(40, No 1) (Mathematical Surveys and Monographs, 40, No 1) Daniel Gorenstein, Richard Lyons, Ronald Solomon - The Classification of The Finite Simple Groups-Amer Mathematical Society (1994)

http://dx.doi.org/10.1090/surv/040.
Selected Titles in This Series

40.1 D a n i e l G o r e n s t e i n , Richard Lyons, a n d R o n a ld S o l o m o n , The classification of the
finite simple groups, number 1, 1994
39 Sigurdur H e l g a s o n , Geometric analysis on symmetric spaces, 1994
38 G u y D a v i d and S t e p h e n S e m m e s , Analysis of and on uniformly rectifiable sets, 1993
37 Leonard L e w i n, Editor, Structural properties of polylogarithms, 1991
36 J o h n B. Conway , The theory of subnormal operators, 1991
35 S h r e e r a m S. Abhyankar , Algebraic geometry for scientists and engineers, 1990
34 V i c t o r Isakov, Inverse source problems, 1990
33 Vladimir G. Berkovich, Spectral theory and analytic geometry over non-Archimedean
fields, 1990
32 Howar d J a c o b o w i t z , An introduction to CR structures, 1990
31 P a u l J. Sally, Jr. and D a v i d A . V o g a n , Jr., Editors, Representation theory and
harmonic analysis on semisimple Lie groups, 1989
30 T h o m a s W . Cusick a nd M a r y E . F l a h i v e , The MarkofT and Lagrange spectra, 1989
29 A l a n L. T . P a t e r s o n , Amenability, 1988
28 Richard B e a l s , P e r c y Deift, and Carlos T o m e i , Direct and inverse scattering on the
line, 1988
27 N a t h a n J. F i n e , Basic hypergeometric series and applications, 1988
26 Hari Bercovici, Operator theory and arithmetic in H°°, 1988
25 Jack K. H a l e , Asymptotic behavior of dissipative systems, 1988
24 Lance W . Small, Editor , Noetherian rings and their applications, 1987
23 E . H . R o t h e , Introduction to various aspects of degree theory in Banach spaces, 1986
22 Michael E. Taylor, Noncommutative harmonic analysis, 1986
21 A l b e r t B a e r n s t e i n, D a v i d Drasin , P e t e r D u r e n , and A l b e r t M a r d e n , Editors,
T he Bieberbach conjecture: Proceedings of the symposium on the occasion of the proof,
1986
20 K e n n e t h R. G o o d e a r l, Partially ordered abelian groups with interpolation, 1986
19 G r e g o ry V . C h u d n o v s k y , Contributions to the theory of transcendental numbers, 1984
18 Frank B . K n i g h t , Essentials of Brownian motion and diffusion, 1981
17 Le B a r o n O. Ferguson, Approximation by polynomials with integral coefficients, 1980
16 O. T i m o t h y O'Meara, Symplectic groups, 1978
15 J . D i e s t el a n d J . J . U h l , J r . , Vector measures, 1977
14 V . G u i l l e m i n a nd S. Sternberg, Geometric asymptotics, 1977
13 C. Pearcy , Editor , Topics in operator theory, 1974
12 J. R. Isbell, Uniform spaces, 1964
11 J. Cronin, Fixed points and topological degree in nonlinear analysis, 1964
10 R. A y o u b , An introduction to the analytic theory of numbers, 1963
9 A r t h u r Sard, Linear approximation, 1963
8 J. Lehner, Discontinuous groups and automorphic functions, 1964
7.2 A . H . Clifford and G. B . P r e s t o n , The algebraic theory of semigroups, Volume II, 1961
7.1 A . H. Clifford an d G. B . P r e s t o n , The algebraic theory of semigroups, Volume I, 1961
6 C. C. Chevalley , Introduction to the theory of algebraic functions of one variable, 1951
This page intentionally left blank
The Classification
of the Finite
Simple Groups
Mathematical
Surveys
and
Monographs
Volume 40.1
The Classification
of the Finite
Simple Groups
Daniel Gorenstein
Richard Lyons
Ronald Solomon
jWrMAr
A m e r i c a n M a t h e m a t i c a l Society
Providence, Rhode Island
T h e a u t h o r s were s u p p o r t e d in p a r t by N S F g r a n t # D M S 89-03124, by D I M A C S , a n
N S F Science a n d Technology C e n t er funded u n d e r c o n t r a c t STC-88-09648, a n d by N S A
grant #MDA-904-91-H-0043.
2000 Mathematics Subject Classification. P r i m a r y 20D05, 20E32.
ABSTRACT. This is the first monograph in a series devoted to a revised proof of the classification
of the finite simple groups.
Library of Congress C a t a l o g i n g - i n - P u b l i c a t i on D a t a
Gorenstein, Daniel.
The classification of the finite simple groups / Daniel Gorenstein, Richard Lyons, Ronald
Solomon.
p. cm. — (Mathematical surveys and monographs, v. 40, number 1-)
Includes bibliographical references and index.
ISBN 0-8218-0334-4
1. Finite simple groups. I. Lyons, Richard, 1945- . II. Solomon, Ronald. III. Title. IV.
Series: Mathematical surveys and monographs; no. 40, pt. 1-.
QA177.G67 1994
512 / .2—dc20 94-23001
CIP
C o p y i n g an d reprinting. Individual readers of this publication, and nonprofit libraries

acting for them, are permitted to make fair use of the material, such as to copy a chapter for use
in teaching or research. Permission is granted to quote brief passages from this publication in
reviews, provided the customary acknowledgment of the source is given.
Republication, systematic copying, or multiple reproduction of any material in this publication
is permitted only under license from the American Mathematical Society. Requests for such
permission should be addressed to the Assistant to the Publisher, American Mathematical Society,
P. O. Box 6248, Providence, Rhode Island 02940-6248. Requests can also be made by e-mail to
reprint-permissionQams.org.
© 1994 by the American Mathematical Society. All rights reserved.

Reprinted with corrections, 2000.
The American Mathematical Society retains all rights
except those granted to the United States Government.
Printed in the United States of America.
@ The paper used in this book is acid-free and falls within the guidelines
established to ensure permanence and durability.
Visit the AMS home page at URL: http://www.ams.org/
10 9 8 7 6 5 4 3 2 05 04 03 02 01 00
To the memory of
Pearl Solomon (1918-1978)
and
Marvin Lyons (1903-1975)
Contents
Preface xiii
Preface to the Second Printing xv
Part I, Chapter 1: Overview 3
Introduction to the Series 3
A. The Finite Simple Groups 6
1. Simple groups 6
2. DC-groups 12
B. The Structure of Finite Groups 12
3. The Jordan-Holder theorem and simple groups 12
4. The generalized Fitting subgroup and quasisimple groups 15
5. p'-cores and p-components 19
6. The embedding of p'-cores and p-components 20
7. Terminal and p-terminal p-components 22
8. ;>-constrained and p-solvable groups 24
C. Classifying Simple Groups 27
9. Internal analysis: targeting local structure 27
10. Internal analysis: passing from global to local information 29
11. Identifying simple groups 31
12. A capsule summary of this series 35
13. The existing classification proof 38
14. Simplifying the classification theorem 41
D. The Background Results 44
15. Foundational material 44
16. The initial assumptions 46
17. Background Results: basic material 47
18. Background Results: revised portions of the proof and
selected papers 48
E. Sketch of the Simplified Proof 51
19. Centralizers of semisimple elements 51
20. Uniqueness subgroups 52
21. The sets £JP(G) and groups of even type 53
22. Generic simple groups and neighborhoods 55
23. The main case division 58
24. Special simple groups 59
25. Stages of the proof 61
26. Generic simple groups 63
27. The identification of G 68
x CONTENTS
F. Additional Comments 72
28. The length of the proof 72
29. The X-group environment 75
30. The term G « G* 76
31. Reading this series 77
Part I, Chapter 2: Outline of Proof 79
Introduction 79
A. The Grids 80
1. Some basic terminology 80
2. The uniqueness grid 83
3. The classification grid 83
B. The Uniqueness Grid 87
4. 2-uniqueness subgroups 87
5. Groups of restricted even type with |M(5)| = 1 90
6. Component preuniqueness subgroups 90
7. The odd uniqueness theorem 92
8. Some technical definitions 93
9. Groups with a strongly embedded subgroup 95
10. Theorem M(5) 96
11. Theorem U(a) 98
C. The Classification Grid: Generic and Special Simple Groups 99
12. e p -groups 99
13. Sp-groups and T p -groups 102
14. Groups of special odd type 103
15. Groups of special even type 105
16. Groups of generic type 106
D. The Classification Grid: The Stages of the Proof 106
17. Theorem Ci 106
18. p-terminal p-components 108
19. Centralizer of involution patterns 109
20. Theorem C2 110
21. Theorem C3 114
22. Theorem C4 114
23. Theorem C5 116
24. Theorem C6 118
25. Vertical neighborhoods 118
26. Theorem C7 120
E. Principal Techniques of the Proof 122
27. Fusion 122
28. The Bender method 123
29. Signalizer functors and ^-balanced groups 124
30. Lv>-balance, the ^-property, and pumpups 127
31. The signalizer functor method 128
32. Near components, pushing up, and failure of factorization 129
33. The amalgam method 131
34. Local analysis at two primes 134
35. Character theory and group order formulas 135
36. Identification of the groups of Lie type 137
CONTENTS xi
37. Properties of DC-groups 138

F. Notational Conventions 139
Background References 140
Expository References 141
Glossary 148
Index 155
Preface
Though elated at the successful completion of the classification of finite simple

groups in the early 1980's, Danny Gorenstein nevertheless immediately appreciated
the urgency of a "revision" project, to which he turned without delay. Before long
we had joined him in this effort, which is now into its second decade. Danny always
kept the project driving forward with relentless energy, contagious optimism and his
unique global vision of finite simple group theory. He inspired other collaborators—
Richard Foote and Gemot Stroth—to contribute theorems designed to fit our re-
vised strategy. Their work forms a vital part of this project. The conception of
these volumes is unmistakably Danny's, and those who know his mathematics and
his persuasive way of explaining it should recognize them everywhere. Since his
death in 1992, we have tried to maintain his standards and we hope that what-
ever changes and additions we have made keep the vigorous spirit which was his
trademark. Of course the responsibility for any stumbling or errors must remain
with us. To accompany him on this mathematical journey was a privilege and as
we continue, our debt to our teacher, colleague and loyal friend is hard to measure.
Thanks, Danny.
This monograph is Number 1 of a projected dozen or so volumes, and contains
two of the roughly thirty chapters which will comprise the entire project. Not all
of the chapters are completely written at this juncture, but we anticipate that the
publication process, now begun, will continue at a steady pace. In the first section
of Chapter 1, we discuss the current status of the work in some more detail; later
in that chapter we delineate the background results which form the foundation for
all the mathematics in the subsequent volumes.
When Danny began, there was already a well-established tradition of "revision-
ism" in finite group theory. Indeed, beginning in the late 1960's, Helmut Bender
produced a series of "revisions" whose beauty and depth profoundly influenced fi-
nite group theory. During the final decade of the classification proof, several more
group-theorists showed how to develop deep mathematics while re-addressing some
of the fundamental theorems in the theory of simple groups. While it is some-
times difficult to draw a distinct line between revisionist and other mathematics,
the clear successes of Bender, Michel Enguehard, George Glauberman, Koichiro
Harada, Thomas Peterfalvi, and Bernd Stellmacher in various revisionist projects
have inspired us.
We are extremely grateful to Michael Aschbacher, Walter Feit, George Glauber-
man, Gary Seitz, Stephen Smith, and John Thompson for their enthusiastic sup-
port and interest in our endeavor, and specifically for having read some of the
manuscripts and made valuable comments. To our collaborators Richard Foote
and Gemot Stroth, our special thanks. The ideas and comments of many col-
leagues have been most useful and will appear in some form in the volumes to
xiii
xiv THE CLASSIFICATION OF THE FINITE SIMPLE GROUPS
come. Indeed, this work has ingredients of several types: there is new mathe-
matics; there is exposition of unpublished work of our colleagues, ranging from
short arguments to entire case analyses which they have generously shared with
us, and which we shall acknowledge specifically as we go along; and finally there
are reworkings of published papers. For now, in addition to all the people already
mentioned, we must also thank Jonathan Alperin, Michel Broue, Andrew Cher-
mak, Michael Collins, K. M. Das, Alberto Delgado, Paul Fong, Robert Gilman,
David Goldschmidt, Kensaku Gomi, Robert Griess, Robert Guralnick, Jonathan
Hall, William Kantor, Martin Liebeck, Ulrich Meierfrankenfeld, Michael O'Nan,
Lluis Puig, Geoffrey Robinson, Jan Saxl, Ernie Shult, Franz Timmesfeld, Jacques
Tits, John Walter, Richard Weiss, and Sia K. Wong.
We are happy to acknowledge the financial support of the National Science
Foundation and the National Security Agency, and the support of DIM ACS and
the Institute for Advanced Study. We extend our sincere thanks to our skillful and
patient typists Adelaide Boulle, Lisa Magretto, Ellen Scott, Pat Toci, and Dorothy
Westgate.
Finally, to our wives Lisa and Myriam, to our families, and to Helen Gorenstein,
we gratefully acknowledge the love and commitment by which you have endured
the stresses of this long project with good humor, grace and understanding.
Richard Lyons and Ronald Solomon

June, 1994
Preface to the Second Printing
Six years have passed since the first printing of this book and sixteen since
the detailed classification strategy presented in Chapter 2 was first adumbrated.
At present the first four books in our series have been published and the fifth and
sixth are well on their way to completion. Since the first printing, some impor-
tant second-generation pieces of the classification of the finite simple groups have
also appeared, in particular two volumes [BG1,P4] presenting the revised proofs
by Bender, Glauberman and Peterfalvi of the Odd Order Theorem of Feit and
Thompson. We have taken the Feit-Thompson Theorem as a Background Result,
but now could therefore remove [FTl] from the list of Background Results (see
pp. 48-49) and replace it with Peterfalvi's revision of Chapter V of that paper
[P4]. In addition, Peterfalvi's treatment of the Bender-Suzuki Theorem (Part II of
[PI]) has now appeared in an English version[P5]. We have taken the opportunity
of this second printing to add these new references, and to correct some misprints
and other minor slips in the first printing.
We are nearing a milestone in our series - the completion of our revised proof
of the Classification Theorem for DC-proper simple groups of odd type following the
strategy carefully outlined in this volume, and in particular in our Theorems 62 ?
C3 and C7.
During these same years there has been considerable research activity related to
simple groups of even type from the so-called "unipotent" point of view, that is, by
means of the analysis of the structure of 2-constrained 2-local subgroups and their
interplay with one another. With the maturing of the theory of amalgams, such
notable results have emerged as Stellmacher's second-generation proofs of the clas-
sifications of iV-groups and thin groups, first proved by Thompson and Aschbacher,
respectively. Furthermore an ambitious program proposed by Meierfrankenfeld is
currently being pursued by Meierfrankenfeld, Stellmacher, Stroth, Chermak and
others; it is directed towards the classification of groups of characteristic p-type for
an arbitrary prime p, excepting those satisfying uniqueness conditions akin to the
existence of a strongly p-embedded subgroup. This promises to remain an area of
intense activity for another decade and one can at best speculate from this distance
concerning what major theorems will emerge once the dust settles. Nevertheless,
since it is the amalgam approach which we have proposed for our Theorem 64,
recent results have naturally had an impact on the appropriateness of the strategy
presented in this volume.
In the seminal work of Thompson on TV-groups, the parameter e(G) was in-
troduced to denote the maximum value of mp(H) as p ranges over all odd primes
and H ranges over all 2-local subgroups of G. A crucial division in the set of all
groups of characteristic 2-type was made between those groups G with e(G) < 2
XV
xvi THE CLASSIFICATION OF THE FINITE SIMPLE GROUPS
and those with e(G) > 3, the former being dubbed quasithin1 groups. This sub-
division persisted in the full classification of finite simple groups of characteristic
2-type (simple groups G of 2-rank at least 3 such that F*(H) = 02(H) for every
2-local subgroup H of G) undertaken during the 1970's. It was however discovered
by Gorenstein and Lyons in the work leading up to their Memoirs volume [GL1]
that the case e(G) = 3 presented unique difficulties deserving special treatment.
Hence [GL1] treats simple groups of characteristic 2-type with e(G) > 4, and the
case e(G) = 3 was handled independently by Aschbacher in [A13]. Both treatments
took a "semisimple" rather than a "unipotent" approach, studying p-signalizers and
p-components of p-local subgroups for primes p other than the "characteristic", in
these cases for p > 2.
The Quasithin Case (e(G) < 2) was subdivided further into the case e(G) =
1 (the "Thin Case") and the case e(G) = 2. The former was treated by As-
chbacher [A 10], while the latter was undertaken and almost completed by G. Ma-
son. A unified treatment of the Quasithin Case is nearing completion by Aschbacher
and S. D. Smith and will be published in this AMS series. When published, the
Aschbacher-Smith volumes will represent a major milestone: the completion of the
first published proof of the Classification of the Finite Simple Groups. But it is also
pertinent to our strategy, since Aschbacher and Smith have relaxed the hypotheses
somewhat. Rather than restricting themselves to the set of all simple groups of
characteristic 2-type, they have partly accommodated their work to the strategic
plan outlined in this volume by having their theorem encompass the larger class
of all simple groups of even type. Indeed their main theorems have the following
immediate consequence, in our terminology.
T H E e(G) < 2 T H E O R E M . Let G be a finite X-proper simple group of even
type. Suppose that every 2-local subgroup H of G has p-rank at most 2 for every
odd prime p. Then one of the following conclusions holds:
(a) G G £hev(2), and G is of twisted Lie rank at most 2> but G is not isomorphic
to U5 (q) for any q > 4; or
(b) G 2* L 4 (2), L 5 (2), Spe(2), A 9 , L 4 (3), U4(3), G 2 (3), M12, M22, M23, M 2 4 ,
J\> J2, J$, J±i HS, He, or Ru.
However it must be noted that the Aschbacher-Smith Theorem will not com-
plete the classification of "quasithin" groups of even type as this term is defined
in the current volume (p. 82), the reason being that during the 1980's we chose to
redefine "quasithin" to include the e(G) = 3 case for groups of even type. This was
motivated both by the peculiar difficulties of the e(G) — 3 case noted above and by
conversations concerning the Amalgam Method which suggested to us that treating
groups with e(G) < 3 via the Amalgam Method would not be substantially more
difficult than treating groups with e(G) < 2.
In any event the recent work of Aschbacher and Smith has now reopened the
possibility of restoring "quasithin" more or less to its original meaning, or more
precisely to the hypotheses of the e(G) < 2 Theorem above. The groups of even
type with e(G) = 3 then could be treated by a "semisimple" strategy akin to the
one for the proof of Theorem CQ discussed briefly in Section 24 of Chapter 2.
Such a change in strategy would entail a different case division defining rows
4, 5 and 6 of the classification grid (p. 85), i.e., altered hypotheses for Theorems
lr
The word "quasithin" has been used in a number of distinct senses by various authors. In
this paragraph and the next, we stick to the original meaning e(G) < 2.
PREFACE TO THE SECOND PRINTING xvii
C4, C5 and C6 (pp. 105-106). Namely the e(G) < 2 Theorem would take the place
of Theorem 64, and in rough terms the dichotomy mp(M) > 4 or mp{M) < 3 for
certain 2-local subgroups would be replaced throughout by the dichotomy mp(M) >
3 or rrip(M) < 2. Consequently there would be changes in the sets X4, DC5 and
%Q of target groups, the first one shrinking and the other two expanding (and all
would be nonempty). In addition the uniqueness theorems M(5) and U(a) would
have to be strengthened to accommodate weakened hypotheses on the p-ranks of
2-local subgroups of G. The authors are now contemplating and investigating the
possibility of adopting this different approach.
Richard Lyons and Ronald Solomon

August, 2000
Part I
PRELIMINARIES
http://dx.doi.org/10.1090/surv/040.1/01
PART I, CHAPTER 1
OVERVIEW
INTRODUCTION TO THE SERIES
The existing proof of the classification of the finite simple groups runs to some-
where between 10,000 and 15,000 journal pages, spread across some 500 separate
articles by more than 100 mathematicians, almost all written between 1950 and
the early 1980's. Moreover, it was not until the 1970's that a global strategy was
developed for attacking the complete classification problem. In addition, new sim-
ple groups were being discovered throughout the entire period—in succession the
Chevalley groups, the Steinberg variations, the Suzuki and Ree groups and the
twenty-one "modern" sporadic groups—so that it was not even possible to state
the full theorem in precise form before the constructions of the last two sporadic
groups F\ and J4 in the early 1980's. Then, too, new techniques for studying sim-
ple groups were steadily being developed; but earlier papers, essential to the overall
proof, were of necessity written without benefit of many of these later methods.
Under such circumstances it is not surprising that the existing proof has an
organizational structure that is rather inefficient and that its evolution was some-
what haphazard, including some duplications and false starts. As a result of these
various factors, it is extremely difficult for even the most diligent mathematician,
not already versed in its intricacies, to obtain a comprehensive picture of the proof
by examining the existing literature.
Considering the significance of the classification theorem, we believe that the
present state of affairs provides compelling reasons for seeking a simpler proof, more
coherent and accessible, and with clear foundations. This series of monographs has
as its purpose an essentially self-contained presentation of the bulk of the classifi-
cation of the finite simple groups-primarily of that portion of the total proof that
can be achieved by what is known as local group-theoretic analysis. The arguments
we give will depend only on existing techniques and the complete project will cover
between 3,000 and 4,000 pages.
Our major simplification is obtained as a result of a global strategy that differs
in several key respects from that of the existing proof. This strategy enables us, on
the one hand, to bypass completely a number of general and special classification
theorems that were integral to the present proof and, on the other, to treat in a
uniform way certain portions of the proof that were originally handled by separate
arguments. A second simplification occurs because we assume in effect at the very
outset that the simple group G under investigation is a minimal counterexample to
the classification theorem, which implies that all proper simple sections of G are of
known type. In the earlier period, when there was a prevalent belief that additional
simple groups remained to be discovered, such an initial hypothesis would have been
entirely inappropriate.
3
4 PART I, CHAPTER 1: OVERVIEW
Finally we note that we are intentionally adopting a fuller style of exposition

than would be appropriate for a journal article, for our primary aim is to provide
as readable a proof as we are able to attain. Some reduction in our projected
3,000-4,000 pages could certainly have been made (perhaps as much as 20%) had
we chosen a more compact style; and the argument would still have been viewed
as self-contained. On the other hand, it is our belief that it will not be possible to
construct a substantially shorter classification of the finite simple groups without
first developing some radically new methods.
This first monograph is preliminary, presenting the conceptual framework un-

derlying the simplified proof of the classification of the finite simple groups that
will appear in the series. Chapter 1 is a relatively non-technical overview. After the
statement of the Classification Theorem, it gives a description of important general
aspects of the structure of finite groups particularly relevant to the classification
proof. (Though some of these properties have been proved only assuming that all
simple sections of the groups in question are of known type, they become properties
of all finite groups after the proof of the classification theorem.) This is followed
by some general comments about classifying simple groups and a short discussion
and critique of the existing classification proof. The Background Results which
we permit ourselves to assume—including some portions of the classification proof
itself—are then listed. The chapter concludes with a brief sketch of the proof to be
presented in this series and some miscellaneous remarks.
In Chapter 2 we provide a more detailed outline of the proof of the Classification
Theorem. The summary includes definitions of the key terms in which the proof
is expressed, a description of the major case divisions, and a diagrammatic chart
of the various stages of the analysis for each case. Finally, we briefly discuss the
principal group-theoretic techniques underlying the proof.
This series of monographs is divided into five major parts:
Part I Preliminaries
Part II Uniqueness theorems
Part III Generic simple groups
Part IV Special odd simple groups
Part V Special even simple groups.
Moreover, each part is itself subdivided into a number of chapters and its full
exposition will require several monographs.
In particular, Part I consists of four chapters. The two chapters of the current
monograph will be designated [Ii] (this Overview) and [I2] (the Outline of Proof),
respectively. Chapter 3 of Part I, designated [I^], will cover essentially all the
general (primarily local) group-theoretic results needed for the classification proof,
while Chapter 4, designated [1^], will cover basic properties of almost simple DC-
groups 1 , which underlie the more detailed results about OC-groups needed for the
analysis of Parts II, III, IV, and V. Both of these chapters will of course rely heavily
on the Background Results named in the Overview.
1
A DC-group is a group such that all the composition factors of all its subgroups are known
simple groups—i.e., simple groups in the conclusion of t h e Classification Theorem. See Section 2
of Chapter 1.
INTRODUCTION TO THE SERIES 5
The proof of the Classification Theorem proceeds by induction, focusing on a

minimal counterexample G. All proper subgroups of G are then DC-groups, and the
proof requires a very elaborate theory of DC-groups. Each of Parts II, III, IV, and V
will contain initial chapters establishing the X-group results needed specifically for
that part. To enhance the organizational structure each part will have two chapters
consisting of
1. Principal properties of DC-groups
2. Specialized properties of DC-groups
to be designated [Up], [Us], [Hip], etc. By "principal properties" we mean those
more central conceptually to that part.
Some general group-theoretic results will be omitted from [IQ] and deferred to
the parts of the proof in which they are used. This material will be included in a
preliminary chapter of the respective part designated [IIG], [HIG], e t c - Similarly
there will be chapters [IIA]? etc.
Our intention is to make explicit the statements and proofs of all group-
theoretic and DC-group results needed for the proof of the classification theorem,
using only material from the explicitly listed Background Results as references.
As has been indicated, the proof of the Classification Theorem will run to
several thousand pages. Because of this, the principal authors have been able to
induce some of our colleagues with expertise in specific areas to work on certain
major portions of the proof, and we gratefully acknowledge their collaboration in the
project at the outset. The topics covered by their contributions are the following.
The terms involved and the meaning of the results are discussed more fully in the
two chapters to follow.
A. In Part II, the classification of groups of even type in which |JVC(G; S)\ = 1
is joint work with Richard Foote, who wrote Chapter 4 of Part II.
B. In Part II, the nonexistence of groups of even type that contain a p-unique-
ness subgroup for each p E c(G) is established in Chapter 6 of Part II,
written by Gemot Stroth.
It is appropriate here to comment on the status, at the moment of this writing,

of the mathematics and the manuscripts for this lengthy project. There is one major
case, whose proof is visualized by the amalgam method, which has not yet been
fully analyzed—the quasithin type case. We should perhaps call it the "revised"
quasithin case because it is not the same as the quasithin case (e(G) < 2, G of
characteristic 2 type) of the original classification proof, but is defined by a different
condition (cr(G) = 0, G of even type) to be made precise below, and it includes in
particular the case e(G) < 3, G of characteristic 2 type. The envisioned analysis
of this case is rather complex, but the results obtained so far, including a global
C(G, 5)-theorem to appear in Part II, and amalgam-theoretic results of Stellmacher
and Delgado, make us reasonably confident that it will be completed along the lines
described below in Chapter 2 (in row 5 of the classification grid and in section 22).
Nevertheless, for the present the reader should regard the strategy described there
as provisional. It would be possible, if absolutely necessary, to assemble parts of
the original classification proof—including the "original" quasithin case—to analyze
our "revised" quasithin case, but the newer amalgam method is appropriate and
extremely well-suited to the problem. We note that because the theory of amalgams,
or "weak (£,iV)-pairs", is presently an active field of general interest, it is likely
that a substantial portion of the work on the quasithin case will be published outside
of this series; if so, it will simply be added to the Background Results.
Turning to the project as a whole, many of the chapters exist now in final
drafts, but not all; some exist only in earlier drafts. Of course, if changes turn out
to be necessary upon final revision, the effects upon the global strategy and the
grids of Chapter 2 will be duly noted. Nevertheless, we are ready to present the
whole strategy at this time, confident that it will stand essentially as is.
A. T H E FINITE SIMPLE G R O U PS
1. Simple groups
It is our purpose in these monographs to prove the following theorem:
CLASSIFICATION T H E O R E M . Every finite simple group is cyclic of prime order,
an alternating group, a finite simple group of Lie type, or one of the twenty-six
sporadic finite simple groups.
A precise statement of the Background Results which we allow ourselves to
assume in the course of the proof will be given in sections 15-18. This material
includes some particular cases of the Classification Theorem as well as a consider-
able amount of general finite group theory and the theory of the particular groups
in the conclusion of the theorem. In addition the current status of the proof of the
theorem is discussed in the preceding Introduction to the Series.
We shall presently discuss the structure of arbitrary finite groups in the context
of the Classification Theorem, but first we shall give a brief introductory sketch of
the principal characters of our story, the finite simple groups. A more detailed
description of these groups without proofs is given in [G3, Chapter 2].
The most familiar examples of finite simple groups are the cyclic groups Zp of
prime order and the alternating groups An consisting of all the even permutations
of the symmetric groups E n , n > 5. However, the bulk of the set of finite simple
groups consists of finite analogues of Lie groups, including analogues of real forms.
These are called the finite simple groups of Lie t y p e , and naturally form 16 infinite
families. In addition, there exist precisely twenty-six sporadic finite simple groups
that are not members of any of these or any reasonably defined infinite families of
simple groups.
The oldest of the finite groups of Lie type are the classical groups: linear,
symplectic, orthogonal, and unitary [Dil, D l , Ar2, A l , KlLil, C h i ] . The
general linear group GLn(q) is the multiplicative group of all n x n nonsingular
matrices over the finite field Fq with q elements, q a prime power. The special lin-
ear group SLn(q) is its normal subgroup consisting of all matrices of determinant
1, and the projective special linear group Ln(q) = PSLn{q) is the quotient
group of SLn(q) by its central subgroup of scalar matrices. If n > 2, then the
1. SIMPLE GROUPS 7
group PSLn(q) is simple except for the two cases n = 2, q = 2, and n = 2, g = 3,

and SLn(q) is the commutator subgroup of GLn(q) except for the case n = 2, q = 2.
The symplectic and orthogonal groups can be defined as subgroups of GLn(q)
consisting of the matrices leaving invariant a given nondegenerate alternating bi-
linear form or quadratic form, respectively, on the underlying n-dimensional vector
space over Fq on which GLn{q) naturally acts. By passage to the commutator sub-
group and the quotient by scalar matrices, they lead to further families PSpn(q)
and Pft^(q) of finite simple groups. We note that over the complex numbers all
nondegenerate quadratic forms in a given dimension are equivalent, but over arbi-
trary fields, this is far from the case; for example, in any even dimension n over Fq,
there are exactly two such inequivalent forms, so we obtain two families of finite
orthogonal groups in this way. (When n is odd only one group arises, PQn(q).)
The complex unitary group GUn can be defined as the subgroup of the complex
general linear group GLn consisting of all matrices fixed by the automorphism a of
GLn defined by
a(X) = ((XY)-\
where X denotes the matrix obtained from X G GLn by replacing each entry by
its complex conjugate and X1 denotes the transpose of the matrix X. The finite
unitary groups GUn(q) are obtained in the same way from GLn{q2) as the fixed
points of the corresponding automorphism a, where now X denotes the matrix
obtained from X e GLn{q2) by raising each of its entries to the qth power. Again
this leads to the family of simple groups Un(q) = PSUn(q) via the special unitary
group SUn(q) = GUn(q) n SLn(q2).
Some of these finite classical groups, namely the linear groups, the orthogonal
groups P^2n+i(tf) 5 the symplectic groups and the orthogonal groups P^2n(g), are
analogues of the complex Lie groups An, Bn, Cn and Dn. Although finite analogues
of some of the exceptional Lie groups had been constructed by Dickson by the
early part of the century, it was not until 1955 that Chevalley [Ch2] showed by
a general Lie-theoretic argument that there exist finite analogues C(q) of every
semisimple complex Lie group C for every finite field Fq. In particular, he proved
simultaneously the existence of the five families G^tf)? E^q), Ee(q), E^{q) and
Eg(q) of exceptional simple groups of Lie type. At approximately the same time,
Tits was giving geometric constructions of several of these families (see [Til]).
These groups C(q), C = An, Bn, C n , D n , £?2, -^4, EQ, E? and Eg^ are called
the untwisted groups of Lie type. They are also called the Chevalley groups,
although this term sometimes is applied to all groups of Lie type.
The general construction of the twisted groups followed soon after. The
complex unitary groups are often referred to as compact real forms of the linear
groups. It was known also that real forms of the orthogonal groups P£l2n = Dn
and EQ exist, connected with a "graph" automorphisms r of period 2, induced from
a symmetry of the associated Dynkin diagram of type Dn and EQ^ respectively. In
1959 Steinberg [St3] showed that Chevalley's construction of the finite analogues
of complex Lie groups could be extended to give analogues of their real forms, thus
obtaining systematically the families 2Dn{q) (the second family of even-dimensional
orthogonal groups) and 2EQ(q) over Fq for every g, as well, of course, as the unitary
groups, which in the Lie notation are the twisted linear groups 2An_i(q).
The classical orthogonal group D4 is the only Lie group having a graph auto-
morphism a of period 3. However, as the complex field possesses no automorphism
TABLE I — T H E FINITE SIMPLE G R O U P S
Group ( n o t e s ) Other names Order
4 , n >5 A.ltn \n\
PSLn+1(q) = Ln+1(q) n+1

An{q), n > 1 W
= L++1(q) = AZ(q)
2 PSUn+1(q) = Un+1(q)
An(q), n >2 «
= L-+1(q) = A-(q)
i=2
#„(<?), n > 2 (2) P&2n+\{q) = ^2n+l(q)
2
252(9) (D.(3) Sz{q) = B2(^q) ? 2 ( ? - i ) ( « 2 + i)
C»(«), n > 2 W PSp2n{q)
£>«(«), n > 3 Pntn(q) = D+{q)
2
L>„(g), n > 2 P^n(q) = D-(q) T3 _l_ y9 n(„-l) (g „ + 1 ) j j ( g 2 i _ 1 )
t=l
12
(g -i)(? + g + i)(96-i)
2 8 4
3
A(g) ?
G2(q) (2) g V-i)(?6-i)
2
G 2 (g) W.W R(q) = 2G2(VQ) <73(<z-i)(<?3 + i)
g2V-i)(96-i)(98-i)(912-i)
2
F4(q) (2).(3) 2
^4(V9) q12(q-l)(q3 + l)(qA-l)(q6 + l)
36 2 5 6
(3,^ I)9 (? -l)(9 -l)(? -l)-
£6(<z) E£(q) (g8-l)(g9-l)(g12-l)
36
2 (3.^ I j 9 ( ? 2 - 1 ) ( ? 5 + 1)(96-1)-
£6(<?) Ee(q) («« - l ) ( g 9 + !)(<,« - 1 )
63
(2.^ T y g (?2-1)(96-1)-
£7(9) («* - 1)(9 10 - 1)(9 12 - 1)"
(g14-l)(g18-l)
q120(q2 ~ l)(q8 ~ 1)(<Z12 - 1)(<Z14 - 1)-
Es(q) ( g i8_ 1 ) ( ? 20_ 1 ) ( g 24_ 1 ) ( 9 30_ 1 )
Continued
1. SIMPLE GROUPS 9
TABLE I — CONTINUED
Group Other names Order
Mu 2 4 • 3 2 • 5 • 11
M12 2 6 • 3 3 • 5 • 11
M22 2 7 • 3 2 • 5 • 7 • 11
M23 27-32-5-7-ll-23
M24 2 1 0 • 3 3 • 5 • 7 • 11 • 23
Jl
2 3 - 3 - 5 - 7 - 1 1 - 19
h HJ 27 • 3 3 • 52 • 7
h HJM 2 7 • 3 5 • 5 • 17 • 19
Ji 221.33-5-7.113-23-29-31-37-43
HS 29-32-53.7-ll
He HHM = F7 2io . 3 3 . 52 . 7 3 . 17
Mc 27.36-53-7-ll
Suz Sz 213.37.52.7.11-13
Ly LyS 2 8 • 3 7 • 5 6 • 7 • 11 • 31 • 37 • 67
Ru 2 1 4 • 3 3 • 5 3 • 7 • 13 • 29
O'N O'S 2 9 • 3 4 • 5 • 7 3 • 11 • 19 • 31
Cox •1 2 2i . 3 9 • 5 4 • 7 2 • 11 • 13 • 23
C02 •2 2 1 8 • 3 6 • 5 3 • 7 • 11 • 23
Co3 •3 2 1 0 • 3 7 • 5 3 • 7 • 11 • 23
Fi22 M(22 ) 2 1 7 • 3 9 • 5 2 • 7 • 11 • 13
Fi2s M(23) 2is . 313 . 52 . 7 . n . 13 . i 7 . 23
M(24)' 2 2i . 3 i 6 . 52 . 7 3 . n . 13 . 17 . 2 3 . 29
Fi'24
F5 HN 2 14 . 36 . 56 . 7 . n . ig
F3 Th 2i5 . 310 . 53 . 72 . 13 . 19 . 31
F2 B = BM 2 4i . 313 . 56 . 72 . n . 13 . 17 . 19 . 2 3 . 3 1 . 47
2 46 . 320 . 59 . 76 . n 2 . 1 3 3 . 1 7 . 19 . 23 • 29 • 31 •41-
M
Fl 47 • 59 • 71
Notes: (1) Ai(2), Ai(3), 2 A 2 ( 2 ), and 2B2(2) are solvable.

