Algebra And Applications Volume 2 Combinatorial

Algebra And Hopf Algebras Abdenacer Makhlouf

download

https://ebookbell.com/product/algebra-and-applications-
volume-2-combinatorial-algebra-and-hopf-algebras-abdenacer-
makhlouf-46707626

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Banach Algebras And The General Theory Of Algebras Volume 2 Algebras

Encyclopedia Of Mathematics And Its Applications Vol 2 1st Theodore W
Palmer

https://ebookbell.com/product/banach-algebras-and-the-general-theory-
of-algebras-volume-2-algebras-encyclopedia-of-mathematics-and-its-
applications-vol-2-1st-theodore-w-palmer-2446932

Linear Algebra And Optimization With Applications To Machine Learning

Volume Ii Fundamentals Of Optimization Theory With Applications To
Machine Learning 1st Edition Jean Gallier

https://ebookbell.com/product/linear-algebra-and-optimization-with-
applications-to-machine-learning-volume-ii-fundamentals-of-
optimization-theory-with-applications-to-machine-learning-1st-edition-
jean-gallier-50947542

Linear Algebra And Optimization With Applications To Machine Learning

Volume I Linear Algebra For Computer Vision Robotics And Machine
Learning 1st Edition Jean H Gallier

https://ebookbell.com/product/linear-algebra-and-optimization-with-
applications-to-machine-learning-volume-i-linear-algebra-for-computer-
vision-robotics-and-machine-learning-1st-edition-jean-h-
gallier-51410252

Algebra And Applications 1 Nonassociative Algebras And Categories 1st

Edition Abdenacer Makhlouf Coordinator

https://ebookbell.com/product/algebra-and-
applications-1-nonassociative-algebras-and-categories-1st-edition-
abdenacer-makhlouf-coordinator-49198566
Universal Algebra And Applications In Theoretical Computer Science 1st
Edition Klaus Denecke Author Shelly L Wismath Author

https://ebookbell.com/product/universal-algebra-and-applications-in-
theoretical-computer-science-1st-edition-klaus-denecke-author-shelly-
l-wismath-author-12056244

Further Algebra And Applications 1st Edition P M Cohn Ma Phd

https://ebookbell.com/product/further-algebra-and-applications-1st-
edition-p-m-cohn-ma-phd-4210808

Multigraded Algebra And Applications 1st Ene Viviana Miller Ezra

https://ebookbell.com/product/multigraded-algebra-and-
applications-1st-ene-viviana-miller-ezra-7106722

Geometric Algebra And Applications In Physics 1st Edition Venzo De

Sabbata

https://ebookbell.com/product/geometric-algebra-and-applications-in-
physics-1st-edition-venzo-de-sabbata-891920

Further Algebra And Applications 1st Edition Paul M Cohn

https://ebookbell.com/product/further-algebra-and-applications-1st-
edition-paul-m-cohn-1823480
Algebra and Applications 2
 SCIENCES
Mathematics, Field Director – Nikolaos Limnios
Algebra and Geometry, Subject Head – Abdenacer Makhlouf

Algebra and Applications 2

Combinatorial Algebra and Hopf Algebras

Coordinated by
Abdenacer Makhlouf
First published 2021 in Great Britain and the United States by ISTE Ltd and John Wiley & Sons, Inc.

Apart from any fair dealing for the purposes of research or private study, or criticism or review, as
permitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced,
stored or transmitted, in any form or by any means, with the prior permission in writing of the publishers,
or in the case of reprographic reproduction in accordance with the terms and licenses issued by the
CLA. Enquiries concerning reproduction outside these terms should be sent to the publishers at the
undermentioned address:

ISTE Ltd John Wiley & Sons, Inc.

27-37 St George’s Road 111 River Street
London SW19 4EU Hoboken, NJ 07030
UK USA

www.iste.co.uk www.wiley.com

© ISTE Ltd 2021

The rights of Abdenacer Makhlouf to be identified as the author of this work have been asserted by him
in accordance with the Copyright, Designs and Patents Act 1988.

Library of Congress Control Number: 2021942616

British Library Cataloguing-in-Publication Data

A CIP record for this book is available from the British Library
ISBN 978-1-78945–018-7

ERC code:
PE1 Mathematics
PE1_2 Algebra
PE1_5 Lie groups, Lie algebras
PE1_12 Mathematical physics
 Contents

Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Abdenacer M AKHLOUF

Chapter 1. Algebraic Background for Numerical Methods, Control

Theory and Renormalization . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Dominique M ANCHON
1.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2. Hopf algebras: general properties . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1. Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.2. Coalgebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.3. Convolution product . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.4. Bialgebras and Hopf algebras . . . . . . . . . . . . . . . . . . . . . . 7
1.2.5. Some simple examples of Hopf algebras . . . . . . . . . . . . . . . . 8
1.2.6. Some basic properties of Hopf algebras . . . . . . . . . . . . . . . . 9
1.3. Connected Hopf algebras . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.1. Connected graded bialgebras . . . . . . . . . . . . . . . . . . . . . . 10
1.3.2. An example: the Hopf algebra of decorated rooted trees . . . . . . . 13
1.3.3. Connected filtered bialgebras . . . . . . . . . . . . . . . . . . . . . . 14
1.3.4. The convolution product . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.5. Characters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.3.6. Group schemes and the Cartier–Milnor–Moore–Quillen theorem . . 19
1.3.7. Renormalization in connected filtered Hopf algebras . . . . . . . . . 21
1.4. Pre-Lie algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
1.4.1. Definition and general properties . . . . . . . . . . . . . . . . . . . . 24
1.4.2. The group of formal flows . . . . . . . . . . . . . . . . . . . . . . . . 25
vi Algebra and Applications 2

1.4.3. The pre-Lie Poincaré–Birkhoff–Witt theorem . . . . . . . . . . . . . 26

1.5. Algebraic operads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.5.1. Manipulating algebraic operations . . . . . . . . . . . . . . . . . . . 28
1.5.2. The operad of multi-linear operations . . . . . . . . . . . . . . . . . 29
1.5.3. A definition for linear operads . . . . . . . . . . . . . . . . . . . . . 31
1.5.4. A few examples of operads . . . . . . . . . . . . . . . . . . . . . . . 32
1.6. Pre-Lie algebras (continued) . . . . . . . . . . . . . . . . . . . . . . . . 35
1.6.1. Pre-Lie algebras and augmented operads . . . . . . . . . . . . . . . 35
1.6.2. A pedestrian approach to free pre-Lie algebra . . . . . . . . . . . . . 36
1.6.3. Right-sided commutative Hopf algebras
and the Loday–Ronco theorem . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.6.4. Pre-Lie algebras of vector fields . . . . . . . . . . . . . . . . . . . . 40
1.6.5. B-series, composition and substitution . . . . . . . . . . . . . . . . . 42
1.7. Other related algebraic structures . . . . . . . . . . . . . . . . . . . . . . 44
1.7.1. NAP algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
1.7.2. Novikov algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.7.3. Assosymmetric algebras . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.7.4. Dendriform algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
1.7.5. Post-Lie algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
1.8. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

Chapter 2. From Iterated Integrals and Chronological Calculus

to Hopf and Rota–Baxter Algebras . . . . . . . . . . . . . . . . . . . . . . 55
Kurusch E BRAHIMI -FARD and Frédéric PATRAS
2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.2. Generalized iterated integrals . . . . . . . . . . . . . . . . . . . . . . . . 58
2.2.1. Permutations and simplices . . . . . . . . . . . . . . . . . . . . . . . 59
2.2.2. Descents, NCSF and the BCH formula . . . . . . . . . . . . . . . . . 64
2.2.3. Rooted trees and nonlinear differential equations . . . . . . . . . . . 67
2.2.4. Flows and Hopf algebraic structures . . . . . . . . . . . . . . . . . . 71
2.3. Advances in chronological calculus . . . . . . . . . . . . . . . . . . . . 74
2.3.1. Chronological calculus and half-shuffles . . . . . . . . . . . . . . . . 75
2.3.2. Chronological calculus and pre-Lie products . . . . . . . . . . . . . 79
2.3.3. Time-ordered products and enveloping algebras . . . . . . . . . . . 81
2.3.4. Formal flows and Hopf algebraic structures . . . . . . . . . . . . . . 83
2.4. Rota–Baxter algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.4.1. Origin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
2.4.2. Definition and examples . . . . . . . . . . . . . . . . . . . . . . . . . 91
2.4.3. Related algebraic structures . . . . . . . . . . . . . . . . . . . . . . . 95
2.4.4. Atkinson’s factorization and Bogoliubov’s recursion . . . . . . . . . 101
2.4.5. Spitzer’s identity: commutative case . . . . . . . . . . . . . . . . . . 103
2.4.6. Free commutative Rota–Baxter algebras . . . . . . . . . . . . . . . . 107
 Contents vii

2.4.7. Spitzer’s identity: noncommutative case . . . . . . . . . . . . . . . . 108

2.4.8. Free Rota–Baxter algebras . . . . . . . . . . . . . . . . . . . . . . . 111
2.5. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

Chapter 3. Noncommutative Symmetric Functions, Lie Series

and Descent Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
Jean-Yves T HIBON
3.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.2. Classical symmetric functions . . . . . . . . . . . . . . . . . . . . . . . . 120
3.2.1. Symmetric polynomials . . . . . . . . . . . . . . . . . . . . . . . . . 120
3.2.2. The Hopf algebra of symmetric functions . . . . . . . . . . . . . . . 122
3.2.3. The λ-ring notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.2.4. Symmetric functions and formal power series . . . . . . . . . . . . . 125
3.2.5. Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
3.3. Noncommutative symmetric functions . . . . . . . . . . . . . . . . . . . 129
3.3.1. Basic definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.3.2. Generators and linear bases . . . . . . . . . . . . . . . . . . . . . . . 131
3.3.3. Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
3.3.4. Solomon’s descent algebras . . . . . . . . . . . . . . . . . . . . . . . 136
3.4. Lie series and Lie idempotents . . . . . . . . . . . . . . . . . . . . . . . 139
3.4.1. Permutational operators on tensor spaces . . . . . . . . . . . . . . . 139
3.4.2. The Hausdorff series . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
3.4.3. Lie idempotents in the descent algebra . . . . . . . . . . . . . . . . . 143
3.5. Lie idempotents as noncommutative symmetric functions . . . . . . . . 144
3.5.1. Noncommutative power-sums . . . . . . . . . . . . . . . . . . . . . . 144
3.5.2. The Magnus expansion . . . . . . . . . . . . . . . . . . . . . . . . . 146
3.5.3. The continuous BCH expansion . . . . . . . . . . . . . . . . . . . . 148
3.5.4. Another proof of the Magnus expansion . . . . . . . . . . . . . . . . 150
3.5.5. The (1 − q)-transform . . . . . . . . . . . . . . . . . . . . . . . . . . 150
3.5.6. Hopf algebras enter the scene . . . . . . . . . . . . . . . . . . . . . . 151
3.5.7. A one-parameter family of Lie idempotents . . . . . . . . . . . . . . 152
3.5.8. The iterated q-bracketing and its diagonalization . . . . . . . . . . . 153
3.6. Decompositions of the descent algebras . . . . . . . . . . . . . . . . . . 155
3.6.1. Complete families of minimal orthogonal idempotents . . . . . . . . 155
3.6.2. Eulerian idempotents . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
3.6.3. Generalized Eulerian idempotents . . . . . . . . . . . . . . . . . . . 160
3.7. Decompositions of the tensor algebra . . . . . . . . . . . . . . . . . . . 160
3.8. General deformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
3.9. Lie quasi-idempotents as Lie polynomials . . . . . . . . . . . . . . . . . 163
3.9.1. The left derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
3.9.2. Multilinear Lie polynomials . . . . . . . . . . . . . . . . . . . . . . . 164
3.9.3. Decompositions on other bases . . . . . . . . . . . . . . . . . . . . . 167
viii Algebra and Applications 2

3.10. Permutations and free quasi-symmetric functions . . . . . . . . . . . . 169

3.10.1. Free quasi-symmetric functions . . . . . . . . . . . . . . . . . . . . 169
3.11. Packed words and word quasi-symmetric functions . . . . . . . . . . . 171
3.12. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175

Chapter 4. From Runge–Kutta Methods to Hopf Algebras

of Rooted Trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Ander M URUA
4.1. Numerical integration methods for ordinary differential equations . . . 179
4.1.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
4.1.2. Runge–Kutta methods . . . . . . . . . . . . . . . . . . . . . . . . . . 180
4.2. Algebraic theory of Runge–Kutta methods . . . . . . . . . . . . . . . . 182
4.2.1. The order conditions of RK methods . . . . . . . . . . . . . . . . . . 182
4.2.2. The independence of order conditions . . . . . . . . . . . . . . . . . 186
4.2.3. Proof of necessary and sufficient order conditions . . . . . . . . . . 188
4.2.4. Composition of RK methods, rooted trees and forests . . . . . . . . 191
4.2.5. The Butcher group . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195
4.2.6. Equivalence classes of RK methods . . . . . . . . . . . . . . . . . . 197
4.2.7. Bibliographical comments . . . . . . . . . . . . . . . . . . . . . . . . 198
4.3. B-series and related formal expansions . . . . . . . . . . . . . . . . . . . 198
4.3.1. B-series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
4.3.2. Backward error analysis, the exponential and the logarithm . . . . . 199
4.3.3. Series of linear differential operators . . . . . . . . . . . . . . . . . . 203
4.3.4. The Lie algebra of the Butcher group . . . . . . . . . . . . . . . . . 205
4.3.5. The pre-Lie algebra structure on g . . . . . . . . . . . . . . . . . . . 206
4.3.6. Bibliographical comments . . . . . . . . . . . . . . . . . . . . . . . . 209
4.4. Hopf algebras of rooted trees . . . . . . . . . . . . . . . . . . . . . . . . 209
4.4.1. The commutative Hopf algebra of rooted trees . . . . . . . . . . . . 210
4.4.2. The dual algebra H∗ and the dual Hopf algebra H◦ . . . . . . . . . 212
4.4.3. B-series and series of differential operators revisited . . . . . . . . . 213
4.4.4. A universal property of the commutative Hopf algebra
of rooted trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215
4.4.5. The substitution law . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
4.4.6. Bibliographical comments . . . . . . . . . . . . . . . . . . . . . . . . 217
4.5. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217

Chapter 5. Combinatorial Algebra in Controllability and

Optimal Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Matthias K AWSKI
5.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
5.1.1. Motivation: idealized examples . . . . . . . . . . . . . . . . . . . . . 223
 Contents ix

