Groups Languages and Automata 1st

Edition Derek F. Holt

Visit to download the full and correct content document:
https://textbookfull.com/product/groups-languages-and-automata-1st-edition-derek-f-h
olt/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Notes on formal languages automata computability and

complexity Gallier J

https://textbookfull.com/product/notes-on-formal-languages-
automata-computability-and-complexity-gallier-j/

Machine Vision Algorithms in Java Techniques and

Implementation Whelan Paul F Molloy Derek

https://textbookfull.com/product/machine-vision-algorithms-in-
java-techniques-and-implementation-whelan-paul-f-molloy-derek/

Theory of Groups and Symmetries Finite groups Lie

groups and Lie Algebras 1st Edition Alexey P. Isaev

https://textbookfull.com/product/theory-of-groups-and-symmetries-
finite-groups-lie-groups-and-lie-algebras-1st-edition-alexey-p-
isaev/

Automata and Computability: A Programmer’s Perspective

Ganesh Gopalakrishnan

https://textbookfull.com/product/automata-and-computability-a-
programmers-perspective-ganesh-gopalakrishnan/
Cellular Learning Automata: Theory and Applications
Reza Vafashoar

https://textbookfull.com/product/cellular-learning-automata-
theory-and-applications-reza-vafashoar/

Cellular Automata and Discrete Complex Systems 20th

International Workshop AUTOMATA 2014 Himeji Japan July
7 9 2014 Revised Selected Papers 1st Edition Teijiro
Isokawa
https://textbookfull.com/product/cellular-automata-and-discrete-
complex-systems-20th-international-workshop-automata-2014-himeji-
japan-july-7-9-2014-revised-selected-papers-1st-edition-teijiro-
isokawa/

Theory of Groups and Symmetries: Representations of

Groups and Lie Algebras, Applications 1st Edition
Alexey P. Isaev

https://textbookfull.com/product/theory-of-groups-and-symmetries-
representations-of-groups-and-lie-algebras-applications-1st-
edition-alexey-p-isaev/

Language and Automata Theory and Applications Shmuel

Tomi Klein

https://textbookfull.com/product/language-and-automata-theory-
and-applications-shmuel-tomi-klein/

Cybercrime and Digital Forensics An Introduction Thomas

J. Holt

https://textbookfull.com/product/cybercrime-and-digital-
forensics-an-introduction-thomas-j-holt/
 LONDON MATHEMATICAL SOCIETY STUDENT TEXTS

Managing Editor: Professor D. Benson,

Department of Mathematics, University of Aberdeen, UK

49 An introduction to K-theory for C*-algebras, M. RØRDAM, F. LARSEN & N. J. LAUSTSEN

50 A brief guide to algebraic number theory, H. P. F. SWINNERTON-DYER
51 Steps in commutative algebra: Second edition, R. Y. SHARP
52 Finite Markov chains and algorithmic applications, OLLE HÄGGSTRÖM
53 The prime number theorem, G. J. O. JAMESON
54 Topics in graph automorphisms and reconstruction, JOSEF LAURI & RAFFAELE
SCAPELLATO
55 Elementary number theory, group theory and Ramanujan graphs, GIULIANA DAVIDOFF,
PETER SARNAK & ALAIN VALETTE
56 Logic, induction and sets, THOMAS FORSTER
57 Introduction to Banach algebras, operators and harmonic analysis, GARTH DALES et al
58 Computational algebraic geometry, HAL SCHENCK
59 Frobenius algebras and 2-D topological quantum field theories, JOACHIM KOCK
60 Linear operators and linear systems, JONATHAN R. PARTINGTON
61 An introduction to noncommutative Noetherian rings: Second edition, K. R. GOODEARL &
R. B. WARFIELD, JR
62 Topics from one-dimensional dynamics, KAREN M. BRUCKS & HENK BRUIN
63 Singular points of plane curves, C. T. C. WALL
64 A short course on Banach space theory, N. L. CAROTHERS
65 Elements of the representation theory of associative algebras I, IBRAHIM ASSEM, DANIEL
SIMSON & ANDRZEJ SKOWROŃSKI
66 An introduction to sieve methods and their applications, ALINA CARMEN COJOCARU &
M. RAM MURTY
67 Elliptic functions, J. V. ARMITAGE & W. F. EBERLEIN
68 Hyperbolic geometry from a local viewpoint, LINDA KEEN & NIKOLA LAKIC
69 Lectures on Kähler geometry, ANDREI MOROIANU
70 Dependence logic, JOUKU VÄÄNÄNEN
71 Elements of the representation theory of associative algebras II, DANIEL SIMSON &
ANDRZEJ SKOWROŃSKI
72 Elements of the representation theory of associative algebras III, DANIEL SIMSON &
ANDRZEJ SKOWROŃSKI
73 Groups, graphs and trees, JOHN MEIER
74 Representation theorems in Hardy spaces, JAVAD MASHREGHI
75 An introduction to the theory of graph spectra, DRAGOŠ CVETKOVIĆ, PETER
ROWLINSON & SLOBODAN SIMIĆ
76 Number theory in the spirit of Liouville, KENNETH S. WILLIAMS
77 Lectures on profinite topics in group theory, BENJAMIN KLOPSCH, NIKOLAY NIKOLOV &
CHRISTOPHER VOLL
78 Clifford algebras: An introduction, D. J. H. GARLING
79 Introduction to compact Riemann surfaces and dessins d’enfants, ERNESTO GIRONDO &
GABINO GONZÁLEZ-DIEZ
80 The Riemann hypothesis for function fields, MACHIEL VAN FRANKENHUIJSEN
81 Number theory, Fourier analysis and geometric discrepancy, GIANCARLO TRAVAGLINI
82 Finite geometry and combinatorial applications, SIMEON BALL
83 The geometry of celestial mechanics, HANSJÖRG GEIGES
84 Random graphs, geometry and asymptotic structure, MICHAEL KRIVELEVICH et al
85 Fourier analysis: Part I – Theory, ADRIAN CONSTANTIN
86 Dispersive partial differential equations, M. BURAK ERDOĞAN & NIKOLAOS TZIRAKIS
87 Riemann surfaces and algebraic curves, R. CAVALIERI & E. MILES
88 Groups, languages and automata, DEREK F. HOLT, SARAH REES & CLAAS E. RÖVER
 London Mathematical Society Student Texts 88

Groups, Languages and Automata

D E R E K F. H O LT
University of Warwick

SARAH REES
University of Newcastle upon Tyne

C L A A S E . R Ö V E R
National University of Ireland, Galway
 University Printing House, Cambridge CB2 8BS, United Kingdom
One Liberty Plaza, 20th Floor, New York, NY 10006, USA
477 Williamstown Road, Port Melbourne, VIC 3207, Australia
4843/24, 2nd Floor, Ansari Road, Daryaganj, Delhi - 110002, India
79 Anson Road, #06-04/06, Singapore 079906

Cambridge University Press is part of the University of Cambridge.

It furthers the University’s mission by disseminating knowledge in the pursuit of
education, learning and research at the highest international levels of excellence.

www.cambridge.org
Information on this title: www.cambridge.org/9781107152359
DOI: 10.1017/9781316588246

c Derek F. Holt, Sarah Rees and Claas E. Röver 2017
This publication is in copyright. Subject to statutory exception
and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written
permission of Cambridge University Press.
First published 2017
Printed in the United Kingdom by Clays, St Ives plc
A catalogue record for this publication is available from the British Library.
ISBN 978-1-107-15235-9 Hardback
ISBN 978-1-316-60652-0 Paperback
Cambridge University Press has no responsibility for the persistence or accuracy of
URLs for external or third-party Internet Web sites referred to in this publication
and does not guarantee that any content on such Web sites is, or will remain,
accurate or appropriate.
 Contents

Preface page ix

PART ONE INTRODUCTION 1

1 Group theory 3
1.1 Introduction and basic notation 3
1.2 Generators, congruences and presentations 5
1.3 Decision problems 7
1.4 Subgroups and Schreier generators 8
1.5 Combining groups 10
1.6 Cayley graphs 20
1.7 Quasi-isometries 23
1.8 Ends of graphs and groups 25
1.9 Small cancellation 26
1.10 Some interesting families of groups 28
2 Formal languages and automata theory 36
2.1 Languages, automata and grammars 36
2.2 Types of automata 38
2.3 More on grammars 50
2.4 Syntactic monoids 51
2.5 Finite state automata, regular languages and grammars 52
2.6 Pushdown automata, context-free languages and gram-
mars 60
2.7 Turing machines, recursively enumerable languages
and grammars 81
2.8 Linearly bounded automata, context-sensitive lan-
guages and grammars 85

v
vi Contents

2.9 Turing machines and decidability 87

2.10 Automata with more than one input word 91
3 Introduction to the word problem 97
3.1 Definition of the word problem 97
3.2 Van Kampen diagrams 98
3.3 The Dehn function 102
3.4 The word problem as a formal language 105
3.5 Dehn presentations and Dehn algorithms 110
3.6 Filling functions 114