(2) For G = B2{2) = C 2 (2), G = G 2 (2), G = 2 G 2 ( 3 ), and G = 2 F 4 ( 2 ) , G' = [G,G] is simple and
its index in G is q = 2, 2, 3, 2, resp.
(3) q = 2 2 n + 1 only (4) q = 3 2 n + 1 only
TABLE II
ISOMORPHISMS A M O N G T H E G R O U P S IN TABLE I
B2(q) = c2 (g) 1
DM s Aa(«)
2 2
I>3(5) = A3(q)
2
D2(q) <* Ai(q2)
m
Bn(2m) * n C (2 ) J
2
A5 = ^ i ( 4 ) £ ^ ( 5 ) * £>2 (2), of order 60
^ i ( 7 ) = ^ 2 (2), of order 168
Ai(8)'- S 2 G 2 (3)' , of order 504
A 6 S Ax (9) £ B 2 (2)' = C 2 (2)', of order 360
2
A 2 (3) ^ G 2 (2)' of order 6048
As = A3(2), of order 20160 |
2 2
\ A3(2) <* D3(2) = B 2 (3) = C 2 (3), of order 25920
of period 3, it is not possible to use this automorphism to construct a twisted form

of D4. On the other hand, in the finite case, one can define an automorphism a of
Dîq3) by the equation a(X) = (X)a, where now X denotes the matrix obtained
from X G D±(q3) by raising each of its entries to the qth power. Again the fixed
elements of a form a simple group, 3DA_(q), known as "triality D4". Steinberg's
construction covered this case as well. The groups 2An(q), 2 £> n (g), 3D±{q), and
2
Ee(q) are sometimes called the Steinberg variations.
Although Suzuki's discovery [Su2, Su3] of his family of simple groups 5z(2 n ),
n odd, n > 1, was not made Lie-theoretically, but rather in the course of solving a
group-theoretic classification problem, Ree noted that they could in fact be obtained
by a variant of the Steinberg procedure as a twisted form of the orthogonal groups
.£?2(2n). Indeed, it was known earlier that because of certain degeneracies in the
multiplication coefficients, the groups i?2(2 n ), G?2(3n) and F4(2 n ) possess an "extra"
automorphism. In the case that n is odd, this automorphism can be taken to have
order 2. The Suzuki groups are then twisted versions of B2 obtained using this
automorphism: Sz(2n) = 2 £?2(2 n). In 1961 Ree carried through his variation of the
Steinberg construction to produce the Ree groups, which were two new families
of twisted groups, 2 G 2 (3 n ) and 2 F 4 (2 n ), n odd [Rl, R2].
There are then 16 families of finite groups of Lie type altogether, twisted and
untwisted. The theories of linear algebraic groups and of (£?, AQ-pairs, which were
also flowering during this postwar period, gave uniform viewpoints for all the fam-
ilies. A general proof of simplicity was given by Tits [Ti3] (noting the handful of
exceptions, including the fact that 2 F 4 (2) is not simple, but its commutator sub-
group 2 F4(2)', which has index 2, is simple). In 1968, Steinberg [St2] gave a uniform
construction and characterization of all the finite groups of Lie type as groups of
fixed points of endomorphisms of linear algebraic groups over the algebraic closure
of a finite field.
1. SIMPLE GROUPS 11
The twenty-six sporadic groups (see [A2, CCPNW1]) arose from a variety of
group-theoretic contexts. 2 The first five were constructed by Mathieu in the 1860's
as highly transitive permutation groups [Ml, M2 , M3]. The remaining twenty-
one were discovered between the mid-1960's and 1980. Moreover, apart from the
three Conway groups [Col], which are connected with the automorphism group of
the 24-dimensional Leech lattice, all the others emerged either directly or indirectly
from one of the classification theorems being considered during this period.
The first of the twentieth-century sporadic groups was constructed by Janko
[Jl] as an exceptional solution to the problem of determining all simple groups in
which the centralizer of every involution (i.e., element of order 2) has the form
Z Q = 3 n ,
n odd, n > 1, and the single exception is Janko's group J\ for q = 5.
Altogether the initial evidence for eight sporadic groups came from an analysis
of centralizers of involutions: each of Janko's four groups, the Held, Lyons and
O'Nan groups, and the Fischer-Griess Monster group F\ [J2, J3, H i M c K l , N o l ,
H e l , He2, L2, S i l , O N 3 , A n l , Gr3].
Janko's second group J2 or HJ was ultimately constructed by M. Hall and
D. Wales [HaWal] as a rank 3 primitive permutation group (i.e., a transitive per-
mutation group in which a one-point stabilizer is a maximal subgroup and possesses
precisely 3 transitive constituents). This led to an intensive investigation of such
groups and the resulting discovery of four more sporadic groups: the Higman-Sims,
Suzuki, McLaughlin and Rudvalis groups [HiSil, Su5, M c L l , R u l , CoWal].
Fischer's three groups Fi22, ^ 2 3 and Fi24 were discovered in the course of
his characterization of the groups which are generated by a conjugacy class of
involutions the product of any two of which has order 1, 2, or 3 [Fil, Fi2]. In
addition to the previously known examples (which include the symmetric groups
and certain classical groups), he found the three additional ones, Fim, m = 22, 23,
24, which he constructed as rank 3 primitive permutation groups. Fischer's "Baby
Monster" F2 arose in turn from his study of the broader class of groups generated
by a conjugacy class of involutions the product of any two of which has order 1,2,
3 or 4 [LeSil].
Finally, the initial evidence for the Thompson and Harada groups F3 and F5
came from the structure of the centralizers of elements of order 3 and 5, respectively,
in the Monster [HI, T3]. However, the starting point for their actual analysis was
the structure of the centralizers of their involutions.
The finite simple groups are listed in Table I. The isomorphism question among
all these groups was largely settled by Artin [Arl] in the mid-1950's, who deter-
mined all coincidences of orders of the simple groups known to exist at that time.
Moreover, Artin's methods were extendable without difficulty to include the newer
groups of Lie type as they were discovered (see for example [Til] or [KLST1]).
The isomorphism question for sporadic groups was settled by proofs of the unique-
ness of each group as it was discovered, subject to various conditions—order, local
structure, existence of a certain permutation representation, etc. The complete list
of isomorphisms among the groups in Table I is given in Table II.
2
We give only an early reference or two for each group. Extensive bibliographies can be found
in [A2] and [ C C P N W 1 ] .
2. 3C-groups
At the time the classification effort was drawing to a close—the late 1970's and
early 1980's—the groups in Table I were naturally referred to as the "known simple
groups" or the "simple DC-groups", and we shall continue to use this terminology
for convenience. Of course, the gist of the Classification Theorem is that the word
"known" can be omitted.
In attempting to prove the classification, it is natural to proceed indirectly
by induction, focusing attention on a minima l counterexample—that is, on a
simple group G of least order not isomorphic to one of the known simple groups.
Establishing the theorem then becomes equivalent to reaching a contradiction by
showing that G itself must be isomorphic to a known simple group.
If H is a proper subgroup of such a minimal counterexample G, then every
simple section 3 of H has order less than that of G and hence by the minimality
assumption on G is necessarily isomorphic to one of the known simple groups. This
leads to the following fundamental definitions.
DEFINITION 2.1. A DC-group is a group X such that every simple section of
X is isomorphic to a known simple group, that is, to a group in Table I.
DEFINITION 2.2. A group G is DC-proper if and only if every proper subgroup
of G is a DC-group.
We have adopted the letter "3C" as an abbreviation for "known". The first
definition implies that if X is a DC-group and X is simple, then X is a known simple
group.
In this terminology, the statement of the Classification Theorem is proved in
the following equivalent form:
If G is a X-proper finite simple group, then G is a %-group.
The proof of the Classification Theorem involves a comprehensive study of the
proper subgroups of G. Because of this, it is necessary to develop an elaborate and
detailed theory of DC-groups.
On the other hand, clearly the term "DC-group" is of significance only as part of
the proof the classification theorem. Indeed, that theorem states that every simple
group is isomorphic to a known simple group and it follows that every finite group
is a %-group. Thus the theory of DC-groups developed for the classification proof
ultimately becomes incorporated into the general theory of finite groups.
B. T H E STRUCTURE OF F I N I TE G R O U P S
3. T h e J o r d a n - H o l d e r t h e o r e m a n d simple g r o u ps
The Jordan-Holder theorem, one of the oldest results in the theory of finite
groups 4 , gives a procedure for associating with an arbitrary finite group X a finite
3
A section of a group H is a quotient of a subgroup of H.
4
Most of the results stated in the next 6 sections, and in particular in sections 6-8, will
be discussed in more detail in subsequent chapters covering basic group-theoretical material.
3. THE JORDAN-HOLDER THEOREM AND SIMPLE GROUPS 13
sequence of finite simple groups, known as the composition factors of X, which

are uniquely determined by X up to isomorphism and the order in which they are
specified. Since every finite group is thereby a "composite" of finite simple groups,
the theorem shows that an understanding of simple groups 5 is of fundamental im-
portance for understanding the structure of arbitrary groups.
The composition factors of a given group X are constructed from a composi-
tion series of X, which is defined to be a sequence of subgroups 1 = Xo < X\ <
• • • < Xn = X of X such that
(1) Xi is a proper normal subgroup of Xi+i, 0 < i < n — 1; and
(2) Each Xi is maximal in -Xi+i, subject to (1).
The maximality of Xi implies that the n factor groups Xi+i/Xi, 0 < i < n — 1,
are all simple. These factor groups are called the composition factors of the
given series and the integer n is called the length of the series.
Thus the composition factors of X are sections of X obtained via a composition
series of X. The Jordan-Holder theorem asserts the following:
T H E O R EM 3.1. For any group X, any two composition series of X have the
same lengths, and for suitable orderings of the sequences of composition factors,
corresponding factors are isomorphic.
In particular, if every composition factor of X is of prime order, X is called a
solvable group.
At the other extreme, a group X is simple if X ^ 1 and the identity subgroup
is the only proper normal subgroup of X, in which case X possesses a unique
composition series 1 = Xo < X\ — X.
The theorem has a natural extension to groups with operators that enables one
to obtain a fuller picture of the role of simple groups in determining the structure
of an arbitrary finite group X. Indeed, if A is any set of endomorphisms of X, then
an A-composition series of X is by definition a sequence of subgroups Xi of X,
0 < i < n, such that Xi < Xi+i for all 0 < i < n, each Xi is A-invariant, and each
Xi is maximal in Xi+\ subject to these conditions. The groups Xi+i/Xi are called
the ^-composition factors of X.
The general form of the Jordan-Holder theorem asserts that for any set A of
endomorphisms of X, any two Acomposition series of X have the same lengths,
and if the lists of Acomposition factors arising from the two composition series are
suitably reordered, then corresponding Acomposition factors are Aisomorphic.
Thus the special form of the theorem stated above corresponds to the case in
which A is empty.
Of special importance to us here is the additional case in which A is the group
Inn(X) of all inner automorphisms of X. The condition that each subgroup Xi be
A invariant is then equivalent to the assertion that each Xi be normal in X. Such
Acomposition series are called chief series of X and the corresponding factor
groups Xi+i/Xi are called the chief factors of X. Thus likewise the chief factors
of X are uniquely determined by X up to isomorphism (preserving the conjugacy
action of X) and the sequence in which they occur in a chief series.
Moreover, any terms not defined here will be defined later. Much of the material can be found in
basic texts [ A l , G l , H u l , H u B l , S u l ] and the papers [ G W 2 , G W 4 , G L 1 ].
5
T o avoid endless repetition, let us agree tha t henceforth in this series, "group" means "finite
group" unless explicitly stated otherwise.
Furthermore, in a chief series, it follows from the maximality of Xi in Xi+i

that each Xi+i/Xi is a minimal normal subgroup of X/Xi and in particular,
Xi is a minimal normal subgroup of X. On the other hand, one easily establishes
the following result concerning the structure of minimal normal subgroups.
PROPOSITION 3.2. Every minimal normal subgroup of a group X is the direct
product of isomorphic simple groups.
It follows therefore that each chief factor of X is the direct product of isomorphic
simple groups. If these direct factors have prime order p, then the given chief
factor is elementary abelian and can be identified with the additive group of a
finite-dimensional vector space over the field Fp of p elements. In particular, if X
is solvable, then every chief factor of X has this form. We use the symbol Epn to
denote an elementary abelian p-group of rank n. In the case n = 1, we generally
use the more customary term Zp.
Likewise, it is easy to show by induction that every normal subgroup of X can
be included in a chief series of X, and, in particular, any minimal normal subgroup
can be so included. Furthermore, by considering the action of X on such a minimal
normal subgroup V by conjugation, one obtains an initial picture of the general
shape of the group X.
Indeed, suppose first that V is an elementary abelian p-group for some prime
p and hence a vector space over Fp, in which case for each x G l , the conjugation
map Int(x) : v H-> VX = x~xvx, V G V, is in fact a linear transformation on V.
Moreover, Int is a homomorphism of X into GL(V) whose kernel is precisely the
subgroup Cx{V) consisting of all elements of X that centralize V. Thus we have
PROPOSITION 3.3. If V is a minimal normal subgroup of X and V is abelian,
then via the conjugation action of X on V, X/Cx(V) is identified with a subgroup
of the general linear group GL(V).
In this case, since V itself is contained in Cx(V), the group X/Cx(V) is a
proper homomorphic image of X.
On the other hand, if the minimal normal subgroup V of X is the direct product
of nonabelian simple groups V*, 1 < i < r, then it is straightforward to establish that
the Vi are the only minimal normal subgroups of V and consequently each x E X
induces by conjugation a permutation of the set V = {Vi | 1 < i < r}. Thus there is
a natural homomorphism /3 of X into the symmetric group on the set V. The kernel
Y of (3 is the subgroup consisting of all elements of X leaving each Vi invariant.
Clearly VCx{V) < Y and Y/Cx(V) is isomorphic to a subgroup of Aut(V) leaving
each Vi invariant. It follows easily from this last condition that Y/Cx(V) is in fact
isomorphic to a subgroup of the direct product Aut(Vi) x Aut(Vz) x • • • x Aut(Vr).
Moreover, as V < Y, the image of V in Y/CX(V) is isomorphic to the group Inn(V)
of inner automorphisms of V. Hence in this case we have
PROPOSITION 3.4. Let V be a minimal normal subgroup of X which is the
direct product of nonabelian simple groups Vi, 1 < i < r. Let Y be the subgroup
consisting of all elements of X leaving each Vi invariant, 1 < i < r. Then
(i) Through the action of X permuting V i , . . . , Vr by conjugation, X/Y is iden-
tified with a subgroup o / E r ; and
(ii) Through the action ofY on V by conjugation, Y/Cx(V) is identified with
a subgroup of Aut(Vi) x Aut(V2) x • • • x Aut(Vr) containing Inn(V).
4. THE GENERALIZED FITTING SUBGROUP AND QUASISIMPLE GROUPS 15
4. The generalized Fitting subgroup and quasisimple groups

Unfortunately, the pictures of X given by the approaches just discussed have
serious inadequacies. On the one hand, even if one knows the isomorphism types
of the composition factors of the group X: the isomorphism type of X is far from
determined, and the "extension problem"—determining all X with given composi-
tion factors—has such depth and subtlety that to date it has been solved in only a
handful of special cases. On the other hand, if one tries to understand X by means
of its action on a minimal normal subgroup V, one has no information about the
subgroup Cxiy) whose action on V is trivial. Moreover, Cx(V) may include a
large portion of X. Indeed, V may well be in the center Z(X) of X, in which case
Cxiy) — X. For example, this will occur if V is an abelian direct factor of X (i.e.,
X = V x XQ for some subgroup Xo of X).
As a first attempt to eliminate this deficiency, one might replace V by the
subgroup V* generated by all minimal normal subgroups of X and then consider
the action of X on V* (this would at least exclude the preceding example, since
clearly V* will never be a proper direct factor of X). Moreover, the structure of
V* is very similar to that of a minimal normal subgroup. Indeed, if V and W are
any two distinct minimal normal subgroups of X, then as V H W is also normal in
X , their minimality forces V n W = 1, and it follows that (V, W) = VW = V xW.
This argument is easily extended to yield:
PROPOSITION 4.1. IfV* is the subgroup of X generated by all minimal normal

subgroups of X, then
V* = V1xV2x-..xVm
for suitable minimal normal subgroups V i , . . . , Vm of X.
However, it still may happen that V* is in the center of X. For example, if
X = SL,2(q), the group of 2 x 2 matrices of determinant 1 with coefficients in
Fq, and q is odd, then X possesses a unique minimal normal subgroup, which is
generated by the scalar matrix —L Thus in this case V* = V = Z(X).
On the other hand, there is an important special case in which Cxiy*) = 15
in which case X does act faithfully on V*. Indeed, if X contains no nontrivial
solvable normal subgroups, then every minimal normal subgroup is a direct product
of nonabelian simple groups and so therefore is V*. But then CV*(V*) — 1 and
consequently CX(V*) n V * = l. Since CX(V*) < X, the definition of V* now
forces CxiV*) = 1, as asserted.
This leads to the following natural question: Does there exist a canonically
defined normal subgroup W of X having both a relatively uncomplicated structure
and the property Cx{W) < Wl
As the preceding discussion makes clear, central subgroups of X are initial
obstructions to locating such a subgroup W in X. Moreover, the class of groups
constructed inductively from central subgroups is that of the nilpotent groups,
which suggests that nilpotent normal subgroups may well be a key to the question's
answer.
There are several equivalent definitions of a nilpotent group. In the spirit of the
present discussion, we take the following: X is said to be nilpotent if X possesses
a chief series 1 = Xo < X\ < • • • < Xn = X such that
Xi+1/Xi < Z(X/Xi) for all f, 0 < i < n - 1.

In particular, each Xi+i/Xi must then be abelian. In fact, as Xi+i/Xi is a

minimal normal subgroup o£X/Xi, each Xi+i/Xi must be of prime order. It follows
therefore that a nilpotent group is necessarily solvable.
We state two basic facts concerning the structure of nilpotent groups.
PROPOSITION 4.2. Every group of prime power order is nilpotent.
PROPOSITION 4.3. X is nilpotent if and only if X is the direct product of its
Sylow subgroups.
Considering nilpotent normal subgroups of the group X, it is not difficult to
show that the subgroup generated by any two such subgroups is itself a nilpotent
normal subgroup of X, which immediately yields the following results.
PROPOSITION 4.4. X contains a unique normal subgroup maximal subject to
being nilpotent.
This maximal nilpotent normal subgroup of X is called the Fitting subgroup
of X and is denoted by F(X). Thus F(X) is the subgroup of X generated by
all nilpotent normal subgroups of X. In particular, F(X) contains every abelian
normal subgroup of X. Clearly F(X) is characteristic in X, that is, invariant
under every automorphism of X.
The following theorem of Philip Hall shows that the Fitting subgroup provides
a satisfactory answer to the above question in solvable groups.
THEOREM 4.5. If X is solvable, then CX{F(X)) < F(X).
Note that if X is solvable (and nontrivial), then F(X) ^ 1.
Bender has extended Hall's theorem to arbitrary finite groups. The form of
his result reveals the fundamental role that perfect central extensions (or covering
groups) of simple groups play in the theory of groups. We make the following
definition:
DEFINITION 4.6. The group X is quasisimple if and only if X is perfect
(i.e., X = [X,X]) and X/Z(X) is simple.
In particular, quasisimple groups are necessarily nonsolvable. Also it is obvious
that nonabelian simple groups are quasisimple. The groups SL2(q), q odd, q > 5,
afford examples of quasisimple groups which are not simple; Z(SL2(q)) is of order
2, generated by the matrix —I. We note that the group 6X2(3) is solvable and
hence not quasisimple.
It is natural to consider central products of quasisimple groups. Thus X is said
to be semisimple if X is the product of quasisimple groups Xi, 1 < i < m, such
that
[Xi,Xj] = 1 for alii ^ j ,
that is, X{ and Xj centralize each other elementwise for i ^ j .
For completeness, the identity group is included among the semisimple groups.
If X is semisimple, then the quasisimple factors Xi of X can be characterized
as the minimal nonsolvable normal subgroups of X and so are in fact uniquely
determined by X.
We mention two important properties of a semisimple group X. First, if Y < X
and Y covers X/Z(X), then Y = X. Second, if a group A acts on the quasisimple
4. THE GENERALIZED FITTING SUBGROUP AND QUASISIMPLE GROUPS 17
group X, and acts trivially on X / Z ( X ) , then it acts trivially on X, This is a simple

consequence of the three subgroups lemma 6 .
There exists a direct analogue of Proposition 4.4 for semisimple normal sub-
groups of a group X, since it easily shown that the subgroup generated by any two
such subgroups is itself a semisimple normal subgroup of X. Thus we have
PROPOSITION 4.7. X contains a unique normal subgroup which is maximal
subject to being semisimple.
This maximal semisimple normal subgroup is called the layer of X and is
denoted by E(X). Clearly likewise E(X) is characteristic in X.
If E(X) 7^ 1, the uniquely determined quasisimple factors of X are called
the c o m p o n e n t s of X. In particular, it follows therefore that each element of
X induces by conjugation a permutation of the set of components of X. More-
over, the components can be equivalently described as the quasisimple subnormal
subgroups of X. Furthermore, Z(E(X)) is an abelian normal subgroup of X,
and E(X)/Z(E(X)) has no trivial normal nilpotent subgroups, so E(X)C\F(X) —
Z(E(X)). Since E[X) and F(X) are both normal in X, [E(X),F(X)] < Z(E(X)),
so by the second property above, E(X) centralizes F(X).
Bender has called the group E(X)F(X) the generalized Fitting subgroup
of X and denoted it by F*(X). His theorem provides an elegant extension of Hall's
theorem to arbitrary finite groups, thereby justifying use of this terminology.
T H E O R EM 4.8. (Bender) For any group X, we have
(i) Cx{F*(X))<F*(X);and
(ii) If W is a normal subgroup of X such that Cx(W) < W, then E(X) < W.
Statement (ii) is included only to show the inevitability of considering E(X).
Statement (i) is the main point, and implies that
CX(F*(X)) = Z{F*{X)) = Z(E(X))Z(F(X)) = Z(F(X)),
so that by the conjugation action, X/Z(F(X)) embeds in Aut(F*(X)). In partic-
ular, if X is nontrivial, then so is F*(X).
Of course, if X is solvable, then E(X) = 1, whence F*(X) = F(X). At
the opposite extreme, if X contains no nontrivial solvable normal subgroups, then
F*(X) = E{X) is the direct product of nonabelian simple groups.
Because of Theorem 4.8, for many purposes the analysis of a group X reduces
to an analysis of the structure of F*(X) and of its automorphism group. The group
extensions essential to understanding F*(X) are of limited types, since F*(X) is
the product of pairwise commuting p-groups, for various primes p, and quasisimple
groups. For this reason, the problem of determining all quasisimple groups is a
critical one, and so the extension problem which is arguably the most vital in
finite group theory is to determine all perfect central extensions of a given simple
group. This subtle problem has been solved for all the known simple groups by the
combined work of many authors, principally Schur, Steinberg and Griess. Indeed,
Schur showed that a (finite) simple group has a universal covering group which
is itself finite—that is, a finite perfect central extension of which all others are
homomorphic images. For all the simple groups, the universal covering groups and
6
This lemma states that if A, B, and C are subgroups of a group and [A, B, C] = [B, C,A] = 1,
then [C, A,B] = 1. Here as always, [A, B, C] = [[A, B], C] and [A, B] = ([a, b] | a G A, b e B).
their centers (called the Schur multipliers) have been calculated (see [Schl, St5,
G r l , Gr2, C C N P W 1 ] . )
For reasons to be discussed presently, although finite nilpotent groups can be
the direct product of p-groups for arbitrarily many primes p, for most of the groups
X playing a critical role in the classification theory either E{X) ^ 1 or F*(X) is
a p-group for some prime p. In any case, since p-groups can be very complicated,
it is desirable to obtain faithful action of X/Z(F(X)) on a suitable homomorphic
image of F*(X) having simpler p-subgroup structure than F*(X) itself.
Recall that in any group X, the intersection of all maximal subgroups of X is
called the Frattini subgroup of X and is denoted by &(X). It, too, is clearly a
characteristic and hence normal subgroup of X. For groups of prime power order,
one has the following basic properties of the Frattini subgroup.
PROPOSITION 4.9. If X is a p-group for some prime p, then we have
(i) X/$(X) is an elementary abelian p-group; and
(ii) / / a is an automorphism of X whose order r is relatively prime to p, then
a induces an automorphism of X/<&(X) of order r.
Since a nilpotent group is the direct product of its Sylow subgroups, it follows if
X is nilpotent that X/$(X) is an abelian group of square-free exponent, isomorphic
to the direct product of the Frattini factor groups of its Sylow subgroups.
We have noted already that if X is a semisimple group and a is an automor-
phism of X of order r, then a induces an automorphism of X/Z(X) of order r.
Combining this observation with Proposition 4.9, one can establish the following
result.
PROPOSITION 4.10. For any group X, if we set X = X/Z{E(X))&(F(X)),
then
F*{X) = F*(X).
(Here we have used the bar convention, that the image of a subset or element
Y of X under the canonical projection X —> X is denoted by Y. We shall use this
convention throughout.)
By definition of X, E(X) is the direct product of nonabelian simple groups and
F(X) is the direct product of elementary abelian groups. Thus C-^(F*(X)) = F(X)
and X/F(X) (= X/F(X)) is faithfully represented as a group of automorphisms
onF*(X) = E(X) xF(X).
Through Bender's theorem, many questions about the structure of the general
finite group X reduce to questions about subgroups of X that leave invariant a com-
ponent of X and in turn to the study of groups Y such that F*(Y) is quasisimple. 7
Aschbacher has termed such a group Y almost simple.
If X is an almost simple group, Proposition 4.10 implies that F*(X) is simple
and so by Theorem 4.8, X is isomorphic to a subgroup of Aut(F*(X)) containing
Inn(F*(X)). Thus, an almost simple group is constructed from a simple group
by "decorating" at the bottom and the top, i.e., by permitting perfect central
extensions and outer automorphisms.
7
There are analogous structure theorems of Aschbacher, O'Nan and Scott in the theories of
permutation groups and linear groups, leading to useful reduction paradigms in these theories
[A17, S c l ] .
5. p'-CORES AND p-COMPONENTS 19
5. p'-cores and p-components

We have taken considerable length to motivate the generalized Fitting subgroup
F*(X) of group X. One must now ask how effective this subgroup is for studying
the group X itself.
On the positive side, one can at least assert that F*(X) bounds the order of
X inasmuch as X/Z(F(X)) is isomorphic to a subgroup of Aut(F*(X)). On the
other hand, the structure of X/F*(X) may be arbitrarily complex. We illustrate
the point with solvable groups.
Suppose then that X is solvable and X ^ 1 and define the Fitting series
Fi(X) of X inductively as follows:
(1) F^X) = F{X); and
(2) Fi+1(X)/Fi(X) = F(X/Fi(X)).
Thus F{+i(X) is the complete inverse image of F{X/Fi{X)) in X. Since ho-
momorphic images of solvable groups are solvable and since the Fitting subgroup
of a nontrivial solvable group is nontrivial, it is immediate that the Fitting series
terminates in X itself. We therefore obtain a series:
F(X) = Fi(X ) < F2{X) < < Fh(X) = X.
The integer h is called the Fitting length of X. By the following simple result,
solvable groups exist with arbitrarily large Fitting length.
PROPOSITION 5.1. If Y is a group, then there exists a group X such that
Y^X/F(X).
Despite the possible complexity of the structure of the group X/F*(X), the
significance of the generalized Fitting subgroup rests on the fact that F* (X) itself
controls a key portion of the local structure of X. In the study of a simple group
G the application of this fact to local subgroups makes possible the transfer of
information among various local subgroups in identifiable chief factors of these
subgroups. We shall describe this control and transfer of information in the next
two sections. Here we introduce the general terms in which that description will be
expressed. We first formalize the key definition of a local subgroup.
DEFINITION 5.2. A local subgroup of X is a subgroup Y of X which is the
normalizer Y = Nx(Q) of some nontrivial solvable subgroup Q of X. If Q is a
nontrivial p-group for some prime p, Y is called a p-local subgroup of X.
Often, we shall simply say "Y is a local" to abbreviate "Y is a local subgroup
of X", and we use a similar contraction with the expression "Y is a p-loeal".
Next, for any set n of primes, X contains a unique normal subgroup On(X)
that is maximal subject to the condition that 07r(X) has order divisible only by
primes in 7r. Likewise X contains a unique normal subgroup On(X) minimal such
that X/07T(X) has order divisible only by primes in n. We also denote by n' the
complementary set of primes to n, and if n = {p}, we write Op(X), Opr(X), Op(X),
P
OP'(X) for 0{p}(X), 0{p},{X), 0* >(X), O W ' ( X ) , respectively. Op,{X) is called
the p'-core of X. When p = 2, we call OîX) the core of X and denote it by
0(X). Thus 0(X) is the unique largest normal subgroup of X of odd order. By
the Feit-Thompson Odd Order Theorem, 0(X) is solvable.
The following facts are direct consequences of the definitions.
LEMMA 5.3. For any set of primes n, we have