5.1.2. Controlled dynamical systems . . . . . . . . . . . . . . . . . . . . . 225

5.1.3. Fundamental questions in control . . . . . . . . . . . . . . . . . . . . 226
5.2. Analytic foundations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228
5.2.1. State-space models and vector fields on manifolds . . . . . . . . . . 228
5.2.2. Chronological calculus . . . . . . . . . . . . . . . . . . . . . . . . . . 230
5.2.3. Piecewise constant controls
and the Baker–Campbell–Hausdorff formula . . . . . . . . . . . . . . . . . 233
5.2.4. Picard iteration and formal series solutions . . . . . . . . . . . . . . 235
5.2.5. The Chen–Fliess series and abstractions . . . . . . . . . . . . . . . . 237
5.3. Controllability and optimality . . . . . . . . . . . . . . . . . . . . . . . . 241
5.3.1. Reachable sets and accessibility . . . . . . . . . . . . . . . . . . . . 241
5.3.2. Small-time local controllability . . . . . . . . . . . . . . . . . . . . . 243
5.3.3. Nilpotent approximating systems . . . . . . . . . . . . . . . . . . . . 247
5.3.4. Optimality and the maximum principle . . . . . . . . . . . . . . . . 251
5.3.5. Control variations and approximating cones . . . . . . . . . . . . . . 255
5.4. Product expansions and realizations . . . . . . . . . . . . . . . . . . . . 262
5.4.1. Variation of parameters and exponential products . . . . . . . . . . 263
5.4.2. Computations using Zinbiel products . . . . . . . . . . . . . . . . . 267
5.4.3. Exponential products and normal forms for nilpotent systems . . . . 269
5.4.4. Logarithm of the Chen series . . . . . . . . . . . . . . . . . . . . . . 273
5.5. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279

Chapter 6. Algebra is Geometry is Algebra – Interactions Between

Hopf Algebras, Infinite Dimensional Geometry and Application . . 287
Alexander S CHMEDING
6.1. The Butcher group and the Connes–Kreimer algebra . . . . . . . . . . . 288
6.1.1. The Butcher group and B-series from numerical analysis . . . . . . 288
6.1.2. Beyond the Butcher group . . . . . . . . . . . . . . . . . . . . . . . . 291
6.2. Character groups of graded and connected Hopf algebras . . . . . . . . 292
6.2.1. The exponential and logarithm . . . . . . . . . . . . . . . . . . . . . 294
6.3. Controlled groups of characters . . . . . . . . . . . . . . . . . . . . . . . 297
6.3.1. Conventions for this section . . . . . . . . . . . . . . . . . . . . . . . 297
6.3.2. Combinatorial Hopf algebras and the inverse factorial character . . 304
6.4. Appendix: Calculus in locally convex spaces . . . . . . . . . . . . . . . 305
6.4.1. C r -Manifolds and C r -mappings between them . . . . . . . . . . . . 306
6.5. References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 306

List of Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
 Preface
Abdenacer M AKHLOUF
University of Haute Alsace, Mulhouse, France

The aim of this series of books is to report on the new trends of research in algebra
and related topics. We provide an insight into the fast development of new concepts
and theories related to algebra and present self-contained chapters on various topics,
with each chapter combining some of the features of both graduate-level textbooks
and research-level surveys. Each chapter includes an introduction with motivations
and historical remarks, as well as the basic concepts, main results and perspectives.
Moreover, the authors have commented on the relevance of the results in relation to
other results and applications.

In this volume, the chapters encompass surveys of basic theories on

non-associative algebras like Lie theories, using modern tools and more recent
algebraic structures like Hopf algebras, which are related to Quantum groups
and Mathematical Physics. The algebraic background of pre-Lie algebras, other
non-associative algebras (non-associative permutative, assosymmetric, dendriform,
etc.) and algebraic operads is presented. This volume also deals with noncommutative
symmetric functions, Lie series, descent algebras, chronological and Rota–Baxter
algebras. We focus on the increasing role played by Combinatorial algebra and Hopf
algebras, as well as some non-associative algebraic structures in iterated integrals,
chronological calculus, differential equations, numerical methods and control theory.
It turns out that the Hopf algebra of rooted trees is an adequate tool, not only for
vector fields, but also for studying the numerical approximation of their integral
curves. Runge–Kutta methods form a group (called the Butcher group), which is
the character group of the Connes–Kreimer Hopf algebra. The algebraic theory
of Runge–Kutta methods B-series and related formal expansions are considered.
Algebraic structures underlying calculus with iterated integrals lead naturally to

Algebra and Applications 2,

coordinated by Abdenacer M AKHLOUF.
© ISTE Ltd 2021.
xii Algebra and Applications 2

the notions of descent (Hopf) algebra, as well as permutation Hopf algebra. In

this volume we discuss the Lie-theoretic perspective and advances of chronological
calculus. Chronological algebras and time-ordered products appear in an (almost)
uncountable number of places, especially in theoretical physics and control theory.
Noncommutative symmetric functions are applied to the study of formal power series
with coefficients in a noncommutative algebra, in particular to Lie series. Moreover,
Lie idempotents, Eulerian idempotents and Magnus expansion are considered. In
addition, the interaction of algebra and (infinite-dimensional) geometry in the guise
of Hopf algebras and certain associated character groups is examined. It turns out
that fundamental concepts in control theory are inherently linked to combinatorial and
algebraic structures. It is shown how modern combinatorial algebraic tools provide
deeper insight and facilitate analysis, computations and design. The emphasis is on
exhibiting the algebraic structures that map combinatorial structures to geometric and
dynamic objects.

I thank Kurusch Ebrahimi-Fard for suggesting these topics, presented at Benasque

Intensive School, and express my deep gratitude to all of the contributors of this
volume and to ISTE for their support.

August 2021
 1
Algebraic Background for
Numerical Methods, Control
Theory and Renormalization

Dominique M ANCHON
University of Clermont Auvergne, Clermont-Ferrand, France

1.1. Introduction

Since the pioneering work of Cayley in the 19th century (Cayley 1857), we have
known that rooted trees and vector fields on the affine space are closely related.
Surprisingly enough, rooted trees were also revealed to be a fundamental tool for
studying not only the integral curves of vector fields, but also their Runge–Kutta
numerical approximations (Butcher 1963).

The rich algebraic structure of the k-vector space T spanned by rooted trees
(where k is some field of characteristic zero) can be, in a nutshell, described
as follows: T is both the free pre-Lie algebra with one generator and the free
non-associative permutative algebra with one generator (Chapoton and Livernet 2001;
Dzhumadil’daev and Löfwall 2002), and moreover, there are two other pre-Lie
structures on T, of operadic nature, which show strong compatibility with the first
pre-Lie (respectively the NAP) structure (Chapoton and Livernet 2001; Calaque et al.
2011; Manchon and Saidi 2011). The Hopf algebra of coordinates on the Butcher
group (Butcher 1963), that is, the graded dual of the enveloping algebra U(T) (with
respect to the Lie bracket given by the first pre-Lie structure), was first investigated

Algebra and Applications 2,

coordinated by Abdenacer M AKHLOUF.
© ISTE Ltd 2021.
Algebra and Applications 2: Combinatorial Algebra and Hopf Algebras,
First Edition. Abdenacer Makhlouf.
© ISTE Ltd 2021. Published by ISTE Ltd and John Wiley & Sons, Inc.
2 Algebra and Applications 2

in Dür (1986), and intensively studied by Kreimer for renormalization purposes in

Quantum Field Theory (Connes and Kreimer 1998; Kreimer 2002), see also Brouder
(2000).

This chapter is organized as follows: the first section is devoted to general

connected graded or filtered Hopf algebras, including the renormalization of their
characters. The second section gives a short presentation of operads in the symmetric
monoidal category of vector spaces, and the third section will treat pre-Lie algebras in
some detail: in particular, we will give a “pedestrian” proof of the Chapoton–Livernet
theorem on free pre-Lie algebras. In the last section, Rota–Baxter, dendriform and
NAP algebras will be introduced.

1.2. Hopf algebras: general properties

We choose a base field k of characteristic zero. Most of the material here is

borrowed from Manchon (2008), to which we can refer for more details.

1.2.1. Algebras

A k-algebra is by definition a k-vector space A together with a bilinear map

m : A ⊗ A → A which is associative . The associativity is expressed by the
commutativity of the following diagram:

m⊗I
A⊗A⊗A / A⊗A

I⊗m m
 m

A⊗A / A

The algebra A is unital if there is a unit 1 in it. This is expressed by the

commutativity of the following diagram:

u⊗I I⊗u
k⊗A / A⊗A o A⊗k
II
II ∼ ∼ uuu
II m u
II uu
II  uuu
$ zu
A

where u is the map from k to A defined by u(λ) = λ1. The algebra A is commutative
if m ◦ τ = m, where τ : A ⊗ A → A ⊗ A is the flip , defined by τ (a ⊗ b) = b ⊗ a.
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 3

A subspace J ⊂ A is called a subalgebra (respectively a left ideal, right ideal

and two-sided ideal ) of A if m(J ⊗ J) (respectively m(A ⊗ J), m(J ⊗ A),
m(J ⊗ A + A ⊗ J)) is included in J.

With any vector space V, we can associate its tensor algebra T (V ). As a vector
space, it is defined by:

T (V ) = V ⊗k ,
k≥0

with V ⊗0 = k and V ⊗k+1 := V ⊗ V ⊗k . The product is given by the concatenation :

m(v1 ⊗ · · · ⊗ vp , vp+1 ⊗ · · · ⊗ vp+q ) = v1 ⊗ · · · ⊗ vp+q .

The embedding of k = V ⊗0 into T (V ) gives the unit map u. The tensor

algebra T (V ) is also called the free (unital) algebra generated by V . This algebra
is characterized by the following universal property: for any linear map ϕ from V
to a unital algebra A, there is a unique unital algebra morphism ϕ  from T (V ) to A
extending ϕ.

Let A and B be the unital k-algebras. We put a unital algebra structure on A ⊗ B

in the following way:

(a1 ⊗ b1 ).(a2 ⊗ b2 ) = a1 a2 ⊗ b1 b2 .

The unit element 1A⊗B is given by 1A ⊗ 1B , and the associativity is clear. This
multiplication is thus given by:

mA⊗B = (mA ⊗ mB ) ◦ τ23 ,

where τ23 : A ⊗ B ⊗ A ⊗ B → A ⊗ A ⊗ B ⊗ B is defined by the flip of the two

middle factors:

τ23 (a1 ⊗ b1 ⊗ a2 ⊗ b2 ) = a1 ⊗ a2 ⊗ b1 ⊗ b2 .

1.2.2. Coalgebras

Coalgebras are the objects which are somehow dual to algebras: axioms for
coalgebras are derived from axioms for algebras by reversing the arrows of the
corresponding diagrams:

A k-coalgebra is by definition a k-vector space C together with a bilinear map

Δ : C → C ⊗ C, which is coassociative . The coassociativity is expressed by the
commutativity of the following diagram:
4 Algebra and Applications 2

Δ⊗I
C ⊗C ⊗C o C ⊗C
O O
I⊗Δ Δ
Δ
C ⊗C o C

Coalgebra C is counital if there is a counit ε : C → k, such that the following

diagram commutes:

ε⊗I I⊗ε
k⊗C o C ⊗C / C⊗k
dII O u:
II ∼ ∼ uu
II Δ u
II uu
II uu
uu
C

A subspace J ⊂ C is called a subcoalgebra (respectively a left coideal, right

coideal and two-sided coideal ) of C if Δ(J) is contained in J ⊗J (respectively C ⊗J,
J ⊗ C, J ⊗ C + C ⊗ J) is included in J. The duality alluded to above can be made
more precise:

P ROPOSITION 1.1.–
1) The linear dual C ∗ of a counital coalgebra C is a unital algebra, with product
(respectively unit map) the transpose of the coproduct (respectively of the counit).
2) Let J be a linear subspace of C. Denote by J ⊥ the orthogonal of J in C ∗ .
Then:
- J is a two-sided coideal if and only if J ⊥ is a subalgebra of C ∗ .
- J is a left coideal if and only if J ⊥ is a left ideal of C ∗ .
- J is a right coideal if and only if J ⊥ is a right ideal of C ∗ .
- J is a subcoalgebra if and only if J ⊥ is a two-sided ideal of C ∗ .

P ROOF .– For any subspace K of C ∗ , we will denote by K ⊥ the subspace of those

elements of C on which any element of K vanishes. It coincides with the intersection
of the orthogonal of K with C, via the canonical embedding C −−→ C ∗∗ . Therefore,
for any linear subspaces J ⊂ C and K ⊂ C ∗ we have:

J ⊥⊥ = J, K ⊥⊥ ⊃ K.

Suppose that J is a two-sided coideal. Take any ξ, η in J ⊥ . For any x ∈ J, we

have:

< ξη, x >=< ξ ⊗ η, Δx >= 0,

 Algebraic Background for Numerical Methods, Control Theory and Renormalization 5

as Δx ⊂ J ⊗ C + C ⊗ J. Therefore, J ⊥ is a subalgebra of C ∗ . Conversely if J ⊥ is a

subalgebra, then:
ΔJ ⊂ (J ⊥ ⊗ J ⊥ )⊥ = J ⊗ C + C ⊗ J,
which proves the first assertion. We leave the reader to prove the three other assertions
along the same lines. 
Dually, we have the following:

P ROPOSITION 1.2.– Let K be a linear subspace of C ∗ . Then:

– K ⊥ is a two-sided coideal if and only if K is a subalgebra of C ∗ .
– K ⊥ is a left coideal if and only if K is a left ideal of C ∗ .
– K ⊥ is a right coideal if and only if K is a right ideal of C ∗ .
– K ⊥ is a subcoalgebra if and only if K is a two-sided ideal of C ∗ .

P ROOF .– The linear dual (C ⊗ C)∗ naturally contains the tensor product C ∗ ⊗ C ∗ .
Take as a multiplication the restriction of tΔ to C ∗ ⊗ C ∗ :
m =tΔ : C ∗ ⊗ C ∗ −→ C ∗ ,
and put u =t ε : k → C ∗ . It is easily seen, by just reverting the arrows of the
corresponding diagrams, that the coassociativity of Δ implies the associativity of m,
and that the counit property for ε implies that u is a unit. 

Note that the duality property is not perfect: if the linear dual of a coalgebra is
always an algebra, the linear dual of an algebra is generally not a coalgebra. However,
the restricted dual A◦ of an algebra A is a coalgebra. It is defined as the space of linear
forms on A vanishing on some finite-codimensional ideal (Sweedler 1969).

The coalgebra C is cocommutative if τ ◦ Δ = Δ, where τ : C ⊗ C → C ⊗ C is

the flip. It will be convenient to use Sweedler’s notation :

Δx = x1 ⊗ x2
(x)

Cocommutativity then expresses as:

 
x1 ⊗ x2 = x2 ⊗ x1
(x) (x)

In Sweedler’s notation coassociativity reads as:


(Δ ⊗ I) ◦ Δ(x) = x1:1 ⊗ x1:2 ⊗ x2
(x)

= x1 ⊗ x2:1 ⊗ x2:2 = (I ⊗ Δ) ◦ Δ(x)
(x)
6 Algebra and Applications 2

We will sometimes write the iterated coproduct as:


x1 ⊗ x2 ⊗ x3
(x)

Sometimes, we will even mix the two ways of using Sweedler’s notation for the
iterated coproduct, in the case where we want to partially keep track of how we have
constructed it (Dǎscǎlescu et al. 2001). For example,

Δ3 (x) = (I ⊗ Δ ⊗ I) ◦ (Δ ⊗ I) ◦ Δ(x)

= (I ⊗ Δ ⊗ I)( x1 ⊗ x2 ⊗ x3 )
(x)

= x1 ⊗ x2:1 ⊗ x2:2 ⊗ x3 .
(x)

With any vector space V , we can associate its tensor coalgebra T c (V ). It is

isomorphic to T (V ) as a vector space. The coproduct is given by the deconcatenation :

n 
Δ(v1 ⊗ · · · ⊗ vn ) = (v1 ⊗ · · · ⊗ vp ) (vp+1 ⊗ · · · ⊗ vn ).
p=0

The counit is given by the natural projection of T c (V ) onto k.

