the face the energy has a saddle point: it is minimal
parallel to the face, and maximal normal to it. (3) The energy is maximal at the
corners of the Brillouin zone, and can be presented there in the quadratic form
E ≈ Emax − Γ1 (k − G11 )2 a2 where G11 = (±π/a, ±π/a). Similar results are found
in the nearly-free electron model, see problem 6.19.

Problem 6.21.
Consider a one-dimensional lattice of lattice constant a, with overlap energy Γ1 > 0
for nearest neighbors and Γ2 = −ζΓ1 for the next nearest-neighbors. The bottom of
the band is at the energy E 0 = Γ1 .
a. Plot schematically the energies in the first Brillouin zone for ζ = 1/8 and for
ζ = 1/4.
b. As in Eq. (6.31), the density of states is defined such that Lg(E)dE equals the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 470

470 Introduction to Solid-State Physics

number of discrete levels in the range {E, E + dE} [note though, that this density of
states is defined per unit length, not like the definition in Chapter 5, e.g., after Eq.
(5.56)]. Derive a general expression for the density of states, and draw schematically
g(E) for the present case.
c. Assume that the Fermi level of the electron gas, EF , is such that EF − Γ1  Γ1 .
Find the total energy of the gas at zero temperature, and compare it with the result
for a free electron gas.

Problem 6.22.
a. Find the wave function in the ground state of an electron in a one-dimensional
potential well described by a delta function, U (x) = −u0 δ(x), u0 > 0.
b. A one-dimensional system is described by a periodic Kronig-Penney potential,
U (x) = −u0 m δ(x − ma). Exploit the result of part (a) to obtain the width of the
lowest energy band in the tight-binding approximation.

Several states in the unit cell. The calculation above produces a single
energy band because it assumes a single quantum state on each atom, and a single
atom in each unit cell. Quite often, however, the atomic energy levels are degenerate;
for instance, three orbital states, according to the quantum number m, belong to
the 2p atomic shell, and each such state can be occupied by two electrons with
opposite spins. It follows that the energy of the 2p state is six-fold degenerate.
Moreover, as encountered in chapter 4, the 2s states are quite close to the 2p ones,
and may hybridize with them to form eight states in the second shell. In addition,
the unit cell may well contain more than one atom.
In order to account for such configurations, the wave functions required for the
description of all the states within a single unit cell centered around the lattice site
(0
R are denoted as ϕiλ (r−R), where the index λ represents the degenerate (or almost
degenerate) states on the atom located at ri in the unit cell. The corresponding
(0)
eigenenergies are Eiλ . The solutions of the Schrödinger equation are written as
[extending Eq. (6.96)]
 (0)
ϕ(r) = aiλ (R)ϕiλ (r − R) . (6.120)
R i,λ

Inserting this form into the Schrödinger equation, multiplying it on the left by
[ϕi λ (r − R )]∗ and integrating the result over r yields
(0)

 
αi λ ,iλ (R − R)aiλ (R) = − γi λ ,iλ (R − R)aiλ (R) ,
(0)
(E − Ei λ )
R i,λ R i,λ
(6.121)
where
−γi λ ,iλ (R − R) = ϕi λ (r − R )|ΔUi (r)|ϕiλ (r − R) ,
(0) (0)

with ΔUi (r) comprising all interactions with other ions, and where αi λ ,iλ (R −
R) = ϕi λ (r − R )|ϕiλ (r − R) are the overlap integrals (including for brevity the
(0) (0)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 471

Electrons in solids 471

term R = R , treated separately before). For nB atoms in the unit cell, each pos-
sessing nS states, there are N nB nS linear equations for all the coefficients {aiλ (R)},
and these have a nontrivial solution when the determinant vanishes. Again gener-
alizing the previous discussion, it is noted that according to the Bloch theorem
aiλ (R) = Aiλ exp[ik · R]. Then
 
αi λ ,iλ (R − R)Aiλ eik·(R −R)
(0)
(E − Ei λ )
R i,λ
 
=− γi λ ,iλ (R − R)Aiλ eik·(R −R) . (6.122)
R i,λ

In this way, separate equations are obtained for the nB nS coefficients {Aiλ }. For
nB = nS = 1 there is solely a single equation, that reproduces Eq. (6.110).

Problem 6.23.
Calculate the energy bands of a one-dimensional lattice with two atoms in
the unit cell, A at the origin and B at the point x = a/2. The lattice constant
is a. Each atom when “free” has a single energy level; the wave function of an
(0) (0)
electron on A is ϕA (x − ma) and the energy is EA , and the wave function on B
(0) (0)
is ϕB (x − ma − a/2) and the energy is EB . Ignore the overlap integrals α between
different atoms, and keep only the overlaps γ of nearest neighbors.

Problem 6.24.
a. Calculate the energy bands of graphene which are formed from atomic states of π
type (wave functions that are normal to the plane of the lattice). Neglect the overlap
integrals α and take into account only the overlaps γ between nearest neighbors.
b. Show that at the vicinity of each corner in the Brillouin zone, e.g., the point
marked K in Fig. 5.9(a), the energies of the two bands appear as two cones joined
at their apexes.

Problem 6.25.
Consider the following way to derive Eq. (6.115). The Hamiltonian of a single
electron in the lattice is given by a matrix whose diagonal elements are R|Ĥ|R =
(0)
E = E (0) − β and the nondiagonal ones are real and depend on the distance
between the atoms, R|Ĥ|R  = R |Ĥ|R = −γ(|R − R |), for R = R .
a. Under which assumptions does this matrix reflect the tight-binding approxima-
tion?
b. A one-dimensional chain contains N atoms at the sites R = nax̂, n =
1, 2, 3, . . . , N ; γ(n − m) = γδn−m,1 . Assuming periodic boundary conditions,
ψ(n) = ψ(n + N ), show that the eigenstates and the eigenvalues of the matrix
√ (0)
are ψ (n) = exp[ink a]/ N a, E(k ) = E − 2γ cos(k a), where k = 2π /(N a)
and = 0, 1, 2, . . . , N − 1. This result coincides with Eq. (6.118).
c. A foreign atom is located at the origin (n = 0) of the one-dimensional lattice
discussed in part (b). Assume that the chain is infinite, N → ∞ (see also Sec.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 472

472 Introduction to Solid-State Physics

5.6). The matrix elements of the Hamiltonian are those of part (b), except for the
(0)
elements 0|Ĥ|0 = E imp , 0|Ĥ|1 = 1|Ĥ|0 = −γimp . Show that in addition to
the wavy solutions, the Schrödinger equation for this Hamiltonian can possess a
localized solution, of the form ψ(n) = A exp[−κ|n|] for |n| ≥ 1, where | exp[κ]| > 1,
and find the conditions for the existence of such a solution.

6.8 Metals, insulators, and semiconductors

Introduction. The previous three sections include various examples of electronic

band-structure calculations. An exact solution of the Kronig-Penney model is pre-
sented in Sec. 6.5; the main advantage of this model is the possibility to derive
energy bands for any strength of the potential that binds the electron to a certain
ion in the lattice, U0 . The band structure in the empty-lattice limit is displayed
in Fig. 6.10(a): a structure with no energy gaps, i.e., the bands are touching each
other. A small U0 gives rise to the energy structure shown in Fig. 6.11(a) with
narrow gaps in-between the bands. A similar result is depicted in Fig. 6.12(b),
calculated within the nearly-free electron approximation, or the weak-binding ap-
proximation. As U0 increases, the gaps widen while the bands’ widths narrow down.
An intermediate configuration is illustrated in Fig. 6.8(b). Very narrow bands are
found for large values of U0 , as in Fig. 6.11(b). Such bands are also derived in the
tight-binding approximation, in Sec. 6.7. As U0 approaches infinity, the bands are
infinitesimally narrow, see Fig. 6.10(b).
The calculations within the weak-binding approximation produce expressions
for the energy gaps, and clarify the reason for the widening of those gaps with
the increasing strength of the attractive potential. As opposed, calculations within
the tight-binding model focus on the bands themselves and show that the bands
widen as the overlap of wave functions on nearby neighbors increases. This width is
defined as the difference between the maximal and the minimal energy of the band,
Wn = En (max) − En (min). In one-dimension, and when only the overlap of nearest
neighbors is accounted for, Eq. (6.118) gives Wn = 4Γn ; as shown, Γn decays
exponentially with the distance between nearest-neighbor ions. In two dimensions
Eq. (6.119) yields Wn = 8Γn , and in the general case [Eq. (6.115)] the band widens
as overlaps with more neighbors are included, and narrows down as a function of the
distance between the neighbors. Figure 6.21(a) illustrates the edges of the band for
a one-dimensional chain of hydrogen atoms, Eq. (6.118), with ka = 0 and ka = π
[the figure exploits the explicit expressions given in Eqs. (6.104) and (6.113)]: the
energy band shrinks as the distance between nearest-neighbor ions is increased.
In higher dimensions the energies involve all components of k, see e.g., Fig.
6.5(b). However, even there the energies are independent of k when there are a few
atoms located very far away from each other, and the bands widen gradually until
touching each other or even penetrating into one another [see e.g., the square lattice
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 473

Electrons in solids 473

with a weak potential, discussed at the end of Sec. 6.6]. Figure 6.21(b) illustrates
the bands of neon and sodium (in three dimensions) as a function of the distance
in-between the ions. As seen, the bands are narrow and are well separated when
the ions are far away from each other; as these distances are shortened, certain
gaps are closed, and there is even an overlap between the bands. Recall that such
overlaps are impossible in one dimension, but may well appear in higher dimensions.
In particular, the overlap between the 2s band and the 2p one is related to the
hybridization between these two types of states, as discussed in length in Sec. 4.3.

(a)

(b) (c)

Fig. 6.21: (a) The boundaries of the energy bands of a one-dimensional chain,
comprising hydrogen atoms in their ground state, with overlap between nearest
neighbors, as a function of the distance between those neighbors (distances in units
of the Bohr radius, energy in units of the ground-state energy of the hydrogen atom).
(b and c) A schematic illustration of the energy bands of neon (b) and sodium
(c). The vertical dashed lines indicate the measured distance in the materials, and
the horizontal ones are located at the Fermi energies.

Filling of energy bands. At zero temperature, when the electron gas is in

its ground state, the electrons occupy all energy levels (two of them in each level),
starting at the lowest one, until their number is exhausted. The upmost filled level
is termed the Fermi level. In contrast with the Sommerfeld picture (Sec. 6.3), in
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 474

474 Introduction to Solid-State Physics

which the energy of the levels increases with the wave number, in the periodic solid
the energy levels form bands separated by gaps. As seen in Eq. (6.27), there are
precisely N = N1 N2 N3 discrete values of lattice momentum k in the first Brillouin
zone. The number of k values is the number of quantum states in each band: there
is a wave function and an energy level for each discrete value of k. As a quantum
state can accommodate two electrons, each band (whose “parent” is a single atomic
level) can accommodate at most 2N electrons. In the example of neon illustrated
in Fig. 6.21(b), each atom donates 10 electrons. The lowest band, marked by 1s2 ,
indeed contains 2N electrons, and hence is full. Similarly, the band denoted 2s2
contains 2N electrons as well, and is completely full too. As the three atomic levels
in the 2p shell are degenerate, the band based on them can contain 6N electrons. All
in all, the lowest three energy bands of neon are fully occupied with electrons, and
the Fermi level is thus in the gap between the 2p6 band and the empty band above
it, corresponding to the atomic level 3s. In contrast, each sodium atom donates 11
electrons. Like in the case of neon, 10 electrons of each atom fill the three lowest
bands. The remaining N electrons (one from each atom) occupy the band based on
the 3s states. This band is half filled, and the Fermi level, marked by the dashed
line in Fig. 6.21(c) is at the middle of this band. The shaded areas in Figs. 6.21(b)
and 6.21(c) indicate the energy levels occupied with electrons.
Metals and Insulators. When each atom in the lattice contributes an odd
number of electrons, i.e., 2M + 1, the number of electrons in the lattice is Ne =
N (2M + 1). Assuming that the energy bands do not overlap (i.e., the atomic states
are not degenerate), each band contains up to 2N electrons. In the ground state,
each of the lowest M bands is filled. These bands are termed “valence bands”; the
electrons residing in them essentially do not contribute to physical processes, mainly
because they do not have nearby (in energy) empty states to move into (e.g., either
because the temperature is increased, for a determination of the specific heat, or
an electric field is applied for a study of the conductivity). As seen in the examples
displayed in Figs. 6.21(b) and 6.21(c), the electrons in the valence bands are bound
mainly to inner atomic shells, and thus remain bound to the atoms that donated
them. The overlap of the “inner” wave functions is small, and the corresponding
bands are narrow. The remaining N electrons occupy the (M + 1)th band. This
band is only half filled, as it can accommodate 2N electrons. This is the situation
in sodium, shown in Fig. 6.21(c). The upmost band, which is not fully occupied, is
called the “conduction band”. The electrons in it usually come from the outmost
atomic shell. As the band is half filled, there are empty states nearby (in energy)
to the occupied ones; this implies that the behavior of the specific heat and the
electrical conductance of the material are similar to those analyzed in Sec. 6.3.
The calculations have to be modified, though, to include the dispersion relation of
the electrons, E(k), and the relevant density of states that are determined by the
band structure. This task is carried out in Sec. 6.9. However, as explained below,
the electrons in the conduction band are relatively free to move over the lattice,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 475

Electrons in solids 475

rendering the material to be a metal.

A different result emerges when the number of electrons in each atom in the
lattice is even, 2M . The total number of electrons is then Ne = 2M N , and they
fully occupy M valence bands, leaving an energy gap in-between the electrons in
the upmost band and the next one, which is the conduction band. This is the
configuration of neon, Fig. 6.21(b). As at zero temperature there are no nearby (in
energy) empty states to which the electrons can move, such materials are usually
insulators: their electrical conductance is zero at zero temperature. The energy
bands of an insulator and a metal are illustrated in the right and left panels of Fig.
6.22.

Fig. 6.22: Schematic illustration of energy bands’ occupations, depicted by the

rectangles, at low temperatures. The grey areas are filled with electrons, and the
white ones are empty. The energy increases upwards. The upmost rectangles show
the conduction band, and the lower ones the valence bands. The dashed line is at
the Fermi energy. Right: insulator; middle: semiconductor; left: metal.

Semiconductors. Assuming that at zero temperature the upper valence band

is fully occupied and the conducion band above it is empty, as in the right panel of
Fig. 6.22, there is a finite probability to excite electrons from the upper part of the
upmost valence band to the bottom of the conduction band; these electrons become
conducting. Since they evacuate states in the valence band, the electrons remaining
there can also move around and hence become conducting as well. This latter
motion can be described as that of holes, which represent the empty states. At low
temperatures the number of electrons that pass between the bands is quite small; it
is proportional to exp[−Eg /(2kB T )], where Eg = Ec − Ev is the energy gap between
the bottom of the conduction band, Ec , and the top of the valence one, Ev . For
typical insulators Eg is of the order of several electron volts, and therefore is much
larger than the thermal energy at room temperature; the number of electrons in
the conduction band is minute, leading to a negligible conductance. The material
is still termed an “insulator” even at finite temperatures. Semiconductors, as
opposed , are insulators for which Eg is relatively small (parts of electron volts as
compared to several electron volts). At room temperature there are electrons at the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 476

476 Introduction to Solid-State Physics

bottom of the conduction band, that moved there from the top of the valence band,
as in the central panel of Fig. 6.22. The details of the electrical conductance in
this configuration are discussed in Sec. 6.9. Contrary to the Matthiessen rule [Eq.
(6.49)] that predicts a decrease of the conductance as the temperature is raised,
the one of a semiconductor increases with the temperature, since there are more
free charge carriers at higher temperatures. The distinction between an insulator
and a semiconductor is not clear cut: the electrical conductance of both materials
vanishes at zero temperature and increases as the temperature is raised; but in the
insulator this increase is rather slow, such that the conductance is tiny even at room
temperature.
Intrinsic and extrinsic semiconductors. In a semiconductor as described
above, the number of electrons in the conduction band balances precisely the number
of empty states (called “holes”) in the valence band. This is so since at zero
temperature the valence band is fully occupied, and the conducion band is totally
empty. This type of semiconductors is called “intrinsic”. Another way to inject
electrons into the conduction band is to replace part of the atoms in an insulator,
that have 2M electrons per atom, by other atoms (“impurities”), which have 2M +1
electrons each. These atoms are called “donors”, since they add surplus electrons
to the states in the conduction band. The added charge carriers are negative; for this
reason such a semiconductor is called n-type. Alternatively, part of the atoms may
be replaced by atoms with 2M −1 electrons. In this case, electrons from the upmost
valence band can move to these impurities, leaving behind them empty states in
the valence band. These foreign atoms are called “acceptors”, as they capture
electrons and create “holes” in the valence band. The holes behave as positive charge
carriers; the semiconductor is then termed a p-type semiconductor. Both the n-
type and the p-type semiconductors are called extrinsic. As shown in Sec. 6.9, the
energy levels of the donors (or acceptors) are close to the conduction band (or the
valence one), and electronic transitions between them and the bands significantly
enhance the conductance.
Double-valence metals. The above description suffices to explain the metal-
licity of crystals built of single-valence atoms, like the alkali metals (sodium, potas-
sium), copper or silver. Each atom donates a single electron to the conduction band,
which is then half filled (see the left panel of Figs. 6.22). The electrons can reach
quite easily empty states and thus contribute to, e.g., the electrical conductance and
the specific heat. This picture remains valid also for numerous elements that con-
tribute an even number of electrons, like diamond, silicon, or germanium, in which
the conduction band is empty at zero temperature, and is separated by a finite gap
from the valence band below. Such materials are insulators (or semiconductors).
Nonetheless, solids comprising double-valence atoms are at times more complicated.
Figure 6.22 illustrates the simplest configuration, where the bands do not overlap
(as indeed happens in one-dimensional lattices). At higher dimensions the bands
widen as the distance between atoms is diminished (as occurs, e.g., in the tight-
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 477

Electrons in solids 477

binding approximation); it is hence plausible that two neighboring bands overlap

with one another, as shown in Fig. 6.21(b). The gap in-between them closes, and
there appear degenerate states in the two bands that may hybridize. This situation
occurs, for example, in certain systems containing carbon, due to the hybridization
of 2s states with 2p ones (see Sec. 4.3). Moreover, in dimensions higher than 1 the
atomic states are frequently degenerate (e.g., due to different values of the quantum
number of the angular momentum). In both cases, the number of electrons available
to populate the joint band is still even, and is 2N gM , where gM is the number of
degenerate states of each original atomic energy level, or the number of overlapping
bands. For gM > 1 the 2N electrons that can occupy the conduction band from
double-valence atoms, like the alkaline earths (calcium, barium, and strontium) or
iron, fill the conduction band only partially, leading to a metallic character of these
materials.
At dimensions higher than one it may happen that two neighboring bands, which
do not touch one another, contain states of identical energies, as illustrated in Fig.
6.23(a). A hypothetical case with this configuration is discussed in problem 6.13.
It can occur, e.g., in a square lattice, for which the lower band is separated from
the upper one at the middle of the face of the first Brillouin zone, e.g., at (π/a, 0);
the lower band though, possesses states on the diagonal of the Brillouin zone; an
energy gap is created only around the corner (π/a, π/a), see Figs. 6.14, 6.16, and
6.18. Within the weak-potential approximation, the second band, illustrated in Fig.
6.16(b), digs a “dip” in the first band along each edge of the Brillouin zone. The
energy levels along the edge are displayed schematically in Fig. 6.23(b). With the
Fermi energy as given by the horizontal line in the figure, electrons occupy states
in the two bands, leaving empty states in the upper parts of both bands. Such a
material, though metallic, is called semimetal.
The periodic table. Figure 6.24 presents the periodic table; the various colors
indicate the different types of crystals built of the corresponding elements. The
alkali atoms in the first column have a valence of 1 and create metallic crystals
(save hydrogen, which under standard ambient conditions forms a gas comprising
diatomic molecules, and is an insulator). The alkaline earths in the second column
have a valence of 2, but their bands overlap significantly, and therefore they create
metallic crystals as well. The same is true for the transition metals in which the inner
electronic shell nd is full, but they donate electrons from the (n + 1)th electronic
shell to the conduction band. At the other end of the table, the lattices formed
by the noble gases are insulators, as each of the atoms possesses an even number
of electrons, that remain relatively close to their parent nuclei. It follows that the
overlap of their wave functions is small [see the example of neon in Fig. 6.21(b)]
and the energy gaps are wide. The valence bands are completely full with an even
number of electrons and the conduction one is empty. According to the table, the
atoms in the upper right triangle are insulators. This fact is usually related to
details of the crystalline structure, and not only to the number of electrons in the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 478

478 Introduction to Solid-State Physics

outmost atomic shell. For instance, the halogens are insulators since the unit cell
of their crystalline structure contains an even number of atoms, and consequently
also an even number of electrons. Carbon is a borderline case: it has four electrons
in the outmost shell, and therefore should be an insulator, provided that the bands
do not overlap. Indeed, diamond is an insulator as it has a large energy gap in-
between the valence and the conduction bands. However, graphite is a semimetal,
since its valence bands touch one another (as is also the case for graphene, see
problem 6.24). The elements on the diagonal below the triangle (marked in blue)
are metalloids; their properties are partially metallic and partially semiconducting
(e.g., germanium and silicon). The other elements in the table crystalize in metallic
structures, for the aforementioned reasons.
The Fermi surface. Possible band structures are illustrated in Fig. 6.22. In
practice, however, the specific heat and the electrical conductivity are determined
by the number of electrons that can pass among the states near the Fermi level. In
lattice-momentum space, these states are at the vicinity of an equal-energy surface
whose energy is the Fermi one. It is called the Fermi surface. In contrast with the
Sommerfeld model, in which the electrons are free and therefore the Fermi surface
is spherical (referred to in Sec. 3.3 as the “Fermi sphere”), the periodic potential
entails a replacement of that sphere by a surface whose geometry may be rather
complicated (it is determined by the energy dependence on the lattice momentum).
As the energy bands are customarily pictured within the reduced spectrum, the
Fermi surface is also presented within the first Brillouin zone.
The Fermi surface of free electrons. In three dimensions, the Fermi momen-
tum of a free particle is given by n = kF3 /(3π 2 ) [Eq. (6.28)] while in two dimensions

(a) (b)

Fig. 6.23: A schematic description of the energy bands in a “semimetal”. (a) A

generalization of Figs. 6.22: the two upper bands are partially overlapping, and
both are not full. (b) The energy levels along an edge of the Brillouin zone of a
square lattice, in which a relatively weak periodic potential is active. The arrow
indicates a transition of an electron from the top of the lower band to the bottom
of the upper band, that penetrates into the lower one.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 479

Electrons in solids 479

Fig. 6.24: The periodic table. The colors distinguish between the various metals
and insulators. Created by the software Atomic PC.

(problem 6.10) n = kF2 /(2π). It follows that the radius of the Fermi sphere (or
circle) increases with the electrons’ density n. When the density is small enough,
the Fermi sphere is contained within the first Brillouin zone; as it increases, the
sphere cuts the borders of the zone and penetrates into the second Brillouin zone.
This is illustrated in Fig. 6.25(a) for the square lattice: when the valence of each
atom is√ (i.e., it donates electrons)
√ the electrons’ density is n = /a2 , and hence
kF a = √2π ; that is, kF a = 2π ≈ 0.798π for = 1 (curve A in the figure) and
kF a = 4π ≈ 1.128π for = 2 (curve C). The Fermi circle cuts the border of
the first Brillouin zone once kF a exceeds π (curve B). Thus curve A, corresponding
to single-valence atoms, is within the Brillouin zone, while curve C, representing
a double-valence material, “overflows” into the second zone. In the first case the
Fermi line is a circle, like for a free particle. In the second case only the arcs that
remained in the first zone represent states there; the ones belonging to states in
the second zone are discussed below. When the neighboring unit cells (in reciprocal
lattice) are drawn as well, the Fermi curve that belongs to the parts of C in the
first zone creates a closed line encircling the corner of the zone, like the thick line
in Fig. 6.25(b). The perimeter of this curve shrinks as the Fermi energy rises, until
it becomes a single point, and the band is completely full.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 480

480 Introduction to Solid-State Physics

The Fermi level of a nearly-free electron. Figure 6.25(c) displays the

equal-energy curves of the first band, for the same four Brillouin zones as above,
when calculated within the nearly-free electron approximation, i.e., in the presence
of a weak periodic potential [see Fig. 6.14(d)]. The periodic potential deforms the
curves: line A, which is far away from the border of the first zone is still roughly
circular, but lines B and C are significantly modified near the boundaries of the
zone. As explained following Fig. 6.14, the periodic potential “smoothens” the
sharp corners of curve C in Fig. 6.25(b), since the equal-energy curves must be
normal to the borderline between the Brillouin zones. A further increase of the
periodic potential, as shown in Fig. 6.25(d), causes curve A to penetrate into the
second zone as well.

(a) (b)

(c) (d)

Fig. 6.25: (a) Three Fermi circles (in the kx − ky plane) of a free particle on a
square lattice. A – each atom donates a single electron, C – each atom donates two
electrons, B – a Fermi circle that touches the boundaries of the first Brillouin-zone
(whose edge length is 2π/a). (b) The four Brillouin zones derived upon displacing
part (a) by reciprocal lattice vectors. (c) Like (b), for energies calculated within
the nearly-free electron model. (d) Like (c), but with a stronger periodic potential.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 481

Electrons in solids 481

The Fermi surface of the second energy band. The circle C in Fig. 6.25(a),
is reconstructed in Fig. 6.26 (right panel). The parts of the circle that penetrate
into the second zone contain states belonging to the second energy band. Indeed, a
displacement of the circle by reciprocal-lattice vectors, as in the mid panel of Fig.
6.26, creates the islands marked by X; the perimeters of these islands are parts
of the Fermi surface. A weak periodic potential displaces these curves, like in the
left panel of Fig. 6.26, and causes the equal-energy curves to cross smoothly the
boundaries of the Brillouin zone, along a direction normal to the faces. As a result,
the Fermi surface consists of the curves marked by C in Fig. 6.25(b) in the first
band, and those appearing in the left panel of Fig. 6.26 in the second band.

Fig. 6.26: Right panel: the dark circle contains the points in lattice-momentum
space that are occupied by free electrons, for double-valence atoms on a square
lattice. The perimeter consists of the arcs C in the first Brillouin zone, pertaining
to the first energy band, and from the arcs that penetrate into the second zone,
and thus refer to the second band. Mid panel: the thick circles represent the circle
of part (a) displaced into the neighboring Brillouin zones. The areas marked by X
contain occupied states in the second energy band; their circumferences are part of
the Fermi surface. Left panel: whereas the right and middle panels describe the
spectrum of free electrons, this panel is drawn for a weak periodic potential. The
equal-energy curves, and in particular the Fermi curve, are slightly shifted. The
potential is mainly effective close to the faces of the Brillouin zone, where these
curves are normal to the faces.

Within the nearly-free electron approximation, the second energy band of the
square lattice penetrates into the first one, and therefore the Fermi surface is con-
tained in the two overlapping bands; as the number of the electrons in the lattice
increases, the electrons occupy both bands, and therefore such a material behaves
as a semimetal. When the periodic potential is strong, the gaps in-between the
bands widen, the lower band is fully occupied for double-valence materials and the
solid is an insulator. Thus, double-valence materials are metals when the electrons
interact weakly with the ions, and are insulators when that interaction is strong.
The equal-energy curves (and in particular the Fermi one), within the tight-binding
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 482

482 Introduction to Solid-State Physics

approximation and for a single band, are illustrated in Fig. 6.20.

Fermi surfaces in three dimensions. The configuration of the band struc-
ture, and consequently of that of the Fermi surface, is quite complex in three
dimensions. A general discussion is beyond the scope of this volume; instead,
several explicit examples based on the nearly-free electron approximation are de-
scribed. The alkali metals in the first column of the periodic table are all of the
BCC type, see Fig. 2.22. There are two atoms in the unit cell of the BCC
and hence the electrons’ density is n = 2/a3 . By Eq. (6.28) n = kF3 /(3π 2 ),
and thus kF = (3π 2 n)1/3 = (6π 2 )1/3 /a ≈ 1.24π/a. The reciprocal lattice of
the BCC is an FCC, with a lattice constant 4π/a, and reciprocal-lattice vec-
tors G = (2π/a)[h(ŷ + ẑ) + k(x̂ + ẑ) + (x̂ + ŷ)], as given in problem 3.8. The
reciprocal-lattice vectors connecting the center of the Brillouin zone with the 12
nearest neighbors in reciprocal lattice are G = (2π/a)(±ŷ±ẑ), G = (2π/a)(±x̂±ẑ),
or G = (2π/a)(±ŷ ± x̂). The first Brillouin zone is the Wigner-Seitz cell of the
FCC, as displayed in Fig. 2.19(a). The boundary planes between the first Brillouin
zone and its neighbors are the normals to those vectors, at their mid points. The
distance from the√origin to each of these planes equals half the length of such a
vector, i.e., k0 = 2π/a (check!). In the present case, kF /k0 ≈ 0.877 and hence the
Fermi surface is entirely contained within the first Brillouin zone, not approaching
its borders. In the nearly-free electron model this surface is roughly spherical (that
of potassium is presented in Fig. 6.27). Indeed, the Fermi surfaces of the alkali met-
als are almost spheres, rendering the Sommerfeld model an adequate description of
their electronic properties. Apparently, this implies that the periodic potential in
these materials is rather weak. On the other hand, the alkaline earths in the
second column of the periodic table have a double valence. For instance, barium
has the BCC structure, with kF = (12π 2 )1/3 /a ≈ 1.56π/a. The Fermi sphere of
the free particles penetrates into the neighboring zones, and the Fermi surfaces for
both bands encircle small islands in lattice-momentum space, as in Figs. 6.25 and
6.26 (see Fig. 6.27). Hence the elements in the second column are semimetals.
Another example is the single-valence transition metals, copper, silver, and
gold. They are all of the FCC type and have similar electronic configurations.
That of copper, for instance, is [Ar]3d10 4s1 (where [Ar]=1s2 2s2 2p6 3s2 3p6 denotes
the electronic configuration of argon); the energy bands are constructed from hy-
bridizations of s-states with d-ones. One of those bands is above the others, and
therefore copper is treated as a single-valence metal, in which this band is half-
filled. In the FCC lattice there are four atoms in the cubic unit cell, and hence
n = 4/a3 . Within the nearly-free electron model the corresponding Fermi momen-
tum is kF = (3π 2 n)1/3 = (12π 2 )1/3 /a ≈ 1.56π/a. The reciprocal lattice is the BCC
one with a lattice constant 4π/a; the Brillouin zone is depicted in Fig. 2.19(b) [and
also in Fig. 5.17(b)]. As seen, the surface of the Brillouin zone contains hexagons
and squares. The distance from the center √ of the zone to a hexagonal surface plane is
a quarter of the cube’s diagonal, k0 = 3π/a, leading to kF /k0 ≈ 1.084; the Fermi
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 483

Electrons in solids 483

sphere cuts these planes and penetrates into the neighboring zones. In contrast,
the distance form the center of the Brillouin zone to a square part of the surface
is k0 = 2π/a, giving kF /k0 ≈ 0.78. All in all, the Fermi surface is close to being
spherical save several narrow “sleeves” that appear along the eight directions from
the center of the zone to the corners of the cubic unit cell, connecting neighboring
zones, see Fig. 6.27.

Fig. 6.27: The Fermi surfaces of copper, barium, and potassium. (Taken from
http://www.phys.ufl.edu/fermisurface.)

Fermi surfaces of other materials may be more complex. Consider for example
the simple-cubic lattice, in which–in the tight-binding approximation–only overlaps
of the wave functions between nearest neighbors are taken into account. Extending
Eqs. (6.115) and (6.119), the energies in the tight-binding approximation are given
by

En (k) ≈ E n + 2Γn (a)[3 − cos(kx a) − cos(ky a) − cos(kz a)] . (6.123)

The Fermi surface of a single-valence metal is at the middle of the band, i.e., at
En = E n + 6Γn , which implies cos(kx a) + cos(ky a) + cos(kz a) = 0. This surface is
shown in Fig. 6.28. In contrast with the nearly-free electron model, for which the
Fermi surface is entirely within the first Brillouin zone [problem 6.27(a)] the tight-
binding approximation produces a surface that penetrates into the second zone,
similar to the surface displayed in Fig. 6.25(d), where also a strong potential opens
“sleeves” in-between the zones.

Problem 6.26.
Double-valence atoms are situated on the sites of a square lattice.
a. Assume a very weak periodic potential. Find the electrons’ densities for which
the electrons occupy states within the first Brillouin zone alone. At which density
they occupy states in the third zone?
b. Exploit the nearly-free electron model, the results of problem 6.19(a) and Fig.
6.18(b), to derive the criterion for the transition from a semimetallic behavior to
an insulating one.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 484

484 Introduction to Solid-State Physics

(a) (b)

Fig. 6.28: The Fermi surfaces of an SC single-atom metal, in the tight-binding

approximation. (a) In the first Brillouin zone. (b) In eight Brillouin zones.

c. Which form do the two lowest bands take in the tight-binding approximation?
(assume overlaps of wave functions only on nearest-neighbors). Is semimetallic
behavior possible here?

Problem 6.27.
a. Consider a single-valence metal on an SC lattice in the presence a very weak
periodic potential. Find the percentage of double-valence atoms that has to be added
to the system in order for the Fermi surface to cross the boundary of the first
Brillouin zone.
b. The band’s energies of an electron on a cubic lattice are given in Eq. (6.123).
What is the upmost energy to be filled such that the Fermi surface is still within the
first Brillouin zone?
c. What is the answer of part (a) for a BCC lattice?

6.9 Density of states, charge-carriers’ density, specific heat

Preface. The occupation of the energy levels is determined by the Fermi-Dirac

distribution, f (E) [Eq. (6.35)], which depends on the temperature. Most electronic
properties are determined by the location of the Fermi energy, i.e., whether it is
within an energy band or in an energy gap. For example, when at zero temperature
the electrons fill all states within a certain band then at a finite temperature they
have to fill states above that band; however, the only available ones are those in the
next band. The energy distance to those can be quite significant, making the cor-
responding Fermi-Dirac distribution exponentially small, f (E) ≈ exp[−β(E − μ)],
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 485

Electrons in solids 485

and consequently decreasing considerably the electrical conductivity, the electronic

specific heat, etc. On the other hand, when at zero temperature the electrons oc-
cupy only part of the band and have easily-available states near the Fermi level, the
finite-temperature behavior becomes similar to that emerging from the Sommerfeld
model.
Density of states. As discussed in Sec. 6.3, the total energy of the electron
gas, from Eq. (6.39), is
 ∞
Etot  = V dEEg(E)f (E) , (6.124)
0

where g(E) is the density of the electrons’ energy levels: V g(E)dE is the number of
eigenenergies of the single-particle Schrödinger equation in the energy range {E, E +
dE}. As opposed to the density of states of a free particle [e.g., Eq. (6.32)], a simple
monotonic function of the energy, the band structure has gaps in which there are
no states at all, implying zero density of states in the gaps. The density of states is
no more a monotonic function of the energy.
The density of states in one dimension. The number of states with wave
numbers in the range {k, k + dk} in one dimension is Lg(k) = 2 × 2 × dk/(2π/L),
where (2π/L) is the length of the segment on the k−axis that contains a single
discrete value of k = 2π /L; consequently, dk/(2π/L) is the number of states in
the segment dk. The density of states as a function of energy, taking into account
only positive energies, is multiplied by 2 since to each state of a wave number k
there corresponds a state with −k. Another factor of 2 accounts for the two spin
states. Save this last factor, the calculation is identical to the one following Eq.
(5.56). Note, though, that here the density of states is defined per unit length
and so the final result is not multiplied by L. Upon transforming a k−integration
into an energy one, one obtains
2 1
g(E) = . (6.125)
π (dE/dk)
As an example, consider the density of states of the bands shown in Fig. 6.5(a).
These bands are derived within the tight-binding approximation, Eq. (6.118). Ex-
ploiting those, one finds that dE/dk = 2aΓn sin(ka) in the range 0 < E −E n < 4Γn .
Using the energy as a variable gives

dE/dk = a (E − E n )(4Γn − E + E n ) .
In-between the bands the density of states vanishes. Inserting these relations into
Eq. (6.125) produces Fig. 6.29. As in the case of phonons in one dimension, there
appear van Hove singularities at the bands’ edges.
The density of states in arbitrary dimensions. The extension of the
calculation of the density of states in one dimension to higher dimensions is similar
to the derivation in Chapter 5, Eqs. (5.61)-(5.66). The density of states is derived
by counting the number of states in the range {E, E + dE} between equal-energy
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 486

486 Introduction to Solid-State Physics

Fig. 6.29: The density of states [in units of 1/(πaΓ1 )] of the energy bands displayed
in Fig. 6.5(a) (energy in units of 2Γ1 ).

surfaces in d − 1 dimensions. The number of states in a volume element dd k in

k−space is 2dd k/[(2π)d /V ], where the factor of 2 counts the spin states. Hence, for
an arbitrary function F [En (k)],
 
V 
2 F [En (k)] = 2 dd kF [En (k)]
(2π)d n
n,k
  
V
= d−1 d dd k dE δ[E − En (k)]F (E) ,
2 π n
and thus 

2 F [En (k)] = V dEg(E)F (E) , (6.126)
n,k

with 
1  1  dS
g(E) = dd k δ[E − En (k)] = . (6.127)
2d−1 π d n
2d−1 π d n
|∇k En (k)|
Here, ∇k En (k) is the gradient of the energy in the n−band with respect to k, such
that |∇k En (k)| is the derivative of the energy with respect to the k−component
perpendicular to the equal-energy surface whose energy is E; the integration is over
this surface (see Fig. 5.16). The final result is obtained upon transforming the
integration to the one over the volume in-between the two equal-energy surfaces,
and using dd k = dSΔk⊥ , where Δk⊥ is the local distance between the two surfaces.
Equation (6.125) is a particular case of Eq. (6.127).
In one dimension, the van Hove singularities near the bands’ edges are caused
by the vanishing of the derivative dE/dk. To explore the analogous behavior at
higher dimensions, consider, e.g., the square lattice. The point k0 = (π/a, δ) is on
the border between the first two Brillouin zones, that is, on the midperpendicular
of the reciprocal-lattice vector G = (2π/a, 0). Since En (k) = En (−kx , ky ) (see
problem 6.12), it follows that ∂kx En (k) = −∂kx En (−kx , ky ). On the other hand,
Eq. (6.64) yields ∇k En (k) = ∇k En (k + G), and hence
∂ ∂ ∂
E (−π/a, δ) = E (π/a, δ) = − E (−π/a, δ) .
∂kx n ∂kx n ∂kx n
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 487

Electrons in solids 487

It follows that
∂ ∂
E (−π/a, δ) = E (k ) = 0 . (6.128)
∂kx n ∂kx n 0
A parabolic behavior of the energies as the surface separating the two Brillouin
zones is crossed normally to it is indeed observed in the solution of problem 6.19.
In such a case ∇k En is parallel to this surface, and the equal-energy surfaces are
normal to it. Similarly, Eq. (6.119) yields that the component of the gradient
parallel to the vector G = (2π/a, 0), i.e., the component normal to the border line
between the two Brillouin zones, is ∂kx En (k) = 2aΓn sin(kx a), that vanishes at all
points on this border line, on which kx = G/2 = π/a.
The density of states of a square lattice. The gradient of the energy, which
vanishes on the zone boundaries, is still in the denominator for dimensions higher
than 1. However, it appears within an integral over the equal-energy surface, see
Eq. (6.127). The integration usually smoothens the divergence, and leaves a weaker
singularity in the density of states (as also happens for the density of states of the
phonon modes). Consider for example the electronic density of states on the square
lattice, derived within the tight-binding approximation. According to Eq. (6.119),
the equal-energy surfaces are given by
En (k) ≈ E n + 4Γn [sin2 (kx a/2) + sin2 (ky a/2)] ,
yielding for the denominator in Eq. (6.127)

|∇k En | = |2Γn a[sin(kx a)x̂ + sin(ky a)ŷ]| = 2Γn a sin2 (kx a) + sin2 (ky a) .

Around the origin En (k) ≈ E n + 2Γn (a)(ak)2 , and therefore the density of states
jumps from zero to the constant g(E) ≈ 1/(2πΓn a2 ) as the energy rises above
Emin = E n . Similarly, near the maximum of the band energy, at the K point (the
corner of the Brillouin zone), the energy is En (k) ≈ E n +4Γn (a)−Γn (a)a2 (k−G11 )2 ,
with G11 = (π/a, π/a). The density of states there jumps from 0 to the same
constant as the energy is reduced through the upper edge of the band, Emax =
E n + 4Γn (a). When the energy is raised from the bottom of the band, or when
it is decreased away from the top of the band, the equal-energy curves deform
gradually, changing from circles to the square obtained at the middle of the band
[see the discussion after Eq. (6.119)]. As a result, the density of states increases
gradually until the middle of the band is reached; there cos(kx a) + cos(ky a) = 0,
i.e., kx a = π − ky a. The equal-energy curve is a straight line, and in such a case
the integral of Eq. (6.127) is
 π/(2a)  π/(2a)
d dkx 8 ln[tan(kx a/2)] kx = 2a
π

8  =8 =  ,
0 2 sin2 (kx a) 0 sin(kx a) a kx =0

where the symmetry of the equal-energy curves under rotations by 45◦ is used
(each integral covers just 1/8 of the Fermi line, i.e., half of the line in a single
quarter); note that d 2 = dkx2 + dky2 for a length element on the equal-energy curve.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 488

488 Introduction to Solid-State Physics

The integral diverges logarithmically near its lower bound, i.e., near the cut of the
middle of the face with the Fermi curve. A somewhat more complicated calculation
yields that the density of states diverges logarithmically near the middle of the
band, as ln[|E − E n − 2Γn (a)|]; this singularity is far weaker than that found in one
dimension (see problem 6.28). Figure 6.30(a) illustrates the density of states of the
square lattice in the tight-binding approximation.

(a) (b)

Fig. 6.30: Schematic illustration of the density of states in the tight-binding

approximation. (a) Square lattice. (b) cubic lattice.

Problem 6.28.
The density of states of energies at the vicinity of the middle of the energy band of
a square lattice is governed by the integral Eq. (6.127) around the point (0, π/a).
Use a Taylor expansion around this point to show that for such energies the density
of states is roughly given by g(E) ∝ ln[|E − E n − 2Γn (a)|].

The density of states in a cubic crystal. The three-dimensional density of

states does not diverge; its derivative, however, does. The parabolic form of the
energy as a function of the crystal momentum
 at the vicinity
 of the band’s edges
yields a density of states proportional to |E − Emin | or to |E − Emax |, with van
Hove singularities that appear at certain planes in the Brillouin zone. For instance,
the density of states of the lower band in a cubic lattice, illustrated in Fig. 6.30(b),
exhibits weak singularities at the reciprocal-lattice vectors (π/a, 0, 0) (the maxima
in the figure) and (π/a, π/a, 0) (the edges of the band).

Problem 6.29.
The energy of the electrons at the bottom of the band may be often (e.g., for semi-
conductors) approximated by
 k2 ky2 kz2 
E(k) = E0 + 2 x
+ + .
2m∗x 2m∗y 2m∗z
This relation between the energy and the crystal momentum is termed “ellipsoid”
dispersion, due to the ellipsoid form of the equal-energy surfaces. The quantities
{m∗ } are called “effective masses”; these are discussed in the next section.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 489

Electrons in solids 489

a. Show that this is the dispersion relation at the bottom of the band of an or-
thorhombic crystal, calculated within the tight-binding approximation, with overlaps
between nearest neighbors.
b. Find the density of states for this case.
c. Find the number of states in the conduction band in the regimes where the
parabolic approximation is valid. Express the density of states in terms of this
number.
d. At the top of the valence band (where the holes in semiconductors are located)
the energy again has an ellipsoid dependence on the crystal momentum,
 k2 ky2 kz2 
E(k) = E0 − 2 x
+ + .
2m∗x 2m∗y 2m∗z
The effective masses that appear here are generally different from those used for the
description of the conduction band. Find the modifications in the answers for parts
(a) and (b) in this region.