(i) On(X/Ow(X)) = 1; and
(ii) IfOn(X) = 1, then F(X) has order divisible only by primes in IT' and every
component of X has order divisible by some prime in irr.
In particular, let X be a group with Op'(X) = 1. Then F(X) = Op(X) and
every component of X has order divisible by p. In this case we say that X is
p-constrained if E(X) = 1, i.e.,
F*(X) = Op(X).
Hence when 0P'(X) = 1, p-constraint can be equivalently defined by the condition
Cx(Op(X)) < Op(X).
More generally, an arbitrary group X is said to be p-constrained if and only if
X/Op' (X) is p-constrained.
Clearly solvable groups are p-constrained for every prime p.
Concerning components of XjOp>[X), we have the following basic result.
PROPOSITION 5.4. Let X be a group, p a prime, and set X = X/Op>(X).
Then X contains a unique normal subgroup L minimal subject to the condition
L = E(X). Moreover, if E(X) =fi 1, then L is the product of normal subgroups Li,
1 (X) for all i^ j , 1 < i, j < r;
(iv) The Li are the only normal subgroups of L—and the only subnormal sub-
groups of X— satisfying (i) and (ii); and
(v) Each element of X induces a permutation of the set { L i , . . . , Lr} by conju-
gation.
For completeness we set r = 0 if E(X) = 1. L is called the p-layer of X and
is denoted by Lp/(X). Furthermore, the subgroups Li, 1 (X) = 1).
The connection between components and p-components is straightforward.
PROPOSITION 5.5. LetX be a group andp a prime, and let K be ap-component
of X. Then the following conditions are equivalent: K is a component of X; K is
quasisimple; K centralizes Op/(K); K centralizes 0P'(X). Conversely, any compo-
nent of X of order divisible by p is a p-component of X.
6. The embedding of p'-cores and p-components

The critical property of the generalized Fitting subgroup is its control of the
embedding in X of the p'-cores and p-layer of every p-local subgroup Y of X—
more precisely, it is the generalized Fitting subgroup of XjOp> (X) that exercises
this control. It is this control that allows one to analyze effectively the subgroup
structure of arbitrary finite simple groups and ultimately force key portions of that
structure to approximate closely the corresponding subgroup structure in one of
the known simple groups.
6. THE EMBEDDING OF p'-CORES AND p-COMPONENTS 21
We fix the group X and the prime p and first consider the embedding of p'-cores
when X is p-eonstrained (in particular, when X is solvable).
PROPOSITION 6.1. If X is p-constrained, then Op>(Y) < Op'(X) for every
p-local subgroup Y of X.
This result is easily proved by passing t o X = X/Op' (X) and invoking Thomp-
son's Ax B lemma to conclude that the image V of Op>(Y) in X centralizes Op(X).
Since F*(X) = Op(X) and C^(F*(X)) < F*{X), this forces V = 1, whence
Op' (Y) < Op' (X). [The Ax B lemma states that if P is a p-group on which Ax B
acts, with A a p'-group and B a p-group, then A centralizes P if and only if A
centralizes Cp(B).]
In the general case, the same argument yields that V centralizes Op(X) and
hence that V < X0 = C^{Op(X)). Again as C X (F*(X)) < F*(X) and F*(X) =
E(X)Op(X), it follows that V acts faithfully on E(X). Ideally, one would like to
be able to assert that V < E(X). One can in fact prove that V normalizes each
component of X; however, in general, V need not induce inner automorphisms on
E(X). This will explain the need to introduce the following term.
Define Lp'(X) to be the subgroup of X consisting of the elements of X that
leave each p-component of X invariant and centralize Op(X/Op'(X)). Clearly
Op'{X)LP'{X) < Lp'(X) and LP'(X) is characteristic and hence normal in X.
THEOREM 6.2. For any p-local subgroup Y of X, 0P'(Y) < LP'(X).
The corresponding general result for p-layers, known as Lp>-balance, has been
given a proof independent of the Classification Theorem only for p = 2. For p > 2,
it has not yet been proved without some assumption on the components K of
X = X/Op'(X). The most convenient such assumption is that each K/Z(K) has
the Schreier property—that is, its outer automorphism group is solvable8. Now
the automorphism groups of the known simple groups have all been calculated,
and observed to have this property conjectured by Schreier. Of course the case in
which the components of X are all covering groups of known simple groups is the
only case that we need, since we shall apply the result only to proper sections of a
DC-proper simple group.
THEORE M 6.3. (Lvi -balance) If the components of X/Op'(X) have the Schreier
property, then LP'(Y) < LP'(X) for every p-local subgroup Y of X.
[In this connection, we note that Theorem 6.2 can also be slightly sharpened
when the components of X are DC-groups. In that case, for example, the im-
age of 0P'(Y) in Xo/E(X) is actually contained in OP'(XQ/E(X)), where Xo =
C^(Op(X)) (cf. Theorem 30.3 of Chapter 2).]
The effect of Theorems 6.2 and 6.3 is that a very large part of the p-local
analysis of a X-proper simple group focuses on the sections LP'(Y)/Op'(Y) for p-
local subgroups Y of G. Although the structure of Y may be extremely complicated,
this section is an an extension of a central p-group by a subgroup of Aut(Li) x • • • x
Aut{Lr) containing Inn(Li) x • • • x Inn(Lr), where L i , . . . , Lr are the components
ofY/Op'(Y).
8
Weaker conditions will do, and when p = 2 a theorem of Glauberman [G13] shows tha t no
extra assumption is necessary.
7. Terminal and p-terminal p-components

In the study of the p-local structure of a X-proper simple group X, the cen-
tralizers of elements of order p play an especially important role. Let us suppose
for this section that the centralizers of some such elements have p-components. In
such a case, those p-components dominate the analysis.
In particular, Theorem 6.3 takes a more precise form for centralizers of com-
muting elements of order p in an arbitrary group X. However, for it to be applicable
for odd p, it is again necessary to assume that
K has the Schreier property for every x G l o f order p and every
component K of Cx{x) = Cx{%) / Op> {Cx(%)) >
We therefore make this assumption throughout the section.
PROPOSITION 7.1. If x and y are commuting elements of order p in a group
X and X satisfies (7.1), then the following conditions hold.
(i) Lp,{Lp,(Cx(x))nCx(y)) = Lv,{Lp,{Cx{y))r\Cx{x)) = Lp,(Cx((x,y)));
and
(ii) If I is a p-component of LP'(Cx{x)) n Cx{y) and J is the normal closure
of I in Lp>(Cx(y)) (note that I < Lp>(Cx{y)) by (i)), then we have
(1) J is either a single x-invariant p-component of Cx{y) or the product of
p p-components of Cx(y) cycled by x; and
(2) In either case, I is a p-component ofCj(x).
One refers to J as the pumpup of / in Cx(y)- According as J is a single
p-component or a product of p p-components, the pumpup is called vertical or
diagonal. Moreover, in the vertical case, either x centralizes or does not centralize
J/Opt(J), and correspondingly the pumpup is said to be trivial or proper.
If the pumpup J of J in Cx(y) is trivial, then x centralizes J/Op'(J) and
it follows that / maps onto the quasisimple group J/Op>(J). Hence in this case,
J = IOp>(J) and consequently IjOp\I) ^ J/Op>(J).
Proposition 7.1(h) links the p-components in the centralizer of one element of
order p in X to those in the centralizer of another (commuting) element of order p,
and enables one to analyze the configuration of p-components of the centralizers of
all the elements of order p in X by means of the pumping-up process. That analysis
in fact revolves around such p-components that are maximal in the sense that they
possess only trivial pumpups. Such p-components are referred to as p-terminal, a
term we now define precisely.
However, to do so requires preliminary notation. First, we use the notation
Cx{A/B), whenever A and B are subgroups of X with B < A, to mean the
subgroup consisting of all y G X such that [y,A] < B; equivalently, Cx(A/B) is
the largest subgroup of Nx(A) Pi Nx(B) acting trivially by conjugation on A/B.
Next, for any element x of order p in X and any p-component K of Cx{x), we let
C{K,x) = Ccx(x)(K/Op'(K))i the subgroup of Cx(x) consisting of all elements
leaving K invariant and centralizing K/Op'(K). Note that if K is quasisimple
(equivalently, if K is a component of Cx(oc)), then C(K,x) = Cx(K) fl Cx{x).
If y G C(K,x) is of order p, then as y centralizes K/Op>(K) and has order
prime to \Opt(K)\, it can be shown that K — CK(y)Op'(K). Hence if we set
/ = Lp>(CK(y)), then likewise / maps onto K/Op>(K) and K = IOp>{K), whence
I/Opl{I)^K/Op>{K).
7. T E R M I N A L AND p - T E R M I N A L p - C O M P O N E N T S 23
Furthermore, we see that / is in fact a p-component of Lp>(Cx{x)) H Cx(y)- In

particular, if K is quasisimple, then clearly I = K.
Thus in any case, Proposition 7.1(h) applies to / . In this case we write Ky for
the pumpup of / in Cx(y), and sometimes call Ky a pumpup of K rather than of
/ . In particular, if Ky is a trivial pumpup of / , it follows that / maps onto both
KjOp\K) and Ky/Op>{Ky), so that
K/Op,{K) * I/Opf{I) * Ky/Op>{Ky).
Note also that if a G C(K,x) centralizes y, then a leaves / invariant and as /
covers K/Op'(K), a centralizes I/Op'(I). Moreover, since Ky is the normal closure
of i" in Lp'(Cx(y)), a must leave Ky invariant. But now if Ky is a trivial pumpup,
then I covers Ky/Op'(Ky), and so it follows that a centralizes Ky/Op'(Ky), whence
a G C(Ky,y). Hence in the trivial pumpup case, we conclude that
C(K,x)nCx(y)<C(Ky,y).
With all these preliminaries, we are now ready to introduce the notion of a
p-terminal p-component.
DEFINITION 7.2. If K is a p-component of Cx(x) for some x £ X of order p,
we say that K is p-terminal in X (more properly, the pair (K, x) is p-terminal in
X) if and only if the following two conditions hold for every y G C(K,x) of order
p:
(1) Lp'{CK{y)) has a trivial pumpup Ky in Cx(y)\ and
(2) If Q e Sylp(C(K,x)) and y G Z(Q), then Q G Sylp(C{Ky,y)).
The preceding discussion shows that if (1) holds, and Q and y satisfy the
hypothesis of (2), then Q < C(Ky,y). Also, we use the notation Sylp(Y) for the
set of Sylow p-subgroups of the group Y.
One establishes the following general result.
THEORE M 7.3. / / the centralizer of some element of X of order p has a p-
component, then there exists an element x G X of order p and a p-component K of
Cx(x) with K p-terminal in X.
One can in fact prove a much sharper result. Indeed, beginning with any
x\ G X of order p and p-component K\ of C x ( ^ i ) , one can reach x and K with
K p-terminal in X by following a suitable sequence of pumpups. Thus every p-
component of the centralizer of an element of order p in X can be "pumped up"
to a p-terminal p-component in X. Special versions of Theorem 7.3, with extra
hypotheses and stronger conclusions, will be established in later volumes of this
series.
A considerable portion of the p-local analysis of the general simple group X is
devoted to eliminating the p'-core "obstruction" of p-terminal p-components and
their pumpups. The desired conclusions are incorporated into the following defini-
tion.
DEFINITION 7.4. If K is a p-terminal p-component of Cx(x) for some x G X
of order p, we say that K is terminal in X if and only if K is a component of
Cx{y) for every element y G C(K, x) of order p.
In particular, the definition implies (with y = x) that K itself must then be
a component of Cx (x). It also implies that K is a component of Cx (y) for every
V € Cx{K) of order p. Indeed, let Q and R be Sylow p-subgroups of C(K,x)

and Cx{K), respectively, with Q < R. Since C(K,x) = Cx(K) n Cx(a;), we have
Q = CR(X). Choose z G Z(-R) of order p. Then z G Q, so that by definition of p-
terminality, Q is a Sylow subgroup of C(if, z). This implies that Q = R. Hence the
desired assertion follows from the definition of terminality. [We prefer "terminal"
to the current term "nonembedded" to describe p-components satisfying the given
conditions, because of the number of distinct usages of "embedded" in finite group
theory.]
The following result, which is a direct consequence of the definitions and Propo-
sition 5.5, expresses the operational conditions for a p-terminal p-component to be
terminal.
PROPOSITION 7.5. If K is a p-component of Cx{x) for some x € X of order
p and K is p-terminal in X, then K is terminal in X if and only if Ky centralizes
Op'{Cx(y)) for every element y G C(K,x) of order p.
The existence of terminal p-components is closely linked to the so-called Bp-
property. A group X with Op'(X) = 1 is said to have the ^ - p r o p e r t y if and
only if every p-component of the centralizer of every element of order p in X is
quasisimple. This general property has been verified directly for almost simple X-
groups, and in this form is needed for the classification proof. Moreover, it has
the following immediate consequence, analogous to L p /-balance. For any group X
define BP'(X) to be the product of all p-components L of X such that L is not
quasisimple (equivalently: L ^ E(X)).
T H E O R E M 7.6. Let X be a group in which all almost simple sections with trivial
p'-core have the Bp-property and the Schreier property. Then Bp/(Y) < Bpr{X) for
every p-local subgroup Y of X.
As noted, once the Classification Theorem is proved, it implies that the hy-
pothesis of Theorem 7.6 holds for any group X.
8. p-constrained and p-solvable groups

As the discussion of the generalized Fitting subgroup suggests, the study of an
arbitrary group X has a natural division into the analysis of its solvable substructure
on the one hand, and its quasisimple substructure on the other. Thus, in order to
analyze the p-local subgroups of X for some prime p, it is necessary to consider
not only the p-components of its p-locals, but also its p-solvable substructure,
p-solvability being a form of solvability "localized at the prime p."
By definition, a group X is p-solvable if and only if every composition factor
of X has order either equal to p or relatively prime to p. Clearly every solvable
group is p-solvable for any prime p. Conversely, in view of the solvability of groups
of odd order, every 2-solvable group is solvable. Furthermore, any p-solvable group
is p-constrained.
In any group X, the subgroup 0P'P(X) is p-solvable. [By definition, Op>p{X)
is the full preimage in X of Op(X/Op>(X)).] The study of the p'-cores of p-locals
relies on various basic results about p-groups acting on p'-groups. Underlying many
of these results is the fundamental Schur-Zassenhaus theorem.
T H E O R E M 8.1. If Y is a normal subgroup of X whose order is relatively prime
to its index in X, then
8. p-CONSTRAINED AND p-SOLVABLE GROUPS 25
(i) Y possesses a complement in X (i.e., there is A < X such that X = YA

and Y C\A= 1), and
(ii) Any two complements to Y are conjugate in X.
The only known proof of (ii) uses the solvability of one of the groups Y and
X/Y, which in turn depends on the solvability of groups of odd order.
The key properties of p-solvable groups are contained in the following variation
of Sylow's theorem, due to Philip Hall. If a is a set of primes, we define a a-
subgroup of H to be a subgroup whose order is divisible only by primes in a. We
define a Hall a-subgroup of H to be a a-subgroup W such that \H : W\ has no
prime divisors in a.
T H E O R E M 8.2. If X = YA with Y normal in X and (\Y\, \A\) = 1, then
(i) For any prime q,
(1) A leaves invariant a Sylow q-subgroup ofY;
(2) Any two A-invariant Sylow q-subgroups of Y are conjugate by an ele-
ment ofCy(A); and
(3) Any A-invariant q-subgroup ofY is contained in an A-invariant Sylow
q-subgroup ofY.
(ii) / / Y is solvable and a is any set of primes, then the assertions of (i) hold
with "(Hall) a-subgroup" in place of "(Sylow) q-subgroup".
One also has the following key generational result in the above situation.
PROPOSITION 8.3. Assume that X = YA with Y normal in X and (\Y\, \A\) =
1. If A is abelian but not cyclic, then
Y=(CY(a)\aeA#).
In the analysis of p-local structure, Theorem 8.2 and Proposition 8.3 are applied
with A a p-group.
Finally, we consider the structure of p-constrained groups Y such that Op> (Y) =
1, that is, groups such that F*(Y) = Op(Y). By a theorem of Borel and Tits
[BuWil] , all p-locals in simple groups of Lie type defined with respect to fields of
characteristic p enjoy this property. Accordingly, when all p-locals in an arbitrary
simple group X have this property, X is said to have characteristic p-type 9 . If X
is simple of characteristic p-type, then the study of p-locals by their p'-cores and p-
components that we have been discussing is vacuous. Instead, the focus of analysis
is on chief factors V = Y1/I2 of p-local subgroups Y with Y\ < Op(Y) (we say that
V is "within" Op(Y)). The aim is to show that unless certain specific types of chief
factors occur, then the structure of Y is controlled by the normalizers of certain
characteristic subgroups of a Sylow p-subgroup P of Y. The purpose is to be able
to link the structures of various p-local subgroups.
Again we have to introduce some terminology.
DEFINITION 8.4. If X is a group with Op(X) = 1 and V is a faithful FPX-
module, then V is a quadratic FpX-module with respect to the elementary abelian
p-subgroup A of X if and only if A ^ 1 and [V, A,A] = 1 in the semidirect product
of V by X.
Returning to a chief factor V of the group Y within Op(Y), we have that V is a
faithful irreducible i^y-module for the group Y = Y/Cy(V), which quickly implies
9
T h e complete definition requires the p-rank to be at least 2, and at least 3 if p = 2.
that Op(Y) = 1. Thus we shall call V quadratic with respect to the elementary
abelian p-subgroup A of Y if and only if V is a quadratic P p y-module with respect
to A in the sense of Definition 8.4; and we simply call V quadratic if and only if
it is quadratic with respect to some elementary abelian p-subgroup A^lofY.
If A = (g) has order p in Definition 8.4, then that definition requires that the
minimal polynomial of g in its action on V be (t — l ) 2 , rather than (t — l)p = tp — 1,
as one might expect. Of course if p = 2 and X has even order, any faithful F2X-
module is quadratic with respect to any subgroup of X of order 2.
Thus the notion is nontrivial only for odd p or | A\ > 4. In these cases, quadratic
modules are rather rare among all isomorphism types. For example, if p > 5 and G
is solvable with Op(G) = 1, then G has no quadratic modules. This is a corollary
of P. Hall and G. Higman's Theorem B [HaHil].
At the start of the theory of simple groups of characteristic p-type, one has the
following two celebrated theorems. For any p-group P , we define the Thompson
subgroup J(P) to be the subgroup of P generated by all the elementary abelian
subgroups of P of largest possible order. We also define Qi(B), for any abelian p-
group B, to be the largest subgroup of B of exponent p. Then J{P) and fii(Z(P))
are characteristic subgroups of P , and both are nontrivial if P is nontrivial.
The Z J-theorem has a number of closely related versions, one of which is the
following.
T H E O R E M 8.5. (Glauberman's Z J-theorem) Let p be an odd prime and let Y
be a group such that F*(Y) = Op(Y). Let P be a Sylow p-subgroup ofY. If no
chief factor ofY within Op(Y) is quadratic, then £l\(Z(J(P))) < Y.
If one weakens the hypotheses of this theorem by allowing p = 2 or more gener-
ally by allowing quadratic modules, one cannot hope to name a single characteristic
subgroup of P which will be necessarily normal in Y. However, the Thompson fac-
torization theorem shows that under some conditions two characteristic subgroups
exist whose normalizers together cover Y. For that theorem, the obstructions are
fewer: they form a subset of the full set of quadratic modules. If V is a faithful
FpX-module, call V a failure of factorization module if and only if X has an
elementary abelian p-subgroup 4 ^ 1 such that
(8.1) \A\ > \V/CV(A)\.
T H E O R E M 8.6. (Thompson) IfF*(Y) = Op(Y) and V = « i ( Z ( 0 p ( y ) ) ) is not
a failure of factorization module for Y/CY{V), then Y = JVy(J(P))Cy(£li(Z(P))).
On the other hand, the Thompson Replacement Theorem implies that if V is
a failure of factorization module for Y = Y/Cy(V), then V is a quadratic FPY-
module with respect to some elementary abelian p-subgroup 4 ^ 1 satisfying (8.1).
This shows the connection between failure of factorization and quadratic modules.
The presence of quadratic or failure of factorization modules complicates the
theory of groups of characteristic p-type—unavoidably so, since quadratic modules
are ubiquitous in the parabolic subgroups of groups of Lie type. In their presence,
the more subtle method of amalgams can be used to link the structures of the p-
local subgroups. This method is discussed in some detail in section 33 of the next
chapter.
9. I N T E R N A L ANALYSIS: T A R G E T I N G L O C A L S T R U C T U R E 27
C. CLASSIFYING SIMPLE G R O U PS
9. Internal analysis: targeting local structure

Prom the more general structural properties of finite groups which we have been
discussing, we begin in this section to narrow our focus toward the classification.
We discuss some important aspects of the way in which the simple groups were
actually classified, including mention of ways to recognize them. We also preview
our general strategy.
At the outset of the proof of the classification theorem, one is faced with a
JC-proper simple group G. Because of Sylow's theorem, G has an abundance of
p-subgroups for various primes p, and hence (by taking normalizers) an abundance
of local subgroups. The bulk of the analysis of G then focuses on the subgroup
structure of G, and in particular on the structure of local subgroups and the re-
lations among their embeddings in G. In a few places certain objects external to
(2, namely linear representations, must be considered. Although these situations
are crucial, they are isolated and occur in cases where G is "small" in some sense.
Overwhelmingly, the classification proof, both in its original form and as revised in
this series, consists of an analysis of the local subgroups of G.
The goal is to show that for some known simple group G*, the local subgroup
structures of G and G* resemble each other so much that one can then prove, as
will be discussed in Section 11, that G = G*.
Because of very strong family similarities among the local structures of all
the simple groups of Lie type, and also among the alternating groups, the "target"
subgroup structure of G* is extremely restricted in nature. Therefore it is a massive
step to pass from the initial data that all proper subgroups of G are DC-groups to
the goal that G and G* locally resemble each other.
We illustrate the magnitude of this passage with some comments about the
structure of centralizers of involutions and other local subgroups in the known sim-
ple groups. If C is the centralizer in the X-proper simple group G of an involution
x, then a priori, C may have arbitrarily many nonsolvable composition factors
of arbitrarily varied isomorphism types and, in addition, C may contain normal
subgroups of odd order of arbitrarily large Fitting lengths.
On the other hand, consider the centralizer C* of an involution x* in one of
the known simple groups G*. As a first example, we take GLn(q), q odd, with x*
in diagonal form with k eigenvalues —1 and n — k eigenvalues + 1 . [It is easier for
expository purposes to make the calculations in GLn{q)\ to obtain the centralizer
in the simple group PSLn (q) of the image of x*, one must of course modify this
calculation slightly.] One computes that the elements of C* are the block matrices
of the form ( R I with A £ GLk{q) and B G GLn_fc(#), which implies that
C* <* GLk(q) x GLn„k(q).
In particular, we see that C* has at most 2 nonsolvable composition factors, which

if they exist are isomorphic to PSLk(q) and P5L n _/ c (g). This latter statement is
true as well for (2* = PSLn{q).
Similarly in the case of the alternating group An with x* the "short" involution
(12) (34), one computes that C* contains a normal subgroup GQ of index 2 of the
form
CQ = E4 x An-4.
Hence in this case, C* has at most 1 nonsolvable composition factor, which if it
exists is isomorphic to A n _4.
The following general result gives the structure of centralizers of involutions
in arbitrary known simple groups; it will be verified in later volumes dealing with
properties of DC-groups.
T H E O R E M 9.1. If G* is a simple X-group, x* is an involution in G*, and
C* = CG*(X*), then the following conditions hold:
(i) C* has at most 2 nonsolvable composition factors, unless G* is an orthogonal
group, in which case C* has at most 4 nonsolvable composition factors;
(ii) If G* is of Lie type of characteristic r, then each nonsolvable composition
factor of C* is of Lie type of characteristic r; and
(hi) 0(C*) is cyclic.
[The doubling factor in the orthogonal case arises because one of the 4-dimen-
sional orthogonal groups is itself a central product of two copies of SL2 (</)•]
Since the automorphism group of a cyclic group is abelian, the cyclicity of
0(C*) yields the ^ - p r o p e r t y as an important corollary.
COROLLARY 9.2. If C* is the centralizer of an involution x* in one of the
known simple groups G*, then
L2,(C*) = E(C*).
Other local subgroups of the known simple groups similarly are restricted in
their structures. Indeed, many of the local subgroups X of G* crucial for the
classification have one of the two following general forms:
(1) X has a subgroup XQ of very small index (usually 1 or 2) such that the
components of F*(X) are normal subgroups of XQ and XQ/F*(X) is abelian;
or
(2) F*(X) = R is a p-group for some prime p (so that X/F*(X) acts faithfully
by conjugation on the elementary abelian p-group R/$(R)).
In (1), XQ has structure analogous to that of a closed connected reductive
subgroup of a linear algebraic group; in (2), X is analogous to a parabolic subgroup
of a linear algebraic group.
Given the extremely strong implications that the isomorphism between the DC-
proper simple group G and the known simple group G* has for the local structure
of G, it is reasonable to ask how likely it will ever be to establish an isomorphism
between G and G* without first forcing critical portions of the subgroup structure
of G to closely approximate those of the target group G*.
However this obviously rhetorical question is eventually resolved, at present the
only known method for achieving the desired isomorphism is by means of a very
long and detailed investigation of the local subgroups of G and their interrelations.
That analysis depends upon a combination of general results of finite group the-
ory coupled with a veritable "dictionary" of properties of DC-groups. As has been
indicated, only after this analysis has forced G to have a local subgroup structure
approximating that of one of the target groups G* has it been possible until now
to produce a presentation for G identical to one for G*.
10. I N T E R N A L ANALYSIS: PASSING F R O M G L O B A L T O L O C A L I N F O R M A T I O N 29
10. Internal analysis: passing from global t o local information

How is one to use the global hypothesis that G is simple to analyze the local
structure of the minimal counterexample G to the classification theorem? The
negative flavor of the hypothesis—G has no nontrivial quotients—makes it difficult
to use. The initial observation that G is a X-proper group is vital but insufficient
in itself. Beyond that, the following obvious consequences of simplicity affect the
local structure of G by means of some important techniques which we shall discuss.
Here S(G) is the largest solvable normal subgroup of G.
(1) G has no proper abelian quotient, i.e., G = [G, G];

(10.1) (2) 5(G) - 1; and
(3) E{G) has no proper direct factor.
Before discussing the techniques to exploit these, we make two digressions. First
we note that in hindsight these three properties form a complete characterization
of simplicity, in the sense that a DC-group G (and hence an arbitrary group) is
simple if and only if it satisfies (10.1). Indeed, conditions (2) and (3) imply that if
we set K = F*(G), then K = E(G) is simple, so that G/K embeds in the outer
automorphism group Out(K) of K. As noted previously, it has been verified that
every simple DC-group has the Schreier property, so G/K is solvable. However, G/K
is perfect by (1), so G = K is a simple group.
Second, we briefly recall the effects of the conditions analogous to (10.1) in
another theory—the Killing-Cart an classification of finite-dimensional simple Lie
algebras over the complex field. If Q is such an algebra, then Cartan's criterion
translates semisimplicity, the analogue of (2), into the nondegeneracy of the Killing
form K. When K is restricted to a Cartan subalgebra f), the resulting form K^
is shown to be nondegenerate and the analysis of the root system begins. One
consequence which is analogous to the B^-property is that for any x G f), the cen-
tralizer 3 fl (#) is reductive, so that it is the direct sum of simple algebras (including
1-dimensional ones). On the other hand, the power of Cartan's criterion makes
the analogue of (1) superfluous in the presence of (2), since it forces a semisimple
algebra to be the direct sum of simple nonabelian ideals. Finally, the analogue of
(3), that Q is indecomposable, implies that the root system is indecomposable.
Now we return to the techniques for exploiting (10.1). The oldest of these,
the theory of transfer and fusion, shows how for each prime p, the structure of the
largest abelian p-factor group of G is completely determined by the structure and
embedding of the normalizers of certain nontrivial subgroups of a fixed Sylow p~
subgroup of G. Accordingly, (10.1)(1) forces restrictions on those local subgroups.
Results of this kind were obtained by Frobenius and Burnside; for example, Burn-
side proved that if G is simple then \G\ is divisible by 4 or p3 for some odd prime
p. The subject has been developed steadily by many authors through this century,
more recently with the notable theories of Alperin and Yoshida [All, Y l ] . In
the original classification proof as well as in our revision, however, transfer comes
into play only in low rank situations and in situations where considerable local
information about G has already been amassed.
By contrast, exploiting the condition 5(G) = 1 is a highly technical and delicate
matter, and central to our problem. The principal method for exploiting it has
been the signalizer functor method pioneered by Gorenstein and Walter [G2,
G W 3 , GW4]. It can be attempted whenever G contains an elementary abelian

p-subgroup A whose p-rank 10 mp(A) is at least 3. The subgroup A may be viewed
as analogous to a Cartan subalgebra of a Lie algebra, although the analogy is weak
in some respects. Under certain conditions, one can associate to each a € A# an
A-invariant p'-subgroup 6{CG{a)) of CG(a) satisfying the "balance" condition
0(CG(a)) H CG(b) = 0(CG{b)) n CG{a) for all a, b e A*.
The mapping 0 is then called an A-signalizer functor on £?, and there are various
signalizer functor theorems asserting that under certain hypotheses, the closure
0(G; A) = (6(CG(a)) \ a G A#) is ap'-group. For example, if 0(CG(a)) is a DC-group
for all a G A#, a theorem of McBride [McB2] gives this conclusion. Furthermore,
in many situations it can also be shown that 6{G\ A), or some appropriately defined
subgroup of it, is normal in G. By the simplicity of G, that subgroup is trivial, which
in turn yields consequences for the structure of CG(a), a G A#, because of the way
9 was originally defined. When the method succeeds, the result generally is that for
every a G A#, CG(a) is "reductive" in the sense that Lp'(CG(a)) < E(CG(a)), or
equivalently every p-component of CG(a) is actually a component. This is a partial
I?p-property, just for the elements of A. [The signalizer functor method really uses
the condition Op>{G) = 1 rather than S(G) — 1. However, this distinction is minor,
and indeed when p = 2, S(G) = 1 implies 0(G) = 1 by the Odd Order Theorem.]
The signalizer functor method cannot be undertaken when G has p-rank at
most 2 for all primes p. In this case and in other low-rank situations, the Bender
method is often useful (e.g,. [Be2, Be5]). It analyzes the structure and embed-
ding of certain maximal subgroups of G and their generalized Fitting subgroups—
particularly those maximal subgroups containing the centralizer of an involution.
For example this method underlies an elegant proof of Burnside's theorem that
groups of order paqb are solvable, not using the theory of characters as Burnside
did (for example, see [Sul]). Until the 1960's no such proof was known.
A further vital way to exploit S(G) = 1 is to use Glauberman's Z*-theorem
[G12], which can be rephrased in the following way. If 0(G) — 1 and an involution
z G G lies in Z(N) for certain 2-local subgroups N olG containing z, then z G Z(G).
Again, in our simple group G, Z(G) — 0(G) = 1 (by the Feit-Thompson theorem),
so the Z*-theorem gives information about various 2-local subgroups N.
The third condition (10.1) (3), that E(G) has only one component, is exploited
primarily through p-component uniqueness theorems, developed by Asch-
bacher and others [A4, G i l , SI, P o T h l ] for p = 2 and to be generalized by
us in [II2] to arbitrary primes p for 3C-proper simple groups. When successfully
combined with the signalizer functor method, these theorems often produce an ele-
ment x of G of order p and a component L of E(CG(x)) such that CG(L) has very
small p-rank, and such that x has a G-conjugate normalizing but not centralizing
L.
These p-component uniqueness theorems, as well as certain parts of the signal-
izer functor method and the Bender method, rely ultimately on the construction of
a strongly embedded subgroup and then on the classification theorem due to
Bender and Suzuki [Be3, Su4] of finite groups with such a subgroup. By definition,
10
T h e p-rank mp(X) of a group X is the largest nonnegative integer n such t h a t X contains
an elementary abelian subgroup of order pn.
11. IDENTIFYING SIMPLE GROUPS 31
the subgroup H of G is strongly embedded in G if and only if H < G, H has even

order, and H D Hx has odd order for all x G G — H.
The Bender-Suzuki theorem asserts that if G possesses a strongly embedded
subgroup and a subgroup isomorphic to E±, then G has a unique composition factor
L of even order, and L is a group of Lie type of Lie rank 1 and characteristic 2.
The importance of the Bender-Suzuki theorem resides in the following fact.
Analysis of a group by its p-local subgroups is most effective when G is generated
by a set of p-local subgroups having a large p-subgroup in common; but is ineffective
at the other extreme, when for some Sylow p-subgroup T of G and some subgroup
H < G, H contains NQ{Q) for every nonidentity p-subgroup Q of T. When this
occurs for p — 2, however, H is strongly embedded in G (as is readily seen from the
definition) and the Bender-Suzuki theorem provides a critical step. Indeed, such
"strongly p-embedded" subgroups H are encountered for odd primes p as well, but
when p is odd their existence can actually be shown in crucial situations to lead to
the existence of strongly embedded subgroups.
Thus the Bender-Suzuki theorem is a pillar of the classification proof. Its
proof is deep and difficult, and like all of the techniques for exploiting simplicity
of the group G, it eventually relies on the basic property that every nontrivial
homomorphism on G has trivial kernel. In particular, it relies on character theory,
that is, analysis of the linear representations of the group in question. Likewise
the proofs of some other basic results, such as the Feit-Thompson theorem and
the Z*-theorem, rely on representation theory, usually in combination with local
analysis.
Other homomorphisms on G, such as transfer homomorphisms or permutation
representations, can be regarded as internal or external to G, depending on one's
point of view. To the extent that linear representations are used, one cannot call the
classification proof "internal." But as already observed, the uses of representation
theory are limited to a few well-defined situations, such as the theorems mentioned
above and a part of the analysis of groups of 2-rank two. Such situations are
even more limited in our revised strategy than in the original proof. Beyond those
situations, the classification proof is truly internal, an intricate study of the local
subgroups of G.
11. Identifying simple groups

The final stage of every classification theorem involves an identification of the
group G under investigation with some known simple group G* by means of a set
of intrinsic conditions which characterize G*. A description of the various sets of
conditions for achieving this recognition appears in [G3]. Here we present only a
summary of that discussion.
There are three primary methods of identifying the simple groups:
A. By presentations (i.e., by generators and relations).