PART TWO FINITE STATE AUTOMATA AND GROUPS 115

4 Rewriting systems 117
4.1 Rewriting systems in monoids and groups 117
4.2 The use of fsa in the reduction process 123
5 Automatic groups 125
5.1 Definition of automatic groups 126
5.2 Properties of automatic groups 129
5.3 Shortlex and geodesic structures 135
5.4 The construction of shortlex automatic structures 135
5.5 Examples of automatic groups 139
5.6 Closure properties 142
5.7 The falsification by fellow traveller property 143
5.8 Strongly geodesically automatic groups 145
5.9 Generalisations of automaticity 146
6 Hyperbolic groups 150
6.1 Hyperbolicity conditions 150
6.2 Hyperbolicity for geodesic metric spaces and groups 153
6.3 Thin bigons, biautomaticity and divergence 154
6.4 Hyperbolic groups have Dehn presentations 158
6.5 Groups with linear Dehn functions are hyperbolic 161
6.6 Equivalent definitions of hyperbolicity 164
6.7 Quasigeodesics 165
6.8 Further properties of hyperbolic groups 166
7 Geodesics 169
7.1 Introduction 169
7.2 Virtually abelian groups and relatively hyperbolic groups 170
7.3 Coxeter groups 173
 Contents vii

7.4 Garside groups 176

7.5 Groups with geodesics lying in some subclass of Reg 182
7.6 Conjugacy geodesics 182
8 Subgroups and coset systems 184
8.1 Rational and quasiconvex subsets of groups 184
8.2 Automatic coset systems 189
9 Automata groups 194
9.1 Introducing permutational transducers 194
9.2 Automata groups 199
9.3 Groups of tree automorphisms 203
9.4 Dual automata 210
9.5 Free automata groups 212
9.6 Decision problems 213
9.7 Resolving famous problems 215

PART THREE THE WORD PROBLEM 219

10 Solubility of the word problem 221
10.1 The Novikov–Boone theorem 221
10.2 Related results 226
11 Context-free and one-counter word problems 228
11.1 Groups with context-free word problem 228
11.2 Groups with one-counter word problem 232
12 Context-sensitive word problems 236
12.1 Lakin’s example 237
12.2 Some further examples 239
12.3 Filling length 240
12.4 Groups with linear filling length 242
13 Word problems in other language classes 249
13.1 Real-time word problems 249
13.2 Indexed word problems 254
13.3 Poly-context-free word problems 254
14 The co-word problem and the conjugacy problem 256
14.1 The co-word problem 256
14.2 Co-context-free groups 257
14.3 Indexed co-word problems 267
14.4 The conjugacy problem 268
viii Contents

References 270
Index of Notation 283
Index of Names 284
Index of Topics and Terminology 287
 Preface

This book explores connections between group theory and automata theory.
We were motivated to write it by our observations of a great diversity of such
connections; we see automata used to encode complexity, to recognise aspects
of underlying geometry, to provide efficient algorithms for practical computa-
tion, and more.
The book is pitched at beginning graduate students, and at professional aca-
demic mathematicians who are not familiar with all aspects of these intercon-
nected fields. It provides background in automata theory sufficient for its ap-
plications to group theory, and then gives up-to-date accounts of these various
applications. We assume that the reader already has a basic knowledge of group
theory, as provided in a standard undergraduate course, but we do not assume
any previous knowledge of automata theory.
The groups that we consider are all finitely generated. An element of a group
G is represented as a product of powers of elements of the generating set X,
and hence as a string of symbols from A := X ∪ X −1, also called words. Many
different strings may represent the same element. The group may be defined by
a presentation; that is, by its generating set X together with a set R of relations,
from which all equations in the group between strings can be derived. Alterna-
tively, as for instance in the case of automata groups, G might be defined as a
group of functions generated by the elements of X.
Certain sets of strings, also called languages, over A are naturally of interest.
We study the word problem of the group G, namely the set WP(G, A) of strings
over A that represent the identity element. We define a language for G to be a
language over A that maps onto G, and consider the language of all geodesics,
and various languages that map bijectively to G. We also consider combings,
defined to be group languages for which two words representing either the
same element or elements that are adjacent in the Cayley graph fellow travel;
that is, they are at a bounded distant apart throughout their length.

ix
x Preface

We consider an automaton to be a device for defining a, typically infinite, set

L of strings over a finite alphabet, called the language of the automaton. Any
string over the finite alphabet may be input to the automaton, and is then either
accepted if it is in L, or rejected if it is not. We consider automata of varying
degrees of complexity, ranging from finite state automata, which define regular
languages, through pushdown automata, defining context-free languages, to
Turing machines, which set the boundaries for algorithmic recognition of a
language. In other words, we consider the full Chomsky hierarchy of formal
languages and the associated models of computation.
Finite state automata were used by Thurston in his definition of automatic
groups after he realised that both the fellow traveller property and the finite-
ness of the set of cone types that Cannon had identified in the fundamental
groups of compact hyperbolic manifolds could be expressed in terms of reg-
ular languages. For automatic groups a regular combing can be found; use of
the finite state automaton that defines this together with other automata that
encode fellow travelling allows in particular a quadratic time solution to the
word problem. Word-hyperbolic groups, as defined by Gromov, can be charac-
terised by their possession of automatic structures of a particular type, leading
to linear time solutions to the word problem.
For some groups the set of all geodesic words over A is a regular language.
This is true for word-hyperbolic groups and abelian groups, with respect to any
generating set, and for many other groups, including Coxeter groups, virtually
free groups and Garside groups, for certain generating sets. Many of these
groups are in fact automatic.
The position of a language in the Chomsky hierarchy can be used as a mea-
sure of its complexity. For example, the problem of deciding whether an input
word w over A represents the identity of G (which, like the set it recognises,
is called the word problem) can be solved by a terminating algorithm if and
only if the set WP(G, A) and its complement can be recognised by a Turing
machine; that is, if and only if WP(G, A) is recursive. We also present a proof
of the well-known fact that finitely presented groups exist for which the word
problem is not soluble; the proof of this result encodes the existence of Turing
machines with non-recursive languages.
When the word problem is soluble, some connections can be made between
the position of the language WP(G, A) in the Chomsky hierarchy and the alge-
braic properties of the group. It is elementary to see that WP(G, A) is regular if
and only if G is finite, while a highly non-trivial result of Muller and Schupp
shows that a group has context-free word problem if and only if it has a free
subgroup of finite index.
Attaching an output-tape to an automaton extends it from a device that
 Preface xi

defines a set of strings to a function from one set of strings to another. We

call such a device a transducer, and show how transducers can be used to de-
fine groups. Among these groups are finitely generated infinite torsion groups,
groups of intermediate growth, groups of non-uniform exponential growth, it-
erated monodromy groups of post-critically finite self-coverings of the Rie-
mann sphere, counterexamples to the strong Atiyah conjecture, and many oth-
ers. Our account is by no means complete, as it concentrates on introducing
terminology, the exposition of some basic techniques and pointers to the liter-
ature.
There is a shorter book by Ian Chiswell [68] that covers some of the same
material as we do, including groups with context-free word problem and an in-
troduction to the theory of automatic groups. Our emphasis is on connections
between group theory and formal language theory rather than computational
complexity, but there is a significant overlap between these areas. We recom-
mend also the article by Mark Sapir [226] for a survey of results concerning
the time and space complexity of the fundamental decision problems in group
theory.
Acknowledgements We are grateful to Professor Rick Thomas for numer-
ous helpful and detailed discussions on the topics covered in this book. Some
of the text in Section 1.2, and the statement and proofs of Propositions 3.4.9
and 3.4.10 were written originally by Rick.
We are grateful also to Professor Susan Hermiller and to an anonymous
reviewer for a variety of helpful comments.
 P A R T ONE

INTRODUCTION
 1
Group theory

1.1 Introduction and basic notation

In this book we are assuming that the reader has studied group theory at un-
dergraduate level, and is familiar with its fundamental results, including the
basic theory of free groups and group presentations. However, in many of the
interactions between group theory and formal language theory, it is convenient
to consider group presentations as special cases of semigroup and monoid pre-
sentations, so we describe them from that aspect here.
We refer the reader to one of the standard textbooks on group theory, such
as [223] or [221] for the definitions and basic properties of nilpotent, soluble
(solvable) and polycyclic groups,
We also include some specific topics, mainly from combinatorial group the-
ory, that will be required later. The normal form theorems for free products
with amalgamation and HNN-extensions are used in the proofs of the insol-
ubility of the word problem in groups, and we summarise their proofs. We
introduce Cayley graphs and their metrical properties, and the idea of quasi-
isometry between groups, which plays a central role in the area and throughout
geometric group theory, and we define the small cancellation properties of pre-
sentations and describe related results.
The final section of the chapter is devoted to a brief introduction to some of
the specific families of groups, such as Coxeter groups and braid groups, that
arise frequently as examples throughout the book. The informed reader may
prefer not to read this chapter in detail, but to refer back to it as necessary.

1.1.1 Some basic notation For g, h in a group, we define the conjugate of g

by h, often written as gh , to be hgh−1 and the commutator [g, h] to be ghg−1 h−1 .
But we note that some authors use the notations gh and [g, h] to mean h−1 gh
and g−1 h−1 gh, respectively.

3
4 Group theory

We recall that a semigroup is a set with an associative binary operation,

usually written as multiplication, a monoid is a semigroup with an identity
element, and a group is a monoid G in which every element is invertible.
We extend the multiplication of elements of a semigroup S to its subsets,
defining T U = {tu : t ∈ T, u ∈ U} and we frequently shorten {t}U to tU, as we
do for cosets of subgroups of groups.