The specific heat of metals. The Fermi level of metals lies within the con-
duction band, and hence the density of states at the Fermi level, g(EF ), differs from
zero. The specific heat of a metal can be found by following the arguments leading
to Eq. (6.41): at low temperatures, the energy change relative to that of the ground
state comes from electrons displaced from the energy range EF − kB T < E < EF to
the range EF < E < EF + kB T . Assuming that the density of states is continuous
and is almost independent of the energy over this narrow range, one concludes that
the number of states there is roughly V g(EF )kB T . As each electron carries the
energy kB T , the total change in the energy is proportional to V g(EF )(kB T )2 , and
hence the electronic specific heat is linear in the temperature. A more accurate
calculation, which extends the one leading to Eq. (6.41), (see also problem 6.9)
yields

V g(EF ) ∞ ∂f (E) π2
CV (electrons) ≈ (E − μ)2 − dE = V g(EF )kB2
T ≡ γT .
T 0 ∂E 3
(6.129)
The electronic specific heat of metals is indeed linear in the temperature, but the
coefficient γ can be very different from that derived within the Sommerfeld model.
This coefficient carries the information on the density of states at the Fermi level.
In particular, much like the density of states, the coefficient γ is maximal wherever
the Fermi surface crosses the boundary of the Brillouin zone (see Fig. 6.30). When
electrons reside in several bands, like in Fig. 6.23, the measured coefficient repre-
sents the sum of the densities of states of the bands that contribute at the Fermi
level. Varying the number of electrons, e.g., by replacing single-valence atoms with
double-valence ones (as described in problem 6.27) facilitates a controlled variation
of the Fermi energy, and the ensuing variation of γ. Thus, measuring this coefficient
provides information on the dependence of the density of states upon the energy.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 490

490 Introduction to Solid-State Physics

At low temperatures the electronic specific heat is smaller than that of the phonons,
and the total specific heat is still given by Eq. (6.42).
There is a pronounced disparity between the specific heat of electrons subjected
to a periodic potential and the one computed within the Sommerfeld model, even
when the energy is approximated by a parabolic form, like in problem 6.29. In this
approximation one finds
γ = (3π 2 nm∗x m∗y m∗z )1/3 kB
2
/(32 ) .
This expression resembles the one in Eq. (6.41), with the mass there replaced here
by the geometrical average of the three effective masses, (m∗x m∗y m∗z )1/3 , see problem
6.29. Since these effective masses reflect the dependence of the energy on the lattice-
momentum vector near the bottom of the band, they are mostly not related to the
mass of a free electron.
The concentration of charge carriers in semiconductors. Assuming that
the electrons’ concentration in the conduction band is low, and that they are mostly
located near the band’s bottom, their energy is generally quadratic in the lattice
momentum, and thus their density of states is given by the answer to problem 6.29,
g(E) = Ce (E −Ec )1/2 , where Ec denotes the energy at the bottom of the conduction

band, and Ce = 2m∗x m∗y m∗z /(π 2 3 ). Exploiting the Fermi-Dirac distribution, and
the expression for Ne , Eq. (6.36), yields
 ∞
√
ne = dEg(E)f (E) = Ce π(kB T )3/2 e−β(Ec −μ) /2 ≡ n0e e−β(Ec −μ) , (6.130)
Ec

for the electrons’ concentration in the conduction band. [This derivation makes
∞ √ √
use of the relation 0 dx x exp[−βx] = π/(2β 3/2 ).] An analogous calculation
gives the number of electrons removed from the valence band, leaving empty states
(called “holes”). The probability to find an empty state of energy E in the valence
band is
1 1
1 − f (E) = 1 − β(E−μ)
= −β(E−μ)
≈ eβ(E−μ) , (6.131)
1+e 1+e
where E is below the top of the upmost valence band, Ev , and it is assumed that
μ − E  kB T . Using part (c) of problem 6.29, the density of states is g(E) =
Ch (Ev −E)1/2 , (Ch includes the effective masses of the holes), and the concentration
of holes is
nh = n0h eβ(Ev −μ) . (6.132)
Note that by multiplying the electrons’ concentration with that of the holes, one
finds
π
ne nh = n0e n0h e−βEg = Ce Ch (kB T )3 e−βEg ≡ n2i , (6.133)
4
where Eg = Ec − Ev is the energy gap between the two bands. This result is a
particular case of the law of mass action, which relates the concentrations of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 491

Electrons in solids 491

different particles held at equilibrium. It is independent of the chemical potential

or of the electrons’ source, and thus enables the detection of the energy gap.
In an intrinsic semiconductor the electrons’ concentration coincides with
that of the holes,
√  √
ne = nh = ni = n0e n0h e−βEg /2 = Ce Ch π(kB T )3/2 e−βEg /2 /2 . (6.134)
Using here Eq. (6.130) yields
μ = (Ev + Ec )/2 + kB T ln[Ch /Ce ]/2 = EF + O(kB T ) . (6.135)
At low temperatures the Fermi energy is located at the center of the gap between
the two bands. The ratio Ch /Ce is determined by the ratio of the effective masses
of the holes and of the electrons.
The electrons’ energy levels in donors and acceptors. Quite generally,
the electron “contributed” by a donor atom is in the outer shell of the latter. This
electron is bound to the nucleus and to the other electrons in the donor. As that
atom is within the lattice, one may assume that the Coulomb energy of the electron
is roughly given by UD ≈ −ze2 /(r), where ze is the charge of the nucleus and
all other nearby electrons, and  is the dielectric constant of the material (this
approximation is valid as long as the electron is relatively far from the nucleus,
such that it “samples” the average dielectric effect of many atoms; it is justified
below). The Schrödinger equation of this electron is
[Ec − 2 ∇2 /(2m∗ ) + UD ]ψ = Eψ
(assuming that the effective mass is isotropic); extending the solution of the hydro-
gen atom yields the energy levels,
E(n) − Ec = −z 2 Ry m∗ /(m2 n2 ) .
Here, Ry = 13.6 eV is the Rydberg constant. The energy is measured relative to
the bottom of the conduction band. The consideration below refers to the ground
state of the donated electron, with the binding energy ED = Ec − E(1). In the
solution of the hydrogen atom, where z = 1, the probability to find the electron at
a distance r from the donor is proportional to exp[−r/r0 ], where r0 = aB m/m∗ .
The dielectric constant of semiconductors is usually quite high, making this distance
much longer than the Bohr radius aB = 0.53Å (and also longer than the lattice
constant, validating the use in the average dielectric constant). The binding energy
ED is usually quite small (of the order of 0.001-0.005 meV); it is thus relatively
easy to ionize the atom and release the electron. Similar conclusions can be drawn
regarding the energies of the electrons in acceptors, measured relative to Ev .
The density of charge carriers in extrinsic semiconductors. Equation
(6.135) implies that the Fermi energy of an intrinsic semiconductor at zero temper-
ature is at the middle of the gap between the highest occupied level and the lowest
empty one [Fig. 6.31, right panel]. In the more general case, the chemical potential
is determined by requiring charge neutrality: the number of free negative charges
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 492

492 Introduction to Solid-State Physics

should compensate the number of positive charges on the donors that contribute
them. As seen, the “outer” electron of the donor has energy within the gap between
the two bands, at a distance ED below the bottom of the conduction band (so that
these electrons are easily excited into the conduction band). The probability to find
an electron on the donor is f (Ec − ED ). Denoting the concentration of donors by
xD and assuming that each of them donates a single electron, then the number of
electrons added to the conduction band is
1 xD
xD [1 − f (Ec − ED )] = xD 1 − = .
1 + eβ(Ec −ED −μ) 1 + e−β(Ec −ED −μ)
Very similar considerations yield that the energy of the “outer” electron in an
acceptor is Ev + EA , and that it is located slightly above the top of the valence
band, such that electrons from that band are easily excited into the levels on the
acceptors. For a concentration xA of acceptors, each accommodating an additional
single electron, the concentration of electrons located on the acceptors is
xA f (Ev + EA ) = xA /(1 + exp[β(Ev + EA − μ)]) .
The total number of electrons obeys
xA xD
ne + = nh + , (6.136)
1 + exp[β(Ev + EA − μ)] 1 + exp[−β(Ec − ED − μ)]
where ne and nh are given in Eqs. (6.130) and (6.132).
The chemical potential μ of an extrinsic semiconductor is determined by Eq.
(6.136). Consider for instance an n-type semiconductor. At low temperatures, such
that all energy differences are larger than kB T , the second term on the right-hand
side may be expanded to obtain
n0e exp[−β(Ec − μ)] = ne ≈ n0h exp[β(Ev − μ)] + xD exp[β(Ec − ED − μ)] .
Ignoring completely the holes’ concentration, then
μ ≈ Ec − ED /2 + kB T ln[xD /n0e ]/2 . (6.137)
Inserting this result into Eq. (6.132) yields nh ≈ n0h exp[β(Eg − ED /2)], which
justifies its aforementioned neglect (the gap energy is larger than kB T ). The Fermi
energy, EF = Ec − ED /2 is at the mid point between the filled level of the donors
and the empty level at the bottom of the conduction band, see the mid panel in Fig.
6.31. As the energy ED (measured relative to the bottom of the conduction band)
is generally smaller than the energy gap, the Fermi level is close to the threshold
of the conduction band, and then the approximations used above are valid as long
as ED  kB T . Otherwise, one has to solve a quadratic equation in the unknown
variable exp[β(μ − Ec )] see problem 6.30. Inserting the result (6.137) into Eq.
(6.130) yields
√
ne  n0e xD e−βED /2 . (6.138)
Analogous results are obtained for p-type semiconductors, left panel in Fig. 6.31.
Here again, a number of electrons reside on the acceptors, leaving holes in the
valence band that contribute to the conductance (check!).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 493

Electrons in solids 493

Fig. 6.31: Right panel: the Fermi energy at the middle of the gap between the full
valence band and the empty conduction one. Mid panel: the Fermi energy at the
mid energy between the level of the electrons in the donors and the bottom of the
conduction band. Left panel: the Fermi energy at the mid energy between the level
of the electrons in the acceptors and the top of the valence band.

Problem 6.30.
a. Find the chemical potential which results from Eq. (6.136), when there are only
donors, without applying any approximations except that kB T  Eg . Discuss the
temperature dependence of the solution.
b. What is the answer when there are only acceptors?

Problem 6.31.
The following data have been obtained for silicone at T = 300K: n0e = 2.8 × 1019
cm−3 , n0h = 1.04 × 1019 cm−3 . The energy gap is 1.14 eV.
a. What is the chemical potential when the concentration of electrons in the con-
duction band is ne = 1017 cm−3 ? What is the concentration of holes in the valence
band in this case?
b. What is the chemical potential when the concentration of holes in the valence
band is nh = 1014 cm−3 ? What is the concentration of electrons in the conduction
band?

The specific heat of semiconductors. In intrinsic semiconductors, the

number of electrons that cross over from the valence to the conduction band is
proportional to exp[−βEg /2] at low temperatures, see Eq. (6.134). Each of these
carries an energy of the order of the gap, Eg , between the bands. It follows that
the electronic specific heat of an intrinsic semiconductor at low temperatures is also
proportional to exp[−βEg /2], and hence is much smaller than the phonons’ specific
heat. A more accurate computation yields that the average energy of the electrons
in the conduction band of an intrinsic semiconductor is
 ∞  ∞
E
= dEg(E)f (E)E = Ec ne + Ce e−β(Ec −μ) dxx3/2 e−βx
V Ec 0

= (Ec + 3kB T /2)ne , (6.139)

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 494

494 Introduction to Solid-State Physics

using the relation


√
dxx3/2 exp[−βx] = 3 π/(4β 5/2 ) , (6.140)

and Eq. (6.130). At low temperatures, the leading-order term in the electronic
specific heat is
CV ≈ Ec (∂ne /∂T ) ∝ β 1/2 exp[−βEg /2] .
A similar term is contributed by the holes in the valence band, but both contri-
butions are much smaller than the phonons’ specific heat. They can though be
measured when the energy gap between the valence and the conduction bands is
not too wide.
The electronic specific heat of an extrinsic semiconductor is also not very
significant. In an n-type semiconductor, the density of states is
xD  g δ[E − E( )] , E < Ec ,
g(E) =  (6.141)
Ce E − Ec , E > Ec ,
where g is the degeneracy of the energy level E , located within the gap. The
average energy is hence
   ∞
E
= dEg(E)f (E)E = g f [E( )]E( ) + Ec nc + Ce e−β(Ec −μ) dxx3/2 e−βx
V 0


= g f [E( )]E( ) + (Ec + 3kB T /2)ne , (6.142)


where the second step exploits Eq. (6.130) and the integral Eq. (6.140). This result
is identical to the one of an intrinsic semiconductor, augmented by contributions
from the electrons on the donors.

6.10 The semiclassical equations of motion for electrons in a peri-

odic potential

Lattice momentum, electrons’ velocity. The band structure induces modifi-

cations in the description of the electrons’ motion as compared to the simplified
calculations within the Sommerfeld model, Sec. 6.3. A conspicuous example is the
average velocity of the electrons. In quantum mechanics, the operator that repre-
sents the momentum is p̂ = −i∇. Its expectation value, in a quantum state that
obeys the Bloch theorem, ψnk = exp[ik · r]unk (r) [Eq. (6.52)], is

pnk = ψnk | − i∇|ψnk  = d3 r[eik·r unk (r)]∗ (−i∇)[eik·r unk (r)] . (6.143)

As
(−i∇){exp[ik · r]unk (r)} = exp[ik · r][k − i∇]unk (r) ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 495

Electrons in solids 495

one finds

pnk = k + d3 ru∗nk (r)(−i∇)unk (r) , (6.144)

using the normalization of the wave functions. Thus, k is not the momentum
of the particle (except for a free particle, for which the second term vanishes).
Nonetheless, in many aspects the quantity k does play the role of a momentum,
and therefore it is called the “lattice momentum”.
To derive an expression for the average momentum of the electron in a Bloch
state, one begins from Eq. (6.62) and considers the variation of its solution as the
vector k is slightly changed,
En (k + δ) = En (k) + δ · ∇k En (k) + O(δ 2 ) . (6.145)
A similar variation of the Hamiltonian operator in the left hand-side of Eq. (6.62))
yields
2 2 δ 2
Ĥ(k + δ) = Ĥ(k) + δ · (−i∇ + k) + . (6.146)
m 2m
Treating the second term as a perturbation, Eq. (C.2) for the eigenenergies of
Ĥ(k + δ) shows that the first-order correction of the energy is the expectation value
of the perturbation,
2
En (k + δ) = En (k) + δ · u | − i∇ + k|unk  + O(δ 2 ) . (6.147)
m nk
A comparison of the two expressions, Eqs. (6.146) and (6.147), reveals that
2
u | − i∇ + k|unk  = ∇k En (k) ,
m nk
and upon inserting into Eq. (6.144), one obtains the average velocity of the electron,
1 1
vnk = pnk = ∇k En (k) . (6.148)
m 
This expression is identical to that for the group velocity of a wave packet in
quantum mechanics [see also Eq. (5.10) and the comments preceding Eq. (6.43)].
Equation (6.148) reproduces the result p = k for a free particle [see also Eq.
(6.144)]. In this case the momentum does not vary in time and is conserved. When
the state of the electron belongs to a band structure, Eq. (6.148) can produce very
different results. In particular, the momentum is not a “good quantum number”,
and therefore is not conserved. As seen in Eq. (6.128), the component of ∇k En (k)
normal to the faces of the Brillouin zones vanishes at the zones’ boundaries, and
therefore the velocity of the electron is zero there as well, though the lattice momen-
tum is not. This conclusion stems from the fact that on the faces of the Brillouin
zones the electron’s wave functions correspond to standing waves. This, in turn,
originates from the Bragg reflection off the periodic potential created by the ions
(Sec. 6.6). The second term in Eq. (6.144) represents the momentum exchange be-
tween the electron and the lattice in this scattering event. [Indeed, substituting unk
from Eq. (6.85) into Eq. (6.144) gives a weighted sum of reciprocal-lattice vectors.]
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 496

496 Introduction to Solid-State Physics

Equations of motion. In quantum mechanics the electron is described by

a wave packet, as in Eq. (5.9). Roughly speaking, this wave packet is concen-
trated around the average momentum in momentum space, and around the average
location in real space. The Ehrenfest theorem relates the time derivative of the ex-
pectation values of the position and momentum operators to the classical equations
of motion (this is the correspondence principle). The wave packets of an electron in
a periodic potential are composed of linear combinations of Bloch functions (that
are the eigenfunctions of the Hamiltonian with a periodic potential). In the pres-
ence of solely the periodic potential, the energy of the electron is En (k) of the
conduction band; the lattice momentum k is a good quantum number, and hence
it is conserved. The average velocity of the electron, Eq. (6.148), is conserved as
well, and does not vary with time. The only way to change this situation is to
apply additional forces on the electron, e.g., an external electric field. When this
force is weak enough so that the width δk of the wave packet that describes the
electron is narrower than the width of the Brillouin zone, the wave packet remains
within a single energy band, and does not include contributions of other bands. The
index n is then a good quantum number, and is not changed, e.g., by collisions. In
real space the wave packet extends over distances of the order of 1/δk, comprising
many unit cells, and the location of the electron, denoted r, is within this region in
space. Under these circumstances, the time dependence of k and r is derived from
the classical equations of motion, with the kinetic energy replaced by En (k). This
approximation is termed semiclassical, since the effect of the periodic potential
due to the ions, which essentially extends over short ranges, is found by a fully-
quantum calculation, while that of the weaker forces, effective over larger distances,
is accounted for classically. The success of this approximation is only partial: For
example, the explanation of the quantum Hall effect (in the next section) requires
quantum corrections.
In the presence of a static electric potential φ, the energy of the electron in the
semiclassical approximation reads
E = En (k) − eφ , (6.149)
where the electric field is E = −∇r φ. Assuming that the wave packet is narrow
enough, the weak electric potential does not modify significantly the energy dis-
tribution among the constituents of the wave packet. Then, differentiating with
respect to the time, and using the (approximate) relation v = ṙ and Eq. (6.148),
one finds
Ė = Ėn (k) − eφ̇ = ∇k En (k) · k̇ − e∇r φ · ṙ = vnk · (k̇ − e∇r φ) . (6.150)
As the energy does not vary in time, the left hand-side vanishes, and hence
k̇ = e∇r φ = −eE . (6.151)
This equation of motion resembles the classical one, ṗ = −eE, which is the ba-
sis of Eq. (6.3) in the Drude theory, or Eq. (6.44) of the Sommerfeld model.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 497

Electrons in solids 497

The only difference is that in Eq. (6.151) the momentum is replaced by the
lattice momentum, k. Though the latter is not the true momentum of the
electron, it plays the role of momentum in the semiclassical equation of motion,
and consequently in all equations derived from momentum conservation, e.g., those
that describe collisions of the electrons with impurities. However, as opposed to
the classical conservation law, the lattice momentum is defined up to a reciprocal-
lattice vector, and hence it may “jump” by such a vector after a collision. This
process is called Umklapp [see also the discussion following Eq. (5.99)]. The excess
momentum is absorbed by the entire lattice.
Ignoring collisions, then in the presence of a constant electric field, the lattice
momentum increases linearly with time, k(t) = k(0) − eEt, precisely as in Eq.
(6.44). Indeed, substituting this time-dependent lattice-momentum in the right-
hand side of Eq. (6.148) yields
1
v(t) = ∇ E [k(0) − eEt/] . (6.152)
 k n
Since the band energy oscillates periodically as a function of k, the electron velocity
oscillates as a function of time. Thus, a constant electric field causes the electrons to
move periodically in time. This periodic motion is named after Bloch, and is called
Bloch oscillations. In particular, when k(t) approaches the boundary between two
Brillouin zones, the velocity in Eq. (6.152) tends to zero [Eq. (6.128)] and changes
its sign as a function of time. This surprising behavior, very different from the linear
increase with time of a free particle, stems directly from the band structure and the
interference of the waves describing the electrons upon their scattering off lattice
points. Combining this result with the one derived from the Drude theory, where
the electron moves freely only during a relaxation time τ , gives rise at equilibrium
to an average velocity,
v = ∇k En [k(0) − eEτ /]/ .
This velocity may have a sign opposite to that of the electric field when the relax-
ation time is long enough. The relaxation times, however, are usually considerably
shorter than the period of the Bloch oscillations (see problem 6.32), and hence this
behavior is not realized.
Equation (6.152) implies that a constant electric field induces a periodic-in-time
motion of the electrons. As the positive charges on the ions (almost) do not move,
it follows that a constant electric field creates an oscillating dipole moment (for
sufficiently long relaxation times). In reality the dipole moment is a sum over
all electrons’ dipole moments in the conduction band. As electrons move in all
directions, the total moment vanishes for a completely filled band. A detailed
calculation is given in problem 6.32.

Problem 6.32.
a. Show that when the electric field E is oriented along a basis vector b in reciprocal
lattice, the electron moves periodically in space, with the period T = b/(e|E|).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 498

498 Introduction to Solid-State Physics

b. Estimate the period in a SC lattice of lattice constant a = 1Å, in the presence

of an electric field |E| = 104 eV/cm. Compare the result with the typical value for
the relaxation time, as estimated in Secs. 6.2 and 6.3. Are there conditions under
which the Bloch oscillations can be detected experimentally?
c. In a one-dimensional semiconductor of length L and lattice constant a, the energy
of the conduction band is given by E(k) = E0 − Γ cos(ka). At zero temperature the
band is filled up to the Fermi level. Find the dipole moment induced by a constant
electric field along the lattice direction.

Acceleration and effective mass. The average acceleration of the electron,

Eq. (6.148), is
d 1 d
ank = vnk = [∇ E (k)] . (6.153)
dt  dt k n
Using Eq. (6.151), the α−component of the acceleration is
1 ∂ ∂ d  1 
aα nk = En (k) kβ = − eE , (6.154)
 ∂kβ ∂kα dt m∗ αβ β
β β

where the tensor of coefficients,

 1  1 ∂ 2 En (k)
= , (6.155)
m∗ αβ 2 ∂kα ∂kβ
replaces the mass in Newton’s second law, a = −(1/m)eE. Admittedly, this nota-
tion is confusing: the tensor m∗ is the inverse of the matrix in the right hand-side
of Eq. (6.155). This is the tensor of the effective mass of the electron. In the
nearly-free electron model, and at the bottom of the conduction band, the energy
is roughly quadratic in k, E(k) ≈ Ec + 2 k 2 /(2m); the effective mass tensor is then
(m∗ )αβ ≈ mδαβ , as in classical mechanics. In other parts of the band structure the
effective mass is not necessarily related to the mass of a free electron. Equation
(6.154) is counterintuitive to classical mechanics: when the tensor is not diagonal,
an electric field may accelerate an electron along a direction normal to the field.
Even when the tensor is diagonal, it may happen that different components of the
electric field induce different accelerations. Note though, that in many cases the
tensor is diagonal, leading in turn to the dispersion relation discussed in problem
6.29.

Problem 6.33.
Exploit second-order perturbation theory [Eq. (C.3)] to express the inverse tensor
of the effective mass in terms of integrals containing the periodic Bloch functions
unk .

Problem 6.34.
Use the results of the nearly-free electron model for the energies of the electron close
to the energy gap [e.g., Eq. (6.80)] to obtain the tensor of the effective mass.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 499

Electrons in solids 499

The conductance of electrons and holes. Within the Drude model, the
electrons’ velocity at constant electric fields is [Eq. (6.154)]
 1 
vα nk = −e E τ . (6.156)
m∗nk αβ β nk
β

This relation assumes that the relaxation time of the electron depends on the band
it occupies, and on its lattice momentum. The total electric current density is the
sum over all occupied states, in all bands,
e 
j=− vnk , (6.157)
V
n,k

where V is the volume of the sample. [The sum over the states divided by the
volume replaces the electronic concentration n in Eq. (6.4); note the different
meaning of the notation n in Eq. (6.4) and in Eq. (6.157).] The valence bands are
fully occupied by electrons, but then they do not contribute to the current, because
for each electron with a certain velocity there is in the band another electron with
the opposite velocity (due to the k ⇔ −k symmetry), and therefore the sum over
such a band vanishes. It follows that the current is carried only by electrons in
partially-filled bands, i.e., by electrons from the conduction bands. One thus finds
j=σ·E , (6.158)
where the conductivity tensor is
e2   1 
σαβ = τ , (6.159)
V m∗nk αβ nk
n,k

and the sum runs over the partially-filled bands. This equation extends Eqs. (6.5)
and (6.6) of the Drude model for an electron in a periodic potential.
At the bottom of the conduction band the energy is minimal as a function of the
lattice momentum; denoting this point by kc and the corresponding energy by Ec ,
then for an isotropic diagonal effective-mass tensor, the energy around this point is
E(k) ≈ Ec + (2 /[2m∗c ])(k − kc )2 ,
from which one finds vc = −eEτc /m∗c . The electrons thus move in the opposite
direction to the field, as expected for negative charges. On the other hand, at
the upper part of the valence band the energy has a maximum, Ev , at the lattice
momentum kv . Assuming that the maximal energy is isotropic, then
E(k) ≈ Ev − (2 /[2m∗v ])(k − kv )2 ,
where m∗v > 0. The effective mass, as defined in Eq. (6.155) is negative, and equals
to −m∗v . The equation of motion gives
vv = −(/m∗v )[k(τv ) − kv ] = eEτv /m∗v .
The electrons thus move in the opposite direction to that of the lattice momentum,
and along the same direction as that of the electric field. This means that they
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 500

500 Introduction to Solid-State Physics

behave like positively-charged particles. For this reason the charge carriers in the
valence band are treated as holes with a positive charge. This may be understood
also intuitively: an electron moving due to an applied electric field occupies an
empty place in the valence band (i.e., it fills a hole). The place left behind is now
occupied by a hole. Thus, the electron and the hole move at opposite directions.
The current in intrinsic semiconductors is carried by the electrons in the conduc-
tion band and the holes in the upper part of the upmost valence band. The effective
mass of the electrons at the bottom of the conduction band is positive; their con-
tribution to the conductivity tensor can be roughly written as ne τe e2 (1/m∗c )αβ . In
contrast, the effective mass of the electrons in the upper part of of the valence band
is negative; they appear as positively-charged particles. This also results from the
following argument. Were the band completely full, then the current carried by
the electrons in it would have vanished. One may therefore express the electrons’
current in the partially-filled valence band as
j = jfull − jedge = −jedge = nh ev , (6.160)
where j = (−e/V ) electrons v is the current of the electrons in the band, jfull = 0
is the current of the fully-occupied band, and jedge = −nh ev is the current of the
missing electrons in the upper edge of the band. The number of these electrons is
the number of states out of which electrons are removed into the conduction band;
each of them carries the current −ev. The right hand-side of Eq. (6.160) appears
as a current of particles that carry the charge +e and move with a velocity v.
Altogether,
e2   1  2
 1   1 
σαβ = τ = e n τ + n τ , (6.161)
V m∗nk αβ nk e e
m∗c αβ h h
m∗v αβ
nk
is the conductivity tensor; the sign of the effective masses of the holes is reversed,
so that they are positive.
Recall that the numbers of the electrons and holes, ne and nh , are exponential
functions of the temperature, see e.g., the discussion in Sec. 6.8, or Eqs. (6.134)
and (6.138). Ignoring the temperature dependence of the relaxation times, the
conductivity of intrinsic semiconductors is such that σ ∝ T 3/2 exp[−βEg /2]. It
thus increases with the temperature, as opposed to that of metals, that decreases
as the temperature is raised, due to the decrease of the relaxation time [Matthiessen
rule, Eq. (6.49)]. In extrinsic semiconductors, the number of charge carriers reaches
the number of donors (or acceptors) at relatively low temperatures and then it is
almost independent of the temperature; the corresponding conductivity decreases
as the temperature is raised due to the decrease of the relaxation times.

6.11 Electrons in a magnetic field

Semiclassical equations of motion. The discussion following Eq. (6.151) in-

dicates that k plays the role of the momentum in the semiclassical equations of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 501

Electrons in solids 501

motion. Indeed, in the presence of a uniform constant magnetic field B, and when
collisions of the electron with various defects are ignored, the semiclassical equation
of motion (6.8) is
d e e
k(t) = − [v × B] = − [∇k E(k) × B] , (6.162)
dt c c
[making use of Eq. (6.148)]. The increment in time of the vector k lies in a plane
perpendicular to the magnetic field. In the same plane, that increment is also
normal to ∇k E. As the gradient of the energy with respect to k is perpendicular
to an equal-energy surface (which is a surface in the three-dimensional momentum
space), the vector k moves on a curve created by the cut of the equal-energy surface
and the plane normal to the magnetic field. For a magnetic field along ẑ, the vector
k moves in the XY plane. Since at low temperatures the electrons are confined
to move on the Fermi surface, it follows that the electron moves on cuts of that
surface with planes perpendicular to the magnetic field.
The periodic motion on the Fermi surface. As seen in Sec. 6.8, the Fermi
surface is roughly a sphere at the vicinity of the extremal points of the bands;
the cuts with a plane are thus approximately circles. Electrons are moving at the
bottom of the band, and holes are moving near the top; in any event, when the cut
is a closed orbit, the motion of a particle around it is periodic. According to Eq.
(6.162), the time dependence of the component of the lattice momentum which is
tangential to the orbit is dk /dt = (eB/[2 c])|∇k E|, and thus the period is
 T  dk
2 c  2 c dA(E, k⊥ )
T= dt = = , (6.163)
0 eB |∇k E| eB dE
&
where the integration marked by is over a scalar function, along the closed orbit
created by the cut of the Fermi surface with a plane perpendicular to the magnetic
field. This integration gives the difference between the two areas encircled by the
equal-energy curves at E and at E + dE, denoted dA/dE. The component of the
lattice momentum parallel to the magnetic field (and thus perpendicular to the
plane of motion) remains constant.

Problem 6.35.
Derive the result (6.163) directly from the equations of motion
eB ∂E eB ∂E
k̇x = , k̇y = − ,
2 c ∂ky 2 c ∂kx
(the magnetic field is along ẑ).

In a two-dimensional lattice the electrons move on the Fermi curve. When the
energy is a paraboloid, E = 2 k2 /(2m∗ ), this curve (close to the extremal points)
is a circle and the velocity is a constant, |∇k E| = vF = 2 kF /m∗ ; the perimeter
of the circle is 2πkF and hence the period is T = 2π/ωc , where ωc = eB/(m∗ c) is
the cyclotron frequency, known from the motion of a free particle in a constant
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 502

502 Introduction to Solid-State Physics

magnetic field [Eq. (6.9) and problem 6.3], with the mass replaced by the effective
mass. As the effective mass of the holes has the opposite sign, it follows that they
move around the same perimeter in the opposite direction. Problem 6.36 shows that
for the ellipsoidal energy introduced in problem 6.29 and a magnetic field along an
arbitrary direction, B = n̂B, the cyclotron frequency is

∗ 2 ∗ 2 ∗ 2
eB mx n̂x + my n̂y + mz n̂z
ωc =  ∗ ∗ ∗ . (6.164)
c mx my mz
Thus, measuring the cyclotron frequency of the electrons and the holes on a periodic
lattice (for instance, by monitoring the resonant absorption of radiation at that
frequency) at various directions of the magnetic field, provides valuable information
on the Fermi surface and on the effective masses.
In three dimensions, the cuts of planes with the Fermi surface may attain diverse
shapes. When the band is almost empty (or almost full) the Fermi surfaces are
approximately spheres, and the cuts are closed orbits (circles or ellipses). In such
cases the motion is periodic; however, as the band is filled up and the Fermi surfaces
reach the boundaries of Brillouin zones this simple picture breaks down. Consider
for instance the Fermi surface depicted in Fig. 6.28, pertaining to a single-valence
metal of a SC structure. The upper row in Fig. 6.32 displays cuts of this surface
with planes normal to a magnetic field oriented along ẑ. As seen, all cuts are
closed orbits, on which the motion is periodic. On the other hand, the lower row
in the figure illustrates the situation where the magnetic field is rotated around the
x̂−axis, in the YZ plane. At small inclinations the orbits are still closed; as the
angle increases there appear open orbits, on which the motion does not repeat itself.

Problem 6.36.
Prove Eq. (6.164).

Problem 6.37.

Equation (6.162) yields the orbit of an electron in lattice-momentum space. Exploit

it to show that the orbit in real space is
r(t) = r(0) − 2
B B̂ × [k(t) − k(0)] + vB tB̂ ,

where B̂ = B/|B|, 2B = c/(eB) is the square of the“magnetic length”, and vB =

ṙ · B̂ is the component of the velocity parallel to the magnetic field (this component
does not vary in time). Show that the projection of this orbit on a plane normal to
the magnetic field is produced by rotating the orbit in lattice-momentum space, and
scaling the coordinates by the multiplicative factor 2B .

Problem 6.38.
a. Find the Hall coefficient when the charge is carried by electrons as well as by
holes. Keep only terms linear in the magnetic field.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 503

Electrons in solids 503

Fig. 6.32: Upper panels: the boundaries between the dark and light areas represent
cuts of the Fermi surface from Fig. 6.28 with planes normal to the ẑ−axis, for
several values of kz a/π. Lower panels: cuts of the same surface with planes passing
through the x̂−axis, making an angle α with the ŷ−axis.

b. Silicon is an intrinsic semiconductor, with ni = 1.5×1010 cm−3 and μe = 3μh =

1450 cm2 V−1 sec−1 . Plot the dependence of the Hall coefficient on the electrons’
concentration in the conduction band. For what value of this concentration does the
Hall coefficient vanish? At what concentration is it extremal?
c. Incorporate terms quadratic in the magnetic field and show that – in contradiction
with Eq. (6.12) – the specific resistivity ρxx does depend on the magnetic field.
This dependence is termed magneto-resistance. Discuss in particular the cases
ne = nh and ne = nh .
Landau levels. Without loss of generality, the magnetic field is chosen to
point along ẑ. The semiclassical picture gives that a free particle of mass m∗ moves
on a circle in the plane normal to the field, with a period given by the cyclotron
frequency, ωc = eB/(m∗ c). As in other cases, quantum mechanics “translates” a
periodic motion into quantized energy levels corresponding to this motion, such that
the level spacing between consecutive levels is E+1 − E = ωc . The quantum-
mechanical derivation (see Appendix E) results in a motion similar to that of a
harmonic oscillator, with the discrete energy levels
1
E = ωc ( + ) . (6.165)
2
These are the Landau levels. In three dimensions, the total energy of the electron
is thus
E( , kz ) = E + 2 kz2 /(2m∗ ) ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 504

504 Introduction to Solid-State Physics

where the second term is the energy of the motion parallel to the magnetic field
(which is unaffected by it; kz is the component of the momentum along the direction
of the field).
Let us denote the area in momentum space, enclosed by the orbit corresponding
to the th energy level, by A . When is large, one may use Eq. (6.163), to find
A+1 − A dA TeB
= = 2 ,
E+1 − E dE  c
and then E+1 − E = ωc gives A+1 − A = 2πeB/(c); consequently, up to an
additive constant, A = 2πeB/(c) + const. The solution to problem 6.39 yields
A = ( + 1/2)2πeB/(c) . (6.166)
The allowed energies of a free particle (whose energy is parabolic in the wave num-
bers) for a certain are located on a cylinder in momentum space, whose base’s
area is A . These levels are filled along the normal direction until the Fermi energy
is reached; there the cylinder cuts the Fermi sphere. Its height there is kzF , as
determined by E( , kzF ) = EF . Figure 6.33(a) illustrates those cylinders within
the Fermi sphere (that in the absence of the magnetic field corresponds to a free
electron). As the Landau level is increased, the radius of the cylinder increases as
well but its height diminishes. The area encompassed by the periodic orbit of the
electron on the Fermi surface is
AF = ( + 1/2)2πeB/(c) = EF 2πm∗ /2
(check!).

(a) (b)

Fig. 6.33: (a) The energy levels of a free electron in a magnetic field along the normal
direction. Each Landau level encompasses the states on one of the cylinders; the
circular cut of the cylinder with the original Fermi sphere gives the corresponding
Fermi energy. (b) The density of the states of (a), Eq. (6.167). The dashed curve
represents the density
√ of states of a free particle in three dimensions, in the absence
of the field, g ∝ E.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 505

Electrons in solids 505

Problem 6.39.
Consider a free electron in a magnetic field. The radius of the th cylinder in Fig.
6.33, K , is determined by E = 2 K2 /(2m∗ ). Find the area A in terms of K and
prove Eq. (6.166).

The density of states in three dimensions. Since the system is finite along
the direction parallel to the field, kz attains discrete values, kz = 2πnz /Lz (where nz
is an integer, and Lz is the length of the sample along the field). These values form
a discrete continuum for a long sample. As the motion along this direction is a free
one in one dimension, the corresponding density of states is C[E − ( + 1/2)ωc ]−1/2
(problem 6.10). Whenever is augmented by 1, another Landau level is “open” for
electrons. It therefore follows that the total density of states is
∞

g(E) = C[E − ( + 1/2)ωc ]−1/2 Θ[E − ( + 1/2)ωc ] , (6.167)
=0

where Θ(x) is the step function, Θ(x) = 0 for x < 0 and Θ(x) = 1 for x > 0. This
density of states is depicted in Fig. 6.33(b). All occupied Landau levels are such
that ( +1/2)ωc ≤ EF (EF is the Fermi energy). As the magnetic field is increased,
the spacing between the Landau levels broadens, and the number of filled Landau
levels diminishes (the cylinder whose radius is the largest moves out of the Fermi
sphere). Whenever the value of

F + 1/2 = EF /(ωc ) = EF m∗ c/(eB)

is reduced by 1, the upmost Landau level, previously occupied, is depopulated, and
the density of states “jumps” from its upmost branch to the branch below [on one
of the vertical lines in Fig. 6.33(b)].
de Haas and van Alphen effect. The magnetic susceptibility of metallic
bismuth was measured by de Haas and van Alphen. It was found that at strong
magnetic fields and low temperatures, the susceptibility is a periodic function in
1/B, where B is the field. This result can be explained from an examination of the
density of states. As seen, the density of states is determined by the geometry of the
cylinders in Fig. 6.33(a): equal areas encompassed by the electron’s orbits on the
Fermi surface in a plane normal to the field give rise to identical densities of states,
and hence to identical results for the physical properties that are dictated by the
density of states. Denoting by B the field that obeys AF = ( + 1/2)2πeB /(c),
and by B+1 the one that is given by AF = ( + 3/2)2πeB+1 /(c) (with the same
area) then the properties measured at these two fields should be the same, and
therefore they are periodic functions of 1/B, with the period
Δ(1/B) = 1/B+1 − 1/B = 2πe/(cAF ) = e/(m∗ cEF ) .
A similar periodicity was found in measurements of the electrical resistance in the
presence of a magnetic field, a phenomenon termed Shubnikov-de Haas effect.
This periodicity is obtained also when the Fermi surfaces are not spherical, e.g.,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 506

506 Introduction to Solid-State Physics

those in Fig. 6.27. As the direction of the magnetic field is changed, the area AF
is changed as well, according to the geometry of the Fermi surface. In this way the
entire surface can be mapped.
Modifications due to the Zeeman interaction. The description given above
is modified when the spin degree of freedom is taken into account. The energy of
an electron in a magnetic field is affected by the Zeeman interaction: each Landau
level splits into two,

E − B · μ = E ± g ∗ μ B B ,

where μ = ±g ∗ μB ẑ are the two possible values of the magnetic moment of the elec-
tron along the magnetic field, μB is the Bohr magneton, and g ∗ is the gyromagnetic
ratio (see Secs. 2.10 and 4.7). The corresponding density of states, that extends
the one in Fig. 6.33(b), is
∞

g(E) = C[(E − E + g ∗ μB B)−1/2 + (E − E − g ∗ μB B)−1/2 ] . (6.168)
=0

Each of the vertical lines in Fig. 6.33(b) splits into two, whose distance from one
another increases as the magnetic field becomes stronger. In a periodic solid, the
mass of the electron is replaced by m∗ , and consequently ωc = eB/(m∗ c). As in
semiconductors m∗ is usually much smaller than m, it turns out that ωc  μB B.
The Zeeman splitting can be detected only at strong magnetic fields.
The density of states in two dimensions. As opposed to the spectrum
in three dimensions, that of a two-dimensional system, when the magnetic field
is perpendicular to the sample, is discrete. Taking into account also the Zeeman
interaction, the density of states is
∞

g(E) = g [δ(E − E + g ∗ μB B) + δ(E − E − g ∗ μB B)] , (6.169)
=0

where g denotes the degeneracy (per unit area) of the th Landau level (without the
spin degree of freedom). This density of states is illustrated in Fig. 6.34(a). In the
absence of the magnetic field, the density of states per unit area (not accounting for
the spin degree of freedom) is g (2d) = m∗ /(2π2 ), problem 6.10. The Ne occupied
single-particle energy levels are represented by the dense points that fill the Fermi
circle (Ne is the number of electrons). In the presence of the field, there are only a
finite number, say ν, of occupied Landau levels up to the Fermi energy. As the total
number of states up to that energy cannot be changed by the magnetic field, each
of the Landau levels should be highly (macroscopically) degenerate. In other words,
the states of the free particle that, in the absence of the field, have (continuous)
energies in the range E − ωc /2 and E + ωc /2 are all confined in the presence
of the field to the single level E . The number of these states (per unit area, and
without the spin degree of freedom) is the product of the two-dimensional density
of states with this energy range, i.e., g = g (2d) ωc = eB/(hc). Denoting the area
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 507

Electrons in solids 507

of the two-dimensional sample by A, then the number of states belonging to each

level (per spin) is
NΦ = g A = (AB)/(hc/e) = Φ/Φ0 ,
where Φ = AB is the magnetic flux through the sample, and Φ0 = hc/e is the
magnetic flux quantum. Thus, the number of states belonging to each level is
the number of flux quanta contained within the total magnetic flux, NΦ . Using the
magnetic length from problem 6.37 gives NΦ = A/(2π 2B ), and therefore the flux
quantum corresponds to the area 2π 2B . From the expression in problem 6.37, one
finds
 
B = c/(2B) = 257Å/ B(Tesla) ,
(check!). Another derivation of the degeneracy of each Landau level appears in
Appendix E. In conclusion, when the sample contains Ne electrons, the number of
levels occupied at zero temperature is
Ne N n
ν≡ = e = e ,
NΦ g A g
where ne = Ne /A is the density of electrons in the plane. The number ν is termed
the filling factor.

(a) (b)

Fig. 6.34: The density of states of a free electron in a magnetic field normal to
the plane, at two dimensions. (a) The Landau levels, each bifurcated due to the
Zeeman interaction. (b) The same as in (a), but in the presence of disorder. The
dark areas contain localized states, that do not contribute to the conductance; the
light ones contain extended states.

The degeneracy of the Landau levels, surface states, and topological

insulators. Within the semiclassical picture, all eigenstates that belong to the same
Landau level (save one) describe each a periodic circular motion, that encompasses
the area 2π 2B . Figure 6.35 illustrates two examples for such orbits (in the center of
the figure). Due to the Pauli principle, each orbit is centered around a different point
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 508

508 Introduction to Solid-State Physics

on the plane, and therefore the plane is covered by circular paths, whose number is
the degeneracy of the level. An electron occupying such an orbit is localized around
the center, and is not free to move away; consequently, it does not contribute to the
electrical transport. However, in addition to these states there are others, near the
edges of the sample. An electron that moves along a circular path centered on the
boundary eventually collides with the surface, and then is reflected, i.e., its velocity
is reversed. It then launches on a new circular path (see the figure). This motion
along the boundary is called a skipping orbit; it is this motion that carries the
electric current along the edges of the sample. The quantum-mechanical treatment
of these edge states is detailed in Appendix E. The electrons in the skipping orbits,
or in the edge states, propagate along a single direction (shown by the arrows in
Fig. 6.35). They cannot be scattered into the opposite direction and therefore
the corresponding electric resistance vanishes. These are topological states: they
extend between the edges of the sample and the vacuum around it, and their number
is independent of the shape of the boundary [for instance, measuring the transverse
resistance of a rectangular sample in quantum-Hall regime (see below) yields the
same result as that of a circular system]. Materials for which the bulk is an insulator
while the surface conducts electricity are called topological insulators.

Fig. 6.35: Circular and skipping orbits of electrons at the Fermi energy, in the
presence of a strong magnetic field directed normal to the plane.

The quantum Hall effect. The “classical” Hall resistivity, Eq. (6.12), is
linear in the magnetic field, ρxy = −RH B. Surprisingly enough, measurements of
the Hall effect of electrons in a thin layer of GaAs [realized in a heterostructure, Sec.
2.9, which includes a thin layer of GaAs grown epitaxially between two layers of
Alx Ga1−x As; the latter is an n-type semiconductor and therefore donates electrons
to the thin layer] reveal a different behavior. At high magnetic fields, of the order
of 1-40 Tesla (i.e., 10-400 kGauss), the Hall resistance has a staircase structure,
at discrete values |ρxy | = h/(νe2 ), where ν is an integer (h = 2π is the Planck
constant). These measurements are presented in the upper part of Fig. 6.36. The
Hall resistance remains constant on the plateaux, while the “usual” (longitudinal)
resistance there is very small, as seen in the lower part of the figure. In the narrow
space in-between steps this resistance is very large. This phenomenon is termed
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 509

Electrons in solids 509

the quantum Hall effect; it is rather amazing, since it enables the detection of
quantum effects on macroscopic length scales.
There are many other measurements of the quantum Hall resistance on other
materials, including graphene. The results are universal, in the sense that they
are independent of the geometry of the sample or its composition. In fact, the
quantization of the Hall resistance reflects that of the Landau levels. In a thin layer
the width along the direction normal to the plane, Lz , is small and hence the cor-
responding wave vector, kz = 2πnz /Lz , is large, implying that at low temperatures
(the measurements shown in Fig. 6.36 for nz = 0 are taken at 8 mK) only the states
with kz = 0 are occupied (and thus the system is effectively two dimensional).

Fig. 6.36: Measurements of the quantum Hall effect on a planar layer of GaAs,
grown on Alx Ga1−x As. The upper plot shows the Hall resistance, RH B = |ρxy |;
the dashed line depicts the classical Hall resistance. The lower part portrays the
electric resistance, R = ρxx . [K. von Klitzing, The quantum Hall effect, Rev. Mod.
Phys. 58, 519 (1986).]