B. By the action of the groups on a suitable geometry.
C. By a primitive permutation representation.
For example, the symmetric group E n is generated by the transpositions x\ =

(12), #2 = ( 2 3 ) , . . . , xn-i = (n — 1, n), which satisfy the conditions
[ rij = 1 if i = j ,
rij
(11.1) (xiXj) = 1, where < rij = 2 if \j — i\ > 1, and
r»j = 3 if \j - i\ = 1.
Moreover, it can be shown that every relation among the x^s is a consequence of
the relations (11.1), so that by definition the x^s together with the relations (11.1)
provide a presentation for E n .
This means that if G is an arbitrary group with generators x'x,x'2,. • • , # n - i
satisfying the corresponding relations (11.1), then the mapping X{ H-> X \ , 1 < i <
n — 1, extends to a homomorphism of E n onto G. Thus the existence of the given
presentation for E n in terms of its generating involutions #i,£2, • • • ?#n-i enables
one to identify G as a symmetric group. A slight variation of the conditions (11.1)
provides a similar presentation for alternating groups.
Every group G(q) of Lie type, where q = p n , p a prime, has a distinguished
presentation [Stl, St4] called the Steinberg presentation, expressed in terms of
relations (the "Steinberg relations") among elements of so-called root subgroups
Xa These Xa are j9-groups.
In particular, when G(q) is an untwisted group, the Xa are isomorphic to the
additive group of Fq and so can be described parametrically as
xa = {xa(t)\teFq},
with xa(t + u) = xa(t)xa(u). Furthermore, the indices a run over the roots of
the indecomposable root system E of the associated complex finite-dimensional
Lie algebra. There is a (non-canonical) ordering < on E, which in particular splits
E into sets of positive and negative roots E + and E~, respectively, so that E is
the disjoint union E + U E~ and E~ = —E + . The indecomposable elements of E +
form a fundamental system II for E. The Dynkin diagram has nodes labelled
by the elements of II; it encodes the geometry of II, which in turn determines E
completely. In particular, II is a basis of HE. (When G(q) is a twisted group,
the Xa have more complicated structures and E is a possibly "non-reduced" root
system.)
The Lie rank 11 of G(q) is defined as the rank of E, which is the dimension
of the ambient Euclidean space RE. Whenever the Lie rank is at least 2 the key
Steinberg relation, other than the relations defining the individual subgroups X a ,
is the Chevalley commutator formula. It applies to any linearly independent
a, (3 G E and to each xa € Xa, xp G Xp. These relations have the following form:
(11.2) [xa,xp] = Y[X7
11
The Lie rank is sometimes also called the t w i s t ed Lie rank. There is a second notion of Lie
rank, sometimes called the u n t w i s t e d Lie rank; the two notions coincide for the untwisted groups.
The untwisted Lie rank of a twisted group G(q) is the subscript in the Lie notation for G(q), or
equivalently the Lie rank of the ambient algebraic group; it is the Lie rank of the untwisted group
which was twisted to form G{q).
11. IDENTIFYING SIMPLE GROUPS 33
Here 7 runs over all elements of E of the form 7 = ia+j/3, with i and j are positive;
the x>y are suitable elements of X1 whose parameters are given explicitly in terms
of those of X& and xp, and the order of the product is given by <.
For groups of Lie rank at least 2, the conditions (11.2) actually provide a
presentation not of the simple group G(q), but of its universal version G(q),
which is a central extension of G(q). In all but a few cases G(q) is actually the
universal covering group of G(q). Thus it follows that an abstract group G which
contains subgroups Xa isomorphic to Xa and indexed by the given root system
E, and satisfying the corresponding relations (11.2) is necessarily isomorphic to a
homomorphic image of G(q); if G is simple, then G = G(q).
It is the latter fact that explains the procedure for identifying the abstract
simple group as a given group G(q) of Lie type. Sufficient information about the
subgroup structure of G must be established to enable one to specify a prime p and
to associate with G a root system E of type G(q) and p-subgroups Xa of G that
behave like the root subgroups of G{q) relative to the root system E—i.e., which
satisfy the corresponding relations of (11.2).
[In practice, the analysis leads to the construction of a subgroup
G0 = (Xa I a e E)
of G isomorphic to the target group G(q) or to a homomorphic image of G(q). There
remains the entirely separate problem of proving that G = Go- This remark applies
equally well to the case in which the target group is an alternating group. This
final step is generally accomplished with the aid of the Bender-Suzuki theorem.]
The Xa (and likewise the Xa) satisfy many conditions that are consequences
of the Steinberg relations. Thus
U = (Xa I a e E + ) and V = (Xa | a e E">
are each Sylow p-subgroups of G(q). Also for any a £ E, (Xa, X-a) is a rank one
group of Lie type; in the untwisted case,
(Xa,X_a)^SL2(q) or PSL2(q).
Furthermore, a Cartan subgroup H of G(q) can be expressed in terms of the
X a ' s . In particular, in the untwisted case if one sets
na(t) = x a ( t ) x _ a ( t _ 1 ) x a ( t ) and ha(t) = n a ( t ) n a ( l ) _ 1 , t e F*,
then
H = (ha{t)\aeZ,teF*).
The group H is abelian and isomorphic to the factor group of the direct product
of m (= rank of E) copies of JPgx by a subgroup isomorphic to Z(G(q)).
In addition, in general
NG(q)(U) = UH and NG{q)(V) = VH, with UnH = VC)H=l.
Likewise the Weyl group W of E is recoverable from the Steinberg relations.
Again in the untwisted case if one sets
N=(na(t)\aeE,teFqx),
one has
H<N and N/H^W.
Here W is a finite Coxeter group and thus a finite group generated by reflec-
tions, i.e., W is generated by m involutions w^ 1 < i < m, where m is the rank
of E, and a presentation for W is afforded by wi,... ,wm and a set of relations
analogous to (11.1), but with the r^ allowed to take values other than 1,2, and 3.
In particular, the symmetric group £ n + i is the Weyl group of An(q) = PSLn+\{q).
We note also that W acts as a group of isometries of £.
N is often called a monomial subgroup of G(q).
We call the subgroup
R= ( n a ( l ) | a e E )
a reduced monomial subgroup of G(q). If q is even, R = W, while if q is odd,
R/E = W, where E = R n H is a normal elementary abelian 2-subgroup of R. In
either case we have
N = HR.
Also R induces a group of permutations on the set £. Analogous results hold for
the twisted groups.
Finally, it follows that G(q) has the Bruhat decomposition
G(q) = BNB,
where B = NG^(U) = UH\ B is a Borel subgroup of G(q).
Indeed, Tits has used the subgroups B, TV, and W of G(q) to give an alternative
description and characterization of the groups of Lie type based on the notion of a
(B,JV)-pair.
DEFINITION 11.1. A group G is said to be a (B, iV)-pair or have a Tits sys-
t e m of rank m if and only if G has subgroups B and N such that:
(i) G=(B,N)-
(ii) H :== B n N is normal in JV;
(iii) W := N/H is generated by involutions wi,..., w m ; and
(iv) If Vi is a representative of Wi in iV, then for each v e N and every i,
v B ^ C £?;£ U BwiB and i^Bu* g 5 .
Moreover, G is said to be split if B = (B D N)U = HU, where U is a normal
nilpotent subgroup of B.
A group G(q) of Lie type is a split (f?, iV)-pair with respect to the subgroups
B, N, H mentioned above.
All simple split (2?, iV)-pairs G (with G finite) have been determined, which
enables one to identify G (more precisely, a particular subgroup G0 of G) as a
group of Lie type by means of its generation by such a pair of subgroups B and
N. When the rank of G is at least 3, this is a consequence of a geometric result of
Tits, classifying suitable buildings [Ti4]. Tits's result can be viewed as a broad
generalization of the fact that for n > 3, any finite n-dimensional projective space
is n-dimensional projective space over Fq, for some q. Partial extensions of Tits's
results have been obtained for (B, iV)-pairs of rank 2, but the complete classification
of split (B, iV)-pairs of ranks 1 and 2 has been obtained only by permitting group-
theoretic and generator-relation type arguments 12 .
Curtis and Tits [Cul, Ti2] have obtained a more efficient method of identifying
the groups of Lie type of rank > 3. In effect, their results assert that a suitable
12
For rank 2, the reference is [FoSel]. For rank 1, the much longer story culminates in the
papers [ H e K a S e l , S h i , B o l ] and is told in [G4] and [Su6].
12. A C A P S U L E S U M M A R Y O F THIS S E R I E S 35
subset of the Steinberg relations (11.2) imply the complete set of relations (11.2).
The vital relations are between two rank 1 subgroups, each generated by root
subgroups corresponding to a fundamental root a G II and its negative. In the
untwisted case, for example, these rank 1 subgroups are isomorphic to SL,2(q) or
PSL2(q), and in practice, it is often easier to verify these (P)SL2(q) conditions
than the full set of Steinberg relations, particularly when q is odd. Other authors
have investigated analogous presentations of certain groups of Lie type by other
types of rank 1 subgroups, which in the classical cases are stabilizers of nonsingular
subspaces on the natural module, for example [Phi, P h 2 , W o l , D a l ] . Again the
relevant relations concern pairs of such subgroups.
Yet a third procedure for establishing the Steinberg relations has been used by
Gilman and Griess in the case q = 2 n [GiGrl]. Their result is expressed in terms
of the structure and embedding of a component K in the centralizer of a suitable
element u of odd prime order p of G(q) (usually, u lies in a Cartan subgroup H)
and a reduced monomial subgroup R of G{q) (R is isomorphic to the Weyl group
of G(q), as q = 2 n ). Under certain compatibility assumptions between K and R,
the Steinberg relations for G(q) are shown to be determined solely from these two
subgroups.
In characterizations of groups G(2) of Lie type over 2*2? Finkelstein, Prohardt
and Solomon used either the Curtis-Tits theorem or a variant of the Gilman-Griess
theorem [FinFrl, F i n S l, FinS2, FinS3].
The three approaches—by (B, iV)-pairs, by generation by rank 1 subgroups,
and by a component and Weyl group, are closely linked, as one would expect. For
simplicity of exposition, we shall refer here to verification of the Steinberg relations
without regard to the particular method used to substantiate them.
Finally, identifications of the sporadic groups have been achieved by a variety
of methods: as highly transitive permutation groups, as primitive rank 3 permu-
tation groups, as suitable primitive permutation groups of rank > 3, as groups of
automorphisms of the Leech lattice, as groups of automorphisms of suitable alge-
bras, as groups of matrices over suitable fields, and more recently in terms of their
2-local geometries. The identification of such a group G as a primitive permuta-
tion group has sometimes required computer calculations to force the uniqueness
of its generating permutations. In the existing classification proof, each sporadic
group has also been characterized in terms of the structure of the centralizers of
its involutions, this information sufficing to yield the above more basic uniqueness
conditions.
12. A capsule summary of this series

At this point it is appropriate to give a cursory view of our overall plan for
this series. This view will, for the purpose of brevity, be somewhat imprecise; the
reader must refer to subsequent sections and the succeeding Outline chapter for a
more precise and detailed description as well as a discussion of the ways in which
it differs from the original proof of the classification.
Let G then be a DC-proper simple group. We must prove that G = G* for some
known simple group G*. For this we make a broad case division depending on
the internal structure of G. In the "generic case" there is a prime p such that G
possesses an elementary abelian p-subgroup T of rank at least 3 (or 4 if p > 2), an
element x G T # and a component L of CQ{X) = CG(X)/Op>(GG(X)) such that L
is "p-generic". This means that L is neither a group of Lie type in characteristic

p, nor a member of a specified family of other small troublesome groups including
among others almost all groups of p-rank 1. This case leads to the conclusion
G = G* where G* is a group of Lie type in characteristic other than p and of
sufficiently high Lie rank or G* is a sufficiently large alternating group. Indeed, if
L is chosen as large as possible, then the isomorphism type of L already determines
the isomorphism type of the target group G* to within a few possibilities.
In the generic case, the subgroup T is analogous to a toral subalgebra of a Lie al-
gebra. The basic idea in this case is (1) to use the signalizer functor method to show
that for all g G T # , CG(g) is "reductive" in the sense that E(CG(g)) = E(CG{g)),
where Cc(g) = CG{g)/Op>(CG{g)) and bars denote reduction modulo O P / ( C G ( # ) ) ;
(2) to use the Steinberg presentation (or a presentation of the alternating groups)
to identify the isomorphism type of G(T) = (E(CG{9)) I 9 € T#) as that of a group
of Lie type or an alternating group; and (3) to prove that G = G(T), eventually by
using the Bender-Suzuki theorem on groups with a strongly embedded subgroup.
We carry out essentially this strategy in Part III. It should be pointed out that we
use the prime p = 2 if possible. Because of this preference, our overall design is con-
structed so that the generic case is encountered for an odd prime p only when it is
also known that G is of even type 1 3 . As a consequence, in this case G turns out to
be a group of Lie type in characteristic 2, as opposed to some other characteristic.
Now consider the case where G is not generic. One extreme case which might
occur is that G is what might be called "quasi-unipotent", that is, for every prime
p and element g G G of order p, CG{9)/OP'(CG(9)) has no component at all. This
is a very difficult case; by contrast, the roughly analogous case of finite-dimensional
complex Lie algebras without semisimple elements is handled elegantly by Engel's
theorem. One prototype of a quasi-unipotent group is a simple split ( 5 , iV)-pair of
rank one, and the recognition of such (£?, iV)-pairs (G = ^2(9)5 Q > 3, Us(q), q > 2,
Sz(22nJrl) or 2 G 2 ( 3 2 n + 1 ) , n > 0) is one of the most difficult chapters in the current
classification proof. (Actually 2 G f 2(3 2n+1 ) is not quasi-unipotent.) Likewise the
Feit-Thompson Odd Order Theorem may be regarded as another hard chapter in
the classification of quasi-unipotent groups. We have no design for improving these
two important results, and take those theorems—whose proofs have already been
improved and streamlined by others—as assumptions for our work. Our entire set
of assumed Background Results will be discussed in Sections 15-18.
Given these assumptions, we may now describe our strategy for the non-generic
case, or the special case, as we call it. First, assume that there exists an elemen-
tary abelian 2-subgroup T of G of rank at least 3, an involution x G T, and a
component L of CQ{X) = CG(X)/0(CG(X)) such that L is not a Chevalley group
in characteristic 2, nor one of a small finite number of specified characteristic 2-like
groups. As L is not 2-generic, L has one of several "thin" isomorphism types of
small 2-rank, and the principal case here is L = L2(q) or SL2(q), q odd. It is con-
venient to include with this analysis all cases in which G has 2-rank 2, regardless
of the structure of centralizers of involutions of G. (By the Feit-Thompson and
Z*-theorems, any nonabelian simple G has 2-rank at least 2.)
13
The definition of "even type" is given in Section 21. A sufficient condition for G to be of
even type is for G to be of characteristic 2-type, that is, 1712(G) > 3 and F*(N) = C>2(N) for
every 2-local subgroup N of G.
12. A CAPSULE SUMMARY OF THIS SERIES 37
The lengthy analysis of this case comprises Part IV of our work. The principal
outcome is to recognize G as a split (B, iV)-pair of rank 1 or 2 over a field of odd
characteristic, although there are a few further possibilities. In the case of (f?, N)-
pairs of rank 2, so much additional information is obtained en route that the final
recognition of G is fairly easy.
Second, we may now assume that G has elementary abelian 2-subgroups T of
order 8, and that for any such T, any x G T # , and any component L of CG(£) =
CG(t)/0(CG(t)), L is a Chevalley group in characteristic 2 or one of the finite list
of characteristic 2-like groups referred to above. In this case the signalizer functor
method yields that 0(CG(X)) = 1 for every involution x, and then that G is of even
type.
Since we are in the special case, if E is an elementary abelian p-subgroup of
£?, p odd, and mp{E) > 4, and if g G E# and L is a component of CG(9) =
CG(9)I'Op'(CG(^)), then L is not p-generic. A refinement of this statement is
established in Part V, ruling out in addition the existence of such L of "p-thin"
type, and as a result, the only possible isomorphism type for such a component L
is a Chevalley group in characteristic p or one of finitely many characteristic p-like
groups.
The analysis of this case hinges on the existence of 2-local subgroups with large
elementary abelian subgroups of odd order. The following set is basic for the case
subdivision.
DEFINITION 12.1.
cr(G) = {p \p is an odd prime, and for some 2-local
subgroup N of G, \G : N\2 < 2 and mp(N) > 4 }.
We consider the cases

(a) a(G) = 0 (the "quasithin case")
(b) <T(G) i=- 0 (the "large sporadic case").
This "revised" quasithin problem is more general than the "classical" quasithin
problem suggested by Thompson and originally considered by Aschbacher and Ma-
son [AlO, M a i , A18], in which it was assumed that for any 2-local subgroup N of
(7, F*(N) = 02{N) and mp(N) < 2 for all odd primes p. Our assumptions are less
restrictive; though G has even type, it may be that F*(N) ^ 02(N); and we only
assume that mp(N) < 3, and for that matter, only for certain 2-local subgroups
AT.14
The "revised" quasithin problem is currently being investigated by the amalgam
method 15 . Stellmacher and Delgado have made inroads; but considerable parts of
the problem remain to be done. Nevertheless, as noted in the Introduction, we are
reasonably confident that this approach will succeed, and if necessary, one could
14
T h e rather technical restriction on the size of a Sylow 2-subgroup of N in the definition of
cr(G) is chiefly for the sake of facilitating the amalgam method in the "revised" quasithin case. But
for that, one could tak e the definition to include the condition tha t N contain a Sylow 2-subgroup
of G.
15
T h e suitability of the amalgam method is evident in Stellmacher's reworking of the original
"thin" problem mp(N) < 1 for all odd p and all 2-locals N [Ste2]. Over the last fifteen years,
the method has steadily been strengthened by the work of Gomi, Hayashi, Rowley, Stellmacher,
Stroth, Tanaka, Timmesfeld and others.
reassemble theorems from a number of papers from the original proof 16 to cover
the "revised" quasithin case. In Part II, we supply the first step of the analysis.
It is a "global C(G, T)-Theorem", proving that a Sylow 2-subgroup of G lies in at
least two maximal 2-local subgroups of G. This result, obtained in collaboration
with Richard Foote, is one of several "uniqueness" theorems comprising Part II.
In the large sporadic case, the signalizer functor method yields one of two pos-
sibilities for each p 6 &{G)\ either G satisfies an analogue of the even type condition
for the prime p, or G has a maximal 2-local subgroup Mp with p-uniqueness prop-
erties; among other things, Mp contains a Sylow p-subgroup of G and contains
NG(Q) for every noncyclic p-subgroup Q of Mp.
In the first case, close analysis of 2-local and p-local subgroups following the
ideas of Klinger and Mason [KMal] leads to the conclusion p = 3 and G = G*,
where G* is one of the sporadic groups Fi, F2, Fi'2±, Fi2s, Fi225 Goi, or one of six
groups of Lie type defined over the field of 2 or 3 elements. The analysis of this
case is in Part V.
The remaining case is that for each p e &(G), G has a p-uniqueness subgroup
Mp as described above. This uniqueness case is shown to lead to a contradiction
in a chapter in Part II written by Gemot Stroth, using techniques of the amalgam
method.
As indicated above, we have collected in Part II a variety of uniqueness theo-
rems which provide the underpinnings for the subsequent argument. In addition to
the global C((7, T)-theorem and Stroth's uniqueness theorem already mentioned,
there is a treatment of ^component uniqueness theorems; a proof for DC-proper
simple groups of the Bender-Suzuki strongly embedded subgroup theorem, quoting
Suzuki's work on split (B, iV)-pairs of rank 1 (see [PI]); several related theorems
about involutions and 2-local subgroups; and some more technical results needed
in Part V connecting uniqueness theorems for p = 2 and odd p.
13. The existing classification proof

Over the thirty-year intensely active period in which the classification of the
finite simple groups was achieved, varied techniques were gradually developed for
studying their local structure, beginning with Brauer's theory of blocks of characters
and the Brauer-Suzuki theory of exceptional characters. The Feit-Thompson Odd
Order Theorem and Thompson's subsequent determination of simple groups each of
whose local subgroups is solvable (known as iV-groups) provided the initial impetus
for what was to become an elaborate theory of local group-theoretic analysis.
At the core of that theory were two basic methods that provided effective
extensions of Feit and Thompson's ideas to a general setting. The Bender method
enabled one to study the structure of maximal subgroups containing the centralizer
of a given involution, while the Gorenstein-Walter signalizer functor method and
the related theory of balanced groups provided a general technique for studying
centralizers of involutions (and later centralizers of elements of arbitrary prime
order) in simple groups. Two results of Glauberman—his fundamental Z*-theorem
for investigating the fusion of involutions and his Z J-theorem for analyzing p-local
structure—became critical auxiliary tools for applications of either method, as did
16
principally [AGL1, A 1 3 , M a i , A 1 8 , A10], together with solutions of certain involution
standard form problems
13. THE EXISTING CLASSIFICATION PROOF 39
several basic results of Aschbacher, Gilman and Solomon dealing with terminal and
2-terminal 2-components of the centralizers of involutions.
Bender's classification of groups G with a strongly embedded subgroup H and
its various extensions by Aschbacher were built upon Suzuki's earlier investigations
of doubly transitive groups in which a one-point stabilizer contains a regular normal
subgroup. These results provided critical underpinnings for all subsequent broad
local group-theoretic analyses. Aschbacher's theory and applications of groups G
with a tightly embedded subgroup H (i.e., a proper subgroup H of G of even
order that intersects each of its distinct (^-conjugates in a group of odd order)
provided the basis for investigating arbitrary simple groups in which the centralizer
of some involution is not 2-constrained.
At the same time, Fischer's general theory of groups generated by a conjugacy
class of transpositions, which had its origins in basic properties of transpositions in
symmetric groups, not only led to the discovery of seven sporadic simple groups,
but also was expanded and developed by Timmesfeld to reach important charac-
terizations of the groups of Lie type of characteristic 2 by internal properties.
During the final years, three important closely related techniques were intro-
duced for the analysis of groups targeted as groups of Lie type of characteristic 2:
(1) the Baumann-Glauberman-Niles theory of pushing up, emerging from a study
of maximal 2-local subgroups of even index in the given group G; (2) Aschbacher's
theory of %-blocks, an extension of properties of 2-components to certain config-
urations of 2-constrained subgroups; and (3) Goldschmidt's theory of amalgams,
dealing with the structure of two subgroups H, K sharing a Sylow 2-subgroup but
with C>2((H,K)) = 1. The last of these had its beginnings in a result of Sims on
primitive permutation groups in which a point stabilizer has an orbit of length 3.
Despite the many duplications and false starts alluded to in the Introduction,
a strategy for classifying the simple groups seemed to evolve inexorably as we
proceeded step by step from one partial classification result to the next, each time
constructing a platform from which to jump off to a greater level of generality. One
can already discern a pattern from four of the first major classification theorems to
be established:
1. The Feit-Thompson proof of the solvability of groups of odd order [FT1].
2. Thompson's classification of simple iV-groups [T2].
3. The Alperin-Brauer-Gorenstein-Walter classification of simple groups con-
taining no E8 subgroups [ABG2, A B G 1 , G W 1 , LI].
4. The Gorenstein-Harada classification of simple groups containing no sections
isomorphic to £32 [GH1].
The Odd Order Theorem showed that every (nonabelian) simple group must
have even order and hence necessarily must contain involutions. The AT-group
theorem demonstrated the power and the potential of local group-theoretic analysis
for treating broad classification theorems. The "2-rank < 2" theorem disposed
once and for all of the "smallest" simple groups. The inductive family of groups of
"sectional 2-rank < 4" had been introduced [MacWl] as a means of treating the
noninductive problem of determing the simple groups with a nonconnected Sylow
2-subgroup (a group S is said to be connected if any two four-subgroups17 of S
can be included in a chain of four-subgroups of S in which every two successive
A four-group is a group isomorphic to Z2 x Zi.

members centralize each other). At the time, nonconnected Sylow 2-subgroups

presented a serious barrier to effective application of the signalizer functor method.
Together these four theorems put group theorists in a position to mount a
full-scale attack on (connected) simple groups in which some involution has a non-
2-constrained centralizer. The 2-rank < 2 and sectional 2-rank < 4 theorems had
already made evident the primary obstructing role of 2'-cores in preventing central-
izers of involutions from approximating those in one of the known target groups.
Attention therefore became focused on establishing the ^ - p r o p e r t y in arbitrary
simple groups, this being the operational meaning of the removal of 2/-core ob-
struction. This turned out to be a very difficult undertaking requiring many years
and involving the combined efforts of a considerable number of group theorists (see
[G4, Wa2]). However, once it was achieved, Aschbacher's component theorem
[A4] then directly yielded the desired approximation of some non-2-constrained
centralizer of an involution to that in one of the target groups, and the path was
now set for the complete classification of all such simple groups. But these cen-
tralizers possessed such a variety of distinct shapes that it required more than fifty
separate, often highly technical analyses over a number of years to finish this major
subproblem of the project. In fact, some of these analyses were among the last
papers written as part of the complete classification of the finite simple groups.
The success of this attack meant that in a minimal counterexample G every 2-
local subgroup is 2-constrained. This represented a watershed on the path towards
a full classification of the finite simple groups, for it meant that the only remaining
known targets G* for G were the groups of Lie type of characteristic 2 and a few
sporadic groups. Since solvable groups are 2-constrained, N-groups necessarily
satisfy the specified conditions; and it turned out that Thompson's proof of the
Af-group theorem provided a framework for the analysis that lay ahead.
First, direct applications of signalizer functor theory eliminated all 2/-cores, so
that F*(H) = 02(H) for every 2-local subgroup H of G, and hence by definition of
the term the minimal counterexample G to the classification theorem was a group
of characteristic 2 type. Just as in Thompson's A/'-group analysis, so the analysis of
the general characteristic 2 type group can be split into the following three major
subcases:
(1) For some maximal 2-local subgroup H of G, 02(H) contains no noncyclic
characteristic abelian subgroups;
(2) No 2-local subgroup of G contains a Epz-subgroup for any odd prime p;
(3) Some 2-local subgroup contains a Epz -subgroup for some odd prime p,
with the solutions of (2) and (3) presupposing the solution of (1). For (1) the target
groups G* are most of the groups of Lie type defined over the prime field JP2 and a
few characteristic 2-type sporadic groups; for (2) (excluding the results of (1)) they
are the groups of Lie type of characteristic 2 and low Lie rank (plus the remaining
characteristic 2-type sporadic groups); and for (3) (again excluding the results of
(1)) they are the "generic" groups of Lie type of characteristic 2.
Despite the complexity of each of these classification problems, the analysis
proceeded much more quickly than in earlier stages of the classification proof, for
most of the necessary tools were in place and finite group theorists had already had
considerable practice in their use. Fischer's theory of transpositions was the key
underlying technique for investigating the "GF(2)-type" groups, the general case
treated by Timmesfeld [Tim3, T i m l ] , with important subcases treated in the work
14. SIMPLIFYING THE CLASSIFICATION PROOF 41
of Aschbacher, S. Smith, F. Smith, and others [A7, A 8 , S m l , Sm2, Sm3, Smil].