Let C and D be the unital k-coalgebras. We put a counital coalgebra structure on

C ⊗ D in the following way: the comultiplication is given by:

ΔC⊗D = τ23 ◦ (ΔC ⊗ ΔD ),

where τ23 is again the flip of the two middle factors, and the counity is given by
εC⊗D = εC ⊗ εD .

1.2.3. Convolution product

Let A be an algebra and C be a coalgebra (over the same field k). Then, there is
an associative product on the space L(C, A) of linear maps from C to A, called the
convolution product . It is given by:

ϕ ∗ ψ = mA ◦ (ϕ ⊗ ψ) ◦ ΔC .

In Sweedler’s notation, it reads:


ϕ ∗ ψ(x) = ϕ(x1 )ψ(x2 ).
(x)
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 7

The associativity is a direct consequence of both the associativity of A and

coassociativity of C.

1.2.4. Bialgebras and Hopf algebras

A (unital and counital) bialgebra is a vector space H endowed with a structure

of unital algebra (m, u) and a structure of counital coalgebra (Δ, ε), which are
compatible. The compatibility requirement is that Δ is an algebra morphism (or
equivalently that m is a coalgebra morphism), ε is an algebra morphism and u is
a coalgebra morphism. It is expressed by the commutativity of the three following
diagrams:
τ23
H⊗H⊗H⊗H / H⊗H⊗H⊗H
O
Δ⊗Δ m⊗m

H⊗H / H / H⊗H
m Δ

ε⊗ε u⊗u
H⊗H / k⊗k H⊗H o k⊗k
O O
m ∼ Δ ∼
 ε
 u
H / k H o k

A Hopf algebra is a bialgebra H together with a linear map S : H → H, called

the antipode , such that the following diagram commutes:
S⊗I
H⊗H / H⊗H
x< FF
Δ xx FF m
xx FF
xx FF
xx F#
ε u
H / k / H
FF x <
FF Δ m x x
FF xx
FF xx
F# xx
I⊗S
H⊗H / H⊗H

In Sweedler’s notation, it reads:

 
S(x1 )x2 = x1 S(x2 ) = (u ◦ ε)(x).
(x) (x)
8 Algebra and Applications 2

In other words, the antipode is an inverse of the identity I for the convolution
product on L(H, H). The unit for the convolution is the map u ◦ ε.

A primitive element in a bialgebra H is an element x, such that Δx = x⊗1+1⊗x.

A grouplike element is a nonzero element x, such that Δx = x⊗x. Note that grouplike
elements make sense in any coalgebra.

A bi-ideal in a bialgebra H is a two-sided ideal, which is also a two-sided coideal.

A Hopf ideal in a Hopf algebra H is a bi-ideal J, such that S(J) ⊂ J.

1.2.5. Some simple examples of Hopf algebras

1.2.5.1. The Hopf algebra of a group

Let G be a group, and let kG be the group algebra (over the field k). It is by
definition the vector space freely generated by the elements of G: the product of G
extends uniquely to a bilinear map from kG × kG into kG, hence, a multiplication
m : kG ⊗ kG → kG, which is associative. The neutral element of G gives the unit
for m. The space kG is also endowed with a counital coalgebra structure, given by:
 
Δ( λi gi ) = λi .gi ⊗ gi

and:
 
ε( λi gi ) = λi .

This defines the coalgebra of the set G: it does not take into account the extra
group structure on G, as the algebra structure does.

P ROPOSITION 1.3.– The vector space kG endowed with the algebra and coalgebra
structures defined above is a Hopf algebra. The antipode is given by:

S(g) = g −1 , g ∈ G.

P ROOF .– The compatibility of the product and the coproduct is an immediate

consequence of the following computation: for any g, h ∈ G, we have:

Δ(gh) = gh ⊗ gh = (g ⊗ g)(h ⊗ h) = Δg.Δh.

Now, m(S ⊗ I)Δ(g) = g −1 g = e and similarly for m(I ⊗ S)Δ(g). But,

e = u ◦ ε(g) for any g ∈ G, so the map S is indeed the antipode. 

R EMARK 1.1.– If G were only a semigroup, the same construction would lead to
a bialgebra structure on kG: the Hopf algebra structure (i.e. the existence of an
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 9

antipode) reflects the group structure (the existence of the inverse). We have S 2 = I
in this case; however, the involutivity of the antipode is not true for general Hopf
algebras.

1.2.5.2. Tensor algebras

There is a natural structure of cocommutative Hopf algebra on the tensor algebra
T (V ) of any vector space V . Namely, we define the coproduct Δ as the unique algebra
morphism from T (V ) into T (V ) ⊗ T (V ), such that:

Δ(1) = 1 ⊗ 1, Δ(x) = x ⊗ 1 + 1 ⊗ x, x ∈ V.

We define the counit as the algebra morphism, such that ε(1) = 1 and ε| = 0.
V
This endows T (V ) with a cocommutative bialgebra structure. We claim that the
principal anti-automorphism:

S(x1 ⊗ · · · ⊗ xn ) = (−1)n xn ⊗ · · · ⊗ x1

verifies the axioms of an antipode, so that T (V ) is indeed a Hopf algebra. For x ∈ V ,

we have S(x) = −x; hence, S ∗ I(x) = I ∗ S(x) = 0. As V generates T (V ) as an
algebra, it is easy to conclude.

1.2.5.3. Enveloping algebras

Let g be a Lie algebra. The universal enveloping algebra is the quotient of the
tensor algebra T (g) by the ideal J generated by x ⊗ y − y ⊗ x − [x, y], x, y ∈ g.

L EMMA 1.1.– J is a Hopf ideal, that is, Δ(J) ⊂ H ⊗ J + J ⊗ H and S(J) ⊂ J.

P ROOF .– The ideal J is generated by primitive elements (according to Proposition

1.5 below), and any ideal generated by primitive elements is a Hopf ideal (very easy
and left to the reader). 

The quotient of a Hopf algebra by a Hopf ideal is a Hopf algebra. Hence, the universal
enveloping algebra U(g) is a cocommutative Hopf algebra.

1.2.6. Some basic properties of Hopf algebras

In the proposition below we summarize the main properties of the antipode in a

Hopf algebra:

P ROPOSITION 1.4.– (see Sweedler (1969, Proposition 4.0.1)). Let H be a Hopf

algebra with multiplication m, comultiplication Δ, unit u : 1 → 1, counit ε and
antipode S. Then:
1) S ◦ u = u and ε ◦ S = ε.
10 Algebra and Applications 2

2) S is an algebra antimorphism and a coalgebra antimorphism, that is, if τ

denotes the flip, we have:

m ◦ (S ⊗ S) ◦ τ = S ◦ m, τ ◦ (S ⊗ S) ◦ Δ = Δ ◦ S.

3) If H is commutative or cocommutative, then S 2 = I.

For a detailed proof, see Kassel (1995).

P ROPOSITION 1.5.–
1) If x is a primitive element, then S(x) = −x.
2) The linear subspace Prim H of primitive elements in H is a Lie algebra.
P ROOF .– If x is primitive, then (ε ⊗ ε) ◦ Δ(x) = 2ε(x). On the contrary,
(ε ⊗ ε) ◦ Δ(x) = ε(x), so ε(x) = 0. Then:

0 = (u ◦ ε)(x) = m(S ⊗ I)Δ(x) = S(x) + x.

Now let x and y be the primitive elements of H. Then, we can easily compute:

Δ(xy − yx) = (x ⊗ 1 + 1 ⊗ x)(y ⊗ 1 + 1 ⊗ y) − (y ⊗ 1 + 1 ⊗ y)(x ⊗ 1 + 1 ⊗ x)

= (xy − yx) ⊗ 1 + 1 ⊗ (xy + yx) + x ⊗ y + y ⊗ x − y ⊗ x − x ⊗ y
= (xy − yx) ⊗ 1 + 1 ⊗ (xy − yx).

1.3. Connected Hopf algebras

We introduce the crucial property of connectedness for bialgebras. The main

interest resides in the possibility of implementing recursive procedures in connected
bialgebras, the induction taking place with respect to a filtration or a grading. An
important example of these techniques is the recursive construction of the antipode,
which then “comes for free”, showing that any connected bialgebra is in fact a
connected Hopf algebra.

1.3.1. Connected graded bialgebras

A graded Hopf algebra on k is a graded k-vector space:


H= Hn
n≥0
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 11

endowed with a product m : H ⊗ H → H, a coproduct Δ : H → H ⊗ H, a unit

u : k → H, a counit ε : H → k and an antipode S : H → H, fulfilling the usual
axioms of a Hopf algebra, and such that:
Hp .Hq ⊂ Hp+q [1.1]

Δ(Hn ) ⊂ Hp ⊗ Hq . [1.2]
p+q=n

If we do not ask for the existence of an antipode H, we get the definition of a

graded bialgebra . In a graded bialgebra H, we will consider the increasing filtration:

n
Hn = Hp .
p=0

It is an easy exercise (left to the reader) to prove that the unit u and the counit ε
are degree zero maps, that is, 1 ∈ H0 and ε(Hn ) = {0} for n ≥ 1. We can also
show that the antipode S, when it exists, is also of degree zero, that is, S(Hn ) ⊂ Hn .
It can be proven as follows: let S  : H → H be defined, so that S  (x) is the nth
homogeneous component of S(x) when x is homogeneous of degree n. We can write
down the coproduct Δ(x) with Sweedler’s notation:

Δ(x) = x1 ⊗ x2 ,
(x)

where x1 and x2 are the homogeneous of degree, say, k and n − k. We then have:

m ◦ (S  ⊗ Id) ◦ Δ(x) = S  (x1 )x2
(x)

= nth component of S(x1 )x2 = ε(x)1. [1.3]
(x)

Similarly, m ◦ (Id ⊗S  ) ◦ Δ(x) = ε(x)1. By uniqueness of the antipode, we then

get S  = S.

Suppose that H is connected , that is, H0 is one-dimensional. Then, we have:


Ker ε = Hn .
n≥1

P ROPOSITION 1.6.– For any x ∈ Hn , n ≥ 1, we can write:



Δx = x ⊗ 1 + 1 ⊗ x + Δx,  ∈
Δx Hp ⊗ Hq .
p+q=n, p=0, q=0

 is coassociative on Ker ε and Δ

The map Δ  k = (I ⊗k−1 ⊗ Δ)(I
 ⊗k−2 ⊗ Δ)...
 Δ 
n−k ⊗k+1
sends H into (H
n
) .
12 Algebra and Applications 2

P ROOF .– Thanks to connectedness we can clearly write:


Δx = a(x ⊗ 1) + b(1 ⊗ x) + Δx
 ∈ Ker ε ⊗ Ker ε. The counit property then tells us that, with
with a, b ∈ k and Δx
k ⊗ H and H ⊗ k canonically identified with H:
x = (ε ⊗ I)(Δx) = bx, x = (I ⊗ ε)(Δx) = ax, [1.4]
hence, a = b = 1. We will use the following two variants of Sweedler’s notation:

Δx = x1 ⊗ x2 , [1.5]
(x)

 =
Δx x ⊗ x , [1.6]
(x)

the second being relevant only for x ∈ Ker ε. If x is homogeneous of degree n, we can
suppose that the components x1 , x2 , x , x in the expressions above are homogeneous
as well, and we then have |x1 | + |x2 | = n and |x | + |x | = n. We easily compute:
(Δ ⊗ I)Δ(x) = x ⊗ 1 ⊗ 1 + 1 ⊗ x ⊗ 1 + 1 ⊗ 1 ⊗ x

+ x ⊗ x ⊗ 1 + x ⊗ 1 ⊗ x + 1 ⊗ x ⊗ x
(x)

 ⊗ I)Δ(x)
+ (Δ 

and
(I ⊗ Δ)Δ(x) = x ⊗ 1 ⊗ 1 + 1 ⊗ x ⊗ 1 + 1 ⊗ 1 ⊗ x

+ x ⊗ x ⊗ 1 + x ⊗ 1 ⊗ x + 1 ⊗ x ⊗ x
(x)

 Δ(x),
+ (I ⊗ Δ) 

hence, the coassociativity of Δ comes from the one of Δ. Finally, it is easily seen by
induction on k that for any x ∈ Hn , we can write:

 k (x) =
Δ x(1) ⊗ · · · ⊗ x(k+1) , [1.7]
x

with |x(j) | ≥ 1. The grading imposes:


k+1
|x(j) | = n,
j=1

so the maximum possible for any degree |x(j) | is n − k. 

 Algebraic Background for Numerical Methods, Control Theory and Renormalization 13

1.3.2. An example: the Hopf algebra of decorated rooted trees

A rooted tree is an oriented graph with a finite number of vertices, one among
them being distinguished as the root, such that any vertex admits exactly one outgoing
edge, except the root which has no outgoing edges. Here is the list of rooted trees up
to five vertices, where the edges are tacitly oriented from top to bottom:

A rooted forest is a finite collection of rooted trees. The Connes–Kreimer Hopf

 (n)
algebra HCK = n≥0 HCK is the Hopf algebra of rooted forests over k, graded by the
number of vertices. It is the free commutative algebra on the linear space T spanned
by nonempty rooted trees. The coproduct on a rooted forest u (i.e. a product of rooted
trees) is described as follows: the set U of vertices of a forest u is endowed with a
partial order defined by x ≤ y if and only if there is a path from a root to y passing
through x. Any subset W of the set of vertices U of u defines a subforest w of u in
an obvious manner, that is, by keeping the edges of u which link two elements of W .
The coproduct is then defined by:

ΔCK (u) = v ⊗ w. [1.8]
V W =U
W <V

Here, the notation W < V means that y < x for any vertex x of v and any vertex
y of w, such that x and y are comparable. Such a couple (V, W ) is also called an
admissible cut , with crown (or pruning) v and trunk w. We have, for example:
 
ΔCK = ⊗1+1⊗ + ⊗
 
ΔCK = ⊗1+1⊗ +2 ⊗ + ⊗

With the restriction that V and W will be nonempty (i.e. if V and W give rise to
an ordered partition of U into two blocks), we get the restricted coproduct:

 CK (u) = ΔCK (u) − u ⊗ 1 − 1 ⊗ u =
Δ v ⊗ w, [1.9]
V W =U
W <V, V,W =∅

which is often displayed as (u) u ⊗u in Sweedler’s notation. The iterated restricted
coproduct, in terms of ordered partitions of U into n blocks, writes:

 n−1
Δ CK (u) = v1 ⊗ · · · ⊗ vn , [1.10]
V1 ···Vn =U
Vn <···<V1 , Vj =∅

and we get the full-iterated coproduct Δn−1CK (u) by allowing empty blocks in the
formula above. The coassociativity of the coproduct follows immediately.
14 Algebra and Applications 2

Note, however, that the relation < on the subsets of vertices is not transitive. The
notation Vn < · · · < V1 is to be understood as Vi < Vj for any i > j, i, j ∈
{1, . . . , n}.