According to Ohm’s law [Eq. (6.7)], the resistance of a d−dimensional cube of

length L is R = ρL2−d . Thus, in two dimensions, R = ρ. When the Fermi level is
located in-between two Landau levels, that is, between two maxima in Fig. 6.34(a)
then all levels below it are populated, and the filling factor ν is an integer. When
ν is an even number, then the upmost Landau level is completely full; when it is
odd, it is only half filled (because of the spin index of the electrons). In both cases,
the density of electrons is ne = νg = νBe/(hc). Exploiting Eqs. (6.11-6.12) yields
ρxy = −RH B = −B/(ne ec) = −h/(νe2 ) . (6.170)
For a constant number of electrons in the sample, the νth Landau level is filled
when Bν = ne hc/(eν), and then there appears a step in Fig. 6.36. The next stair
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 510

510 Introduction to Solid-State Physics

occurs when the field reaches the value Bν+1 = ne hc/[e(ν + 1)]. The difference,
1/Bν+1 − 1/Bν = e/(ne hc), is a constant independent of the field, similar to the
situation in the Shubnikov-de Haas effect.
When the chemical potential (i.e., the Fermi energy) is located in one of the gaps
between the Landau levels the material is an insulator, as indeed confirmed in the
experiment [Fig. 6.22(a)]. The longitudinal conductance vanishes, σxx = σyy = 0.
Problem 6.40 shows that by inverting the conductance tensor one finds ρxx = ρyy =
0. Surprisingly – but again confirmed in the experiment – the diagonal elements in
the conductance tensor and the diagonal elements of the resistance tensor are all
zero, and off-diagonal ones are ρxy = −ρyx = 1/σyx , so that
σxx = σyy = 0 , σxy = −σyx = ν(e2 /h) ,
ρxx = ρyy = 0 , ρxy = −ρyx = 1/σyz = −h/(νe2 ) . (6.171)
Within each of the energy gaps the Hall conductance, σxy , attains discrete constant
values that are integer multiples of the basic conductance unit, e2 /h. As opposed,
when the chemical potential (i.e., the Fermi energy) coincides with one of the Lan-
dau levels, that level is not completely occupied, the electrons there are free to carry
charge, and consequently σxx = σyy is finite; its value is proportional to the number
of states in that level. These are the narrow peaks displayed in the lower part of
Fig. 6.36.

Problem 6.40.
Calculate the resistance tensor from the conductance tensor, and verify Eq. (6.171).

In the earlier experiments on the quantum Hall effect, the strength of the mag-
netic field was kept constant, but the number of the electrons was varied by the
potential applied to the sample [so that electrons are injected into (depleted from)
the system, from (into) the neighboring layers]. At low temperature, this is equiv-
alent to a modification of the Fermi energy. When one of the Landau levels in
Fig. 6.34(a) is full, and the one above it is still empty, then by varying the number
of electrons the Fermi energy may jump into the gap in-between the levels, as in
Fig. 6.31(a). When the electrons’ density attains the appropriate value, the Hall
resistance reaches the discrete value given in Eq. (6.170), and the “conventional”
resistance vanishes (as happens in the experiment). However, in clean systems,
the addition of a single electron may bring the Fermi energy into the next discrete
Landau level, and then the discrete value is achieved only for a single value of the
density. In the experiment whose data are given in Fig. 6.36 the density of the
electrons was fixed and the strength of the magnetic field was varied. One expects
that as the strength of the magnetic field is increased, the energy of each of the
Landau levels will grow as well (and also the Zeeman bifurcation); the upmost filled
level can cross the Fermi energy and depopulate into the lower levels. In this way
the discrete values of the Hall resistance are reached at discrete values of the mag-
netic field. However, in the experiment the discrete values of the Hall resistance
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 511

Electrons in solids 511

pertain to finite ranges of the magnetic field (these are the stairs in the upper part
of Fig. 6.36). A possible solution to this conundrum is to add quantum states whose
energies are in-between the Landau levels. When these states are localized, the elec-
trons that occupy them do not contribute to the longitudinal conductance (at zero
temperature they cannot move under the effect of an electric field) that remains
very low. The density of the electrons that participate in the Hall conductance is
not changed, and the Hall resistance is a constant, as in Eq. (6.170).
What happens when, in addition to the magnetic field, the electrons are also
subjected to a random potential, that is not homogeneous in space? Such a poten-
tial represents “disorder” in the system [see, e.g., part (c) of problem 6.25 and also
Sec. 6.12]. In the presence of disorder each Landau level is broadened and becomes
a discrete continuum, see the curve of the density of states in Fig. 6.34(a). The
energies close to the original Landau level (the light areas in the figure) correspond
to states which are extended over the plane; the electrons residing in them con-
tribute to the electrical conductance. The states belonging to the dark areas are
localized near minima of the random potential; the electrons occupying those do not
contribute to the conductance (at zero temperature), as they cannot move under
the effect of an electric field. Thus, the longitudinal conductance is still zero and
the Hall conductance is fixed at one of its discrete values. The next stair of the Hall
resistance occurs when the Fermi energy is increased and the region of extended
states belonging to the next Landau level is reached. When the magnetic field is
reduced (in the shadowed areas) electrons are transferred from the extended to the
localized states, but the longitudinal conductance is still very low and the Hall con-
ductance has still the discrete values. Once the extended states’ region is reached,
near the consecutive Landau level, the longitudinal conductance is increased, and
the Hall conductance is gradually moving to the next step.
The measurements of the quantum Hall effect yield the most accurate value of
the quantum unit of the conductance,
h
= 25812.807557(18) Ω ∼ 26kΩ .
e2
This quantity, sometimes called Klitzing for obvious reasons, is nowadays the in-
ternational standard of the resistance unit. Such standards are the basis of
the field of metrology, which determines the units of physical quantities. The very
accurate number (the accuracy is 10 digits!) is also relevant for the scrutinization
of the fine-structure constant in quantum electrodynamics, α = e2 /(c).
The fractional Hall effect. The quantum Hall effect as discussed so far
is also known as the integer quantum Hall effect, IQHE. There exist two-
dimensional systems whose Hall resistance contains stairs at non-integer, rational
values of the filling factor, e.g., ν = 3/7, 2/5, 1/3, etc. This phenomenon is
termed the fraction quantum Hall effect, FQHE. Its interpretation is based on
the electronic interactions, that lead to the creation of quasiparticles of fractional
charge, e.g., e/3. These quasiparticles are neither bosons nor fermions, and are
termed anyons. Their discussion is beyond the scope of this volume.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 512

512 Introduction to Solid-State Physics

6.12 A metal or an insulator?

When only single-electron states of an electron in a periodic lattice are

considered, a certain material is an insulator if the valence band is fully populated
and the conduction band is empty (at zero temperature), with a finite energy gap
in-between the two bands. While this picture describes faithfully quite a number of
solids, there are cases in which it fails. That is, a material can be an insulator though
its single-electron upmost band is not fully populated. Such cases are surveyed
tersely below.
Anderson insulator. The electrons’ wave functions do not obey the Bloch
theorem when the potential acting on them is not periodic. In particular, when
that potential is random in space with large variations of its magnitude at various
sites in the lattice (e.g., due to the presence of impurities – foreign atoms located
randomly in space) there appear electronic states centered around locations where
the potential is strongly attractive. Such states are termed localized; the proba-
bility to find an electron in them decays exponentially with the distance from the
local minimum of the potential. A similar situation is described in Sec. 5.6, for
localized states of the lattice vibrations. The similarity is further explored in part
(c) of problem 6.25.
At zero temperature, a localized electron cannot move under the effect of an
electric field. In case there are only localized electronic states around the Fermi
level, the material is an insulator. This type of insulators is named after Anderson.
In less extreme situations, the states whose energies are close to the band’s edges
are localized, while those belonging to energies around the center of the band are
extended. Once the Fermi energy passes from an “extended region” to a “localized”
one, the system changes from being an insulator to being a conductor. This is the
zero-temperature Anderson transition, and it is precisely the transition at the stairs
of the quantum Hall effect. Crossing over from a light region to a shadowed one in
Fig. 6.34(b) illustrates this transition.
The Anderson transition depends on the dimensionality of the system. In the
absence of a magnetic field, any amount of disorder (i.e., randomness) will cause a
total localization of all electrons at one and two dimensions. Thus all such materials
are insulators at zero temperature. In three dimensions, the states that correspond
to energies around the center of the band are extended, provided that the amount
of the disorder is not too high. As the level of randomness is increased, that region
is narrowed down, until all states become localized.
Mott insulatorinsulator. The description in terms of single-electron wave
functions ignores the Coulomb interactions among the electrons. These inter-
actions can be viewed as turning the electron gas into a liquid, termed the Fermi
liquid. The motion of each electron within this liquid affects the electrons around,
in such a way that those screen its electric potential. In other words, each elec-
tron is replaced by a quasiparticle, an entity composed of the electron and the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 513

Electrons in solids 513

“screening cloud” around it. At low energies the quasiparticles behave as indepen-
dent fermions, and their wave functions can be classified according to the Bloch
theorem, with parabolic energies near the band boundaries.
Consider the configuration in which each atom contributes a single electron to
the conduction band. When the atoms in the lattice are far away from each other,
that is, the lattice constant is quite large, the preferred situation is that in which
each electron remains in the atomic state belonging to a certain atom (whose energy
is 0 ), as on the left side in Fig. 4.30. This situation corresponds to a very narrow
band, as in Fig. 6.10(b), or as in the limit of long inter-atomic distance in Fig.
6.21. Another electron added to this system has to join an electron already residing
on a certain atom. The cost in energy is the short-range repulsion energy between
two electrons, denoted U (also called the Hubbard energy). When the lattice
constant is large, the energy of the second electron is 0 + U . In this way a new
band is created, at an energy distance U from the original band (see the left side
of Fig. 6.37). The original band, that can contain 2N electrons, is bifurcated into
two, each of which can contain N electrons. It follows that in its ground state, the
material is an insulator, as each electron remains localized near one of the atoms
in the lattice. This is the Mott insulators insulator. In certain configurations
these localized states also give rise to the exchange interaction (Sec. 4.7), that
causes their magnetic moments to be arranged in an antiferromagnetic pattern.
As the inter-atomic distance is decreased (e.g., due to a pressure applied on the
sample) the overlap energy between wave functions on nearest neighbors increases
and the two bands become wider. This is because the energy width, W , is propor-
tional to the overlap energy between nearest neighbors, Γ, as in Fig. 6.21. As a
result, the gap in-between them becomes narrower, and the two bands may even
overlap. This implies a transition between the insulating state to the conducting
one, called the Mott transition. The dependence of the two bands on the overlap
energy is illustrated in Fig. 6.37.

Fig. 6.37: The two energy bands formed from a single band by the repulsion interac-
tion between the electrons on the same atom in the lattice. As the lattice constant
becomes smaller, the overlap of the wave functions on neighboring atoms increases,
causing the bands to broaden, as illustrated in the figure (from left to right).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 514

514 Introduction to Solid-State Physics

6.13 Answers for the problems in the text

Answer 6.1.
a. The probability for the electron to undergo a collision during the time interval
dt is dt/τ . Hence, the probability that it will not collide during the time t + dt is
the product of the two probabilities, P (t) and (1 − t/τ ), P (t + dt) = P (t)(1 − dt/τ ).
It follows that dP/dt = −P/τ , and hence P (t) = exp[−t/τ ] [it is assumed here that
the initial condition is P (0) = 1].
b. The probability for a collision to occur precisely within the time interval between
t and t + dt is P (t)(dt/τ ). Consequently, the mean time in-between two successive
collisions is
∞
tP (t)dt/τ
t = ∞0
=τ .
0
P (t)dt/τ

Answer 6.2.
The component of the new velocity v normal to the direction of the original one
is still averaged to zero; its component along the direction of the original velocity
is the projection along that direction, that obeys v  ≥ v(τ ) cos β, as the angle
between the old and the new velocities is smaller or equal to β. By inserting this
relation into
v = v  − eEτ /m ,
one finds that
v(1 − cos β) ≥ −eEτ /m .
This result might be interpreted as yielding a generalized relaxation time, obeying
τ  ≥ τ /(1 − cos β). This implies that the effective relaxation time becomes longer as
the scattering angle is decreased, and diverges to infinity when this angle vanishes
and no scattering takes place.

Answer 6.3.
When the magnetic field is parallel to the ẑ−axis, the classical equations of motion
are
ṗx = −ωc py , ṗy = ωc px , ṗz = 0 .
It follows that pz is a constant of motion, and the motion along the ẑ−axis is with
a constant velocity, z(t) = z(0) + pz (0)t/m.
Differentiating the second equation with respect to time and inserting the first
one leads to
p̈y = ωc ṗx = −ωc2 py ,
with the solution
py (t) = py (0) cos(ωc t) .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 515

Electrons in solids 515

As
ẋ = px /m = ṗy /(mωc ) = −[py (0)/m] sin(ωc t) ,
one finds
x(t) = x(0) + [py (0)/(mωc )][cos(ωc t) − 1] .
In a similar way,
ẏ = py /m ⇒ y(t) = y(0) + [py (0)/mωc ] sin(ωc t) .
The last two equations correspond to a circular motion in the XY plane with the
cyclotron frequency ωc , and with the radius py (0)/(mωc ).

Answer 6.4.
The density of lithium and copper atoms (in CGS units) is nLi = 2/(3.49 ×
10−8 cm)3 = 4.7 × 1022 cm−3 , nCu = 4/(3.61 × 10−8 cm)3 = 8.45 × 1022 cm−3 .
Using e = 4.8 × 10−10 esu and c = 3 × 1010 cm/sec yields RH (Li) = −1.48 × 10−24
(CGS units) and RH (Cu) = −0.82 × 10−24 (CGS units).

Answer 6.5.
a. From the equations of motion (6.8) it follows that the equations for the in-plane
momentum components are identical to those analyzed following Eq. (6.8); the one
for the normal component is
ṗz  = −eEz − pz /τ .
Using p(t) = p0 exp[−iωt] yields
− iωp0x = −eE0x − ωc p0y − p0x /τ ,
− iωp0y = −eE0y + ωc p0x − p0y /τ ,
− iωp0z = −eE0z − p0z /τ . (6.172)
The current density, from Eq. (6.4), is
j(t) = j0 exp[−iωt] , j0 = −(en/m)p0 .
Using these relations in Eqs. (6.172) yields the required result,
⎡ ⎤ ⎡ ⎤⎡ ⎤
E0x 1 − iωτ ωc τ 0 j0x
⎣ E0y ⎦ = 1 ⎣ −ωc τ 1 − iωτ ⎦ ⎣ j0y ⎦ ,
0
σ0
E0z 0 0 1 − iωτ j0z
where σ0 = ne2 τ /m. The matrix is the specific-resistivity tensor ρ; its inverse is
the specific-conductivity one,
⎡ 1−iωτ −ωc τ
⎤
(1−iωτ )2 +(ωc τ )2 (1−iωτ )2 +(ωc τ )2 0
⎢ ⎥
σ = ρ−1 = σ0 ⎣ (1−iωτ )2c+(ω τ )2 (1−iωτ
ω τ 1−iωτ
)2 +(ωc τ )2 0 ⎦ .
c
1
0 0 1−iωτ
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 516

516 Introduction to Solid-State Physics

b. At high frequencies,
⎡ iω ωc ⎤
τ (ω 2 −ωc2 ) τ (ω 2 −ωc2 ) 0
⎢ −ωc ⎥
σ ⇒ σ0 ⎣ iω
τ (ω 2 −ωc2 ) τ (ω 2 −ωc2 ) 0 ⎦ .
i
0 0 ωτ
At ω ≈ ωc , the currents induced in the plane normal to the magnetic field are
rather high, even though the electric field magnitude is not. The system approaches
resonance conditions, which enable an efficient conversion of electromagnetic energy
into kinetic energy of the electrons.

Answer 6.6.
a. Assuming that the temperature gradient is along the positive x̂−axis, then
electrons impinging on the origin from the left, from the point −|vx |τ , have the
“thermal velocity” |vx [T (−|vx |τ )]|. The velocity of the electrons coming from the
opposite direction is −|vx [T (|vx |τ )]|. It follows that for a small gradient, the average
thermal velocity is
vEx  = {|vx [T (−|vx |τ )]| − |vx [T (|vx |τ )]|}/2 ≈ −|vx |τ (d|vx |/dT )(dT /dx)
= −(τ /m)[d(mvx2 /2)/dT ](dT /dx) = −(τ /[3m])(du/dT )(dT /dx)
= −(τ c/[3m])(dT /dx) .
Inserting this expression into Eq. (6.4) yields
jQ = −nevE  = (σc/(3e)∇T ,
for the charge current.
b. The absence of an electric current implies an electric field giving rise to a
current flowing along the opposite direction to that computed above, such that the
resultant vanishes, jE + jQ = 0. Exploiting Eq. (6.5) yields the current induced
by the electric field, jE = σE. One therefore finds E = Q∇T , with the Seebeck
coefficient given by Q = −c/(3e). The equipartition theorem yields c = 3kB /2, and
hence Q = −kB /(2e). This estimate exceeds by a factor of 100 the experimentally-
measured values; it is corrected by introducing modifications into the Drude theory.

Answer 6.7.
By Eq. (6.28), kF = (3π 2 n)1/3 = 1.4Å−1 . It follows that the Fermi energy is
EF = (kF )2 /(2m) = 1.25 × 10−11 erg, and the Fermi temperature is TF = 90500
K. Room temperature is far lower than the Fermi temperature.

Answer 6.8.
a. As the particles are identical, all possible arrangements of them among the
gi degenerate states yield the same result. The total number of these possibili-
ties amounts to the number of permutations of ni electrons among gi states, i.e.,
{gi !/[ni !(gi − ni )!]}. The total number of possibilities, W , is the product of the
numbers of all configurations.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 517

Electrons in solids 517

b. It follows that

ln W = {ln[gi !] − ln[ni !] − ln[(gi − ni )!]} .
i

By the Stirling formula,


ln W = {gi ln gi − gi − ni ln ni + ni − (gi − ni ) ln(gi − ni ) + (gi − ni )} ,
i

and

F = E − TS = {ni i + kB T [ni ln ni + (gi − ni ) ln(gi − ni ) − gi ln gi ]} .
i

c. Transferring δn particles from state m to state changes the free energy by

δF = (∂F/∂n − ∂F/∂nm )δn .
It follows from the condition δF = 0 that
∂F/∂n = ∂F/∂nm ,
that is, this derivative is the same for all states, and can be denoted by μ = ∂F/∂nm .
d. Differentiating the expression for F with respect to ni gives
∂F ni
μ= = i + kB T ln ,
∂ni g i − ni
that is
ni = gi /{1 + exp[β(i − μ)]} ,
as in Eq. (6.35) for gi = 1.

Answer 6.9.
a. The derivative of the function in Eq. (6.35) is
−f  = −∂f (E)/∂E = βeβ(E−μ) /[1 + eβ(E−μ) ]2 .
This derivative is displayed in Fig. 6.38 for β = 10, 20, 40, where energies are
measured in units of μ. The function is concentrated symmetrically around the
value E = μ,
−f  (μ − E) = βe−β(E−μ) /[1 + e−β(E−μ) ]2 = −f  (E − μ) ,
where it takes the value −f  (E = μ) = β/4. As the temperature is lowered this
height increases while the width decreases. In particular, at zero temperature the
height tends to infinity while the width shrinks to zero. There are several options
to estimate the width. For instance, the energy at which the height is halved, i.e.,
the solution of
f  (E1/2 )/f  (μ) = 1/2 , i.e. 4x/(1 + x)2 = 1/2 , where x = exp[β(E1/2 − μ)] .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 518

518 Introduction to Solid-State Physics

Fig. 6.38
√ √
One finds x = 3 ± 2 2, and consequently E1/2 − μ = kB T ln[3 ± 2 2] = ±1.76kB T .
The width ΔE1/2 = 3.52kB T is proportional to the temperature, and indeed de-
creases with the temperature. The integral over this function is nonetheless a con-
stant,
 ∞ ∞

dE[−f  (E)] = −f (E) = f (−∞) − f (∞) = 1 ,
−∞ −∞

independent of the temperature. In the T → 0 limit, f  (E) = 0 for any E = μ,

while f  (E = μ) = ∞. As the integral equals 1, the function −f  (E) obeys the
same properties as a delta function; hence −f  (E) → δ(E − μ).
b. The average (per particle) of a certain property A(E) of the electron gas at an
arbitrary temperature T is
 ∞
A(E) = V dEg(E)A(E)f (E)/Ne  ,
−∞
 ∞
Ne  = V dEg(E)f (E) . (6.173)
−∞
We introduce the function
 E
B(E) = V dE  g(E  )A(E  ) ,
−∞
and integrate by parts the numerator in Eq. (6.173),
 ∞ ∞  ∞

dEg(E)A(E)f (E) = B(E)f (E) − dEB(E)f  (E)
−∞ −∞ −∞
 ∞
= dEB(E)[−f  (E)] .
−∞
The first term in the intermediate step vanishes in both limits [B(−∞) = 0 and
f (∞) = 0]. As the function −f  (E) is narrow, and is concentrated around the Fermi
level, the integral on the right hand-side is significant mainly at energies for which
|E − μ| ≤ kB T . Therefore, at low enough temperatures one may use the expansion
B(E) = B(μ) + (E − μ)B  (μ) + (E − μ)2 B  (μ)/2 + . . . ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 519

Electrons in solids 519

to obtain
 ∞
dEB(E)[−f  (E)] = B(μ) + B  (μ)(δE)2 + . . . .
−∞

The first term here is the zero-temperature result, and the second contains (δE)2 =
∞
−∞
dE(E − μ)2 [−f  (E)], which is the standard deviation of the energy off the
chemical potential with the weight function −f  (E). As seen in part (a), the width
of this function is proportional to T , and therefore the leading-order correction at
low temperatures is proportional to T 2 [and indeed, a precise computation of the
integral yields (δE)2 = (π 2 /3)(kB T )2 ]. The same correction appears in the integral
in the expression for Ne ; it follows that the average of any quantity calculated
with the Fermi-Dirac distribution contains corrections to the zero-temperature value
which are quadratic in the temperature. The higher terms include higher powers of
the temperature.

Answer 6.10.
a. The volume of the Fermi sphere in d dimensions is Sd kFd /d and the volume that
corresponds to a discrete “allowed” value of the momentum is (2π)d /V . Each such
state accommodates two electrons. Hence, an extension of Eq. (6.28) is
2Sd kFd /d 2Sd (2mEF )d/2
Ne = d
= V = nV .
(2π) /V d(2π)d
It follows that kF = 2π[nd/(2Sd )]1/d , and the Fermi energy is
EF = (2π)2 [nd/(2Sd )]2/d /(2m) .
The electrons’ density is
n = 2Sd [kF /(2π)]2 /d = 2Sd (2mEF )d/2 /[d(2π)2 ] .
b. Similarly, an extension of Eq. (6.31) to an arbitrary dimension is
2Sd k d−1 dk nd
E 2 −1 dE .
d
V g(E)dE = d
=V d/2
(2π) /V 2(EF )
The total energy of the electron gas at zero temperature is hence
 EF
d
Etot = V Eg(E)dE = V nEF .
0 d+2
c. The average number of the electrons is
 ∞
Ne  = V dEg(E)f (E) .
0

Equating it to N = V n, and using for g(E) the expression from the previous part,
yield
 ∞ 
E 2 −1 d(kB T )d/2 ∞ y 2 −1
d d
d
1= dE = dy . (6.174)
2(EF )d/2 0 1 + eβ(E−μ) 2(EF )d/2 0 1 + ey−βμ
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 520

520 Introduction to Solid-State Physics

The solution of this equation is the chemical potential μ as a function of β and

EF . At very high temperatures the coefficient in front of the last integral tends
to infinity, and consequently the integral should tend to zero, i.e., exp[−βμ] → ∞.
The 1 in the denominator may then be discarded, and the distribution function
becomes approximately f (E) → exp[−β(E − μ)], with the proper normalization.
The integral in Eq. (6.174) becomes
 ∞ d −1
y2
y−βμ
= eβμ Γ(d/2) ,
0 e
 √
where Γ(z) is the Gamma function, Γ(3/2) = π/2, Γ(1) = 1, and Γ(1/2) = π.
The equation for the chemical potential gives that the product (kB T )d/2 exp[βμ] is
independent of the temperature, and thus the leading-order dependence on the tem-
perature, at high temperatures, is μ ≈ −(d/2)(kB T ) ln(kB T ). The factor exp[βμ]
in the distribution function is cancelled in the calculation of any weighed aver-
age; therefore in the high-temperature limit, the results are those of the Maxwell-
Boltzmann distribution, e.g.,
∞ d/2 ∞ d/2 −x
Etot  0
E f (E)dE 0
x e dx kB T
= ∞ d/2−1 = kB T ∞ d/2−1 −x =d .
Ne  0
E f (E)dE 0
x e dx 2
This is the equipartition theorem, that also leads to the Dulong-Petit law.
d. In two dimensions, the equation for the chemical potential (from the previous
part) is
 ∞  ∞
1 1 1
1= dEg(E)f (E) = dE β(E−μ) = ln[1 + eβμ ] ,
0 E F 0 e + 1 βE F
and consequently μ = ln[eβEF − 1]/β.
e. The arguments used in the estimation of the specific heat still hold: the number
of electrons excited above the Fermi level is proportional to T ; each of them carries
an energy of the order of kB T . The specific heat is then proportional to T . However,
as seen after Eq. (5.69), the phonon specific heat is proportional to T d . Hence, the
extension of Eq. (6.42), upon taking into account the contribution of the phonons,
is
C V ≈ γ d T + Ad T d .

Answer 6.11.
a. The density of the electrons is given by n = 6.022 × 1023 z ρm /A, where z
is the number of conduction electrons “donated” by each atom, A is the atomic
mass (in grams) and the coefficient is the Avogadro number (the number of atoms
in one mole). For copper, z = 1 and A ≈ 64, and hence n = 0.84 × 1029 m−3 .
From the expressions following Eq. (6.6) one finds τ = m/(ρne2 ) = 2.7 × 10−14
sec. Equation (6.28) implies that kF = (3π 2 n)1/3 = 1.35 × 108 cm−1 , and then
vF = kF /m = 1.56 × 108 cm/sec, EF = mvF2 /2 = 13.8 eV, TF = EF /kB = 1.6 × 105
K, and = vF τ = 4.2 × 10−6 cm= 420Å.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 521

Electrons in solids 521

b. The relaxation time is τimp = m/(ρme2 ) = 0.6 × 10−14 sec, and then using the
expression before Eq. (6.48), σimp = 1/(τimp vF Nimp ) = 12.5(Å)2 .

Answer 6.12.
a. The function ψnk (r) obeys Eq. (6.59),

ψnk (r + R) = T̂ (R)ψnk (r) = exp[ik · r]ψnk (r) .

Complex conjugating yields

∗ ∗ ∗
T̂ (R)ψnk (r) = ψnk (r + R) = exp[−ik · r]ψnk (r) .
∗
Hence, ψnk (r) is an eigenfunction of T̂ (R), with the eigenvalue exp[−ik · r]. The
function ψn−k (r) is an eigenfunction of T̂ (R) too, with the same eigenvalue. As
the eigenvalues of T̂ (R) for a certain value of n are non-degenerate, it follows
∗
that ψnk (r) = ψn−k (r). Complex-conjugating the Schrödinger equation yields
∗ ∗
Ĥ(r)ψnk (r) = En (k)ψnk (r). The replacement k → −k in the Schrödinger equation
gives Ĥ(r)ψn−k (r) = En (−k)ψn−k (r). Comparing the two equations implies that
En (k) = En (−k).
b. Denote k = (−kx , ky , kz ). From Eq. (6.59),

ψnk (r + a1 ) = T̂ (a1 )ψnk (r) = exp[ikx a1 ]ψnk (r) ,

and upon complex-conjugating,

∗ ∗ ∗
ψnk (r + a1 ) = T̂ (a1 )ψnk (r) = exp[−ikx a1 ]ψnk (r) .

On the other hand,

ψnk (r + a1 ) = T̂ (a1 )ψnk (r) = exp[−ikx a1 ]ψnk (r) .

∗
As the eigenvalues of T̂ (a1 ) are not degenerate, it follows that ψnk (r) = ψnk (r).
Finally, complex-conjugating the Schrödinger equation for ψnk (r) and a comparison
of the same equation for ψnk (r) yields that En (k) = En (k ).

Answer 6.13.
According to the Bloch theorem, it suffices to consider the solutions of the
Schrödinger equation in the first Brillouin zone, −π/a ≤ k < π/a (in-between
the vertical lines in the figure). The one-dimensional Schrödinger equation with
periodic boundary conditions has solely two independent solutions for each energy
En (k). It is found in problem 6.12 that these two solutions are ψnk (x) and ψn−k (x)
∗
[related to one another by ψn−k (x) = ψnk (x)]. Figure 6.6 shows that for the energy
represented by the horizontal line there are four independent solutions, in contra-
diction with the requirement that there are only two independent solutions. As
explained following Eq. (6.62), in higher dimensions there are more independent
solutions, and consequently the bands may well overlap each other.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 522

522 Introduction to Solid-State Physics

Answer 6.14.
a. In the limit specified in the problem, each potential barrier becomes very high
b
and very narrow; the area it covers is a dxU (x) = U0 (a − b) = u0 . This step
has precisely the same properties as the function u0 δ(x − a), from which follows
the comb’s shape. Assuming that the energy of the electron remains finite, i.e.,
E  U0 , one may neglect K 2 as compared to Q2 ≈ −2mU0 /2 , turning Eq. (6.73)
into

cos(ka) = cos(Ka) − [Q2 (a − b)/2K] sin(Ka) ,

and hence

cos(ka) = F (K) = cos(Ka) + φ sin(Ka)/(Ka) .

This reproduces the solutions of a free particle for φ = 0, i.e., ka = Ka + 2 π. The

borders of the bands are obtained from the relation F (K) = ±1; since F (K) =
(−1)n for Ka = nπ where n is an integer, these points are always borders of energy
bands.
b. In the φ  1 limit, the energy bands are narrow because F (K) varies steeply as
a function of K [and therefore also as a function of E = 2 K 2 /(2m)] within each
band. Consequently, close to each border found in part (a) there is another one,
where the function takes the value (−1)n+1 . Assuming that these borders are close
to each other, one defines Ka = nπ + κ, and expands in the small variable κ. The
equation obtained for the second border is then

(−1)n+1 = cos(nπ + κ) + φ sin(nπ + κ)/(nπ + κ) = (−1)n [cos κ + φ sin κ/(nπ + κ)]

≈ (−1)n [1 − κ2 /2 + . . . + φ(κ − κ3 /6 + . . .)/(nπ + κ)] ≈ (−1)n [1 + φκ/(nπ)] .
(6.175)
To leading order κ ≈ −2nπ/φ, and the bands are in the ranges

nπ(1 − 2/φ) < Ka < nπ , for n = 1, 2, 3, . . . .

Fig. 6.39
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 523

Electrons in solids 523

Denoting Kn,max = nπ/a and Kn,min = nπ(1 − 2/φ)/a, one finds that the width of
the n−th band is

ΔEn = 2 (Kn,max
2
− Kn,min
2
)/(2m) ≈ 22 n2 π 2 /(mφ) ,

and hence the bands widen as the energy increases. Within each band Ka =
nπ(1 − x), where 0 < x < 2/φ. Inserting this into the equation for F (K), cos(ka) =
F (K) ≈ (−1)n (1 − φx), yields x ≈ [1 − (−1)n cos(ka)]/φ, and hence the energy in
the band is

E = 2 K 2 /(2m) = 2 (nπ)2 (1 − x)2 /(2ma2 ) .

These energies are plotted in Fig. 6.39, for φ = 20. In the φ → ∞ limit these
are the discrete energies of an infinite potential well of width a; the leading-order
correction to those is proportional to x,

2 K 2 2 (nπ)2 2 
E= = 2
1− 1 − (−1)n cos(ka) .
2m 2ma φ

c. When φ  1 the second term in Eq. (6.175) is a minor correction to the free-
particle solutions, for which ka = Ka + 2 π. As it is found that there exists a
border of a band at each point Ka = nπ, when the right hand-side equals (−1)n ,
one considers another point close to this one, at which the right hand-side has the
same value. That is, denoting Ka = nπ + y and requiring that

(−1)n = cos(nπ + y) + φ sin(nπ + y)/(nπ + y) ≈ (−1)n [1 − y 2 /2 + φy/(nπ)] ,

it follows that y ≈ 2φ/(nπ). Thus, the difference between the two energies on the
opposite sides of the energy gap is

ΔE = 2 [(Ka + y)2 − (Ka)2 ]/(2ma2 ) ≈ 22 φ/(ma2 ) = 2u0 /a .

The bottom of the lowest band is obtained from the solution of the equation

1 = cos(Ka) + φ sin(Ka)/(Ka) ≈ 1 − (Ka)2 /2 + φ[1 − (Ka)2 /6] ,

and therefore (Ka)2 ≈ 2φ, which yields

Emin ≈ 2 (Ka)2 /(2ma2 ) ≈ 2 φ/(ma2 ) = u0 /a .

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 524

524 Introduction to Solid-State Physics

Answer 6.15.
When the periodic potential U (r) is weak, it can be regarded as a small pertur-
bation. In the absence of the potential the Hamiltonian describes a free particle,
Ĥ0 = −2 ∇2 /(2m), whose solution is discussed in Sec. 6.3. Using periodic
√ bound-
(0)
ary conditions, the eigenfunctions of Ĥ0 are Ψk (r) = exp[ik · r]/ V , and the
corresponding eigenenergies are E (0 (k) = 2 k2 /(2m) [with the discrete values of
the wave vectors from Eq. (5.43) as reproduced in Eq. (6.27)]. As explained in
App. C, the corrections to the energy levels and the wave functions are obtained
(0) (0)
from the matrix of the perturbation Hamiltonian, Ψk |U |Ψk . When the wave
vector of the electron is far away from Brillouin-zone boundaries the “original”
states (without the perturbation) are not degenerate (see Sec. 6.6), and then Eq.
(C.3) yields
 |Ψ(0) (0) 2
(0) (0) k |U |Ψk |
E(k) = E (0) (k) + Ψk |U |Ψk  − . (6.176)
k =k
E (0) (k ) − E (0) (k)
The matrix elements of the perturbation are given by

(0) (0) 1 
 (k − k) .
Ψk |U |Ψk  = d3 rei(k−k )·r U (r) ≡ U
V
(0) (0)
The first-order correction is given by the diagonal term, Ψk |U |Ψk  =
d3 rU (r)/V . This is just the average of the potential over the whole crystal;
it is a constant independent of k, and as such can be subtracted from all energy
levels (i.e., it shifts the origin of the energy) and then ignored. As explained in Sec.
3.8, the only nonzero Fourier components of a periodic function are those for which
the wave vector equals one of the reciprocal-lattice vectors, G. Hence, the sum in
Eq. (6.176) comprises only terms for which k − k = G,
  (G)|2
|U
E(k) = E (0) (k) − + ... ,
G=0
E (0) (k − G) − E (0) (k)
reproducing Eq. (6.88).
In a similar fashion App. C gives the correction to the wave function,
(1)
 (0) (1) (0)
  (G)
U (0)
Ψk = Ψk−G |Ψk Ψk−G = Ψ + ... .
G G=0
E (0) (k − G) − E (0) (k) k−G

When the unperturbed states are degenerate (or when the difference between two
“original” energies is not large enough), App. C indicates that the degenerate
states have to be treated separately, choosing linear combinations of them which
are eigenstates of the perturbation Hamiltonian. This calculation yields the first-
order correction to the energy. In the one-dimensional configuration analysed in
Sec. 6.6 there is double degeneracy only near each reciprocal-lattice vector, with
the perturbation matrix
 
E (0) (k) − E (0) (G/2)  (G)
U
 (−G) .
U E (0) (k − G) − E (0) (G/2)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 525

Electrons in solids 525

The eigenvalue equation of this matrix is identical to Eq. (6.89).

Answer 6.16.
The Fourier transform of the potential is U (G) = (1/a) a/2 dx exp[−iGx]U (x) =
−a/2
u0 /a. Hence, the equation for the Fourier coefficients of the wave function is
 2 k 2  u 

− E ψ(k) + 0 ψ(k − G) = 0 .
2m a
G

The second term contains the function f (k) =  − G). It follows that
ψ(k
G

 u f (k)
ψ(k) =− 0 2 2 ,
a  k /(2m) − E
which, for k → k − G yields

 − G) = − u0 f (k − G)
ψ(k .
a  (k − G)2 /(2m) − E
2

As the sum representing f (k) contains all reciprocal-lattice vectors, one may shift
the summation index and prove that
 
f (k − G ) =  − G − G ) =
ψ(k  − G ) = f (k) ,
ψ(k
G G

and therefore
 − G) = − u0
ψ(k
f (k)
.
a 2 (k − G)2 /(2m) − E
It remains to sum this expression over G to obtain
u  1
f (k) = − 0 f (k) ,
a 2 (k − G)2 /(2m) − E
G

which has a solution provided that

 1 2 a
= − , E = 2 K 2 /(2m) .
(k − G)2 − K 2 2mu0
G

The last equation relates K, i.e., the energy, to k.

When u0 is very small, the right hand-side is rather large, and then the sum
on the left side is dominated by the terms with the smallest denominators. In the
u0 → 0 limit this yields E = 2 (k − G)2 /(2m), reproducing the reduced spectrum
of a free particle. When ka is away from the zone boundaries, an expansion in
powers of u0 reproduces the second-order correction, which is also obtained from
perturbation theory (Sec. 6.6). On the border of a Brillouin zone, e.g., at ka = π,
the sum is dominated by two terms, with G = 0 and G = 2π/a; then the result
reproduces the energy gap, proportional to u0 . In the opposite limit, when u0 → ∞,
the right hand-side vanishes, and a plot of the left side, as a function of K, shows
that it is zero for any value of k when Ka = 2πn; this yields the discrete energy
levels of an infinite potential well. (Check!)
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 526

526 Introduction to Solid-State Physics

The equation derived in problem 6.14 is obtained as follows. The denominator

in the left hand-side is (k − G)2 − K 2 = (k − G − K)(k − G + K), leading to
 1 a  1 1
= − .
(k − G) − K
2 2 2K ka − Ka − Ga ka + Ka − Ga
G G

(6.177)
Utilizing the identity given in the problem and other trigonometrical relations, and
G = 2πn/a, gives
 1 a  1 1
= −
(k − G) − K
2 2 4K n a(k − K)/2 − nπ a(k + K)/2 − nπ
G
a ka − Ka ka + Ka a sin(Ka)
= cot( ) − cot( ) = .
4K 2 2 2K cos(Ka) − cos(ka)
The last equation is identical to the result of problem 6.14.

Answer 6.17.
The energy gaps are determined by the Fourier transforms of the potential, Eq.
 (G) = a dx exp[−iGx]U (x)/a, where the integration is carried out over
(6.77), U 0
the unit cell. Inserting here Eq. (6.65) gives
 a
 (G) = U0 U  
U dxe−iGx = 0 e−iGb − e−iGa .
a b iGa
The energy gap at k = G/2 is

 (G)| = 2U0 2(1 − cos[G(a − b)]) = 4U0 | sin[G(a − b)/2]| ,
2|U
|G|a |G|a
where the identity | exp[iα] − exp[iβ]|2 = 2 − 2 cos(α − β) = 4 sin2 [(α − β)/2] is used.
As expected, and indeed found in Sec. 6.5, the gap decreases as |G| increases, that
is, in-between higher-energy bands. In the delta-function limit, U0 (a − b) → u0 , one
finds

2|U (G)| ≈ 4U0 |[G(a − b)/2]| → 2u0 ,

|G|a a
in agreement with the results of problem 6.14(a).

Answer 6.18.
Figure 6.40(a) reproduces Fig. 3.14; the parts of the third Brillouin zone are in-
dicated. Displacing those by reciprocal-lattice vectors shifts them all into the first
Brillouin zone, as shown in Fig. 6.40(b).
Figure 6.41(a) displays the dependence of the energy on the components of k,
once the energies are shifted into the first Brillouin zone. Figure 6.41(b) shows the
equal-energy curves. The minima are located at the corners of the zone, and the
maxima are at the centers of the faces.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 527

Electrons in solids 527

(a) (b)

Fig. 6.40

(a) (b)

Fig. 6.41

Figure 6.42 summarizes the results for the three bands, along a special path in
the first Brillouin zone. As expected, the third and the second bands overlap over
the diagonal of the zone (in much the same way as the second and the first bands
overlap on the faces).

Answer 6.19.
a. Assume that the boundary between two Brillouin zones is given by the vectors q0
that obey Eq. (3.26) [see also Fig. 3.13], 2q0 ·G = G2 where G is a reciprocal-lattice
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 528

528 Introduction to Solid-State Physics

Fig. 6.42

vector. In a three-dimensional system, Eq. (6.91) becomes


(1)  (G)|2 /2 .
Eq± = E (0) (q) + E (0) (q − G) ± [E (0) (q) − E (0) (q − G)]2 + 4|U
(6.178)
For q = q0 it yields
(1)
= E (q0 ) ± |U
Eq 0 ± (0)  (G)|, and therefore the energy gap at
 (G)|.
each point on the border plane (or line) is constant, equals 2|U
b. An expansion around the point q0 , q = q0 + δ, gives
 (G)| + +  δ 2 + δ · (2q − G) ±  (δ · G) .
2 2 2
(1) (1)
Eq± ≈ Eq0 ± (q0 ) ± |U
2m 0
4m|U (G)|
When δ  G then it is perpendicular to the plane between the two zones. In
particular, δ · (2q0 − G) = 0, and consequently
 (G)| +  1 ±  G
2 2 2
(1) (1)
Eq± ≈ Eq0 ± (q0 ) ± |U δ2 ,
2m 2m|U  (G)|
which is analogous to Eq. (6.92) pertaining to the one-dimensional case. For a
weak enough potential it is plausible that 1 < 2 G2 /(2m|U  (G)|). Therefore the
coefficient of the quadratic term is positive for the upper band and negative for the
lower one. As the dependence is quadratic, the energy of the lower band possesses
a maximum on the plane between the two zones, and an equal-energy surface in-
tersects that plane normally (i.e., the energy is not changed when the boundary is
crossed), as indeed seen in Figs. 6.14(c) and (d). The energy of the upper band
is minimal along the same direction, and there again the equal-energy curves are
perpendicular to the boundary plane.
On the other hand, when δ is on the border plane then q = q0 + δ lies there
 (G)|. These are two paraboloids
as well, and the energies are given by E (0) (q) ± |U
shifted one against the other by the energy gap. It follows that at the middle of the
zone’s face the upper band is minimal along all directions; the coefficients of the
parabolas, however, are different along the different directions. The lower band has
a saddle point: a maximum perpendicular to the plane and a minimum within it.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 529

Electrons in solids 529

(a) (b)

Fig. 6.43

c. Figures 6.14(c) and (d) can be deduced from Eq. (6.178) by using an approx-
imation: in each quarter of the Brillouin zone [Fig. 6.15(b)] one uses the vector
q0 corresponding to the boundary separating that quarter from the closest zone.
As expected, away from the zone boundaries the plots are quite similar to those
in Figs. 6.14(a) and (b). The dissimilarities appear near the zone boundaries. In
the vicinity of the zone faces the normal derivative of the energy vanishes. As the
energy of the lower band is smaller than the one of a free electron, the equal-energy
curves are displaced outwards, intersecting perpendicularly the face (as explained
quantitatively in the previous part of this solution).
In the absence of the periodic potential the energy is E (0) = 2 k2 /(2m), and
hence the equal-energy curves are planar circles. A weak periodic potential leaves
them almost unchanged, except close to the zone boundaries. Figure 6.43(a) shows
the energy along the line connecting the center of the Brillouin zone with the center
of one of the faces, e.g., M . As this face is approached, the energy is affected by
the gap there, and then the lower band (parabolic for a free electron) is shifted
downwards, from the thin to the thick line. The intersection of the horizontal line
in the figure, E0 , with the function E(k) indicates the location of the equal-energy
curve whose energy is E0 . The figure shows that the intersection point moves to the
right upon passing from the original parabola to the new energy, that is affected
by the gap. Hence, the points on the equal-energy curves, which are circular for
a free particle, move towards the centers of the faces of the zone, as illustrated by
the arrows in Fig. 6.43(b). The circular curves are gradually deformed, until the
boundary curve turns into a square rotated by 45◦ relative to the Brillouin zone.
Increasing further the energy, the equal-energy curve reaches the maximum of the
thick line where it is normal to the face of the zone.
d. At the corners of the Brillouin zone there is an additional degeneracy, since there
the energy is unchanged with respect to displacements by more reciprocal-lattice
vectors. For example, in a square lattice the energy of the free particle is degenerate
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 530

530 Introduction to Solid-State Physics

at the four corners of the first Brillouin zone, while at the corners of the cubic lattice
the degeneracy is eight. To find the energy gaps in such cases one has to diagonalize
the perturbation matrix among all degenerate states. In the example of the square
lattice, the degenerate energies on the corners are e0 = 2 π 2 /(ma2 ), and the matrix
to be diagonalized, accounting only for the Fourier components mentioned in the
problem, is
⎡ ⎤
e0 u u v
⎢ u e0 v u ⎥
⎢ ⎥  
⎣ u v e u ⎦ , with u = U (kM ) = 0.2 , v = U (kK ) = 0.15 .
0
v u u e0

The rows and the columns represent the corners, k = (π/a)(1, 1), (π/a)(−1, 1),
(π/a)(1, −1) and (π/a)(−1, −1). This matrix has two eigenvectors with eigenvalue
e0 − v, (−1, 0, 0, 1) and (0, −1, 1, 0), an eigenvector with the eigenvalue e0 − 2u + v,
(1, −1, −1, 1), and another one with eigenvalue e0 + 2u + v, (1, 1, 1, 1). It follows
that out of the four degenerate states two remain degenerate, with no change in
their energy, while the other two are shifted upwards and downwards. Assuming
that u > v, the lower band moves further down, to the energy e0 − 2u + v, and the
next band to the energy e0 − v. Thus the energy gap at the point K is 2(u − v).
This gap is smaller than the one at M , that is equal to 2u. Figures 6.14 (c) and (d)
and 6.18(b) are computed by generalizing the above determinant, with the energies
in the diagonal elements that correspond to an arbitrary point in the Brillouin zone.
For instance, when kx > 0 and ky > 0, the energies on the diagonal are
2π 2π 2π 2π
E (0) (kx , ky ) , E (0) (kx − , k ) , E (0) (kx , ky − ) , E (0) (kx − ,k − ).
a y a a y a
Near the center of the Brillouin zone it is enough to keep the first row; near the
M point it suffices to keep the first two rows (as done in the previous parts of the
problem), and near the K point all four rows have to be used.