Likewise Aschbacher treated the major "thin" subcase of the "quasithin" group
problem, while G. Mason, in work eventually completed by Aschbacher, extended
Aschbacher's analysis to the much more difficult general quasithin problem [A10,
M a i , A18]. The theories of x-blocks, pushing up and amalgams were all critical for
treating certain minimal configurations that arose in the course of these analyses.
The "generic" characteristic 2-type case involved a study of the centralizers in
G of elements of odd prime order by means of the signalizer functor method, very
similar in spirit to the treatment accorded centralizers of involutions in the generic
but non-2-constrained case. The analysis of this case was primarily the work of
Aschbacher, Gorenstein and Lyons, and Gilman and Griess [A13, A G L l , GL1,
GiGrl].
Finally there was a rich interplay between the developing classification proof
and the discovery and construction of the sporadic groups during the 1960's and
1970's, which provided an added dimension to the entire endeavor.
14. Simplifying the classification proof

As a result of the evolutionary character of the existing classification proof,
the focus of attention remained localized until the final years. Because a complete
classification of the simple groups was viewed as too vast to be considered as a single
problem and its attainment well beyond reach, individual facets were almost always
investigated without reference to their global implications. In particular, theorems
were formulated in such a way that they could be established independently of the
full classification theorem. Even within these theorems, individual cases were often
treated as independent results.
For example, a result of Alperin asserts that a simple group of 2-rank < 2 has a
Sylow 2-subgroup that is either quaternion, dihedral, semidihedral, wreathed (-2^ I
Z2), homocyclic abelian, or isomorphic to a Sylow 2-subgroup of PSUs(4). In the
original classification proof, each of these five cases was given an entirely separate
treatment. Verification of the ^ - p r o p e r t y constitutes perhaps the most extreme
illustration of this phenomenon. The general case was covered by Aschbacher's
"classical involution theorem"; however, treatment of the various residual cases
spread over nearly a dozen additional difficult papers (see [A9], Wa2).
Beyond this, classification results established in the early years were of course
derived without benefit of techniques developed later. Under all these circum-
stances, it was hardly suprising that the complete classification proof would be
inefficient and accessible only to experts already versed in its intricacies. Whether
new techniques can be developed or a new approach discovered for drastically sim-
plifying the classification theorem remains to be seen. The only question there is
any hope of settling at present is the following: 75 it possible with existing group-
theoretic techniques to construct a coherent, efficient, and accessible proof of the
classification of the finite simple groups? It is the purpose of this series of mono-
graphs to provide an affirmative answer to this question.
Even a superficial examination of the present proof will reveal many areas
capable of consolidation and simplification. Consider, for example, the construction
of the sporadic groups. Originally nineteen of the twenty-one groups other than the
five Mathieu groups were constructed by separate means. However, the Fischer-
Griess monster F\ is known to contain twenty of the twenty-six sporadic groups as
either subgroups or sections. Construction of F\ depends only on the existence of

Conway's largest group Co\, and construction of Co\ only on the existence of the
largest Mathieu group M24. Hence once M24, C01, and F\ have been constructed,
one obtains the existence of 17 additional sporadic groups as a corollary.
Although centralizers of involutions and centralizers of elements of odd prime
order p were each investigated by the signalizer functor method, there are consid-
erable differences in the strategies that were employed. For example, the odd p
analysis was carried out on the basis of only a partial B p -property and also without
any analogue of the sectional 2-rank < 4 theorem for dealing with nonconnected
Sylow p-subgroups. Then, too, in the involution case, the identification of G as a
group of Lie type of odd characteristic was established by means of its "intrinsic
SL,2(q) 2-components" via Aschbacher's classical involution theorem, whereas in
the odd p case, the identification of G as a group of Lie type of characteristic 2 was
achieved via a terminal p-component and a "proper neighbor." It is therefore nat-
ural to ask whether this entire analysis can be carried out uniformly for all primes
simultaneously.
As a still further illustration, we note that in the present proof many of the
groups of Lie type of characteristic 2 are identified three or four times, from each
of the following points of view:
1. Prom suitable terminal 2-components in the centralizers of involutions. (The

only target groups arising here have non-inner automorphisms of order 2.)
2. As groups of GF(2)-type (only certain groups over F2 arise here).
3. Prom Timmesfeld's "root involution" theorem [Tim2].
4. From suitable terminal p-components in the centralizers of elements of odd
prime order p [GiGrl] .
[Almost all of the groups G* of Lie type of characteristic 2 are generated by a

conjugacy class C* of involutions with the property that the order of the product
of any two elements of C* is 1, 2, 4 or odd and in addition, if the product has order
4, then its square is an element of C*. The class C* consists of r o o t elements in the
Lie sense. Timmesfeld's root involution theorem gives a complete determination
of all simple groups generated by a conjugacy class of involutions satisfying these
conditions.]
One can thus ask whether such duplication of effort is necessary—equivalently,
whether it is possible to organize the classification proof in such a way that each
known target group is identified just once in the course of the analysis. Our aim
has been to provide an affirmative answer to this question as well.
Considered from the most general perspective, the overall strategy of our "sec-
ond generation" classification proof is essentially the same as that of the original
proof: a division of groups into "odd" and "even" types, special treatments of the
low rank groups of each type, and a systematic treatment of terminal p-component
problems for centralizers of prime order p.
However, by treating the classification theorem as a single result rather than
as a collection of more or less independent subsidiary theorems, as had been nec-
essary in the original proof, we are able to effect a considerable simplification and
consolidation in its organizational structure. In particular, this approach allows
us to make more systematic use of the %-group assumption on proper subgroups.
Moreover, it leads to significant differences in the way the component parts of the
14. SIMPLIFYING THE CLASSIFICATION PROOF 43
two proofs are "packaged," the effect of which is to enable us to bypass a number
of important subsidiary theorems of the present proof.
Indeed, the proof we shall present here does not directly involve the general
form of any of the following results that were used in the existing classification
proof:
1. Sectional 2-rank < 4 classification theorem [GHl].
2. Classical involution theorem [A9].
3. Solutions of a large number of the terminal 2-component centralizer of in-
volution problems (see [Sel; Se2, pp. 47-53]).
4. Structure of a Sylow 2-group of a tightly embedded subgroup [ASel, Gr-
M a S e l , A6].
5. G?F(2)-type classification theorem [Tim3, Sm3].
6. Root involution theorem [Tim2].
7. Strongly closed abelian 2-subgroup classification theorem [Go5].
[Goldschmidt has generalized the Z*-theorem by determining the simple groups
G in which a Sylow 2-subgroup S contains a nontrivial strongly closed abelian
subgroup A, that is, A9 n S < A for all g e G.]
In the previous paragraph, the terms "directly involve" and "general form"
have been carefully chosen, for it should be clear that any classification of the
simple groups based upon presently available techniques will encounter many of
the same critical configurations. Nor is it reasonable to expect that the proof can
be so organized that all such configurations can be bypassed. Indeed, we are forced
to consider many that were involved in the original proof.
In particular, that is the case for a number of configurations arising in the
sectional 2-rank < 4 analysis. In some cases, our present approach yields a modest
simplification, but in others the initial treatment seems to be optimal. The proof
of the classical involution theorem enters in an even more significant way, for our
analysis of centrahzers of involutions and more generally of elements of prime order
p in the case of generic simple groups G utilizes "3/2-balanced" signalizer functors.
These objects, developed by Goldschmidt and Aschbacher [A9, Go4], formed the
basis for Aschbacher's elimination of 2'-core obstruction in the classical involution
context.
Furthermore, the bulk of Goldschmidt's proof of the strongly closed abelian
theorem is taken up with elimination of 2'-core obstruction in the centrahzers of
involutions. However, we require the result only in the case that the 2'-core of the
centralizer of every involution is assumed to be trivial and therefore need only the
remaining small part of his argument. On the other hand it is precisely the Bender
method for eliminating 2'-core obstruction, as elaborated by Goldschmidt, which
plays a crucial role in the analysis of special groups of odd type. Thus much of
Goldschmidt's proof appears in our work in cannibalized form.
Just as we use the Bender method more extensively than the original proof
did, we should also point out that the proposed treatment of the revised quasithin
problem as well as Stroth's new elimination of the "uniqueness case" for odd primes
makes considerably greater use of the Goldschmidt amalgam method than occurred
in the original proof, for the full power of the method was not developed until
after the classification theorem had been completed. The original treatments of
the quasithin and uniqueness cases used the method of weak closures, invented by
Thompson in the iV-group paper [T2, A14, A16, M a i ] .
Finally, we do not wish to leave the reader with the impression that our ap-
proach is without its own liabilities. One of these occurs precisely because of our
strategy of bypassing many terminal 2-component centralizer of involution prob-
lems. Indeed, although this results in a considerable net saving, it forces us to
permit non-2-constrained 2-locals to occur in situations that in the prior analysis
had been restricted to 2-constrained 2-locals, thereby increasing the number of con-
figurations to be investigated. To be sure, there are also the liabilities that come
with an induction hypothesis as massive as the one we use. Not only does a math-
ematical argument hang by an inductive thread for thousands of pages, but also
a vast and sometimes tedious theory of X-groups is required. We can hope that
future revisionists will find a tidier and more robust proof; for now we have taken
what seems to us the most direct approach.
D. T H E BACKGROUND RESULTS
15. Foundational material

This brings us to a further aspect of the existing classification proof. Because
of its extreme length—both in toto and in many of the individual papers—finite
group theorists were strongly motivated to quote available results whenever possi-
ble. With a few notable exceptions, little effort was made to search for self-contained
arguments. As a consequence, the papers making up the classification proof, to-
gether with the papers to which they refer, together with those to which they refer,
and so on, form a web whose ultimate foundations are difficult to trace. They
include not only group-theoretic results from the turn of the century, but also early
papers in geometry and number theory. Thus not only is the existing proof itself of
inordinate length, but the foundational material on which it is based is presently
in an unorganized state.
Any simplified proof of the classification theorem must also deal with this foun-
dational problem. Clearly for a project of this magnitude ever to reach completion,
some body of group-theoretic material must be accepted at the outset and permit-
ted to be quoted as needed. On the other hand, it is essential that this foundational
background material be made explicit if the resulting proof is to be placed on a
firm base.
Furthermore, as we have already observed, the existing proof depends on a very
large number of properties of DC-groups and especially simple dC-groups. Some of
these are very general, but many involve detailed properties of individual simple
groups or families of simple groups. Similarly the classification theorem requires a
great many properties of general finite groups, ranging from the very broad to the
very specialized.
Indeed, the opening sections of a paper on the classification of the simple groups
typically consist of the verification of those 3C-group and general group-theoretic
properties required for the classification theorem under consideration. For exam-
ple, the 55-page first installment of Thompson's iV-group paper [T2] is largely
devoted to establishing such results, in this case primarily properties of solvable
and nilpotent groups. The most extreme illustration of this phenomenon occurs
in the Gorenstein-Lyons work on groups of characteristic 2 type [GL1], in which
15. F O U N D A T I O N A L M A T E R I A L 45
the first Part (pages 15-425) is taken up essentially solely with the development of
the various X-group and general group-theoretic results needed for the subsequent
analysis. In particular, it includes (without proofs) lists of local properties of each
of the twenty-six sporadic groups [GLl, Part I, §5] (many of which were worked
out for us by O'Nan).
It seems inevitable that any classification of the finite simple groups based on
an internal analysis of the subgroup structure of a minimal counterexample will
require a vast body of such preliminary results. This is therefore as much a feature
of our revised proof as it is of the original one.
One of the most difficult decisions the authors have had to make in formulating
an overall strategy has concerned these preliminary properties of 3C-groups and of
arbitrary finite groups:
(A) What should we be allowed to quote?
(B) What should we be required to prove?
(C) How should the material be organized?
Ideally, our preference would have been the following:
(I) Quote only results available in standard books, monographs, or lecture
notes.
(II) Establish the needed X-group and general finite group properties on the
basis solely of results available in (I).
(Ill) To achieve maximum coherence and efficiency of development, defer orga-
nization of the material in (II) until the revised proof has been completed.
Unfortunately, we have reluctantly concluded that this strategy would delay
publication of any portion of our revision beyond the foreseeable future. Indeed, it
would force us to include several additional volumes pertaining to this preliminary
material.
The most serious problem concerns the sporadic groups, whose development at
the time of the completion of the classification theorem was far from satisfactory.
The existence and uniqueness of the sporadic groups and the development of their
properties form a very elaborate chapter of simple group theory, spread across a
large number of journal articles. Moreover, some of the results are unpublished (e.g.
Sims's computer calculations establishing the existence and uniqueness of the Lyons
group Ly). Furthermore, until very recently, the two principal sources for properties
of the sporadic groups were [ C C N P W l ] and [GLl, Part I, §5] consisting only of
statements of results without proofs.
However, over the past few years attempts to rectify this situation have been
begun by a number of authors, initially focusing on existence and uniqueness of the
sporadic groups. Moreover, Aschbacher [A2] has recently begun a more systematic
development of the sporadic groups which illustrates the verification of many of
their properties as well as existence and uniqueness.
One faces a somewhat different problem regarding the finite groups of Lie type.
Carter's two books and Steinberg's lecture notes [ C a l , C a 2, S t l ] provide an
excellent treatment of many of their general properties, but they do not cover
in sufficient detail certain bread-and-butter material required for the classification
proof, such as the structure of the centralizers of semisimple elements, questions
about balance and generation, or Schur multipliers. The Schur multipliers are
treated in the general case by Steinberg [Stl, St4 , St5], with a number of residual
cases computed by several authors, most of them by Griess [Grl] (see also [Gr2]).
Although most of the standard repertoire of finite group theory needed for local
group-theoretic analysis can be found in books by Gorenstein, Huppert, Blackburn,
Suzuki, and Aschbacher [Gl, H u l , H u B l , S u l , A l ] , nevertheless many impor-
tant local results exist only in the journal literature. A similar situation prevails
regarding representation theory and Brauer's theory of blocks of characters. Most
but not all of the needed background material is covered in standard references
such as those of Feit and Isaacs [ F l , I s l ] .
Thus a systematic treatment of all this preliminary material would require
volumes concerning (a) the sporadic groups, (b) the finite groups of Lie type, and
(c) basic local group theory and character theory. However, we prefer to focus our
efforts on the classification proof per se, for it is here that our major revisions and
simplifications are to be made. We have therefore decided not to proceed in quite
so ambitious a fashion.
16. The initial assumptions

On the other hand, to follow the opposite path with respect to this large body
of preliminary material, allowing ourselves to quote the literature at will, would
defeat one of our primary objectives, which is to provide as self-contained a proof
of the classification theorem as we can feasibly achieve.
We have therefore settled on a middle course—on the one hand practicable,
and on the other, consistent with this objective. We have decided to operate on the
basis of the following two principles. First, in addition to the material contained
in certain published books, monographs, and lecture notes, listed below in section
17, we shall assume a number of further basic results contained in a selected list of
papers to be specified in section 18. This list will be comparatively short; moreover,
within those papers we shall quote only basic results.
Second, regarding the sporadic groups, we shall proceed essentially as in [GL1],
permitting ourselves to quote any of their properties in tables like those of [GL1,
Part I, §5] and to be given in [IA]> In addition, we shall make an assumption con-
cerning their existence and uniqueness. As observed in section 1, a variety of paths
has been followed for constructing and identifying the sporadic groups. However,
an identification of each group has been achieved in terms of the centralizers of its
involutions. To avoid listing the large number of papers on which this assertion is
based, we shall simply assume here that 1 8
the twenty-six sporadic groups exist, and each of them (except Mu)
(16.1) is uniquely determined up to isomorphism as the only simple
group with its centralizer of involution pattern.
[In the case of M u , one must add its order to the given conditions to distinguish
it from L 3 (3). Because of this disparity with the other sporadic groups, we prefer
to identify Mu in the text from the given conditions.]
18
By definition, two groups G and G* have the same centralizer of involution pattern if and
only if there is an isomorphism x *—*• x* from a Sylow 2-subgroup S of G to a Sylow 2-subgroup
S* of G* such that for all involutions x and y in 5, the following two conditions hold: (a) x and
y are G-conjugate if and only if x* and y* are G*-conjugate; and (b) CG(X) =• CQ* (#*)• If only
condition (a) is imposed, G and G* are said to have the same involution fusion pattern.
17. B A C K G R O U N D R E S U L T S : BASIC M A T E R I A L 47
Aschbacher's recent volume goes some distance toward organizing the proof of
(16.1), by constructing the twenty sporadic groups involved in the Monster JFI and
proving uniqueness for five of these groups from conditions even weaker than cen-
tralizer of involution patterns. It also documents some of the background properties
we assume.
The complete list of our assumed material will, for brevity, be referred to as
the Background Results, and is an implicit hypothesis of the main theorem.
This material consists, of course, of proved results; we use the term "hypothesis"
here solely to emphasize the "rules of the game" under which we plan to operate.
Thus, subject to these specific background results, all proofs are intended to be
self-contained, including not only the classification proof proper but all further
necessary supporting properties of X-groups and general finite groups. That is, the
only results we may quote either must come from the Background Results or else
must have been previously established in the course of the proof.
To summarize: the Background Results consist of (16.1) and the references
listed in the next two sections. These references, upon which our proofs depend,
are also listed in a separate bibliography as the Background References. All other
references, included only for the sake of exposition, are listed in a bibliography of
Expository References, and this practice will be followed in all the monographs.
17. Background Results: basic material

We now list the basic books and monographs (including lecture notes and mem-
oirs) from which we shall be permitted to quote in the course of the classification
proof. The aim is to keep this reference list as short as feasible, consistent with
covering the needed background material.
(1) General group theory
[Al] M. Aschbacher, Finite group theory, Cambridge University Press, Cam-
bridge, 1986.
[Gl] D. Gorenstein, Finite groups, Second Edition, Chelsea, New York, 1980.
[Hul] B. Huppert, Endliche Gruppen I, Springer-Verlag, Berlin, New York,
1967.
[HuBl] B. Huppert and N. Blackburn, Finite groups III, Springer-Verlag, Berlin,
New York, 1985.
[Sul] M. Suzuki, Group Theory I, II, Springer-Verlag, Berlin, New York, 1982,
1986.
(2) Character theory
[Fl] W. Feit, The Representation Theory of Finite Groups, North-Holland,
Amsterdam, 1982.
[Isl] I. M. Isaacs, Character theory of finite groups, Academic Press, New
York, 1976.
(3) Groups of Lie type
[Cal] R. Carter, Simple groups of Lie type, Wiley-Interscience, New York,
1972.
[Dl] J. Dieudonne, La geometrie des groupes classiques, Springer-Verlag,
Berlin, 1955.
[Stl] R. Steinberg, Lectures on Chevalley groups, Lecture Notes, Yale Uni-
versity, 1967-68.
[St2] R. Steinberg, Endomorphisms of linear algebraic groups, Mem. Amer.

Math. Soc. 80 (1968).
(4) Sporadic groups
[A2] M. Aschbacher, Sporadic Groups, Cambridge University Press, Cam-
bridge, 1994.
The principal reference for general group theory is [Gl] and to a lesser extent
[Al]; the remaining listed books, and the character theory references, will be re-
ferred to only for certain specific results. For example, the subgroups of PSL2(q)
are determined in [Hul], the Bender-Thompson odd prime signalizer lemma is
proved in [HuBl], and Glauberman's Z*-theorem in [Fl]. The material from the
Lie type references relevant to finite groups will be used freely. Finally, since we are
taking (16.1) as given, we use [A2] only for certain structural properties of sporadic
groups.
The final reference to be listed will be the source for considerable local group
theory and for many properties of the groups of Lie type as well as for structural
assertions concerning the sporadic groups not covered in [A2].
[GL1] D. Gorenstein and R. Lyons, The local structure of finite groups of
characteristic 2 type (Part I, pp. 15-425), Mem. A. M. S. 276 (1983).
Some comments concerning our use of [GL1] are in order, for it is only by
means of this memoir that we are able to keep the list of background references as
compact as it is. For example, our source for the Schur multipliers of the simple
groups of Lie type will be found in the tables on page 72 of [GL1]. However, the
validity of this table is based on the results of at least a dozen papers. This is the
most serious example of this phenomenon, though not the only one. On the other
hand, we supply our own proof of Seitz's fundamental theorem on the generation
of groups of Lie type [Se3, L3].
18. Background Results: revised

portions of t h e proof and selected papers
Revisions of certain portions of the classification proof have already appeared
in print. To avoid undue duplication of effort we shall include a selected list of such
references within our Background Results.
The most important such revisions concern the solvability of groups of odd
order and the recognition of split (B, iV)-pairs of rank 1.
A. Solvability of groups of odd order: Bender and Glauberman have revised
Chapter IV of the original Feit-Thompson proof [FT1] and Peterfalvi [P3] has
shortened Chapter VI. Both of these are included in:
[BeGll] H. Bender and G. Glauberman, Local Analysis for the Odd Order The-
orem, L. M. S. Lecture Notes Series 188, Cambridge Univ. Press, 1994.
In addition, Sibley has revised substantial parts of Chapter V, but his work is
not published. Hence for completeness, we include in the Background Results 19
[FT1] W. Feit and J. Thompson, Solvability of groups of odd order, Pacific J.
Math. 13 (1963), 775-1029,
but only for the material in Chapter V and the preceding material which supports
it, namely, Chapter III and a few elementary lemmas from Chapters I and II.
See the Preface to the Second Printing: this can be replaced by Peterfalvi's revision [P4].
BACKGROUND RESULTS: REVISED PORTIONS OF THE PROOF, SELECTED PAPERS 49
B. Recognition of split (B,N)-pairs of rank 1: Peterfalvi has revised both

Bender and Suzuki's original work [Be3, Su4] in the characteristic 2 case and
O'Nan's characterization of the unitary groups in odd characteristic [ONI, ON2],
and Enguehard has revised the corresponding characterizations by H. Ward, Janko-
Thompson, Thompson, and Bombieri [Warl, J T 1 , T l , B o l ] of the Ree groups
of characteristic 3. We only require Part II of Peterfalvi's work on the Bender-
Suzuki theorem, which is a treatment of Suzuki's original work; we shall establish
separately the main result of Part I under our X-group hypotheses 20 .
[El] M. Enguehard, Obstructions et p-groupes de classe 3; Caracterisation
des groupes de Ree, Asterisque 142-143 (1986), 3-139.
[PI] T. Peterfalvi, Le Theoreme de Bender-Suzuki (Part II only), Asterisque
142-143 (1986), 235-295.
[P2] T. Peterfalvi, Sur la caracterisation des groupes U^{q), pour q impair,
J. Algebra 141 (1991), 253-264.
For a complete treatment of the Ree group problem we must also include the
appendices to Bombieri's paper [Bol] by Hunt and Odlyzko. These independently
each contain the final step of the characterization of the Ree groups—reports of
machine computations eliminating a finite number of exceptional cases of fields of
characteristic 3 not included in Bombieri's or Enguehard's work.
[Odl] A. Odlyzko, The numerical verification of Thompson's identity, Invent.
Math. 58 (1980), 97-98.
[Hunl] D. Hunt, A check of the Ree group conjecture for small fields, Invent.
Math. 58 (1980), 99.
C. Signalizer functor theorems: The theory of signalizer functors, and in partic-

ular the signalizer functor theorems asserting the completeness of various signalizer
functors, play a crucial role in the proof of the classification. The solvable sig-
nalizer functor theorem was perfected in the papers of Goldschmidt, Bender, and
Glauberman [Go2, Go3, Be4, Gil], and building on those, the nonsolvable case
is covered by two papers of McBride [McBl, McB2]. For the solvable theorem,
there is a choice of references, and we choose the original paper of Glauberman; the
theorem is treated also in [Al].
In addition, the authors [GL3, GL4] have recently established a nonsolvable
signalizer functor theorem, which although less general than McBride's is consid-
erably easier to prove and is entirely sufficient for application in the classification
proof. Nevertheless, we have in the end decided to use McBride's papers as back-
ground; though his proofs are more difficult than ours, the statement of the result
itself is cleaner and stronger and its application requires no technical hypotheses to
be verified beyond certain simple properties of %-groups.
[Gil] G. Glauberman, On solvable signalizer functors on finite groups, Proc.
London Math. Soc. 33 (1976), 1-27.
[McBl] P. McBride, Near solvable signalizer functors on finite groups, J. Algebra
78 (1982), 181-214.
[McB2] P. McBride, Nonsolvable signalizer functors on finite groups, J. Algebra
78 (1982), 215-238.
20
I n Part II, Peterfalvi makes occasional reference to technical lemmas from Part I and we
include these implicitly in the above reference.
D . N e a r c o m p o n e n t s a n d local C(G, T ) - t h e o r e m : Underlying Aschbacher's

fundamental local C(G, T)-theorem is a basic result of Baumann, Glauberman,
and Niles [ B a l , G I N i l , N i l ] dealing with the following special configuration:
F*(G) = Op(G), p a prime, and G/Op(G) ^ SL2(pn). Their result gives the
precise structure of G when no nontrivial characteristic subgroup of a Sylow p-
subgroup T of G is normal in G. Using the amalgam method, Stellmacher has
obtained a considerably shorter proof of a version of the Baumann-Glauberman-
Niles theorem, and we include his paper as a background reference. In addition,
we have separately written a revised proof of the local C{G, T)-theorem, under a
X-group hypothesis appropriate for its applications here. We shall therefore also
take that proof as a background reference.
[Stel] B. Stellmacher, Pushing up, Arch. Math. 46 (1986), 8-17.
[GL2] D. Gorenstein and R. Lyons, On Aschbacher's local C(G,T) theorem,
Israel J. Math. 82 (1993), 227-279.
E. C h a r a c t e r t h e o r y : We have had some difficulty deciding how to handle the

character-theoretic portions of the classification proof not available in standard
texts. Character theory is used only in the analysis of simple groups of 2-rank at
most 3 apart from its use in establishing some of the Background Results, such as
the recognition of split (B, iV)-pairs of rank 1 or existence and uniqueness of certain
sporadic groups—primarily those having one conjugacy class of involutions. In the
end, we have decided to include complete proofs involving only ordinary character
theory (for example, the contents of the Bender-Glauberman paper [BeG12], which
provides the character-theoretic basis for the classification of groups with dihedral
Sylow 2-groups), but to leave the development of certain basic modular character-
theoretic material to the Background Results.
This material consists of the analysis of p-blocks with certain specified defect
groups, aimed at certain explicit order formulas for G in terms of the structure of
centralizers of involutions. The cases in question are the principal 2-block in groups
with semidihedral, wreathed or E$ Sylow 2-groups; blocks with defect group E4 and
inertial index 1; and the principal p-block (p odd) in case the Sylow p-subgroup is
abelian with automizer of order 2. Enguehard's revision of the Ree group problem
covers the Eg case. For the other cases, and just in connection with these specific
preliminary modular character-theory results, we include the following papers.
[Brl] R. Brauer, Some applications of the theory of blocks of characters of
finite groups III, J. Algebra 3 (1966), 225-255.
[Br2] R. Brauer, Some applications of the theory of blocks of characters of
finite groups IV, J. Algebra 17 (1971), 489-521.
[Br3] R. Brauer, Character theory of finite groups with wreathed Sylow 2-sub-
groups. J. Algebra 19 (1971), 547-592.
Finally, it may be necessary or desirable to add a few additional papers to
our background list as we proceed. This remark applies as well to possible further
JC-group references. For example, it is quite likely that verification of a few needed
results about sporadic groups will depend on basic properties not listed in [GL1,
Part I, §5], but which appear in the Conway-Curtis-Norton-Parker-Wilson Atlas of
Finite Groups [ C C N P W l j . If so, we shall include these further properties in our
Background Results and add the Atlas to our reference list.
19. CENTRALIZERS OF SEMISIMPLE ELEMENTS 51
E. SKETCH OF THE SIMPLIFIED P R O O F
19. Centralizers of semisimple elements

In 1954, Richard Brauer proposed that an inductive attack on the classification
of the finite simple groups might proceed by passing from an inductive knowl-
edge of the centralizers of involutions to a determination of the groups themselves.
Bolstered by the Odd Order Theorem, this philosophy prevailed until around 1974,
when in the final years of the classification program it became clear that the central
significance of the prime 2 resided primarily in the following facts.
Most finite simple groups are groups of Lie type defined over a field of odd
characteristic r. As such their most natural matrix representations are as matrix
groups over finite fields Fq of characteristic r. In such groups, elements of order 2 are
semisimple and indeed the largest rank among the semisimple abelian subgroups
is achieved, or very nearly achieved, by certain 2-subgroups. Moreover, if the group
has Lie rank at least 3, the centralizers of the (commuting semisimple) elements in
such a 2-group contain subnormal subgroups which are quasisimple or isomorphic to
5L2(3) and which together generate the entire group, and from which the Steinberg
presentation of the group can be recovered.
In groups of Lie type over fields of characteristic 2, it is elements of odd prime
order p that are semisimple, and if one chooses p appropriately (usually to divide
the order of a Cart an subgroup) and an appropriate abelian semisimple p-subgroup,
then the above assertions for commuting involutions carry over to commuting el-
ements of order p. In most of the alternating groups we shall regard elements of
order 2 as semisimple and proceed as in Lie type groups of odd characteristic to
obtain the Coxeter presentation for the groups.
From our point of view, then, the primary objects of study in the classifica-
tion of finite simple groups should be the centralizers of "semisimple" elements of
appropriate prime order p, and more precisely the components of these centralizers.
A group of Lie type is, of course, defined over some field, so at the outset one
knows that elements of order coprime to its characteristic — in particular, of prime
order dividing the order of a Cartan subgroup — are semisimple. On the other
hand, if G is an arbitrary finite simple group, even a minimal counterexample to
the classification theorem, then no underlying field is specified a priori, let alone
a Cartan subgroup, so that determination of the appropriate prime p on which to
focus attention is itself one of the central problems within the analysis.
There is the even more basic decision of whether to take p to be 2 or an odd
prime — i.e., whether the ultimate presentation for G is to emerge from a study
of the centralizers of involutions or from the centralizers of elements of odd prime
order. In general, we know in the abstract how we hope to make this decision:
namely, apart from low rank cases, if G is to turn out to be a group of Lie type of
odd characteristic or an alternating group, we want to take p = 2, whereas if it is to
turn out to be a group of Lie type of characteristic 2, we want to take p to be odd.
On the other hand, for the sporadic groups, low rank groups of Lie type and small
alternating groups, the situation is more complex but generally the identification of
these groups will proceed from 2-local data. What is required then are criteria on
the internal subgroup structure of G that will enable us to determine in advance
which of these possibilities occurs in a given situation (see section 21 below).
20. Uniqueness subgroups

As has been discussed earlier, centralizers of involutions (and more generally
of elements x of any prime order p) in an arbitrary dC-proper simple group may a
priori have a structure far more complex than that of such centralizers in any of the
known target groups G*. Moreover, a large part of the existing proof as well as of
our simplified classification proof is taken up with forcing an approximation of the
structure of these centralizers for appropriate p and x. The strategy for carrying
this out is quite involved, utilizing a number of distinct techniques, the appropriate
one depending upon the ultimate known target group G*.
Despite these differences, however, there is a common feature to all the meth-
ods, which we digress briefly to discuss. The existence of some obstruction to the
similarity of G and G* leads to the construction of a "large" maximal subgroup M
of G — large in the sense that M contains the normalizer in G of many nonidentity
p-subgroups of M, in some cases for a single prime p and in others for all primes p in
some set. Usually, these conditions equivalently assert that for each Q in some suit-
able collection of p-subgroups of M, M is the unique G-conjugate of M containing
Q. For this reason, such subgroups M are frequently called p-uniqueness subgroups
of G. In the classification proof at various points we construct such p-uniqueness
subgroups both for p = 2 and for all p in the set o~(G) of odd primes (Definition
12.1).
Harada has colorfully referred to p-uniqueness subgroups as "black holes," in
the sense that they ensnare all subgroups in their neighborhood. This is certainly
an apt description of their basic properties.
The primary example of a p-uniqueness subgroup is that of a strongly p-
embedded subgroup M , which by definition is a proper subgroup H of G of
order divisible by p such that HC\H9 is a p'-group for all g G G — H. For p = 2, this
is the usual notion of a strongly embedded subgroup. If M is strongly p-embedded
in G, then M = NG(M) and M is the unique G-conjugate of M containing any
nontrivial p-subgroup Q of M. It follows that NQ{Q) < M for each such Q. In
fact, more generally, any p-local subgroup of G containing Q lies in M.
We shall require a slight weakening of this notion for odd primes and introduce
the idea of a strong p-uniqueness subgroup. For the purposes of this chapter,
the reader should think of this as a strongly p-embedded subgroup. The precise
definition, only required for certain odd primes p, will be given in the next chapter.
Underlying the classification proof is a network of results related to uniqueness
subgroups, involving, on the one hand, a complete determination of the possibilities
for G when G is assumed to have a uniqueness subgroup of some type and on the
other, various sufficient conditions for G to possess a uniqueness subgroup.
The Bender-Suzuki theorem [Be3] is the most fundamental example of the
former. Because of our X-proper hypothesis we need only the following weaker
theorem which will be proved in [Hi]. In the form stated here, it incorporates the
Brauer-Suzuki theorem [BrSul, G14] on groups of 2-rank 1, a special case of the
Z*-theorem.
T H E O R E M 20.1. If G is a %-proper simple group containing a strongly embed-

ded subgroup M, then G ^ L2{2n), Sz{2n), or U3(2n), n>2, and M is a Borel
subgroup of G.
21. THE SETS £p(<3) AND GROUPS OF EVEN TYPE 53
An example of the latter type of uniqueness theorem is the following result con-
cerning component uniqueness subgroups, which extends to arbitrary primes
earlier work of Aschbacher, Gilman and Solomon for the prime 2 [A4, G i l , S i ] ,
and will be proved in [II2].
T H E O R E M 20.2. Let G be a X-proper simple group, p a prime, and M a max-
imal subgroup of G. Ifp is odd, assume that M has p-rank at least 4- Let K be
a component of M = M/Op>(M), of p-rank at least 2. If CG(X) < M for every
element x of order p in CM(K), then one of the following holds:
(i) K is normal in M; or
(ii) M is strongly p-embedded in G.
In particular, for p = 2, the theorem yields as a corollary the basic result that
terminal components of centralizers of involutions in G are necessarily standard in
£?, if they have 2-rank at least 2. By definition, a component K in the centralizer
of an involution of G is standard in G if and only if K is terminal and commutes
elementwise with none of its G-conjugates.
A third example of a uniqueness result is a particular case of the C(G,S)
theorem (for S € SyfaiG)), in which it is assumed that NG(S) is contained in a
unique maximal 2-local subgroup of G and various additional side conditions are
satisfied. As in the original C(G, S)-theorem, the classification of such groups G is
achieved by first arguing that G contains an Aschbacher x-block; the subsequent
analysis involves the study of groups containing such a %-block. It is the latter
investigation that has been carried out by Foote. Chapters 3 and 4 of Part II are
devoted to this result. The remainder of Part II is taken up with work on the
"uniqueness" case for groups of even type. This is in two parts. The first consists
of some theorems about 2-local properties of p-uniqueness subgroups, analogous to
[AGL1]; in particular it is shown that for p £ o~(G), if M is a strong p-uniqueness
subgroup of G, and Q is a noncyclic p-subgroup of M, then any 2-local subgroup of
G containing Q lies in M. The second, which uses these technical results, is Stroth's
proof that the uniqueness case does not occur, an analogue of Aschbacher's original
work [A16] but proved by a different approach.
Thus Part II: Uniqueness Theorems should be viewed as a collection of tool
theorems needed for the classification per se.
21. The sets &P(G) and groups of even type