1.3.3. Connected filtered bialgebras

A filtered bialgebra on k is a k-vector space together with an increasing N-indexed

filtration:
H0 ⊂ H 1 ⊂ · · · ⊂ H n ⊂ · · · , Hn = H
n

endowed with a product m : H ⊗ H → H, a coproduct Δ : H → H ⊗ H, a unit

u : k → H, a counit ε : H → k and an antipode S : H → H, fulfilling the usual
axioms of a bialgebra, and such that:
Hp .Hq ⊂ Hp+q [1.11]

Δ(Hn ) ⊂ Hp ⊗ Hq . [1.12]
p+q=n

It is easy (and left to the reader) to show that the unit u and the counit ε are of
degree zero, if we consider the filtration on the base field k given by k 0 = {0} and
k n = k for any n ≥ 1. Namely, u(k n ) ⊂ Hn and ε(Hn ) ⊂ k n for any n ≥ 0.

If we ask for the existence of an antipode S, we get the definition of a filtered Hopf
algebra if the antipode is of degree zero, that is, if:
S(Hn ) ⊂ Hn [1.13]
for any n ≥ 0. Contrarily to the graded case, it is not likely that a filtered bialgebra
with antipode is automatically a filtered Hopf algebra (see, for example, Montgomery
(1993, Lemma 5.2.8), Andruskiewitsch and Schneider (2002) and Andruskiewitsch
and Cuadra (2013)). The antipode is, however, of degree zero in the connected case:
for any x ∈ H, we set
|x| := min{n ∈ N, x ∈ Hn }. [1.14]
We say that the filtered bialgebra H is connected if H is one-dimensional. There
0

is an analogue of Proposition 1.6 in the connected filtered case, the proof of which is
very similar:

P ROPOSITION 1.7.– For any x ∈ Hn , n ≥ 1, we can write:



Δx = x ⊗ 1 + 1 ⊗ x + Δx,  ∈
Δx Hp ⊗ Hq . [1.15]
p+q=n, p=0, q=0

 is coassociative on Ker ε and Δ

The map Δ  ⊗k−2 ⊗ Δ)...
 k = (I ⊗k−1 ⊗ Δ)(I  Δ 
n−k ⊗k+1
sends H into (H
n
) .
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 15

As an easy corollary, the degree of the antipode is also zero in the connected case,
that is, S(Hn ) ⊆ Hn for any n. This is an immediate consequence of the recursive
formulae [1.19] and [1.20] below.

Any graded bialgebra, or the Hopf algebra, is obviously filtered by the canonical
filtration associated with the grading:

n
Hn := Hi , [1.16]
i=0

and in that case, if x is a homogeneous element, x is of degree n if and only if |x| = n.

1.3.4. The convolution product

An important result is that any connected filtered bialgebra is indeed a filtered

Hopf algebra, in the sense that the antipode comes for free. We give a proof of this
fact as well as a recursive formula for the antipode using the convolution product : let
H be a (connected filtered) bialgebra, and let A be any k-algebra (which will be called
the target algebra ): the convolution product on L(H, A) is given by:

ϕ ∗ ψ(x) = mA (ϕ ⊗ ψ)Δ(x)

= ϕ(x1 )ψ(x2 ).
(x)

P ROPOSITION 1.8.– The map e = uA ◦ ε, given by e(1) = 1A and e(x) = 0 for any
x ∈ Ker ε, is a unit for the convolution product. Moreover, the set G(A) := {ϕ ∈
L(H, A), ϕ(1) = 1A } endowed with the convolution product is a group.

P ROOF .– The first statement is straightforward. To prove the second, let us consider
the formal series:
 ∗−1
ϕ∗−1 (x) = e − (e − ϕ) (x)

= (e − ϕ)∗k (x).
k≥0

Using (e − ϕ)(1) = 0, we have (e − ϕ)∗k (1) = 0 immediately, and for any

x ∈ Ker ε:

(e − ϕ)∗k (x) = (−1)k mA,k−1 (ϕ ⊗ · · · ⊗ ϕ)Δ (x). [1.17]
 k−1
ktimes

When x ∈ Hn , this expression vanishes and then for k ≥ n + 1. The formal series
then ends up with a finite number of terms for any x, which proves the result. 
16 Algebra and Applications 2

C OROLLARY 1.1.– Any connected filtered bialgebra H is a filtered Hopf algebra. The
antipode is defined by:

S(x) = (uε − I)∗k (x). [1.18]
k≥0

It is given by S(1) = 1 and recursively by any of the two formulae for x ∈ Ker ε:

S(x) = −x − S(x )x [1.19]
(x)

S(x) = −x − x S(x ). [1.20]
(x)

P ROOF .– The antipode, when it exists, is the inverse of the identity for the
convolution product on L(H, H). We then just need to apply Proposition 1.8 with
A = H. The two recursive formulas come directly from the two equalities:

m(S ⊗ I)Δ(x) = m(I ⊗ S)Δ(x) = 0

fulfilled by any x ∈ Ker ε. 

Let g(A) be the subspace of L(H, A) formed by the elements α, such that
α(1) = 0. It is clearly a subalgebra of L(H, A) for the convolution product. We
have:

G(A) = e + g(A).

From now on, we will suppose that the ground field k is of characteristic zero. For
any x ∈ Hn , the exponential:
 α∗k (x)
e∗α (x) = [1.21]
k!
k≥0

is a finite sum (ending up at k = n). It is a bijection from g(A) onto G(A). Its inverse
is given by:
 (−1)k−1
Log(1 + α)(x) = α∗k (x). [1.22]
k
k≥1

This sum again ends up at k = n for any x ∈ Hn . Let us introduce a decreasing

filtration on L = L(H, A):

Ln := {α ∈ L, α| n−1 = 0}. [1.23]

H
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 17

Clearly, L0 = L and L1 = g(A). We define the valuation val ϕ of an element ϕ

of L as the greatest integer k, such that ϕ is in Lk . In the sequel we will consider the
ultrametric distance on L induced by the filtration:

d(ϕ, ψ) = 2− val(ϕ−ψ) . [1.24]

For any α, β ∈ g(A) let [α, β] = α ∗ β − β ∗ α.

P ROPOSITION 1.9.– We have the inclusion:

Lp ∗ Lq ⊂ Lp+q , [1.25]

and moreover, the metric space L endowed with the distance defined by [1.24] is
complete.

P ROOF .– Take any x ∈ Hp+q−1 , and any α ∈ Lp and β ∈ Lq . We have:


(α ∗ β)(x) = α(x1 )β(x2 ).
(x)

Recall that we denote by |x| the minimal n, such that x ∈ Hn . Since |x1 | + |x2 | =
|x| ≤ p + q − 1, either |x1 | ≤ p − 1 or |x2 | ≤ q − 1, so the expression vanishes. Now,
if (ψn ) is a Cauchy sequence in L, it is seen immediately that this sequence is locally
stationary , that is, for any x ∈ H there exists N (x) ∈ N, such that ψn (x) = ψN (x) (x)
for any n ≥ N (x). Then, the limit of (ψn ) exists and is clearly defined by:

ψ(x) = ψN (x) (x).


As a corollary, the Lie algebra L1 = g(A) is pro-nilpotent, in a sense that it is the
projective limit of the Lie algebras g(A)/Ln , which are nilpotent.

1.3.5. Characters

Let H be a connected filtered Hopf algebra over k, and let A be a k-algebra.

We will consider unital algebra morphisms from H to the target algebra A. When
the algebra A is commutative, we will call them, slightly abusively, characters . We
recover, of course, the usual notion of character when the algebra A is the ground
field k.

The notion of character involves only the algebra structure of H. On the contrary,
the convolution product on L(H, A) involves only the coalgebra structure on H. Let
us now consider the full Hopf algebra structure on H and see what happens to algebra
morphisms with the convolution product:
18 Algebra and Applications 2

P ROPOSITION 1.10.– Let H be any Hopf algebra over k, and let A be a commutative
k-algebra. Then, the characters from H to A form a group G1 (A) under the
convolution product, and for any ϕ ∈ G1 (A), the inverse is given by:

ϕ∗−1 = ϕ ◦ S. [1.26]

P ROOF .– Using the fact that Δ is an algebra morphism, we have for any x, y ∈ H:

f ∗ g(xy) = f (x1 y1 )g(x2 y2 ).
(x)(y)

If A is commutative and if f and g are characters, we get:


f ∗ g(xy) = f (x1 )f (y1 )g(x2 )g(y2 )
(x)(y)

= f (x1 )g(x2 )f (y1 )g(y2 )
(x)(y)

= (f ∗ g)(x)(f ∗ g)(y).

The unit e = uA ◦ ε is an algebra morphism. The formula for the inverse of a

character comes easily from the commutativity of the following diagram:
(f ◦S)∗f
d c b ` _ ^ \ [ Z fX⊗fW
f
H ⊗ Hk j h g
S⊗I
/ H⊗H V T / ⊗A
< FF S R
Δ xx n x l FF m P OEEEE mA
xx o FFF NEEE
xxrx p a ` ` ` ` _ _ _ ^ ^ F
^F ^ ] ] ] \ \ \ [ M EE
x c c b b b a a # K"
H
c / k e / H [ [ [/%- A
FFM
FF NΔ ε u
m xx
x< mA r ys<9
yy
f
FF O xx p y
FF P xx o yy
F# R S xx f ⊗f k l n yy
T V I⊗S
/ H⊗H h j / A⊗A
H⊗H W X Z
[ \ ^ _ ` b c d f g
f ∗(f ◦S)


We call infinitesimal characters with values in the algebra A those elements α of
L(H, A), such that:

α(xy) = e(x)α(y) + α(x)e(y).

P ROPOSITION 1.11.– Let H be a connected filtered Hopf algebra, and suppose that
A is a commutative algebra. Let G1 (A) (respectively g1 (A)) be the set of characters
of H with values in A (respectively the set of infinitesimal characters of H with values
in A). Then, G1 (A) is a subgroup of G, the exponential restricts to a bijection from
g1 (A) onto G1 (A), and g1 (A) is a Lie subalgebra of g(A).
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 19

P ROOF .– Take two infinitesimal characters α and β with values in A and compute:

(α ∗ β)(xy) = α(x1 y1 )β(x2 y2 )
(x)(y)
   
= α(x1 )e(y1 ) + e(x1 )α(y1 ) . β(x2 )e(y2 ) + e(x2 )α(y2 )
(x)(y)

= (α ∗ β)(x)e(y) + α(x)β(y) + β(x)α(y) + e(x)(α ∗ β)(y).

Using the commutativity of A, we immediately get:

[α, β](xy) = [α, β](x)e(y) + e(x)[α, β](y),

which shows that g1 (A) is a Lie algebra. Now, for α ∈ g1 (A), we have:
n  

∗n n ∗k
α (xy) = α (x)α∗(n−k) (y),
k
k=0

as easily seen by induction on n. A straightforward computation then yields:

exp(α)(xy) = exp(α)(x) exp(α)(y),

with
 α∗k α∗2
exp α := =e+α+ + ··· .
k! 2
k≥0

The series above makes sense thanks to connectedness, as explained in section

(−1)j−1 γ ∗j
1.3.4. Now let ϕ = e + γ ∈ G1 (A), and let log(ϕ) = j≥1 j . Set
ϕ∗t := exp(t log ϕ) for t ∈ k. It coincides with the nth convolution power of ϕ
for any integer n. Hence, ϕ∗t is an A-valued character of H for any t ∈ k. Indeed, for
any x, y ∈ H, the expression ϕ∗t (xy) − ϕ∗t (x)ϕ∗t (y) is polynomial in t and vanishes
on all integers, and hence, vanishes identically. Differentiating with respect to t at
t = 0, we immediately find that log ϕ is an infinitesimal character. 

1.3.6. Group schemes and the Cartier–Milnor–Moore–Quillen theorem

T HEOREM 1.1 (Cartier, Milnor, Moore, Quillen).– Let U be a cocommutative

connected filtered Hopf algebra and let g be the Lie algebra of its primitive elements,
endowed with the filtration induced by the one of U, which in turns induces a filtration
on the enveloping algebra U(g). Then, U and U(g) are the isomorphic as filtered Hopf
algebras. If U is graded, then the two Hopf algebras are isomorphic as graded Hopf
algebras.
20 Algebra and Applications 2

P ROOF .– The following proof is borrowed from Foissy’s thesis. The embedding
ι : g → U obviously induces an algebra morphism

ϕ : U(g) −→ U. [1.27]

It is easy to show that ϕ is also a coalgebra morphism. It remains to show that ϕ

is surjective, injective and respects the filtrations. Let us first prove the surjectivity by
induction on the coradical filtration degree:
 n (x) = 0}.
d(x) := min{n ∈ N, Δ [1.28]

Set Un := {x ∈ U, d(x) ≤ n}, and similarly for U(g). We can limit ourselves to
the kernel of the counit. Any x ∈ U1 ∩ Ker ε is primitive, hence ϕ : U(gg)1 → U1 is
obviously a linear isomorphism. Now, for x ∈ Un ∩ Ker ε (for some integer n ≥ 2),
we can write, using cocommutativity:

 n−1 (x) =
Δ x(1) ⊗ · · · ⊗ x(n)
(x)

1   (σ1 )
= x ⊗ · · · ⊗ x(σn )
n!
σ∈Sn (x)

where the x(j) s are of coradical filtration degree 1, hence primitive. But, we also have:

 n−1 (x(1) · · · x(n) ) =
Δ x(σ1 ) ⊗ · · · ⊗ x(σn ) . [1.29]
σ∈Sn

Hence, the element y = x− n!1

(x) x
(1)
⊗· · ·⊗x(n) belongs to Un−1 . It is a linear
combination of products of primitive elements by induction hypothesis, hence so is x.
We have thus proven that U is generated by g, which amounts to the surjectivity of ϕ.

Now consider a nonzero element u ∈ U(g), such that ϕ(u) = 0, and such that
d(u) is minimal. We have already proven d(u) ≥ 2. We now compute:
 
0 = Δ ϕ(u) = (ϕ ⊗ ϕ)Δ(u)
⎛ ⎞

= (ϕ ⊗ ϕ) ⎝u ⊗ 1 + 1 ⊗ u + u ⊗ u ⎠
(x)

= ϕ(u ) ⊗ ϕ(u ).
(u)

By minimality hypothesis on d(u), we then get (u) u ⊗ u = 0. Hence, u is

primitive, that is, d(u) = 1, a contradiction. Hence, ϕ is injective. The compatibility
with the original filtration or graduation is obvious. 
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 21


Now, let H : n≥0 H be a connected filtered Hopf algebra and let A be
n

a commutative unital algebra. We suppose that the components of the filtration

are finite-dimensional. The group G1 (A) defined in the previous section depends
functorially on the target algebra A: in particular, when the Hopf algebra H itself is
commutative, the correspondence A → G1 (A) is a group scheme . In the graded case
with finite-dimensional components, it is possible to reconstruct the Hopf algebra H
from the group scheme. We have indeed:

P ROPOSITION 1.12.–
  ◦
H = U g1 (k) ,

where g1 (k) is theLie algebra of infinitesimal characters with values in the base field
k, where U g1 (k) stands for its enveloping algebra, and (−)◦ stands for the graded
dual.

In the case when the Hopf algebra H is not commutative, it is no longer possible
to reconstruct it from G1 (k).