Answer 6.20.
Displacing r by R in Eq. (6.101) yields

ϕk (r + R ) = Ak exp[ik · R]ϕ(0) (r + R − R) .
R

Setting the dummy index of the sum to be R = R − R gives


ϕk (r + R ) = exp[ik · R ]Ak exp[ik · R ]ϕ(0) (r − R ) ,
R

as required by Eq. (6.54). The function


uk (r) = exp[−ik · r]ϕk (r) = Ak exp[−ik · (r − R)]ϕ(0) (r − R) ,
R
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 531

Electrons in solids 531

is periodic, since a displacement of r by R and a change of the dummy index from

R to R − R give

uk (r + R ) = e−ik·(r+R ) ϕk (r + R )
 
= Ak e−ik·(r+R −R) ϕ(0) (r + R − R) = e−ik·r ϕk (r) = uk (r) ,
R

as required in Eq. (6.53). Displacing k of Eq. (6.101) by a reciprocal-lattice vector

G, and using Eq. (3.17) give
 
ϕk+G (r) = Ak+G ei(k+G)·R ϕ(0 (r − R) = Ak+G eik·R ϕ(0) (r − R) .
R R

By the normalization of the wave function |Ak+G | = |Ak |2 . The wave function
2

is determined up to a phase (that does not change any quantum average); thus
ϕk+G (r) = ϕk (r).

Answer 6.21.
a. The energies in the band are
E(k) = Γ1 {1 + 2[1 − cos(ka)] − 2ζ[1 − cos(2ka)]}
= Γ1 {1 + 4 sin2 (ka/2) − 4ζ sin2 (ka)} .
Figure 6.44(a) portrays these energies in the first Brillouin zone (in unit of Γ1 ), for
ζ = 1/8 (the thick line) and for ζ = 1/4 (the thin line). The leading-order behavior
at the bottom of the band is
E(k) − Γ1 ≈ Γ1 (1 − 4ζ)(ka)2 + O(k 4 ) .
It is parabolic for ζ = 1/8, but E(k) − Γ1 ∝ k 4 for ζ = 1/4. Larger negative values
of Γ2 give rise to four solutions for energies close to zero, which is impossible [see
problem 6.13]. This means that the tight-binding approximation breaks down for
such values.
b. Similar to Eq. (5.56), the density of states is g(E) = (1/π)dk/dE. In the present
case the derivative is
dE
= Γ1 sin(ka)[1 + 4(1 − ζ) − 32ζ sin2 (ka/2)] .
dk
The density of states diverges near the center of the Brillouin zone and at its bound-
aries. This van Hove singularity emerges from the vanishing of the derivative of the
energy with respect to the wave number at these points. The result, in units of
1/(Γ1 a) is exhibited in Fig. 6.44(b) for ζ = 0.1.
c. At zero temperature all levels up to the Fermi energy are filled, and hence Etot =
E
L Γ F dEg(E)E. From the condition EF − Γ1  Γ1 it follows that E − Γ1  Γ1
1
for all filled energies, and thus the parabolic approximation, E(k) − Γ1 ≈ Γ1 (1 −
4ζ)(ka)2 , can be used. As a result,
1 dk 1
g(E) = ≈ ,
π dE 2πΓ1 (1 − 4ζ)ka2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 532

532 Introduction to Solid-State Physics

(a) (b)

Fig. 6.44

leading to
 
EF
E − Γ1 (E − Γ1 )3/2
Etot = N Γ1 + L dE = N Γ1 + N F .
Γ1 2πa Γ1 (1 − 4ζ) 3π Γ1 (1 − 4ζ)
The resemblance to the density of states of a free particle is conspicuous: the density
of states is inversely proportional to the square root of the energy (relative to the
bottom of the band). The difference is in the numerical coefficient.

Answer 6.22.
a. Inserting the function ϕ(0) (x) = A exp[−κ|x|], with κ > 0, into the Schrödinger
equation shows that it is indeed a solution for any x = 0, with the energy E (0) =
−2 κ2 /(2m). At the vicinity of the point x = 0, the Schrödinger equation is
2 d2 ϕ(0)
− − u0 δ(x)ϕ(0) = E (0) ϕ(0) .
2m dx2
Integrating it between − to  yields, in the  → 0 limit,
2 dϕ(0)  dϕ(0) 
−  +−  − u0 ϕ(0) (0) = 0 ,
2m dx x=0 dx x=0−
i.e., −[2 /(2m)][−κA − κA] − u0 A = 0, and hence κ = u0 m/2 . The normalization
of the wave function implies that
 ∞
1= dx|ϕ(0) (x)|2 = |A|2 /κ ,
−∞
√
and consequently A = κ. In this way a unique solution is obtained, and since its
energy is negative, E (0) = −mu20 /(22 ), it is the ground state.
b. The overlap integral between nearest neighbors is
  0  a
∗ (0)
α = dx[ϕ (x)] ϕ (x − a) = κ
(0)
dxe κ(2x−a)
+ dxe−κa
−∞ 0
 ∞
+ dxe−κ(2x−a) = κae−κa + cosh(κa) .
a
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 533

Electrons in solids 533

The integral yielding γ, for nearest neighbors, is

 
∗
γ = − dx[ϕ (x)] ΔU ϕ (x − a) = u0
(0) (0)
dx[ϕ(0) (x)]∗ δ(x − ma)ϕ(0) (x − a)
m=0

−κ|ma| −κ|m−1|a
= −u0 κ e e = −2u0 κ[e−κa + 2e−3κa + 2e−5κa + . . .]
m=0

≈ −2u0 κe−κa .

Likewise,
 
β =− dx[ϕ(0) (x)]∗ ϕ(0) (x) = −u0 |ϕ(0) (x)|2 δ(x − ma)
m=0

−2κ|ma|
= −u0 κ e ≈ −2u0 κe−2κa .
m=0

In the tight-binding approximation κa = u0 am/2  1 and therefore |β|  |γ|.

Discarding the correction to the normalization in the denominator of Eq. (6.110)
gives E(k) ≈ E (0) − β − 2γ cos(ka). The width of this band is 2γ = 4u0 κ exp[−κa].

Answer 6.23.
For the configuration described in this problem, Eq. (6.122) gives rise to two equa-
tions,
(0)
(E − EA )AA = −
γAA (k)AA − γ
BA (k)AB ,
(0)
(E − EB )AB = −
γBB (k)AB − γ
AB (k)AA ,

where γij (k) = R γij (R) exp[−ik·R], and when all overlap integrals save αii (0) =
1 are ignored. The two equations possess a nontrivial solution provided that their
determinant vanishes. Hence,

(0) (0) (0) (0)
E = EA + E B − γ AA − γ BB ± (EA + EB − γ AA − γBB )2 + 4|γAB |2 /2 .

When only nearest-neighbors hopping are retained, as stated in the problem, then
(0) (0)
AB = 2γAB cos(ka/2), where γAB = ϕA (x)|ΔU |ϕB (x−a/2), and γ
γ AA = γBB =
0. With these approximations,

(0) (0) (0) (0)
E = EA + EB ± (EA − EB )2 + 16|γAB |2 cos2 (ka/2) /2 .

Figure 6.45(a) displays the two bands for different atoms; it resembles qualitatively
Fig. 5.9 that portrays the acoustic and optical bands of lattice vibrations in a
crystal with two atoms in the unit cell. When the two atoms are identical, the unit
cell is twice smaller, and the Brillouin zone is doubled, precisely as in the case of
lattice vibrations, see Fig. 6.45(b).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 534

534 Introduction to Solid-State Physics

(a) (b)
Fig. 6.45

Answer 6.24.
a. Recall that the graphene crystal is represented by a triangular lattice with two
atoms in the unit cell. The wave functions on the different atoms can be indexed as
in Fig. 5.10, with unm and vnm representing the wave functions in the two points
in the nm−unit cell. As in problem 6.23, there are two linear equations for the
wave functions on the two sites in the unit cell. Since the atoms are all identical,
(0) (0)
energies are given by E± = E ± | γAB |, with
(0)
EA = EB , and the
−ik·rnn −ik·a1 −ik·a2
AB (k) =
γ γAB e = γAB (e +e + e−ik·(a1 +a2 ) ) ,
nn √
where the sum is over nearest neighbors; a1 = (a, 0) and a2 = (a, 3a)/2. [The
expression resembles the one in Eq. (5.33), due to the similarity of the tight-binding
equations and those for lattice vibrations.] A three-dimensional calculation of the
two energy bands is depicted in Figs. 6.46. Two view points are shown, in order
to emphasize the triangular symmetry and the linear dispersion relations near the
point K in the Brillouin zone.
b. As seen in Fig. 6.46, and similar to the result in Chapter 5, the two bands touch
one another at the six points marked by K in Fig. 5.12. For example, around one
of them, say (4π/(3a), 0), k = (4π/(3a) + k x , k y ) and hence 
√ √
AB (k) = γAB (e−i(4π/3+kx a) + e−i(2π/3+kx a/2+ 3 ky a/2) + e−i(3kx a/2+ 3 ky a/2) .
γ
Expanding in the deviation away√from this point gives
AB (k) = γAB ( 3 − 3i)(k x − ik y )a/4 + O(k2 ) .
γ
The energy
√ is thus linear in the deviation vector, E = ±| γAB (k)| ≈
±(γAB 3/2)|k|a. Each of the signs gives an equation of a cone, symmetric with
respect to rotations around the ẑ−axis, as indeed seen in the figure. Such a linear
dispersion is typical to solutions of the Dirac equation, that describes a particle
and an antiparticle with zero mass. This unique behavior facilitates experimen-
tal explorations of the solutions of the Dirac equation, and entails a very peculiar
electrical behavior of graphene. There is currently much interest in this special
material.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 535

Electrons in solids 535

(a) (b)

Fig. 6.46

Answer 6.25.
a. The matrix in the problem corresponds to an Hamiltonian in which the overlap of
wave functions on different atoms is absent, α(R−R ) = ϕ(0) (r−R)|ϕ(0) (r−R ) =
(0)
δ(R − R ); β, Eq. (6.112), is “absorbed” in the diagonal term, E = E (0) − β.
b. The Schrödinger equation Ĥψ = Eψ takes the form m n|Ĥ|mψ(m) = Eψ(n).
In the present case, it becomes N linear homogeneous equations in N variables,
(0)
E ψ(n) − γψ(n − 1) − γψ(n + 1) = Eψ(n) .

The similarity between this equation and Eq. (5.4) allows one to build on the
arguments used to solve the latter. In particular, using ψ(n) = A exp[ikna], and
eliminating A exp[ikna] on both sides proves that the wavy solution is an accepted
one, provided that
(0)
E=E − γ(eika + e−ika ) .

The periodic boundary conditions, ψ(n) = ψ(n + N ) imply that exp[ikN a] = 1, so

that the discrete wave vectors are k = 2π /(N a), = 0, 1, 2, . . . , N − 1. This result
reproduces Eq. (6.118).
c. In this case, the equations for n = 0 and n = ±1 have to be handled separately,
(0)
E imp ψ(0) − γimp ψ(−1) − γimp ψ(1) = Eψ(0) ,
(0)
E ψ(±1) − γimp ψ(0) − γ ψ(±2) = Eψ(±1) .

All other equations, for |n| > 1, remain as in part (b). Is ψ(n) = A exp[−κ|n|], for
|n| ≥ 1, an accepted solution? The equations of part (a) imply
(0)
E=E − γ(exp[κ] + exp[−κ])] ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 536

536 Introduction to Solid-State Physics

while the equations for n = ±1 give for this solution

(0)
E A exp[−κ] − γimp ψ(0) − γA exp[−2κ] = EA exp[−κ] ,
and the equation for n = 0 yields
(0)
E mp ψ(0) − 2γimp A exp[−κ] = Eψ(0) .
Inserting the expression for the energy into the first equation gives γA = γimp ψ(0).
Substituting the latter and the expression for the energy in the second equation
ensures that this is indeed an accepted solution, provided that
(0) (0)
E imp − E + γ(e−κ + eκ ) = 2γimp
2
e−κ /γ ,
which is a quadratic equation in exp[κ],
(0) (0)
exp[2κ] + [(E imp − E )/γ] exp[κ] + 1 − 2(γimp /γ)2 = 0 .
The solutions are
 (0) (0)
exp[κ] = b ± b2 + 2c − 1 , b = (E imp − E )/(2γ) , c = (γimp /γ)2 > 0 .
The solutions are complex when b2 + 2c − 1 < 0, but then | exp[κ]|2 = b2 + (1 −
2c − b2 ) = 1 − 2c < 1; such solutions are not physical. For b2 + 2c − 1 > 0 the two
solutions are real. They cross the values exp[κ] = ±1 when c = 1 ± b (substitute
exp[κ] = ±1 in the quadratic equation). The condition | exp[κ]| > 1 is fulfilled when
there is a single solution, c = 1 − |b|, and for the two solutions when c > 1 + |b|.

Answer 6.26.
a. As shown in√ the solution of problem 6.10, the Fermi wave vector in two dimen-
sions is kF = 2πn, where n is the planar electrons’ density. The second Brillouin
zone is first filled when the Fermi circle touches the face of the first zone [see Fig.
6.25(a)]. Therefore, it is only the first zone which is occupied as long as kF < π/a,
i.e., n < π/(2a2 ). As the density can be written in the form n = /a2 , one finds
< π/2 ≈ 1.57. The third Brillouin zone is first occupied [see Fig. 6.25(a)]
√ when
the Fermi circle crosses the corner of the first zone, i.e., when kF > 2π/a, or
n > π/a2 , which implies > π ≈ 3.14.
b. As found in the solution of problem 6.19(a) the energy gap in-between the two
bands on the face of the Brillouin zone is 2Δ = 2|U  (G)|. Ignoring the correction
found in problem 6.19(d), the energy at the top of the lower band (at point K, the
corner of the Brillouin zone) is about 2 π 2 /(ma2 ) − Δ; the energy at the center of
the face (point M) is approximately 2 π 2 /(2ma2 ) − Δ. The bottom of the second
band is also at M, with the energy 2 π 2 /(2ma2 ) + Δ. The difference between the
bottom of the upper band and the top of the lower one is
[2 π 2 /(ma2 ) − Δ] − [2 π 2 /(2ma2 ) + Δ] = 2 π 2 /(2ma2 ) − 2Δ .
The transition from the semimetal behavior, which is characterized by the configu-
ration of two overlapping bands, to an insulating one (with a finite gap in-between
the bands) occurs for 2 π 2 /(2ma2 ) = 2Δ.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 537

Electrons in solids 537

c. The band energy, in the tight-binding approximation, Eq. (6.119), is

En (k) ≈ E n + 4Γn (a)[sin2 (kx a/2) + sin2 (ky a/2)] .
The form of the bands is dictated by the sign of Γn . When the signs of both Γ1 and
Γ2 are identical (positive), the two bands reach their maximal and minimal points
at the same locations. The lower part of the upper band is then also at the center
of the Brillouin zone (as opposed to the situation for a free particle!). The material
is thus a semimetal as long as that minimal energy, E2 (0, 0) ≈ E 2 , is lower than
the maximal energy of the lower band,
E1 (π/a, π/a) ≈ E 1 + 8Γ1 (a) .
When the signs of Γ1 and Γ2 are opposite to one another, say Γ2 < 0, the minimum
of the second band is above the maximum of the first one, and the configuration
is similar to that found for a one-dimensional solid: the material is an insulator
unless the bands overlap near the corners of the Brillouin zone. The condition for
this configuration to occur is
E 1 + 8Γ1 ≈ E1 (π/a, π/a) < E2 (π/a, π/a) ≈ E 2 − 8|Γ2 | .

Answer 6.27.
a. The Fermi sphere touches the face of the Brillouin zone when kF = π/a. At three
dimensions and for free electrons, n = kF3 /(3π 2 ) [Eq. (6.28)]. The electrons’ density
is then n = π/(3a3 ). Denoting the concentration of the double-valence atoms by x,
the average number of electrons in each unit cell is = 2x + 1(1 − x) = 1 + x, and
the spatial electrons’ density is n = (1 + x)/a3 . It follows that (1 + x)/a3 = π/(3a3 ),
and x = π/3 − 1 ≈ 4.72%.
b. As seen in Fig. 6.28, the curve formed by the cut of the Fermi surface of
an SC lattice with the face of the Brillouin zone is roughly circular. The face is
characterized by, e.g., kx = π/a, and hence the energy is
En (k) ≈ E n + 2Γn (a)[4 − cos(ky a) − cos(kz a)] .
In the middle of the band, from Eq. (6.123), 0 = cos(kx a) + cos(ky a) + cos(kz a),
which becomes cos(ky a) + cos(kz a) = 1 on the face, and this indeed appears like a
circle. The radius is gradually diminished as cos(ky a) + cos(kz a) increases, and is
reduced to a single point precisely at the center of the face when this sum becomes 2.
For larger values of the sum there are no points on the face that obey this condition,
and the Fermi surface is contained within the first Brillouin zone. Indeed, inserting
2 into Eq. (6.123) results in En (k) ≈ E n + 4Γn (a), i.e., an energy located at the
third of the band width.
c. In a BCC lattice there are two atoms in the unit cell, and therefore the electrons’
density is n = 2(1 + x)/a3 . The reciprocal lattice is an FCC, whose lattice constant
is 4π/a, and whose lattice vectors are given in Eq. (3.21),
G = (2π/a)[h(ŷ + ẑ) + k(x̂ + ẑ) + (x̂ + ŷ)] .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 538

538 Introduction to Solid-State Physics

The center of the zone is connected with its 12 nearest neighbors by the reciprocal-
lattice vectors
G = (2π/a)(±ŷ ± ẑ) , G = (2π/a)(±ẑ ± x̂) , or G = (2π/a)(±x̂ ± ŷ) .
The surfaces separating the first Brillouin zone from its neighbors cut perpendic-
ularly these vectors
√ at their mid points; the distance from the
√ origin to each of
these√planes is 2π/a. It follows that n = kF /(3π ) = ( 2π/a) /(3π 2 ), and
3 2 3

x = 2π/3 − 1 ≈ 48%.

Answer 6.28.
For convenience, denote kx a = ξ and ky a = π − η, and expand in the variables ξ
and η,
En (k) ≈ En(0) − βn − 2γn [cos ξ + cos(π − η)] ≈ E0 + γn (ξ 2 − η 2 ) ,
E0 = En(0) − βn .

the equal-energy curve En (k) ≡ E, dEn =

The point (0, π/a) is a saddle point. On
2γn (ξdξ − ηdη) = 0, and |∇En | ≈ 2γn a ξ 2 + η 2 . The length segment on this curve
obeys
(d )2 = (dkx )2 + (dky )2 = a2 [(dξ)2 + (dη)2 ] = a2 [1 + (ξ/η)2 ](dξ)2 ,
and therefore the integrand is
d dξ dξ
 = =  .
2γn a ξ 2 + η 2 2γn η 2γn ξ 2 − (E − E0 )/γn

Since ξ 2 + η 2 > 0, it follows that ξ > ξ0 = (E − E0 )/γ. As the integral is
dominated by the surrounding of ξ ≈ ξ0 , the other bound of the integration is
of no importance, and the integral can be evaluated in the range ξ0 < ξ < π/2.
An equivalent contribution comes from the two sides of the four points (0, ±π/a),
(±π/a, 0), leading to
 π/2
dξ 4 E − E0
g(E) ≈ 8  ≈ − ln ,
ξ0 2γn ξ 2 − (E − E0 )/γn γn γn π
√ √
where the relation dx/ x2 − a = ln[x + x2 − a] is used; it is also assumed that
E is very close to the middle of the band, E − E0  γn .

Answer 6.29.
a. Denote the lattice constants of the orthorhombic lattice by ax , ay , and az , and
the overlap integrals between nearest neighbors along the corresponding directions
by Γx , Γy , and Γz . Then
En (k) ≈ E n + 4Γx sin2 (kx ax /2) + 4Γy sin2 (ky ay /2) + 4Γz sin2 (kz az /2) .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 539

Electrons in solids 539

A Taylor expansion around the origin yields

En (k) ≈ E n + Γx (kx ax )2 + Γy (ky ay )2 + Γz (kz az )2 .
Introducing the definition 2 /(2m∗i ) = Γi a2i results in the expression given in the
problem.
b. Equation (6.127) gives g(E) = [1/(4π 3 )] d3 k n δ[E − En (k)]. Changing
√
variables, kα / m∗α = k α , yields
 ∗ ∗ ∗
mx my mz 2 2 2
g(E) = dk x dk y dk z δ[E − E0 − (k x + k y + k z )/2] .
4(π)3
Transforming back to spherical coordinates,
 ∗ ∗ ∗   ∗ ∗ ∗
mx my mz 2 2 2mx my mz 
g(E) = 4π dk dkδ[E − E − k /2] = E − E0 .
4(π) 3 0
π 2 3
This result resembles the one given in Eq. (6.31), where the mass there is replaced
here by the geometrical mean of the three effective masses that appear in the ex-
pression for the energy.
c. The density of the electrons in the conduction band is
 EF  ∗ ∗ ∗
Ne 2mx my mz 2
n= = dEg(E) = 2 3
(EF − E0 )3/2 ,
V E0 π 3
and thus, generalizing Eq. (6.32),
3n(E − E0 )1/2
g(E) = .
2(EF − E0 )3/2
As mentioned, this result pertains to holes in the upper part of the valence band.
 √
d. Similar manipulations yield g(E) = 2m∗x m∗y m∗z Emax − E/(π 2 3 ), and hence
 EF  ∗ ∗ ∗
Ne 2mx my mz 2
n= = dEg(E) = (E − EF )3/2 .
V E0 π 2 3 3 max

Answer 6.30.
a. The required equation is
xD
n0e e−β(Ec −μ) = n0h eβ(Ev −μ) + −β(E
.
1+ exp c −ED −μ)

Denoting x = exp[−β(Ec − μ)] yields

n xD
n0e x = 0h e−βEg + .
x 1 + x exp[βED ]
The assumption kB T  Eg validates the neglect of the first term on the right-hand
side. This results in a quadratic equation, whose only positive solution is

x = {−1 + 1 + 4eβED xD /n0e }/(2 exp[βED ]) .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 540

540 Introduction to Solid-State Physics


For kB T  ED it becomes x ≈ exp[−βED /2] xD /n0e , and thus

μ ≈ Ec − ED /2 + kB T ln[xD /n0e ]/2 ,

in accordance with the solution in the text. On the other hand, for
ED / ln[n0e /xD ]  kB T  Eg , one finds x ≈ xD /n0e , and then μ ≈ Ec +
kB T ln[xD /n0e ]. In this limit ne ≈ xD , i.e., all electrons are transferred from the
donors to the conduction band. At sufficiently low temperatures xD > n0e , since
n0e ∝ T 3/2 . Thus, as the temperature rises, the chemical potential rises as well,
from μ(T = 0) = EF = Ec − ED /2 towards Ec . As long as xD > n0e , the chemical
potential may even surpass Ec . At higher temperatures xD < n0e , and the chemical
potential decreases as the temperature is increased. In any case, since ED  Eg ,
the chemical potential is always around the threshold of the conduction band, and
the neglect of the holes’ concentration (the first term on the right hand-side of the
original equation) is justified.
b. For xD = 0, Eq. (6.136) yields

n0e exp[−βEg ]/x + xA /(1 + x exp[βEA ]) = n0h x , x = exp[β(Ev − μ)] .

Discarding the first term on the left produces a quadratic equation with the solution

x = {−1 + 1 + 4 exp[βEA ]xA /n0h }/(2 exp[βEA ]) .

It follows that

μ ≈ Ev + EA /2 − kB T ln[xA /n0h ]/2 , for kB T  EA

μ ≈ Ev − kB T ln[xA /n0h ], for EA / ln[n0h /xA ]  kB T  Eg .

The chemical potential is always around the top of the valence band.

Answer 6.31.
a. The chemical potential is found from Eq. (6.130),

Ec − μ = kB T ln[n0e /ne ] = 0.026 (eV) ln[2.8 × 1019 /1017 ] = 0.146 eV .

As by Eq. (6.133)

n2i = ne nh = n0e n0h e−βEg

= 2.8 × 1019 × 1.04 × 1019 e−1.14/0.026 cm−6 = 2.6 × 1019 cm−6 ,

one finds nh = n2i /ne ≈ 260 cm−3 . The same result is obtained from a direct
calculation, nh = n0h exp[−β(μ − Ev )].
b. In this case

μ − Ev = kB T ln[n0h /nh ] = 0.026 (eV) ln[1.04 × 1019 /1014 ] = 0.31 eV ,

and therefore ne = n2i /nh ≈ 2.6 × 105 cm−3 .

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 541

Electrons in solids 541

Answer 6.32.
a. As the motion is solely along the b−axis, it suffices to consider the vectors’
components along this direction. From Eq. (6.148), vb = ∂En /(∂kb ) and hence
 t
rb (t) = rb (0) + (1/) [∂En (k)/∂kb ]dt .
0
According to Eq. (6.151) one may write dt = −dkb /(eE), and hence
 kb (t)
1 ∂En (k) E [k (t)] − En [kb (0)]
rb (t) = rb (0) − dkb = rb (0) − n b .
eE kb (0) ∂kb eE
(The components of k other than kb remain unchanged.) The energy is periodic
in the reciprocal-lattice vectors, and thus returns to the same value once kb (t) =
kb (0) + b. The period is found by integrating,
  T   kb (T) b
 
T= dt = dkb /(eE) = .
0 kb (0) eE
b. In a cubic lattice b = 2π/a, and thus T = h/(eEa). Using E = 104 V/cm and
a = 1Å gives T = 10−10 sec. This implies that T  τ ∼ 10−14 sec, and the electron
is scattered before it reaches the boundary of the Brillouin zone. In order to detect
Bloch oscillations, one has to increase considerably the lattice constant , e.g., by
utilizing superlattices, increase the magnitude of the electric field, and extend the
relaxation time, e.g., by using high quality crystals at very low temperatures.
c. The lattice momentum of each electron, in the presence of a constant electric
field, is k(t) = k(0) − eEt; the induced dipole moment is
kF  kF
P = −e x[k(0), t] = −e(L/π) dk(0)x[k(0), t] ,
k(0)=−kF −kF

where x[k(0), t] = −(E[k(t)] − E[k(0)])/(eE) is the location at time t of an electron

with an initial momentum k(0), and the sum runs over all electrons in the conduction
band (two electrons in each state). It follows that
ΓL
P = [sin(kF a − eEat/) + sin(kF a + eEat/) − 2 sin(kF a)]
πaE
4ΓL
=− sin(kF a) sin2 [eEat/(2)] .
πaE
A constant electric field induces an oscillating dipole moment! this moment vanishes
when the band is fully occupied, i.e., kF a = π.

Answer 6.33.
One begins from Eq. (6.146), treating (2 /m)(−i∇ + k) · δ as a perturbation.
Then Eq. (C.3) yields a generalization of Eq. (6.147): up to second order in the
perturbation,
2
En (k + δ) = En (k) + δ · unk | − i∇ + k|unk 
m
 2
 δ
2 2 |unk | 2m (−i∇ + k) · δ|un k |2
+ + . (6.179)
2m 
En (k) − En (k)
n=n
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 542

542 Introduction to Solid-State Physics

Next, one notes that a continuation of the expansion of Eq. (6.155) gives
1  ∂ 2 En (k)
En (k + δ) = En (k) + δ · ∇k En (k) + δ δ + ... . (6.180)
2 ∂kα ∂kβ α β
α,β

Comparing Eqs. (6.179) and (6.180) and using Eq. (6.155) yields
 1  1 ∂ 2 En (k)
≡
m∗ αβ 2 ∂kα ∂kβ
1  unk | − i∇α |un k un k | − i∇β |unk  + {α ⇔ β}
= δαβ + ,
m 
En (k) − En (k)
n=n

where the orthogonality of the wave functions is exploited in the constant terms
with δ · k, and where {α ⇔ β} indicates the previous term in the numerator, with
interchanged cartesian components.

Answer 6.34.
The solution of problem 6.19 gives that for the wave vector q = q0 + δ, where q0 is
on the boundary between two Brillouin zones [i.e., 2q0 · G = G2 ], the energies are
2  
(1) (1)
Eq± ≈ Eq0 ± (q0 ) ± |U (G)| +  δ 2 + δ · (2q0 − G) ± 2 (δ · G)2 /[4m|U  (G)|] .
2m
This expression is quadratic in the momentum only at the points q0 = G/2. Dif-
ferentiating it there, one obtains
 1  1 ∂ 2 E± (q) 1 
≡ = δ ±  2
G G (1 + δ )/(8m|  (G)|) ,
U
m∗ αβ 2 ∂qα ∂qβ m αβ α β αβ

where the two signs refer to the two bands.

Answer 6.35.
Using dt = dkx /k̇x and
eB ∂E
k̇x = ,
2 c ∂ky
yields
   
T
2 c dkx 2 c ∂ky 2 c d
T= dt = = dkx = ky dkx ,
0 eB |∂E/∂ky | eB ∂E eB dE
and the integral is over the area encircled by the orbit.

Answer 6.36.
The energy is
kx2 ky2 k2
E(k) = E0 + 2 ∗
+ ∗
+ z∗ . (6.181)
2mx 2my 2mz
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 543

Electrons in solids 543

The momentum component parallel to the field does not vary with time, k · n̂ = k0 .
Equation (6.181) describes a cone whose apex is at the origin; the area of its base is
√
A and its volume is Ω = Ak0 /3. Transforming to the coordinates kα / m∗α = k α ,
the equal-energy surface becomes the paraboloid
1 2 2 2
E(k) = E0 + [k x + k y + k z ] ,
2
and the cone equation is

k · n = k0 , nα = nα m∗α / .

In these coordinates Ω = Ωc m∗x m∗y m∗z /3 , where Ωc is the volume of the cone in
the spherical coordinates.
 The base of the cone is the cut of a plane with a sphere
of radius R = 2(E − E0 ), and its height is

h = k0 /|n| = k0 / m∗x n̂2x + m∗y n̂2y + m∗z n̂2z .
√
It follows that the radius of the cone’s base is R2 − h2 , and its volume is Ωc =
π(R2 − h2 )h/3. Thus,

A = 3Ω/k0 = π m∗x m∗y m∗z [2(E − E0 ) − h2 ](h/k0 )/3 ,
and hence
 ∗ ∗ ∗
∂A mx my mz
= 2π  .
∂E 2 m∗ n̂2 + m∗ n̂2 + m∗ n̂2
x x y y z z

Equation (6.163) yields the period, as given in Eq. (6.164). For instance, for a

magnetic field along the ẑ−axis it yields ωc = eB/[c m∗x m∗y ]. For an isotropic
effective mass one obtains for the cyclotron frequency an expression similar to that
of a free particle; in the general case the cyclotron frequency depends on the weighed
effective mass.
The result above may be obtained by an alternative method. For a field along
the ẑ−axis, the equations of motion are
k̇x = (eB/[m∗y c])ky , k̇y = −eB/([m∗x c])kx .
It follows that k̈x = −[(eB)2 /(m∗x m∗y c2 )]kx ; this equation describes a periodic mo-

tion with the cyclotron frequency ωc = eB/[c m∗x m∗y ].

Answer 6.37.
The vector product of Eq. (6.162), k̇ = −(e/c)ṙ × B, with B yields
e e
B × k̇ = − B × [ṙ × B] = − [ṙB2 − B(B · ṙ)] .
c c
Hence ṙ = −[c/(eB)]B̂ × k̇ + vB B̂, and upon carrying out the temporal integration
one obtains the required result. It shows that the vector
c
r⊥ (t) − r⊥ (0) = − B̂ × [k(t) − k(0)]
eB
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 544

544 Introduction to Solid-State Physics

lies in a plane perpendicular to B. It is also normal to the vector [k(t) − k(0)], and
its length is
|r⊥ (t) − r⊥ (0)| = B |k(t)
2
− k(0)| ,
Thus, the orbit [r⊥ (t) − r⊥ (0)] can be derived from the orbit [k(t) − k(0)] by a 90◦ −
rotation, and a multiplication by 2B . This conclusion coincides with the result of
problem 6.3 (check!).

Answer 6.38.
a. One begins by extending Eq. (6.8) to account for particles of charge qj (i.e.,
qe = −e for electrons and qh = e for holes), a positive effective mass m∗j , a relaxation
time τj , a density nj , and an average velocity vj :
m∗j (v̇j + vj /τj ) = qj (E + vj × B/c) .
The magnetic field is along the ẑ−direction, while the electric field and the particles
are assumed to be in the XY plane. In a stationary state
vjx  B vjy  B
m∗j = qj Ex + vjy , m∗j = qj Ey − vjx .
τj c τj c
Solving the two equations (discarding quadratic terms in the magnetic field) yields
q j τj  q j τj B  q j τj  q j τj B 
vjy ≈ ∗ Ey − ∗ Ex , vjx ≈ ∗ Ex + E .
mj mj c mj m∗j c y
The current density is given by

j= nj q j v j .
j

The component jy vanishes in the geometry of Fig. 6.1, leading to

 τj   τ 2 BE
j
nj qj2 ∗ Ey − nj qj3 ∗
x
=0.
j
m j j
m j c

Similarly, the current along the x̂−direction is

 τj   τ 2 BE
j y
jx = nj qj2 ∗ Ex + nj qj3 ∗ .
j
m j j
m j c

As Ey ∝ Ex B, the second term in this expression is quadratic in the magnetic field

and thus is ignored; then
Ey j nj qj (qj τj /m∗j )2 nh μ2h − ne μ2e
RH = ≈ = ,
jx B c[ j nj (qj2 τj /m∗j )]2 ce(ne μe + nh μh )2
where μj = eτj /m∗j are the mobilities of the two types of charge carriers. This result
reproduces Eq. (6.11), RH = −1/(ne ec), when the density of the holes vanishes.
In contrast, when the charge carriers are holes, then RH = 1/(nh ec), as can be
expected for positively-charged particles. In an intrinsic semiconductor ne = nh ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 545

Electrons in solids 545

and the sign of the result depends on the sign of μh − μe . In the more general case,
the sign, as well as the numerical factor, can be arbitrary. This removes one of the
caveats of the Drude theory. 
b. Introducing the notations b = μe /μh and x = ne /ni = ne /nh , part (a) yields
x 1 − b2 x 2
RH = .
ceni (1 + bx2 )2
This function is plotted in Fig. 6.47 (in units of 1/(ceni )). As seen, RH = 0 when
ne /ni = x = 1/b = 1/3, and hence ne = ni /3 = 5 × 109 cm−3 . The extremal points
of RH are derived from dRH /dx = 0, i.e.,

3b(1 + b) ± 9b2 (1 + b)2 − 4b3
b x − 3b(1 + b)x + 1 = 0 ⇒ x =
3 4 2 2
.
2b3
The two solutions are ne /ni = x = 1.14, 0.17, and nh /ni = 1/x = 0.88, 2.5, which
correspond to the maximal and minimal RH .

Fig. 6.47

c. When higher orders in the magnetic field are taken into account Eq. (6.13)
pertains separately for each type of particles. The current is the sum over the
currents carried by each type of particles, and therefore Eq. (6.14) is generalized to
be
 σj τj ωcj  σj
σxy = −σyx = 2
, σ xx = σ yy = ,
j
1 + (τj ωcj ) j
1 + (τj ωcj )2

where σj = nj qj2 τj /m∗j and ωcj = qj B/(m∗j c). The resistivity tensor is the inverse
of the conductivity one,

ρyx = −ρxy = σxy /[σxx

2 2
+ σxy 2
] , ρxx = ρyy = σxx /[σxx 2
+ σxy ].

Equation (6.12) is reproduced upon assuming that there is only one type of parti-
cles, and then the diagonal element, ρxx = ρ is independent of the magnetic field.
Otherwise, the diagonal elements do depend on the magnetic field. In particular,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 546

546 Introduction to Solid-State Physics

for very strong fields, with qe = −e for the electrons and qe = e for the holes, one
finds
σxx = σyy = (ne /μe + nh /μh )ec2 /B2 , σxy = −σyx = (ne − nh )ec/B .
In intrinsic semiconductors, for which ne = nh , the diagonal resistivity is quadratic
in the magnetic field, ρxx ∝ B2 , producing very large magneto-resistance. As
opposed, for ne = nh the magnetic field is cancelled out, and ρxx approaches a
finite value.

Answer 6.39.
The cylinder cuts the Fermi sphere at a circle of radius K , whose area is A = πK2 .
Since K2 = 2m∗ E /2 , one finds
2πm∗ E 2πm∗ ωc 1 2πeB 1
A = = ( + )= ( + ).
 2  2 c 2

Answer 6.40.
The inverse of the matrix in Eq. (6.13) is
     
ρxx ρxy 1 σyy −σxy 0 1/σyx
= = .
ρyx ρyy σxx σyy − σyx σxy −σyx σxx 1/σxy 0
The last step stems from the substitution σxx = σyy = 0. Comparing the left
hand-side with the right one yields ρxy = ρyx = 1/σyx and ρxx = ρyy = 0.

6.14 Problems for self-evaluation

s.6.1.
Exploit the Drude model to compute the average energy that the electron loses within
a unit time, and use the result to prove the Joule law for the energy loss per unit
volume per unit time, W = σE 2 .

s.6.2.
Consider a two-dimensional hexagonal lattice, as the one displayed in Fig. 6.48,
with a lattice constant a = 3Å. Each atom in the lattice donates a single electron of
mass m, that is free to move in the plane. The total area of the sample is A. The
measured conductivity of this sample is σ = 10−2 /Ω.
a. How many free electrons are there in the sample?
b. What is the density of states?
c. Determine the Fermi energy.
d. Find the average energy of an electron in the ground state of this system.
e. Find the specific heat at low temperatures and at high temperatures.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 547

Electrons in solids 547

f. Determine the mean free-path of the electron.

g. How many Brillouin zones are occupied by the electrons? Suppose that each
atom releases electrons. At which value will the second Brillouin zone begin to
be occupied?

Fig. 6.48

s.6.3.
A three-dimensional gas of free electrons, contained in a volume V is subjected to
a constant magnetic field, B. The energies of electrons whose moment is parallel
(antiparallel) to the field are E −μB B (E +μB B), where μB is the magnetic moment
resulting from the spin of the electron.
a. Exploit the Sommerfeld theory to write down expressions for the number of
electrons whose magnetic moment is parallel or is antiparallel to the field.
b. Prove that at low temperatures, and for a weak magnetic field, the total magnetic
moment of the system is M = 3N μ2B B/(2EF ). [Hint: use the results of problem 6.9].

s.6.4.
Suppose that the potential steps of the Kronig-Penney model are replaced by potential
wells. This is accomplished by assuming that U0 of Eq. (6.65) is negative.
a. Describe the wave functions for energies in the range U0 < E < 0. What is the
relation between the energy and the wave number in this case?
b. Find the width of the lowest band in the delta-function limit, U0 → −∞, (a−b) →
0, U0 (a−b) → u0 < 0, and compare it with the calculation in the tight-binding limit,
problem 6.22.

s.6.5.
A material is described by a one-dimensional lattice with three atoms in each prim-
itive unit cell, whose length is a. The potential that the atoms exert on the electrons
is

U (x) = U0 a [2δ(x − na) − δ(x − (n + 1/4)a) − δ(x − (n + 3/4)a)] .
n

It is assumed that U0  2 π 2 /(2ma2 ), where m is the free electron mass.

a. Find the value of the energy gap at the end of the first Brillouin zone, close to
the energy 2 π 2 /(2ma2 ).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 548

548 Introduction to Solid-State Physics

b. Determine the effective mass of the electrons at the bottom of the band located
slightly above 2 π 2 /(2ma2 ). What is the answer for the holes at the top of the band
located slightly below this energy?
c. Find the value of the energy gap at the center of the first Brillouin zone, at the
vicinity of 22 π 2 /(ma2 ).
d. Each atom donates electrons to the system. For which values of is the
material an insulator? and for which ones is it conducting?

s.6.6.
A free electron of mass m is moving in a square lattice of lattice constant a.
a. Calculate the energy of the electron at the corners and at the faces’ centers of
the first Brillouin zone.
b. Assuming that each atom donates electrons, calculate the radius of the Fermi
circle. How many Brillouin zones are occupied when = 1? when = 2? in these
two cases, are the charge carriers electrons or holes?
c. The electron is subjected to a weak two-dimensional potential of the form
2πx 2πy
U (x, y) = −4U0 cos cos ,
a a
as shown in Fig. 6.49. This potential can be artificially created in the lab, by
combining oscillating electric fields produced by lasers. The experimentalists can
capture atoms in such a potential, and thus explore the behavior of electrons in a
periodic potential. Estimate the energy gaps created at the corners and at the mid
points of the edges of the first Brillouin zone.

Fig. 6.49

s.6.7.
Consider an electron gas in a single-atom lattice of a BCC structure.
a. Exploit the tight-binding approximation to obtain the energies in a single band,
as functions of the lattice momentum. The overlap of the wave functions between
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 549

Electrons in solids 549

nearest neighbors (nn) is parametrized by the energy γ1 , and between next-nearest

neighbors (nnn) by the energy γ2 .
b. Find the acceleration along the ŷ−direction caused by an external electric field
along the x̂−direction.
c. Calculate the effective mass tensor at the bottom of this band.
d. Assuming that the band is the conduction band, and that all bands below are
completely full, derive the electric conductivity tensor in the Drude approximation.
The (low) electrons’ density in the band is n and their relaxation time is τ .

s.6.8.
The energy of electrons in a three-dimensional crystal, at the bottom of the conduc-
tion band, is
2 2 α 4
E = E0 + k + k ,
2m∗ 4
where the effective mss m∗ and the coefficient α are positive. The material is placed
in an electric field of frequency ω, E = E 0 cos(ωt). At time t = 0 there is an
electron at the bottom of the band, with k = 0. Exploit the Semiclassical picture for
the electron motion to find
a. the location of the electron at time t;
b. the frequencies and the amplitudes of the electron’s motion;
c. the amplitude of the field E 0 required to make the effect of the quartic term in
the energy on the electron motion at frequency ω comparable to that of the quadratic
one.

s.6.9.
The energy gap of crystalline silicon (whose valence is 4) is 1.14 eV, and the effective
masses of the electrons and of the holes are 0.2 m and 0.3 m, respectively (m is the
free-electron mass). What is the concentration xD of arsenic atoms, which are five-
valence donors, that has to be added to the crystal such that at room temperature
the extrinsic conductivity exceeds the intrinsic one by a factor of 104 ? Assume that
the binding energy of the electron, ED , is small, ED  kB T .

s.6.10.
A material is described by a square lattice of lattice constant a. The dispersion
relation in the first energy band is E(k) = −γ cos[(|kx | + |ky |)a/2], with γ > 0. This
system is placed in a magnetic field B = Bẑ. The Fermi energy of the electrons is
EF = aγ.
a. Find the Fermi curve.
b. Find the velocity of an electron at the Fermi energy within the semiclassical
approximation.
c. Find the force exerted on the electron by a magnetic field along an arbitrary
direction.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 550

550 Introduction to Solid-State Physics

d. What is the time needed for the electron to complete a full circle around the
Fermi curve?

s.6.11.
a. Find the effective-mass tensor of an electron in the two energy bands of graphene.
Exploit the tight-binding approximation, confining yourselves to overlap couplings
between nearest neighbors. In particular, determine the effective mass at the center
of the Brillouin zone (the Γ point) and at the center of the hexagonal face (the M
point).
b. What is the density of states in graphene around the Γ and the K points of the
Brillouin zone? (the latter is at the corner of the hexagonal face).

s.6.12.
Figure 6.50 displays the density of states in a material, in which the density of
electrons is 5×1028 m−3 .
a. Determine the Fermi energy. Is this material a metal, an insulator, or a semi-
conductor? Explain your choice.
b. Determine the specific heat of this material at low temperatures.
c. How are the answers modified when the electrons’ density is 7×1028 m−3 ?