1.1.2 Strings and words Strings over a finite set are important for us, since
they are used to represent elements of a finitely generated group.
Let A be a finite set: we often refer to A as an alphabet. We call the elements
of A its letters, and we call a finite sequence a1 a2 · · · ak of elements from A a
string or word of length k over A. We use these two terms interchangeably. We
denote by ε the string of length 0, and call this the null string or empty word.
For a word w, we write |w| for the length of w.
We denote by Ak the set of all strings of length k over A, by A∗ the set (or
monoid) of all strings over A, and by A+ the set (or semigroup) of all nonempty
strings over A; that is

∞ 
∞
A∗ = Ak , A+ = Ak = A∗ \ {ε}.
k=0 k=1

For w = a1 a2 · · · ak and i ∈ N0 , we write w(i) for the prefix a1 a2 · · · ai of w

when 0 < i ≤ k, w(0) = ε and w(i) = w for i > k.
In this book, A often denotes the set X ∪ X −1 of generators and their inverses
for a group G; we abbreviate X ∪ X −1 as X ± . In this situation, we often refer
to words in A∗ as words over X even though they are really words over the
alphabet A.
For g ∈ G, a word w over X of minimal length that represents g is called
a geodesic word over X, and we denote the set of all such geodesic words by
G(G, X). If w is an arbitrary word representing g ∈ G, then we write |g| or |w|G
(or |g|X or |w|G,X if X needs to be specified) for the length of a geodesic word
over X that represents g. Similarly, we use v = w to mean that the words v
and w are identical as strings of symbols, and v =G w to mean that v and w
represent the same element of the group.
We call a set of strings (i.e. a subset of A∗ ) a language; the study of languages
is the topic of Chapter 2. It is convenient at this stage to introduce briefly the
notation of a language for a group.

1.1.3 Languages for groups For a group G generated by X, we call a subset

of (X ± )∗ that contains at least one representative of each element in G a lan-
guage for G; if the set contains precisely one representative of each element we
 1.2 Generators, congruences and presentations 5

call it a normal form for G. We shall be interested in finding good languages for
a group G; clearly we shall need to decide what constitutes a good language.
Typically we find good examples as the minimal representative words under a
word order, such as word length or shortlex, <slex , defined below in 1.1.4. The
shortlex normal form for a group selects the least representative of each group
element under the shortlex ordering as its normal form word. The set G(G, X)
of all geodesic words provides a natural language that is not in general a normal
form.

1.1.4 Shortlex orderings Shortlex orderings (also known as lenlex orderings)

of A∗ arise frequently in this book. They are defined as follows. We start with
any total ordering <A of A. Then, for u, v ∈ A∗ , we define u <slex v if either
(i) |u| < |v| or (ii) |u| = |v| and u is less than v in the lexicographic (dictionary)
ordering of strings induced by the chosen ordering <A of A.
More precisely, if u = a1 · · · am , v = b1 · · · bn , then u <slex v if either (i)
m < n or (ii) m = n and, for some k with 1 ≤ k ≤ m, we have ai = bi for i < k
and ak <A bk .
Note that <slex is a well-ordering whenever <A is, which of course is the case
when A is finite.

1.2 Generators, congruences and presentations

1.2.1 Generators If X is a subset of a semigroup S, monoid M or group G,
then we define SgpX, MonX or X to be the smallest subsemigroup, sub-
monoid or subgroup of S, M or G that contains X. Then X is called a semi-
group, monoid or group generating set if that substructure is equal to S, M or
G respectively, and the elements of X are called generators.
We say that a semigroup, monoid or group is finitely generated if it possesses
a finite generating set X.

1.2.2 Congruences If S is a semigroup and ∼ is an equivalence relation on S,

then we say that ∼ is a congruence if

s1 ∼ s2 , t1 ∼ t2 =⇒ s1 t1 ∼ s2 t2 .

We then define the semigroup S /∼ to be the semigroup with elements the

equivalence classes [s] = {t ∈ S : t ∼ s} of ∼, where [s1 ][s2 ] = [s1 s2 ].
6 Group theory

1.2.3 Presentations for semigroups, monoids and groups For a semigroup

S generated by a set X, let R = {(αi , βi ) : i ∈ I} be a set of pairs of words from
X + with αi =S βi for each i. The elements of R are called relations of S . If ∼
is the smallest congruence on X + containing R, and S is isomorphic to X + /∼,
then we say that R is a set of defining relations for S, and that SgpX | R is a
presentation for S . In practice, we usually write αi = βi instead of (αi , βi ). (This
is an abuse of notation but the context should make it clear that we do not mean
identity of words here.) Similarly the monoid presentation MonX | R defines
the monoid X ∗ / , for which is the smallest congruence on X ∗ containing R.
For groups the situation is marginally more complicated. If G is a group
generated by a set X and A = X ± , then G is isomorphic to A∗ /∼, where ∼ is
some congruence on A∗ containing {(aa−1 , ε), (a−1 a, ε) : a ∈ X}. We define a
relator of G to be a word α ∈ A∗ with α =G ε. Let R = {αi : i ∈ I} be a set of
relators of G. If ∼ is the smallest congruence on A∗ containing

{(α, ε) : α ∈ R} ∪ {(aa−1 , ε) : a ∈ X} ∪ {(a−1 a, ε) : a ∈ X},

and if G is isomorphic to A∗ /∼, then we say that R is a set of defining relators

for G and that X | R is a presentation for G. Rather than specifying a relator
α, so that α represents the identity, we can specify a relation β = γ (as in the
case of monoids or semigroups), which is equivalent to βγ−1 being a relator.
We say that a semigroup, monoid or group is finitely presented (or, more
accurately, finitely presentable) if it has a presentation in which the sets of
generators and defining relations or relators are both finite.

1.2.4 Exercise Let G = X | R and let A = X ± . Show that

G  MonA | IX ∪ R,

where IX = {(xx−1 , ε) : x ∈ X} ∪ {(x−1 x, ε) : x ∈ X} and R = {(w, ε) : w ∈ R}.

1.2.5 Free semigroups, monoids and groups If S is a semigroup with pre-

sentation SgpX | ∅ (which we usually write as SgpX |), then we say that S
is the free semigroup on X; we see that S is isomorphic to X + in this case. Sim-
ilarly, if M is a monoid with presentation MonX |, then we say that M is the
free monoid on X, and we see that M is then isomorphic to X ∗ . If S = X + and
L ⊆ S, then SgpL = L+ ; similarly, if M = X ∗ and L ⊆ M, then MonL = L∗ .
If F is a group with a presentation X |, then we say that F is the free group
on X; if |X| = k, then we say that F is the free group of rank k (any two free
groups of the same rank being isomorphic). We write F(X) for the free group
on X and Fk to denote a free group of rank k.
 1.3 Decision problems 7

1.2.6 Exercise Let G = X | R be a presentation of a group G. Show that the

above definition of G, which is essentially as a monoid presentation, agrees
with the more familiar definition X | R = F(X)/RF(X) , where RF(X)  de-
notes the normal closure of R in F(X).

1.2.7 Reduced and cyclically reduced words In F(X), the free group on X,
every element has a unique representation of the form w = x11 x22 . . . xnn , where
n ≥ 0, xi ∈ X and i ∈ {1, −1} for all i, and where we do not have both xi = xi+1
and i = −i+1 for any i; in this case, we say that the word w is reduced. Each
word v ∈ A∗ is equal in F(X) to a unique reduced word w.
If w is a reduced word and w is not of the form x−1 vx or xvx−1 for some
x ∈ X and v ∈ A∗ , then we say that w is cyclically reduced. Since replacing
a defining relator by a conjugate in F(X) does not change the group defined,
we may (and often do) assume that all defining relators are cyclically reduced
words.

1.3 Decision problems

In his two well-known papers in 1911 and 1912 [75, 76], Dehn defined and
considered three decision problems in finitely generated groups, the word, con-
jugacy and isomorphism problems. While the word problem in groups is one
of the main topics studied in this book, the other two will only be fleetingly
considered. A good general reference on these and other decision problems in
groups is the survey article by Miller [192].

1.3.1 The word problem A semigroup S is said to have soluble word prob-
lem if there exists an algorithm that, for any given words α, β ∈ X + , decides
whether α =S β. The solubility of the word problem for a monoid or group
generated by X is defined identically except that we consider words α, β in X ∗
or (X ± )∗ . For groups, the problem is equivalent to deciding whether an input
word is equal to the identity element. The word problem for groups is dis-
cussed further in Chapter 3 and in Part Three of this book. Examples of finitely
presented semigroups and groups with insoluble word problem are described
in Theorems 2.9.7 and 10.1.1.