1.3.7. Renormalization in connected filtered Hopf algebras

In this section we describe the renormalization à la Connes–Kreimer (Connes

and Kreimer 1998; Kreimer 2002) in the abstract context of connected filtered Hopf
algebras: the objects to be renormalized are characters with values in a commutative
unital target algebra A endowed with a renormalization scheme , that is, a splitting
A = A− ⊕ A+ into two subalgebras. An important example is given by the minimal
subtraction scheme of the algebra A of meromorphic functions of one variable z,
where A+ is the algebra of meromorphic functions which are holomorphic at z = 0,
and A− = z −1 C[z −1 ] stands for the “polar parts”. Any A-valued character ϕ admits
a unique Birkhoff decomposition :

ϕ = ϕ∗−1
− ∗ ϕ+ ,

where ϕ+ is an A+ -valued character, and ϕ(Ker ε) ⊂ A− . In the MS scheme case

described above, the renormalized character is the scalar-valued character given by the
evaluation of ϕ+ at z = 0 (whereas the evaluation of ϕ at z = 0 does not necessarily
make sense).

Here, we consider the situation where the algebra A admits a renormalization

scheme , that is, a splitting into two subalgebras:

A = A− ⊕ A+

with 1 ∈ A+ . Let π : A −→
→ A− be the projection on A− parallel to A+ .
22 Algebra and Applications 2

T HEOREM 1.2.–
1) Let H be a connected filtered Hopf algebra. Let G(A) be the group of those
ϕ ∈ L(H, A), such that ϕ(1) = 1A endowed with the convolution product. Any
ϕ ∈ G(A) admits a unique Birkhoff decomposition:

ϕ = ϕ∗−1
− ∗ ϕ+ , [1.30]

where ϕ− sends 1 to 1A and Ker ε into A− , and ϕ+ sends H into A+ . The maps
ϕ− and ϕ+ are given on Ker ε by the following recursive formulae:
  
ϕ− (x) = −π ϕ(x) + ϕ− (x )ϕ(x ) [1.31]
(x)
  
ϕ+ (x) = (I − π) ϕ(x) + ϕ− (x )ϕ(x ) . [1.32]
(x)

2) If the algebra A is commutative and if ϕ is a character, the components ϕ− and

ϕ+ occurring in the Birkhoff decomposition of χ are characters as well.

P ROOF .– The proof goes along the same lines as the proof of Theorem 4 from Connes
and Kreimer (1998): for the first assertion, it is immediate from the definition of π that
ϕ− sends Ker ε into A− , and that ϕ+ sends Ker ε into A+ . It only remains to check
equality ϕ+ = ϕ− ∗ ϕ, which is an easy computation:
  
ϕ+ (x) = (I − π) ϕ(x) + ϕ− (x )ϕ(x ) .
(x)

= ϕ(x) + ϕ− (x) + ϕ− (x )ϕ(x )
(x)

= (ϕ− ∗ ϕ)(x).

The proof of assertion 2 goes exactly as in Connes and Kreimer (1998) and relies
on the following Rota–Baxter relation in A:
 
π(a)π(b) = π π(a)b + aπ(b) − π(ab), [1.33]

which is easily verified by decomposing a and b into their A± -parts. Let ϕ be a

character of H with values in A. Suppose that we have ϕ− (xy) = ϕ− (x)ϕ− (y) for
any x, y ∈ H, such that |x|+|y| ≤ d−1, and compute for x, y, such that |x|+|y| = d:

ϕ− (x)ϕ− (y) = π(X)π(Y ),

 
with X = ϕ(x) − (x) ϕ− (x )ϕ(x ) and Y = ϕ(y) − (y) ϕ− (y  )ϕ(y  ). Using the
formula:

π(X) = −ϕ− (x),

 Algebraic Background for Numerical Methods, Control Theory and Renormalization 23

we get:
 
ϕ− (x)ϕ− (y) = −π XY + ϕ− (x)Y + Xϕ− (y) ,

hence:

ϕ− (x)ϕ− (y) = −π ϕ(x)ϕ(y) + ϕ− (x)ϕ(y) + ϕ(x)ϕ− (y)
  
+ ϕ− (x )ϕ(x ) ϕ(y) + ϕ− (y)
(x)
 
+ ϕ(x) + ϕ− (x) ϕ− (y  )ϕ(y  )
(y)
 
+ ϕ− (x )ϕ(x )ϕ− (y  )ϕ(y  ) .
(x)(y)

We have to compare this expression with:


ϕ− (xy) = −π ϕ(xy) + ϕ− (x)ϕ(y) + ϕ− (y)ϕ(x)
 
+ ϕ− (x y)ϕ(x ) + ϕ− (x )ϕ(x y)
(x)
 
+ ϕ− (xy  )ϕ(y  ) + ϕ− (y  )ϕ(xy  )
(y)
 
+ ϕ− (x y  )ϕ(x y  ) .
(x)(y)

These two expressions are easily seen to be equal using the commutativity of the
algebra A, the character property for ϕ and the induction hypothesis. 

R EMARK 1.2.– Assertion 2 admits a more conceptual proof (see the notes by
Ebrahimi-Fard in the present volume), which is based on the following recursive
expressions for the components of the Birkhoff decomposition: define the Bogoliubov
preparation map as the map b : G(A) → L(H, A), recursively given by:

b(ϕ)(x) = ϕ(x) + ϕ− (x )ϕ(x ). [1.34]
(x)

Then, the components of ϕ in the Birkhoff decomposition read:

ϕ− = −π ◦ b(ϕ), ϕ+ = (I − π) ◦ b(ϕ). [1.35]

The Bogoliubov preparation map also writes in a more concise form:

b(ϕ) = ϕ− ∗ (ϕ − e). [1.36]

24 Algebra and Applications 2

Plugging equation [1.36] into equation [1.35] and setting α := e − ϕ, we get the
following expression for ϕ− :

ϕ− = e + P (ϕ− ∗ α) [1.37]
     
= e + P (α) + P P (α) ∗ α + · · · + P P ...P ( α) ∗ α · · · ∗ α + · · ·

n times

[1.38]

where P : L(H, A) → L(H, A) is the projection defined by P (α) = π ◦ α. The

renormalized part ϕ+ satisfies an analogous recursive expression:

ϕ+ = e + P̃ (ϕ− ∗ α) [1.39]
= e + P̃ (ϕ+ ∗ β) [1.40]
     
= e + P̃ (β) + P̃ P̃ (β) ∗ β + · · · + P̃ P̃ . . . P̃ ( β) ∗ β · · · ∗ β + · · ·

n times

with β := ϕ−1 ∗ α = e − ϕ−1 , and where P̃ = I − P is the projection on L(H, A)

defined by P̃ (α) = (I − π) ◦ α.

1.4. Pre-Lie algebras

Pre-Lie algebras are sometimes called Vinberg algebras , as they appear in the work
of Vinberg (1963) under the name “left-symmetric algebras” on the classification
of homogeneous cones. They appear independently at the same time in the work
of Gerstenhaber (1963) on Hochschild cohomology and deformations of algebras,
under the name “pre-Lie algebras”, which is now the standard terminology. The
term “chronological algebras” has also been used sometimes, for example, in the
fundamental work of Agrachev and Gamkrelidze (1981). The notion itself can,
however, be traced back to the work of Cayley (1857) which, in modern language,
describes the pre-Lie algebra morphism Fa from the pre-Lie algebra of rooted trees
into the pre-Lie algebra of vector fields on Rn , sending the one-vertex tree to a given
vector field a. For a survey emphasizing geometric aspects, see Burde (2006).

1.4.1. Definition and general properties

A left pre-Lie algebra over a field k is a k-vector space A with a bilinear binary
composition that satisfies the left pre-Lie identity:

a (b c) − (a b) c=b (a c) − (b a) c, [1.41]
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 25

for a, b, c ∈ A. Analogously, a right pre-Lie algebra is a k-vector space A with a

binary composition  that satisfies the right pre-Lie identity:

(a  b)  c − a  (b  c) = (a  c)  b − a  (c  b). [1.42]

The left pre-Lie identity is rewritten as:

L[a,b] = [La , Lb ], [1.43]

where La : A → A is defined by La b = a b, and the bracket on the left-hand side
is defined by [a, b] := a b − b a. As an easy consequence, this bracket satisfies
the Jacobi identity: If A is unital (i.e. there exists 1 ∈ A, such that 1 a = a 1 = 1
for any a ∈ A), it is immediate thanks to the fact that L : A → End A is injective.
If not, we can add a unit by considering A := A ⊕ k.1 and extend L : A → End A
accordingly. As any right pre-Lie algebra (A, ) is also a left pre-Lie algebra with
product a b := b  a, we can stick to left pre-Lie algebras, which we will do unless
specifically indicated.

1.4.2. The group of formal flows

The following is taken from the paper by Agrachev and Gamkrelidze (1981).
Suppose that A is a left pre-Lie algebra endowed with a compatible decreasing
filtration, namely, A = A1 ⊃ A2 ⊂ A3 ⊃ · · · , such that the intersection of the
Aj ’s reduces to {0}, and such that Ap Aq ⊂ Ap+q . Suppose, moreover, that A is
complete with respect to this filtration. The Baker–Campbell–Hausdorff formula:
1 1
C(a, b) = a + b + [a, b] + ([a, [a, b]] + [b, [b, a]]) + · · · [1.44]
2 12
then endows A with a structure of a pro-unipotent group. An example of this situation
is given by A = hB[[h]], where B is any pre-Lie algebra, and Aj = hj B[[h]]. This
group admits a more transparent presentation as follows: introduce a fictitious unit 1,
such that 1 a = a 1 = a for any a ∈ A, and define W : A → A by:
1 1
W (a) := eLa 1 − 1 = a + a a+ a (a a) + · · · . [1.45]
2 6
The application W is clearly a bijection. The inverse, denoted by Ω, also appears
under the name “pre-Lie Magnus expansion” in Ebrahimi-Fard and Manchon (2009b).
It verifies the equation:
LΩ(a) 
Ω(a) = a= Bi LiΩ(a) a, [1.46]
e − Id
LΩ(a)
i≥0

where the Bi s are the Bernoulli numbers. The first few terms are:
1 1 1
Ω(a) = a − a a + (a a) a+ a (a a) + · · · [1.47]
2 4 12
26 Algebra and Applications 2

By transferring the BCH product by means of the map W , namely:

  
a#b = W C Ω(a), Ω(b) , [1.48]
 
we have W (a)#W (b) = W C(a, b) = eLa eLb 1 − 1, hence W (a)#W (b) =
W (a) + eLa W (b). The product # is thus given by the simple formula:
a#b = a + eLΩ(a) b. [1.49]


The inverse is given by a#−1 = W −Ω(a) = e−LΩ(a) 1−1. If (A, ) and (B, )
are two such pre-Lie algebras and ψ : A → B is a filtration-preserving pre-Lie algebra
morphism, we should immediately check that for any a, b ∈ A we have:
ψ(a#b) = ψ(a)#ψ(b). [1.50]
In other words, the group of formal flows is a functor from the category of
complete filtered pre-Lie algebras to the category of groups.

When the pre-Lie product is associative, all of this simplifies to:

a#b = a b+a+b [1.51]
and
1 
a#−1 = −1= (−1)n an . [1.52]
1+a
n≥1

1.4.3. The pre-Lie Poincaré–Birkhoff–Witt theorem

This section exposes a result by Guin and Oudom (2005).

T HEOREM 1.3.– Let A be any left pre-Lie algebra, and let S(A) be its symmetric
algebra, that is, the free commutative algebra on A. Let ALie be the underlying Lie
algebra of A, that is, the vector space A endowed with the Lie bracket given by [a, b] =
a b−b a for any a, b ∈ A, and let U(A) be the enveloping algebra of ALie , endowed
with its usual increasing filtration. Let us consider the associative algebra U(A) as a
left module over itself.

There exists a left U(A)-module structure on S(A) and a canonical left

U(A)-module isomorphism η : U(A) → S(A), such that the associated graded linear
map Gr η : Gr U(A) → S(A) is an isomorphism of commutative graded algebras.

P ROOF .– The Lie algebra morphism

L : A −→ End A
a −→ (La : b → a b)
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 27

extends by the Leibniz rule to a unique Lie algebra morphism L : A → Der S(A).
Now we claim that the map M : A → End S(A) defined by:

Ma u = au + La u [1.53]

is a Lie algebra morphism. Indeed, for any a, b ∈ A and u ∈ S(A) we have:

Ma Mb u = Ma (bu + Lb u)
= abu + aLb u + La (bu) + La Lb u
= abu + aLb u + bLa u + (a b)u + La Lb u.

Hence

[Ma , Mb ]u = (a b−b a)u + [La , Lb ]u

= M[a,b] u,

which proves the claim. Now M extends, by universal property of the enveloping
algebra, to a unique algebra morphism M : U(A) → End S(A). The linear map:

η : U(A) −→ S(A)
u −→ Mu .1

is clearly a morphism of left U(A)-modules. It is immediately seen by induction that

for any a1 , . . . , an ∈ A, we have η(a1 · · · an ) = a1 · · · an + v, where v is a sum of
terms of degree ≤ n − 1. This proves the theorem. 

R EMARK 1.3.– Let us recall that the symmetrization map σ : U(A) → S(A), uniquely
determined by σ(an ) = an for any a ∈ A and any integer n, is an isomorphism for
the two ALie -module structures given by the adjoint action. This is not the case for the
map η defined above. The fact that it is possible to replace the adjoint action of U(A)
on itself by the simple left multiplication is a remarkable property of pre-Lie algebras,
and makes Theorem 1.3 different from the usual Lie algebra PBW theorem.

Let us finally note that, if p stands for the projection from S(A) onto A, for any
a1 , . . . , ak ∈ A, we easily get:
  
p ◦ η(a1 · · · ak ) = La1 · · · Lak 1 = a1 a2 · · · (ak−1 ak )... [1.54]

by a simple induction on k. The linear isomorphism η transfers the product of the

enveloping algebra U(A) into a noncommutative product ∗ on S(A) defined by:
 
s ∗ t = η η −1 (s)η −1 (t) . [1.55]
28 Algebra and Applications 2

Suppose now that A is endowed with a complete decreasing compatible

filtration as in section 1.4.2. This filtration induces a complete decreasing filtration
S(A) = S(A)0 ⊃ S(A)1 ⊃ S(A)2 ⊃ · · · , and the product ∗ readily extends to the

completion S(A). For any a ∈ A, the application of equation [1.54] gives:

p(e∗a ) = W (a) [1.56]


as an equality in the completed symmetric algebra S(A).

According to equation [1.48], we can identify the pro-unipotent group {e∗a ,


a ∈ A} ⊂ S(A) and the group of formal flows of the pre-Lie algebra A by means of
the projection p, namely:

p(e∗a )#p(e∗b ) = p(e∗a ∗ e∗b ) [1.57]

for any a, b ∈ A.

1.5. Algebraic operads

An operad is a combinatorial device which appeared in algebraic topology (May

1972), coined for coding “types of algebras”. Hence, for example, a Lie algebra is
an algebra over some operad denoted by L IE, an associative algebra is an algebra
over some operad denoted by A SSOC, a commutative algebra is an algebra over some
operad denoted by C OM and so on.