Fig. 6.50

s.6.13.
The wave functions of electrons in the d-shell decay strongly with the distance from
their parent atom; consequently, the overlap between wave functions localized on
neighboring atoms is small, and the energy band (in the tight-binding approximation)
is narrow. Assume that the center of the band is on the atomic energy Ed , and that
the density of states can be taken as a rectangular function, of width W .
a. Determine the height of the rectangular function.
b. Find the total energy, at zero temperature, of the electrons in the band, when
each atom contains N d-electrons (ignore the other bands).
c. Assume that the energy found in part (b) is the binding energy of the lattice.
Find the conditions for this energy to be maximal.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 551

Electrons in solids 551

s.6.14.
The model of Su, Schrieffer, and Heeger is a generalization of problem 6.23: elec-
trons in a one-dimensional lattice. The atoms on the lattice sites are identical and
it is assumed that the energies on the atoms vanish. The coupling energies of the
tight-binding model between nearest neighbors alternate,

γ2n−1,2n = γ2n,2n−1 = γ− = γ0 (1 − Δ) ,
γ2n+1,2n = γ2n,2n+1 = γ+ = γ0 (1 + Δ) .

Such a model can describe, e.g., the lattice of polyyne, Fig. 1.7(b), which comprises
two types of bonds. Su, Schrieffer, and Heeger designed their model to describe
polyethylene, [C2 H2 ]n , where each carbon on the chain is coupled also to a hydrogen,
and the bonds between the carbons alternate between single and double bonds; see
the discussion concerning polyyne, Sec. 4.3. The shift Δ can result also from elastic
longitudinal deformation of the lattice, of the type um = u0 exp[iπm]. When u0 is
positive, the atoms at m = 2n move slightly to the right, and those with m = 2n − 1
move to the left, thus forming a static phonon at the edge of the Brillouin zone.
The basis for this model is the dependence of γ on the interatomic distance. The
pairs of atoms that move closer together are called dimers.
a. Ignore the overlap of the atomic wave functions, and show that there is an energy
gap when Δ is finite.
b. Assume that the gain in elastic energy due to the deformation is ΔEd = N xΔ2 ,
where the lattice contains 2N atoms. Show that when each atom donates a single
electron the material is an insulator, and the energy is minimal for both possibilities
Δ = ±Δ0 . It follows that the state with Δ = 0, where the material is a metal, is
unstable: the material is “dimerized” into the insulating phase. This instability is
named after Peierls, and is known as the Peierls instability.
c. What are the two states of the infinite system when |Δ| = 1? What happens in
a finite system comprising 2N atoms?

6.15 Answers for the self-evaluation problems

Answer s.6.1.
In-between collisions, the equation of motion of the electron is p(t) = −eEt, and
consequently its energy at time t is (eEt)2 /(2m). The probability to reach the
time t without undergoing an additional collision is exp[−t/τ ], and therefore the
probability to collide and lose all the energy between t and t + dt is exp[−t/τ ]dt/τ .
Hence, the average energy the electron loses at each collision is
 ∞
[(eEt)2 /(2m)]e−t/τ dt/τ = (eEτ )2 /m .
0
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 552

552 Introduction to Solid-State Physics

The energy loss per unit time, per unit volume, is found by multiplying this result
by the electron density n and dividing it by the mean free-time τ ; then W =
(ne2 τ /m)E 2 = σE 2 .

Answer s.6.2. √ √
a. The area of each hexagon is S = 6 × ( 3a2 /4) = 3 3a2 /2. Each hexagon
contributes two electrons (as each atom around it belongs to three hexagons, there
are two atoms per each hexagon). Hence, the planar density of the electrons is
√
n = 2/S = 4/(3 3a2 ) .
The total number of electrons in the sample is
√
nA = 2A/S = 4A/(3 3a2 ) .
Another option to obtain this result
√ is to note that the √
unit cell of the hexagonal
lattice is a rhombus whose area is a2 3/2, where a = a 3 is the lattice constant
of the triangular lattice. This unit cell contains two atoms, and hence
√ √
n = 2/(a2 3/2) = 4/(3 3a2 ) .
b. Consider a circle of radius k. Its area is πk 2 , and therefore the number of states
within it (including the spin degree of freedom) is 2 × (πk 2 )/[(2π)2 /A] = Ak 2 /(2π).
Using E = 2 k 2 /(2m) results in AmE/(π2 ) for the number of states. It follows
that the number of states per unit area between the energies E and E + dE is
g(E)dE = mdE/(π2 ) .
c. The Fermi energy is given by the relation
 EF
n= g(E)dE ,
0
and therefore
n = mEF /(π2 ) , ⇒ EF = nπ2 /m .
Inserting this into the result of part (b) gives g(E) = n/EF , in accordance with the
solution of problem 6.10.
d. The average energy per particle is
 EF
E = A Eg(E)dE/N = (1/n)(n/EF )EF2 /2 = EF /2 ,
0
again in accordance with the solution of problem 6.10.
e. According to the solution of problem 6.9, the specific heat at low temperatures
is linear in the temperature (in any dimension). At high temperatures it obeys the
Dulong-Petit law, CA = 2N kB .
f. From Eq. (6.6), τ = mσ/(ne2 ). The mean free-path is
= vF τ = (kF /m)τ = kF σ/(ne2 ) .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 553

Electrons in solids 553

√
Using kF = 2mEF and EF = nπ2 /m yields

= 2π2 /n(σ/e2 ) .
√
As n = 2/S = 4/(3 3a2 ), one finds
 √
= 3π 3/2(aσ/e2 ) ,

and with the numerical value of the conductivity, this becomes ≈ 354Å.
√ √
g. Exploiting the expressions above yields kF = 2πn = 8π/(3 3)/a. The
closest point to the center of the Brillouin zone is located at the mid point of the
reciprocal-lattice basis vector.√ The length of the reciprocal-lattice vector in the
triangular lattice is b = 4π/(a 3) (see, e.g., Fig. 3.11). Hence, all occupied states
are within the first Brillouin
 zone provided that kF < b/2, which is indeed the case
√
here, since 2kF /b = 2/(π 3) < 1. When each atom contributes electrons, then
 √
2kF /b = 2 /(π 3); the Fermi circle cuts the boundary of the Brillouin zone for
√
= π 3/2 ≈ 2.72.

Answer s.6.3.
a. The energies of the electrons are shifted by the magnetic field, and thus the
Fermi-Dirac distribution [Eq. (6.35)] for the two types of electrons, are f (E ∓μB B),
where E is the kinetic energy, Eq. (6.27). As the total number of electrons is N ,
the density of states of each type is
√ 3/2
g(E)/2 = 3n E/(4EF ) ,

[Eq. (6.32), where n = N/V ]. The average number of electron of each type is
 ∞
N± = (V /2) dEg(E)f (E ∓ μB B) .
0

b. When the magnetic field is weak, one may expand the distributions, f (E ∓
μB B) ≈ f (E) ∓ f  (E)μB B. At low temperatures f  (E) ≈ −δ(E − EF ) [problem
6.9]. Hence,
 ∞
V N V
N± = dEg(E)[f (E) ∓ f  (E)μB B] ≈ ± g(EF )μB B .
2 0 2 2
The total moment is
3N 2
M = (N+ − N− )μB ≈ V g(EF )μ2B B = μ B,
2EF B
where in the last step use has been made of Eq. (6.32). To leading order, the mag-
netic moment is linear in the field; this phenomenon is called the paramagnetism
of the electron gas.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 554

554 Introduction to Solid-State Physics

Answer s.6.4.
a. Using Eq. (6.67), one finds that in the present case Q2 > 0, but
K 2 = 2mE/2 = −κ2 < 0 .
It follows that the wave functions oscillate within the potential wells, but decay
(or increase) exponentially, A exp[−κx] + B exp[κx] in-between the wells. Equation
(6.73) still pertains, but one has to insert there K = iκ; consequently
−κ2 + Q2
cos(ka) = cos[Q(a − b)]cosh(κb) − sin[Q(a − b)]sinh(κb) .
2κQ
b. In the limit specified in the problem,
Q2 mu0 a sinh(κa)
cos(ka) ≈ cosh(κb) − (a − b)sinh(κb) → cosh(κa) + .
2κ 2 κa
In the tight-binding limit one may approximate cosh(κa) ≈ sinh(κa) ≈ exp[κa]/2.
The band boundaries are at the points where cos(ka) = ±1. There
1  mu 
±1 ≈ eκ± a 1 + 2 0 ,
2  κ±
that is,
κ± ≈ κ0 (1 ± 2 exp[−κ± a]) , κ0 = −mu0 /2 .
The band width is the difference between the corresponding energies, i.e.,
2 2  2 κ0  2 κ0 42 κ20 −κ0 a
ΔE = (κ+ − κ2− ) ≈ (κ+ − κ− ) ≈ 4κ0 e−κ0 a = e .
2m m m m
Inserting here κ0 = m|u0 |/2 shows that this is precisely the gap obtained in the
solution of problem 6.22.

Answer s.6.5.
The Fourier transform of the potential is

 1 a
U (G) = dxU (x)e−iGx
a −a
 a  
= U0 dx [2δ(x − na) − δ[x − (n + 1/4)a] − δ[x − (n + 3/4)a]e−iGx .
−a n
The first two terms in the integrand contribute solely for n = 0, while the third one
necessitates n = −1. It follows that
 (G) = U0 [2 − e−iGa/4 − eiGa/4 ] = 2U0 [1 − cos(Ga/4)] .
U
a. The required gap separates the first two bands, and is due to the degeneracy
 (G1 )| = 4U0
E (0) (G1 /2) = E (0) (−G1 /2), where G1 = 2π/a. The energy gap is 2|U
[see Eq. (6.90)].
b. The energies near the boundary of the first Brillouin zone, by Eq. (6.92), are
2 
(1) 2 G21  (G1 )| +  1 ± 2 G21  2
E± (k) ≈ ± |U δ + ....
8m 2m  (G )|
2m|U 1
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 555

Electrons in solids 555

The coefficient of (δ)2 /2 in the last term defines the inverse effective mass, 1/m∗ ,
 2 G21 
m∗± = m/ 1 ± .
8mU0
The upper sign corresponds to electrons in the conduction band, and the lower one
corresponds to holes in the valence band. The effective mass in the valence band is
negative because of the assumption that U0 < 2 G21 /(4m).
c. The required gap is in-between the second and the third bands, owing to the
 (G2 | =
degeneracy E (0) (G2 /2) = E (0) (−G2 /2), with G2 = 4π/a. Its value is 2|U
8U0 .
d. The electrons’ density is n = Ne /L = 3 /a. The total number of electrons is
related to the Fermi wave vector,
Ne = 2(2kF )/[2π/L] = 2kF L/π , ⇒ kF = πn/2 = 3π /(2a)
When = 1 then kF = 3π/(2a). This value is at the middle of the second Brillouin
zone, between G1 /2 and G2 /2. Therefore, this band is half-full, and the material
is metallic. For = 2 kF = 3π/a, exactly on the border between the third and the
fourth Brillouin zones. Hence the material is an insulator. Generally speaking, the
material is metallic for odd ’s and is an insulator for even ones.

Answer s.6.6.
a. The mid point of the edge of the first Brillouin zone is, e.g., at (π/a, 0), and the
corner is at (π/a, π/a). The corresponding energies are E(π/a, 0) = (π/a)2 /(2m)
and E(π/a, π/a) = (π/a)2 /m.
b. The Fermi wave vector in two dimensions is
√ √
kF = 2πn = 2π /a .

When = 1, then kF /(π/a) = 2/π < 1, and therefore all states within the Fermi
sphere are contained in the √first Brillouin zone. On the other hand, when = 2 then
√
1 < kF /(π/a) = 2/ π < 2, and the circle penetrates into the second Brillouin
zone, but does not reach the third one. Figure 6.51 displays the two Brillouin zones
and the two Fermi circles. As seen, in the first case the Fermi circle lies away from
the edges of the zone, and therefore the charge carriers are nearly-free electrons.
In the second case there are empty places in the first Brillouin zone only near its
corners, where electrons in the first band have maximal energy. Thus, the charge
carriers in this case are holes. The electrons that “overflow” into the second zone
are near the mid points of the edges of the zone, that is, they are near the minimal
energy of the second band. It follows that they behave like electrons, but with a
different effective mass.
c. The minima of the potential are located on a square lattice of lattice constant
a. By writing
U (x, y) = −U0 e2πix/a + e−2πix/a e2πiy/a + e−2πiy/a ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 556

556 Introduction to Solid-State Physics

it is seen that U (G) = −U0 only for the four reciprocal-lattice vectors G =
(±2π/a, ±2π/a). The only points in the first Brillouin zone that can be connected by
one of these vectors are located at the opposite corners of the zone, e.g., (π/a, π/a)
and (−π/a, −π/a). The degeneracy of pairs of such corners gives rise to an energy
gap in-between the lowest two bands. Extending Eq. (6.91) that has been discussed
in problem 6.19, leads to
2
(1)  (G)| = (π) ± |U0 | .
Eq0 ± = E (0) (π/a, π/a) ± |U
ma 2
On the other hand, the energy at points along the edges of the Brillouin zone is not
degenerate with any of the points displaced by one of the vectors G(±2π/a, ±2π/a),
and therefore the energy gaps created there are small, of the order of |U0 |2 /[E(q) −
E(q + G)].

Fig. 6.51

Answer s.6.7.
a. Using a cubic unit cell comprising two atoms, there are eight nearest neigh-
bors, at (±a, ±a, ±a)/2, and six next nearest-neighbors, at (0, 0, ±a), (0, ±a, 0),
and (±a, 0, 0). The band energy is hence
E(k) = E − β − 2γ2 [cos(kx a) + cos(ky a) + cos(kz a)] − 2γ1 {cos[(kx + ky + kz )a/2]
+ cos[(kx + ky − kz )a/2] + cos[(kx − ky + kz )a/2] + cos[(−kx + ky + kz )a/2]}
= E0 − β − 2γ2 [cos(kx a) + cos(ky a) + cos(kz a)]
− 8γ1 cos(kx a/2) cos(ky a/2) cos(kz a/2) .
b. Equations (6.154) and (6.155) yield
ay  = −(1/m∗ )yx eEx ,
(1/m∗ )yx = 2γ1 a2 sin(kx a/2) sin(ky a/2) cos(kz a/2)/2 .
c. At the bottom of the band
0 + (γ1 + γ2 )a2 k2 .
E(k) ≈ E0 − β − γ2 (6 − a2 k2 ) − γ1 (8 − a2 k2 ) = E
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 557

Electrons in solids 557

Hence,
 1 
= δαβ 2(γ1 + γ2 )a2 /2 .
m∗ αβ
d. The Drude formula gives
 1 
σαβ = ne2 τ = δαβ 2ne2 τ (γ1 + γ2 )a2 /2 .
m∗ αβ

Answer s.6.8.
a. The velocity, in the semiclassical approximation, is
vnk = ∇k En (k)/ = [(/m∗ ) + αk2 ]k .
The equations of motion are
k̇ = −eE 0 cos(ωt) , ⇒ k(t) = −[e/(ω)]E 0 sin(ωt) ,
where it has been assumed that k(t = 0) = 0. Using the expression for the average
velocity yields
e αe3 2
v(t) = − E 0 sin(ωt) − E E 0 sin3 (ωt) .
m∗ ω 3 ω 3 0
By using the identity sin3 x = [3 sin x − sin(3x)]/4, the velocity becomes
e  3αe2 m∗ E02  αe2 m∗ 2
v(t) = − ∗ E 0 1 + sin(ωt) − E sin(3ωt) .
m ω 3
4 ω 2 43 ω 2 0
Finally, integrating over the time gives, up to a constant,
e  3αe2 m∗ E02  αe2 m∗ 2
r(t) = ∗ 2 E 0 1 + cos(ωt) − E cos(3ωt) .
m ω 3
4 ω 2 123 ω 2 0
b. As seen from the expressions above, the oscillations are at the two frequencies
ω and 3ω, with the amplitudes
e  3αe2 m∗ E02  αe3 E02
1 + , − ,
m∗ ω 2 43 ω 2 123 ω 4
respectively.
c. The contribution of the quartic term in the energy to these results is proportional
to α. Thus, both parts of the dispersion relation are of a comparable significance
at frequency ω once 3αe2 m∗ E02 /(43 ω 2 ) = 1, i.e., E02 = 43 ω 2 /(3αe2 m∗ ).

Answer s.6.9.
Assume that the conductivities of the electrons and of the holes are given by the
Drude formula Eq. (6.6), with the appropriate concentrations and masses, and with
the same relaxation time. The concentration of electrons in an extrinsic semicon-
ductor is xD ; hence
σint = (ne /me + nh /mh )e2 τ = ne (1/me + 1/mh )e2 τ ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 558

558 Introduction to Solid-State Physics

and
σext = xD e2 τ /me .
By the conditions of the problem,
104 = σext /σint = (xD /ne )mh /(me + mh ) = 3xD /(5ne ) .
Equation (6.134) implies that
 √
ne = nh = Ce Ch π(kB T )3/2 exp[−βEg /2]/2 ,
where (from the explicit expressions in problem 6.29),
Ce Ch = 2(me mh )3/2 /(π 4 6 ) .
Substituting the value of the energy gap, and in addition kB T ≈ (1/40) eV at
room temperature, m = 0.511 MeV/(3 × 108 m/sec)2 , and  = 6.582 eV sec, gives
ne = 3.62 × 1014 m−3 . Hence, xD = 6 × 1018 m−3 .

Answer s.6.10.
a. The Fermi curve consists of the straight segments given by the equations
(|kx | + |ky |)a/2 = arccos(−α) .
In particular, the equation for the segment in the first quarter of the Brillouin zone
is (kx + ky )a/2 = arccos(−α). It follows that the surface is a square whose corners
are at (±2arccos[−α]/a, 0) and (0, ±2arccos[−α]/a).
b. In the absence of the magnetic field, the velocity in the first quarter is
1 aγ
v= ∇k E = sin[(kx + ky )a/2](1, 1, 0) .
 2
On the Fermi curve
EF = −γ cos[(|kx | + |ky |)a/2] = αγ ,
and thus the velocity of the electron is constant along the line and is given by
aγ 
v= 1 − α2 [sign(kxF ), sign(kyF ), 0] .
2
In the first quarter of the Brillouin zone the velocity is

v = [aγ/(2)] 1 − α2 (1, 1, 0) ,
perpendicular to the Fermi curve. The same conclusion pertains for the other three
quarters.
c. The equation of motion is
k̇ = −(e/c)v × B .
Inserting here the velocity from part (b) gives

k̇ = −[2aγB/(2c)] 1 − α2 (−1, 1, 0) .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 559

Electrons in solids 559

The force is parallel to the Fermi segment and therefore the lattice momentum
varies linearly with time along the segment, with a rate given by
√ 
k̇ = −[eaγB/( 2c)] 1 − α2 .
d. The duration of the motion around the Fermi curve is four times the duration
of motion in each quarter, that is
√ arccos(−α) arccos(−α)
T = 4[2 2]  = 162 c √ .
eaγB (1 − α )/2/( c)
2 2 eaγB 1 − α2
One obtains the same result from Eq. (6.163).

Answer s.6.11.
a. According to problem 6.24, the energies of the electron in the two bands are
E = E (0) ± |
γAB (k)| ,

AB (k) =
γ γAB e−ik·rnn = γAB (e−ik·a1 + e−ik·a2 + e−ik·(a1 −a2 ) ) ,
nn
that is

E =E (0) ± |γ|AB | 3 + 2 cos(k · a1 ) + 2 cos(k · a2 ) + 2 cos(k · (a1 − a2 ))
 √
= E (0) ± |γ|AB | 3 + 2 cos(kx a) + 4 cos(kx a/2) cos(ky a 3/2) .
Assume for convenience that E (0) √ = 0 and denote  = E/|γAB |, and u = 3 +
2 cos(kx a) + 4 cos(kx a/2) cos(ky a 3/2). Then, the derivatives are
∂ u ∂ uy
x = = x , y = = ,
∂kx 2 ∂ky 2
∂2 u u2 ∂2 uyy u2y ∂2 uxy ux uy
xx = 2
= xx − x3 , yy = 2
= − 3 , xy = = − ,
∂kx 2 4 ∂ky 2 4 ∂kx ∂ky 2 43
where u = ∂u/∂k , um = ∂ 2 u/(∂k ∂km ),
√
ux = −2a[sin(kx a) + sin(kx a/2) cos(ky a 3/2)] ,
√ √
uy = −2a 3 cos(kx a) sin(ky a 3/2)] ,
√
uxx = −a2 [2 cos(kx a) + cos(kx a/2) cos(ky a 3/2)] ,
√
uyy = −3a2 cos(kx a/2) cos(ky a 3/2) ,
√ √
uxy == − 3a2 sin(kx a/2) sin(ky a 3/2) .
Inserting these expressions in Eq. (6.155) yields
 1  |γ |
∗
= ± AB  .
m αβ 2 αβ
At the Γ point k = 0; the effective-mass tensor is a diagonal matrix, proportional
to the unit matrix, i.e.,
(m∗ )αβ = ∓22 /(|γAB |a2 )δαβ .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 560

560 Introduction to Solid-State Physics

The effective mass at the bottom of the lower band is positive, and hence corre-
sponds to electrons; the effective mass close to the maximum of the√upper band is
negative, i.e., it represents holes. At the M point k = [π/a, −π/(a 3)]. It follows
that
 √   √ 
1 |γAB | 2 1 3/2 2 0√ 2/ 3
=± 2 a √ ∗
, m =± .
m∗  3/2 0 |γAB |a2 2/ 3 −4/3
The eigenvalues of the effective mass are
22 22
m∗ = ± , ∓ ,
3|γAB |a2 |γAB |a2
and the eigenvectors are
√ √
(−1, 3)/2 , ( 3, 1)/2 .
b. Near the Γ point E = 2 k2 /(2m∗ ); one may therefore follow the calculation in
problem 6.10 to obtain g(E) = 2m∗ /(π2 ). Near the K point, according to problem
6.24,
√
γAB (k)| ≈ ±(γAB 3/2)|k|a .
E = ±|
The equal-energy curves are√circles in the plane of the lattice momentum. On each
such curve |∇k E| = |γAB |a 3/2. It then follows from Eq. (6.127) that
2π|k| 4E
g(E) = = .
2π 2 |∇k E| 3π|γAB |2 a2

Answer s.6.12.
a. The density of the electrons can be obtained from the zero-temperature expres-
E
sion n = 0 F dEg(E). In the units of Fig. 6.50, the density of the electrons that
occupy the first band is 1028 m−3 (this is the area below the density of states of this
band). The first step of the second band adds 2 × 1028 m−3 , and the second stair
of this band can add an additional 3 × 1028 m−3 . Since the total density is 5 × 1028
m−3 , the electrons fill only 2/3 of the last step. It follows that the Fermi energy is
at 5.667×10−19 J; as this value is within the band, the material is a metal.
b. At low temperatures the specific heat is given by Eq. (6.129),
π2 −3
cV = 2
g(EF )kB T = 188 T (Jm K−1 ) ,
3
where the temperature is in K and kB = 1.38 × 10−23 JK−1 .
c. When the total density is 7 × 1028 m−3 , the second band is full, and the Fermi
energy lies in-between this and the third band. The gap between these two is
1 × 10−19 J, much larger than the room-temperature energy kB T ≈ 4 × 10−21 J,
and hence the material is an insulator. The specific heat, for a three-dimensional
sample, is determined by Eq. (6.139). This specific heat is rather small, cV ∝
β 1/2 exp[−βEg /2], with βEg /2 ≈ 25. The low-temperature specific heat is tiny.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 561

Electrons in solids 561

Answer s.6.13.
a. The density of states is finite in the range Ed − W/2 < E < Ed + W/2. As there
are 5 orbital states in the band, each capable of accommodating two electrons, the
density of states (per atom) in this range is g = 10/W .
b. At zero temperature the band is filled up to the energy Emax , where
 Emax
10  W
N= dEg(E) = Emax − Ed + ,
Ed −W/2 W 2
from which follows that
Emax = Ed + (N − 5)W/10 .
The additional energy (to the atomic energy) due to the finite width of the band is
 Emax
5  2 
ΔE = dEg(E)E = Emax − (Ed − W/2)2 = N Ed + N W (N − 10)/2 .
Ed −W/2 W
c. The first term on the right-hand side represents the additional energy in the
limit of zero band-width (each electron adds its energy to the atomic energy). The
next term is negative, and is the “gain” in energy due to the finite band width,
which allows the electron to move among the ions. Hence, the binding energy is
−(ΔE − N Ed ) = −W N (N − 10)/2 ,
and it is maximal when N = 5.

Answer s.6.14.
a. Similar to problem 6.23, the nth unit cell, whose length is a, contains two sites;
the wave functions there are ψ2n−1 and ψ2n , and the Schrödinger equations in the
tight-binding approximation become
Eψ2n = −γ− ψ2n−1 − γ+ ψ2n+1 ,
Eψ2n−1 = −γ− ψ2n − γ+ ψ2n−2 .
With the Ansatz ψ2n−1 = A exp[ikna] and ψ2n = B exp[ikna], one obtains
EA = −γ+ Be−ika − γ− B = −B(γ− + γ+ e−ika ) ,
EB = −γ− A − γ+ Aeika = −A(γ− + γ+ eika ) .
The solution is nontrivial provided that the determinant vanishes, which yields
E 2 − (γ− + γ+ e−ika )(γ− + γ+ eika ) = 0 ,
that is
 
E± = ± 2 + γ 2 + 2γ γ cos(ka) = ±2γ
γ− + − + 0 cos2 (ka/2) + Δ2 sin2 (ka/2) .

As in the solution to problem 6.23, there are two energy bands (of positive and of
negative energies); the energy gap in-between them vanishes only when Δ = 0.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 562

562 Introduction to Solid-State Physics

Fig. 6.52

b. Since E− (Δ) < E− (Δ = 0) electrons in the lower band gain energy when Δ = 0.
In the “half-filled” case, all levels in the lower band are occupied, and therefore the
total electronic energy is
 
N a π/(2a)
Etot (Δ) = E− (k) = E (k) .
2π −π/(2a) −
k
As a result, the system has the energy Etot (Δ) + Ed , which depends only on Δ2 ,
i.e., it is even in Δ. Consequently, the states with Δ and with −Δ are degenerate.
The gain in energy is maximal when
dEtot /d(Δ2 ) = −N x . (6.182)
One can verify that the integral in the expression for dEtot /d(Δ2 ) diverges at Δ = 0,
rising from −∞ at Δ = 0 and reaching zero as |Δ| increases. As a result, Eq. (6.182)
has always two solutions, Δ = ±Δ0 .
c. When |Δ| = 1, the state of the system attains depends on the sign of Δ.
When it is negative, γ+ = 0; consequently, dimers (i.e., pairs of atoms) are created
within each unit cell. Dimers belonging to different unit cells are not coupled. As
is the situation with the hydrogen molecule [Eq. (4.14)], the eigenstates of each √
dimer are the bonding state and the anti-bonding one, Ψ± = (ψ2n−1 ± ψ2n )/ 2,
whose energies are ∓2γ0 . On the other hand, Δ = 1 implies that γ− = 0. The
dimers comprise the right atom in a certain cell with the left atom in the cell on its
right. Again the dimers are not coupled. In the infinite system the two states are
degenerate, but not in the finite system. Assume that the atoms are located in the
sites 1,2,3,. . . ,2N . The upmost row in Fig. 6.52 shows the lattice before the Peierls
deformation takes place. The situation for Δ = −1 is illustrated in central row; that
of Δ = 1 is shown in the lower row. There is no physical disparity between the two
configurations in the bulk of the lattice, but there is a significant difference at the
edges. In the first case all electrons belong to the dimers, while in the second there
are two “single” electrons, residing in surface states (similar to those described in
the context of the quantum Hall effect). The energy of each electron in the dimer is
E− (1) = −2γ0 (as the energy band is flat), but the energy of an electron at the edge
is 0, and hence that electron is in an excited state. Remarkably, this state persists
also when |Δ| deviates from 1. Such edge states appear in topological insulators
(for instance, in the quantum Hall effect, see Fig. 6.35).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 563

Appendix E

The Schrödinger equation for electrons in

a magnetic field

Hamilton’s equations in classical mechanics. The classical motion of a par-

ticle can be described in terms of Hamilton’s equations of motion. In this
formulation a canonical momentum, p, is conjugated to the coordinate of the par-
ticle r. When the spin degree of freedom is ignored, the motion is described by a
function of both p and r, denoted H(p, r), such that the equations of motion of the
particle are
ṙ = ∇p H , ṗ = −∇r H . (E.1)
The function H(p, r) is called the Hamiltonian of the particle, and Eqs. (E.1)
are Hamilton’s equations of motion. In the absence of a magnetic field, the
Hamiltonian of a particle in the field of an electric potential that does not vary in
time, ϕ(r), is the total energy of the particle,
H(p, r) = [p2 /(2m)] − eϕ(r) ,
and Eqs. (E.1) give
ṙ = p/m , ṗ = e∇r ϕ = −eE ,
where E = −∇ϕ is the electric field. In this case the canonical momentum of the
particle coincides with mechanical momentum, p = mv = mṙ, and the correspond-
ing equation of motion is just the second law of Newton, mr̈ = mv̇ = ṗ = −eE.
When the particle is also subjected to a magnetic field B, the Lorentz force depends
on the velocity; consequently, the canonical momentum does no longer coincide with
the mechanical one. The classical equations of motion, mv̇ = −e(E + v × B/c) [Eq.
(6.8)], are reproduced by Hamilton’s equations when the Hamiltonian takes the
form
1 e 2
H(p, r) = p + A − eϕ(r) , (E.2)
2 c
where A is the vector potential, which depends on the coordinate and on the time.
The electric and the magnetic fields are
E = −∇ϕ − Ȧ/c , B=∇×A .

563
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 564

564 Introduction to Solid-State Physics

Equation (E.1) yields

v = ṙ = ∇p H = [p + (e/c)A]/m , ⇒ p = mv − (e/c)A .
The canonical momentum p differs from the mechanical momentum mv. Dif-
ferentiating the canonical momentum with respect to time gives
ṗ = mv̇ − (e/c)[∂A/∂t + (v · ∇)A]
(the second term includes both the explicit derivative and the chain derivative due
to the the dependence of r on t). On the other hand, the second of Hamilton’s
equations becomes
∂H ∂ϕ  e  e ∂A ∂ϕ  e  ∂A
ṗi = − =e − p+ A · =e −v· .
∂xi ∂xi mc c ∂xi ∂xi c ∂xi
Comparing the two expressions for pi , and substituting v̇i = ẍi show that
e ∂Ai ∂A ∂ϕ
mẍi = + v · ∇Ai − v · +e . (E.3)
c ∂t ∂xi ∂xi
Problem E.1. demonstrates that this equation is identical to the classical equation
of motion (6.8).

Problem E.1.
Use the identity
v × [∇ × A] = ∇(v · A) − (v · ∇)A ,
(check!) and the expressions for the electric and the magnetic fields to show that
Eq. (E.3) reproduces the classical equation of motion, with the Lorentz force.

The Schrödinger equation with a constant magnetic field. The quan-

tum behavior of the electron is obtained from the solution of the Schrödinger equa-
tion, Ĥψ = Eψ, where Ĥ is the Hamiltonian operator. The latter is derived from
the classical Hamiltonian of the electron, replacing the canonical momentum p
by the operator p̂ = −i∇; the coordinate operator is the same as the coor-
dinate of the electron, i.e., r̂ = r. Thus, in the absence of the magnetic field
Ĥ = p̂2 /(2m) + V̂ (r) = −2 ∇2 /(2m) + V (r). In the presence of a magnetic filed,
the quantum Hamiltonian is
1  e 2
Ĥ = − i∇ + A − eϕ(r) . (E.4)
2m c
Choosing the direction of the magnetic field as the ẑ−axis, one may use the Landau
gauge, A = −Byx̂ (check that indeed B = ∇ × A = Bẑ). In the absence of the
electric potential, the Schrödinger equation takes the form
1  ∂ e 2  ∂2 ∂2 
− i − By − 2 2
+ 2 ψ(r) = Eψ(r) . (E.5)
2m ∂x c ∂y ∂z
As the coordinates x and z do not appear in the Hamiltonian, the motion along
these two directions is that of a free particle. In other words, substituting ψ(r) =
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 565

The Schrödinger equation for electrons in a magnetic field 565

χ(y) exp[i(kx x + kz z)], where kx and kz are constants (to be determined from the
boundary conditions), in Eq. (E.5) yields
2 d 2 1 2 2
− χ(y) + mω 2
c (y − y 0 ) 2
χ(y) = E − k χ(y) , (E.6)
2m dy 2 2 2m z
where ωc = eB/(mc) is the cyclotron frequency, Eq. (6.9). This is the Schrödinger
equation of a one-dimensional harmonic oscillator, vibrating along the ŷ−direction,
around the point y0 = ckx /(eB) = 2B kx , where 2B = c/(eB) is the square of the
magnetic length. The energy levels of the particle are
2 2
E( , kz ) = k + ωc ( + 1/2) . (E.7)
2m z
The last term, derived from the energy levels of the harmonic oscillator, represents
the Landau levels, and confirms Eq. (6.165). The first term represents the motion
of the electron parallel to the magnetic field (as in the classical case, the motion par-
allel to the field is not affected by the field, see problem 6.3). For a two-dimensional
system, the first term in Eq. (E.7) does not exist. The solution of Eq. (E.6) is
1 −[(y−y0 )/B ] 2
χ (y) =  √ e
/2
H [(y − y0 )/ B] ,
2 ! B π
where H is the Hermite polynomial of order . This function is localized within a
region of width of the order B around the point y0 . Thus, in the two-dimensional
system, the wave function is a plane wave along the x̂−axis times a narrow function
along ŷ.
The degeneracy of the Landau levels in a two-dimensional motion.
Assume that the electron is moving in a planar rectangular sample, of dimensions
Lx and Ly . Periodic boundary conditions along x imply that kx Lx = 2πnx . where
nx is an integer, and therefore y0 (nx ) = c2πnx /(eBLx ). Since 0 ≤ y0 ≤ Ly , it
follows that
Lx Ly B Φ
nx ≤ = ,
(ch/e) Φ0
where Φ = Lx Ly B is the magnetic flux through the entire sample, and Φ0 = ch/e is
the quantum flux, precisely the result derived before Fig. 6.34. The degeneracy
of each Landau level (i.e., the number of eigenstates of the momentum along x̂)
is identical to the number of magnetic-flux quanta contained in the area of the
sample. This huge degeneracy is not in contradiction with the Pauli principle, since
the wave functions corresponding to different values of the quantum number nx
describe oscillators that vibrate around different centers y0 (nx ).

Problem E.2.
a. Prove that the expressions for the magnetic filed, B = ∇ × A, and for the
electric field, E = −∇ϕ − Ȧ/c, remain unchanged also when the vector potential
is displaced by a gradient of an arbitrary scalar function, A → A + ∇f (r, t), and
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 566

566 Introduction to Solid-State Physics

the scalar potential by the time-derivative of the same function, ϕ → ϕ − f˙(r, t)/c.
This is the gauge freedom.
b. Prove that a constant magnetic field along ẑ can be also described by the sym-
metric gauge, As = B×r/2. Prove that the difference between the vector potentials
in the symmetric gauge and in the Landau one obeys the result in part (a), and that
both gauges obey the gauge conditions of Coulomb, ∇ · A = 0.
Prove that for ∇ · A = 0, Eq. (E.4) yields
2 2 ie 1  e 2 2
Ĥ = − ∇ − A·∇+ A − eϕ .
2m mc 2m c
Substitute the symmetric gauge and prove that
2 2 eB m
Ĥ = − ∇ + L̂ + ωc2 (x2 + y 2 ) − eϕ .
2m 2mc z 8
The second term represents the “normal” Zeeman effect, that is, the coupling of
the orbital magnetic moment, μz = −eLz /(2mc) with the magnetic field. The third
term is the potential of a two-dimensional harmonic oscillator.
d. Using the symmetric gauge, express the Hamiltonian in cylindrical coordinates,
r = (ρ cos θ, ρ sin θ, z), and show that the eigenstates of the Schrödinger equation
for ϕ = 0 are of the form
ψnM kz (r) = RnM (ρ)eiM θ eikz z ,
where M is the eigenvalue of the angular-momentum component perpendicular
to the plane, L̂z = −i∂/∂θ, and where RnM (ρ) is the radial wave function of a
planar harmonic oscillator. Find the differential equation for RnM (ρ), and show
that R ∼ ρ|M | near the origin, while R ∼ exp[−(ρ/ B )2 /4] far away from the origin.
Explain why this function describes the classical circular motion, as in Fig. 6.35.

The quantum Hall effect. The Hall effect pertains to the geometry portrayed
in Fig. 6.1: electrons are moving along the x̂−direction, in the presence of a
magnetic field along ẑ and an electric field along ŷ (both fields are constant in
space and in time). As E = −∇ϕ, it follows that ϕ = −Ey; this implies that the
Hamiltonian in Eq. (E.4) is augmented by V (y) = eEy. The solution is still that of
an harmonic oscillator, because the algebraic manipulation
mωc2 1 eE eE
(y − y0 )2 + eEy = mωc2 y − y0 + + eE y0 −
2 2 mωc2 2mωc2
shows that the center of the harmonic oscillator is simply shifted from y0 to y0 −
eE/(mωc2 ), and the energy is changed into
2 2
E( , kz , kx ) = k + ωc ( + 1/2) − eEy0 + O(E 2 ) . (E.8)
2m z
The Landau levels attain a dependency on y0 = ckx /(eB), and thence on kx . Since
semiclassically
vx = ∂E/∂px = (∂E/∂kx )/ ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 567

The Schrödinger equation for electrons in a magnetic field 567

[Eq. (6.148)] one finds vx = −Ec/B, and a current density along x̂, jx = −nevx =
(nec/B)E = σxy E, in agreement with the classical result (6.15). In a similar fashion,
one finds that vy = (∂E/∂ky )/ = 0, which implies that the longitudinal conduc-
tance vanishes, σyy = 0.
Surface states. Obviously, the sample has a finite width along ŷ; the surface
there can be described by a potential U (y) which repels the electron (and prevents
it from leaking out). As this potential can be ignored in the bulk of the sample, the
wave functions corresponding to the Landau levels are ψ (kx ), that are localized
along ŷ around the centers y0 (kx ). The energies of states localized close to the
surface, in the lowest order of perturbation theory, are
E → E( , kx ) ≈ E + ψ (kx )|U (y)|ψ (kx ) ≈ E + U [y0 (kx )] , (E.9)
where it is assumed that the width of ψ (kx ) (which is of the order of the magnetic
length B ) is small compared to the length over which U (y) varies. (Note that ψ
is normalized.) The velocity of the electron, along the tangential to the surface, is
1 ∂E( , kx ) 1 ∂U [y0 (kx )] ∂y0 c ∂U
v x ( , kx ) = ≈ = . (E.10)
 ∂kx  ∂y0 ∂kx eB ∂y0
This velocity is positive at the vicinity of the right-side edge, where the potential
increases as the surface is approached (along the positive y−direction), and is nega-
tive along the opposite edge, where the potential decreases (again along the positive
y−direction). When the Fermi energy is in-between two Landau levels the only oc-
cupied states are those for which EF = E( , kx ). These states are localized near the
edges of the sample, and therefore they correspond to the classical surface states
depicted in Fig. 6.35. As in that figure, the electrons in such a state move only
along one direction. When the sample is wide enough for the two opposite edges
to be far away from one another, the probability of scattering from a state moving
near one of the edges into a state moving near the other is small, even when there
are foreign atoms in the sample that cause scattering. It follows that the electrical
resistance vanishes, as indeed is confirmed in the experiment.
Gauge transformation of the Schrödinger equation, the Aharonov-
Bohm effect. The gauge transformation of the potentials is specified in problem
E.2. It is shown there that the transformation leaves the electric and magnetic
fields intact. The modification caused in the quantum wave functions as a result
of this transformation is found as follows. One begins with the time-dependent
Schrödinger equation,
∂ψ 1  e 2
i = − i∇ + A − eϕ ψ , (E.11)
∂t 2m c
and applies the transformation
ϕ → ϕ = ϕ − f˙(r, t)/c , A → A = A + ∇f (r, t) ,
to obtain
∂ψ  1  e 2
i = − i∇ + A − eϕ ψ , (E.12)
∂t 2m c
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 568

568 Introduction to Solid-State Physics

where (as shown in problem E.3.),

ψ  (r, t) = ψ(r, t) exp[ief (r, t)/(c)] . (E.13)
Problem E.3.
Show that the wave function in Eq. (E.13) indeed solves Eq. (E.12).

Equation (E.13) demonstrates that the sole effect of the gauge transformation
is a change in the phase of the wave function: the transformation does not modify
the probability density of the electron’s location, |ψ|2 = |ψ  |2 . The modification
caused in the phase can be measured in situations in which the wave functions
interfere. This can happen when the electron moves along an orbit enclosing a
region where a magnetic field is applied. The component of the field normal to the
plane where the electron is moving can be described by a vector potential that exists
everywhere, also in the regions where the magnetic field vanishes. For instance, the
magnetic field created by an infinite solenoid is active only within the cylinder,
but the corresponding vector potential is a nonzero constant over circular orbits
enclosing the cylinder that extend over the entire space. Though the electron may
move in a region where the magnetic field vanishes, it is still affected by it through
the vector potential; this is a purely quantum-mechanical phenomenon, with no
classical analogue.
Assume that the wave function ψ  corresponds to the situation where A = 0,
that is, the situation in which the magnetic field vanishes. Assume also that the
electric potential vanishes as well, and that the system is not subjected to time-
dependent forces. The vector potential is then A = ∇f , where the function f is
given by an integration over a path, say path “1”, that connects two points,
 r 
f1 (r) = dr · A(r ) ≡ dr · A(r ) . (E.14)
r0 1
Why is this useful? consider the two-slit experiment. The electron emerges from
a point S, then passes through the two slits via two paths, 1 and 2, and finally
impinges on a screen, see Fig. E.1. A magnetic field B is enclosed between the
two paths, in the circle marked in the figure; it creates a vector potential in the
entire plane which contains the two paths. In particular, the vector potential takes
different values on the two paths. The wave functions on these two paths are thus
ψ1 (r) = ψ  (r) exp[ief1 (r)/(c)] , and ψ2 (r) = ψ  (r) exp[ief2 (r)/(c)] , (E.15)
where the function f1 is given in Eq. (E.14), and an analogous equation gives the
function f2 . The probability for finding the electron at a point r on the screen is
|ψ(r)|2 = |ψ1 (r) + ψ2 (r)|2
= |ψ1 (r)|2 + |ψ2 (r)|2 + 2Re[ψ1∗ (r)ψ2 (r)] = 2|ψ  (r)|2 [1 + cos(φ)] ,
where   
e   e
 
φ= dr · A(r ) − dr · A(r ) = dr · A(r ) . (E.16)
c 2 1 c
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 569

The Schrödinger equation for electrons in a magnetic field 569

The difference between the two path integrals is the integral over a closed path,
shown by arrows in Fig. E.1. By the Stokes theorem,
  
 
dr · A(r ) = d rn̂(r) · [∇ × A(r)] =
2
d2 rB⊥ (r) = Φ ,
S S
where n̂(r) is a unit vector normal to the surface on which the integral is calculated,
and B = ∇ × A, and thus B⊥ is the field component normal to the surface. The
integral of this component over the area is precisely the total magnetic flux pene-
trating the surface, Φ. Since the field is active only within the circle in Fig. E.1,
the surface integral yields the same result for any path enclosing this circle. Using
Eq. (E.16), one obtains
e Φ
φ = Φ = 2π , (E.17)
c Φ0
where Φ0 = hc/e is the magnetic flux quantum, discussed in the context of the
quantum Hall effect. The interference pattern in the two-slit experiment displays
a periodic dependence of the intensity on the magnetic flux penetrating the closed
path, with a full period attained whenever the flux is increased by a flux quantum.
These oscillations, as a function of the magnetic flux, are named after Aharonov
and Bohm. They serve to detect the coherence of the wave functions in mesoscopic
systems, since they disappear when coherence is destroyed (recall that coherence is
necessary for constructive interference of the wave functions).

Fig. E.1: The two-slit experiment for an electron in a magnetic field.

Problem E.4.
Figure E.2 illustrates a metallic ring encompassing a solenoid.
a. Calculate the vector potential along the tangential to the ring; write down the
Schrödinger equation for the wave functions of an electron moving on the ring.
b. Find the eigenfunctions and the eigenenergies of the Schrödinger equation.
c. Show that the average charge-current density around the ring, in an eigenstate
with an energy En , is jθ  = −c∂En /∂Φ. Explain this result in terms of the relation
between the current and the induced magnetic moment.
d. Find the total energy of all electrons in the ring at zero temperature. Find the
total current around the ring, as a function of the magnetic flux penetrating the
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 570

570 Introduction to Solid-State Physics

ring, and show that it is finite and does not depend on time as long as there is a
magnetic field within the solenoid. This current is termed persistent current.
For which values of the flux does it vanish?

Fig. E.2: The inner cylinder represents a solenoid of radius a, where there is a
magnetic field B along the axis. The circle C describes a ring of radius b containing
electrons.

E.1 Answers

Answer E.1.
The vector identity in the problem gives
(v · ∇)Ai = ∇i (v · A) − (v × [∇ × A])i = v · ∇i A − (v × B)i ,
using ∇i v = 0. Inserting this into Eq. (E.3) gives
e ∂Ai e 
mẍi = − (v × B)i + e∇i ϕ = −eEi − [v × B] ,
c ∂t c i
which is Eq. (6.8) (without the scattering term).