We return to the discussion of section 19, which clearly indicates that in a
DC-proper simple group G the following sets will be of critical importance for the
analysis.
DEFINITION 21.1. For each prime divisor p of the order of G,
HP(G) = {K I K is a component of CG{X)/OP>{CG{X))
for some x G G o f order p}
Since G is DC-proper, every element of Lp{G) is a quasisimple DC-group — that

is, a perfect central extension of a known simple group — whose center is a p-group.
It is important for us to consider all isomorphism types of such groups:
DEFINITION 21.2. DC is the set of all quasisimple DC-groups, and
DCP = {K € DC I K is quasisimple and Op'(K) = 1}.
Thus, &P(G) C Xp. Of course, &P(G) may be empty for a given p. For example,
this will certainly be the case if the centralizer in G of every element of order p is
solvable.
The question of which elements of G will play the role of semisimple elements—
involutions or elements of odd prime order—can be almost settled by the nature of
the sets &P(G) for various primes p. Indeed, in a group G of Lie type of characteristic
r, the composition factors (if any) of the centralizers of semisimple elements are
themselves groups of Lie type of the same characteristic r, while the centralizers
of elements of order r are r-constrained and so have no r-components. Therefore
for any prime p, £JP(G) consists of groups of Lie type in the same characteristic
r, and is actually empty if p = r. On the other hand, apart from a few low
degree alternating groups and a few sporadic groups, standard components of the
centralizers of involutions in all other known simple groups are not groups of Lie
type of odd characteristic.
Thus, the existence of a single element of &2(G) that is a group of Lie type of
odd characteristic in our 3C-proper simple group G should in general be sufficient
to guarantee that G will turn out to be a group of Lie type of odd characteristic.
Hence when £>2(G) has a component of such an isomorphism type, we shall consider
involutions to be semisimple elements and attempt to recover a Steinberg presen-
tation for G by studying centralizers of involutions. As remarked earlier, we also
plan to consider involutions to be "semisimple" elements when the target group for
G is an alternating group. On the other hand, for most target sporadic groups G,
even though &2(G) ^ ®> w e do n ° t wish to consider involutions to be "semisimple"
elements of G.
To distinguish the two outcomes we divide the set %2 into two subsets: one
called 62, consisting of those elements of %2 which are groups of Lie type of char-
acteristic 2 as well as a few other "characteristic 2-like" groups, and the comple-
mentary set consisting of the "odd characteristic-like" groups, including sufficiently
large alternating groups. In an analogous way, for each prime p we define a subset
Gp of %p consisting of the groups of Lie type of characteristic p as well as a few
"characteristic p-like" groups 21 . These sets Cp are denned precisely in the next
chapter.
In this terminology, we shall thus regard involutions as semisimple elements
whenever some element of H>2{G) is not in 62-
There is a further situation in which involutions should be viewed as semisimple
elements. Indeed, the only known simple groups in which the centralizer C of some
involution has the property Oy (C) ^ 1 are certain alternating groups and groups of
Lie type of odd characteristic. Hence we also want to view involutions as semisimple
elements when 0 2 ' ( C G ( # ) ) 7^ 1 for some involution x of our DC-proper simple group
G.
Thus, only when the above two conditions fail is it appropriate to focus atten-
tion on centralizers of elements of odd prime order. Furthermore, the analysis of
groups of 2-rank at most 2 also concentrates on centralizers of involutions. This
leads to a fundamental division of all simple groups that is basic to our proof of
the classification theorem.
21
Actually, for p = 2 and 3, a handful of bizarre covering groups of groups of Lie type of
characteristic p are excluded from Gp.
22. GENERIC SIMPLE GROUPS AND NEIGHBORHOODS 55
DEFINITION 21.3. A 3C-proper simple group G is said to be of even typ e if

and only if
(1) Every element of £J2(G) lies in 62; and
(2) 02>{CG(x)) = 1 for every involution x of G\ and
(3) G has 2-rank at least 3. 2 2
This gives a purely internal set of conditions on G that enables one to decide
in advance whether to take p = 2 or p odd; namely, if G is not of even type, focus
on centralizers of involutions, while if G is of even type, focus on centralizers of
elements of odd prime order. In the latter case, however, the question of which
odd prime p to choose remains open. In fact the question is moot in the important
quasithin case (cr(G) = 0) where the analysis remains focused on 2-local subgroups
throughout. When cr(G) ^ 0, the choice of odd prime is resolved only at a later
stage of the analysis in terms of a suitable maximality condition on elements of
LP(G). [In groups of Lie type of characteristic 2 — the primary target groups when
G is of even type — the maximal p-rank for odd primes p is realized by primes
dividing the order of either a Cart an subgroup or one other closely related torus.]
It should be noted that if G is of even type with £J2(G) empty, then G is
of characteristic 2 type — i.e., F*(H) = 02(H) (equivalently, CH(02(H)) <
02(H)) for every 2-local subgroup H of G. Moreover, in the original classification
proof, it was the latter set of conditions which determined the point at which
one shifted attention from centralizers of involutions to centralizers of odd prime
order. The advantage of separating groups of even type from the others rather than
drawing the dividing line at groups of characteristic 2 type is that we are thereby
able to bypass a large number of technically difficult standard component problems.
Finally, we organize the global strategy for classifying groups of even type so
that, if possible, we only examine centralizers of elements x of odd prime order p
when these centralizers have p-rank > 4. We thus make the following definitions.
DEFINITION 21.4. For any prime p,
XP(G) = {x e G I x has order p},
T°(G) = {xe IP(G) I mp(CG(x)) > 4} if p is odd,
ZJ(G) = {xe I2(G) I m2(CG(x)) > 3}, and
L°p(G) = {L I L is a component of CG(x)/Opf(CG(x)) for some x <E Z°{G)}.
The definition of 12(G) has been made in this fashion for uniformity. In reality,
an application of the Thompson transfer lemma shows that if m2(G) > 3, then
m2(CG(x)) > 3 for every involution xeG. Thus 12{G) equals T2{G) ifm2(G) > 3,
and is empty otherwise. Likewise &2{G) = &2(G) if m2(G) > 3, and otherwise
£J2{G) is empty.
22. Generic simple groups and neighborhoods

As observed in section 19, a presentation of a group of Lie type of odd char-
acteristic and Lie rank at least 3 can be recovered from the components of the
centralizers of suitable commuting involutions, and a corresponding assertion holds
22
Condition (1) is the primary one; in fact (2) turns out to be redundant. Were it not for
pathologies in some small C2-groups, the redundancy of (2) would be easy to prove.
in characteristic 2 for components of the centralizers of commuting elements of

suitable odd prime order p.
In fact, the centralizers of families of commuting semisimple elements contain
redundant information, and in many cases the desired presentation is completely
determined from only two appropriately chosen commuting elements x, y of order
p and single components if, J in their respective centralizers with K n J a "large"
subgroup of both K and J. An analogous assertion holds for alternating groups
of sufficiently high degree (with p = 2) and leads to their classical generation by
involutions.
However, there are examples of distinct groups of Lie type that possess isomor-
phic pairs if, J. In such cases it is necessary to consider further elements u € (x,y)
and components L of their centralizers to distinguish the two groups. The resulting
configuration of components is the basis of the key notion of a neighborhood of
(x,K). To avoid technicalities, we shall not attempt to make the definition precise
here, but limit ourselves to two illustrative examples in which the two components
K, J suffice for the desired presentation. Further discussion of neighborhoods ap-
pears in sections 26 and 27 below.
First, consider G = SLn(q), q — r m , r a prime, with n > 4 and q > 3. If r
is odd, take p = 2, while if r = 2, take p to divide q — 1. For such a choice of p,
commuting elements of order p can be simultaneously diagonalized. Let us assume
for simplicity that (n, q— 1) = 1. Then G is simple, and there exist A, /i G F* with
A ^ / z , A/i n _ 1 = 1, and Xp = \xv = 1. The diagonal elements
/A \ (V-
M and y •
V
have order p and determinant 1 and so lie in G. Moreover, one computes that
CQ{X) and Cc{y) have respective components K and J satisfying the following
conditions:
(1) K^J^SLîq);
(2) I:=E(KnJ)^SLn-_2(q);
(3) CG{K) and CG(J) have cyclic Sylow p-subgroups; and
(4) / is contained in a single component of CQ{U) for all u £ (x,y).
A large part of a set of defining relations for G is already visible in the three
groups K, J, K fl J, and it is possible to recover the rest from corresponding data
in the centralizers of G-conjugates (indeed {K, J)-conjugates) of x and y. In this
situation, the 5-tuple (x, y, K, J, J) suffices to describe the neighborhood of (x, K).
The inclusion mappings of I into J and K are considered part of the information
contained in the 5-tuple.
Next, consider the case G = An, n > 13, and take x — (n — 3, n — 2)(n — 1, n)
and y = (12)(34), so that x and y are involutions of G. This time CQ{X) and Cc(y)
have respective components K and J satisfying the following conditions:
(1) K^JÂn-A;
(2) I = E(K H J) ^ An_8;
(3) CQ{K) and CQ{J) have E± Sylow 2-subgroups; and
(4) J is a component of CQ (xy).
22. G E N E R I C S I M P L E G R O U P S A N D N E I G H B O R H O O D S 57
Again the desired presentation for G is determined from K, J,K f) J and

corresponding information in the centralizers of G-conjugates of x and y. Here
(x, y, K, J, I) suffices to describe the neighborhood of (#, K).
Note that in the SLn(q) case, conditions (1), (2) and (3) also hold in the
wreath product G — SLn-i(q) I Zp with #, y suitable commuting elements of order
p in G that cycle the p SLn_i(q) components of G. But in that case, (x,y) D
E(G) 7^ 1, and for any u G (x, y) C\E(G), CG(U) contains a subgroup isomorphic to
SLn-2(q) I Zp, with I embedded diagonally, so condition (4) fails in this example.
Thus in the general definition an analogue of condition (4) is critical for avoiding
wreathed solutions for G, and is included in the definition of what we call a vertical
neighborhood.
It should also be clear that this approach will be effective only for sufficiently
large values of n. Indeed, if K PI J is too small (solvable, for example), there may
be insufficient interplay between the groups K and J to recover the presentation
for G. For SLn(q), we require n > 4 for K D J to have a component, while for An,
we require n > 13 to achieve this condition. In terms of the components K and
J, it means that correspondingly we must have K = SLk(q), k > 3, or Ak, k > 9.
Thus, the definition of a neighborhood of (x, K) must also include some minimal
size restrictions on the given components K and J .
What we have just described is the "end game" for reaching presentations of
the generic known simple groups. It provides a target for the global strategy of the
analysis of our DC-proper simple group G: generally, our aim will be to prove the
existence of a neighborhood in G of an appropriate (#, K), where x is an element
of G of suitable prime order p and K is a component of CG{X); and by means
of this neighborhood to demonstrate that G is isomorphic to the corresponding
known simple group. However, as the above discussion has indicated, this strategy
can be successful only if &P(G) includes "sufficiently large" groups not in Cp. For
example, if £>2(G) contains no alternating groups of degree k > 9, it will obviously
be impossible to construct an alternating type neighborhood within G.
This leads to a further fundamental division, to be made precise in the next
chapter, of the set of groups not in Cp into the set Sp of "p-generic groups", i.e.,
those that are large enough to eventually yield presentations for G via the existence
of appropriate neighborhoods, and the complementary set 7p of "p-thin groups".
Recall that for each prime p, %p is the set of all known quasisimple groups K with
Op'{K) = 1. Thus we actually have a tripartite partition
JVp — L»p LJ Jp <J jp.
Furthermore, groups in Cp, T p and Sp will often be called Cp-groups, Tp -groups,

and Sp-groups. For example, for p = 2, the definition will stipulate that all groups
of Lie type of odd characteristic are S2-groups other than SL2(q) and ^2(^)5 q
odd, 1,3(3), tr 3 (3), PSp4{3), 1,4(3), I74(3), 2174(3), 2 G 2 (3) / and G 2 (3). SL2{q) is an
OVgroup for all odd q and L2{q) is an T2-group except for q a Fermat or Mersenne
prime or 9, and the rest of the listed exceptions are (^-groups.
We can now define the basic notion of a group of Lp-generic type. We em-
phasize that it requires that the group be of even type if p is odd and not of even
type if p = 2. Recall also that -C^Ô — &2(G) or 0 according as m2(G) > 3 or not.
D E F I N I T I ON 22.1. Let p be a prime. We say that G is of £ p -generic type if
and only if one of the following holds:
(1) p = 2 and some element of ^(G) is a S2-group (whence by definition

™<2(G) > 3 and G is not of even type); or
(2) p is odd, G is of even type, and some element of &p(G) is a Sp-group.
Thus it is for the groups of £ p -generic type that the neighborhood strategy
described above is applicable. In case (2), p will in fact be further restricted to the
set cr(G) defined in Definition 12.1.
23. The main case division

We have now introduced the elements of the main logic of the proof of the
Classification Theorem. As the discussion of the preceding section indicates, the
proof divides into a "generic" case and the residual "special" cases. Moreover, our
arguments to classify £ p -generic groups work most smoothly when their p-rank is
large; and one alternative conclusion of the analysis is always that the group has
a strong p-uniqueness subgroup. 23 Accordingly the generic argument deals with a
X-proper simple group G and a prime p for which G has no strong p-uniqueness
subgroup and for which the p-rank is large in a suitable sense which we now define.
For any S £ SyîG) let Mi(S) be the set of all maximal 2-local subgroups M
of G such that \S : M n S\ < 2. Recall from Section 12 that a(G) is the set of
all odd primes p such that mp{M) > 4 for some M e M i ( 5 ) . This definition is of
interest only when G is of even type. It is obviously independent of 5, by Sylow's
theorem.
DEFINITION 23.1.
&o(G) = {p G c(G) \G has no strong p-uniqueness subgroup}.
The term "generic" is then made precise as follows:

DEFINITION 23.2. A DC-proper simple group G is generic, or of generic type,
if and only if one of the following holds:
(a) G is of /^-generic type;
(b) G is of even type 2 4 , CTQ(G) ^ 0, and G is of £ p -generic type for all p G cr0(G).
Finally, G is of special type if and only if it is not generic. (Of course this term
has nothing to do with the common notion of a "special" p-group.)
For the special cases we introduce the following terminology. We say that G
is of ^2-special type if and only if no element of ^(G) is a S2-group. For an
odd prime p, we say that G is of £ p -special type if and only if G is of even type
and no element of £ £ ( £) is a S p -group. Thus a group is either of ^-generic or
^-special type (but not both). Likewise, for any odd prime p, a group G of even
type is either of £ p -generic or £ p -special type, but not both.
Therefore in considering our DC-proper simple group G, there are four mutually
exclusive and exhaustive cases, as indicated by Table III. We call G of odd type
23
See Section 20.
24
We shall actually introduce in the next chapter a notion of "restricted" even type, a slight
and technical strengthening of the notion of even type; and our analysis will divide into cases
according as G is or is not of restricted even type. Thus in our definitions here "even type" should
be replaced by "restricted even type". However, for the expository purposes of this chapter we
shall ignore this minor distinction. For the case division for generic groups, moreover, there is no
particular advantage to the refined definition; it is of value only for special groups.
24. SPECIAL SIMPLE GROUPS 59
TABLE III: THE FOUR-PART CASE DIVISION
ODD TYPE EVEN TYPE
GENERIC G has a o ( G ) ^ 0 , and

TYPE £>2-generic type G has £ p -generic type for all p G 00(G)
(PART III) (PART III)
SPECIAL G has odd type and <T0(G) = 0, or

^2-special type G has £ -special type for some p G CTQ(G)
TYPE p
(PART IV) (PART V)
if and only if it is not of (restricted) even type. The four cases are to be considered
in the indicated parts of this series, with the two generic type cases unified to the
extent possible.
We emphasize that this entire analysis proceeds with the benefit of our unique-
ness results from Part II. For example, we may assume the following:
(1) G contains no strongly embedded subgroup.
(2) If G is of even type and a(G) ^ 0, then a 0 (G) ^ 0.
In addition to (1), there are other types of "uniqueness subgroups" which G is
forbidden to contain because of results in the first two chapters of Part II. Statement
(2) will be established in Chapters 5 and 6 of Part II.
24. Special simple groups

We now describe briefly the analysis of groups of special type. Although most
of the known simple groups fall within the generic category, unfortunately the
largest portion of the proof is devoted to the analysis of the special cases (which,
in particular, include all sporadic groups). In fact, it is necessary to make a care-
ful subdivision of these non-generic cases according to the method to be used for
investigating the subgroup structure of G and establishing an appropriate set of
internal conditions in terms of which G is ultimately to be identified.
Moreover, these internal conditions themselves differ widely from group to
group. To emphasize their importance in the analysis, we introduce the symbol
GnG*
to indicate the fact that the DC-proper simple group G under investigation satisfies
the same set of internal conditions as the known simple group G*.
Thus the meaning of " « " is dependent on the group G*. For instance, when
G is of generic type, G* is an alternating group or a group of Lie type and the
definition of G « G* requires G and G* to have isomorphic neighborhoods (see
Section 26 below). On the other hand, for a sporadic group G* it is precisely the
structure of the centralizers of its involutions together with the involution fusion
pattern that defines the term G « G* (so that once it is proved that G « G*, our
Background Results imply immediately that G = G*).
The point is that a presentation for each finite simple group, generic or special,
is to be derived via an appropriate set of conditions on its subgroup structure, these
forming the basis for the definition of the term G & G*. In all cases, the global
strategy of the classification proof is designed to force G to satisfy one of these

possible sets of conditions.
The analysis of groups of special type in fact splits into six major cases, three
for groups of odd type and three for groups of even type. To describe these, we
require some more terminology.
First, if S € Syl2(G)> then M(S) = M(G;S) is the set of maximal 2-local
subgroups of G containing NG(S). Next, the groups SL2(q)^ q odd, 2An, 7 <
n < 11, (the coverings of An by a center of order 2) and the coverings of 1/3(4)
by a center of exponent 4 are called ^ - g r o u p s ('"B" for Bender, since when these
groups occur the "Bender method" is used). With this terminology, the special
type case division is as follows.
Special odd type: G is of £,2-special type.

Case 1. G has odd order.
Case 2. Either G has 2-rank at most 3 or some element of ^(G) is a 32-group.
Case 3. G has 2-rank at least 4 and no element of /C2(Gf) is a 232-group.
Special even type: ao(G) — 0, or G is of <Cp-special type for some p G <To(G).
Case 4. a(G) is empty. (By the uniqueness result at the end of the last section, this
is equivalent to (JQ(G) = 0.)
Case 5. For some p e ao(G), every element of &p(G) is a Cp-group.
Case 6. Case 5 does not hold, but for some p £ cro(G), some element of &°(G) is a
Tp-group.
Likewise no fewer than six major methods enter into the analysis of these
various cases (some used in more than one case)—three limited to groups that are
of odd type, two to groups of even type, and one applicable to both types. We limit
ourselves to listing them by name, adding a few brief comments concerning where
they are used—a more detailed discussion appears in the next chapter.
Special odd type:

A. Character theory
B. Analysis of 2-fusion
C. The Bender method
Special even type:
D. The Goldschmidt amalgam method
E. The Klinger-Mason method
Both types:
F. &-balanced signalizer functors
Exceptional character theory is used in both the Odd Order Theorem and in
the analysis of Case 2; while block theory is used in Case 2.
In Case 2 an analysis of 2-fusion forces the possible isomorphism types of a
Sylow 2-subgroup of G as well as the involution fusion pattern in G.
The Bender method is used in Case 2 to eliminate 2'-core obstruction and is
also used in the revised version of the Odd Order Theorem to study the maximal
subgroups of G.
Case 4 is the quasithin case, or more distinctly the "revised" quasithin case.
Here the possible structures of the elements of M(5) are to be determined by the
25. STAGES OF THE PROOF 61
amalgam method. One of the uniqueness theorems of Part II allows us to conclude

that |3Vt(5)| > 1, so the method is indeed applicable.
In the even type case, once the quasithin case has been treated, one is reduced
to the case in which <Jo(G) ^ 0. Now G contains a 2-local subgroup of p-rank at
least 4 for every p G cr(G), so in particular Tp{G) ^ 0. It is for this reason that we
are able to restrict the analysis in Cases 5 and 6 (as well as in the £ p -generic type
case for odd p) to the study of centralizers of elements of T°(G).
The Klinger-Mason method [KMal, GL5] is the least known of the six listed
techniques. It is applicable in groups G that are simultaneously "characteristic
2-like" and "characteristic p-like" for some odd prime p, under the assumption that
G contains a "large" {2,p}-subgroup—thus, in particular, it is applicable in Case 5.
This is an interesting case, for p is forced to be 3 and the largest sporadic groups,
including the Monster, occur as solutions as well as a half-dozen moderately large
groups of Lie type over the fields of 2 and 3 elements. Furthermore, the local
analysis involves the structure of the centralizers of commuting elements of order
2 and 3, and by analogy with the notion of a neighborhood for the prime p these
form the basis of the definition of a ZQ X ^ - n e i g h b o r h o o d in G.
Finally, the signalizer functor method is used in both Cases 3 and 6 to eliminate
p'-core obstruction. It is also used in Case 2 to treat the subcase in which a Sylow
2-subgroup contains a nontrivial strongly closed abelian subgroup.
25. Stages of the proof

The analysis in each of the six special type cases just described as well as in
the generic type case involves several distinct stages, the total number varying from
case to case. These can be viewed as landmarks on the road to the identification
of G. It is our intention to make these critical junctures in the proof explicit, so
that a flow chart of the entire classification proof will appear in the next chapter
as a grid whose rows consist of the seven different cases and whose columns consist
of the conclusions of the separate stages of the analysis. An analogous flow chart
will be provided for the various uniqueness results which underlie the classification
proof.
In the next chapter, we describe the individual stages for the grids. Actually
the stages for the generic type case will be discussed as well in the next section.
Here as an illustration we give a sketchier description of one of the special type
cases. We consider the following subcase of Case 2 above:
1. Every element of &2(G) is either a C2-group or isomorphic to PSL,2(q) or
SZ/2(#), q odd; and
2. Some element of &2(G) is isomorphic to SL2(q) for some odd q.
In this situation the target groups G* are the following groups (here q = pn, p
an odd prime, with q > 3 except in the case of 3D4(q)):
G* = L3(q), U3(q), PSp4(q), G2(q), 3D4(q), or L\(q),q £ e(mod 8), e = ± 1 .
If we ignore certain groups defined over F3, then these families, together with
the groups £2(2) and 2G2{q), q odd, consist in fact precisely of those simple groups
of Lie type of odd characteristic in which no element of H>2(G) is a S2-group. Note,
too, that they include the groups L±(q), q ^ 1 mod 8, which are of Lie rank 3.
On the other hand, when G = L±(q), q = \ mod 8, some element of ^{G) is
62 P A RT I, C H A P T E R 1: O V E R V I E W
isomorphic to L3(q), which is a S 2 -group. This illustrates how delicately one must
make the divisions among the various cases.
At the other end of the spectrum, the groups U^(q) are the only ones on the list
of Lie rank 1. Because groups of Lie rank 1 have a much tighter internal structure
than do groups of higher Lie rank, the configurations leading to the Us (q) solutions
require very lengthy analysis.
All results stated below are derived using the uniqueness results of Part II. In
particular, we may assume that G does not contain a strongly embedded subgroup.
Stage 1. On the basis of an analysis of 2-fusion in G, establish the existence of a
target group G* on the above list with the following properties:
1. A Sylow 2-subgroup S of G is isomorphic to one of G*;
2. G and G* have the same involution fusion pattern; and
3. If z is an involution of Z(S) and C = CG(Z), then the product of the com-
ponents of C/02'{C) is approximately the same as that in G* = G G * ( Z * ),
z* a 2-central involution of G*.
It is during this stage that signalizer functor theory is used to eliminate a par-
ticular configuration in C involving a nontrivial strongly closed abelian 2-subgroup.
Stage 2. Use the Bender method to study a maximal subgroup M of G containing
C and prove that one of the following holds:
1. G and G* have 2-rank at least 3, C has the ^ - p r o p e r t y 2 5 , and 02'(G) is
cyclic of order dividing |C>2'(Cr*)|;
2. M = G, G* = Us(q), and C has the £?2-property;
3. M > G, G* = L3(q), q = pn, p a prime, and Op (M) is isomorphic to
QP (M*) for some maximal parabolic subgroup M* of G*.
In case (3), we have shown at this point that both the centralizer of an involu-
tion and a maximal p-local subgroup of G approximate the corresponding subgroups
in the groups Ls(q), and in addition that G has only one conjugacy class of invo-
lutions. We take these conditions as the definition of the term G « £3(2), from
which the presentation for G is to be derived.
Therefore, at the end of stage 2 one of the target families has already emerged
from the analysis. A further bifurcation of the flow chart occurs at this juncture,
one path leading to the family Us(q), the other to the remaining families.
Stage 3. Assume that G* = Us(q) and use local group-theoretic methods to pin
down the structure of Oy (C) and its embedding in G to the extent possible.
[Without character theory, it does not seem to be possible to prove that Oy (C)
must be cyclic, as it is in the groups Us(q).]
Stage 4. Assume that G* = U$(q) and use block-theoretic methods to establish
the following results:
1. C has approximately the same structure as G* = Co* (2*), z* an involution
ofG*;
2. If P G Sylp(G), then B = NG(P) has approximately the same structure as
a Borel subgroup of G*; and
3. G acts doubly transitively on the set ft of right cosets of B with P acting
regularly on Q — {B}.
25
I n this context, this means that every component of C/0(C) is the image of a component
of C. See Section 7.
26. G E N E R I C S I M P L E G R O U P S 63
[It is essentially these conditions that form the basis of the term G « U^{q).
In particular, G is then a split (£?, iV)-pair of rank 1. Now it follows from the
Background Results that G = U^(q) in this case.]
Stage 5. Assume that G* ^ Ls(q) or Us(q) and use a combination of local analysis
and further block-theoretic calculations to prove that G has subgroups if, N and M
of approximately the same structure as a Cartan subgroup H* of G*, its normalizer
JV*, and a maximal parabolic subgroup M* of G* (except when G* = L±(q) or
U^q), in which case M approximates the centralizer of an involution in G*.)
[These conditions together with the given information about C (plus the already
determined involution fusion pattern) define the term G « G* in these cases. It is
analogous to the definition of the term G w L^{q).}
Stage 6. Assume that G* ^ U3(q) and prove next that (1) if G* = LA(q) or U4(q),
then G is generated by subgroups Lj, i = 1,2,3 such that (Li,L,2) = (S)Ls(q) or
(S)Us(q) for i = 1 and 3, and [ L i , ^ ] = 1; (2) otherwise, G is a split (B,N)-paii
of rank 2 of the same general shape as G*. Now one is in a position to identify G,
using the Curtis-Tits Theorem 26 or a unitary variation of it in (1), and proving in
(2) that the multiplication table of B is uniquely determined and in turn uniquely
determines that of G, forcing G = G*. In case (2), much less elaborate analysis is
needed than in the Fong-Seitz classification of split (B, iV)-pairs of rank 2, since so
much additional information is already known.
26. Generic simple groups

Because most of the known finite simple groups are of £ p -generic type (groups
of Lie type of large Lie rank or alternating groups of degree at least 13), it is
especially important to describe the various phases of the analysis that lead to
their presentations by generators and relations.
We thus assume that G is a 3C-proper simple group satisfying the following
conditions:
1. Either p = 2 and some element of &2(G) is a S2-group, or G is of even type,
p G <J{G) and some element of ££(G) is a Sp-group; and
2. G does not contain a strong p-uniqueness subgroup.
For brevity, for any x G XP (G), we write Cx for CG{X). Also, recall that
L°2{G) = £ 2 (G) if L°2(G) is not empty.
Because some element of ££(G) is a Sp-group, our general results on p-terminal
p-components yield the existence of x G 1p{G) such that Cx contains ap-component
K with K p-terminal in G and KjOv'{K) a Sp-group. We call such an (x,K) a
p-terminal Sp-pair. Moreover, among all such pairs, we choose x and K so that
K/Op'(K) is maximal in a suitable ordering of the elements of ££(G) according to
isomorphism type; and our analysis focuses on such a pair (x, K).
The case p = 2, K/Op'{K) = 2A m , is an anomaly which is easily eliminated
by an analysis of 2-fusion.
The global strategy is then aimed at showing for suitable p, x and K that G
possesses a neighborhood of (x, K) of the same shape as in one of the target groups
G*. To describe the process by which the desired neighborhood emerges from the
analysis, we need some preliminary terminology.
Set Cx = Cx/Op'(Cx), so that K is a component of Cx.
See section 27.
DEFINITION 26.1. Let y e Ip(NCx(K)) with y acting nontrivially on K and

let / be a p-component of C#(y). If the image of i" in K(y)/CK{y){K) is terminal
in K(y)/Cx(y}{K), we call (y,I) a s u b t e r m i n a l (#, if)-pair.
[When p = 2, there are small cases in which one must extend the definition to
allow / to be a solvable 2-component, so that 1/0(1) = 5L2(3).]
DEFINITION 26.2. If (y, I) is a subterminal (x,i^)-pair, then for any u E
(a;, 7/)^, the pumpup Iu of / in Cu (that is, the subnormal closure of i" in Cu) is
called a neighbor of K (more precisely, a (y, jQ-neighbor of (x,K)). Moreover,
if each p-component of Iu is quasisimple, we call the neighbor Iu semisimple.
Note that in particular, K itself is a neighbor of K (arising from u — x).
Furthermore, by Lp>-balance, each Iu is either a single ^-invariant p-component of
Cu or the product of p such p-components cycled by x.
The goal of the first two stages of the analysis is to show that every neighbor
of K is in fact semisimple, thus establishing a partial I? p -property for centralizers
of suitable elements of order p in G. This is achieved via signalizer functor theory,
using our assumption that G does not possess a strong p-uniqueness subgroup. We
focus on a suitably chosen Epa-subgroup A of Cx called an (#, K) p-source, and
in the first stage of the analysis verify that G is 3/2-balanced with respect to
A—that is, for all a, a' e A* and all B < A with B = Ep2, we have
1. AG(B)nCa <Opf(Ca), where AG(B) = f| 0^{Ch)\ and
beB*
2. [Opf{Ca,)C\Ca,A]<Op,{Ca).
[See section 29 of the next chapter for a fuller discussion.]
When p — 2 and K = K/Oy^K) is of Lie type over Fq, q odd, of sufficiently
high Lie rank, K contains the direct product H of three "root" SL2(#)-subgroups,
in which case we take A to map onto Z(H). It is then straightforward to verify
3/2-balance on the basis of L2'-balance and the ^ - p r o p e r t y in proper sections of
G. However, when such a choice of A is not available, verification of 3/2-balance is
considerably more complicated. This occurs when K = An or when p is odd, and
especially when p = 2 and K is of Lie type of low Lie rank, the most troublesome
case. The precise definition of an (x, K) p-source for all K is too technical to present
here.
We are now ready to state the objectives of the various stages of the analysis,
but in somewhat simplified form.
Stage 1. Establish the following two results:
1. By an analysis of 2-fusion, show that K/0^(K) ^ 2An when p = 2; and
2. In the remaining cases, verify that G is 3/2-balanced with respect to any
(x, K) p-source A.
R E M A R K. For 2An to be a S2-group, one must have n > 12 by the definition
ofS2.
Stage 2. Using signalizer functor theory, relative to A, show that suitable neighbors
of K are semisimple. In particular, K itself is quasisimple.
R E M A R K S . In Stage 2, on the basis of the signalizer functor theorem, it follows
from Stage 1 that the associated "3/2-functor" 63/2(^5^) on Epi-subgroups of A
26. GENERIC SIMPLE GROUPS 65
is a p'-group (see section 29 of the next chapter). The analysis splits into two parts
according as G3/2(G;A) = 1 or 0 3 / 2 (G;A) ^ 1. In the first case, we argue that
suitable neighbors of K are semisimple, as desired, while in the second, with the
aid of our p-component uniqueness theorems, we show that G possesses a strong
p-uniqueness subgroup, contrary to assumption.
In Stage 3, we establish a variety of results, which taken together show that

the internal structure of G approximates that of one of the desired target groups.
We need the following definition.
DEFINITION 26.3. Let (y, / ) be a subterminal (a^jFQ-pair and for u G (x,y)#,
let Iu be the pumpup of / in Cu. The (y, /)-neighborhood DSf of (x,K) consists
of the group (x,y), the group / , and the groups Iu, u G (x,y}#. Furthermore, we
say that N is vertical provided:
(1) Iu is quasisimple for every u G (#, y)^\ and
(2) I < Iu for some u G (x, y) — (x).
Stage 3. For some choice of the prime p and of the p-terminal Sp-pair (x, K) and
the subterminal (x, K)-paii (y, L), show that the (y, L)-neighborhood N of (x, K)
is vertical and that one of the following holds:
1. p = 2, K = An, n > 9, and either
(a) CG(K) = E4 and a root four-subgroup27 of K centralizes a G-conjugate
of K\ or
(b) m2(CG(K)) = 1;
2. K is of Lie type defined over Fq and the following conditions hold:
(a) mp(CG(K)) = 1;
(b) If p = 2, then q is odd, and if p is odd, then q = 2 n ;
(c) p divides q± 1;
(d) Either every component of N is defined over Fq or K is of low Lie rank;
and
(e) If K is a linear or unitary group, then every component of Ji of the
same untwisted Lie rank as K is a linear or unitary group; or
3. p = 2, K is sporadic, and 777-2(C^(if)) = 1.
REMARKS . We make a number of comments concerning the rather elaborate
conclusions of Stage 3. First, these results together yield that mp{CG(K)) = 1
unless p = 2 and K = An. The proof of this assertion depends on generational
properties of K together with our p-component uniqueness theorems. Likewise if p
is odd, K is necessarily of Lie type of characteristic 2, the proof of which depends
critically on the fact that G is of even type when p is odd.
The assertion that the neighbors in Ji are quasisimple is a consequence of our
maximal choice of (x,K). Indeed, if false for some u G (#, y)#, we pump up a
component of Iu to a p-terminal S p -pair (x*,K*) and compare (x*,K*) to (x,K)
in our ordering for a contradiction. The proof that N is in fact vertical (whence
I < Iu for some u e (x, y) — (x)) depends on the following generational property of
centralizers of elements of order p acting on S p-groups: If B = Ep2 acts faithfully
on the (quasisimple) S p -group L and b G B#, and we set Lo = {CL(bf)\bf G B — (6)),
then generally (b) is not weakly closed in a Sylow p-subgroup of Lo with respect to
A root four-subgroup of An is a four-subgroup of An fixing n — 4 points.