1.5.1. Manipulating algebraic operations

Algebra starts, in most cases, with some set E and some binary operation ∗ :
E × E → E. The set E shows some extra structure most of the time. Here, we will
stick to the linear setting, where E is replaced by a vector space V (over some base
field k), and ∗ is bilinear, that is, a linear map from V ⊗ V into V . A second bilinear
map is deduced from the first by permuting the entries:

a ∗op b := b ∗ a. [1.58]

It also makes sense to look at tri-, quadri- and multi-linear operations, that is, linear
maps from V ⊗n to V for any V . For example, it is very easy to produce 12 tri-linear
maps starting with the bilinear map ∗ by considering:

(a, b, c) → (a ∗ b) ∗ c,
(a, b, c) → a ∗ (b ∗ c),

and the others deduced by permuting the three entries a, b and c. We could also
introduce some tri- or multi-linear operations from scratch, that is, without deriving
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 29

them from the bilinear operation ∗. We can even consider 1-ary and 0-ary operations,
the latter being just distinguished elements of V . Note that there is a canonical 1-ary
operation, namely, the identity map e : V → V . At this stage note that the symmetric
group Sn obviously acts on the n-ary operations on the right by permuting the entries
before composing them.

The bilinear operation ∗ is not arbitrary in general: its properties determine the
“type of algebra” considered. For example, V will be an associative or a Lie algebra
if for any a, b, c ∈ V , we have respectively:

(a ∗ b) ∗ c = a ∗ (b ∗ c), [1.59]
(a ∗ b) ∗ c + (b ∗ c) ∗ a + (c ∗ a) ∗ b = 0, a ∗ b − b ∗ a = 0. [1.60]

The concept of operad emerges when we try to rewrite such relations in terms
of the operation ∗ only, discarding the entries a, b, c. For example, the associativity
axiom equation [1.59] informally expresses itself as follows: composing the operation
∗ twice in two different ways gives the same result . Otherwise said:

∗ ◦1 ∗ = ∗ ◦2 ∗. [1.61]

The Lie algebra axioms (equation [1.60]), involving flip and circular permutations,
are clearly rewritten as:

(∗ ◦1 ∗) + (∗ ◦1 ∗) ◦ σ + (∗ ◦1 ∗) ◦ σ 2 = 0, ∗ + ∗ ◦ τ = 0, [1.62]

where τ is the flip (21) and σ is the circular permutation (231). The next section will
give a precise meaning to these “partial compositions”, and we will end up giving the
axioms of an operad, which is the natural framework in which equations like [1.61]
and [1.62] make sense.

1.5.2. The operad of multi-linear operations

Let us now look at the prototype of algebraic operads: for any vector space V , the
operad Endop(V ) is given by:

Endop(V )n = L(V ⊗n , V ). [1.63]

The right action of the symmetric group Sn on Endop(V )n is induced by the left
action of Sn on V ⊗n given by:

σ.(v1 ⊗ · · · ⊗ vn ) := vσ−1 (1) ⊗ · · · ⊗ vσ−1 (n) . [1.64]

30 Algebra and Applications 2

Elements of Endop(V )n are conveniently represented as boxes with n inputs

and one output: as illustrated by the graphical representation below, the partial
composition a ◦i b is given by:

a ◦i b(v1 ⊗ · · · ⊗ vk+l−1 ) :=
 
a v1 ⊗ · · · ⊗ vi−1 ⊗ b(vi ⊗ · · · ⊗ vi+l−1 ) ⊗ vi+l ⊗ · · · ⊗ vk+l−1 .
[1.65]

b
··· ···
1 i k
a

Partial composition a ◦i b

The following result is straightforward:

P ROPOSITION 1.13.– For any a ∈ Endop(V )k , b ∈ Endop(V )l and c ∈ Endop(V )m ,

we have:

(a ◦i b) ◦i+j−1 c = a ◦i (b ◦j c), i ∈ {1, . . . , k}, j ∈ {1, . . . , l}

(nested associativity property),
(a ◦i b) ◦l+j−1 c = (a ◦j c) ◦i b, i, j ∈ {1, . . . , k}, i < j
(disjoint associativity property).

The identity e : V → V satisfies the following unit property:

e◦a=a [1.66]

a ◦i e = a, i = 1, . . . , k, [1.67]

and finally, the following equivariance property is satisfied:

a.σ ◦i b.τ = (a ◦σi b).ιi (σ, τ ) [1.68]

where ιi (σ, τ ) ∈ Sk+l−1 is defined by letting τ permute the set Ei = {i, i + 1, . . . , i +

l − 1} of cardinality l, and then by letting σ permute the set {1, . . . , i − 1, Ei , i +
l, . . . , k + l − 1} of cardinality k.
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 31

The two associativity properties are graphically represented as follows:

···

c
··· ···
1 j l 1 l
···
··· b ··· b c
1 i k 1 i j
a a

Nested associativity Disjoint associativity

1.5.3. A definition for linear operads

We are now ready to give the precise definition of a linear operad:

D EFINITION 1.1.– An operad P (in the symmetric monoidal category of k-vector

spaces) is given by a collection of vector spaces (Pn )n≥0 , a right action of the
symmetric group Sn on Pn , a distinguished element e ∈ P1 and a collection of partial
compositions :

◦i : Pk ⊗ Pl −→ Pk+l−1 , i = 1, . . . , k
(a, b) −→ a ◦i b

subject to the associativity, unit and equivariance axioms of Proposition 1.13.

The global composition is defined by:

γ : Pn ⊗ Pk1 ⊗ · · · ⊗ Pkn −→ Pk1 +···+kn

   
(a, b1 , . . . , bn ) −→ ... (a ◦n bn ) ◦n−1 bn−1 · · · ◦1 b1

and is graphically represented as follows:

b1 b2 ··· bn

Global composition γ(a; b1, b2, . . . , bn)

32 Algebra and Applications 2

The operad P is augmented if P0 = {0} and P1 = k.e. For any operad P,

a P-algebra structure on the vector space V is a morphism of operads from P to
Endop(V ). For any two P-algebras V and W , a morphism of P-algebras is a linear
map f : V → W , such that for any n ≥ 0 and for any γ ∈ Pn , the following diagram
commutes,
f ⊗n
V ⊗n / W ⊗n

γ γ
 
V / W
f

where we have denoted by the same letter γ the element of Pn and its images in
Endop(V )n and Endop(W )n .

Now let V be any k-vector space. The free P-algebra is a P-algebra FP (V ) endowed
with a linear map ι : V −−→ FP (V ), such that for any P-algebra A and for any linear
map f : V → A, there is a unique P-algebra morphism f : FP (V ) → A, such that
f = ι ◦ f . The free P-algebra FP (V ) is unique up to isomorphism, and we can prove
that a concrete presentation of it is given by:

FP (V ) = Pn ⊗Sn V ⊗n , [1.69]
n≥0

with the map ι being obviously defined. When V is of finite dimension d, the
corresponding free P-algebra is often called the free P-algebra with d generators.

There are several other equivalent definitions for an operad. For more details about
operads, see, for example, Loday (1996) and Loday and Vallette (2012).

1.5.4. A few examples of operads

1.5.4.1. The operad A SSOC

This operad governs associative algebras. A SSOCn is given by k[Sn ] (the algebra
of the symmetric group Sn ) for any n ≥ 0, whereas A SSOC 0 := {0}. The right action
of Sn on A SSOCn is given by linear extension of right multiplication:
 
 
λi σi σ := λi (σi σ). [1.70]
i i

Let σ ∈ A SSOC k and τ ∈ A SSOC l . The partial compositions are given for any
i = 1, . . . , k by:

σ ◦i τ := ιi (σ, τ ), [1.71]
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 33

with the notations of equation [1.68]. The reader is invited to check the two
associativity axioms, as well as the equivariance axiom which reads:

(σσ  ) ◦i (τ τ  ) = (σ ◦σ (i) τ )(σ  ◦i τ  ) [1.72]

for any σ, σ  ∈ A SSOC k and τ, τ  ∈ A SSOC l . Let us denote by ek the unit element in
the symmetric group Sk . We obviously have ek ◦i el = ek+l−1 for any i = 1, . . . , k.
In particular,

e2 ◦1 e2 = e2 ◦2 e2 = e3 . [1.73]

Now let V be an algebra over the operad A SSOC , and let Φ : A SSOC → Endop(V )
be the corresponding morphism of operads. Let μ : V ⊗ V → V be the binary
operation Φ(e2 ). In view of equation [1.73] we have:

μ ◦1 μ = μ ◦2 μ. [1.74]

In other words, μ is associative. As ek can be obtained, for any k ≥ 3, by

iteratively composing k − 2 times the element e2 , we see that any element of A SSOC k
can be obtained from e2 , partial compositions, symmetric group actions and linear
combinations. As a result, any k-ary operation on V , which is in the image of Φ, can be
obtained in terms of the associative product μ, partial compositions, symmetric group
actions and linear combinations. Summing up, an algebra over the operad A SSOC is
nothing but an associative algebra. In view of equation [1.69], the free
A SSOC -algebra
over a vector space W is the (non-unital) tensor algebra T + (W ) = k≥1 W ⊗k .

In the same line of thoughts, the operad governing unital associative algebras is
defined similarly, except that the space of 0-ary operations is k.e0 , with ek ◦i e0 = ek−1
for any i = 1, . . . , k. The unit element u : k → V of the algebra V is given by
u = Φ(e0 ). The free unital algebra over a vector space W is the full tensor algebra
T (W ) = k≥0 W ⊗k .

1.5.4.2. The operad C OM

This operad governs commutative associative algebras. C OMn is one-dimensional
for any n ≥ 1, given by k.en for any n ≥ 0, whereas C OM 0 := {0}. The right action
of Sn on C OMn is trivial. The partial compositions are defined by:

ek ◦i el = ek+l−1 for anyi = 1, . . . , k. [1.75]

The three axioms of an operad are obviously verified. Let V be an algebra over
the operad C OM, and let Φ : C OM → Endop(V ) be the corresponding morphism of
operads. Let μ : V ⊗ V → V be the binary operation Φ(e2 ). We obviously have:

μ ◦1 μ = μ ◦2 μ, μ = μ.τ, [1.76]
34 Algebra and Applications 2

where τ ∈ S2 is the flip. Hence, μ is associative and commutative. Here, any k-ary
operation in the image of Φ can be obtained, up to a scalar, by iteratively composing
e2 with itself. Hence, an algebra over the operad C OM is nothing but a commutative
associative algebra. In view of [1.69], the free C
OM -algebra over a vector space W is
the (non-unital) symmetric algebra S + (W ) = k≥1 S k (W ).

The operad governing unital commutative associative algebras is defined similarly,

except that the space of 0-ary operations is k.e0 , with ek ◦i e0 = ek−1 for any
i = 1, . . . , k. The unit element u : k → V of the algebra V is given by
u = Φ(e0). The free unital algebra over a vector space W is the full symmetric algebra
S(W ) = k≥0 S k (W ).

The map ek → ek is easily seen to define a morphism of operads Ψ : A SSOC →

C OM . Hence, any C OM -algebra is also an A SSOC -algebra. This expressed the fact
that, forgetting commutativity, a commutative associative algebra is also an associative
algebra.

1.5.4.3. Associative algebras

Any associative algebra A is some degenerate form of operad: indeed, defining
P(A) by P(A)1 := A and P(A)n := {0} for n = 1, the collection P(A) is obviously
an operad. An algebra over P(A) is the same as an A-module.

This point of view leads to a more conceptual definition of operads: an operad

is nothing but an associative unital algebra in the category of “S-objects”, that is,
collections of vector spaces (Pn )n≥0 with a right action of Sn on Pn . There is a
suitable “tensor product”  on S-objects, however not symmetric, such that the global
composition γ and the unit u : k → P1 (defined by u(1) = e) make the following
diagrams commute:

γI
PPP / PP

Iγ γ
 γ 
PP / P,

uI Iu
kP / PP o Pk .
II u
II ∼ ∼ uu
II γ u
II uu
II  uuu
$ zu
P.
 Algebraic Background for Numerical Methods, Control Theory and Renormalization 35

These two diagrams commute if and only if e verifies the unit axiom and the partial
compositions verify the two associativity axioms and the equivariance axiom (Loday
and Vallette 2012).

1.6. Pre-Lie algebras (continued)

1.6.1. Pre-Lie algebras and augmented operads

1.6.1.1. General construction

We adopt the notations of section 1.5. The sum of the partial compositions yields
 structure on the free P-algebra with one generator, more
a right pre-Lie algebra
precisely on FP+ := n≥2 Pn /Sn , namely:


k
a  b := a ◦i b. [1.77]
i=1

Following Chapoton (2002), we can consider the pro-unipotent group GeP

associated with the completion of the pre-Lie algebra FP+ for the filtration induced by
the grading. More precisely, Chapoton’s group GP is given by the elements g ∈ F P ,
such that g1 = 0, whereas GeP is the subgroup of GP formed by elements g, such that
g1 = e.

Any element a ∈ Pn gives rise to an n-ary operation ωa : FP⊗n → FP , and for

any x, y1 , . . . , yn ∈ FP+ , we have1(Manchon and Saidi 2011):

n
ωa (y1 , . . . , yn )  x = ωa (y1 , . . . , yj  x, . . . , yn ). [1.78]
j=1

1.6.1.2. The pre-Lie operad

Pre-Lie algebras are algebras over the pre-Lie operad , which has been described in
detail by Chapoton and Livernet (2001) as follows: PLn is the vector space of labeled
rooted trees, and the partial composition s ◦i t is given by summing all of the possible
ways of inserting the tree t inside the tree s at the vertex labeled by i. To be precise,
the sum runs over the possible ways of branching on t the edges of s, which arrive on
the vertex i.

Thefree left pre-Lie algebra with one generator is then given by the space
T = n≥1 Tn of rooted trees, as quotienting with the symmetric group actions

1. We thank Muriel Livernet for having brought this point to our attention.
36 Algebra and Applications 2

amounts to neglect the labels. The pre-Lie operation (s, t) → (s → t) is given by

the sum of the graftings of s
on t at all vertices of t. As a result of [1.78], we have two
pre-Lie operations on T  = n≥2 Tn , which interact as follows (Manchon and Saidi
2011):

(s → t)  u = (s  u) → t + s → (t  u). [1.79]

The first pre-Lie operation  comes from the fact that PL is an augmented operad,
whereas the second pre-Lie operation → comes from the fact that PL is the pre-Lie
operad itself! Similarly:

T HEOREM 1.4.– The free pre-Lie algebra with d generators is the vector space of
rooted trees with d colors, endowed with grafting.

1.6.2. A pedestrian approach to free pre-Lie algebra

In this section we give a direct proof of Theorem 1.4 without using operads. It is
similar to the proof of the main theorem in Chapoton and Livernet (2001) about the
structure of the pre-Lie operad, except that we consider unlabeled trees. We stick
to d = 1 (i.e. one generator), the proof for several generators being completely
analogous. Let T be the vector space spanned by rooted trees. First, the grafting
operation is pre-Lie, because for any trees s, t and u in T, the expression:

s → (t → u) − (s → t) → u [1.80]

is obtained by summing up all of the possibilities of grafting s and t at some vertex

of u. As such, it is obviously symmetric in s and t. Now let (A, ) be any left pre-lie
algebra, and choose any a ∈ A. In order to prove Theorem 1.4 for one generator, we
have to show that there is a unique pre-Lie algebra morphism Fa : T −→ A, such that
Fa ( ) = a. For the first trees, we easily obtain:

Fa ( ) = a
Fa ( ) = a a

Fa ( ) = (a a) a
Fa ( ) = a (a a) − (a a) a.