Answer E.2.
a. Using the shifted potentials in the expressions for the fields, one finds
B → ∇ × [A + ∇f ] = ∇ × A + ∇ × [∇f ] = B ,
E → −∇[ϕ − f˙/c] − [Ȧ + ∇f˙]/c = −∇ϕ − Ȧ/c = E .
b. The symmetric gauge gives
As = B[ẑ × r]/2 = B + (−yx̂ + xŷ)/2 ,
leading to
∇ × As = ∇ × (−yx̂ + xŷ)B/2 = Bẑ ,
∇ · As = ∇ · (−yx̂ + xŷ)B/2 = 0 .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 571

The Schrödinger equation for electrons in a magnetic field 571

In the Landau gauge AL = −Byx̂, and thus

∇ · AL = ∇ · (−yx̂B) = 0 .
The difference between the two gauges is
As − AL = (yx̂ + xŷ)B/2 = ∇f , f = Bxy/2 .
c. Expanding the square in Eq. (E.4) gives
e 2 ie    e 2
− i∇ + A = −2 ∇2 − ∇·A+A·∇ + A2 .
c c c
Applying the central term on the wave function ψ results in
(∇ · A + A · ∇)ψ = ψ(∇ · A) + 2(A · ∇)ψ = 2(A · ∇)ψ ,
where the result of part (b), ∇ · A = 0, is used. The contribution of this central
term to the Hamiltonian is
e eB eB eB
−i [B × r] · ∇ = [ẑ × r] · p = ẑ · [r × p] = L .
2mc 2mc 2mc 2mc z
In the symmetric gauge A2 = B2 (x2 + y 2 )/4; thus the required expression is re-
trieved.
d. In cylindrical coordinates
∂2 1 ∂ 1 ∂2 ∂2 ∂
∇2 = 2
+ + 2 2
+ 2
, L̂z = −i ,
∂ρ ρ ∂ρ ρ ∂θ ∂z ∂θ
and therefore the Schrödinger equation in these coordinates is
2  ∂ 2 1 ∂ 1 ∂2 ∂2  ∂ 1  eB 2 2
− + + + − iμ B B + ρ ψ = Eψ ,
2m ∂ρ2 ρ ∂ρ ρ2 ∂θ2 ∂z 2 ∂θ 8m c
where μB = e/(2mc) is the Bohr magneton. Using for the wave function the form
ψ(ρ, θ, z) = R(ρ)Θ(θ)Z(z)
separates the equation into three equations. First,
2 ∂ 2 Z 2 kz2
− = E z Z ⇒ Z = e ikz z
and E z = .
2m ∂z 2 2m
It follows that for Θ(θ) = exp[iM θ] one finds L̂z Θ = −i∂Θ/∂θ = M Θ. The
remaining equation is
 ∂2 1 ∂ M2  1 2m
+ − 2 − 2 ρ2 R(ρ) = − 2 (E − Ez − μB M )R(ρ) .
∂ρ 2 ρ ∂ρ ρ 4 B 
The Ansatz R ∼ ρu for small ρ, with u ≥ 0, yields (u2 − M 2 )ρu−2 + O(ρu ) = 0,
and hence u = |M |. For very large ρ, one tries R ∼ exp[−aρ2 ], which gives for the
left-hand side the expression
[(2aρ)2 − 4a − M 2 /ρ2 − ρ2 /( 4
B )]R ,
while the right-hand side is proportional to R. Hence, a = 1/(4 2B ), namely, R ∼
exp[−(ρ/ B )2 ]/4. it follows that the range in which the wave function is significant
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 572

572 Introduction to Solid-State Physics

is of the order of the magnetic length, ρ ∼ B . For M = 0, the function is finite at

the origin and decays at large distances. For M = 0 the wave function is zero at the
origin, then reaches a maximum at a distance of the order of the magnetic length,
and finally decays. The result is a circular region around the origin in which the
wave function is significant, similar to the classical cyclotron radius.

Answer E.3. Substituting the gauge transformation and Eq. (E.13) in Eq. (E.12)
gives
∂ ef 1  e 2 ef
i [ψei c ] = − i∇ + (A − ∇f ) − e(ϕ + f˙/c) [ψei c ] . (E.18)
∂t 2m c
Using
∂ ef ef
i + e(ϕ + f˙/c) [ψei c ] =ei c iψ̇ − ψef˙/(c) + ψe(ϕ + f˙/c)
∂t
ef ∂
= ei c i + eϕ ψ ,
∂t
and
ef ef
− i∇ + e(A − ∇f )/c [ψei c ] =ei c − i∇ψ + ψe∇f /c + ψe(A − ∇f )/c
ef
= ei c − i∇ + eA/c ψ ,

one finds that Eq. (E.18) is indeed obeyed when ψ is the solution of Eq. (E.11).

Answer E.4.
a. Due to the rotational symmetry around the solenoid, the vector potential is
tangential to the ring and is independent of the location there. The line integral
around the ring is thus

dr · A = 2πbAθ .

On the other hand, by the Stokes theorem this integral is the magnetic flux through
the ring, i.e., 2πbAθ = Φ = πb2 B. It follows that Aθ = Φ/(2πb) = a2 B/(2b), and
then Eq. (E.4) becomes
1  e 2 1  1 ∂ eΦ 2
Ĥ = − i∇θ + Aθ = − i + ,
2m c 2m b ∂θ 2πcb
where θ is the angular location of the electron on the ring. The eigenstates are
ψ(θ) exp[−iEt/], with
1  1 ∂ eΦ 2
− i + ψ(θ) = Eψ(θ) .
2m b ∂θ 2πcb
√
b. The solution is ψn = exp[inθ]/ 2πb, as can be shown by substituting this form
in the Schrödinger equation; the eigenenergy is
2  Φ 2  Φ 2
En = n + = e 0 n + .
2mb2 Φ0 Φ0
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 573

The Schrödinger equation for electrons in a magnetic field 573

The continuity of the function round the ring, ψ(θ) = ψ(θ + 2π), implies that n is
an integer.
c. The classical equations of motion of Hamilton yield that the velocity is v =
(p+eA/c)/m [see the discussion following Eq. (E.2)]. Therefore, the charge current
carried by each electron is
j = −ev = −e(p + eA/c)/m .
In the present case the motion is tangential to the ring, and therefore the current
density per electron is
jθ = −evθ = −e(pθ + eAθ /c)/m .
In the quantum-mechanical description
pθ → p̂θ = −i∂/(b∂θ) .
√
The quantum average of the current in the state ψn = exp[inθ]/ 2πb is
e  Φ
jθ n = −evθ  = −eψn |p̂θ + eAθ /c|ψn  = − n + .
2πmb2 Φ0
A straightforward differentiation shows that this expression is identical to
−c∂En /∂Φ. Indeed, the relation between the current around the ring and the
induced magnetic moment leads to this identification: the magnetic moment is
M = −∂En /∂B, and the current is jθ  = cM/A.
d. At zero temperature, all energies up to the Fermi energy are filled, and hence
the total energy is

Etot = En ,
En ≤EF

while the total current is


j = −c∂Etot /∂Φ = −2e0 (n + Φ/Φ0 ) ,
n

where each term is multiplied by 2 to account for the two spin states, and the sum
runs over all levels for which En ≤ EF . In the absence of the magnetic flux one
may join together terms with n = ±|n|, and thus the sum vanishes. This means
that the persistent cutrent flows only in the presence of a magnetic flux. However,
it vanishes when this flux equals an integer product or a half integer product of the
flux quantum. For instance, when Φ = N Φ0 the current is
  
j = −2e0 (n + N ) = −2e0 n , −Nm ≤ n = n + N ≤ Nm = EF /e0 ,
n n

and again the contributions of n = ±|n | cancel each other. In a similar fashion,
when Φ = (N + 1/2)Φ0 then
 
Nm
j = −2e0 (n + N + 1/2) = −2e0 (n + 1/2) , EF = e0 (Nm + 1/2)2 .
n n =−N m −1
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 574

574 Introduction to Solid-State Physics

The contribution of the term with n = |n | cancels that of n = −|n | − 1, and the
current vanishes. Consequently, the persistent current is finite only when Φ/Φ0 is
not an integer or a half integer. Denoting Φ/Φ0 = N + x, with −1/2   x  1/2,
one finds that there are occupied states when − EF /e0 ≤ n + N + x ≤ EF /e0 .
If Nm − N is the integer part of EF /e0 , then the current is
 
Nm
j = −2e0 (n + N + x) = −2e0 (n + x) .
n n =−Nm

The terms in the sum can be arranged in pairs, with n = ±|n |, and then j =
−2e0 x(1 + 2Nm ). It follows that the current is a periodic function of the flux, which
decreases linearly between Φ/Φ0 = N − 1/2 and Φ/Φ0 = N + 1/2, etc, around each
integer value of N .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 575

Chapter 7

Selected topics

7.1 Preface

This chapter overviews several topics that require certain extensions of the concepts
introduced in the previous chapters. It includes two main parts: (i) topics in soft
condensed-matter, i.e., materials that are not translationally invariant, and (ii)
topics in mesoscopic physics, that is the study of electronic properties in tiny systems
for which the macroscopic descriptions presented in Chapter 6 are not sufficient.
Section 7.2 presents aspects of the physics of soft condensed-mater, e.g., liquid
crystals, which are periodic along a certain direction in space and behave like liq-
uids along the other directions. Another example are polymers; a ubiquitous simple
picture for the latter envisages them as one-dimensional sinuous chains, whose ar-
rangement in space is quite complex. Another subject introduced in Sec. 7.2 are
fractals, structures that cannot be mapped onto themselves by translations; they
do “return to themselves” when viewed on different length scales, under a transfor-
mation called “inflation”. An example of fractal structures appears in percolation
theory, which describes random mixtures of components with different properties.
Section 7.3 is devoted to mesoscopic physics, which describes systems whose
typical sizes are in-between the microscopic, atomic scale and the macroscopic one.
Though tiny crystals can be represented by finite-size lattices (with the appropriate
conditions on their boundaries) their properties, in particular the electric ones, are
dominated by quantum mechanics; the semiclassical theory of Drude [even in its
quantum-mechanical version, Eq. (6.45)] is not sufficient to capture their behavior.
For instance, quantum dots behave as artificial atoms, the electrical conductance
of a point contact attains discrete values, and the electrical conductance of quan-
tum wires is specific for each one of them. In contrast with macroscopic samples,
whose physical properties are independent of the specific specimen (due to self av-
eraging of the properties among the various parts of the sample), measurements
taken on mesoscopic samples vary from one to another, though the disparate values,
i.e., the fluctuations, follow certain universal properties.

575
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 576

576 Introduction to Solid-State Physics

7.2 Soft condensed-matter

An important phase of matter which is not a gas, liquid, or solid, are liquid crys-
tals, that are built of long organic molecules. Liquid crystals possess unique optical
properties, mostly because the long molecules modify the polarization of the light
passing through them; the modification varies with the temperature, or under the
effect of electric and magnetic fields. For this reason liquid crystals are used as
screens (liquid-crystal displays, LCD), in thermometers, and so on. At low enough
temperatures even these linear molecules arrange themselves in periodic solid crys-
tals. Examples of such crystals are displayed in Figs. 4.23, 4.24, and 4.25. In
many cases there is no direct phase transition from the solid to the liquid as the
temperature is increased. Rather, there appear intermediate phases, with partial
periodicity.

Fig. 7.1: Phases of a liquid crystal; from right to left: liquid, nematic liquid-crystal,
smectic liquid-crystal A, smectic liquid-crystal C, solid.

The nematic phase. Intermediate phases between the solid and liquid ones
are shown in Fig. 7.1. At high temperatures the molecules are arranged as in a
conventional fluid, that is, both the locations of the molecules and the directions
of their axes are random. At most there are short-range correlations among the
locations and orientations of close-by molecules (see the right panel). When the
density is raised, or the temperature is reduced, there appears at times the nematic
phase (nematic is the Greek word for a thread), in which the centers of mass of the
molecules are still like in a normal fluid, but the axes tend to be parallel to each
other, as in the second panel of Fig. 7.1, along a vector n̂ called the “director”. As
opposed to the magnetic order, where the magnetic dipole moments are parallel or
are antiparallel to a certain vector, there is no such distinction in this case, and the
pattern resembles an arrangement of quadrupoles parallel to each other. In other
words, whereas the order parameter of a ferromagnet is given by the average (over all
magnetic moments) of cos(ϑ) (ϑ is the angle between n̂ and the magnetic moment),
and this average vanishes in the paramagnetic (disordered) phase and increases
upon cooling in the ferromagnetic phase, the order parameter of the nematic phase
is the average (over all molecules) of

[3 cos2 (ϑ) − 1]/2 . (7.1)

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 577

Selected topics 577

Here ϑ is the angle between n̂ and the molecule’s axis. In the fully-ordered nematic
phase ϑ vanishes, and the order parameter is 1. In the disordered phase the molecule
is free to be oriented along an arbitrary axis, which renders the average of cos2 (ϑ) to
be 1/3 (check!), namely the order parameter vanishes. As explained in Sec. 4.4, the
van der Waals potential energy is lower in configurations in which long molecules
are parallel to each other. This energy gain explains the origin of the nematic phase.
The smectic phases. A further reduction in the temperature may lead to the
appearance of the smectic phase (the Greek word for soap or lather), the third
or the fourth panel in Fig. 7.1. In this phase the molecules are located on planes,
whose average distance from each other is fixed; the molecules are still free to move
within each plane as in a liquid. Denoting the unit vector normal to planes by
N̂, then the correlation function along N̂ is that of a solid, (third panel in Fig.
1.3), while within the plane it is that of a liquid (second panel in that figure).
Quantitatively, the order parameter that describes the smectic phase is the average
of the local density of the molecules, n(r). This average is independent of r in the
nematic phase, but is periodic along the N̂ direction in the smectic phases,
n(r) ≈ n0 + 2nq cos(q N̂ · r) + . . . , (7.2)
where q = 2π/ and is the distance between the planes. n0 and nq are constants
which represent the mean density of the material and the amplitude of the first
harmonic of the oscillations along the normal to the planes, respectively, while
. . . represents higher harmonics of the wave vector (integer multiples of q) that
appear at lower temperatures. The resulting spatial pattern is a combination of the
nematic order parameter, Eq. (7.1), and the spatial one, Eq. (7.2). At sufficiently
low temperature the material becomes a periodic solid, the left panel in Fig. 7.1,
and then n(r) is periodic along any of the basis vectors.
Figure 7.1 exhibits a specific series of transitions among various phases of a liquid
crystal. However, other phases are possible in Nature. The third panel (from right)
in the figure shows a special case of the smectic phase, with the molecules’ axes
normal to the plane. This phase is usually marked by A: the director n̂ is parallel
to N̂. The next panel exhibits yet another smectic phase, marked by C. In this
phase the axes of the molecules (along n̂) make a finite angle with the direction of
N̂. Other smectic phases are shown in Fig. 7.2 (by a side view, and a single plane
viewed from above). The upper row displays phases that are liquid-like in each
plane. On the left of the figure the molecules are normal to the plane (the smectic
A phase), and on the right they are inclined, as in the smectic C. The two lower
rows show solid phases with a periodic order. The second row, on the other hand,
illustrates the hexatic phases: the pattern of the molecules is not periodic, but
the orientations of the lines that join neighboring molecules are correlated; those
tend to point along specific directions. The points in the figure denote the periodic
lattice, and the circles (or the triangles, or the ellipses) indicate the projections of
the molecules on the plane. The centers of the molecules in the solid phases are
congruent with the triangular lattice; in the hexatic phase they are not located
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 578

578 Introduction to Solid-State Physics

on a periodic lattice but the lines that join neighboring molecules are parallel to
the directions of the triangular lattice, with angles that are multiples of 60◦ (this
is the origin of the term “hexatic”). The hexatic phase appears usually between
the smectic phase and the solid one. This topological phase replaces at times the
periodic one in two dimensions (see Sec. 5.5).

Fig. 7.2: Various phases of liquid crystals. The left column illustrates the order
within the planes (liquid, hexatic phase-with correlations among the bonds, or a
solid). The next two columns show “normal” states, in which the molecules are
oriented normal to the planes, or are inclined with respect to it. [J. D. Brock, R.
J. Birgeneau, J. D. Litster, and A. Aharony, Liquids, crystals, and liquid crystals,
Physics Today 42, 52 (1989).]

The cholesteric phases. The axes of the molecules in the smectic phases are
normal to the plane or are inclined, making a finite angle with the normal to the
plane. Figure 7.3 illustrates the cholesteric phase, in which the molecules lie
in the plane, such that their axes are parallel to each other (like in the nematic
phase), but their locations are random, as in a liquid. This is the phase adopted by
molecules derived from cholesterol, and hence the name. The molecules may rotate
from one plane to the other (clockwise or anticlockwise). When the corresponding
angle is a rational fraction times 180◦ , the structure repeats itself after a finite
number of planes, and thus there is a finite unit cell along the direction normal to
the planes (the structure in the figure includes four planes). Such a pattern, which
depends on the direction of a rotation (of the form of a screw) is called helical or
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 579

Selected topics 579

chiral. Along the normal to the planes, the director n̂ rotates around N̂. That is,
for N̂ along ẑ,
n̂ = cos(qz)x̂ + sin(qz)ŷ . (7.3)
The distance between successive planes (which determines the wave vector q), and
the size of the unit cell along the normal to the planes, depend on the temperature.
These lengths dictate the wave length of the light that will be reflected from the
liquid crystal. Consequently, the liquid crystal changes its color as the temperature
varies; this property serves to make it a thermometer.

Fig. 7.3: A cholesteric liquid crystal; the molecule are parallel to each other within
the planes but are rotated by 45◦ between neighboring planes.

Problem 7.1.
a. The symmetries of a liquid crystal are determined by the symmetries of the
average density, n(r). Determine the symmetries of the phases in Figs. 7.1, 7.2,
and 7.3.
b. Determine the lattice structures of the solid phases in Fig. 7.2.

Polymers. Another research field in soft condensed-matter is centered around

the properties of polymers. Simple models for those are based on one-dimensional
chains. Each chain is built of basic units, called monomers; the one-dimensional
chain is sometimes augmented by short side-branches. When immersed in a solution,
the chain attains various forms, determined by the forces the atoms in the solution
exert on it, and also by self-exerted forces: different parts apply forces on other
parts. The optimal form of the chain at a given temperature is dictated by the
entropy. Polymers, for instance, proteins or the DNA and the RNA molecules, are
key objects in biology. Ensembles of polymers create a diversity of flexible materials
as rubber and plastic systems. When polymers arrange themselves on a periodic
lattice, the diffraction pattern contains information on the structure factor of each
unit cell (see Sec. 3.9), and hence on the structure of a single polymer.
Random walks. The simplest model of an isolated polymer is based on ran-
dom walks: beginning with the monomer at one end of the polymer, the successive
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 580

580 Introduction to Solid-State Physics

monomer points at a random direction in space. The next monomer points along
another random direction and so on, till the other end of the polymer. A random
walk along such a path is called diffusive motion; under certain conditions this
is the motion of a particle moving in space and colliding with random scatterers.
This is also the motion, i.e., the diffusion, of an electron in the Drude picture [Sec.
6.2, or Eq. (6.45)], since the direction of the motion is changed randomly after each
scattering event.
Let us denote each of the random steps between the successive monomers by
Δrn . After N steps the walk reaches the point

N
r(N ) = Δrn .
n=1

When the steps are independent (i.e., they are not correlated), then
Δrn  = 0 , Δrn · Δrm  = a2 δnm (7.4)
(the angular brackets indicate an average over all possible configurations, and a is
the average length of the step; δnm is the Kronecker delta function). In this case
r2 (N ) = a2 N , (7.5)
(see problem 7.2). The typical distance between the two ends of the polymer, and
therefore also the characteristic linear dimension of the region in space covered by
the monomers in the chain, is

L ∝ r2 (N ) = aN 1/2 , or N = ALD , (7.6)
where A is a dimensionless coefficient, and D = 2. This power is independent of
the spatial dimension of the space in which the random walk is located.

Problem 7.2.
a. Prove the relation r2 (N ) = a2 N , given Eq. (7.4).
b. Each pair of two consecutive steps, in a random-walk of N monomers where N
is an even number, is replaced by a single step,
Δrn = Δr2n + Δr2n−1 .
The result is a random walk of N/2 steps, that begins at the origin and ends at the
same point as the original walk, r (N/2) = n=1 Δrn = r(N ). Prove that
N/2

Δrn · Δrm  = (a )2 δnm ,

and obtain the average length of the new step a . Use the result to prove that
[r (N/2)/a ]2  = N/2 .
Extend the result to a walk for which each step is a sum of basic steps. This
mapping, where the information pertaining to short distances is averaged over to
yield the average behavior at longer distances, is called “renormalization”.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 581

Selected topics 581

Problem 7.3.
A biased random walk is defined by
Δrn  = r0 , (Δrn − r0 ) · (Δrm − r0 ) = a2 δnm .
The bias, in the case of a particle motion, can be caused by e.g., a non-homogeneous
distribution of the scatterers, or an external force applied along a certain direction
(for instance, gravity). In the Drude picture the bias is due to an electric field.
Assume that the deviation away from r0 is spatially symmetric, i.e.,
(Δrn − r0 )α (Δrm − r0 )β  = a2 δαβ δnm /3 ,
α, β = 1, 2, 3 are the cartesian components.
a. Calculate r(N ), [r(N )]2 , and the mean standard deviation, [r(N )−r(N )]2 .
b. Prove that the probability per unit volume that the end of a random walk of
N  1 monomers is at a distance R from the origin, i.e., the initial point, is
−3[R− R(N )2 ]
exp 2N a2
P (R, N ) = .
(2πN a2 /3)3/2
This is a Gaussian (in three dimensions). This formula expresses the central limit
theorem. Find the probability to return to the origin. [Hint: prove that

P (R, N ) = d3 R P (R , N − 1)P (R − R , 1) .

The consequence is that the Fourier transforms obey

P(k, N ) = [P(k, 1)]N .
Exploit the Taylor expansion of P(k, 1), and derive the inverse transform.]
c. Use the results of part (b) to find [r(N )]2 .

Fractals. Equation (7.6) is a power law that relates the “mass” of the polymer,
which is proportional to N , to the linear length, L, of the region where the polymer
resides. Such a relation characterizes fractals; the power D is called the fractal
dimension of this structure. The name originates from the similarity between Eq.
(7.6) and the relation for the volume of a d−dimensional cube, V = Ld , where d is
the spatial dimension. The fractal dimension need not be an integer. The density
of the fractal is ρ = N/V = (A/ad )(L/a)D−d . When D < d (for instance, a random
walk in three dimensions) the average density decreases as the sample is increased.
As the number of elementary units increases, the fractal occupies a diminishing
part of the volume. On the other hand, for D > d (e.g., a random walk in one
dimension), the path “visits” each point numerous times, and the density increases
with the number of elementary units.
Consider a sphere centered around the initial point of the random walk (taken
as the origin). The number of points within such a sphere of radius r is proportional
to (r/a)D , and therefore the radial correlation function in a d−dimensional space is
g(r) ∝ (r/a)D−1 /(r/a)d−1 = (r/a)D−d ,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 582

582 Introduction to Solid-State Physics

[Eq. (1.1)]. The scattering intensity off such a structure, from Eq. (3.48), is
proportional to the Fourier transform of the correlation function, and hence to

Γ(q) ∝ (qa)−D , problem 7.4.

Problem 7.4.
According to problem 3.18 (part (b)), the scattering intensity off a three-dimensional
sample, in which the radial correlation function is Γ(r) = g(r) is

 4π ∞
Γ(q) = rdrΓ(r) sin(qr) .
q 0
For a finite fractal, that contains N = A(L/a)D elementary units in a region whose
linear size is L, the correlation function is g(r) ∝ (r/a)D−d for r ≤ L, and h(x)
decays to zero for r ≥ L. Let us hence assume
g(r) ∝ (r/a)D−d h(r/L) ,
with h(x) ≈ 1 for x < 1 and decays to zero as x > 1.
a. Prove that when qL  1, then

Γ(q) = 1 − (qRg )2 /3 + O[(qL)4 ] ,

r2  r2
Rg2 = = d2 rg(r) ∝ L2 .
2 2
Rg is the average “radius” of the fractal.

b. Prove that when qL  1, then Γ(q) ∝ q −D .

As opposed to solid crystals, which are mapped on themselves under translations,

fractals are mapped on themselves when the length scale on which they are examined
is varied. For instance, in part (b) of problem
√ 7.2, a reduction of the instantaneous
picture of a random walk by a factor of 2 produces another picture that appears
very much like certain fragments of the original random walk. Figure 7.4 illustrates
the same phenomenon: when every four steps out of the 400 of a certain random
walk are joined together into a single step, the figure is “squeezed” by a factor
of 2. Comparing with the random-walk picture obtained after 100 steps of this
random walk, one cannot see a significant disparity. This property is termed “self-
similarity”. Another example for such a system is displayed in Fig. 2.31: different
generations of the Fibonacci lattice appear the same when measured with different
“rulers”.
Self-avoiding random walks. In one-dimension, a random walker moves
forwards and backwards, so that every point is visited many times, and the length
covered is far shorter than the number of steps, (L/a) ∝ N 1/2  N . A real
polymer, on the other hand, is composed of monomers. Each monomer occupies a
finite volume in space, and two of them cannot occupy together the same volume
in space. (This can be also understood as the result of the repulsion between the
electronic clouds at short distances, due to the Pauli principle, see Chapter 4.)
Consequently, in one dimension the length of a polymer (which cannot return back,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 583

Selected topics 583

(a) (b)

(c) (d)
Fig. 7.4: Random walks with steps of length a = 1 and random planar orientations.
(a) One hundred steps. (b) Four hundred steps. (c) As in (b), with each step
representing four original steps. (d) The pattern of (c), doubly reduced.

in contrast with a random walker) of N monomers must be L = aN , and the

random-walk model does not describe it faithfully. This is not the case in higher
dimensions. There, the probability that a random walker hits the same point more
than once is proportional to the product of the probabilities to be in that point,
namely to the square of its density, ρ2 ∝ (LD−d )2 . The number of points visited
more than once is proportional to the product of this probability with the volume,
that is, Ld ρ2 ∝ L2D−d . Hence, for very large L  a, in dimensions higher than
du = 2D = 4, the path of the random walker does not collide with itself. The
dimensionality du , above which this simple behavior appears, is called the upper
critical dimension. Such critical dimensions are ubiquitous in other physical
problems; this is the reason why certain problems are investigated as functions of
their dimensionality. The random walk model can serve as a faithful model for the
polymer in dimensions d > du . In contrast, in lower dimensions i.e., for d < du ,
there is a finite probability (at large L) for the path of the walker to collide with
itself and another model is needed. Such a model is the self-avoiding random
walk. For 1 ≤ d ≤ du = 4, this model produces fractal dimensions less than 2,
approximately DSAW = (d + 2)/3 (problem 7.5). As the dimension of the fractal is
smaller than 2, the density of the polymer is smaller than that of the simple random
walk. In one dimension, as mentioned, one expects that N = La; the dimension
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 584

584 Introduction to Solid-State Physics

DSAW = 1 is indeed equal to the (approximate) formula for d = 1. For d in-between

1 and 4 the fractal dimension of the self-avoiding walk increases linearly from 1 to
2.

Problem 7.5.
a. Explain why the number of neighboring monomers of each monomer whose
volume is of the order of ad in a polymer described by a random walk, is ad (N/Ld ).
This implies that the repulsion energy between pairs of monomers is approximately
E = CN 2 /(L/a)d , where C is the repulsion energy of a single pair.
b. Show that by multiplying the Boltzmann weight factor with the probability derived
in part (b) of problem 7.3 gives the probability to find a walker that “avoids” colli-
sions, and prove the Flory approximation: The maximal value of this probability,
as a function of L, is when N ∝ L(d+2)/3 .

Percolation. Another description of polymers is based on a dilute lattice: every

lattice site is occupied by a monomer which may form bonds with its neighbors.
The concentration p of such bonds increases with time. This process happens, for
instance, while cooking an egg, or among molecules that create a gel. When these
bonds are replaced by tubes through which water vapor can flow among coffee grains
that fill a container, the process mimics a percolator. For this reason the model
is termed percolation. Figure 7.5 illustrates examples for clusters of bonds, as a
function of p. For p  1 there are only small clusters in which each monomer is
connected to at least one of its neighbors. The clusters increase as p grows, and at
a certain probability pc , called the percolation threshold, there appears a cluster
that connects opposite faces of the sample, even when the latter is very large (the
mid panel in Fig. 7.5). A further increase of p leads to a further increase of that
cluster, which now contains islands with finite clusters that are not attached to it
(the left panel). As p increases, the finite clusters gradually diminish in size; the
fully-periodic lattice is reached when p = 1. When p < pc the monomers in the gel
move as in a viscous liquid. In the example of the percolator, water steam cannot
reach from one end of the sample to the other. Once p > pc , the sample behaves like
a flexible gel, or a rubber, and is easily deformed. At p = pc there are clusters of
all sizes; the largest cluster has the geometry of a fractal, with a fractal dimension
which is 1 at d = 1, and reaches 4 for d > 6.
Another example for percolation is a lattice of metallic wires. The complete
lattice is capable of conducting electricity between its opposite edges. One may
calculate the electrical resistance of this lattice by solving the Kirchhoff equations
for a lattice of resistors. When the wire bonds between nearest neighbors are
cut at random, the conductance decreases. Increasing the concentration of broken
bonds leads to the appearance of clusters of bonds that are disconnected from the
biggest one, which connects the two edges of the lattice. Once the concentration of
broken bonds in a very large sample exceeds 1 − pc , that cluster is dismantled and
the conductance vanishes. This is the classical metal-insulator transition.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 585

Selected topics 585

Fig. 7.5: Percolation on the square lattice: the probability to form a bond is p.
Right - p = 0.4, middle - pc = 0.5927 . . ., left - p = 0.8. The figures are created by
choosing for each possible bond a random number between 0 and 1. If that number
is less than p the bond is formed; otherwise, it does not exist.

The percolation model describes also the dynamics of fluids in porous media.
The sample is viewed as a periodic collection of boxes, some filled with solid material
while the others are empty (or filled with a fluid, e.g., water or oil). The permeability
of such a system to the flow of the fluid in the voids is large when the concentration
of empty boxes is large, and it decreases to zero as that concentration tends to the
percolation threshold.
The percolation is also applied to lattices of magnetic ions, in which the
exchange interaction among neighboring ions is effective only when the bond in-
between them is occupied by an additional ion (for instance, the lattice of lanthanum
cuprate [Fig. 2.21(b) and problem s.2.17]. The antiferromagnetic interactions be-
tween nearest-neighbors coppers is established only through the oxygen that resides
on the bond connecting them. Diluting the oxygen ions decreases the magnetic or-
der. In the simplest classical model, the antiferromagnetic order totally disappears
once the concentration of the oxygens is less than pc .
This reduction in the antiferromagnetic moment as a function of the concentra-
tion resembles the diminishing of that moment as the temperature is being raised
from zero to the transition temperature of the magnetic order in the complete lat-
tice. This similarity leads to investigations of the percolation problem by methods
developed for the study of phase transitions.
Fractal growth. The discussion following Fig. 2.3 and the one in Sec. 2.9
describes the growth of periodic crystals. In many cases though, the absorbed
atoms are not joined together to form a crystalline cluster. Rather, they create a
fractal. For instance, an electrode immersed in a solution collects on its surface ions
from the solution that arrive there at random; the electric potential of those prevents
additional atoms from moving into the space between the “branches” already grown
on the cluster. Figure 7.6 illustrates a cluster grown from a solution on a planar
surface; the measured fractal dimension (found by counting the number of atoms
in a series of circles of increasing radii, and plotting these numbers as a function of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 586

586 Introduction to Solid-State Physics

the radii of the circles) is D ≈ 1.66. A similar growth in three dimensions yields
D ≈ 2.5. In both cases the fractal dimension is smaller than the spatial one, which
implies that the cluster density diminishes as its size increases.
A famous numerical procedure to mimic such growths is to consider diffusing
particles; once the first one hits randomly the origin, it “sticks” to it. A second
particle arriving there is glued to the first, and so on. Since the particles hit the
growing cluster at random orientations and then they are not free anymore to reach
an equilibrium (as opposed to the description displayed in Fig. 2.3), the growing
cluster is not crystalline. The growing branches prevent additional particles to
approach the origin, leading thus to a fractal structure, like the one in Fig. 7.6.
This numerical procedure is termeddiffusion limited aggregation (DLA).

Fig. 7.6: Fractal cluster of zinc atoms, grown from a solution onto a plane.
[M. Matsushita, M. Sano, Y. Hayakawa, H. Honjo, and Y. Sawada, Fractal Struc-
tures of Zinc Metal Leaves Grown by Electrodeposition, Phys. Rev. Lett. 53, 286
(1984).]

7.3 Mesoscopic physics

Preface. As mentioned in Chapter 1, many recent new technologies are based on

small units of materials, of typical sizes of several nanometers up to a few hundreds
of them (1 nanometer = 10−9 meter = 10 Å = 10−3 micron). As this scale
is in-between the microscopic (atomistic) one and macroscopic scale, this field of
research is termed mesoscopic physics, or nanotechnology. One of its main
features is the fact that the electrons need to be treated quantum-mechanically, as
the semiclassical picture of Drude is not applicable there.
Section 1.2 describes properties of systems whose dimensions is less than 3. In
the extreme situation, a system is one-dimensional (or is two-dimensional) once its
widths along the two directions normal to it (or along the perpendicular direction)
comprise a single layer of atoms. Note though, that in problem 5.21 it is found that
the specific heat of the phonons can appear as if it belongs to a two-dimensional
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 587

Selected topics 587

system even when the width of the system is finite, and that there is a crossover
from a two-dimensional behavior to a three-dimensional one as the width increases.
The electronic properties of mesoscopic samples depend on their linear dimensions
as well: the physical properties of one-dimensional and two-dimensional systems
can differ quite significantly from those of the macroscopic three-dimensional ones.
For instance, phonon excitations can destroy the solid phase at low dimensions (Sec.
5.5). Likewise, a tiny amount of impurities turns a conductor into an insulator in
dimensions less than three due to repeated coherent scattering of the electrons off
the impurities, a process which is less detrimental in three dimensions; there only
at a finite concentration of impurities a metallic system becomes an insulator (at
zero temperature). The density of states is another property that is sensitive to the
dimensionality of the sample; this leads to unique quantum phenomena at one and
two dimensions (Sec. 6.11).
Consider a system of linear size L (along all directions), in particular when L
is larger than all other characteristic lengths of the electron motion. The Drude
picture (Sec. 6.2) relates the electric resistance to collisions of electrons with cer-
tain imperfections in the material. The electron motion resembles that of a random
walker (Sec. 7.2), that is, it moves diffusively. The distance between consecutive
scatterers is of the order of the mean-free-path, . At low temperatures, the col-
lisions are mostly with impurities (Fig. 6.4), and are elastic. The corresponding
mean-free-path, e (which is determined by the impurities’ concentration) is of the
order of 1-100 microns.
In the opposite limit, where L is shorter than e , the motion is ballis-
tic, that is, without collisions; the motion of the electrons has to be treated
quantum-mechanically. The wave function of a free electron is either a plane wave,
∝ exp[ik · r], with a definite phase determined by the optical path of this wave,
or a Bloch function, which also possesses a definite phase (related to the lattice
momentum). Upon an elastic scattering event, the electron retains its quantum
coherence. One has to solve the Schrödinger equation in the presence of the scat-
terers to obtain the wave function in the entire space. The phase of this wave
function is determined by the sum of the optical paths (the products of the wave
number with the distances) accumulated between the collisions, augmented by co-
herent contributions from the scattering); as a result, the motion of the electron is
affected by quantum interference. In contrast, the collisions with phonons, or
with other electrons, are generally inelastic, and thus affect also the surround-
ing of the electron, for instance the phonon gas. Consequently, the phase of the
wave function is modified, and the electron loses its coherence. The mean-free-path
that corresponds to the typical distance between such events is denoted by Lφ (the
subscript φ indicates the phase), it is of the order of the mean-free-path in-between
inelastic collisions, and is growing as the temperature is reduced. At low tem-
peratures, and in very clean samples, Lφ can exceed by several orders of magnitude
e . In samples larger than Lφ the electrons lose the quantum coherence, and then
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 588

588 Introduction to Solid-State Physics

their motion may be described by the semiclassical formalism, like in Sec. 6.10.
The term mesoscopic physics refers to the intermediate region, e  L  Lφ ,
where there appear deviations away from the semiclassical picture.
When a three-dimensional sample is significantly thicker than Lφ along two
spatial directions but is in the mesoscopic regime along the third one, the energy
levels of Eq. (6.27) are discrete continua along the first two directions, and are
discrete along the third one. For a thin enough third dimension, the spacing between
the discrete levels, which is proportional to L−2 in Eq. (6.27), is larger than kB T ;
the Boltzmann probability to occupy these levels is quite small. Hence these levels,
except the lowest one, can be ignored and the system is described by the two-
dimensional spectrum. Slightly larger thicknesses imply that several discrete levels,
that represent the perpendicular coordinate, have to be taken into account. Similar
considerations apply to a system which is very long along one direction, but finite
along the other two. Such a system is described by discrete continuum along the first
direction, and a finite number of discrete levels along the perpendicular directions.
This is the appropriate description of the energy levels of electrons in the nanowires
of Fig. 1.8, or in the nanotubes of Fig. 1.9(a). (In the first case there are free
surfaces along the narrow direction; in the second, the nano stripe is rolled over to
form a cylinder, and so the states along the direction perpendicular to the axis of
the tube obey periodic boundary conditions, see problems 7.6 and s.7.4.)

Problem 7.6.
A two-dimensional electronic system is described by a square lattice as shown in
Fig. 7.7(a). The energy bands, derived within the tight-binding approximation, are
given in Eq. (6.119).
a. The planar sheet is folded to form a cylinder, such that the point at the end
of the vector L1 coincides with the origin, and the dashed lines are congruent to
one another. The pentagonal prism in Fig. 7.7(b) displays the cylinder formed for
L1 = 5a1 . Find the unit cell and the energy spectrum when this lattice is viewed as
a one-dimensional lattice.
b. Find the modifications in the answers to part (a) when L1 is replaced by L2 =
2a1 + a2 [the cylinder in Fig. 7.7(b)] and the dotted lines are congruent to one
another.

Several phenomena based on the coherent motion of electrons are mentioned

in Chapter 6. The Quantum Hall effect, which determines the quantum of the
electrical resistance h/e2 , results from a coherent ballistic motion of electrons along
one-dimensional surfaces of a two-dimensional sample. As seen in Sec. 6.11, the
absence of scattering is due to the spatial separation between the surfaces,
 because
the width of the sample is longer than the magnetic length, B = c/(eB) (B
is the magnitude of the magnetic field). Due to this separation, the scattering of
an electron that moves to the right on the lower surface in Fig. 6.35 to a state
moving to the left on the opposite surface can be ignored. The fact that both
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 589

Selected topics 589

(a) (b)

Fig. 7.7

expressions, for the quantum of the resistance and for the magnetic length, contain
the Planck constant h (or ) marks these phenomena as due to quantum origins.
The same is true for quite a number of mesoscopic phenomena. For instance, the
Aharonov-Bohm effect, surveyed in Appendix E and the related phenomenon
of persistent currents flowing around rings under the effect of magnetic fields.
Both effects stem from the quantization of the magnetic flux in units of the flux
quantum Φ0 = hc/e, and both are detected in experiments carried on mesoscopic
rings in which the coherence of the wave functions is maintained.
Discrete energy levels, quantum dots. Another length scale in the elec-
tronic motion is the wave length of de Broglie at the Fermi energy [see the discussion
preceding Eq. (6.28)], λF = 2π/kF = h/pF = h/(m∗ vF ). In semiconductors like
GaAs, the effective mass and the density of the conduction electrons are small; then
λF ≈ 1000 Å=0.1 μm. The energy levels of an electron [Eq. (6.27)] in a sample
of size L form a discrete continuum provided that L  λF . But when L is small,
the spacings between the energy levels increase; once they exceed kB T , the phys-
ical properties depend on details of the spectrum, and not only on the density of
states (a concept suitable for the description of a continuum or of a discrete con-
tinuum of levels). A system whose dimensions are all mesoscopic [for instance, a
small cube (or a square in two dimensions), or the Bucky ball in Fig. 1.9(b)] is a
zero-dimensional system, or a quantum dot. Because the energy levels of the
electron in such a system are discrete, it is sometimes called an artificial atom.
The reason for this is related to the Pauli principle according to which each dis-
crete state can accommodate at most two electrons, and to the degeneracy of the
energy levels. For instance, the energy levels of a spherically-symmetric potential
are characterized by the quantum numbers {n m}, but the energy is independent
of m [the degeneracy is (2 + 1)]. The energy levels in an atom depend only weakly
on , leading to the structure of the periodic table, where shells of 2n2 states are
gradually filled. The single-electron energy levels in each shell are close to one an-
other, while the spacing between different shells is larger. Other symmetries of the
potential give rise to different degeneracies, but the gradual filling of the discrete
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 590

590 Introduction to Solid-State Physics

levels with electrons enables the construction of a periodic table corresponding to

each quantum dot, see Fig. 7.8.

(a)

(b)

Fig. 7.8: (a) The current through a two-dimensional circular quantum dot as a func-
tion of the gate voltage, and the increment of the electrons’ energy as a function of
the number of electrons (in the inset). (b) The electronic shells in a two-dimensional
harmonic potential (upper part) and the corresponding periodic table (lower part).
[L. Kouwenhoven and C. Marcus, Quantum Dots, Physics World, June 1998, p. 35.]

Coulomb blockade. What is the energy of N electrons occupying the discrete

energy levels 1 ≤ 2 ≤ 3 ≤ . . . ≤ N in a quantum dot? Suppose that an electrode,
whose voltage is Vg , is placed near the quantum dot. This voltage is termed gate
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 591

Selected topics 591

voltage. It gives the energy −eVg to each electron on the dot. For N  1,
the Coulombic repulsion energy of the N electrons is (N e)2 /(2C) where C is the
capacitance of the dot. This capacitance is proportional to the linear size of the
quantum dot. The total energy of the N electrons the quantum dot is

(N e)2 N
E(N ) = − N eVg + n . (7.7)
2C n=1

An electron added to the quantum dot increases the total energy by

E(N + 1) − E(N ) = (2N + 1)e2 /(2C) − eVg + N +1 .

When the quantum dot is coupled to a bath of electrons whose chemical potential
is μ, an extra electron can be added to the dot provided that μ = E(N + 1) − E(N ),
that is

Vg(N +1) = (2N + 1)e/(2C) + (N +1 − μ)/e . (7.8)

where μ is the mean chemical potential of the reservoirs. Hence when Vg (or μ)
is varied, the number of electrons residing on the dot does not change until one
of the discrete values of this equality is reached. The fact that an electron cannot
be added to the dot at other values is called Coulomb blockade: the Coulomb
repulsion prevents the addition of another electron. This device is also referred to
as a single-electron transistor, since it enables the passage of electrons one by
one.
The energy spectrum of a quantum dot can be measured by connecting it to two
electronic reservoirs, whose chemical potentials differ from one another only slightly.
At low temperatures a current can flow in-between the two reservoirs only when an
electron is able to pass through the quantum dot; this happens at certain discrete
value of the gate voltage, i.e., when Eq. (7.8) is obeyed. Figure 7.8(a) shows the
results of such experiments. The measured system was a circular planar quantum
dot. The current displays narrow peaks, at discrete values of the gate voltage, and
it almost vanishes everywhere else. Equation (7.7) implies that the spacing between
two consecutive peaks is

Vg(N +1) − Vg(N ) = e/C + (N +1 − N )/e .