1.3.2 The conjugacy and isomorphism problems The conjugacy problem

in a semigroup S is to decide, given two elements x, y ∈ S, whether there exists
z ∈ S with zx = yz. Note that this relation is not necessarily symmetric in x and
8 Group theory

y, but in a group G it is equivalent to deciding whether x and y are conjugate

in G.
Since the word problem in a group is equivalent to deciding whether an el-
ement is conjugate to the identity, the conjugacy problem is at least as hard as
the word problem, and there are examples of groups with soluble word problem
but insoluble conjugacy problem. A number of such examples are described in
the survey article by Miller [192], including Theorem 4.8 (an extension of one
finitely generated free group by another), Theorem 4.11 (examples showing
that having soluble conjugacy problem is not inherited by subgroups or over-
groups of index 2), Theorem 5.4 (residually finite examples), Theorem 6.3 (a
simple group), Theorem 7.7 (asynchronously automatic groups), and Theorem
7.8 (groups with finite complete rewriting systems) of that article.
The isomorphism problem is to decide whether two given groups, monoids
or semigroups are isomorphic. Typically the input is defined by presentations,
but could also be given in other ways, for example as groups of matrices. There
are relatively few classes for which the isomorphism problem is known to be
soluble. These classes include polycyclic and hyperbolic groups [232, 234, 72].

1.3.3 The generalised word problem Given a subgroup H of a group G, the

generalised word problem is to decide, given g ∈ G, whether g ∈ H. So the
word problem is the special case in which H is trivial. We shall encounter
some situations in which this problem is soluble in Chapter 8. As for the con-
jugacy problem, the survey article [192] is an excellent source of examples (in
particular in Theorems 5.4 and 7.8 of that article), in this case of groups with
soluble word problem that have finitely generated subgroups with insoluble
generalised word problem.

1.4 Subgroups and Schreier generators

Let H be a subgroup of a group G = X, and let U be a right transversal of H
in G. For g ∈ G, denote the unique element of Hg ∩ U by g. Define
 
Z := uxux−1 : u ∈ U, x ∈ X .
Then Z ⊆ H.
1.4.1 Theorem With the above notation, we have H = Z.
Our proof needs the following result.
−1
1.4.2 Lemma Let S = {ux−1 ux−1 : u ∈ U, x ∈ X}. Then Z −1 = S .
 1.4 Subgroups and Schreier generators 9

Proof Let g ∈ Z −1 , so g = (uxux−1 )−1 = uxx−1 u−1 . Let v := ux ∈ U. Then,

since the elements vx−1 and u are in the same coset of H, we have vx−1 = u,
−1
and g = vx−1 vx−1 ∈ S .
−1
Conversely, let g = ux−1 ux−1 ∈ S, so g−1 = ux−1 xu−1 . Let v := ux−1 . Then
vx = u, so g−1 = vxvx−1 ∈ Z and g ∈ Z −1 . 

Proof of Theorem 1.4.1 Let U ∩ H = {u0 }. (We usually choose u0 = 1, but

this is not essential.) Let h ∈ H. Then we can write u−1 0 hu0 = a1 · · · al for some
ai ∈ A := X ± . For 1 ≤ i ≤ l, let ui := a1 · · · al . Since u−1
0 hu0 ∈ H, we have
ul = u0 . Then
h =G (u0 a1 u−1 −1 −1
1 )(u1 a2 u2 ) · · · (ul−1 al ul ).

Note that ui+1 = a1 · · · al+1 is in the same coset of H as ui ai+1 , so ui ai+1 = ui+1 ,
and
h =G (u0 a1 u0 a1 −1 )(u1 a2 u1 a2 −1 ) · · · (ul−1 al ul−1 al −1 ). (†)

Each bracketed term is in Z if ai ∈ X, and in Z −1 if ai ∈ X −1 by Lemma 1.4.2.

So H = Z. 

1.4.3 Corollary A subgroup of finite index in a finitely generated group is

finitely generated.

1.4.4 Rewriting The process described in the above proof of calculating a

word v over Z from a word w over X that represents an element of H is called
Reidemeister–Schreier rewriting. We may clearly omit the identity element
from the rewritten word, which results in a word over Y = Z \ {1}, which
we denote by ρX,Y (w). From the proof, we see immediately that:

1.4.5 Remark If 1 ∈ U, then |ρX,Y (w)| ≤ |w|.

1.4.6 Schreier generators and transversals The above set Y of non-identity

elements of Z is called the set of Schreier generators of H in G. Of course, this
set depends on X and on U.
The set U is called a Schreier transversal if there is a set of words over X
representing the elements of U that is closed under taking prefixes. Note that
such a set must contain the empty word, and hence 1 ∈ U. By choosing the
least word in each coset under some reduction ordering of A∗ (where A = X ± ),
it can be shown that Schreier transversals always exist. Reduction orderings
are defined in 4.1.5. They include the shortlex orderings defined in 1.1.4.
It was proved by Schreier [228] that, if G is a free group and U is a Schreier
transversal, then the Schreier generators freely generate H.
10 Group theory

The following result, known as the Reidemeister–Schreier Theorem, which

we shall not prove here, provides a method of computing a presentation of
the subgroup H from a presentation of the group G. Note that it immediately
implies the celebrated Nielsen–Schreier Theorem, that any subgroup of a free
group is free. As with many of the results stated in this chapter, we refer the
reader to the standard textbook on combinatorial group theory by Lyndon and
Schupp [183] for the proof.

1.4.7 Theorem (Reidemeister–Schreier Theorem [183, Proposition II.4.1])

Let G = X | R = F/N be a group presentation, where F = F(X) is the free
group on X, and let H = E/N ≤ G. Let U be a Schreier transversal of E in
F and let Y be the associated set of Schreier generators. Then Y | S  with
S = ρX,Y (uru−1 ) : u ∈ U, r ∈ R is a presentation of H.

1.4.8 Corollary A subgroup of finite index in a finitely presented group is

finitely presented.

1.5 Combining groups

In this section we introduce various constructions that combine groups. We
leave the details of many of the proofs of stated results to the reader, who is
referred to [183, Chapter IV] for details.

1.5.1 Free products Informally, the free product G ∗ H of the groups G, H is

the largest group that contains G and H as subgroups and is generated by G
and H. Formally, it can be defined by its universal property:

(i) there are homomorphisms ιG : G → G ∗ H and ιH : H → G ∗ H;

(ii) if K is any group and τG : G → K, τH : H → K are homomorphisms,
then there is a unique homomorphism α : G ∗ H → K with αιG = τG and
αιH = τH .

As is often the case with such definitions, it is straightforward to prove

uniqueness, in the sense that any two free products of G and H are isomor-
phic, and it is not hard to show that G ∗ H is generated by ιG (G) and ιH (H). But
the existence of the free product is not immediately clear.
To prove existence, let G = X | R and H = Y | S  be presentations of G
and H. Then we can take

G ∗ H = X ∪ Y | R ∪ S ,
 1.5 Combining groups 11

where ιG and ιH are the homomorphisms induced by the embeddings X →

X ∪ Y and Y → X ∪ Y; we tacitly assumed that X and Y are disjoint.
It is not completely obvious that ιG and ιH are monomorphisms. This follows
from another equivalent description of G ∗ H as the set of alternating products
of arbitrary length (including length 0) of non-trivial elements of G and H,
with multiplication defined by concatenation and multiplications within G and
H. With this description, ιG and ιH are the obvious embeddings, and G and
H are visibly subgroups of G ∗ H, known as the free factors of G ∗ H. The
equivalence of the two descriptions follows immediately in a more general
context from Proposition 1.5.12.
The definition extends easily to the free product of any family of groups.
The following result, which we shall not prove here, is used in the proof of the
special case of the Muller–Schupp Theorem (Theorem 11.1.1) that torsion-free
groups with context-free word problem are virtually free.

1.5.2 Theorem (Grushko’s Theorem [183, IV.1.9]) For a group G, let d(G)
denote the minimal number of generators of G. Then d(G ∗ H) = d(G) + d(H).

1.5.3 Direct products The direct product G×H of two groups G, H is usually
defined as the set G × H with component-wise multiplication. We generally
identify G and H with the component subgroups, which commute with each
other, and are called the direct factors of G×H. Then each element has a unique
representation as a product of elements of G and H. It can also be defined by a
universal property:

(i) there are homomorphisms πG : G × H → G and πH : G × H → H;

(ii) if K is any group and τG : K → G and τH : K → H are homomorphisms,
then there is a unique homomorphism ϕ : K → G × H with τG = πG ◦ ϕ
and τH = πH ◦ ϕ.

If G = X | R and H = Y | S  are presentations, then G × H has the

presentation

G × H = X ∪ Y | R ∪ S ∪ {[x, y] : x ∈ X, y ∈ Y}.

We can extend this definition to direct products of families of groups as fol-

lows. Let {Gω : ω ∈ Ω} be a family of groups. Then the (full) direct product,

also known sometimes as the Cartesian product, ω∈Ω Gω of the family con-
sists of the set of functions β : Ω → ∪ω∈ΩGω for which β(ω) ∈ Gω for all
ω ∈ Ω, where the group operation is component-wise multiplication in each
Gω ; that is, β1 β2 (ω) = β1 (ω)β2 (ω) for all ω ∈ Ω.
12 Group theory

The elements of ω∈Ω Gω consisting of the functions β with finite support
(i.e. β(ω) = 1G for all but finitely many ω ∈ Ω) form a normal subgroup of

ω∈Ω G ω . We call this subgroup the restricted direct product of the family
{Gω : ω ∈ Ω}. It is also sometimes called the direct sum of the family to
distinguish it from the direct product.