Can we continue like this? We proceed by double induction, first, on the number
of vertices, second, on the number of branches, that is, the valence of the root. Write
any tree t with n vertices as t = B+ (t1 , . . . , tk ), where the tj s are the branches and
B+ is the operator that grafts the branches on a common extra root. By the induction
hypothesis on n, the images Fa (tj ) are well-defined.
Random documents with unrelated
content Scribd suggests to you:
Loch Lomond, where they found "a few small stone cabins, some fat
bairns, abundance of ale, and a sufficiency of capital whisky." After
crossing to Tarbet and examining both the head and the foot of the
lake, they went on from Balloch to Dumbarton by stage, and thence
by steamer to Glasgow; there they spent a few days, and returned
to Edinburgh by way of Dumbarton and Lanark. Steamers and
coaches, slow as they then were, were all too fast for Audubon on
this journey, and he declared that if ever again he visited the
Highlands, it should be on foot, "for no man, with nerve and will,
and an admirer of the beauties of nature, can ever truly enjoy the
pleasures of travelling, unless he proceed in this manner."

Mrs. Audubon's health had not improved by the journey, for

shortly after their return she was again taken ill; she was placed in
the care of Dr. John Argyle Robertson, for whose efficient aid and
"kind and gentle treatment," said Audubon, "we can never cease to
cherish the most lively feelings of affection." "It is a curious part of
my history," he continues, "that during the whole of my sojourn in
Britain, none of the principal medical advisers whom we had
occasion to employ would receive any recompense from us."

By the 5th of November, 1838, but a few days after the issue of
the fourth volume of his "Biographies," the printing of the fifth and
last had already begun. The following letter[152] is interesting for its
personal details, and in showing that his confidence in the existence
of the "bird of Washington"[153] had not been shaken:

Audubon to Edward Harris

Edinburgh December 19th., 1838.

My dear Friend:
 Your letter of the 13th. instant to Victor reached us this
morning, and glad were we all to hear from you.

My object in writing to you is, for the purpose of assuring

you that I feel great [pleasure] in preparing a box of bird skins
for you according to your desire. It is true, however, that I am
now on the eve of commencing my synoptical arrangement of
our birds, I shall not be able to show you as many of my
specimens [as] I could otherwise have done, but you must take
the will for the deed. The Box will be taken as far as London by
Victor, who will leave us on the first of January. You will find in it
a list of the contents, and I trust such bird skins as may answer
your purpose well. Besides these Victor will also attend to your
request as soon as in New York and will ship to you by way of
Havre as you direct.

I am glad that you should have seen what you conceive to

be the great rara avis F. Washingtonii. I am sorry you could not
have pocketed it, but who knows if it is not left yet in store for
you—and I to shoot a pair of these noble birds at The West, and
that, after having satisfactorily examined its habits, its eggs, or
its young! Bonaparte, between you and I, is exceedingly
ignorant as regards our birds, as I found to my cost when he
was in London, and where he pumped me sadly too much, but
it is now over and I forgive him as I do all others who have or
who may try to injure me.

John Bachman wrote to me that he had left in commission

to Trudeau, the purchase for me of a copy of Vieillot's Oiseaux
de l'Amérique Septentrionale, and also a copy of Boié or Bojé
work[154] on birds generally, but I have received neither books
or promises of them from Trudeau as yet, perhaps you would
undertake the task yourself, and show to Havell as soon as
possible, for I shall be sadly in want of them in a few weeks
more. I should also like you to try to find Mr. Augustus
Thorndike of Boston, to whom Victor wrote a few days ago, with
the view to inquire from that Gentleman when he wished his
copy of the Birds of America to be delivered. Victor addressed
his letter to "his Hotel" or to the care of the "American
Embassy." We are not sure, however, whether he is in Paris at
present, and let me know what discoveries you have made as
regards this.—Victor will remain ten days in London and wishes
you, should you write to him there, to put your letter under
cover to Havell and request him to keep it. Victor will write you
from thence.

I cannot account why Trudeau has not written to me in

answer to my last, now full two months old? Should you
perchance discover a specimen of the Bird of Washington in
Paris and purchase the same, I should like you to send it to me
on loan to enable me to compare it with mine, and the
Immature of the F. Albicilla of Europe!

I have got twelve sheets of the 5th. Vol. of Biographies

already printed, and I expect to have quite finished by the 1st.
of April next. I have decided on the Trichas resembling Sylvia
Philadelphia of Wilson. It is a distinct species, but what will
probably surprise you more, the S. Agilis of the same author is
also perfectly distinct from either. All this you will plainly see
when you read their separate descriptions and compare the
three species.

I wish you would ask Trudeau whether he recollects the

specimen of an Eagle send by Townsend in his first collection,
numbering 54 and which the latter has lost, though he
considered it as a new species. It was procured in California.
Townsend speaks sorrowfully of the loss of this specimen. It
never came under my eye, did it come under yours? Ask
 Trudeau whether he ever saw my Hirundo Serripennis in
America. Bachman wrote to me that Trudeau thought he had in
the skins of Frederick Ward. I think Trudeau will be pleased with
the anatomy of our birds, as it opens misteries hitherto
unknown in connection with the relative affinities of some
species toward others and assists in the formation of groups
&c., in what some day or other, will be called a Natural
arrangement!
I wish I could have spent a few weeks in Paris with you and
Trudeau, as I readily imagine that some new species of North
American birds, may yet be found there unknown to the World
of Science. I have written to Mr. Chevalier and to Townsend, but
will not, I dare say, hear anything more of the former until
through Victor, who intends to see him very shortly after his
arrival in America.

My Dear Wife is much better than when you saw her, and I
hope that when once again she has been safely landed on our
shores and enjoyed the warmth of our own Summers, her
health will be quite restored.

The Little Lucy has grown as fat as butter, and the rest of us
are well.

We all unite in kindest best wishes to you and to Trudeau,

and I remain as ever, my Dear Friend,

Yours,
John J. Audubon.

6 Alva Street.

In May, 1839, Audubon's fifth and last volume of the

Ornithological Biography, consisting of 704 pages, was issued. It was
followed almost immediately by A Synopsis of the Birds of North
America, in which the efficient aid of MacGillivray was again enlisted.
On May 4 Audubon wrote to Havell that this work was in press and
would be ready in about a month's time; again, on the 30th of June
he announced that it was finished and in the hands of the binder.
With this methodical catalogue of the birds of North America then
known and described, to the number of 491, fifty-two of which were
new, Audubon's life and labors in England were brought to a close.

The introduction to the last volume of his "Biographies" begins

as follows:

How often, Good Reader, I have longed to see the day on

which my labours should be brought to an end! Many times,
when I had laid myself down in the deepest recesses of the
western forests, have I been suddenly awakened by the
apparition of dismal prospects that have presented themselves
to my mind. Now, sickness, methought, had seized me with
burning hand, and hurried me away, in spite of all my fond
wishes, from those wild woods in which I had so long lingered,
to increase my knowledge of the objects which they offered to
my view. Poverty, too, at times walked hand in hand with me,
and on more than one occasion urged me to cast away my
pencils, destroy my drawings, abandon my journals, change my
ideas, and return to the world.

Later on he says: "You may well imagine how happy I am at this

moment, when ... I find my journeys all finished, my anxieties
vanished, my mission accomplished;" and he concludes: "I have
pleasure in saying that my enemies have been few, and my friends
numerous. May the God who granted me life, industry, and
perseverance to accomplish my task, forgive the former, and forever
bless the latter!"
 Audubon's introductions to the five volumes of his "Biographies,"
from which we have frequently quoted, are characteristic; in them
he cheers his subscribers, calls all his helpers and correspondents by
name, and takes the public into his confidence by recording the acts
which marked the steady progress of his work. Frequent appeals to
the "good" and "gentle reader" have gone out of fashion, but in this
instance they seem in keeping with the style and character of the
man, and they were not made in vain. Audubon's belief in his
mission was so plainly sincere, his power so manifest and his
enthusiasm so ardent, that there were few who did not gladly
acclaim the extraordinary success of the man who twelve years
before had landed in Liverpool poor and unknown.

In the winter and spring of 1839, while Audubon was engaged in

Edinburgh and Victor was in America, the settlement of his business
affairs in London was entrusted mainly to Robert Havell, his
engraver. At that time Havell was also pulling up roots, for he had
caught the spirit of his patron and had decided to emigrate with his
family to the United States; this involved disposing of his stock and
breaking up his engraving and printing establishment at 77 Oxford
Street. Havell had acquired distinction as well as a competence
through his long engagement with Audubon, and being then in his
forty-sixth year, he doubtless looked to America as a field for the
fuller expression of his artistic aspirations and talents. How anxious
Audubon was at this juncture regarding the disposition of the
residual stock of his plates, his drawings, and his books, then in
Havell's hands, is seen by the following letter,[155] written at
Edinburgh, in the winter of this year.

Audubon to Robert Havell

Edinburgh, Feb. 20th, Monday, 1839.

My dear Mr. Havell

 I perceive by the date of your letter of the 16th instant that
you must have been some days beyond my expectations, in the
receiving of my parcel to you, and that on that account my
letter of Saturday last crossed yours of the same date. I thank
you for what you say as regards the balance in my favor at
Wright and Co.

Does Henry sail from London Docks or from Portsmouth?

and pray what is the name of the Captain of the "Wellington"? I
do not precisely understand what you mean by the loose sets
which you desire to know how they should be packed? let me
hear what they are and how many of them by return of mail.
The five perfect sets I think might all go into one case, tinned
as usual and insured of course to the full amount of their value,
as well as all others and to which I pray you to attend as if for
your own self. It is impossible for me to go to London at present
and indeed I cannot exactly tell when I will, and I trust to you
entirely for the seeing that all the volumes are fair and good
and passed through your own inspection of them before they
are packed. No volumes of Biographies must be put in the same
boxes.

When you have disposed of your business, what will you do

with what you have on hand belonging to us? This requires an
answer from you at once. You have a great number of volumes
of Biographies, Pictures, &c. &c., a regular list of which you
ought to send me. I cannot yet say when the 5th vol. of
Biographies will be finished, but will let you know as soon as I
can. I received yesterday morning a letter from a gentleman
who has procured a copy of the work through Mr. Eame the
bookseller, he says that he has called upon you to say that he is
missing one plate and begs to have a copy of the plate struck
and remitted to Mr. Eame who will pay you whatever price the
extra trouble on this account may amount to, but he does not
 say what plate it is, and I therefor suppose that you do? If so as
he is the brother-in-law of Mr. Walker of Ravensfield Park, one
of our good subscribers I would say do it for him! My wife begs
of you to save all the loose prints which were returned to you by
our son Victor, as well as any others whatever, perhaps among
them you might find one to send Mr. Eame's subscriber?

On the 4th of this month the 'Great Western' was nearly half
way across the Atlantic!! Sir William Jardine has published a
capital review of the work! What a strange world we do learn in!
Be sure to let me know about the original drawings at Henry's, if
he has finished them, where they are &c. &c. We all remain as
usual with kind good wishes to you all,

Your friend
John J. Audubon.
6 Alva St.
[Superscribed]
Robert Havell Esq.
Engraver.
77 Oxford st
London.
How fully Audubon's injunctions in regard to his residual stock,
and particularly to scattered plates of the Birds, were followed, is not
known, but it is certain that a part of this residuum remained in
England, where it is occasionally turned up even at the present day.
In a considerable number of the original plates which were found in
a bookshop in New Oxford Street in August, 1912, twelve were in
the uncolored state, and several had the appearance of rejects;
moreover, in a collection of these plates received from England in
1910, there were nine copies of the same subject, the Painted
Bunting (No. 11, Plate 53). Though a complete set of the plain
plates is known,[156] and a considerable number were probably
dispersed in America, they are very rare.

Audubon wrote to Havell again on the 13th of March, when he

complained of the gross mistakes made by the "idle rascals" who
were employed in filling orders in his shop, and who had so sadly
mixed matters that no less than twelve numbers of his Birds had
come back to him, some of them containing one, three, and five
copies of the same plate, instead of a "Number," and mixtures of the
most incongruous sort; he thought that "a clever young man as a
clerk was worth a hundred thick heads," and begged Havell again to
send him "a correct list of what he shipped to America on his
account, and that list not made by any other person than either
himself or Mrs. Havell." His next injunction, on May 4, was to insure
his copper-plates of The Birds of America for £5,000, and to send
them to either Victor Audubon or Mr. N. Berthoud, Number 2
Hanover Street, New York. At that moment Audubon was planning to
return to America with his family by the Great Western on July 6. In
writing again on June 30, he remarked that he was not at all certain
that Havell, who was then visiting at his native Reading, in
Berkshire, would really sail on the 25th of July, since he had already
postponed the journey so many times; he added that it would not
even surprise him if his work on the Quadrupeds of America might
not be out before they could fish and shoot together in his "native
land."

Havell eventually came to America with his wife and daughter on

the ship Wellington, in September, 1839; they landed at New York
after the 15th of that month, and for a time were the guests of the
Audubons at Number 86 White Street. His brother, Henry,[157] who
visited the United States in 1829, returned at about this time and
established a print shop in Broadway, New York, but according to
Robert's biographer, his enterprise was ruined by a fire, when he
went back to England and he died there about 1840. After a brief
residence in Brooklyn, Robert Havell settled at Sing Sing, now
Ossining, at a beautiful spot on the Hudson, overlooking the
Palisades, which he named "Rocky Mount." There he devoted himself
with characteristic energy to painting and sketching, but he also
engraved and published a number of excellent views of his favorite
river, the Hudson, as well as of New York and other American cities.
In 1857 he established himself at Tarrytown, where he built a house
and studio, and where in his later years he produced many
meritorious works in oils. "He never tired," says his biographer, "of
the great, broad, sweeping Hudson, and propped up in bed, that he
might gaze at will on this mighty river," he died at the age of eighty-
five, November 11, 1878.[158]

Havell has been described as quite the opposite of Audubon in

many of his characteristics, calm, deliberate, not easily discouraged,
and fully his equal in industry, perseverance and determination.
Audubon sometimes complained of his friend's lax business habits,
but their long sustained and cordial relations were never broken
during life, and their mutual debt was great. The engraver's first
son, who lived but a year, was named Robert Audubon, and the
naturalist, who was his godfather, held the child at its baptism at old
St. James Church, Oxford Street, in 1827. A descendant of Luke
Havell, who was a drawing master at Reading, uncle of Robert the
second, possesses a silver loving-cup which Audubon presented to
his engraver upon the completion of the second volume of his
illustrations; it is inscribed "To Robert Havell, from his friend J. J. A.
1834."