The figure shows that this spacing is almost a constant for the fourth, fifth, and
sixth electrons, i.e., the energy levels of these electrons are quite close to each
other. Similarly, the spacings are almost constant for the 7 − 12 electrons. Hence,
the ground state is doubly degenerate, the second level is four times degenerate,
the degeneracy of the third one is 6, and so on. When each such level represents
an electronic “shell”, the result is the filling of shells and the “periodic table” as
described in Fig. 7.3(b). Once a shell is completely full, the spacing between it and
the next peak is larger, due to the addition of the difference of energies between the
shells.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 592

592 Introduction to Solid-State Physics

It has been noted that the “magic” numbers 2, 6, 12, . . . are obtained also when
the potential confining the motion of the electrons to the quantum dot is harmonic,
mω 2 ρ2 /2, rather than spherical. (ρ is the radius of the motion in the plane.) The
solution of the Schrödinger equation in problem (E.2.) yields the energy levels
En = ω(n + 1), with (n + 1) degeneracy (ignoring the spin, the ground-state level
is not degenerate, the second level corresponds to the angular momentum M = ±1
and hence is doubly degenerate, in the third level M = 0, ±2, and so on). In this
potential the nth level can be occupied by 2(n + 1) electrons (including the spin
degree of freedom), and thus the first level is filled after 2 electrons, the second shell
is filled after 6 electrons, the third shell is filled after 2 + 4 + 6 = 12 electrons, and
so on, as is indeed found experimentally. Hence, the harmonic potential describes
quite faithfully the confining potential of the electrons in the circular quantum dot.
Indeed, the harmonic potential, which follows from a Taylor expansion around the
minimum of any potential, gives a good approximation for the lower energy levels.
Problem 7.7.
The single-electron levels in a quantum dot are 1 < 2 < . . . < n < . . .; the
degeneracy of the nth level is gn . The repulsion energy [see for example Eq. (7.7)]
between each pair of electrons residing in the nth level is denoted Un , and that of
electrons in the nth and the mth levels is Un,m . The repulsion energies are assumed
to be much smaller than the spacings between the single-particle energies.
a. Find the total energy of N electrons on this dot, and the differences between the
values of the gate potential at which an additional electron can enter the dot.
b. Discuss the result of part (a) for g = 1 for all ’s. What is the result when
U,n = U ? Compare it with Eq. (7.7). Is the approximation in that equation valid?
Qubits and quantum computation. The mode of operation of a “classical”
computer is based on bits which can assume one of two values, e.g., 0 or 1. In recent
years there is a vast interest in quantum computing, based on qubits. The qubit
is a superposition of two basis states, with complex coefficients. This superposition
can be mapped on others by quantum operations; in this way the two complex
coefficients are replaced by another pair of complex coefficients. This is the principle
of quantum computing. Since a complex number contains much more information
than the classical bit, it is expected that quantum computers will be much faster as
compared to the classical one. An example of a qubit is the spinor that describes the
state of the spin of an electron. The spinor is a superposition of the two basis states
in which a specific component of the spin angular momentum is ±/2. A possible
realization of a quantum computer is built of a network of quantum dots, such that
each dot contains a single electron. The application of a magnetic field causes the
spin magnetic-moment to point along a desired direction, and thus to determine its
spinor, that is, the qubit it represents. Magnetic interactions among the electrons
on neighboring dots enable in principle complicated quantum computations. There
are other proposals for quantum computers based on quantum systems with two
basis states.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 593

Selected topics 593

Two-dimensional electron gas and quantum wires. A typical mesoscopic

system is described in Sec. 6.11. The quantum Hall effect is usually measured on
such a system: a thin layer of GaAs, prepared by an epitaxial growth (Sec. 2.9).
When the layer is thin enough, the spacings between the energy levels corresponding
to the motion along the normal to the film are larger than kB T , and then at low
enough temperatures the motion along that direction is at the ground state of that
direction, and the electrons move only in the plane of the film. The electron gas
there is called a two-dimensional electron gas, abbreviated as 2DEG. This type
of thin films is used in many experiments in the mesoscopic regime. The thin film
described in Sec. 6.11 is covered by a relatively thick layer of GaAlAs, which is an
insulator. Metallic electrodes attached to this layer create in the thin film regions of
negative potential, which prevents the electrons from moving there. In this way the
rectangular region that serves for the quantum Hall effect measurements is formed.
The energy levels of an electron in the two-dimensional film are

E(n, k) = [2 /(2m∗ )](kx2 + ky2 ) + E n ,

where k is the wave vector in the plane (that belongs to the discrete continuum
spectrum) and E n , n = 0, 1, 2, 3, . . ., are the energy levels of the one-dimensional
potential confining the motion of the electrons along the ẑ−axis (e.g., a rectangular
potential well, or an harmonic oscillator). The first term represents the dense
levels; for each value of n their density of states (including the spin degeneracy) is
g (2d) = m∗ /(π2 ) [problem 6.10 and the discussion following Eq. (6.169)]. When
the energy E n is not degenerate, the total density of the states is
∞

g(E) = g (2d) Θ(E − E n ) . (7.9)
n=0

Since Θ(x) is the step function, there appear discrete steps in the density of states
as a function of the energy. At low temperatures, an increase in the Fermi energy
(for instance, by deepening the gate voltage that attracts the electrons from the
nearby bath to the wire, or by changing the Fermi energy in this bath as to add
more electrons into the wire) entails jumps in the density of states, and consequently
also in the electronic specific heat [Eq. (6.129)].
Quantum point-contacts. Gate voltages also serve to confine the electrons’
motion to a narrow stripe (whose width is of the order of the electron’s wave length,
λF ). In this case the spacing between the energy levels that correspond to the
motion perpendicular to the stripe can be large, making the motion of the electrons
one-dimensional, like in a quantum wire. Such a wire, that connects two wider
regions as in Fig. 7.9(a), is called a quantum point-contact, QPC. When its
width W is small, W < e < Lφ , the electronic motion along the normal to the wire
is ballistic (with no collisions), and is described in terms of discrete energy levels.
The conductance of this device as a function of the gate voltage is depicted in Fig.
7.9(b). When the gate voltage is very negative, it blocks completely the motion of
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 594

594 Introduction to Solid-State Physics

the electrons, and then the effective width vanishes, W = 0. As the gate voltage
is reduced, the effective width W gradually increases, and current can flow. The
measured conductance does not follow Ohm’s law, or the Drude’s result, Eq. (6.7)
with A ⇒ W : the conductance is not linear in the width. Rather, it has a staircase
structure, with steps of 2e2 /h, the same quantum conductance unit that appears
in the quantum Hall effect (the factor 2 is due to the spin as the energy levels are
degenerate in the absence of a magnetic field). Similar experiments carried out in
the presence of a magnetic field show that each step is split into two. Experimental
results like those displayed in Fig. 7.9(b) can serve to deduce the exact value of
the quantum of the conductance, and thence the fine-structure constant (see Sec.
6.11).

(a) (b)

Fig. 7.9: (a) A quantum point-contact. The two-dimensional electron gas 2DEG is
in the mid layer, in-between two insulating layers. Metallic electrodes, i.e., gates,
are placed on the upper layer. These exert negative electric potential on the electron
gas and create a narrow passage, i.e., a quantum wire, between its two parts, QPC.
A small potential drop between the two contacts at the two edges induces current
and enables the measurement of the electrical conductance of this device. (b) The
conductance of the device as a function of the gate voltage. [B. J. van Wees, H.
van Houten, C. W. J. Beenakker, J. G. Williamson, L. P. Kouwenhoven, D. van
der Marel, and C. T. Foxon, Quantized conductance of point contacts in a two-
dimensional electron gas, Phys. Rev. Lett. 60, 848 (1988).]

A simple model that describes this phenomenon envisages the wire as a rectan-
gular stripe of width W and length L (W  L < e < Lφ ). The spacings between
the levels E ny decrease as W is increased. For example, when the potential well
is very deep, the transverse wave functions are standing waves, with wave length
λ = 2W/ny and energy E ny ≈ 2 (ny π/W )2 /(2m∗ ) (check!). Assuming that the mo-
tion normal to the plane is in its ground state (as the first excited level of that motion
is significantly larger than kB T ) and that the Fermi energy is EF ≈ 2 kF2 /(2m∗ ),
all the states for which ny π/W ≤ kF are occupied at low temperatures. Each such
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 595

Selected topics 595

transverse state represents a one-dimensional energy band (corresponding to the

motion along the passage); N = W kF /π = 2W/λF such bands are occupied at low
temperatures. When a potential drop V is applied between the two wider regions
of the electron gas, the respective Fermi energies differ by δμ = eV . The density
of electrons that participate in the transport (moving from the higher potential to
the lower one) is half that of total one, i.e., ρ(1d) (E)δμ/2 = 2eV /(hv) [using the re-
sults of part (b) of problem 6.10 for one dimension, ρ(1d) (E) = n(dk/dE) = 4/(hv)
where v is the velocity of the electrons; see also Eq. (6.125) for a free particle]. Each
electron carries a charge e. Multiplying the density by the velocity, the current is
j1 = 2e2 V /h. The velocity v is cancelled, and hence this contribution to the current
does not depend on the band’s index. It follows that the current is multiplied by
N and the conductance is

G = N j1 /V = N (2e2 /h) . (7.10)

Increasing gradually the width increases in turn N , and adds additional steps to the
conductance, as shown in Fig. 7.9(b). The quantum effects become less important
when N  1 and the conductance is large.
Scattering of electrons and the Landauer formula. Equation (7.9) is
based on the assumption that the quantum wire is a perfect conductor, in which the
electrons do not undergo collisions. That is, any electron that enters the wire leaves
it with the same wave function. When the wire contains scatterers, an electron can
traverse it with a probability T (T ≤ 1), called the transmission coefficient, or it
can be reflected with a probability R (R ≤ 1), termed the reflection coefficient.
Charge conservation dictates that R + T = 1. A simple one-dimensional system
in which these coefficients are used is presented in Sec. 5.6: the elastic scattering
of a one-dimensional wave of a phonon (i.e., acoustic wave). The wave functions
on the two sides of the scatterer are displayed in Eqs. (5.87) and (5.88): the
impinging wave, A exp[i(kna − ωt)], the reflected one, rA exp[i(−kna − ωt)], with
the complex reflection amplitude r, and the transmitted wave, tA exp[i(kna − ωt)],
with the transmission amplitude t (all for n ≥ 0). The explicit calculation yields
R = |r|2 and T = |t|2 , and confirms the conservation law R + T = 1 (check!). A
similar calculation for A exp[i(−kna − ωt)] for n ≥ 0 gives the transmitted wave
t A exp[i(−kna−ωt)] and the reflected one r A exp[i(kna−ωt)] with R = |r |2 = |r|2
and T = |t |2 = |t|2 . Problem 7.8 presents a similar solution for the scattering of
an electron from a point-like scatterer on a one-dimensional lattice, within the
tight-binding approximation. Since the probability of the electron to traverse the
scatterer, at zero temperature, is T , the ideal conductance 2e2 /h is multiplied by T .
At zero temperature, when the the Fermi energies of the electrons’ reservoirs on the
two sides differ only slightly, the only electrons that pass the sample are those at
the (common) Fermi energy. The Landauer formula states that the conductance
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 596

596 Introduction to Solid-State Physics

of a wire connected to two leads, each with N transversal channels, is

2e2 
N
G= Tαβ , (7.11)
h
α,β=1

where Tαβ = |tαβ |2 is the transmission coefficient from channel α in one wire to
channel β in the other wire, calculated at the Fermi energy. For a perfect transmis-
sion, Tαβ = δαβ Eq. (7.9) is retrieved. The Landauer formula (and its extension
to finite temperatures) is the main tool to analyze the conductance of mesoscopic
systems.

Problem 7.8.
An electron of energy E = EF , moves on a one-dimensional chain, modeled in
the tight-binding approximation, with the Hamiltonian defined in problem 6.25(b),
except for

0|Ĥ|0 = 1|Ĥ|1 = E  ,
0|Ĥ|1 = 1|Ĥ|0∗ = −γ  .
(0)
Energies are measured relative to E = 0.

Fig. 7.10

a. Show that the wave functions that appear in Eqs. (5.87) and (5.88) solve the
tight-binding equations of this example, find the transmission and the reflection
coefficients, and verify that they obey charge conservation, T + R = 1.
b. Repeat part (a) for an electron impinging from the opposite direction, and verify
that the resulting transmission and reflection coefficients do not change.
c . Assume that the site 0 and 1 are also connected to a third atom (or a quantum
dot), A (see Fig. 7.10),

A|Ĥ|A = A ,
0|Ĥ|A = A|Ĥ|1 = −γA e−iφ/2 .

Assume also that E  = 0 and that γ  = γ0 is real (it results from the overlap of
real wave functions). The phase φ represents the magnetic flux, in units of the flux
quantum, through the triangle 1A0 due to the Aharonov-Bohm effect, Eq. (E.17).
Calculate the transmission coefficient of this system, and show that T (φ) = T (−φ)
and that the transmission coefficient is periodic in φ. Determine the period.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 597

Selected topics 597

Weak localization. The derivation of the Drude expression for the conduc-
tivity, Eqs. (6.6) or (6.45), assumes that successive collisions of the electrons with
the scatterers are not correlated. In other words, repeated scattering events by the
same scatterer are ignored (as it is assumed that such events are rare). However,
this approximation is not always valid. Consider the probability that an electron
reaches a point B from a point A; both points are located within the same sample.
There are various paths connecting A with B, as illustrated in Fig. 7.11(a). In the
quantum-mechanical description, each such path has a finite “width” of the order of
the Fermi wave length λF = 2π/kF ≈ a. At sufficiently low temperatures one may
ignore the inelastic scattering events, and assume that λF , e  L < Lφ . To each of
the n paths between A and B there corresponds an amplitude, An = |An | exp[iφn ],
where the phase φn is determined by the optical path-length and the elastic scat-
tering along the path. The total probability P to reach point B from point A is the
absolute value squared of the total amplitude, i.e.,
  
P = |A|2 = | An | 2 = |An |2 + 2 |An ||Am | cos(φn − φm ) . (7.12)
n n n<m

The first term on the right-hand side represents the probabilities of the independent
paths, that is, the classical probability; the second describes quantum interference
among the paths. This term is ignored in the Drude’s formulation; the argument is
that various paths possess different phases (e.g., due to their various lengths). As
a result the sum in the second term comprises positive and negative contributions
that make the sum far smaller than the one in the first term. There are, though,
paths for which this argument breaks down. These are the ones in which the two
points A and B coincide, Fig. 7.11(b). The electron can move along the path
A → C → D → E → B, (with amplitude A1 = |A1 | exp[iφ1 ]) or via the reverse
path, A → E → D → C → B, (with amplitude A2 = |A2 | exp[iφ2 ]). The second
path is obtained from the first one by the operation of time reversal. Assuming
that this operation does not change the Hamiltonian of the system then the elec-
tron traverses the same optical path and experiences the same scattering events.
It follows that the probability of the electron to come back to its original starting
point is P0 = |A1 + A2 |2 = 4|A1 |2 , twice the classical probability for the same
process, Pcl = |A1 |2 + |A2 |2 = 2|A1 |2 . Hence, the quantum-mechanical diffusion is
slower than the classical one, due to constructive interference of paths that return to
their origin. This interference causes a reduction in the electrical conductance. The
electron tends to remain at the vicinity of the origin point, a phenomenon termed
“weak localization”. Weak localization, observed at relatively small concentra-
tion of scatterers, is a weaker version of the more dramatic phenomenon of strong
localization, Sec. 6.12.
According to the (classical) Matthiessen rule [Eq. (6.49) and Fig. (6.4)], the
resistance of a metal increases as the temperature is raised. This relation, though,
is derived for macroscopic systems by ignoring quantum interference effects. How-
ever, as the temperature is decreased, Lφ increases. At distances smaller than Lφ
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 598

598 Introduction to Solid-State Physics

the weak localization phenomenon takes place, and the resistance increases as the
temperature is lowered, in contrast with the Matthiessen rule.

(a) (b)

Fig. 7.11: (a) Three different paths between the points A and B. (b) The two points
coincide and the path returns to its origin, A=B.

The increase in the probability to return to the origin stems form the constructive
interference of the paths, which, in turn, requires that the Hamiltonian is symmet-
ric with respect to the operation of time reversal. Magnetic fields break
this symmetry; indeed, when a magnetic field is applied to the system, say normal to
the plane in Fig. 7.11, and when the Zeeman interaction with the electron’s spin is
ignored, the effect of the field is to induce an additional phase on the amplitude due
to the Aharonov-Bohm effect. This phase, φ [which is proportional to the magnetic
flux through the loop, Eq. (E.17)] is added to the phase of one of the paths, and is
subtracted from the phase of the other. The probability to return to the origin is
then P0 = |A1 exp[iφ] + A2 exp[−iφ]|2 = 2|A1 |2 [1 + cos(2φ)]. Consequently, the elec-
trical conductance is periodic in the magnetic flux. The experimental observation
of these periodic oscillations formed the proof for the existence of the phenomenon
of weak localization. For a single path, the contribution to the conductance is
in the range {0, 4|A1 |2 }, which implies that the oscillations of the conductance
are of the same order of magnitude as the conductance itself. This is one of the
sources of the universal conductance fluctuations discussed below. This is also
the order of magnitude of the disparity in the conductances measured in the pres-
ence of a magnetic field on different samples with the same mean concentration of
impurities (the impurity configurations leading to the closed paths are different in
each sample). As the measurement averages over many closed loops that contain
various magnetic fluxes, the effect of the weak localization is weakened.
The spin-orbit interaction, which couples the motion of the electron with its
spin state, is significant in certain semiconductors. In such materials, the interfer-
ence of the electronic paths is modified. The reason is that due to the spin-orbit
interaction, the angular momentum of the orbital motion of the electron induces an
effective magnetic field (somewhat similar to the magnetic field created by an elec-
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 599

Selected topics 599

tric current in a loop or in a solenoid), which operates on the spin of the electron.
Then, when the electron moves in that effective magnetic field, its spin direction
rotates. It turns out that the wave function of the spin reverses its sign when the
electron completes a full cycle around a closed path. This implies that the sign of
the interference term is different for the two paths in Fig. 7.11(b) when the spins
rotate in opposite directions, leading to destructive interference which reduces to
probability to return to the origin. As a result, the conductance increases (as com-
pared to its reduced value due to the weak-localization phenomenon). This effect
is termed weak anti-localization.
Self-averaging. As mentioned, in samples whose linear size is larger than Lφ ,
the electrons lose their quantum coherence, and then their motion can be described
within the semiclassical theory. In particular, the conductance of a one-dimensional
wire longer than Lφ obeys Ohm’s law. Consider a wire whose length obeys L  Lφ ,
and divide it into N = L/Lφ segments, each of length Lφ . Because of the scatterers
that are randomly located in the wire, the resistances of these segments, {rn , n =
1, . . . , N } are distributed at random. The resistance of the wire is the sum of these
N
resistances, R = . Denote by r = rn  the average resistance of a single
1 rn
segment, and by Δr = (rn − r)2  its standard deviation. Since the electronic
wave functions in the different segments lack quantum coherence, the resistances of
the various segments are not correlated and therefore (rn −r)(rm −r) = (Δr)2 δnm .
When the total resistance R is averaged over an ensemble of wires, all with the same
mean density of scatterers, then

R = N r , (ΔR)2 = (R − R)2  = (rn − r)(rm − r) = N (Δr)2 ,
n,m

leading to
ΔR Δr 1
= √ .
R r N
This is the statistical law of large numbers. Note the similar result obtained for
a random walker, Eq. (7.1). In both cases the law of large numbers follows from
the addition of many independent, un-correlated, quantities. The mean-square
deviation expresses the typical disparity of two samples: though the microscopic
compositions are different, the relative difference of the resistances diminishes as the
sizes of the samples increase. A large enough sample comprises a large number of
the “basis” resistors (of length Lφ ), and consequently their sum implies an averaging
over almost all possible values. This is the meaning of self averaging. In fact,
other extensive variables, obtained by summations over contributions from various
parts of the sample (which are not correlated), are expected to behave similarly.
In dimensions d higher than 1, a sample whose linear size is L is divided into
N = (L/Lφ )d cubes, each of volume Ldφ . In the absence of any correlations among
√
the cubes, the mean-square deviation of the total resistance is proportional to N ∼
(L/Lφ )d/2 , and again the mechanism of self averaging takes place. This argument
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 600

600 Introduction to Solid-State Physics

yields

ΔG ΔG0  Lφ d/2
= ,
G G0  L

where ΔG0 /G0  is the relative mean-square deviation of the conductance of a cube
of linear size Lφ .
Self averaging is realized in any system that possesses a basic length scale L0 ,
such that there are no correlations between the physical properties of cubes of
volume Ld0 . When the linear size of the system is such that L  L0 , the relative
disparities between various samples are proportional to (L/L0 )−d/2 , and therefore
their macroscopic properties are rather similar though microscopically they are quite
different. For instance, macroscopic samples of a glass made of the same chemical
material have usually the same specific heat, though the microscopic configurations
of the atoms in them differ.
Universal conductance fluctuations. The self-averaging feature of the con-
ductance pertains to macroscopic samples, for which Lφ  L. Smaller samples,
all with the same mean scatterers’ concentration, have different conductances, since
their scatterers’ configurations differ. The differences in the conductances of samples
of size Lφ are scaled by ΔG0 /G0 . At low temperatures, where the conductance of
a mesoscopic system is determined by the diffusive paths of the electrons, its mea-
sured values fluctuate among samples prepared under identical conditions. Slight
modifications in the Fermi energy (which change the de Broglie wave length of the
electron, and consequently the lengths of the optical paths), the application of weak
magnetic fields (that vary the magnetic fluxes and thus the Aharonov-Bohm phases
in closed loops), or displacing the locations of the scatterers, induce significant
changes in the conductance, because the interference is sensitive to the phases of
the wave functions. Although this disparity affects considerably the quantum in-
terference, and therefore the measured conductance, the standard deviation of the
conductance, that is the average deviation from the average conductance, or the
amplitude of the fluctuations in the conductance as a function of the Fermi energy
or of a magnetic field, are always of the order of the quantum unit of the con-
ductance, e2 /h, independent of the size of the samples, of the spatial dimension,
or of the impurity concentration.
In contrast to the situation in large samples, where self averaging is active, the
properties of various parts of a mesoscopic system of size Lφ are correlated. These
correlations originate in the interference of wave functions corresponding to paths
of the diffusion motion at distances shorter than Lφ . According to the Landauer
formula, Eq. (7.11), the mean conductance (per spin) is

e2  e2   
N N
G=  Tαβ  = N− |rαβ |2  , (7.13)
h h
α,β=1 α,β=1
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 601

Selected topics 601

where the condition of charge conservation


N 
N
|tαβ |2 = N − |rαβ |2 , (7.14)
α,β=1 α,β=1

is used. Here, rαβ is the reflection amplitude of a wave impinging in channel α and
reflected in channel β. The mean-square deviation of the conductance is found from
an estimation of the mean-square deviations of the various terms in Eq. (7.13). It
turns out that the correlations among the reflection coefficients are weaker than
those among the transmissions, as there are less scattering events which return the
electron to the channel where it came from.1 From Eq. (7.13), the square of the
mean deviation is
  N 2  
(ΔG)2 = (e2 /h)2 Δ |rαβ |2 = (e2 /h)2 N 2 |rαβ |4  − |rαβ |2 2 . (7.15)
α,β=1

The reflection amplitude can be written as a sum over the amplitudes for returning
back to the origin, rαβ = m Bm . In this representation
 
|rαβ |2  = Bn∗ Bm  = |Bn |2  .
n,m n

On the other hand

 
|rαβ |4  = Bn∗ Bm Bi∗ Bj  = |Bn |2 |Bi |2 (δnm δij + δnj δmi ) = 2|rαβ |2 2 .
n,m,i,j n,m,i,j

Correlations between different paths are ignored in the last step, as their number
is very small compared to the total number of paths. It follows that
 2
Δ|rαβ |2 = |rαβ |4  − |rαβ |2 2 = |rαβ |2 2 .

It remains to estimate |rαβ |2 . To this end, consider the Drude formula for the
conductivity in two dimensions,
e 2 ne τ e2
σ= = k , (7.16)
m∗ h F e
where the relations ne = kF2 /(2π) (from part (a) of problem 6.10) and τ = e /vF =
∗
e m /(kF ) are used. The dimensionless number kF e specifies the material. A
material with kF e  1 is a conductor, while when this number is of the order of
1 the conductance is of the order of the quantum unit, and one has to take into
account the quantum effects discussed above. When kF e  1 the material ia an
insulator. The condition kF e  1 for metallic behavior is called the Ioffe-Regel
criterion. This condition can be understood qualitatively: the semiclassical picture
is valid as long as the wave packet that represents the electron is narrow enough,
Δk  kF . By the uncertainty relation, Δk ∼ 1/Δr ∼ 1/ e : the semiclassical
picture is relevant when kF e  1 and then material is a conductor.
1 P. A. Lee, Universal conductance fluctuations in disordered metals, Physica A140, 169 (1986).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 602

602 Introduction to Solid-State Physics

According to Ohm’s law, Eq. (7.16) the states that the conductance of a planar
sample of width W and length L is
W e2 π e
=2
G=σ N , (7.17)
L h 2L
where the relation N = W kF /π is used [see Eq. (7.10)]. For spinless electrons, Eq.
(7.11) yields
e2 2
N Tαβ  .
G =
h
Comparing it with Eq. (7.17) results in
π e
|tαβ |2  = Tαβ  =
2LN
This result is inserted in Eq. (7.14), and then
1 π 
|rαβ |2  = 1− e .
N 2L
It follows from Eq. (7.15) that when L  e ,
 
N 2  2
Δ |rαβ |2 = N 2 Δ|rαβ |2 = N 2 |rαβ |2 2 ≈ 1 .
α,β=1

Hence ΔG ≈ e /h, as indeed observed in experiments. Exact calculations replace

2

the 1 with a number of order 1, which depends on the geometry of the sample.
Thouless energy. The wave function of an electron moving on the nth diffusive
path [Fig. 7.11 and Eq. (7.12)] is multiplied by the factor exp[−iEn t/], where En
is the energy of the moving electron. (This time dependence results from the time-
dependent part of the Schrödinger equation.) In case the electron’s energies on two
paths differ slightly from each other, say by δE, then after time t1 a phase difference
δEt1 / is accumulated between the two paths. Once this phase difference is of the
order of 1, the constructive interference is lost. During the time t1 ≈ /δE, the
electron accomplishes randomly N = t1 /τ diffusion
√ steps,
 and therefore according
√
to Eq. (7.5) it traverses the distance L1 ∼ e N ∼ 2 t /τ ∼
e 1 Dt 1 , where the
diffusion coefficient corresponds to the Fermi energy, D = D(EF ) = vF2 τ /d =
vF /d. It follows that once the energy difference is such that δE ≈ D/L21 , phase
coherence (and consequently, the constructive interference) is lost. This typical
energy, for a sample of length L, is called the Thouless energy, ET = D/L2 .
The time spent by an electron diffusing from one side of the sample to the other,
τT = L2 /D, is the Thouless time; it is related to the Thouless energy by the
uncertainty relation, ET τT ∼ . The destruction of the constructive interference
causes a spatial decay of the wave function and consequently a reduction in the
conductance. In fact, the ratio of the Thouless energy to the spacing between the
energy levels around the Fermi energy conveys information about the character
of these wave functions, whether extended (and then the conductance is high) or
localized (which corresponds to a poor conductance).
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 603

Selected topics 603

Using the expression for the densities of states (problem 6.10),

dne dn
g(EF ) = = ∗ e2 ,
2EF m vF
in the expression for the Drude conductance [for instance, Eq. (6.161)] results in
the Einstein relation
e 2 ne τ
σ= = e2 g(EF )D(EF ) . (7.18)
m∗
In d dimensions, the conductance is [using Eq. (7.18)]
e 2 ET
G = σLd−2 = , (7.19)
 δ
where δ = 1/(g(EF )Ld ) is the mean spacing between energy levels around the Fermi
energy. When ET  δ the energy levels around the Fermi energy are very dense;
then the conductance is much higher than the quantum unit. On the other hand,
when ET  δ, the spacing between the levels is large, which usually signals that
they are localized at certain regions in the system, and are not extended all over it.
Indeed then the conductance is far smaller than the quantum unit.

Problem 7.9.
The wave function of an electron is substantial within a “tube” of width λF , the wave
length at the Fermi momentum (why?), and therefore the volume (in d dimensions)
it occupies when the electron moves a time dt is of the order of λd−1F vF dt. On the
other hand, during a time t, a diffusive electron reaches the distance (Dt)1/2 , and
hence it covers a volume of the order of (Dt)d/2 .
a. Explain why the probability of the path to return to its original point is
 τx
λd−1 vF
dt F d/2 ,
τ (Dt)
where τx = L2x /D. What is the maximal length Lx , for Lφ  L and for Lφ  L?
Compute explicitly the integral, and discuss the result as a function of the size of
the system, L.
b. Assume that the probability found in part (a) is small. Explain why in this
limit the relative correction to the conductivity, Δσ/σ, due to weak localization
effects is proportional to this probability. Estimate this correction in one, two, and
three dimensions. What can be deduced about the conductivity of metals at low
temperatures in arbitrary dimensions?

Artificial lattices. The technological developments of recents years enable the

construction of artificial latices. For example, Fig. 7.12 displays periodic arrays
of quantum dots, whose lattice constants are of the order of hundreds of nanome-
ters. Figure 7.13 shows a network of nano-scale transistors; these are manufactured
commercially by Intel, for future use in quantum computers.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 604

604 Introduction to Solid-State Physics

Fig. 7.12: Periodic arrays of quantum dots. Taken from http://www.rciqe.

hokudai.ac.jp/movpe/en/k-naiyou/index.html.

Fig. 7.13: Wafer of transistors whose size is tens of nanometers. Taken from http:
//qbnets.wordpress.com/2011/05/07/intel-waffles.

Another branch of current research of periodic structures is discussed in problem

s.6.6, part (c). Figure 6.49 displays an electric potential shaped as an egg tray. It
is created by the interference of two electromagnetic waves perpendicular to one
another. The interference constructs a two-dimensional standing wave; both the
height of the potential barriers in-between the minima and the maxima and their
locations can be controlled by a suitable choice of the electromagnetic waves (pro-
duced by monochromatic laser radiation). This potential can serve as a substrate
that captures atoms from the gas above it. It is termed an optical lattice.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 605

Selected topics 605

7.4 Answers for the problems in the text

Answer 7.1.
a. The right panel in Fig. 7.1 displays long molecules that are arranged in a liquid
phase, that is, without any apparent order. The average density of this pattern is
homogeneous; this means that on the average the system has continuous symmetry
under translations and under rotations. The correlation function n(0)n(r) is the
one shown in Fig. 1.2(b). The next panel in Fig. 7.1, in the nematic phase, the
director vector has on the average a preferred direction. This breaks the continuous
symmetry with respect to rotations around that direction. In the third panel from
the right, that shows the smectic phase A, the continuous symmetry with respect
to translations is broken as well. A crystalline pattern is formed along the ẑ−
direction: the symmetry with respect to translations along this axis is discrete, i.e.,
n(r + ẑ) = n(r), but is continuous in the XY plane, as in the liquid phase.
In the fourth panel that displays the smectic phase C the molecules are inclined
with respect to N̂, which itself is normal to the planes. As a result, the rotation
symmetry around this axis is broken. The left panel shows the solid phase of the
inclined smectic crystal. It has discrete translational symmetry as the tetragonal
lattice, but lacks the rotational symmetry.
The upper line in Fig. 7.2 shows the smectic phases A (on the left) and C (on
the right). The symmetries of these phases are discussed above. The next line
displays the hexatic phase, which breaks another continuous rotational symmetry,
within the planes. In this phase there is no symmetry with respect to translations
in the XY planes, but the symmetry with respect to rotations by 60◦ around the
ẑ−axis is maintained.
The cholesteric phase is exhibited in Fig. 7.3. In addition to the broken sym-
metries of the smectic phases, here the translational symmetry along the direction
normal to the planes is by a base vector four times longer than the distance between
neighboring planes.
b. The lattice denoted by BC in Fig. 7.2 is a triangular one. A side view reveals
that neighboring planes are displaced relative to each other, such that they are
ordered in the form ABABAB. . . (see Fig. 2.24). This is a hexagonal lattice (Fig.
2.28) as opposed to the HCP lattice shown in Fig. 2.24. The basis vector, normal
to the planes, is longer than the vector that appears in the dense packing of spheres.
The unit cell of this lattice comprises a rhombus in the plane, and the normal to
the plane, of length 2 . The phase E as well has periodicity of two planes along the
ẑ−axis, but within each plane the lattice is rectangular, with two molecules in each
rectangular unit (at the center and at the corner).
The lattices J, G, K, and H have all periodicity of a single plane. J and G
are triangular in the plane and therefore are simple hexagonal lattices. The planes
that build the lattices K and H appear as rectangular; the axes of the rectangle
are inlined by 60◦ relative to the horizontal axis in the figure. These lattices are
orthorhombic.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 606

606 Introduction to Solid-State Physics

Answer 7.2.
N
a. Using r(N ) = n=1 Δrn gives


N 
N 
N
[r(N )]2  = Δrm · Δrn  = a2 = N a2 ,
m=1 n=1 n=1

since Δrn  = 0 and Δrn · Δrm  = a2 δnm .

b. In order to describe a “new” step built of two “old” ones, one first examines the
correlation of two new steps in terms of that of two old ones,

Δrn · Δrm  = Δr2n−1 · Δr2m−1  + Δr2n−1 · Δr2m 

+ Δr2n · Δr2m−1  + Δr2n · Δr2m  .

This correlation is nonzero only when n = m, and then its value is 2a2 ,

[Δrn ]2  = 2a2 .
√
The mean value of the new length is a = 2a. Hence

[r (N  )]2  = (N/2)2a2 = N  (a )2 , with N  = N/2 .

For successive steps one may repeat these arguments to find that
√
a = a,

and

[r (N  )]2  = (N/ ) a2 = N  (a )2 , with N  = N/ .

The functional dependence of the mean distance on the number of steps is not
changed.

Answer 7.3.
a. According to the information given in the problem,


N
r(N ) = Δrn  = N r0 ,
n=1

N
[r(N ) − N r0 ]2  = (Δrn − r0 ) · (Δrm − r0 ) = N a2 ,
n,m=1

and then, with r(N ) = N r0 + [r(N ) − N r0 ], one obtains

[r(N )]2  = (N r0 )2 + N a2 .

b. The probability per unit volume to reach from the origin to the point R = r(2)
after two steps is obtained from the sum of the probabilities to reach the point Δr1
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 607

Selected topics 607

after one step, with the probability per unit volume P (Δr1 , 1); then to do the step
Δr2 with the probability per unit volume P (Δr2 , 1),

P (R, 2) = d3 (Δr1 )P (Δr1 , 1)P (R − Δr1 , 1) .

Similarly, one finds


P (R, N ) = d3 R P (R , N − 1)P (R − R , 1) .

This is a convolution of two probabilities; the Fourier transform is then, according

to Eq. (B.11), the product of the Fourier transforms, and therefore
P(k, N ) = P(k, N − 1)P(k, 1) = · · · = [P(k, 1)]N .
The Fourier transform of the probability of a single step is

1 1
P(k, 1) = 3
d3 Re−ik·R P (R, 1) = e−ik·R  ,
(2π) (2π)3
where the right-hand side is an average over all possible steps. By using a Taylor
expansion, one finds
e−ik·R  = 1 − ik · R − (k · R)2 /2 + O(k 3 )
= exp[−ik · R − [k · (R − R)]2  + O(k 3 )] .
Since the steps are symmetric,

3
[k · (R − R)]2  =  kα (R − R)α kβ (R − R)β  = k 2 a2 /3 .
α,β=1

Carrying out the inverse Fourier transform gives the desired probability,
 
1 ik·R −ik·R N 1 2 2 3
P (R, N ) = 3
d ke e  = d3 keik·(R− R ) e−N k a /6+O(k )
(2π)3 (2π)3
√ √
substitution of k ⇒ q/ N , R − R ⇒ u N gives
 √
1 2 2 3
P (R, N ) = √ d3 qeiq·u e−q a /6+O(q / N ) .
(2π N )3
In the large N −limit the last term in the exponent may be ignored, and then one
obtains the Fourier transform of the Gaussian distribution,

1 2 2 2 2
P (R, N ) = √ d3 qeiq·u e−q a /6 = (2πa2 N/3)−3/2 e−3(R− R ) /(2a N ) .
(2π N )3
The probability to return to the origin, in three dimensions, is

P (0, N ) = (a 2πN/3)−3 .
b. The mean of the squared distance is

[r(N )]  = 2π R2 dR R2 P (R, N ) .
2

The outcome of this integration reproduces the results in part (a).

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 608

608 Introduction to Solid-State Physics

Answer 7.4.
a. In this limit one may use
sin(qr) (qr)2
=1− + ... .
qr 6
One then finds (exploiting the normalization of the correlation function,
d3 rg(r) = 1) the desired result.
b. Denoting qr = y, then
 ∞

Γ(q) ∝ q −D
dyy D−2 sin(y)h[y/(qL)] .
0

As the integral here converges, one deduces that Γ(q) ∝ q −D .

Answer 7.5.
a. The number of monomers that exist in a volume ad around a certain monomer
is ad ρ = ad N/Ld , where L is the linear size of the entire chain. Each of these
monomers “collides” with the given monomer, and hence the total energy of the
collisions equals this number, times the energy of a single collision event, times the
number of monomers,
E = CN (ad N/Ld ) = CN 2 /(L/a)d .
b. From part (b) of problem 7.3, one finds that the probability to find a random walk
whose two edges are at a distance L apart is proportional to exp[−3L2 /(2a2 N )]. The
Boltzmann weight of the collision energy is exp[−E/(kB T )]. The total probability
is hence exp[−F (L)/(kB T )], with the “free energy”
F (L) = 3kB T (L/a)2 /(2N ) + CN 2 /(L/a)d .
The probability is maximal when the free energy as a function of L is minimal,
which indeed happens when L ∝ N 3/(d+2) .

Answer 7.6.
The energy bands of a square lattice, calculated within the tight-binding ap-
proximation that accounts for overlaps of wave functions on nearest-neighbors ions
alone, are given in Eq. (6.119),
En (k) ≈ E n + 4Γn (a)[sin2 (kx a/2) + sin2 (ky a/2)] .
Wrapping the two-dimensional sheet adds periodic boundary conditions around the
circular direction. The cylinder can be described as a one-dimensional lattice along
the axis.
a. The cylinder is a one-dimensional lattice, with a lattice vector a2 . The unit cell
comprises five lattice sites, rm = ma1 , m = 0, 1, 2, 3, 4. The boundary condition
along the lattice vector a1 is ψ(r + L1 ) = ψ(r). The component of the lattice
momentum kx can assume the discrete values kx = 2πn/(5a), with n = 0, 1, 2, 3, 4;
the values of ky are still continuous. Each discrete value of kx represents an energy
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 609

Selected topics 609

band, and thus apparently there are five bands. However, the energies are identical
for n ↔ 5 − n, and therefore in reality there are only three energy bands, where
the bands of n = 1, 2 are doubly degenerate. These bands are illustrated in Fig.
7.14(a).

(a) (b)

Fig. 7.14

b. In this case the unit vector of the one-dimensional lattice is 2a2 − a1 , perpen-
√
dicular to the vector L2 . The unit cell is a square whose edge length is |L2 | = a 5;
it comprises five sites: rm = 0 , a2 , 2a2 , a1 + a2 , a1 + 2a2 . The periodicity along
the normal to the cylinder’s axis implies that ψ(r + L2 ) = ψ(r). In order to use
Eq. (6.119) it is convenient to rotate the axes such that they point along the lattice
vectors L2 and 2a2 − a1 . The rotation is with the angle
φ = arctan[L2y /L2x ] = arctan[1/2] ,
leading to
kx = kx cos(φ) − ky sin(φ) , ky = kx sin(φ) + ky cos(φ) .
√
Using these values in the expression for the energy, with kx = 2πn/( 5a) and
continuous ky yields five different energies, derived by displacements along ky ;
they are shown in Fig. 7.14(b).

Answer 7.7.
a. One begins by filling the ground-state energy, to obtain the energy E(1) =
1 −eVg . Adding more electrons gives the additional energies E(2) = 2(1 −eVg )+U1 ,
E(3) = 3(1 − eVg ) + 3U1 , etc., until
E(g1 ) = g1 (1 − eVg ) + g1 (g1 − 1)U1 /2 .
The next electron occupies the level 2 ; hence
E(g1 + 1) = g1 (1 − eVg ) + 2 − eVg + g1 (g1 − 1)U1 /2 + g1 U1,2 ,
etc. Finally,
E(g1 + g2 ) = g1 (1 − eVg ) + g2 (2 − eVg ) + g1 (g1 − 1)U1 /2
+ g1 g2 U1,2 + g2 (g2 − 1)U2 /2 .
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 610

610 Introduction to Solid-State Physics

Once the n level is filled up, the total energy is


n
E(N ) = [g (n − eVg ) + g (g − 1)U /2] + m(n − eVg ) + m(m − 1)Un /2
=1

n−1 
n−1
+ g g U, + mg U,n , N = g1 + g2 + . . . + gn−1 + m , 0 < m ≤ gn .
, =1 =1

In this situation,

n−1
E(N + 1) − E(N ) = n − eVg + mUn + g U,n .
=1

b. For non-degenerate energy levels g = 1. Then m = 0, and hence


N
E(N + 1) − E(N ) = N +1 − eVg + U,n .
=1

As seen, the assumption U,n = U leads to E(N + 1) − E(N ) = N +1 − eVg + N U .

The same result is obtained from Eq. (7.7) when the first term there is replaced by
N (N − 1)U/2. The approximation in Eq. (7.7) is valid as long as N  1.

Answer 7.8.
a. The tight-binding equations for this situation are
Eψ(n) = −γ[ψ(n − 1) + ψ(n + 1)] , n = 0, 1, (7.20)
  ∗  
(E − E )ψ(1) = −(γ ) ψ(0) − γψ(2) , (E − E )ψ(0) = −γ ψ(1) − γψ(−1) .
For an electron impinging from the left, the solution is of the form
ψ(n) = A(eikna + re−ikna ) , n ≤ 0 ,
ψ(n) = Ateikna , n ≥ 1 .
These functions solve the first of Eqs. (7.20) with the dispersion relation
E(k) = −2γ cos(ka) .
The other two equations of (7.20) give
2iγ sin(ka)(γ  )∗ e−ika 2iγ sin(ka)(E − E  + γeika )
t=− , 1 + r = .
(E − E  + γeika )2 − |γ  |2 (E − E  + γeika )2 − |γ  |2
When γ = γ  and E  = 0, one finds that r = 0 and t = 1. In all other cases it can
be checked that T + R = |t|2 + |r|2 = 1.
b. For an electron impinging from the opposite direction, the solution is of the form
ψ(n) = A(e−ikna + r eikna ) , n ≥ 1 , ψ(n) = At e−ikna , n ≤ 0 .
Using the equations in part (a) give
2iγ sin(ka)γ  e−ika 2iγ sin(ka)(E − E  + γeika )
t = − , 1 + r  2ika
e = .
(E − E  + γeika )2 − |γ  |2 (E − E  + γeika )2 − |γ  |2
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 611

Selected topics 611

As seen, |r |2 = |r|2 and |t |2 = |t|2 .

c. The tight-binding equations for the sites 0 and A are
(E − E  )ψ(0) = −γ  ψ(1) − γψ(−1) − γA eiφ/2 ψ(A) ,
(E − E  )ψ(1) = −(γ  )∗ ψ(0) − γψ(2) − γA e−iφ/2 ψ(A) ,
(E − A )ψ(A) = −γA [e−iφ/2 ψ(0) + eiφ/2 ψ(1)] . (7.21)
Using the third of Eqs. (7.21) to express ψ(A) in terms of ψ(0) and ψ(1), and
substituting it in the other equations there, reproduce the equations in part (a),
with the replacements
γ  ⇒ γ  − (γA )2 eiφ /(E − A ) , E  ⇒ E  + (γA )2 /(E − A ) .
As seen, these replacements turn the triangle into an effective bond, whose matrix
element includes the interference between the (previous) direct bond and the path
that passes through the point A. The energies on the sites 0 and 1 are modified
as well. One can now use the results obtained in part (a) to find the reflection
and transmission coefficients. In particular, the latter depends on the phase φ only
through the expression
|γ  |2 ⇒|γ0 − (γA )2 eiφ /(E − A )|2
= (γ0 )2 + (γA )4 /(E − A )2 − 2γ0 (γA )2 cos(φ)/(E − A ) ,
where it is assumed that the original γ  (denoted here by γ0 for clarity) is real.
Hence, the transmission coefficient is an even periodic function of the magnetic
flux, T (φ) = T (φ + 2π). It follows from Eq. (E.17) that the transmission coefficient
returns to the same value when Φ ⇒ Φ + Φ0 ; the periodicity is the quantum unit
of the flux, Φ0 = hc/e.

Answer 7.9.
a. The uncertainty in the momentum of an electron along the direction normal to
the direction of the motion is rather small, of the order of the Fermi momentum.
This implies that the uncertainty in the location of the electron along that direction
is of the order of λF , the Fermi wave length. The probability of the electron to
“cover” the volume λd−1F vF dt at time t is that volume, divided by the total volume
d/2
at that time, (Dt) . This probability is summed over the entire motion of the
electron. The minimal relevant time for this sum is τ , i.e., the time for a single
collision to occur. On the other hand, the electron’s wave function remains coherent
as long as the electron does not collide with the surfaces of the sample, or traverses
the length Lφ . Since the motion is diffusive (a random walk) the maximal allowed
time for the coherent motion is
τx = [min(L, Lφ )]2 /D .
Once the integration (i.e., the sum) is carried out, one finds that the probability to
return to the point where the motion started is
λd−1 v
p = d/2F F (τ 1−d/2 − τ 1−d/2 ) .
D (1 − d/2) x
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 612

612 Introduction to Solid-State Physics

In two dimensions, d = 2, the result is (λF vF /D) ln(τx /τ ). When Lφ  L, the

probability p is independent of the system’s size L. In contrast, at low temperatures
Lφ  L; the probability p is dominated by

τx1−d/2 ∝ L2−d ⇒ L , ln[L/ ] , L−1 , for d = 1 , 2 , 3 , respectively .

b. The probability to return to the origin (the point where the motion began)
expresses the loss of conductance. One therefore may argue that Δσ/σ = −p, and
at low temperatures

Δσ ∼ −(e2 /h)L , Δσ ∼ −(e2 /h) ln[L/ ] , Δσ ∼ −(e2 /h)/L ,

in d = 1 , 2 , 3.