1.5.4 Semidirect products Let N and H be groups, and let φ : H → Aut(N)

be a right action of H on N. We define the semidirect product of H and N,
written N φ H or just N  H, to be the set {(n, h) : n ∈ N, h ∈ H} equipped
with the product
φ(h−1
1 )
(n1 , h1 )(n2 , h2 ) = (n1 n2 , h1 h2 ).
We leave it as an exercise to the reader to derive a presentation of N φ H from
presentations of H and N and the action φ. We note that sometimes the notation
H  N is used for the same product. We identify the subgroups {(n, 1) : n ∈ N}
and {(1, h) : h ∈ H} with N and H, and hence (n, h) with nh, so that the
expression above reads
φ(h−1
1 )
n1 h1 n2 h2 = n1 n2 h1 h2 .
The direct product N × H is the special case when φ is the trivial action.
The semidirect product is itself a special case of a group extension, which is a
group G with normal subgroup N and G/N  H. Unfortunately roughly half of
the set of mathematicians refer to this as an extension of N by H, and the other
half call it an extension of H by N. An extension is isomorphic to a semidirect
product if and only if N has a complement in G (that is, G has a subgroup K,
with N ∩ K = {e}, G = NK), in which case it is also called a split extension.
Note that we can also define a semidirect product of two groups N and H,
from a left action of H on N.

1.5.5 Wreath products Let G and H be groups and suppose that we are given
a right action φ : H → Sym(Ω) of H on the set Ω. We define the associated
(full) permutational wreath product G  H = G φ H as follows.

Let N = ω∈Ω Gω , where the groups Gω are all equal to the same group
G. So the elements of N are functions γ : Ω → G. We define a right action
−1
ψ : H → Aut(N) by putting γψ(h) (ω) = γ(ωφ(h ) ) for each γ ∈ N, h ∈ H, and
ω ∈ Ω. We then define G φ H to be the semidirect product N ψ H. So the
elements have the form (γ, h) with γ ∈ N and h ∈ H. As in 1.5.4, we identify
{(γ, 1) : γ ∈ N}, {(1, h) : h ∈ H} with N and H, and hence (γ, h) with the
product γh.
If we restrict elements of N to the functions γ : Ω → G with finite support,
 1.5 Combining groups 13

then we get the restricted wreath product, which we shall write as G R H or

G Rφ H.
The special case in which φ is the right regular action of H (i.e. Ω = H and
φ(h2 )
h1 = h1 h2 for h1 , h2 ∈ H) is known as the standard or restricted standard
wreath product. This is the default meaning of G  H or G R H when the action
φ is not specified.
Finally we mention that, if we are given a right action ρ : G → Sym(Δ),
then we can define an action ψ : G φ H → Sym(Δ × Ω) by setting (δ, ω)ψ(γ,h) =
(δρ◦γ(ω) , ωφ(h) ). This right action plays a central role in the study of imprimitive
permutation groups, but it will not feature much in this book.

1.5.6 Exercise Show that the restricted standard or permutational wreath prod-
uct G R H is finitely generated if both G and H are finitely generated. Verify
also that G  H is not finitely generated unless H is finite and G is finitely
generated.

1.5.7 Graph products Let Γ be a simple undirected graph with vertices la-
belled from a set I, and let Gi (i ∈ I) be groups. Then the graph product of the
Gi with respect to Γ can be thought of as the largest group G generated by the
Gi such that [Gi , G j ] = 1 whenever {i, j} is in the set E(Γ) of edges of Γ.
If Xi | Ri  is a presentation of Gi for each i, then
∪i∈I Xi | ∪i∈I Ri ∪ {[xi , x j ] : xi ∈ Xi , x j ∈ X j , i, j ∈ I, {i, j} ∈ E(Γ)}
is a presentation of the graph product.
Note that the right-angled Artin groups (see 1.10.4) can be described equiv-
alently as graph products of copies of Z.

1.5.8 Free products with amalgamation The amalgamated free product gen-
eralises the free product. Suppose that G and H are groups with subgroups
A ≤ G, B ≤ H, and that there is an isomorphism φ : A → B.
Informally, the free product G ∗A H of G and H amalgamated over A (via φ)
is the largest group P with G, H ≤ P, G, H = P, and a = φ(a) for all a ∈ A.

1.5.9 Example Suppose that Γ = G, H and G ∩ H = A, where A is a sub-

group of both G and H with |G : A| ≥ 3 and |H : A| ≥ 2. Suppose also that Γ
acts on the left on a set Ω and that Ω1 , Ω2 are subsets of Ω with Ω1  Ω2 , such
that

(1) (G \ A)(Ω1 ) ⊆ Ω2 and (H \ A)(Ω2 ) ⊆ Ω1 ;

(2) A(Ωi ) ⊆ Ωi for i = 1, 2.
14 Group theory

Then Γ  G ∗A H.
This result is often known as the ping-pong lemma. It is proved in [74,
IIB.24] for the case A = 1 but essentially the same proof works for general
A. The reader could attempt it as an exercise, using Corollary 1.5.13 below.

1.5.10 Exercise Let Γ = SL(2, Z) = x, y with

   
0 1 1 −1
x= and y = .
−1 0 1 0
Let G = y, H = x, and A = G ∩ H = y3  = x2 . Show that Γ  G ∗A H by
taking Ω = Z2 , Ω1 = {(x, y) ∈ Ω : xy < 0} and Ω2 = {(x, y) ∈ Ω : xy > 0}.

1.5.11 Example As a consequence of the Seifert–van Kampen Theorem [189,

Chapter 4, Theorem 2.1], we see that, for a topological space X = Y ∪ Z for
which Y, Z and Y ∩ Z are open and path-connected, and the fundamental group
π1 (Y ∩ Z) embeds naturally into π1 (Y) and π1 (Z), the fundamental group of X
is isomorphic to the free product with amalgamation π1 (Y) ∗π1 (Y∩Z) π1 (Z) (see
[183, IV.2]).
Formally, G ∗A H can be defined by the following universal property:
(i) there are homomorphisms ιG : G → G ∗A H and ιH : H → G ∗A H with
ιG (a) = ιH (φ(a)) for all a ∈ A;
(ii) if K is any group and τG : G → K, τH : H → K are homomorphisms with
τG (a) = τH (φ(a)) for all a ∈ A, then there is a unique homomorphism
α : G ∗A H → K with αιG = τG and αιH = τH .
The uniqueness of G ∗A H up to isomorphism follows easily, but not its exis-
tence, which is most conveniently established using presentations, as follows.
Let G = X | R and H = Y | S  be presentations of G and H. For each
element of a ∈ A, let wa and va be words over X and Y representing a and φ(a),
respectively, and put T := {wa = va : a ∈ A}. Then, as in [183, IV.2], we define
G ∗A H := X ∪ Y | R ∪ S ∪ T ,
and it is straightforward to show, using standard properties of group presenta-
tions, that G ∗A H has the above universal property, where ιG and ιH are defined
to map words in G and in H to the same words in G ∗A H.
Note that, in the definition of T , it would be sufficient to restrict a to the
elements of a generating set of A so, if G and H are finitely presented and A is
finitely generated, then G ∗A H is finitely presentable.
But we have still not proved that G and H are subgroups of G ∗A H; that
is, that ι1 and ι2 are embeddings. We do that by finding a normal form for the
 1.5 Combining groups 15

elements of G ∗A H. Let U and V be left transversals of A in G and B ∈ H,

respectively, with 1G ∈ U, 1H ∈ V. From now on, we shall suppress the maps
ιG , ιH and just write g rather than ιG (g).
1.5.12 Proposition Every element of G ∗A H has a unique expression as
t1 · · · tk a for some k ≥ 0, where a ∈ A, ti ∈ (U \ {1G }) ∪ (V \ {1H }) for 1 ≤ i ≤ k
and, for i < k, ti ∈ U ⇔ ti+1 ∈ V.
In particular, since distinct elements of G and of H give rise to distinct
expressions of this form, G and H embed into G ∗A H as subgroups, and
G ∩ H = A = φ(A).
Proof By definition, each f ∈ G ∗A H can be written as an alternating product
of elements of G and H, and working from the left and writing each such
element as a product of a coset representative and an element of A (which has
been identified with φ(A) = B), we can write f in the specified normal form.
Let Ω be the set of all normal form words. We define a right action of G ∗A H
on Ω, which corresponds to multiplication on the right by elements of G ∗A H.
To do this, it is sufficient to specify the actions of G and of H, which must of
course agree on the amalgamated subgroup.
Let α = t1 · · · tk a ∈ Ω and g ∈ G. If k = 0 or tk ∈ V, then we define
αg = t1 · · · tk tk+1 a , where tk+1 a =G ag. Otherwise, k > 0 and tk ∈ U, and we
put αg = t1 · · · tk−1 tk+1 a , where tk+1 a =G tk ag. In both cases, tk+1 ∈ U, a ∈ A
and we omit tk+1 if it is equal to 1. We define the action of H on Ω similarly.
It is easy to see that these definitions do indeed define actions of G and H
on Ω that agree on the amalgamated subgroup A, so we can use them to define
the required action of G ∗A H on Ω. This follows from the universal property of
G ∗A H. It is also clear from the definition that, taking α = ε ∈ Ω, and f to be
the element of G ∗A H defined by the normal form word t1 · · · tk−1 tk a, we have
α f = t1 · · · tk−1 tk a. So the elements of G represented by distinct normal form
words have distinct actions on Ω, and hence they cannot represent the same
element of G ∗A H. 
1.5.13 Corollary Suppose that f = f1 f2 · · · fk ∈ G ∗A H with k > 0, where
fi ∈ (G \ A) ∪ (H \ B) for 1 ≤ i ≤ k and, for i < k, fi ∈ G ⇔ fi+1 ∈ H. Then f
is not equal to the identity in G ∗A H.
Conversely, suppose that the group F is generated by subgroups (isomorphic
to) G and H with G ∩ H = A, where a =F φ(a) for all a ∈ A, and that f  1
for every element f = f1 f2 · · · fk ∈ F with k > 0, fi ∈ (G \ A) ∪ (H \ B) for
1 ≤ i ≤ k and, for i < k, fi ∈ G ⇔ fi+1 ∈ H. Then F  G ∗A H.
Proof The assumptions ensure that, when we put f into normal form as de-
scribed in the above proof, the resulting expression has the form t1 · · · tk a with
16 Group theory