It is in the verification of (2c) (which we express by saying that p splits K)

that a choice of the prime p is made when G is of even type. To prove that p can
be chosen to split K, it is necessary to consider the set 7(G) of all odd primes p
for which no p-uniqueness subgroup exists, but for which there exists a p-terminal
Sp-pair (x, K), and then choose p G 7(G), x and K so that K is as large as possible
in our ordering, with p splitting K if possible. If, for such a choice of p, x and K,
it turns out that p does not split K, we shift attention to an odd prime r that does
split K and argue on the basis of our maximal choice of p, x and K that G contains
a r-uniqueness subgroup M. Using this result, we go on to show that M is in fact
a p-uniqueness subgroup, contrary to hypothesis.
Likewise the proof of (2d) (which we express by saying that 3Sf is level) depends
on our maximal choice of (x,K). Furthermore, if (2e) fails, so that some neighbor
Iu of the same untwisted Lie rank as K is not linear or unitary, we argue that the
pair (u, Iu) is also maximal in our ordering and then replace (x, K) by (it, Iu).
Finally, each target group possesses a neighborhood satisfying the conclusions
of Stage 3. However, there exist possibilities for K and CG(X) in Stage 3 that
correspond to no target group G*. For example, if p = 2, there is no G* if K =
An with rri2{CG(K)) = 1, or if K is sporadic, or if K is one of certain proper
homomorphic images of SLn(q), q odd. Thus it is not until the next stage of the
analysis that we are able to associate a specific target group G* with G.
On the other hand, when K is of Lie type, the fact that J$ is vertical and (in
most cases) level does enable us to eliminate the troublesome nonsimple "wreathed"
and "field automorphism" cases, which in the original classification proof required
rather elaborate analyses, with a final contradiction to the simplicity of G obtained
by a transfer argument.
Indeed, in the wreathed case G locally resembles J(x), where J = J\ x J2 x
••• x Jp with J\ = K, x cycling the Ji, 1 < i < p, and K = Cj(x) (whence
CG*{X) = K x (x)). However, in this case one checks that for suitable u, the
pumpup Iu of I in CQ{U) is the product of p components cycled by x, contrary to
the fact that Iu is quasisimple as Ji is vertical. Hence G* does not have this form.
Similarly in the field automorphism case, G resembles J(x), where J is defined
over FqP, x induces a nontrivial field automorphism on J, and K = E(Cj(x)). This
time one sees that Iy is defined over an extension field of FqP, contrary to the fact
that X is level, so G* does not have this form either.
Now fix the prime p, the p terminal Sp-pair (x, K), the subterminal (x, K) pair
(y, I) and the corresponding neighborhood Ji to satisfy the conclusion of Stage 3.
[There are cases in which there is more than one choice for (y, / ) . Usually y is taken
to lie in K, but in some situations the appropriate (y, / ) requires y £ K.)
We define the span of 3Sf to be the subgroup
G0(X) = (Iu\ue(x,y)#).
The next major step in the analysis is to identify Go(3\f) as a quasisimple X-

group.
Since the Lie type case is considerably more complicated than the alternating
and sporadic case, we treat it separately.
Stage 4a. If K is of Lie type defined over Fq, show that Go(N) is a central
extension of G* for some group G* of Lie type defined over Fq.
26. GENERIC SIMPLE GROUPS 67
REMARKS. The goal of the analysis is to show that Go(N) satisfies the same
set of Steinberg relations as hold in the target group G*, from which the desired
isomorphism follows. Moreover, this is to be achieved either via the Curtis-Tits
or Gilman-Griess approach. Many of the desired relations are already visible in K
inasmuch as the Lie rank of K is either 1 or 2 less than that of G*; and the remaining
rank 1 subgroups needed to identify Go (3Sf) are constructed by an analysis of the
neighbors in N (in some cases also of components of centralizers of other elements
of order p in CG((X, y))). Although it is probably possible to follow either approach
in proving that GQ(N) satisfies the Steinberg relations for G*, it appears to be
most efficient to use the Curtis-Tits theorem or a variant of it when p = 2 (and
correspondingly K is of odd characteristic) and the Gilman-Griess theorem when p
is odd (and K is of characteristic 2). There are several reasons why the first method
is better adapted for the p = 2 analysis and the second for the p odd analysis.
First, if p = 2 and v is a classical involution in K (i.e., CK(V) contains a
component or solvable component J = 6X2(9) with v £ 5), then we are able to
"keep J in the picture" and to determine its embedding in G. There is no analogous
result for odd p.
On the hand when p is odd, there exists a unique (up to conjugacy) elementary
abelian p-subgroup B of maximal rank in CG({X, y)), and using the embedding of B
in both K and a suitable neighbor of K one can determine a subgroup R of NQ{B)
generated by involutions acting as reflections on B and such that R is isomorphic
to the Weyl group W* of G* (to only a proper subgroup of W* in some cases when
q = 2).
A corresponding result for p = 2 would be much more difficult to derive since
one would then have to show that R is isomorphic to a reduced monomial subgroup
R* of G* (R* contains a nontrivial elementary abelian normal 2-subgroup H* and
R*/H* = W*). This would involve (a) identification of an appropriate elementary
abelian 2-subgroup B of CG((x,y)) with B ^ # * ; (b) proof that R/B 2* W*\ and
(c) determination of the isomorphism type of the extension of R/B by B.
In the course of the foregoing analysis, we eliminate those possibilities for K of
Lie type that correspond to no target simple DC-group G*. Also the low Lie rank
generic cases require special treatment.
We express the conclusion of Stage 4a by writing G ?x G*.
We carry out a similar analysis in the alternating and sporadic cases.
Stage 4b. If p = 2 and K is not of Lie type, show that

(i) KÂn, n>9;
(ii) CG{K) £* £4; and
(iii) G0(N)^,4n+4.
REMARKS. If K = An and CQ{K) = E±, we identify G 0 (N) via the classical

presentation for A n + 4. Hence to establish the conclusions of Stage 4b, it remains
to eliminate the cases rri2(Cc(K)) — 1 with K alternating or sporadic.
If K = An, one argues first that Go(N){x) = £ n + 2 5 tins time using the classical
presentation for E n + 2 (see the next section) and then uses transfer to contradict
the simplicity of G. [The case n — 10 is exceptional since Aut(F$) possesses an
outer automorphism with a Z2 x E10 centralizer. This case is also eliminated by
transfer.]
On the other hand, if K is sporadic, then as K is a S2-group,
K 9* J i , Mc, Ly, ON, He, Co3 or F5.
In each case, one derives a contradiction by playing off the existence of the vertical
neighborhood U\T against the structure of appropriate local subgroups of G.
We express the conclusion of Stage 4b by writing G « A n + 4.
Thus Stages 4a and 4b together yield that G « G* for some G*.

Finally in Stage 5, we argue that Go(N) = G, thereby completing the proof of
our theorem in the generic case.
Stage 5. Show that G « G* implies G = G* for the G* arising in Stage 4.
REMARKS . In the contrary case, GQ(N) < G as Go(3Sf) is a central extension
of G*; and we argue that M = -/V"G(GO(3VT)) is a strong p-uniqueness subgroup of
G, contrary to our hypothesis. The first step in the proof is the assertion that
CG(U) < M for every u G (x,y)^, which follows readily from the definition of
Go(N) and the structure of CG(U) for u £ (x,y)^. [It is at this step that the
Aut(Fs) and £12 cases separate.]
We hope that this rather detailed outline provides a useful summary of the
generic case analysis.
27. The identification of G

As repeatedly emphasized, the overall objective of the classification proof is
to derive a set of internal conditions on the group G under investigation that will
lead to a presentation for it identical to a presentation in one of the known simple
groups G*, thereby enabling us to "recognize" G as the target group G*. There is
great variation in the complexity of the passage from such internal conditions to
the desired presentation. At one end of the spectrum, the existence of a vertical
neighborhood in G leads to the identification of G as a generic group of Lie type
via the Steinberg relations or to an alternating group via the classical presentation.
At the other end, corresponding presentations for some sporadic groups such as Ly
and ON have required computer calculations, as noted earlier.
We conclude this sketch of the classification proof by first outlining the pasage
from subgroup information to generators and relations data in a specific example in
which the steps are particularly transparent. Following that, we state the Curtis-
Tits and Gilman-Griess theorems. We shall take as G* the almost simple group
S n , n > 11, and we shall assume that our 3C-proper simple group G possesses
a "neighborhood" (x,y,K,J,I) of "symmetric" type—i.e., satisfying the following
conditions:
1. Cx = (x) x K, where K ^ E m , m > 9;
2. Cy = (y) x J, where J = E m ;
(27.1) 3. y £ K is a transposition and a: € J is a transposition;
4. CxnCy = (x, y) x i*, where I = K D J = E m _ 2 ; and
5. I1 is a component of Cxy.
27. THE IDENTIFICATION OF G 69
Note that if G = S m + 2 , m > 9, and x and y are the transpositions (12) and
( m + l m + 2), then (x,y) is a four-group satisfying the conditions (27.1). [Since
m > 9, all automorphisms of K and J are inner, so it is a well-defined notion for
an element of one of these groups to be a transposition.]
Although S m + 2 is not simple, the specified configuration of subgroups corre-
sponds nevertheless to a "real-life" situation. Indeed, it seems that the best way to
eliminate the possible existence in G of such a configuration is first to prove that G
has a presentation identical to the classical presentation of S m + 2 by transpositions,
whence G = £ m +2; a n d only after this has been established to use the simplicity
of G to force a contradiction.
Observe that E m _2 has two embeddings in E m up to conjugacy: one has image
in Am, fixing none of the ra points, while the other, which we call "canonical", has
image fixing 2 points. At the outset, the precise nature of the embeddings of I in
J and K needs to be settled; questions of this nature about embeddings are typical
at this stage of the proof. In this case, a brief fusion argument using ra > 9 shows
that
(27.2) I is canonically embedded in both J and K.

We shall outline the proof of the following illustrative result.
THEOREM 27.1. Ifx,y,K,J,I satisfy (27.1) and (27.2), then
(K, J) = S m + 2 .
As indicated, the goal of the proof will be to show that the subgroup {K, J)
of G has a presentation identical to that of the classical presentation of £ m +2 in
terms of its transpositions stated in section 11.
Thus our task is to produce nontrivial elements a^, 1 1, and r^- = 3 if \j — i\ = 1. Indeed, it will
then follow that the map ( i i + 1) t—• a^, 1 < i < m -h 1, extends to an isomorphism
of S m + 2 with (K, J).
Identifying K = S m with the symmetric group on the letters { 1 , 2 , . . . , m}, we
have the corresponding canonical generation of K by transpositions c^ = (i i + 1),
1 < i < m — 1. Moreover, as y is a transposition in X, we can order the underlying
set so that y = a\ — (12).
Then since CK{V) — (y) x I and J is canonically embedded in K, I is generated
by elements 03,04,... , a m _ i corresponding to the transpositions (34), (45), . . . ,
(m — 1 m).
On the other hand, I x (x) < J with this embedding being canonical. We
can therefore similarly view J as acting on the set { 3 , 4 , . . . , m + 2} with x corre-
sponding to the transposition a m +i = ( m + 1 ra + 2). Furthermore, there exists
a transposition a m = (ra m -f 1) in J such that the elements 03,04,... a m +i are a
canonical generating set for J.
In particular, clearly (K,J) = {a\, 02, •.. , a m + i ) . Furthermore, as (x,y) cen-
tralizes / with x centralizing K and y centralizing J, all relations (a,iaj)rij = 1 hold
whenever both ai and aj lie in K x (x) or both lie in J x (y), since they hold in
these groups. However, the only a^ from K x (x) not contained in J x (y) is 02,
while the only one from J x (y) not contained in K x (x) is am. Thus to show
that the a^, 1 < i < m + 1, form a canonical generating set yielding the desired
presentation for (K, J ) , it remains only to prove that the order of the product a2drn
is the same as that for the corresponding generators o / £ m + 2 - That is, (a2<2m)2 = 1,
or equivalently, a<i centralizes am.
Thus the following lemma will complete the proof of the theorem.
LEMMA 27.2. a2 centralizes am.
For simplicity of notation, put u = a<i, v = a m , t = 03, and w = a m _ i . Then

by observation of K x (x) and J x (y), we know all the relations among y, u, t and
x,v,w, respectively, and, in addition that (y,t) centralizes (x,v,w) while (x,w)
centralizes (y,u,t). Furthermore, these six elements are all distinct.
We now choose k with 4 < k < m — 2, and consider the transposition z =
afc. Such a choice of k is possible as m > 9. Furthermore, the six involutions
x,v,w,y,u,t are all contained in CQ{Z). On the other hand, J has only one conju-
gacy class of transpositions so z = xa for some a G J . Set L = Ka, so that
CG(z) = (z) x L, where L ^ E m .
Since a € J, y = ya, and as y is a transposition in K, it is then also one in L.
Now Cx(^) — £ m _ 2 x ^2 (as z is a transposition of K = S m ) . Furthermore,
y, u and £ are all CK(z)-conjugate, and so all are transpositions in L. Similarly,
considering Cj(z), we obtain that also x, v, w map to transpositions in L(z)/(z) =
L. Observe that by inspection of Cy and C x , the six elements y, u, t, x, v, w; remain
distinct modulo (z) (except for a possible coincidence of u and v, which would in
any case imply [u,v] = 1).
To prove that u and v centralize each other we work in the natural represen-
tation of L(z)/(z) on m letters and show that their supports are disjoint. Suppose
false. Since u ^ v, their supports thus have a point (3 in common, as indicated by
the following diagram (i.e., uv is a 3-cycle):
<* u & v 1
o o o
On the other hand, in K we see that tu has order 3, whence the support of t
has a point in common with the support of u. But t centralizes v, and t and v are
distinct modulo (z), so this common point must be a. Similarly w and v have the
point 7 as common support, thus giving the following diagram (where UJ ^ S since
[£, w] = 1, as can be seen in K)\
o o o o o
Finally consider the elements y and x. Since yu has order 3, the support of y
has a point in common with that of u. But y centralizes t, and y and t are distinct
modulo (z), so this common point is not a. Hence it must be /3. Similarly it follows
that /3 is one of the points of the support of x (as xv has order 3 and x centralizes
w). However, since x ^ y mod (z), we conclude that the projection of xy on L is
a 3-cycle, contrary to the fact that x and y are commuting involutions.
This contradiction establishes that u and v commute modulo z. Therefore
[u,v] G (z) Pi (L(z)Y = 1, which proves the lemma and completes the proof of the
theorem.
Finally we turn to the Curtis-Tits and Gilman-Griess theorems. The precise
statement of the former [Cul, Ti2] is expressed in terms of relations among certain
root subgroups of a group of Lie type.
27. THE IDENTIFICATION OF G 71
T H E O R E M 27.3. (Curtis-Tits) Let G* be a finite group of Lie type. In a Bruhat

decomposition of G*, let E be the root system and Xa (a G T>) the corresponding
root subgroups. Let II be a fundamental system in E and for each a G II let
G^ = (Xa^X-a). Assume that \U\ > 3. Then the relations among the elements
of the groups G* holding in the groups (G*, G*p), a, (3 G II, form a set of defining
relations for a central extension of G*. Consequently, if G is any group generated
by subgroups Ga (a G li), and if there are homomorphisms $& : G£, —> Ga and
<f>a/3 • {G^GV) —> (Ga,Gp) for all a, (3 G II, which are coherent in the sense that
<f>ap — 4>(3a and 0C*/3|G* = $<* for aM a and P i>n H> then G/Z(G) is a homomorphic
image ofG*/Z(G*). "
In the applications, G is simple and G* quasisimple, and s o G = G*/Z(G*).
In practice, hypotheses which are superficially somewhat weaker are sufficient
to establish the same conclusion. For example, in the simplest case of untwisted
groups with only single bonds in the Dynkin diagram, the theorem has the following
corollary.
COROLLARY 27.4. Let q be an odd prime power and let G* be one of the simple
groups An(q), Dn(q), E^(q), Ej(q) or E%(q), so that the Dynkin diagram A o/G*
has only single bonds. Suppose that G is a simple group generated by subgroups
Ga = SL2{q), one for each node a G A, and let za be the involution in the center
of Ga. Assume that the following conditions hold.
(1) (GaiGp) = (P)SLs(q) or [Ga,Gp] = 1 according as a and /? are joined or
are not joined in A;
(2) [za, z@] = 1 for all a^ (3 G I I ; and
(3) / / we set Ho = (za \ a G A) (so that Ho is an elementary abelian 2-group),
then for each a, \Ho : G# 0 (G a ) | < 2.
ThenG^G\
In the applications, when the Ga are constructed in a natural way from a ter-
minal component and its neighbors, the second and third conditions above emerge
naturally. We shall also use variants of the above result applicable to twisted groups
G* and allowing the possibility (Ga, Gp) = (P)SUs(q). Theorems of this type were
investigated by Phan and more recently by Das [ P h i , P h 2 , D a l ] .
To state the Gilman-Griess theorem [GiGrl], we first make the following def-
inition.
DEFINITION 27.5. Two groups G(q) and Gi(q) of Lie type of characteristic 2
are said to be F-compatible if and only if the following conditions hold:
(1) G(q) has Lie rank at least 4 and G\{q) has Lie rank at least 2;
(2) G(q) and Gi(q) are either both twisted or both untwisted;
(3) If E and Ei are root systems of G{q) and Gi(q), respectively, then Ei can
be identified with a proper subsystem (also called Ei) of E; and
(4) If a G Ei, then the root subgroup Xa of G(q) is isomorphic to the root
subgroup Xa of Gi(q);
(5) Y is a subgroup of the Weyl group of E, and for every pair of roots a, (3 G E,
there is an element w G Y such that {a,(3} w C Ei .
THEOREM 27.6. (Gilman-Griess) Let G(q) and G\{q) be Y-compatible groups
of Lie type of characteristic 2, as in the preceding definition. Identify Ei with a
subset o/E ? and let W\ and W be their Weyl groups, so that W\ is identified with a
subgroup ofW and Y < W. In a Bruhat decomposition of G(q), let {Xa \ a G Ei }
be the set of root subgroups generating G\{q)y with H\ and N\ the corresponding
Cartan and monomial subgroups of G\(q).
Suppose that Go is a group generated by G\ (q) and a subgroup R satisfying the
following conditions:
(a) R^Y;
(b) When R is identified with Y as in (a), Rf)Gi(q) = Y C\ W\; and
(c) For each a £ Ei, the stabilizer Ra of a in R normalizes Xa, and the group
Ra/Cii^Xa) is abelian.
Under these conditions, G0/Z(G0) = G(q)/Z(G(q)).
F. ADDITIONAL COMMENTS
28. The length of the proof

The preceding sketch provides a summary of our several-thousand-page multi-
volume proof of the classification of the finite simple groups. It is reasonable to ask
why any single mathematical theorem should require such an excessively long argu-
ment. The only answer we can give is that the present proof consists of the shortest
argument we have been able to devise. Perhaps the matter should rest there, but
this response seems somehow inadequate, especially as the general mathematical
community believes on the basis of the historical record that in due course this re-
sult, like other major mathematical theorems, will undergo dramatic simplification.
Even though our "second generation" classification represents roughly a three-fold
reduction of the original proof, it is far too long to satisfy this requirement.
Obviously it is impossible to predict the minimum length of a proof of any
mathematical theorem. Nevertheless, one may ask whether there are features of
the classification of the finite simple groups which indicate heuristically whether or
not a short proof—on the order of a few hundred pages—is attainable.
Consider first the very existence of the known simple groups. Elegant construc-
tions of the groups of Lie type exist via the theory of algebraic groups. Moreover,
the existence of M24 suffices for the construction of Go\ and that of Go\ in turn
for the existence of F\, and embedded within these three groups are twenty of the
twenty-six sporadic groups. However, a complete proof of these two assertions, plus
the construction of the remaining six sporadic groups, requires a substantial body
of argument, and we have not yet begun the main body of the classification proof.
But existence is not the central issue—in fact, a reasonable case can be made
that the question of existence is entirely separate from that of the classification
of the finite simple groups, which should be concerned solely with proving that
the list has not missed any groups. Rather the key factor that will dominate any
inductive proof of the classification theorem is the a priori structure of a minimal
counterexample G. Why should the structure of G bear any resemblance to that of
some known simple group G*; yet how is one to imagine a proof that G is isomorphic
to G* which does not require one to show first that G and G* have comparable
subgroup structures?
28. T H E L E N G T H OF T H E PROOF 73
The most common suggestion of an alternative classification strategy is to try

to associate some mathematical object on which G acts as a group of automor-
phisms and then to derive the classification theorem from a classification of the
corresponding "geometries". It certainly sounds plausible, particularly as there
exist important geometric classification theorems, notably Tits's classification of
spherical buildings of rank at least 3, yielding as a corollary a characterization of
the finite groups of Lie type of Lie rank at least 3.
Perhaps we can provide insight into the classification of the simple groups
by examining this idea in some detail. For simplicity we limit the discussion to
an important but very special case. We consider the problem of determining the
minimal simple groups, which by definition consist of the nonabelian simple groups
G each of whose proper subgroups is solvable. Our remarks apply also to the
slightly more general class of AT-groups, in which each local subgroup is assumed
solvable. Moreover, by replacing the term "solvable" by "3C-group" and the short
list of groups in (28.1) below by the full set of known simple groups, our remarks
can be readily adapted to a minimal counterexample to the complete classification
theorem.
Whatever approach one may take in defining an associated geometry T(G)
of G, in order for G to act as a group of automorphisms of T(G), clearly some
structural properties of G must be involved in the definition. However, until one
proves otherwise, why doesn't the subgroup structure of G resemble that of an
arbitrary solvable group G*l If this is the case, then T(G) will resemble 37(G*), the
associated geometry of the general solvable group!
On the other hand, the goal is to prove that G is isomorphic to one of the
known simple AT-groups:
(28.1) L2(q), Sz(2n), L 3 (3), U3(3), 2

F 4 (2)', A7, or Mn,
excluding the solvable exceptions £2(2), £2(3), and Sz(2). Equivalently, in our
geometric formulation, our goal is to show that T(G) is isomorphic to the geometry
of one of these groups. [Note that, in particular, G must have even order, so
that there are no minimal simple groups of odd order. Thus the group-theoretic or
geometric analysis must include a proof of the Feit-Thompson Odd Order Theorem!]
Observe that apart from A? and M n , the groups in (28.1) are of Lie type of
Lie rank at most 2 (or of index 2 in such a group), and the remaining two groups
are doubly transitive permutation groups. Thus the desired structure of T(G) is
extremely restricted compared to that of the geometry of an arbitrary solvable
group. The restricted nature of T(G) for the groups in (28.1) is reflected in their
subgroup structure. Consider, for example, a Sylow p-subgroup of G for some prime
p. If p = 2, then except for 2 F4(2)', P is either abelian, of nilpotency class 2, or of
2-rank 2; if p > 7, then P is cyclic; and if p — 3 or 5, then P has rank at most 2.
Thus introduction of T(G) has shifted the focus of the classification problem
but has not necessarily eased the task. Clearly the assumption of simplicity must
be used to show that the geometry of G in no way approximates that of a general
solvable group, but is in fact close to that of one of the groups in (*). This would
appear to require an effort comparable to showing group-theoretically that the
subgroup structure itself of G is severely limited, approximating that of one of the
above groups.
In fact, the group-theoretic approach has a decided advantage at the present

time since powerful tools have been developed over the past century for studying
the subgroup structure of G:
A. There exists a well-developed theory of solvable groups.
B. Representation and character theory (ordinary and modular) have consider-
able consequences for the subgroup structure (and order) of a simple group.
C. There exist many non-simplicity criteria established by means of the transfer
homomorphism.
D. Deep investigations have been made in the theory of doubly transitive per-
mutation groups.
No corresponding theory has as yet been developed in connection with the geometry
of a general solvable group.
This discussion has, of course, been predicated on the assumption that the
classification of the finite simple groups is to be carried out by induction, with the
group G under investigation taken as a counterexample of least order. However,
for some time finite group-theorists (and perhaps the mathematical community as
well) have dreamed of a non-inductive proof that depended solely on the simplicity
of G, independent of any DC-group conditions on its proper subgroups. At least two
appealing ideas have been floated:
1. The Sims conjecture, which asserts that in a primitive permutation group
G (i.e., a transitive group on a set Q in which the stabilizer Ga of a point
a G ft is maximal in G) the order \Ga\ is bounded above by a function of
the minimum length of an orbit of Ga in its action on Q — {a}.
2. The conjecture that there is an a priori way to establish an upper bound
to the order of a sporadic simple group (which would, of course, imply that
there are only a finite number of sporadic groups).
Despite some effort, a viable independent approach to the Sims conjecture has
not been developed. A proof of either conjecture would very likely yield a much
shorter proof of the classification theorem, and one can never preclude the possibil-
ity that one of them will be verified some time in the future. [The Sims conjecture
has been shown to be a consequence of the classification theorem [CPSS1].]
However, the more likely prognosis is that simplifications in the classification
proof will continue to occur as they have in the past, relative to specific subcases.
For example, in the years since the Odd Order Theorem was first established in 1962,
its proof has been reduced by a factor of two with the aid of new techniques and
some improved arguments. On the other hand, the original Feit-Thompson overall
strategy has remained intact. Likewise, Bender and Glauberman have achieved
considerable simplifications of the the original classifications of simple groups with
dihedral or abelian Sylow 2-subgroups [ G W 1 , W a l , B e G l 2 , Be2].
The proof we present here is in the same spirit. Our global strategy is essentially
the same as that of the existing proof, but with some important repackagings and
consolidations which produce their own substantial simplifications. However, most
of our improvements are in the treatment of various major subcases of the theorem.
No doubt this process will continue long after our classification proof has been
published, steadily chipping away at its overall length. But the number of subcases
seemingly requiring separate treatment is so large that no matter how successful
this effort turns out to be, it is difficult to visualize that it will ever reduce the clas-
sification of the finite simple groups to the desired several-hundred-page argument.
29. THE IK-GROUP ENVIRONMENT 75
29. The X-group environment

As has been repeatedly stressed, our intent is to establish the Classification
Theorem as a single result rather than as a consequence of a collection of largely
independent subsidiary theorems, as was true of the original proof. The implication
of this viewpoint is that we make no special attempt to derive any such subsidiary
result in any more general form than needed for our arguments. On the contrary, the
aim throughout will be to strive for simplicity rather than generality of exposition.
Certainly other approaches to the classification of the simple groups are equally
valid, but we feel that an attempt to construct a "geodesic" proof is justified by the
fact that even then the argument will require several thousand pages to elucidate.
The principal manner in which this single-minded perspective manifests itself
is that in our set-up, all proper subgroups of G are X-groups. We invoke this
continually throughout the analysis. To illustrate the point, consider the case
in which the simple group G under investigation contains a strongly embedded
subgroup M. The Bender-Suzuki theorem gives a complete classification of such
simple groups G without any X-group assumptions on its proper subgroups. On
the other hand, if one assumes that M is a X-group, one can prove at the outset
that one of the following holds:
v(1)
;M is solvable; or
(29.1)
V ;
(2) E(M/02>(M)) ^ L2(q) for some q = 3 mod 4
[In particular, G has dihedral Sylow 2-subgroups in the latter case.]

Indeed, a preliminary property of groups with a strongly embedded subgroup
asserts the following:
(29.2)
(1) M has a single conjugacy class of involutions; and
(2) M = CM(Z)D for some involution z G M and subgroup D of odd order.
But if one analyzes X-groups satisfying the conditions of (29.2), one finds that
M must have one of the two forms specified in (29.1). Hence our Classification
Theorem requires the Bender-Suzuki theorem only under the assumption that the
strongly embedded subgroup M possesses one of these two structures, in which case
the analysis is easier than in the general case; and it is this approach that will be
presented here, in [Hi].
This DC-group approach has certain significant consequences. First, by shifting
portions of the argument to preliminary X-group properties, as with the strongly
embedded subgroup theorem above, it makes the essential lines of the proof of a
particular subsidiary result more readily transparent. Second, again as in the above
example, it leads to the investigation of properties of X-groups which may be of
independent interest. Finally, it often allows for extensions to arbitrary primes of
results previously proved only for the prime p = 2.
An example of the latter is the Aschbacher-Gilman theorem [A4, Gil] that
a terminal (i.e., nonembedded) component K of 2-rank at least 2 is necessarily
standard in a simple group G. The original result, for general G but p = 2, depends
upon special properties of involutions. On the other hand, under the assumption
that G is X-proper, an analogous assertion holds for odd primes p with K a p-
component of the centralizer of some element of G of order p and K terminal in
G. Moreover, the proof for odd p is equally valid for p = 2, thus leading to a
considerable conceptual simplification of the global strategy.
The strongly embedded subgroup theorem is an example of a simplification re-
sulting solely from our DC-group hypothesis. Other simplifications occur because we
permit ourselves to derive subsidiary results in less than full generality, using any
convenient extra hypotheses which will be available when we apply them. Exam-
ples of this are our treatment of Goldschmidt's strongly closed abelian 2-subgroup
theorem and of the classification of split (2?, iV)-pairs of rank 2 (see sections 14 and
25).
Furthermore, as indicated, an important effect of our global DC-group assump-
tion is that it enables us to package the classification proof in a manner quite
distinct from the existing proof. This package is composed of a number of major
subtheorems about DC-proper simple groups G, which will be described in detail
in the next chapter. Within this X-group context, they can be viewed as parallel
results, which taken together yield the Classification Theorem.
Finally, many of the properties of DC-groups that we require are general results.
On the other hand, as in the existing classification proof, our arguments require a
great many detailed, specialized properties of individual quasisimple DC-groups or
families of such groups, which for the most part have no intrinsic interest beyond
the specific context in which they are needed. The seemingly excessive length of
our DC-group "dictionary" is in part at least a consequence of strict adherence to the
following principle: A property of the known simple groups, or of %-groups, can be
used in the course of the proof only if it has been explicitly stated in a preliminary
X-group lemma. This is in marked contrast to much of the existing literature, in
which many such properties are invoked, often implicitly, without proof or reference.
We conclude with a comment only indirectly related to the DC-group hypothesis.
This is the fact that we have endeavored to utilize methods of local group-theoretic
analysis wherever possible. As a result, our treatments of simple groups with semidi-
hedral or wreathed Sylow 2-subgroups and of groups with Sylow 2-subgroups of
type Us(4) differ in significant ways from their existing classification. Indeed, the
original proofs can be viewed as making maximum use of block-theoretic methods,
whereas ours are designed to replace such arguments where feasible by ordinary
character-theoretic or local arguments. Although this approach does not necessar-
ily yield shorter proofs, nevertheless our taste is for internal arguments based on
local subgroup structure.
30. The term G « G*

As already described, the thrust of the classification proof is first to force the
internal structure of the DC-proper simple group G under investigation to approxi-
mate that of some known simple group G*, this conclusion designated by the symbol
G « G*, and then to turn this approximation into a presentation for G identical
to one that holds in G*, thereby proving that G = G*. It is clearly critical for
understanding the overall strategy to have a good picture of the various sets of
internal conditions that characterize the term G « G*.
As already stated, in the generic Lie type case the term G « G* is defined by
the condition that G possess a vertical neighborhood generating a central extension
of the generic target group G*. Likewise for generic alternating groups (of degree at
least 13) the term G « G* is expressed by the existence of a vertical neighborhood
31. READING THIS SERIES 77
identical to that determined by a root four-subgroup of G* and generating an

alternating group. Analogous conditions define the term G & An, 9 < n < 12.
On the other hand, in the remaining special type cases the corresponding defi-
nitions have four possible shapes, three of which have been indicated in the course
of the preceding discussion. Historically, the oldest is that given by the involution
fusion pattern together with the structure of the centralizers of involutions, and we
take these conditions as the definition of the term G « G* when the target group
G* is either sporadic or alternating of degree 7. [As remarked earlier, in the case
G* = M n , we must add the condition \G\ = \Mu | to distinguish G from the group
1,3(3), which has the same involution fusion pattern and centralizer of involution
structure. Also A$ = 1/2(4), A& = 1/2(9), and As = L±(2) are best identified as
groups of Lie type.]
Next for the groups of Lie type of Lie rank 1, we define the term G « G* by
the conditions that G be doubly transitive with the structure of a point stabilizer
closely approximating that of a Borel subgroup of G*. [In particular, this covers
A5 and ^6.]
There remain (special) groups of Lie type which with a small number of excep-
tions are of Lie rank 2 and 3. Here definitions differ according as the target group
G* is of odd characteristic r or of characteristic 2. In the odd characteristic case,
for most groups G* of Lie rank 2, the term G « G* is defined by conditions on the
involution fusion pattern, the structure of the centralizer of a 2-central involution,
the structure of the "Weyl group" of G, and also the structure of a certain r-local
subgroup containing a Sylow r-subgroup of G; for groups of larger Lie rank and
U±(q) and Us(q), it is defined by the existence of an appropriate vertical neighbor-
hood.
Finally, it is the quasithin type case that gives rise to most of the special groups
of Lie type of characteristic 2. Since this case is to be treated by the amalgam
method, the term G « G* is defined by conditions on the structure of maximal
2-local subgroups of G containing a given Sylow 2-subgroup. For the few groups of
Lie type arising from the Klinger-Mason case, on the other hand, the term G & G*
is defined by the existence of an appropriate vertical neighborhood.
31. Reading this series