7.5 Problems for self-evaluation

s.7.1.
Figure 7.15 illustrates the molecules of a liquid crystal from a side view. Discuss
the order parameter that describes the locations of the centers of the molecules, and
the one that corresponds to the directions of their axes. This phase is referred to as
“chiral smectic”. Explain why, and determine when this term is indeed valid.

Fig. 7.15

s.7.2.
Many of the organic molecules are planar, for instance that of benzene, Fig. 4.2.
Such molecules form liquid crystals that are called “discotic”. Describe several
examples of discotic liquid crystals, that are built of thin (with width much smaller
than the diameter) discs.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 613

Selected topics 613

s.7.3.
A particle “walks” randomly on an infinite lattice, beginning the walk at the origin.
At each time-step it “jumps” from one lattice site to one of the nearest-neighbor
sites, with the probability w = 1/z, where z is the coordination number of the lattice
(i.e., the number of nearest neighbors).
a. Express the probability to reach the site R after N steps, P (R, N ), in terms of
the probability to reach one of the nearest neighbors after (N − 1) steps. Confin-
ing yourselves to one dimension, compare this relation to the ones that appear in
previous chapters of the main text.
b. Derive the continuum limit of the relation found in part (a) (i.e., the limit where
the lattice constants approach zero, while the number of steps is very large). Show
that the Gaussian distribution (problem 7.3) solves this relation.
c. Obtain the equation for the Fourier transform, P(k,  N ), show that P(k,
 N) =
[P(k, 1)] , and find P(k, 1).
N

d. As in problem 7.3, show that for N  1 one may expand P (k, 1) in a Taylor
series, and use this expansion and the inverse Fourier transform to find an approx-
imate expression for P (R, N ). Compare the result with the one found in problem
7.3.

s.7.4.
Figure 1.9(a) shows a nanotube. When such a tube is cut along a line parallel to the
axis of the cylinder, a rectangular piece of graphene is created. Figure 7.16 displays
two vectors on the graphene lattice, L1 and L2 . This lattice is cut along parallel
lines that are normal to each of these vectors at the edges, and is rolled to form
nanotubes, such that the parallel lines are congruous to each other.
a. Determine the unit cell of the one-dimensional lattice that is created in each of
the cases. How many lattice sites are contained in this unit cell?
b. Find an expression for the energy bands within the tight-binding approximation
for the lattice formed by L1 (assume that the bonds of the tight-binding model are
between nearest neighbors).
c. In the case where the nanotube is formed by L2 , explain qualitatively how should
the band structure be calculated, and how many bands are there.

s.7.5.
The transmission coefficient of the ring in Fig. 7.17 is T = 1 − R. The ring is
threaded by a magnetic flux Φ, which is created by a magnetic field perpendicular to
the plane of the ring. The reflection amplitude of an electron impinging on the ring
comes from the following contributions: r0 for the path reflected from the impinging
point (it is assumed to be real), r1 exp[i(δ + φ)] from the path that encircles the
ring once in the clockwise sense and then returns, (the phase δ is due to elastic
scattering by impurities randomly located on the ring, and φ = 2πΦ/Φ0 is called the
Aharonov-Bohm phase), r1 exp[i(δ − φ)] from the path that encircles the ring once
in the anti-clockwise sense and then returns, and so on.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 614

614 Introduction to Solid-State Physics

Fig. 7.16

a. Explain these expressions, and add the expressions for the paths that encircle the
ring more times. Use the results to derive the conductance through the ring, and
show that it is even and periodic in φ. Identify the contributions of the first two
harmonics of the conductance.
b. A famous experiment measured the conductance of a cylinder between two lines
parallel to its axis. Figure 7.17 illustrates the projection of the cylinder along that
axis. Assuming that the experiment averaged over the conductances of many parallel
rings, what should it measure? Determine the leading harmonic of the measured
conductance.
c. Part (c) of problem 7.8 solves for the reflection coefficient of a system in which
a triangle replaces the ring in Fig. 7.17. Show that an expansion of the reflection
coefficient found in problem 7.8 resembles the solution for the reflection from the
ring.

Fig. 7.17

7.6 Answers for the self-evaluation problems

Answer s.7.1. If one accounts only for the centers of mass of the molecules, then
Eq. (7.2) for the order parameter holds along the direction normal to the planes,
with q = 2π/ . An examination of the directions of the director vectors reveals that
they return to themselves only at each second plane; hence the order parameter
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 615

Selected topics 615

of Eq. (7.1) behaves in an antiferromagnetic fashion along N̂, with periodicity 2 .

The projections of these vectors on the planes, however, behave as in the cholesteric
phase, Fig. 7.3, with periodicity 2 . One concludes that this phase is a combination
of a smectic phase with a cholesteric (or a chiral) one.

Answer s.7.2.
The right panel in Fig. 7.18 shows a nematic phase, in which the director is per-
pendicular to the discs. The next panel displays a columnar structure, in which
the discs are arranged periodically in separate columns such that each of them is
a periodic one-dimensional lattice; the columns themselves are scattered randomly
in space. In the third panel the columns are parallel to each other, but arranged at
random in the plane perpendicular to their direction. In the left panel the discs are
perpendicular to the planes which by themselves are arranged periodically. Within
each plane the discs are parallel to each other, and their directions in neighboring
planes form a certain finite angle. This is a cholesteric phase.

Fig. 7.18

Answer s.7.3.
a. The probability that the particle is at R in the N −th time step due to the jumps
to this point from nearest-neighbor ones is

P (R, N ) − P (R, N − 1) = w [P (R + δ, N − 1) − P (R, N − 1)] , (7.22)
δ
where the sum runs over the z nearest neighbors of R, located at R+δ. For instance,
there are two nearest neighbors, at δ = ±a, in the one-dimensional lattice, four in
the square lattice at δ = ±a1 , ±a2 , etc.
In one dimension, the right hand-side of Eq. (7.22) resembles that of Eq. (5.3),
under the mapping P (na, N ) ⇒ un (t). At N  1 the left-hand side can be ap-
proximated by P (R, N ) − P (R, N − 1) ≈ ∂P/∂N . Assuming that each step occurs
during the short time t0 , then t = N t0 , and then ∂P/∂t = (1/t0 )∂P/∂N , i.e., a first
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 616

616 Introduction to Solid-State Physics

derivative with respect to time in contrast with the second derivative with respect
to time that appears in Eq. (5.3). Similar equations appear in Sec. 5.9.
Still confining oneself to one dimension, one notes that the right-hand side of
Eq. (7.22) resembles the one that appears in the solution of problem 6.25, for wave
functions in the tight-binding approximation. That equation is derived from the
Schrödinger equation for the stationary solution exp[−iHt/]. The energy E re-
places the time derivative in the Schrödinger equation, i∂/∂t. Hence, the equation
for the random walk resembles the tight-binding approximation for the Schrödinger
equation, save for the factor i in the latter. This resemblance persists also at di-
mensions higher than 1.
b. In the continuum limit, the right-hand side of Eq. (7.22) in one dimension is
P (x + a) + P (x − a) − 2P (x) ≈ a2 ∂ 2 P/∂x2 .
In higher dimensions, the second derivative approaches ∇2 P , as in the wave equa-
tion, the Schrödinger equation, etc. Then, replacing the left-hand side
∂P/∂N ⇒ t0 ∂P/∂t ,
one finds
∂P/∂t = (a2 w/t0 )∂ 2 P/∂x2 ,
which is the diffusion equation.2 In one dimension, the solution of problem 7.3
becomes
2
/(2a2 N )
P (x, N ) = (2πa2 N )−1/2 e−x ,
and then
∂P/∂N = [x2 /(2a2 N 2 ) − 1/(2N )]P , ∂P/∂x = −[x/(a2 N )]P ,
from which follows that
∂P/∂N = (a2 /2)∂ 2 P/∂x2 ,
which is the diffusion equation with w = 1/2. A similar result can be derived in
three dimensions.
c. Applying the Fourier transform

 1
P (k, N ) = d3 Re−ik·R P (R, N ) ,
(2π)3
on the equation (with a shift of the variables in certain integrals), gives
  N
P(k, N ) = w P(k, N − 1)eik·δ ⇒ P(k, N ) = w eik·δ ,
δ δ
2 Fick’slaw states that the current of matter is proportional to the gradient of its density, i.e.,
J = −D∇n where D is the diffusion coefficient. The continuity equation, ∂n/∂t + ∇ · J = 0, then
yields ∂n/∂t = ∇ · (D∇n). The usual diffusion equation is obtained when D is a constant.
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 617

Selected topics 617

where the initial condition P (R, 0) = δ(R), which implies P(k, 0) = 1, is used. The
expression in the square brackets resembles expressions for the frequencies of lattice
vibrations, and for band energies in the tight-binding approximation.
d. The Taylor expansion is

P(k, 1) = w [1 + ik · δ − (k · δ)2 /2 + . . .] .
δ
This is similar to the expansion used in the solution of problem 7.3. In a Bravais
lattice δ δ = 0; δ (k·δ)2 /z replaces the expression (k·R)2  that appears there.
For hyper-cubic lattices in d dimensions
 
d
(k · δ)2 /z = 2 kα2 a2 /(2d) = k 2 a2 /d
δ α=1

(check!). For other lattices one obtains different quadratic expressions. For instance,

(k · δ)2 /z = 2[(k · a1 )2 + (k · a2 )2 + (k · (a1 − a2 ))2 ]/6 = k 2 a2 /4 ,
δ
in the triangular lattice. A change of variables in the integral of the inverse Fourier
transform, as in problem 7.3, justifies the neglect of the higher-order terms in the
expansion, and thus leads again to the Gaussian distribution. Save for the coefficient
of k 2 , which determines [R(N )]2 , this distribution is universal.

Answer s.7.4.
a. In order to identify the cylindrical structure which is formed, one has to add lines
perpendicular to the vectors L1 and L2 , at their edges, see Fig. 7.19. These are the
edges of the sheet that needs to be wrapped. The perpendicular line ends at a lattice
site that hosts the same point as the origin of the line (recall that the hexagonal
lattice contains two basis points in each rhombohedral unit cell). The latter lattice
site belongs to the next one-dimensional unit cell of the cylinder, and the vector
that reaches it from the origin is the basis vector of the one-dimensional lattice.
The one-dimensional unit cell contains all points within the rectangle created by
this vector and the original vector. These unit cells are displayed in Fig. 7.19 by
the dashed lines. The cell built on L1 contains eight lattice sites (four points of each
type of the original hexagonal lattice), and the one built on L2 contains 76 lattice
sites (38 of each type).
b. The energy spectrum of the graphene lattice is calculated in problem 6.24,
within the tight-binding model, and accounting only for overlaps between √ nearest
neighbors. √The basis vectors used there are marked in Fig. 7.19, a1 = a(3x̂+ 3ŷ)/2
and a2 = a 3ŷ. The energies are
E± (k) = E (0) ± |
γAB (k)| ,

AB (k) =
γ γAB e−ik·Rnn = γAB (e−ik·a1 + e−ik·a2 + e−ik·(a1 +a2 ) ) .
nn
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 618

618 Introduction to Solid-State Physics

Fig. 7.19

Explicitly,
√ √
γAB (k)|2 = |γAB |2 [3 + 2 cos( 3ky a) + 4 cos(3kx a/2) cos( 3ky a/2)] .
|
In the case of the cylinder built on the vector L1 = 4a1 − 2a2 = 6ax̂, one has to
add the periodic boundary condition along the x̂−axis, kx L1 = 6kx a = 2πn, and
therefore there are six discrete bands as a function of ky (which remains continuous),
on the two sides of E (0) .
c. In this case one expresses kx and ky in terms of kx and ky , that are obtained
from the rotation in the√ plane that transfers the x̂−axis into the direction of L2 =
5a
√ 1 − 2a 2 = (15x̂ + 3ŷ)a/2. The periodic boundary condition yields kx L2 =
57kx a = 2πn.

Answer s.7.5.
a. In the absence of the magnetic flux, the reflection amplitude for traversing the
ring is independent of the sense of the motion; i.e., it is r1 exp[iδ] for both the
clockwise motion and the anti-clockwise one. On the other hand, the magnetic
Aharonov-Bohm phase reverses its sign when the direction of motion is reversed.
This is the reason for the two different reflection amplitudes for the two directions
of motion. When the path encircles the ring m times the two reflection amplitudes
are rm exp[im(δ ± φ)]. It follows that the reflection coefficient is
∞

R = |r0 + rm eimδ (eimφ + e−imφ )|2
m=1

= |r0 | + 2|r1 |2 + 4|r0 ||r1 | cos(δ) cos(φ) + 2|r1 |2 cos(2φ) + . . . .

2

The reflection coefficient is indeed periodic in Φ, with periodicity Φ0 ; the explicit

expression contains all harmonics.
b. Different rings contain different realizations of scatterers (but with the same
mean concentration), and therefore the phase δ attains many different values. It
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 619

Selected topics 619

follows that cos(δ) is averaged to zero in an ensemble of many rings. Glancing at

the expression obtained in part (a) shows that as a result the odd harmonics (in
the flux) of the mean reflection coefficient disappear. The leading harmonics comes
from the term with cos(2φ), which is a periodic function of periodicity Φ0 /2. This
is indeed observed experimentally.
c. The reflection amplitude found in part (c) of problem 7.8 was
2iγ sin(ka)(E − E  + γeika )
r= −1 ,
(E − E  + γeika )2 − |γ  |2
with
γ  = γ0 − (γA )2 eiφ /(E − A ) , E  = E0 + (γA )2 /(E − A ) ,
where γ0 and E0 are the “bare” values of γ  and of E  . The denominator in the
expression for r depends on the parameters of the triangle only via the expression
|γ  |2 = (γ0 )2 + (γA )4 /(E − A )2 − γ0 γA
2
(eiφ + e−iφ )/(E − A ) .
Each of the terms on the RHS represents a “motion” of the electron from site 0 to
site 1 and back: the first term represents motion via the lower side of the triangle,
the second term represents the motion from 0 to A to 1 and back, and the last term
represents the clockwise motion, 0 to A to 1 to 0 or the anticlockwise motion, 0 to
1 to A to 0.
If one ignores the last term in |γ  |2 then the resulting expression for r contributes
only to r0 . Expanding r in a power series in γ0 γA 2
(eiφ + e−iφ )/(E − A ), it is easy to
see that the first term yields r1 . The second term is proportional to (e2iφ +e−2iφ +2).
The first two term contribute to r2 , and the last term (which results from a motion
clockwise followed by a motion anticlockwise) contributes to r0 . Similarly, the m’s
term in the series contributes to all the rn ’s with 0 ≤ n ≤ m.
December 12, 2014 15:40 BC: 9334 – Bose, Spin and Fermi Systems 9334-book9x6 page 2

This page intentionally left blank

July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 621

Index

acceptors, 476, 491–493, 500 Bohr magneton, 72, 271, 506, 571
acoustic branch, 319, 320, 332, 343, 348, bonding energy, 219–222, 229, 243, 248,
358, 368, 379, 380, 387, 393, 396, 400, 251, 258, 269
402, 408, 414 bonding wave function, 238
acoustic solutions, 314 Born-Oppenheimer approximation, 232,
Aharonov-Bohm effect, 567, 589, 596, 598 255, 259, 309
alkali atoms, 37, 216, 477 Born-von Karmann condition, 335
allotropes, 10, 11, 250, 291 Bragg law, 123, 139, 143
alloys, 8, 15, 58, 65, 164, 196 Bragg peaks, 113, 123–128, 137, 142, 147,
amorphous materials, 6 149, 150, 152, 154, 155, 157, 158, 161,
Anderson insulator, 512 163, 164, 166, 168–171, 173, 175, 176,
anharmonic terms, 311, 316, 317, 322, 192, 193, 196, 199, 202–205, 354, 361,
352, 363–365 362, 433
anti-bonding, 237, 265, 266, 562 Bravais lattice, 15–17, 23–25, 28, 30, 31,
anti-ferromagnet, 70, 72, 74, 90, 95, 96 35–39, 43, 50–52, 54, 56, 58, 61, 65, 68,
anti-Stokes scattering, 363 80, 82, 91, 98, 124, 125, 127, 129, 132,
artificial atom, 589 138, 139, 152, 164, 194, 211, 324, 466,
atomic force microscope (AFM), 114 617
Brillouin scattering, 363
ballistic motion, 588 Brillouin zone, 135, 136, 180, 313–315,
base, v, 15–19, 22–28, 30, 31, 33–39, 41, 318–321, 325–327, 330–337, 339, 340,
43, 46–49, 52–56, 67, 69, 75, 76, 78–82, 344–348, 360, 361, 363, 366, 367, 371,
87–91, 93–98, 101–104, 106, 107, 126, 372, 377, 378, 382, 385, 394, 395,
130, 131, 141, 142, 147, 151–153, 155, 398–400, 402, 404–406, 408, 440, 441,
157–159, 163, 171, 173–177, 185, 188, 445, 448–461, 466–469, 471, 474,
189, 192, 196, 201–205, 225, 229, 240, 477–479, 481–484, 487–489, 496, 521,
249, 252, 264, 266, 280, 282–284, 291, 525–527, 529–531, 533, 534, 536–538,
297–299, 411, 418, 430, 504, 543, 605 541, 547, 548, 550, 551, 553–556, 558
benzene molecule, 218, 219, 247, 248, 269, bulk modulus, 231, 260, 280, 290, 292,
335, 462, 464 293, 427
bi-partite lattice, 74
Bloch oscillations, 497, 498, 541 canonical momentum, 563, 564
Bloch theorem, 312, 356, 416, 436–440, carbon, 8, 10–13, 16, 19, 26, 29, 41, 44,
442, 444, 445, 449, 451, 455, 462–464, 88, 89, 102, 168, 185, 194, 199, 200,
471, 494, 512, 513, 521 217–219, 241, 245–251, 254, 282, 287,
body-centered cubic lattice, 36 299, 335, 353, 464, 477, 551

621
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 622

622 Introduction to Solid-State Physics

central limit theorem, 581 188, 190, 195, 199, 200, 202, 323, 326,
cesium-chloride (CsCl), 34, 36, 80 484, 488, 530, 541
charge conductivity, 423 cumulene, 11, 12, 16, 250
chemical potential, 427–431, 491–493, 510, cyclotron frequency, 420, 501–503, 515,
519, 520, 540, 591 543, 565
cholesteric phases, 578
classical equations of motion, 311, 324, de Haas-van Alphen effect, 505
394, 401, 432, 496, 514, 563, 573 Debye approximation, 341–343, 348–350,
close packed hexagonal lattice (HCP), 46 353, 355, 387, 389, 394–397, 402, 406,
close packing, 44, 45 407, 434
co-valent bond, 217, 218, 232 Debye frequency, 341, 348–350, 367, 370,
collisions, 364, 365, 367, 415–420, 426, 386, 387, 395, 396, 402, 406, 408
432–435, 496, 497, 501, 514, 551, 584, Debye temperature, 341, 344, 350, 355,
587, 593, 595, 597, 608 366, 395, 407, 430, 435
commensurate, 66–69, 90 Debye-Waller factor, 159, 161–164, 310,
compressibility, 231, 314 351, 355
condensed matter, v, 7, 58, 388 density of charge carriers, 491
conduction band, 371, 474–477, 489–493, density of states, 340–350, 385, 389,
496, 497, 499, 500, 503, 513, 539–541, 394–396, 399, 402, 406, 407, 416, 426,
549, 555 430, 431, 469, 470, 474, 485–490, 494,
conduction electrons, 72, 415, 424, 435, 504–507, 511, 531, 532, 546, 550, 553,
520, 589 560, 561, 587, 589, 593
continuum limit, 313, 314, 328, 411, 440, diamond, 8, 10–12, 26, 41, 42, 44, 54, 92,
613, 616 141, 152, 154, 157, 188, 201, 217, 220,
coordination number, 20, 21, 34, 78, 217, 221, 250, 251, 254, 266, 267, 270, 298,
613 476, 478
correlation function, 1, 4–6, 149, 150, 158, diatomic molecule, 218, 295
577, 581, 582, 605, 608 diffraction pattern, 58–60, 65, 113, 118,
Coulomb blockade, 590, 591 119, 123, 129, 133, 140, 144, 147, 150,
Coulomb force, 21, 216, 220 153, 154, 157, 158, 164, 166, 169, 172,
crystal, 1, 6, 8, 13, 15, 17–22, 24–28, 33, 173, 192, 194, 579
35–38, 40–44, 48, 55–60, 66, 67, 70–72, diffusion, 580, 597, 600, 602, 616
78, 80, 81, 85, 92, 94, 101, 102, 113, diffusion limited aggregation (DLA), 586
114, 116–119, 122–124, 137, 140, diffusive motion, 580
143–146, 148, 151–159, 163, 164, dimension, 1, 12, 15, 19, 51, 64, 126, 129,
166–168, 170, 172, 174–176, 189, 190, 146, 179, 207, 250, 309, 310, 314, 324,
192, 194, 195, 197, 201, 202, 215–217, 326, 328, 333, 335, 340, 341, 345, 346,
219–222, 225, 227–232, 248, 251, 253, 348, 351, 352, 355, 358, 360, 368, 372,
254, 258, 259, 262, 266, 269, 273, 279, 378, 399, 442, 448, 449, 453, 455, 461,
280, 291, 292, 298, 309–311, 314, 316, 472, 473, 485, 486, 488, 505, 519, 552,
317, 319, 321, 322, 324, 332, 335, 338, 580–586, 588, 595, 600, 613, 615, 616
343, 345, 349, 352–354, 358, 359, 361, dimers, 551, 562
363, 365, 367, 369, 371, 374, 387, 389, dispersion relation, 312, 315, 316, 318,
391, 394, 395, 398, 416, 437, 442, 448, 319, 327, 329, 337, 340, 346–350, 353,
466, 488, 489, 524, 533, 534, 549, 576, 356, 357, 359, 361, 362, 368, 370, 371,
577, 579, 605, 612 373, 376, 377, 384, 386, 387, 392, 393,
crystal growth, 21, 66 395–397, 404, 409, 467, 474, 489, 498,
cubic lattice, 52, 53, 65, 80, 83, 87, 88, 91, 549, 557, 610
94, 99, 101, 119–121, 132, 136, 138, 139, dodecahedron, 40, 53, 65, 109–111, 181
141, 154, 155, 157, 172–175, 177, 186, donors, 476, 491–494, 500, 540, 549
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 623

Index 623

double bond, 247, 250, 287 Fermi-Dirac distribution, 427–429, 431,

Drude theory, 415, 416, 423, 424, 428, 484, 490, 519, 553
433, 497, 516, 545 ferrimagnet, 73, 74
Dulong-Petit law, 339, 341, 343, 344, 349, Ferroelectricity, 57
350, 366, 386, 388, 392, 403, 424, 430, ferromagnet, 10, 57, 70, 73, 289, 299, 369,
431, 520, 552 370, 392, 409, 576
Fibonacci lattice, 61–64, 86, 164, 582
edge states, 508, 562 Fibonacci series, 60
effective mass, 431, 467, 491, 498–500, filling factor, 507, 509, 511
502, 543, 548–550, 555, 560, 589 flexural displacement, bending, buckling,
Einstein approximation, 343, 351, 355, 317
358, 403 Flory approximation, 584
Einstein relation, 603 fluctuations, 7, 14, 425, 575, 600
elastic scattering, 362, 595, 597, 613 Fourier law, 364
electron scattering, 116, 166
Fourier transform, 5, 6, 59, 113, 150, 158,
electronegativity, 217, 218 165, 209, 212, 214, 226, 315, 324, 333,
energy band, 441, 447, 456, 457, 460, 466, 336, 350, 368, 369, 449, 450, 454, 455,
468–470, 472, 481, 484, 488, 496, 549, 525, 554, 582, 607, 613, 616, 617
550, 562, 595, 609
fractal, 62, 575, 581–586
energy gap, 447, 449, 454, 457, 461, 475,
fractal dimension, 581, 583
477, 478, 484, 490–494, 498, 512, 523,
fractal growth, 585
525, 526, 528, 530, 536, 547–549, 551,
free energy, 7, 56, 73, 259, 337, 339, 365,
554, 556, 558, 561
412, 427–429, 517, 608
entropy, 7, 10, 56, 216, 259, 267, 268, 274,
276, 429, 579 frustration, 73, 75, 273, 276, 289
epitaxial adsorption, 15 fullerene, 12
equations of motion, 3, 309, 311, 314, 316,
318, 322, 324, 331, 332, 335, 336, 356, gas, 1–5, 7, 13, 68, 73, 150, 155, 216, 219,
369–371, 373–377, 379–383, 389, 391, 221, 255, 265, 366, 371, 416, 418,
392, 394, 396, 397, 399, 401, 408, 411, 424–432, 470, 473, 477, 485, 512, 518,
416, 419, 432, 494, 496, 500, 501, 514, 519, 547, 548, 553, 576, 587, 593–595,
515, 543, 557, 563, 573 604
equilibrium state, 322, 377 gate voltage, 590, 591, 593, 594
Evjen method, 223, 224 gauge, 564, 566–568, 570–572
Ewald method, 223 gel, 584
Ewald’s sphere, 192 glass, 1, 7–9, 73, 353, 600
exchange energy, 240, 270, 271, 273, 277 golden ratio, 61, 62, 74, 86, 164, 190
exchange forces, 72 Grüneisen parameter, 365
extended spectrum, 455–457, 459 granular materials, 48
extrinsic semiconductor, 492, 494, 557 graphene, 2, 9, 11–13, 15, 26–30, 66, 68,
71, 77, 93, 114, 119, 152, 169, 185, 217,
face-centered cubic lattice (FCC), 37, 38 244, 250–252, 268, 296, 317, 330–333,
family of lattice planes, 136 352, 353, 381, 395, 406, 471, 478, 509,
Fermi level, 269, 425, 429–431, 470, 473, 534, 550, 613, 617
474, 478, 480, 485, 489, 492, 498, 509, graphite, 9–12, 26–30, 44, 54, 68, 69, 71,
512, 518, 520 88, 93, 119, 153, 168, 169, 188, 250, 251,
Fermi momentum, 425, 431, 434, 478, 482, 266, 353, 478
603, 611 group velocity, 314, 315, 340, 343,
Fermi surface, 478, 481–484, 489, 501–506, 345–348, 371, 372, 385, 386, 393, 397,
537 405–407, 431, 495
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 624

624 Introduction to Solid-State Physics

Hall coefficient, 420, 421, 436, 502, 503 Ising model, 275, 276, 289
Hall effect, 2, 415, 416, 419, 421, 496,
508–512, 562, 566, 569, 588, 593, 594 Kepler conjecture, 47, 48
halogen atom, 216, 221 kinetic theory, 366, 416, 422
harmonic approximation, 161, 309, 311, Klitzing, 509, 511
314, 316, 322, 323, 335, 336, 352, Kronig-Penney potential, 442, 446–448,
363–365 454, 470
heat conductivity, 433
Heisenberg model, 273, 275, 276, 367 Landau levels, 503, 505–507, 509–511,
Heitler-London approximation, 239, 253 565–567
helical order, 311, 578 Landauer formula, 595, 596, 600
heterostructures, 71, 72 lanthanum cuprate, 42, 43, 54, 96, 272,
hexagonal lattice, 21, 22, 26–30, 34, 44, 585
45, 51, 54, 57, 71, 77, 90, 94, 104, 105, lattice, 8, 15, 16, 18, 19, 24, 25, 28–38, 40,
132, 139, 140, 152, 153, 169, 178, 185, 43, 47–58, 60–69, 71–106, 111, 113–115,
198, 201, 244, 250, 330, 331, 412, 546, 117–133, 135–143, 145–161, 163, 164,
552, 605, 617 166–204, 206, 207, 209, 211, 212, 215,
hexatic phases, 577 217, 219, 221–231, 243, 244, 250, 251,
holes, 354, 475, 476, 489–494, 499–502, 254, 258–262, 264, 266, 267, 269–272,
539, 540, 544, 546, 548, 549, 555, 557, 275, 276, 278–280, 287–292, 296–299,
560 309–336, 338, 340–342, 344, 346–348,
honeycomb lattice, 26 351, 352, 354–356, 358–364, 366–372,
Hooke’s law, 311, 323, 334, 411, 412 376, 379–381, 392–396, 400–404, 406,
hybridization, 218, 229, 239, 244, 245, 408, 409, 411, 412, 414, 415, 436, 439,
247, 252, 253, 268, 282, 283, 287, 292, 461, 462, 464, 466–468, 474, 546, 552,
296, 473, 477 579, 582, 584, 585, 587, 588, 595, 603,
hydrogen bond, 219–221, 263–266, 292, 605, 608, 609, 613, 615, 617
298 lattice constants, 16, 20, 33, 51, 66, 67,
hydrogen molecule, 217, 232, 239, 240, 69, 70, 73, 87, 93, 94, 96, 98, 99, 104,
261, 271, 272, 287, 462, 464, 562 105, 175, 207, 210, 310, 419, 421, 426,
427, 433, 465, 466, 541
ice, 3, 9, 220, 266–268, 276 lattice momentum, 360–362, 366, 437, 449
icosahedron, 65, 109–111 lattice vector, 15–18, 20, 22–25, 28, 30, 32,
impurities, 355, 358, 364, 367, 434, 435, 34, 42, 47, 49, 51, 52, 67, 97, 98, 102,
476, 497, 512, 587, 598, 613 179, 211, 366, 437, 438, 464
incommensurate, 67–69, 73, 176, 289, 300 Lennard-Jones potential, 219, 257, 258,
inelastic scattering, 167, 360, 362, 363, 324
396, 597 Lindemann criterion, 163, 190, 206, 259,
inflation, 62, 86, 575 310, 314, 351, 352, 354, 355, 389
insulator, 10, 11, 70, 475–478, 481, 508, liquid, v, 1–5, 7, 8, 14, 57, 150, 163, 251,
510, 512, 513, 537, 548, 550, 551, 555, 258, 259, 262, 310, 338, 354, 355, 512,
560, 584, 587, 593, 601 576–579, 584, 605, 612
intrinsic semiconductor, 491, 493, 494, liquid crystals, 8, 355, 575, 576, 578, 612
503, 544 localized states, 269, 359, 507, 511–513
inversion, point reflection, 50, 51, 82, 85, lock-in, 68, 69
323, 371 longitudinal waves, 313, 327, 336, 377
Ioffe-Regel criterion, 601 Lorentz force, 419, 563, 564
ionic bonding, 216, 221
ionic radius, 216, 226 Madelung constant, 222, 224, 225, 228,
ionic states, 233 278
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 625

Index 625

magnetic flux, 507, 565, 569, 572, 573, order-disorder transitions, 7

589, 596, 598, 611, 613, 618 organic chemistry, 158
magnetic flux quantum, 507 orthorhombic lattice, 53, 57, 130, 132,
magnetic structure factor, 172, 173, 175, 142, 146, 157, 182, 189, 194, 198, 210,
192 222, 265, 297, 440, 538
magnons, 367, 370, 371, 392
Matthiessen’s rule, 476, 500, 598 packing ratio, 19, 20, 22, 23, 26, 30, 34,
mean free path, 366, 367, 419, 426, 35, 37, 39, 41, 45–48, 54, 75–77, 79–81,
433–435, 547, 552, 587 83, 84, 92, 94, 100, 101, 103, 105
melting temperature, 3, 8, 163, 197, 206, paramagnetism, 553
220, 221, 259, 351, 354, 355, 389 Pauli principle, 216, 217, 219, 226, 229,
Mermin-Wagner theorem, 351, 352 238, 240, 241, 248, 257, 269, 272, 416,
mesoscopic physics, 13, 575, 586, 588 425–427, 432, 442, 507, 565, 582, 589
meta-stable state, 7 Peierls instability, 551
metal, 218, 221, 248, 262, 269, 272, 364, Peierls insulator, 562
371, 415, 422, 427, 449, 475, 482–484, Penrose tiles, 64
489, 502, 512, 550, 551, 560, 584, 597 percolation, 575, 584, 585
metallic bond, 218, 220, 247, 248, 270 periodic crystal structure, 215, 219, 354
metalloids, 478 periodic crystals, v, 8, 13, 15, 124, 164,
Miller indices, 137–141, 156, 157, 167, 166, 585
168, 175, 180, 187, 188, 193, 196, 203 periodic lattice, 1, 15, 16, 20, 21, 51, 65,
mobility, 417 66, 87, 113, 207, 434, 577, 578
molecular biology, 158 periodic table, 8, 10, 37, 43, 44, 47, 155,
molecular crystals, 292 216, 217, 221, 226, 241, 244, 245, 248,
molecular dynamics, 3 251, 252, 254, 265, 448, 477, 479, 482,
molecular orbitals, 239 589–591
molecular wave function, 243 perovskite structure, 42, 57, 93, 196, 290
monochromatic radiation, 118, 141, 143, persistent current, 570, 574
157 phase problem, 158, 187
monoclinic lattice, 54, 198 phase transition, 2, 3, 42, 57, 58, 96, 106,
Mott insulator, 416, 512, 513 196, 270, 271, 355, 576
multiplicity of family of planes, 138, 139 phase velocity, 312, 313
phases, 1–4, 7–10, 70, 125, 158, 196, 203,
nano-ribbon, 11, 12 220, 248, 258, 265–267, 354, 380,
nano-tube, 250, 353, 388 576–579, 597, 600, 605
nanotechnology, 13, 16, 586 phonon, 345, 360, 362, 363, 365–367, 390,
nearly-free electron, 416, 461, 468, 469, 396, 408, 434, 487, 520, 551, 587, 595
472, 480–483, 498 phonon collisions, 364–366, 415–417, 419
nematic phase, 576–578, 605, 615 phonon scattering, 363, 365
neutron scattering, v, 113, 169–171, 347 pi (π)-bond, 282
noble gases, 8, 219, 254, 257–260, 268, 477 pi (π)-bonds, 283
plasma frequency, 422
oblique lattice, 51–54, 83, 84, 97, 297 Plato’s bodies, 109
octahedron, 42, 43, 51, 65, 109, 110, 181, point group, 50, 51, 53
252 polar molecules, 263
Ohm’s law, 415, 416, 418, 420, 421, 509, polyhedra, 65, 109, 110
594, 599, 602 polymers, v, 8, 575, 579, 584
optical branch, 319, 320, 332, 343, 344, polymorphs, 9, 10, 26, 216, 225
375, 380, 396, 403 polyyne, 11, 12, 19, 250, 322, 551
optical lattice, 604 porous media, 585
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 626

626 Introduction to Solid-State Physics

powder scattering, 118, 141, 157, 171 582

primitive cell, 36, 37, 39, 43, 81, 98 scattering of radiation, 117, 361
projection, 35, 36, 38, 42, 49, 50, 62, 64, Seebeck effect, 423
65, 94, 103, 106, 128, 134, 181, 191, 192, self-averaging, 600
203, 502, 514, 614 self-avoiding walks (SAW), 582, 583
self-similarity, 582
quantum computing, 592 semiconductor, 475, 476, 491–494, 498,
quantum dot, 589–592, 596 503, 508, 544, 550, 557
quantum Hall effect, 508, 509 Shubnikov-de Haas effect, 505, 510
quantum interference, 587, 597, 600 sigma (σ)-bond, 282
quantum point contact, 593 silicene, 29, 71, 250, 251
quantum tunneling, 114 simple-cubic lattice (SC), 34, 35, 80, 140,
quantum wire, 593–595 193, 273, 483
quasi-crystals, 15, 58–60, 62, 64, 65, 103, single electron transistor, 421
113, 164, 166 singlet, 240, 271, 272
qubit, 592 smectic phases, 577, 578, 605
snow flakes, 9, 220
radial correlation function, 4, 6, 581, 582 solid, v, 1–4, 6–9, 12–14, 24–26, 34, 43,
Raman scattering, 363 56, 57, 66, 78, 86, 96, 103, 104, 106,
random walk, 580–584, 608, 611, 616 113, 115, 150, 163, 195, 232, 236, 258,
reciprocal lattice, 113, 124–140, 143, 259, 264, 265, 287, 292, 319, 342, 354,
146–149, 151–154, 156, 157, 167, 168, 376, 390, 398, 399, 415, 432, 433, 455,
173, 176, 178–180, 211, 433, 437, 440, 474, 481, 506, 537, 576–579, 582, 585,
448–450, 452, 482 587, 605
rectangular centered lattice, 52, 152, 186, Sommerfeld theory, 416, 424, 433, 436, 547
268, 288, 401 sound velocity, 313–315, 320, 327, 330,
rectangular lattice, 51–53, 97, 106, 120, 331, 340, 350, 354, 366, 371, 372, 377,
121, 130 378, 380, 384, 393, 394, 397, 400, 406,
reduced spectrum, 455–457, 459, 478, 525 411
relaxation time, 416, 417, 419, 421, 433, sp-type hybrid, 242, 282, 283, 317
435, 497–500, 514, 521, 541, 544, 549, sp2 -type hybrid, 244, 245
557 sp3 -type hybrid, 245, 283
renormalization, 580 sp3 d-type hybrid, 252
residual resistivity, 434, 435 sp3 d2 -type hybrid, 252
resistivity ratio, 435 space group, 50, 51
rotations, 36, 38, 50–52, 54, 56, 58, 59, 64, specific heat, 309, 336, 338–345, 348–350,
83, 88, 89, 96, 109–111, 164, 198, 244, 358, 365–367, 370, 384, 387, 388, 394,
261, 323, 328, 332, 405, 487, 534, 605 396, 397, 399, 402, 403, 407, 416, 423,
424, 429, 430, 433, 464, 474, 476, 478,
scanning tunneling microscope (STM), 484, 485, 489, 490, 493, 494, 520, 546,
113, 167 550, 552, 560, 586, 593, 600
scattering amplitude, 118, 119, 124, 127, spin glass, 73
148, 164, 170, 172, 176, 188, 202, 203, spin waves, 309, 369–371, 392, 396, 408
362 square lattice, 23, 51, 121, 467, 472, 487,
scattering angle, 117–119, 122, 124, 139, 488, 555
141, 143, 151, 156, 163, 169, 182, 187, Stokes scattering, 363
194, 199, 201, 514 strain tensor, 411
scattering from anti-ferromagnets, 70, 173 stress tensor, 411
scattering intensity, 125–127, 141, 143, structure factor, 113, 147, 149, 151–158,
147, 152, 153, 170, 183, 185, 199, 200, 161, 163, 168, 169, 172, 173, 175, 184,
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 627

Index 627

185, 188, 189, 192–195, 199, 200, 130–132, 134, 138, 141–143, 145–155,
202–205, 226, 579 157, 158, 160, 163, 168, 171–179, 181,
subgroup, 56 184–186, 188, 189, 192, 194–196, 198,
super-exchange, 272 199, 203–205, 207, 211, 212, 221–223,
superconductivity, 2, 70 225–231, 251, 260, 262, 269, 271, 274,
superconductor, 10, 43 278–280, 289, 290, 295, 298, 299,
symmetries, 8, 15, 48, 49, 54–56, 58, 65, 318–322, 325, 330–332, 336, 337, 339,
66, 109, 132, 144, 157, 164, 166, 233, 343, 344, 351, 353, 370, 374, 375, 380,
234, 254, 323, 325, 405, 579, 589, 605 381, 394–396, 400–402, 405–407, 427,
symmetry group, 49–53, 55, 56, 74, 83, 89, 436, 439, 443, 450, 462, 470, 471, 478,
90, 109, 110 482, 483, 526, 533, 534, 537, 547, 552,
symmetry operation, 48, 49, 58, 144 556, 561, 562, 578, 579, 588, 605, 608,
synchrotron radiation, 116, 158 609, 613, 617
universal conductance fluctuations, 598,
table salt (NaCl), 9, 21, 37, 74, 90, 101, 600, 601
155, 171, 175, 202, 216, 223, 227 upper critical dimension, 583
tetragonal lattice, 53, 54, 82, 83, 157, 182,
189, 192, 196, 198, 202, 203 valence band, 371, 475, 476, 489, 490,
tetrahedron, 41, 42, 45, 47, 65, 101, 492–494, 499, 500, 512, 539, 540, 555
109–111, 194, 200, 245, 246, 249, 251, valence electrons, 216–219, 241, 242,
263, 266, 279, 280 244–246, 250, 254, 269, 270, 287, 415,
thermal conductivity, 422, 423 448
thermal expansion, 309, 363, 364 van der Waals bond, 219, 221, 254
thermoelectric effect, 423 van der Waals force, 219, 255
thin layer, 167, 508, 509, 593 van Hove singularity, 531
Thouless energy, 602 von Laue equations, 128
tight binding, 416, 446, 461, 466, 470–472, von Laue law, 122, 123
477, 481, 483–485, 487–489, 531, 534, von Laue method, 143, 144
537, 548, 550, 551, 554, 561
topological insulators, 507, 508, 562 water, 3, 20, 219, 220, 248, 249, 266, 267,
topological phase, 355, 578 294, 295, 584, 585
translation operator, 438 wave function, 160, 162, 217, 218,
translations group, 24, 30, 49, 50, 62, 439 232–234, 236–242, 248, 250, 253, 256,
transverse waves, 327, 377, 413 257, 269, 270, 286, 301–304, 314, 320,
triangular lattice, 20, 21, 23, 24, 26, 33, 325, 326, 335, 358, 359, 361, 413, 424,
44–47, 49, 51, 74, 75, 77, 78, 97, 176, 436, 437, 443, 445, 448–450, 452, 454,
177, 180, 552, 577, 578 462–465, 467, 470, 471, 474, 524, 525,
triclinic lattice, 54, 197, 198 531, 532, 565, 566, 568, 571, 572, 587,
trigonal lattice, 103, 197 595, 599, 602, 603, 611
triple bond, 247, 282, 292, 376 wave packet, 117, 315, 431, 495, 496, 601
triplet, 39, 179, 180, 182, 189, 201, 241, weak anti-localization, 599
271, 272, 281 weak localization, 597, 598, 603
two-dimensional electron gas (2DEG), 13, Wiedemann-Franz law, 423, 424
593, 594 Wigner-Seitz cell, 22, 23, 32–34, 40, 75,
76, 78, 80, 92, 97, 99, 135, 150, 152,
Umklapp collision, 366, 367, 497 180, 181
Umklapp process, 367 Wurtzite, 157
unit cell, 16–18, 20–22, 24–43, 46–49,
52–55, 57, 60, 64, 67–70, 73–85, 87–91, X-rays, 13, 113, 115, 116, 142, 144, 147,
93–100, 102–104, 106, 121, 122, 128, 155, 156, 164, 167, 169, 170, 194, 195
July 4, 2018 17:35 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 628

628 Introduction to Solid-State Physics

Young’s modulus, 314

zero-point motion, 162, 259, 309, 384

zinc blende, 40, 41, 48, 92, 101, 157, 188,
227, 229, 251, 253, 279
July 18, 2018 14:34 ws-book975x65-961x669 Introduction to Solid state physics-BC-bk040 OURBOOK page 629

Name Value CGS SI Notation

Planck’s constant 6.6262 𝑥×𝑥 10−27 𝑒𝑟𝑔 ⋅ 𝑠𝑒𝑐 10−34 𝐽 ⋅ 𝑠 ℎ

−15
in Ev 4.1357 𝑥×𝑥 10 𝑒𝑉 ⋅ 𝑠

Reduced Planck’s constant 1.0546 𝑥×𝑥 10−27 𝑒𝑟𝑔 ⋅ 𝑠𝑒𝑐 10 −34

𝐽⋅𝑠 ℏ
𝑥×𝑥 −16
in Ev 6.5822 10 𝑒𝑉 ⋅ 𝑠

Boltzman’s constant 1.3807 𝑥×𝑥 10−16 𝑒𝑟𝑔 ⋅ 𝐾 −1 10−23 𝐽 ⋅ 𝐾 −1 𝑘𝐵

in Ev 8.6173 𝑥×𝑥 10−5 𝑒𝑉 ⋅ 𝐾 −1

Electron charge 1.6021 𝑥×𝑥 10−12 𝑒𝑟𝑔 ⋅ 𝑒𝑉 −1 10−19 𝑐𝑜𝑙𝑜𝑢𝑚𝑏 𝑒

9.1095 −28 −31
Electron mass 𝑥×𝑥 10 𝑔 10 𝑘𝑔 𝑚
−24 −1
Bohr magneton 9.2740 𝑥×𝑥 10−21 𝑒𝑟𝑔 ⋅ 𝐺 −1 10 𝐽⋅𝑇 μ𝐵
−5
in Ev 5.7883 𝑥×𝑥 10 𝑒𝑉 ⋅ 𝑇 −1

Bohr radius 5.2917 𝑥×𝑥 10−9 𝑐𝑚 10−11 𝑚 𝑎𝐵

Rydberg’s constant 13.605 𝑥×𝑥 𝑒𝑉 𝑅𝑦

10 −1 8 −1
Speed of light in vacuum 2.9979 𝑥×𝑥 10 𝑐𝑚 ⋅ 𝑠𝑒𝑐 10 𝑚 ⋅ 𝑠 𝑐
23 −1
Avogadro’s number 6.0221 𝑥×𝑥 10 𝑚𝑜𝑙 𝑁𝐴

Klitzing constant 2.5812 𝑥×𝑥 104 Ω 𝑅𝐾

Permittivity of vacuum 8.8541 𝑥×𝑥 10−12 𝐹 ⋅ 𝑚−1 ϵ0

4π −7 −1
Permeability of vacuum 𝑥×𝑥 10 𝐻 ⋅ 𝑚 μ0