the same k and, since we are assuming that k > 0, this is not the representative
ε of the identity element.
For the converse, observe that the hypothesis implies that the normal form
expressions for elements of F described in Proposition 1.5.12 represent distinct
elements of F, and so the map α : G ∗A H → F specified by (ii) of the definition
of the G ∗A H is an isomorphism. 
A product f = f1 f2 · · · fk as in the above corollary is called a reduced form
for f . It is called cyclically reduced if all of its cyclic permutations are reduced
forms. Every element of G ∗A H is conjugate to an element u = f1 · · · fn in
cyclically reduced form, and every cyclically reduced conjugate of u can be
obtained by cyclically permuting f1 · · · fn and then conjugating by an element
of the amalgamated subgroup A [183, page 187].
If f1 f2 · · · fk is a reduced form with k > 1 and ( f1 f2 · · · fk )n is not a reduced
form for some n > 0, then fk f1 cannot be reduced, and so f1 f2 · · · fk cannot be
cyclically reduced. This proves the following result.
1.5.14 Corollary [183, Proposition 12.4] An element of finite order in G∗A H
is conjugate to an element of G or to an element of H.

1.5.15 HNN-extensions Suppose now that A and B are both subgroups of the
same group G, and that there is an isomorphism φ : A → B. The corresponding
HNN-extension, (due to Higman, Neumann and Neumann [141]) with stable
letter t, base group G and associated subgroups A and B, is roughly the largest
group G∗A,t that contains G as a subgroup, and is generated by G and an extra
generator t such that t−1 at = φ(a) for all a ∈ A.

1.5.16 Example Analogously to Example 1.5.11, suppose that we have a path-

connected topological space Y with two homeomorphic open subspaces U and
V, of which the fundamental groups embed into that of Y, and suppose that we
form a new space X by adding a handle that joins U to V using the homeo-
morphism between them. Then it is a consequence of the Seifert–van Kampen
Theorem that the fundamental group π1 (X) of X is isomorphic to the HNN-
extension π1 (Y) ∗π1 (U),t ; see [183, IV.2].
We can also define an HNN-extension of G = X | R via a presentation.
Again, for a ∈ A, we let wa and va be words over X representing a and φ(a),
respectively, and define
G∗A,t := X, t | R ∪ T , where T := {t−1 wa t = va : a ∈ A}.
Again, we can restrict the elements a in T to a generating set of A, so G∗A,t
is finitely presentable if G is finitely presented and A is finitely generated.
 1.5 Combining groups 17

There is a homomorphism ι : G → G∗A,t that maps each word over X to

the same word in G∗A,t . Once again, we can use a normal form to prove that
ι embeds G into G∗A,t , and we shall henceforth suppress ι and write g rather
than ι(g). Let U and V be left transversals of A and B in G, respectively, with
1G ∈ U, 1H ∈ V.

1.5.17 Proposition Every element of G∗A,t has a unique expression as

t j0 g1 t j1 g2 · · · gk t jk gk+1 for some k ≥ 0, where

(i) gi ∈ G for 1 ≤ i ≤ k + 1 and gi  1 for 1 ≤ i ≤ k;

(ii) ji ∈ Z for 0 ≤ i ≤ k and ji  0 for 1 ≤ i ≤ k;
(iii) for 1 ≤ i ≤ k, we have gi ∈ U if ji > 0, and gi ∈ V if ji < 0.

In particular, since distinct elements of G give rise to distinct expressions of

this form with k = 0 and j0 = 0, G embeds as a subgroup of G∗A,t .

Proof Clearly each f ∈ G∗A,t can be written as t j0 g1 t j1 g2 · · · gk t jr gk+1 for

some k ≥ 0 such that (i) and (ii) are satisfied. If j1 > 0, then we write g1 as g1 a
with g1 ∈ U and a ∈ A and, using the relation t−1 at = φ(a), replace at in the
word by tφ(a). Similarly, if j1 < 0, then we write g1 as g1 b with g1 ∈ V and
b ∈ B, and replace bt−1 in the word by t−1 φ−1 (b). By working through the word
from left to right making these substitutions, we can bring f into the required
normal form (i.e. satisfying (i), (ii) and (iii)).
Let Ω be the set of all normal form words. We define a right action of G∗A,t
on Ω, which corresponds to multiplication on the right by elements of G∗A,t .
To do this, it is sufficient to specify the actions of G and of t provided that, for
each a ∈ A the action of a followed by that of t is the same as the action of
t followed by that of φ(a). In fact it is more convenient to specify the actions
of t and t−1 separately and then check that they define inverse mappings. Let
α = t j0 g1 t j1 g2 · · · gk t jk gk+1 ∈ Ω.
If g ∈ G, then we define αg := t j0 g1 t j1 g2 · · · gk t jk gk+1 , where gk+1 = gk+1 g.
We need to subdivide into three cases for the action of t.

(a) If gk+1  A, then we write gk+1 = gk+1 a with 1  gk+1 ∈ U and a ∈ A, and
αt := t j0 g1 t j1 g2 · · · gk t jk gk+1 tφ(a);
(b) If gk+1 ∈ A and jk  −1, then αt := t j0 g1 t j1 g2 · · · gk t jk +1 φ(gk+1 );
(c) If gk+1 ∈ A and jk = −1, then αt := t j0 g1 t j1 g2 · · · t jk−1 gk , where
gk = gk φ(gk+1 ) (and gk = 1 if k = 0).

We have the corresponding three cases for the action of t−1 .

(a) If gk+1  B, then we write gk+1 = gk+1 b with 1  gk+1 ∈ V and b ∈ B, and
−1
αt := t j0 g1 t j1 g2 · · · gk t jk gk+1 t−1 φ−1 (b);
Another random document with
no related content on Scribd:
The Project Gutenberg eBook of Infiltration
This ebook is for the use of anyone anywhere in the United States
and most other parts of the world at no cost and with almost no
restrictions whatsoever. You may copy it, give it away or re-use it
under the terms of the Project Gutenberg License included with this
ebook or online at www.gutenberg.org. If you are not located in the
United States, you will have to check the laws of the country where
you are located before using this eBook.

Title: Infiltration

Author: Algis Budrys

Release date: November 4, 2023 [eBook #72026]

Language: English

Original publication: New York, NY: Royal Publications, Inc, 1958

Credits: Greg Weeks, Mary Meehan and the Online Distributed

Proofreading Team at http://www.pgdp.net

*** START OF THE PROJECT GUTENBERG EBOOK

INFILTRATION ***
 INFILTRATION

By ALGIS BUDRYS

If werewolves exist, they don't necessarily

conform to all the superstitions people have.
They may even know fear....

[Transcriber's Note: This etext was produced from

Infinity October 1958.
Extensive research did not uncover any evidence that
the U.S. copyright on this publication was renewed.]
 I
Sunset. They're coming for me, tonight, he knew as he woke.
Sunset. Not really—if he were to get dressed now, and go out on the
street, the red globe would still be hanging over the cliffs of New
Jersey. But the shadow of the building next door had fallen over his
apartment windows, and he sleepily pushed a cigarette between his
numb lips and swung his feet over the side of the bed, fumbling with a
match as he walked over to the small radio on the windowsill and
turned it on. There was a double-header between the Giants and
Cincinnati—the first game was probably in its last inning.
Sunset—odd, how the conditioning worked. Was it conditioning? Or
were the old wives' tales not so absurd, after all? But he could go out
in the sunlight—had done it many times. His tan proved it. He
touched silver and cold iron countless times each day, crossed
running water—and he'd gone to church every Sunday, until he was
twelve. No, there was a core of truth under the fantastically complex
shell of nonsense, but the old limitations were not part of it. He
shrugged. Neither were most of the powers.
Still, he liked to sleep in the daytime. His schedule seemed to gain an
hour at night, lose one in the morning, until, almost unnoticeably, it
had slipped around the clock.
He went into the bathroom while the worn tubes in the radio warmed
up slowly, and washed his face, brushed his teeth, shaved. He
combed his hair, then paused thoughtfully. Wouldn't do any harm. No
full moon in here, either, he thought, looking up at the circular
fluorescent tube in the ceiling, but he noticed no impediment as he
coalesced, dropped to all fours, and ran his pelt against the curry-
combs he had screwed to the bathroom door. He did a thorough job,
enjoying it, and, after he had realigned, walked out of the bathroom in
time to hear the Giants making their final, fruitless out of the first
game. Five-Zero, Cincinnati, and he grimaced in disgust. Four shut-
outs in the last five games.
He laughed at himself, then, for actually being annoyed. Still and all, it
wasn't the first time a man became emotionally involved in a mirage.
Was it a mirage? True, there weren't really any such things as the
San Francisco Giants—but a man could certainly be expected to
forget that, occasionally, if he were part of the same illusion at least
half the time. And certainly, such stuff as dreams are made of is solid
enough when you are yourself a dream.
He went out in the kitchen and started coffee, then came back and
sat down next to the radio, hardly listening to the recap of the game.