When we consider the size of Audubon's plates, which required

for the portrayal of his largest subjects, such as the Whooping Crane
or Wild Turkey, an area of no less than five square feet, it will be
seen that his engraver was compelled to adopt the most expeditious
methods. This and kindred difficulties were overcome by Havell's
skillful union of aquatint with etching and line engraving, but some
of his smaller figures, as the Snow Birds (Plate 13), appear to have
been etched in the usual way, with but slight use of either aquatint,
dry-point or burin. In aquatinting the plate was usually bitten to the
desired depth for the softer shading of feathers or foliage, or for the
entire expression of sky, water or landscape. Says George Alfred
Williams:[159]

Aquatint proper consists entirely of gradations of tone

produced by biting with aquafortis into the copper through a
resinous ground broken into a multitude of fine granules, that
render the personal touch practically negligible, and in
consideration of this we can appreciate the exceptionally skillful
use Havell, Junior, made of the difficult process. The graining of
the aquatint grounds is produced by allowing fine dust particles
to settle upon the freshly prepared plate. It is to these grainings
of different degrees of fineness that the engraver must look for
the subtlety of the tonal surfaces, but strength is obtained
usually through the use of the etched line. The chief limitation
of the aquatint process lies in the great difficulty of getting more
than a few differences of shade, as the ground goes to pieces
rather rapidly under successive bitings, and the transitions from
one tone to another are very few, so that half tones are not
readily obtainable. It is in the economical use of these half
tones that Havell, Junior, achieved so much and thereby
produced a chiaroscuro seldom, if ever, equaled in aquatint.

Artists have sometimes frowned upon this combination of

aquatint with other forms of engraving, especially when executed
with the burin, for "like the permanent misery of a quarrelling
married couple, they may ruin everything with discord"; but any
such lack of harmony, when not overcome by Havell's skill, was
usually completely subdued by the color which was subsequently
applied to the printed sheet. This is seen by a comparison of the
plain proof of such an elaborate plate as that of the rattlesnake
attacking the mocking birds (Plate xxi) with the result attained in the
finished impression. In replying to a criticism of Havell's work in
1830, Audubon expressed his conviction that "no birds were ever so
beautifully and softly represented on copper," and any hardness,
which his plates rarely showed, was not due to aquatinting, the
inherent quality of which is softness. To quote our authority again:

In Havell's hands aquatint gave the essential structure of

forms by a judicious use of the process called "feathering." This
he often did upon the bare copper plate without a ground,
allowing the acid to bite its own granular surface. In this way he
produced the soft gradations and telling accents so necessary to
the portrayal of birds, but by a further judicious use of line the
accurate forms of both bird and plant life were given with great
force and delicacy.

When Audubon first proposed to present the Congressional

Library at Washington with a copy of his Birds, he asked Havell to do
all the coloring himself; Havell

colored his prints in the usual manner by flowing washes of

pure water-color tints over the monochrome proof which was
printed from the copper plate. In this phase of the process the
great charm of his genuine talent for water-color painting
asserted itself. Aside from the first crude washes, put on by
artists or colorists employed for the purpose, he himself applied
the salient tones and all the more delicate tints.

Much misunderstanding has arisen in regard to the engraving

and publication of Audubon's earlier plates owing to the complex
relations which existed between Lizars, the two Havells of the same
name, and the naturalist himself; this involved the reissue of the first
two numbers of the work, and a confusing series of legends upon
the plates themselves, occasioned partly by a division of labor
between father and son, and by the death of Robert Havell, Senior,
in 1832. The errors into which some commentators have fallen,
however, are due to their examination of but one set of plates. The
account which follows is based upon a collation of complete copies
at the British Museum, the Radcliffe Library, Oxford, the Jardin des
Plantes, Paris, and the Boston Society of Natural History, and of
numerous scattered plates. Lizars engraved the first ten plates only
of the large folio, but before the summer of 1827 a considerable
number of these early impressions had been distributed. The
Havells, as we have seen, started the work anew, and Robert, the
younger, retouched the greater part of Lizars' plates, so that their
reissue in London constitutes for the bibliophile a second, and in
some cases a third, edition. Moreover, the plates which were
eventually executed by the younger Havell, to the number of 425,
were repeatedly printed from to meet the requirements of new
subscribers; on such occasions errors were corrected, minor changes
in the artist's or engraver's name introduced, and the nomenclature
of the birds and plants more or less completely revised. Frequently
the Whatman water-marks, or, indirectly, Audubon's titles, postdate
the year of publication as printed on the plate itself. Conflicting
legends, particularly on the earlier plates, which have often puzzled
collectors, are mainly due to one or another of the conditions
named. Nearly every plate underwent alterations of some sort, but
the various legends show conclusively whether the print belongs to a
first or a subsequent impression. In the British Museum copy, the
younger Havell's first independent legend appears on Plate lxxvi
(Virginia Partridge), as "Engraved, Printed, & Coloured by R. Havell
Junr. 1830." The word "London" was later added, but was eventually
discarded. After the senior Havell's death in 1832, the son naturally
dropped the suffix from his own name; thereafter his brief
designation of name and date became gradually stereotyped. It
should be noticed, however, that when Robert Havell needed to
reprint from plates which were executed before his father's death,
he would frequently remove only the "Junior," and let the original
date stand, so that legends like the following are not uncommon:
"Engraved, Printed, & Coloured by R. Havell, —— London—1831"; to
avoid the lacuna, designation of place and date were, at times,
erased also. The naturalist's name underwent a different evolution
on the plates, but after 1830, the date of his election to membership
in the Royal Society, his designation gradually settled into the
phrase: "Drawn from nature by J.J. Audubon, F.L.S. F.R.S." (as in
Plate lxxxi, 1830); accordingly, all plates which bear these titles, but
an earlier date, are second or later editions. The previous hints, with
examples to be given presently, will enable collectors to determine
whether a given plate is a first or subsequent impression.
 PLATE CCCI

Canvas backed Duck

FULIGULA VALLISNERIA, Steph.
1 2 Male, 3 Female View of Baltimore

Drawn from nature by Engraved Printed &

J.J. Audubon F.R.S. Coloured by R. Havell.
F.L.S. 1836

Lizars' first edition of plates, Numbers i to x, bore no dates, and,

so far as known, the first date of the entire series was "1828," which
was added to Plate 31 (the White-headed Eagle), but was erased
from later issues. When Robert Havell, Junior, retouched Lizar's
engraving of the Turkey Hen (No. II, Plate vi), he added the date
"1829," but in a third or later edition the name was changed and the
date removed. This same date appears also on the retouched Plate
ii, of the Yellow-billed Cuckoo, and on that of the Purple Grackle or
Crow Blackbird (Plate vii). A curious error crept into the first
impressions of the former plate, which was mislabeled "Black-billed
Cuckoo," a bird then recognized as quite distinct.[160] Plates xxxi to
lxxv (1828-29) were all dated in the first issue, and bore the names
of both Havells; the legends are: "Engraved, Printed & Coloured by
R. Havell & Son, London, 1828," and "Engraved by Robt. Havell,
Junr. Printed & Coloured by R. Havell Senr. London, 1828," or "1829"
(British Museum copy), but when later printed by Robert, the
younger, the dates were erased. All subsequent dating of the plates
was somewhat irregular until 1834 (Plate ccii), but from that point to
the end of the series, the year of issue was consistently added to
each plate. Ornithologists are specially interested in the time of
publication, since forty-seven new specific names occur on
Audubon's large plates, and should date from them and not from the
letterpress which followed.

To illustrate what has been said of successive editions of

Audubon's plates, we shall give the legends of two or three of the
most famous, taken from copies in the British Museum and Boston
Society of Natural History libraries or from detached plates obtained
in London, in 1903:

Plate i (1st edit., Lizars; Brit. Mus.).—"Great American Cock

Male—Vulgo (Wild Turkey) Meleagris Gallopavo. Drawn by J. J.
Audubon, M. W. S. Engraved by W. H. Lizars Edinr."

Plate i (2nd or later edit., Havell; Boston Soc. Nat. Hist.).

—"Wild Turkey. Meleagris Gallopavo. Linn. Male American Cane.
Miegia macrosperma. Drawn from nature by J. J. Audubon
F.R.S., F.L.S. Engraved by W. H. Lizars Edinr. Retouched by R.
Havell Junr."

Plate vi (1st edit., Lizars; Brit. Mus.).—"Great American Hen

& Young, Vulgo, Female Wild Turkey—Meleagris Gallopavo.
Drawn from Nature by John J. Audubon F.R.S.E. M.W.S.
Engraved by W. H. Lizars Edinr."
 Plate vi (2nd or later edit., Havell; detached plate).—The
same with the exception of "F.L.S.," added to Audubon's name,
and "Retouched by R. Havell. Junr. London, 1829."

Plate vi (3rd or later edit., Havell; Bost. Soc. Nat. Hist.).

—"Wild Turkey, Meleagris Gallopavo. Linn. Female and Young."
(Artist's name apparently cut off by binder.) "Engraved by W. H.
Lizars. Retouched by R. Havell. Junr."

Plate 31. (1st edit., Havell. Brit. Mus.).—"White-headed

Eagle, male. Falco Leucocephalus. Fish Fulgo—Yellow mud Cat.
Drawn from Nature & Published by John J. Audubon, F.R.S.E.,
F.L.S., M.W.S. Engraved, Printed & Coloured by R. Havell & Son,
London, 1828."

Plate xxxi (2nd or later edit., Havell; Bost. Soc. Nat. Hist.).
—"White-headed Eagle, Falco Leucocephalus. Linn. Male. Yellow
Cat-fish. Drawn from Nature and Published by John J. Audubon.
F.R.S. F.L.S. Engraved, Printed & Coloured, by R. Havell."

Plates which were wholly the work of Lizars have naturally

become extremely rare; they were evidently disregarded by
Audubon when he recorded on July 2, 1827, that he had given Mr.
Children a proof of his first number, which he called "the first in
existence," and declared that the two guineas then received was the
first money that had been returned to his hands. Lizars' initial
number had actually been finished in the previous winter, and a copy
of this is recorded as having been given to the daughter of Sir
Walter Scott on the 9th of March, 1827.

When Audubon had finally closed all his business affairs in

Edinburgh and London, late in the summer of 1839, he returned to
America, with the remaining members of his family, and settled in
New York, where he purchased a house at Number 84 White Street,
then in the uptown district.
 An anonymous writer in the London Athenæum[161] in giving a
final review of Audubon's labors in 1839, paid this interesting
tribute:

It seems but as yesterday that we were walking about with

a transatlantic stranger, picturesque enough in his appearance
and garb, to arrest the eye of every passing gazer; a tall
stalwart man, with hair sufficiently long to qualify him to serve
as a model to Gray's "Bard," and trousers ample almost as
petticoats of "good Harmony cloth," so absorbed in the
enthusiastic prosecution of his gigantic plan—a life's labour—as
to be heedless of the singularity of those meteoric locks, and
those liberal nether garments. Some dozen of years, however,
have elapsed since that day; the American Woodsman's hair—
long since cut short—has grown white; his magnificent
undertaking is completed, and he is now on the point of quitting
England, to settle himself for the remainder of his days whether
by the side of a bayou, in some forest clearing, or as an
inhabitant of one of the American cities which have learned to
know his value, report saith not.

We shake hands with the author, tendering him our hearty

congratulations on the completion of a task almost as arduous
as has ever been proposed to a literary man....

The confidential simplicity of Mr. Audubon's own prefaces

would make yet more personal leave-takings and farewells, on
the critic's part, natural and graceful,—but it must suffice to say,
that few have quitted England, carrying with them a larger
portion of honest regard and sincere good wishes.

Possibly it was the same writer who gave this striking picture of
Audubon in the pages of the same journal, thirty years later:[162]
 We can remember when his portfolio excited delight in
Edinburgh, London, and Paris, rivalling in smaller circles a new
Waverley novel. The man also was not a man to be seen and
forgotten, or passed on the pavement without glances of
surprise and scrutiny. The tall and somewhat stooping form, the
clothes not made by a West-end but by a Far West tailor, the
steady, rapid, springing step, the long hair, the aquiline features,
and the glowing angry eyes,—the expression of a handsome
man conscious of ceasing to be young, and an air and manner
which told you that whoever you might be he was John
Audubon, will never be forgotten by anyone who knew or saw
him.

We will add to this the musings of an anonymous American

writer[163] in the North American Review for the following year
(1840):

It must have been with mingled and varied feelings that

Audubon published his concluding volume. He was sure then
that he had raised an imperishable monument to commemorate
his own renown. All anxieties and fears which overshadowed his
work in its beginning had passed away. The prophecies of kind
but overprudent friends, who did not understand his self-
sustaining energy, had proved untrue; the malicious hope of his
enemies, for even the gentle lover of nature has enemies,—had
been disappointed; he had secured a commanding place in the
respect and gratitude of men; he had secured a treasure of rich
and glowing recollections, to warm his own heart in his
declining years, and to kindle enthusiasm in his children's
children....

On the other hand he had lost an employment which for

years had kept all the powers of body and mind in healthy
though intense exertion; whatever else he might do, the great
 work of his intellectual life was finished.... His trumpet of victory
at the result must have given an uncertain sound, partly
exulting in his success, and partly lamenting that his great work
was finished.

It has often been asked, how many complete sets of Audubon's

folio of The Birds of America were distributed, and how many are in
existence today. No definite answer can be given to either question.
His final lists, appended to the last volume of his "Biographies" in
1839, and reproduced in Appendix III to the present work, gave the
number of standing names as 161 (calling for 166 copies), of which
79 (with 84 copies) pertained to Europe, and 82 were American;
inasmuch as 118 subscribers had dropped off, with incompleted sets
on their hands, this brought the total number of original patrons at
that time to 279, and the number of copies which had been
originally ordered to 284. On the other hand, the following
advertisement, evidently from the hand of Audubon himself,
appeared in The Athenæum, under date of "London, November 1,
1837," and was reprinted in America a few weeks later:[164]

The number of perfect copies at present subscribed for does

not exceed 190, of which upwards of 80 are subscribed for in
America; and the expense of getting them up is so great, that
not more than ten or fifteen copies, above the number
subscribed for, will be prepared.

The Establishment necessary for its publication will be

broken up when the last Number is coloured; and any
application for the Work must be made to the Author, 4,
Wimpole—street, London; or Mr. R. Havell, Engraver, 77 Oxford
—street, before the first of May next, as after that time no
subscription can be received.
 Assuming that Audubon's final published lists, calling for 166
copies, were correct, and that ten additional sets were prepared, this
would bring the total original number known to have been complete
to 176, which agrees with an estimate attributed to Victor Audubon,
who once expressed the belief that "about 175 copies" were in
existence, of which about 80 were in America.[165] It is probable,
however, that of the 118 persons who allowed their subscriptions to
lapse, a number completed their sets from the residual stock, which
must have been great, though the later plates would be the hardest
to obtain; in this event the number of perfect sets of Audubon's
Birds probably never exceeded 190 or 200.

The proprietor of a well known Philadelphia bookshop[166] has

stated that during his experience as an antiquary, he has had
personal knowledge of forty or fifty copies of the folio edition of
Audubon's Birds in America, and he thought it probable that a single
New England print dealer, in the course of twenty years, had broken
up thirty or forty volumes for the purpose of selling the plates. This
is not surprising, since from the sale of a single volume upwards of
$1,500 might be realized in this way, but no reputable dealer would
now think of breaking up an unimpaired set.

Mr. Ruthven Deane, who has compiled a careful record of copies

of The Birds of America known to exist in the United States,
recorded in 1908 that he had ascertained the resting-place of
seventy-five sets which, with few exceptions, were complete and in
good condition. "A set in the library of the Mechanics-Mercantile
Institute, San Francisco, California, which had been there for some
thirty years, and another set in the San Francisco Art Association,
presented in 1894 by Mr. Edward F. Searles, Methuen,
Massachusetts, were both destroyed by the disastrous earthquake
and fire which visited that city April 18, 1906."[167]
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com