Odd, how it had all started. Being suddenly marooned on this planet,
forced to survive, somehow, through the long years while waiting for
rescue. How many years had it been, now? Some five hundred
thousand, in the subjective reference for this particular universe. He
knew the formula for conversion into objective time—it all worked out
to the equivalent of about six months—but that wasn't what mattered,
as long as they'd all had to survive in this universe.
Sleep—suspended animation, if you wanted to call it that—had been
the only answer. And they couldn't do that, directly. They'd had to
resort to chrysalids.
He smiled to himself, got up, and turned down the fire under the pot
until the coffee was percolating softly.
The original plan had snowballed, somewhat.
Resolving chrysalids was one thing—making them eternal was
another, and unnecessary. It was far simpler to arrange the chrysalids
so they'd be able to reproduce themselves. And, of course, in order to
survive, and take care of itself, a chrysalis had to have some
independent intelligence.
And, so it worked. The chrysalis housed a sleeper, operating
unawares and completely independent of him—or her—until the
chrysalis wore out. Then the sleeper was passed on to a new
chrysalis, with neither of the chrysalids involved—nor, for that matter,
the sleeper—conscious of the transfer. So it would continue, through
the weary, subjective years; generation upon generation of
chrysalids, until, finally, the paramathematical path drifted back to
touch this universe, and the sleepers could wake, and continue their
journey.
And if the human race chose to speculate on its origins in the
meantime, well, that was part of the snowball.
He got up again, and turned off the flame under the coffeepot. Now, if
I were a sorcerer—as defined by Cotton Mather's ilk, of course—, he
thought, I should be able to (a) turn the fire off without getting up, or,
(b) generate the flame without the use of Con Edison's gas, or, (c), if I
had any self-respect at all, conjure hot coffee out of thin air. His lips
twisted with nausea as he thought that nine out of ten people would
expect him to be drinking blood, as a matter of course.
He sighed with some bitterness, but more of resignation. Well, that
was just another part of the snowball.
Because the chrysalids had done a magnificent job in all three of its
subdivisions. They had kept the sleepers safe—and reproduced, and
used their intelligence to survive. They had survived in spite of
pestilence, famine, and flood—by learning enough to wipe out the
first two, and control the third. It would seem that progress was not a
special quality to be specially desired. Most of the chrysalids were
consumed by a fierce longing for the Good Old Days, as a matter of
fact. It was merely the inescapable accretion to sheer survival.
And so came civilization. With civilization: recreation. In short, the
San Francisco Giants, and—He reached over, suddenly irritated at
the raspy-voiced and slightly frantic recapitulation of the lost
ballgame, and changed the station. And Beethoven.
He relaxed, smiling slightly at himself once again, and let the music
sing to him. Chrysalids, eh? Well, they certainly weren't his kind of
life, free to swing from star to star, riding the great flux of Creation
from universe to universe. But whence Beethoven? Whence
Rembrandt, Da Vinci, and Will Shakespeare, hunched over a mug of
ale and dashing off genius on demand, with half an eye on the
serving wench?
He shook his head. What would happen to this people, when the
sleepers woke?
The snowball. Ah, yes, the snowball. That was a good part of it—and
he and his kind were another.
If we had known, he thought. If we had known how it would be...?
But, they hadn't known. It had been just a petty argument, at first.
Nobody knew, now, who had started it. But there were two well-
defined sides, now, and he was an Insurgent, for some reason. The
winning side gives the names that stick. They were Watchers—an
honorable name, a name to conjure up trust, and duty, and loyalty.
And he was an Insurgent. Well, let it stand. Accept the heritage of
dishonor and hatred. Somewhere, sometime, a gage was flung, and
he was heir to the challenge.
The chrysalids solved the problem of survival, of course. But the
problem of rescue had remained. For rescue, in the sense of help
from an outside agency, would be disastrous. When the path shifted
back, they had to learn of it themselves, and go on of their own
accord—or go into slavery. For there is one currency that outlives
document and token. Personal obligation. And, if they were so
unlucky as to have an actual rescuer, the obligation would be high—
prohibitively so.
The solution had seemed simple, at first. In each generation of
chrysalids, there would be one aware individual—one Watcher, to
keep guard, and to waken the rest should the path drift back in the
lifetime of his chrysalis. Then, when that particular chrysalis wore out,
the Watcher would be free to return to sleep, while another took his
place.
His mouth twisted to one side as he took a sip of coffee.
A simple, workable plan—until someone had asked, "Well and good.
Excellent. And what if this high-minded Watcher realizes that we,
asleep, are all in his power? What if he makes some agreement with
a rescuer, or, worse still, decides to become our rescuer when the
path drifts back? What's to prevent him, eh? No," that long-forgotten,
wary individual had said, "I think we'd best set some watchers to
watch the Watcher."
Quis custodiet?
What had it been like? He had no way of knowing, for he had no
memory of his exact identity. That would come only with Awakening.
He had only a knowledge of his heritage. For all he knew, it had been
he who raised the fatal doubt—or, had been the first delegated
Watcher. He shrugged. It made no difference. He was an Insurgent
now.
But he could imagine the voiceless babel among their millions—the
argument, the cold suspicion, the pettiness. Perhaps he was passing
scornful judgment on himself, he realized. What of it? He'd earned it.
So, finally, two groups. One content to be trustful. And the other a
fitful, restless clan, awakening sporadically, trusting to chance alone,
which, by its laws, would insure that many of them were awake when
the path drifted back. The Insurgents.
So, as well, two basic kinds of chrysalids. The human kind, and the
others. Wolves, bears, tigers. Bats, seals—every kind of living thing,
except the human. The Insurgent kind.
And so the struggle began. It was a natural outgrowth of the
fundamental conflict. Which side had tried to over-power the first
chrysalis? Who first enslaved another man? he thought, and half-
snarled.
That, too, was unimportant now. For the seed had been planted. The
thought was there. Those who are awake can place those who sleep
under obligation. Control the chrysalids, and you control the sleepers
within. But chrysalids endure for one generation, and then the
sleepers pass on.
What then? Simplicity. Group your chrysalids. Segregate them. Set
up pens for them, mark them off, and do it so the walls and fences
endure through long years.
This is my country. All men are brothers, but stay on your side of the
line, brother.
Sorry, brother—you've got a funny shape to your nose. You just go
live in that nice, walled-off part of my city, huh, brother?
Be a good fellow, brother. Just move to the back of the bus, or I'll
lynch you, brother.
And the chrysali die, the sleepers transfer—into another chrysalis in
the same pen. SPQR. Vive, Napoleon! Sieg Heil!

Some of the time it was the Watchers. Some of the time it was the
Insurgents. And some of the time, of course, the chrysalids evolved
their own leaders, and imitated. For, once the thing had begun, it
could not be stopped. The organization was always more powerful
than the scattered handsful. So, the only protection against
organization was organization.
But it was not organization in itself that was the worst of it. It was the
fact that the only way to control the other side's penned chrysalids
was to break down a wall in the pen, or to build a larger pen including
many of the smaller ones.
And, again, it was too late, now, to decide who had been at fault.
Who first invented War?
The way to survive war is to wage more decisive war. The chrysalids
had to survive. They learned. They ... progressed? ... by so doing.
They progressed from bows to ballistas to bombs. From arbalests to
aircraft to A-bombs. Phosphorus. Chlorine. HE. Fragmentation.
Napalm. Dust, and bacteriological warfare. Thermopylae, Crecy, the
Battle of Britain, Korea, Indo-China, Indonesia.
And try to believe as you sit here, Insurgent, that none of this is real,
that it is all a phase, acted out by dolls of your own creation in a sham
battle that is really only a bad dream in the unfamiliar bed of a lodging
for the night!
Chrysalids they might be, Insurgent, he lashed himself, but it was the
greed and suspicion of all your kind—Insurgent and Watcher alike—
that set this juggernaut to rolling!
He took another sip of coffee, and almost gagged as he realized it
had grown cold. He got up and walked into the kitchen with the cup in
his hand. He threw the rest of the coffee in the sink, washed out the
cup, and turned on the burner under the coffeepot.
One more thing—one more development, born of suspicion.
For the original one-Watcher plan had been abandoned, of course.
And here, again, there was no telling whose blame it was. Quis
custodiet ipsos custodes? Who will watch the Watchers? There had
been many Watchers to a generation—how many, no one knew. They
balanced each other off, and they checked the random number of
Insurgents who awoke in each generation. So, more Insurgents
awoke to check the Watchers—and, more.
In spite of what the Transylvanians believed, a wolf is no match for a
man, except under special conditions. A tiger can pull a man down—
but cannot fire back at the hunters. A seal is prey to the Eskimo.
So, "werewolves." Child of fear, of Watcher propaganda, and of one-
tenth fact. The animals were Insurgent chrysalids, right enough. But,
for an awake Insurgent to compete with a Watcher, the Insurgent, too,
had to be a man—or something like it.
The coffee had warmed up. He poured himself a fresh cup, and
added cream and sugar absently. The refrigerator was empty. He
reached in and turned it off. No more need for that, after tonight.
So, that was the power the Insurgents had. The only power, and the
Watchers had it, as well. They could resolve their chrysalids into any
form they chose—realign. A wolf could become a man—without hair
on his palm, and with garlic on his breath, if he so chose. A man—a
Watcher, of course—could become a wolf.
Thus, the final development. Espionage and counter-espionage.
Infiltration. Spying, if you chose.
The Insurgent smiled bitterly, and drained the cup. And propaganda,
of course. Subtle—most of it indirect, a good deal of it developed by
the chrysalids themselves, but propaganda, nevertheless. Kill the evil
ones—kill the eaters of dead flesh, the drinkers of blood. They are the
servants of the Evil One.
He almost retched.
But, you could hardly blame them. It was a war, and, in a war, you
play all your cards, even if some of them were forced into your hand.
Yes, and I've played genuine werewolf on occasion, when I had to.
He started to wash the coffeepot and the cup—then, threw both into
the garbage can. He walked back to the radio and dialed it away from
Eroica and back to baseball. The Giants were losing, Three-Zero, in
the third inning.
The house phone buzzed. He went to it slowly, picked it calmly off the
hook.
"Yes, Artie?"
"Mister Disbrough, there's a couple of guys coming up to see you. I'm
not supposed to tell you about it, but.... Well, I figured ..." the
doorman said.
"All right. Thanks, Artie," he answered quietly. He almost hung up,
then thought of something. "Artie?"
"Yes, Mister Disbrough?"
"There'll be a couple of fifths of Dewar's in my cupboard. I won't be
back for a while. You and Pete are welcome to them. And thanks
again."
He hung up and began to dress, realigning his chrysalis to give him
the appearance of clothing. The doorbell rang, and he went to open it
for the two men from the FBI.

II
What difference did it make, what particular pen he represented?
Rather, since the sober-faced men knew very well which pen it was,
why should it be so necessary to them for him to confirm what they
already knew without a shadow of a doubt?
"Now, then, Mister Disbrough," one of the FBI men said, leaning his
hands on the edge of the table at which the Insurgent was sitting, "we
know who sent you."
Good. Why bother me, then?
"We know where you got your passport, we know who met you at the
dock, we know your contacts. We have photographs of everyone
you've met and talked to, we have tapes of every telephone call
you've made or received. We also know that you are the top man in
your organization here."
And? They were chrysalids, every one of them. Perhaps there was no
Watcher behind them—perhaps. But he'd been picked up a little too
quickly. The net had folded itself around him too soon. No—there had
to be a Watcher. He wished they'd stop this talking and bring him out.
"Now, I'd simply like to point out to you that this is an airtight case. No
lawyer in the world will be able to break it down. You'll retain counsel,
of course. But, I'd simply like to point out to you that there'll be no
point to any denial you may make to us. We know what you've been
doing. I'd suggest you save your defense for the trial."
He looked up at him and smiled ruefully. "If you've got a list of
charges," he said, "I'll be glad to confess to all of them—provided, of
course, that it is a complete list."
I'm sure it doesn't list me as a werewolf, he thought. I wonder what
the sentence would be—death by firing squad equipped with silver
bullets?
But, then, he wasn't going to confess to that, anyway.
"Um!" The FBI man looked suspicious. Obviously, he'd expected
nothing of the kind.
"No strings," the Insurgent reassured him. "The job's over, and it's
time to punch the clock."
Which was just about the way it was. But he wanted that Watcher. If
he was in the office at all, he'd almost have to come out to witness
the confession. After all, the Insurgent was supposed to be a pretty
big fish.
The FBI man went into a cubicle office set off to one side. When he
came out, carrying a sheaf of paper, the Watcher was with him.
The Insurgent felt the hackles standing up on the back of his neck,
and something rumbled inaudibly at the base of his throat. He knew.
He could tell. He could smell Watcher every step of the way, from the
day he had docked until now, when the scent—half there, half the
pure intuition of instinct—rose up before him in an over-powering
wave.
Then he saw the look of distaste crawl across the Watcher's face,
and he barked a laugh that drew curious looks from the men in the
office. Hello, brother.
He saw the bulge of the hip holster on the Watcher's belt, and
laughed again. So, we play the game, he thought. We add up scores,
and, in the end, the side with the most points wins. Forget that there
should be no sides, that every point, no matter for whom scored, is a
mark of shame and disgrace.
He wondered, briefly, whether the Watcher was of his kind by choice,
or whether it was simply something that had happened, as it was with
him. Probably. Two separate heritages had met, represented by
identical individuals who happened to have awakened in dissimilar
chrysalids.
Will we remember? he wondered. When we awaken, will we
remember this? How we battled, blinded, in the shadows of our own
casting? Or was there more mercy in Creation than they, themselves,
had shown to the chrysalids? He had three brothers among the
sleepers. When they woke, would they embrace, not remembering
that each had killed the other countless times? Or forgetting that they
had stood together, on some battlefield? Would all the old comrades,
all the bitter enemies, be wiped from memory? He hoped so. With
every segment of his being, he hoped so, for there was no peace,
through eternity, if it was otherwise.
He stood up, lightly, tensing the muscles in his calves. The FBI men,
suddenly alert, began to move for him, but he'd maneuvered things
so that none of them were close enough to him.
The Watcher went pale.
"Shall I coalesce, brother?" the Insurgent asked, the words rumbling
out of his throat, a grin of derision baring his teeth.
"No!" The Watcher was completely frightened. Words could be
explained away, particularly if they sounded like nonsense to the
other men in the room. But a werewolf, fanging the throat of a
Watcher who would have to fight back with his spectacular
weapons.... Nothing in the world could keep the rumors from
spreading. The chrysalids might even learn, finally and irrevocably,
the origin of their species.
"Your obligation, brother," the Insurgent half-laughed, and kept
stalking toward the Watcher. Perhaps he is my brother.
And if he is...?
No difference. The shadows are thick and very dark. One of the other
men shot him in the side, but he sprang for the Watcher, carefully
human, to hold the Watcher to his debt, and the Watcher shot him
three times in the chest, once in the throat, and once in the stomach.
The shape of a cross? Did he believe it himself? Was it true? A plus
sign, cancelling a negative force? Who knew? Shadow, shadow, all is
darkness.
He fell to his knees, coughing, in victory. Score one more for the
Insurgents, and a Watcher, at that!
"Thank you, brother," the Insurgent murmured, and fell into the long
sleep with a grateful sigh.
*** END OF THE PROJECT GUTENBERG EBOOK INFILTRATION
***

Updated editions will replace the previous one—the old editions will
be renamed.

Creating the works from print editions not protected by U.S.

copyright law means that no one owns a United States copyright in
these works, so the Foundation (and you!) can copy and distribute it
in the United States without permission and without paying copyright
royalties. Special rules, set forth in the General Terms of Use part of
this license, apply to copying and distributing Project Gutenberg™
electronic works to protect the PROJECT GUTENBERG™ concept
and trademark. Project Gutenberg is a registered trademark, and
may not be used if you charge for an eBook, except by following the
terms of the trademark license, including paying royalties for use of
the Project Gutenberg trademark. If you do not charge anything for
copies of this eBook, complying with the trademark license is very
easy. You may use this eBook for nearly any purpose such as
creation of derivative works, reports, performances and research.
Project Gutenberg eBooks may be modified and printed and given
away—you may do practically ANYTHING in the United States with
eBooks not protected by U.S. copyright law. Redistribution is subject
to the trademark license, especially commercial redistribution.

START: FULL LICENSE

THE FULL PROJECT GUTENBERG LICENSE
 PLEASE READ THIS BEFORE YOU DISTRIBUTE OR USE THIS WORK

To protect the Project Gutenberg™ mission of promoting the free

distribution of electronic works, by using or distributing this work (or
any other work associated in any way with the phrase “Project
Gutenberg”), you agree to comply with all the terms of the Full
Project Gutenberg™ License available with this file or online at
www.gutenberg.org/license.

Section 1. General Terms of Use and

Redistributing Project Gutenberg™
electronic works
1.A. By reading or using any part of this Project Gutenberg™
electronic work, you indicate that you have read, understand, agree
to and accept all the terms of this license and intellectual property
(trademark/copyright) agreement. If you do not agree to abide by all
the terms of this agreement, you must cease using and return or
destroy all copies of Project Gutenberg™ electronic works in your
possession. If you paid a fee for obtaining a copy of or access to a
Project Gutenberg™ electronic work and you do not agree to be
bound by the terms of this agreement, you may obtain a refund from
the person or entity to whom you paid the fee as set forth in
paragraph 1.E.8.

1.B. “Project Gutenberg” is a registered trademark. It may only be

used on or associated in any way with an electronic work by people
who agree to be bound by the terms of this agreement. There are a
few things that you can do with most Project Gutenberg™ electronic
works even without complying with the full terms of this agreement.
See paragraph 1.C below. There are a lot of things you can do with
Project Gutenberg™ electronic works if you follow the terms of this
agreement and help preserve free future access to Project
Gutenberg™ electronic works. See paragraph 1.E below.