Because of the extreme length of the proof of the Classification Theorem and
hence of this series of monographs, it will be useful to give a few guidelines for
reading the text.
First, the next chapter is a prerequisite for all that follows, for the detailed
outline that it includes provides the context for the specific results to be established
in each of the later chapters.
Furthermore, the introductory material of chapters [I^], [1^], [HA], [Up]? [HsL
e
[IIG]» "tc, will include basic notation needed for the subsequent chapters, and so
it too is required reading. On the other hand, although many of the results within
these chapters are of independent interest, a large percentage, especially in the
3C-group chapters [II5], [His], etc., are highly technical and their hypotheses are
difficult to grasp independently of the context in which they arise. Thus these
chapters are best viewed primarily as reference material for the body of the text.
Beyond these caveats, we have attempted as much as possible to write each
chapter of Parts II, III, IV, and V so that it can be read independently of any of
78 P A R T I, C H A P T E R 1: O V E R V I E W
the other chapters. However, occasionally there will be references to a result in a

prior chapter — in most cases to the principal result of the earlier chapter.
http://dx.doi.org/10.1090/surv/040.1/02
PART I, CHAPTER 2
OUTLINE OF PROOF
INTRODUCTION
As stated in the preceding chapter, the purpose of this series of monographs is

to provide a proof of the Classification Theorem, assuming the Background Results
listed there. The theorem can be formulated equivalently as follows.
CLASSIFICATION T H E O R E M . Every %-proper finite simple group is cyclic, a
group of Lie type, an alternating group, or one of the 26 sporadic simple groups.
Recall that a group G is DC-proper if and only if each of its proper subgroups
is a %-group; a group X is a 3C-group if and only if every simple section of X
(including X itself if X is simple) is isomorphic to a known simple group—i.e., to
one of the groups in the conclusion of the Classification Theorem.
Throughout this series, we shall assume the Background Results, and

the letter G will be reserved for a DC-proper noncyclic simple group.
Since both the Odd Order Theorem and Z*-Theorem are included as Back-
ground Results, it follows in particular that G has even order and 2-rank at least
2.
As described in the brief sketch in sections 23-26 of Chapter 1, a good way
to visualize the proof of the Classification Theorem is as a grid whose rows rep-
resent the major case divisions, as dictated by the distinct techniques needed for
the analysis, and whose entries in each row represent the conclusions reached at
successive critical stages of the argument in each of the individual cases. In fact,
it is preferable to split this grid into two separate grids-the uniqueness grid and
the classification grid. This is because the uniqueness results can be viewed
as preparatory, ruling out the existence of certain special "uniqueness" configu-
rations in G, while the results described in the classification grid constitute the
main body of the classification proof, taking seven mutually exclusive and exhaus-
tive possibilities for the structure of G and driving each case to the conclusion
that G is a known simple group. Typically the results from the classification grid
are proved by contradiction, with contradictions arising from appeal to uniqueness
theorems.
The classification grid is itself divided into three major subgrids corresponding
to the respective cases in which G is of special odd type, special even type
and generic type. These terms were discussed in sections 23-26 of the preceding
chapter and will be made precise below. Since the analysis in the generic case
depends on certain results from the special cases, the logic of the total proof is
clearest if the special cases appear in the classification grid prior to the generic
79
80 PART I, CHAPTER 2: OUTLINE OF PROOF
case. On the other hand, because of the obvious importance of the determination
of all "generic" simple groups and because of the extreme length of the analyses in
the special cases, we shall present the generic case before the special cases, stating
explicitly those special case results required for the proof.
Thus this series of monographs is divided into five parts as follows:
Part I: Preliminaries
Part II: Uniqueness results
Part III: Generic simple groups
Part IV: Special simple groups of odd type
Part V Special simple groups of even type
The uniqueness grid dictates the structure of Part II, and the classification grid
dictates that of the subsequent three parts. Part I consists of the two chapters of
this volume and two further chapters: one on general group theory, the other an
introduction to the theory of X-groups.
The uniqueness grid is divided into three parts: first, those uniqueness results
that hold only for the prime 2, then those valid for all primes, and finally those
that hold only for odd primes.
The classification grid is yet more elaborate. Part III corresponds to one row
of the grid, while each of Parts IV and V corresponds to three rows. Thus the
classification grid represents a total of seven broad cases—the generic one plus six
special ones.
In organizing the grids, our aim has been to emphasize conceptual stages of the
analysis of each case. Hence there is no necessary correlation among the lengths
of the arguments required to pass from one stage to the next, either within the
analysis of a given case or among those of different cases.
Although the solvability of groups of odd order is one of the Background Re-
sults, for conceptual completeness we are including the various stages of the analysis
of such groups within the classification grid.
Note that the term G « G* for some known simple group G* appears in both
grids. As described in sections 24 and 30 of Chapter 1, this symbol is intended
to incorporate the exact set of internal conditions on the group G from which a
presentation for G is to be derived which yields the desired conclusion G = G*.
The meaning of G « (7*, which varies from case to case, will be discussed below.
After we present the grids in sections 2 and 3, the balance of the chapter will
be devoted to an explanation of all terms appearing in the grids and a description
of the successive stages of the proofs of each of the principal subsidiary theorems
which taken together provide a proof of the Classification Theorem. This will then
be followed by a brief discussion of the principal techniques used in the analysis.
A. T H E GRIDS
1. Some basic terminology

Some of the basic terms about to be defined were introduced in the preced-
ing chapter, but we repeat the definitions for completeness. As is our convention
throughout, G is a DC-proper simple group.
1. S O M E BASIC T E R M I N O L O G Y 81
DEFINITION 1.1. Let X be an arbitrary finite group and p a prime. A com-

ponent of X is a quasisimple subnormal subgroup of X. (A quasisimple group
is a perfect central extension of a simple group.) Furthermore, we define
TP(X) = {x G X | x has order p}, and
&P(X) = {K | for some x € TP(X), K is a component of Cx(x)/0P'(Cx(x))}
As described in the preceding chapter, the structure of the sets £ P (G) in the
DC-proper simple group G under investigation dominates the global logic of the
classification proof.
DEFINITION 1.2. If a; € XP(X), p a prime, and K i s a component of Cx(x) of
order divisible by p, we say that the pair (x, K) is terminal in X if and only if K
is a component of Cx(y) for every y E Xp(Cx(K)).
When the element x or the prime p is clear from the context, we simply say
that K is terminal in X, and call K a terminal component. This definition
includes Definition 7.4 of Chapter 1 by the remarks made after that definition.
Recall that % is the set of quasisimple X-groups, i.e., the set of perfect central
extensions of simple groups of Lie type, alternating groups, and sporadic groups.
We have
% = Qhev U Alt U Spor, and Qhev = Upehev(p)
where Ghev(p) (resp. Alt or Spor) is the set of all K G % such that K/Z(K) is
simple of Lie type of characteristic p (resp. alternating or sporadic). Moreover, for
a given prime p, %p is the set of quasisimple DC-groups K for which Op'(K) = 1.
In section 12 below we shall precisely define a partition of %p into two sets: Cp,
consisting of those groups which are "characteristic p-like", and Gp>, the comple-
mentary set. The set C2 is required to define the key term "even type". We shall
call a group in 62 (or Cp, etc.) a C2-group (or Cp-group, etc., respectively). We
also use the standard notation mp(X) for the p-rank of X , the rank of the largest
elementary abelian p-subgroups of X.
DEFINITION 1.3. We say that G is of even t y p e if and only if the following
conditions hold:
(1) Every element of £ 2 ((2) is a C2-group;
(2) 02'(CG{X)) = 1 for every involution x of G\ and
(3) m2(G) > 3.
The simple groups of Lie type of characteristic 2 satisfy (1) and (2) of Definition
1.3 (with L2{G) empty) and apart from L 2 (4), C/3(4), L 3 (2), Sp 4 (2)', and G 2 (2) ; ,
they also satisfy (3).
We shall introduce in section 8 below a slightly stronger notion, that of "re-
stricted" even type. By definition G is of restricted even type if and only if it is
of even type and it satisfies some additional conditions related to specific small
elements of ^ ( ^ O - We shall then say that G is of odd t y p e if and only if it is not
of restricted even type; and it is the dichotomy
restricted even type / odd type
which separates three of the special cases in the classification grid from the other
three. However, for almost all purposes the reader need only keep in mind the
simpler definition of even type.
DEFINITION 1.4. Suppose that G is of even type and let S G Syl2(G). Then
3Vt(S) = M(G; 5) is the set of maximal 2-local subgroups of G containing NG(S).
Also Mi (5) = Mi(G; S) is the set of maximal 2-local subgroups M of G such that
\S:SHM\ <2.
DEFINITION 1.5. Let G be of even type and let S e Syl2(G). Then
a(G) = {p \p is an odd prime and mp(M) > 4 for some M G M i ( 5 ) } .
It would be more elegant to use M(5) in the definition of cr(G), but there are
technical advantages to our definition.
DEFINITION 1.6. If G is of even type with cr(G) empty, we say that G is
quasithin 1 (or of quasithin type).
Next we define some notions related to uniqueness subgroups.
DEFINITION 1.7. Let p be a prime, P a p-subgroup of the group X, and k an
integer such that 1 < k < mp(P). Then
rPtk(X) = (NX(Q) I Q < P, mp(Q) > k).
Moreover, Y°P2(X) is the following large subgroup of Tpi2(G):
T°Pt2(X) = (NX(Q) | Q < P, mp(Q) > 2, mp(QCP(Q)) > 3).
DEFINITION 1.8. Let p be a prime divisor of the order of the group X. A

strongly p-embedded subgroup of X is a proper subgroup Y of X such that
T P , I ( X ) < Y for some P G Sylp(X).
The definition has several equivalent formulations. For example, a subgroup Y

of X is strongly p-embedded in X if and only if Y = Nx (Y) and in the permutation
representation of X on the set Q, of X-conjugates of Y, every element of X of order
p fixes exactly one point of ft, but Q itself has more than one point.
As is customary, when p = 2, we suppress p and say simply that Y is strongly
embedded in X.
DEFINITION 1.9. A 2-uniqueness subgroup of G is a proper subgroup Y of
G such that for some S G Syl2(G),
(a) If m 2 (5) > 3, then T°S2(G) < Y; and
(b) If m2(S) = 2, then Ts]i(G) < Y.
As remarked earlier, we must have m2(S) > 2 in any case.
DEFINITION 1.10. Suppose that p G cr(G). Then a p-uniqueness subgroup
of G is a proper subgroup Y of G such that Yp^2{G) < Y for some P G Sylp(G).
In our analysis of groups of even type, when cr(G) ^ 0 we endeavor to construct
p-uniqueness subgroups for various p G cr(G) which satisfy some further similar
conditions. Such subgroups will be called strong p-uniqueness subgroups; the
definition of this term is not given until section 8 below, as it requires considerable
technical detail. We then define
x
This is a new usage of "quasithin"; see the Introduction to the Series.
3. T H E C L A S S I F I C A T I O N G R I D 83
D E F I N I T I O N 1.11. If G is of even type, then a0(G) is the set of all p e a(G)

such that G does not possess a strong p-uniqueness subgroup.
On a first reading, the reader may well ignore the word "strong" and consider
(Jo{G) as the set of all primes p G cr{G) for which Tp2{G) = G for some (hence all)
P € Sylp(G).
Several additional important terms were mentioned in the preceding chapter
without being defined precisely, and we prefer to defer their exact definition until
later in the chapter: the sets Qp and T p (section 13); vertical neighborhood (section
25); G « G* (sections 9, 19-23 and 26).
2. The uniqueness grid

As described in the preceding chapter, the uniqueness grid consists of a va-
riety of results about the existence in G of strongly p-embedded subgroups or p-
uniqueness subgroups for suitable primes p, or subgroups satisfying similar condi-
tions. Some of these results determine the possibilities for G when such subgroups
exist, and others establish sufficient conditions for G to contain such a subgroup.
A few comments concerning the uniqueness grid are in order, applicable to the
classification grid as well. Each row in the grid represents a theorem about G. The
first entry of the row indicates the hypothesis on G] the remaining entries, from
left to right, are the results of the successive stages of the proof. At certain points,
a bifurcation occurs-that is, the specified conclusions at a particular stage of the
analysis consist of two distinct alternatives. Both are listed in the grid and the
subsequent stages of the analysis of each are separately included. The listing of
the stages of the proof of a uniqueness result can terminate in several ways: (a)
G is shown to be isomorphic to some known simple group G*, or to satisfy some
other desired conclusion (the existence of a standard component or a strongly p-
embedded subgroup, for example); (b) the conclusion at a given stage is identical
to that reached at some stage of a previously listed uniqueness result, or (c) a
contradiction is reached.
We note that the analysis of groups of even type with |M(5)| = 1 uses re-
sults concerning 2-component preuniqueness subgroups. As a consequence, our
p-component preuniqueness theorems will be established before those on groups
with |M(5)| = 1.
The precise statements of the theorems represented by the uniqueness grid will
be given in sections 4-8 below, and the stages of their proofs will be discussed in
sections 9-11.
3. The classification grid

In this section we present the seven-case classification grid, but again begin
with a few preliminary comments. In the uniqueness grid, the points of bifurcation
arise from exceptional configurations, the conclusion at such a stage including a
"singular" solution. This phenomenon occurs in the classification grid, too, but
most of the bifurcations occur for a more significant reason.
For example, when G is of £!B2-type (case 2 of the special odd type portion
of the grid), the ultimate target groups G* include (a) most of the groups of Lie
type of odd characteristic and Lie rank 1, (b) most of the groups of Lie type of
odd characteristic of Lie rank 2 (plus certain linear groups of Lie rank 3), and (c)
A? and several sporadic groups. On the other hand, the definition of the term
UNIQUENESS GRID
Hypothesis stage 1 stage 2 stage 3 stage 4
p=2 |
G has strongly G is doubly M is 2-trans. of G^ 1

Suzuki type: £ 2 (2 n ) ,
^-embedded transitive on strongly
subgroup M the cosets of M embedded G w L 2 ( 2 n ) , Sz(2"),
Sz{2n),U3(2n) ^3(2")
G has a a) M is str.
G has a proper
2-uniqueness ^-embedded
2-generated core
subgroup M b) G w Ji b) G ^ Ji
G even type, a) G has a strongly
strongly closed ^-embedded subgroup
abel. 2-subgrp b) G « Ji
G even type,M M is strongly
weakly Z-emb. 2,-embedded
G of even type, a) G has a near a) C has a a) contra-
|M(5)| = 1; component 2-local standard diction
quasithin or uniqueness subgroup near
uniqueness b) G has a 2-unique- component
case ness subgroup
p arbitrary; M, K satisfy p-component preuniqueness hypothesis (6.1)
a) M is strongly
M controls p-embedded
rank 1 fusion b) p odd, M is almost
strongly p-embedded
a) M controls
M controls rank 1 fusion
rank 2 fusion b) p odd, M is almost
strongly p-embedded
\M not standard M controls rk 2 fusion
K is terminal, a) G has a strongly a) contra-
m2(K)>l, embedded subgroup diction
p=2 b) K is standard
p odd; G is of even type, cr(G) ^ 0
p-uniqueness {2,P}- a) 2-amalgam a) contra-
subgroups uniqueness type, or diction
Vp G <T(G) subgroups b) |M(S)| = 1
3. THE CLASSIFICATION GRID 85
CLASSIFICATION G R ID
Case stage 1 stage 2 stage 3 stage 4 stage 5
Special Odd Type
odd order a*(G)- {p,q}- contra-

1. G has
uniqueness uniqueness parabolic diction
odd order
type type type
a)G»Mi2,
Mc, Ly, O'JV,
L 3 (g), ^ odd
2-terminal b) 2-central G*-
2. G has Gr r& Gr G^G*
£232-type G*-type type, ra2(G*) > 3
G* G %W G* G K (2 )
G* e ft(2)* c) 2-maximal G*-
type mod cores,
m2(G*) = 2
(G* G X^2)*)
3. G has 2-terminal G«G* G^G*

£72-type £7 2 -type G* G ft(3) G* G X^3)
Special Even Type
4. G has 2-amalgam
2-amalgam GnG* G^G*
quasithin G*-type,
type G* G OC^) G* G 3CW
type G* e X ( 4 )
5. G has wide quasisymplectic G«G* G^G*

<Cep-type £e p -type type (p = 3) G* G OCW G* G ft(5)
6. G has p-terminal contradiction

£7 p -type £7 p -type
Generic Type
7. G has proper
|-balanced semisimple Gr W Gr G^G*
generic semisimple
type type G* e 3C G* € 3C<7)
(7)
type type
G & G* differs sharply according as G* is one of the groups in (a), (b), or (c). As
a consequence, the paths leading to the identification of G inevitably diverge into
three subcases at some stage of the analysis.
Next, as remarked earlier, the conclusions derived at the various stages are
based on the earlier uniqueness theorems. For instance, in the case that G is of odd
type, the critical such result is that G has no 2-uniqueness subgroup. When G has
even type, on the other hand, the corresponding result—that G has no strong p-
uniqueness subgroup for the odd prime p on which we focus—is arranged in advance
by our choosing p E O~Q(G) if possible. When 00(G) = 0, then Theorem U(cr) (row
10 of the uniqueness grid) implies that CF(G) = 0, so that G is quasithin.
The seven-case division is intended to split the classification proof into essen-
tially disjoint parts, so that each known simple group—except for the groups L 2 (2 n ),
Sz(2n), U3{2n) and J i , which possess 2-uniqueness subgroups- -occurs exactly once
as a target group for G.
The term "essentially disjoint" requires some clarification. For compactness of
exposition, it is convenient to have a designation for the sets of simple groups that
occur as solutions for G in each of the seven cases. We introduce in Definition 3.1
the important partition of the set of simple X-groups into the sets X^,... , X^
of target groups of the seven cases of the classification grid, and one additional set
X^ consisting of the simple groups arising from the uniqueness grid.
The separation of cases is almost but not quite absolute. In a few cases (i.e.,
rows of the grid), the full hypothesis of the row is needed only at a particular
stage of the analysis, so that subsequent results established in that case will be
more generally valid. For example, the special odd type assumption when G is of
£23 2 -type of 2-rank > 3 (case 2 of the grid) is used only in stage 1. Hence, once it
is shown that G is of 2-terminal 3C(2)-type, the subsequent analysis applies to an
arbitrary 3C-proper simple group—in particular, even if G is of generic type. This
means that results about groups of 2-terminal 3C(2)-type will be applicable to the
analysis of the generic case. It is for this reason that the clearest flow chart of the
proof proceeds from the special to the generic cases.
Likewise the solutions for G in cases 2 and 3 include a few groups of even type,
as will be seen below from the definition of the sets X^ and X^3\ despite the
stated assumption that G is special of odd type. Again this arises as a matter of
convenience in the proof, and is codified in the overall logic by the formulation of
a sharper but ad hoc notion of "restricted even type" (see section 8). These extra
groups of even type are not of restricted even type. However, for simplicity, we
omit all such refinements from the grid and confine them to the definitions given
in the text.
DEFINITIO N 3.1. We partition the set of simple IK-groups into the following
eight disjoint subsets X^\ 0 = {£ 2 (2 n ), *73(2n), Sz{2n) \ n > 2} U {J 1 }.
X™ = 0.
X& = {L2(q) (q > 5), L3(q), U3(q), PSpA{q){q > 3), U{q){q £ 1 mod 8,
q > 3), U4(q)(q ^7 mod 8, q > 3), G2(q)(q > 3), 2G2(q)(q = &n+\
q > 3), 3D4(q), with q odd throughout; A7, M1U M12, Mc, Ly, ON}.
X&= {A9, A10, An}.
XW = {L 3 (2 n )(n > 1), L 4 (2 n )(n > 1), U4(2n)(n > 1), U5(2n)(n > 1),
PSp4(2n)(n > 1), PSp6(2n)(n > 1), G2(2n)(n > 1), 2FA{2n)f(n > 1),
2
D4(2n)(n > 1), 3 £> 4 (2 n )(n > 1), L 5 (2), L 6 (2), L 7 (2), C76(2), Sp 8 (2),
£>4(2), F 4 (2), L 4 (3), *74(3), G 2 (3), A12, M 2 2 , M 2 3 , M 2 4 , J 2 , J 3 , J 4 ,
HS, He, Ru, Suz, Co3, Co2, F 5 , F3}.
X&= {Pn7(3),Pnt(3),2D5(2),U7(2),2E6(2),Fi22,Fi23lFi,24,CouF2lF1}.
X^ = 0.
X^ = {All groups of Lie type not already listed} U {An \ n > 13}.
4. 2 - U N I Q U E N E SS S U B G R O U P S 87
Of the twenty-six sporadic groups, note that J\ arises in a uniqueness theorem,

and the other twenty-five arise in only one special odd and two special even cases:
£2 2 -type : M1X,M12, Mc, Ly, ON
Quasithin type : M2 2, M23, M24, J2, Jz, J±, HS, He, Ru, Suz, Co 3 , Co2l F 5 , F3
£C 3 -type : Fi22, Fi23, Fif24, Co1 ,F2,F1.
The classification grid represents the sequence of seven theorems:
T H E O R E M 6» (i = 1 , . . . , 7). Let G be a X-proper simple group satisfying the
assumptions of Case i, 1 < i < 7. Then G = G* for some G* e X^l\
A more exact statement of each of the seven Theorems 6^, 1 < i < 7, is given
below in sections 14-16. Of course, Theorem Ci is a Background Result.
This series of monographs is taken up with the proofs of Theorems 6^, 2 < i < 7,
as well as of the various supporting theorems listed in the uniqueness grid. Together,
these results and the Background Results will constitute a proof of the Classification
Theorem.
A few comments about our notational conventions may help to clarify the
rather formidable array of terms appearing in the grids. First, the major cases into
which the analysis is divided are denoted by specifying that the given X-proper
simple group G is of a particular t y p e . Thus we have the terms "uniqueness type",
"even type", "quasithin type", "generic type", and "special type". Moreover, script
characters are used to specify various sets of quasisimple X-groups, some of which
may occur as proper sections of £?, such as Cp/-groups, Sp -groups, and T p-groups
(arbitrary p) and 232-groups (p = 2); others of which are target groups (the X^).
Script letters are also used to denote subcases; for example, among the special type
cases, we have £T p -type (arbitrary p), £!B2-type (p = 2), and £Cp-type (p odd),
indicating that certain elements of &P(G) lie in the appropriate set 7p, tB2, etc.
B. T H E UNIQUENESS GRID
4. 2-uniqueness subgroups
In the next eight sections we give a more detailed description of the uniqueness
grid. We begin with the statements of the principal theorems to be proved, covering
the p = 2 portion of the grid in this section.
The notion of a strongly Z-embedded subgroup, first considered by Aschbacher
in his study of groups having a proper 2-generated core [A3], represents a nat-
ural generalization of the notion of a strongly embedded subgroup. Underlying
this extension is the following basic proposition concerning groups with a strongly
embedded subgroup (cf. [Gl, §9; Be3]). We limit its statement to our group G.
Recall first that if X < G and T is a p-subgroup of X, p a prime, then X is
said to control the G-fusion (resp. strong G-fusion) of the subset A of T if
and only if for every g e G such that A9 C T, there i s x G l such that A9 = Ax
(resp. a9 = ax for all a e i ) ; and X is said to control (strong) G-fusion in T if
and only if X controls the (strong) G-fusion of every subset of T. Thus X controls
G-fusion in T if and only if any two subsets of T are conjugate in G if and only if
they are conjugate in X.
PROPOSITION 4.1. If M is a strongly embedded subgroup of G, then the fol-
lowing conditions hold:
(i) CG(x) < M for every x e T2{M);
(ii) M controls G-fusion in a Sylow 2-subgroup of M;
(iii) G and M each have only one conjugacy class of involutions; and
(iv) IfH < G with \HnM\ even and H £M, and if we set Hi = (12(H)), then
Hi ^ M and Hi DM is strongly embedded in Hi.
In particular, (iii) implies that every involution of M is 2-central in G, that
is, lies in the center of a Sylow 2-subgroup of G.
We note also that if M is a proper subgroup of G of even order satisfying
conditions (i) and (ii), then it is easily shown that M is strongly embedded in G.
Strong ^-embedding is defined by analogous conditions on a subgroup M of G,
relative to a single conjugacy class Z of 2-central involutions of G. For any subset
X of G, we define
zx-2;nx
In particular, if X is a subgroup of G, then (Zx) < X.
DEFINITION 4.2. Given a conjugacy class Z of 2-central involutions of G, we
say that a proper subgroup M of G is strongly ^-embedded in G if and only if
the following conditions hold:
(1) £ M ^ 0 ;
(2) CG(z) < M for every z € ZM]
(3) M controls G-fusion in a Sylow 2-subgroup of M; and
(4) If H is a proper subgroup of G with ZHDM nonempty and H jt M, and if
we set Hi = (ZH), then Hi ^ M and Hi D M is strongly embedded in Hi.
By the proposition, if M is strongly embedded in G, then M is strongly Z-
embedded with Z = 22(G).
The first row of the uniqueness grid gives the flow chart for the fundamental
Aschbacher-Bender-Suzuki theorem classifying groups with a strongly Z-embedded
subgroup. [As with our other results, it will be proved only under our assumption
that G is a OC-proper simple group.]
Because of the importance of the strongly embedded subcase, we split the
statement of the principal result into two separate theorems. Proper subgroups
of G containing a strongly (Z-)embedded subgroup are easily seen to be strongly
(Z-)embedded, so there is no loss in taking M maximal.
T H E O R E M SZ. (Uniqueness grid, row 1, stages 1 and 2) Let M be a maximal
subgroup of G. If M is strongly Z-embedded in G for some conjugacy class Z of
2-central involutions of G, then the following conditions hold:
(i) M is strongly embedded in G; and
(ii) G acts doubly transitively by conjugation on the set of G-conjugates of M.
This result covers the first two stages of the analysis. In particular, it shows
that if G contains a strongly embedded subgroup M with M a maximal subgroup
of G, then G is necessarily doubly transitive on the set of G-conjugates of M.
4. 2-UNIQUENES S S U B G R O U P S 89
The second result deals with the strongly embedded case and thus covers the
remaining stages of the analysis of groups with a strongly ^-embedded subgroup.
T H E O R E M SE. (Uniqueness grid, row 1, stages 3 and 4) Let M be a maximal
subgroup of G. If M is strongly embedded in G, then for some n > 2, G = Z/2(2n),
Sz(2n) or Uz(2n), and M is a Borel subgroup of G.
Next is an extension of Aschbacher's proper 2-generated core theorem [A3].
T H E O R E M U(2). (Uniqueness grid, row 2) Let M be a maximal subgroup of G.
If M is a 2-uniqueness subgroup, then one of the following holds:
(i) M is strongly Z-embedded in G for some conjugacy class Z of 2-central
involutions of G (and hence G ^ L2{2n), Sz{2n) or U3(2n)); or
(ii) G ^ J i .
The first part of the proof shows that the hypothesis r ^ 2 ( G ) < M leads to
the stronger condition Ts,2(G) < M if rri2(S) > 3. This stage of the analysis was
originally investigated by Harada [H2]. Now since Ts,2(G) is a proper subgroup
of G, G has a proper 2-generated core by definition, and like Aschbacher we verify
the hypotheses of Theorem SZ for G.
Finally, with the aid of Theorem SZ, we establish two additional important
results needed for the classification of groups of even type. The first of these
is Goldschmidt's strongly closed abelian 2-group theorem [Go5], limited here to
groups of even type. Recall that a subgroup A of 5 G Syl2(G) is said to be
strongly closed in 5 with respect to G if and only if whenever a e A, g £ G
and a9 e 5, then a9 G A.
T H E O R E M SA. (Uniqueness grid, row 3) Assume that G is of even type and
let S G Syl2{G). If S contains a nontrivial abelian subgroup that is strongly closed
in S with respect to G, then one of the following holds:
(i) G contains a strongly Z-embedded maximal subgroup M for some conjugacy
class Z of 2-central involutions of G (whence again G = L2(2n), Sz(2n) or
U3{2n)); or
(ii) G^JX.
The second result is Holt's theorem in the particular case of groups of even
type [Hoi].
T H E O R E M SF. (Uniqueness grid, row 4) Assume that G is of even type, S e
Syh(G), and M is a proper subgroup of G containing S with the following proper-
ties:
(a) M controls G-fusion in S; and
(b) CG(Z) < M for some involution z of Z(S).
Then one of the following holds:
(i) M is strongly Z-embedded with respect to the conjugacy class Z of involutions
containing z (whence G ^ L2(2n), Sz(2n) or U3{2n)); or
(ii) GÂ9.
Because G is assumed to be of even type, the possibilities G = An, n odd,
n > 9, of Holt's theorem are excluded here. In the particular circumstances when
Theorem SF is applied, the AQ conclusion will be clearly impossible. Hence we
do not include A$ in %(°\ Accordingly, the term "G has a weakly Z-embedded
subgroup M" in the uniqueness grid means by definition that the hypotheses of
Theorem SF hold, but G ^ A9.
5. Groups of restricted even typ e with |M(5)| = 1

Application of the amalgam method in a group G of even type requires a Sylow
2-subgroup S of G to be contained in at least two maximal 2-local subgroups. Thus
before using amalgams, one must first dispose of the case in which |M(5)| = 1. In
this case, if we let M be the unique maximal 2-local subgroup of G containing
NG(S), it follows that
(5.1) NG(T) < M for every 1 ^ T < S with T< NG(S).
In particular, it follows that
, x C(G, S) < M, where
v ;
C{G, S) = (NG{T) 11 ^ T char S)
is the "characteristic 2-core".
Thus the condition |3VC(5)| = 1 implies that the corresponding maximal 2-
local subgroup M satisfies (5.2), a weak type of 2-uniqueness. In the original
classification proof this situation was analyzed for arbitrary groups of characteristic
2 type. Fortunately, it turns out that the analysis is required for us only under
certain conditions on the p-structure of G for odd primes p, namely when either
&{G) is empty, or there are odd primes p for which mp(M) > 4 and for each such
p, M is a p-uniqueness subgroup. [By definition any such p lies in cr(G).} The
objective is again to prove that M must be strongly embedded in G, so that G is
determined from Theorem SE.
We can therefore state the principal result to be proved in this connection as
follows.
T H E O R E M M(S). (Uniqueness grid, row 5) Assume that G is of restricted even
type with \M(S)\ = 1 for S E SyfaiG), and let M be the unique element ofM(S).
Suppose either that &(G) = 0 or that there exist odd primes p for which mp(M) > 4,
and for every such p, M is a p-uniqueness subgroup. Then M is strongly embedded
in G (whence G =* L2(2n), Sz(2n), or U3(2n) for some n > 2).
[The definition of restricted even type will be given in section 8.]
6. Component preuniqueness subgroups

We next describe the portion of the grid that deals with p-component pre-
uniqueness subgroups, a notion we proceed to define.
DEFINITION 6.1. Let M be a proper subgroup of G, p a prime, and set M =
M/Op'(M). We say that M is a p-component preuniqueness subgroup of
G (with respect to K) if and only if M contains a p-component K such that
CG{x) < M for every x € XP(CM(K)).
[The p-component K will usually not have to be mentioned as its identity will
be clear from the context. Note that if K is a component of M, then ZP(CM{K)) =
XP(CM(K)).}
An important example of such a preuniqueness subgroup arises in connection
with a terminal component K in G (Definition 1.2). Indeed, let Q be the set of
G-conjugates of K and define a relation ~ on Ct by setting / ~ J for 7, J e ft if
and only if either [7, J] = 1 or I = J. Using the condition of terminality, it is easy
to show that ~ is an equivalence relation on Ct. We denote by [K] the product
6. COMPONENT PREUNIQUENESS SUBGROUPS 91
of all elements in the equivalence class containing K, so that [K] is a semisimple

group—i.e., a central product of quasisimple groups. Moreover, K is a component
of [K]. Again using the terminality of if, one directly obtains the following result.
PROPOSITION. If we set M = NG([K]), then K is a component of M and
CG(X) < M for every x G TP(CM{K)), SO that M is a p-component preuniqueness
subgroup. Moreover, M permutes the components of [K] transitively by conjugation.
We consider various conditions on a p-component preuniqueness subgroup M.
DEFINITION 6.2. Let p, M and K be as in Definition 6.1, let Q be a Sylow
p-subgroup of CM{K)I a n ^ expand Q to P 6 Sylp(M). We say that M controls
rank 1 fusion (with respect to K) if and only if M controls the strong G-
fusion in P of every element of TP(Q), and that M controls rank 2 fusion (with
respect to K) if and only if M controls the strong G-fusion in P of every Epi-
subgroup of Q. In addition, we say that M is standard (with respect to K) if
K is normal in M.
Use of the term "standard" here is justified by the fact that if K is terminal,
then K is standard in G if and only if K = [K] and hence if and only if K <
NG{[K\).
As the terminology suggests, we shall aim to prove that a p-component pre-
uniqueness subgroup is actually a p-uniqueness subgroup (Definition 1.10), indeed
that it is strongly p-embedded in G. However, since we wish our p-component pre-
uniqueness results to be valid for arbitrary p, we prefer not to impose at this time
the additional assumption that G is of even type when p is odd (which is the only
situation in which the odd prime results are to be applied). Unfortunately, without
this assumption, it turns out not to be possible to establish strong p-embedding in
every case. As a consequence, for odd p we are forced to introduce the notion of an
almost strongly p-embedded subgroup M, which incorporates the exceptional
configurations.
The precise definition is rather technical, and will be given in Definition 8.4. It
should be remarked that part (4) of that definition arises not from the preuniqueness
results we are discussing, but in the study of generic groups. Hence we only discuss
the remaining alternatives (l)-(3) of the definition here. In addition to the condition
that M be a p-uniqueness subgroup, the definition has essentially three critical parts
in the exceptional configurations. The first restricts the isomorphism type of K:
if m(Q) > 2, the only case of interest at the moment, then mp(K) = 1. The
second gives the structure of F*(M) and K = (K ), while the third gives the
approximation to strong p-embedding that can be proved in these residual cases,
asserting at least that TR^(G) < M for some explicit large p-subgroup R of M.
To state our p-component preuniqueness results, we assume the following con-
ditions, which we term the p-component preuniqueness hypothesis:
(1) M is a p-component preuniqueness subgroup of G with respect to K;

(2) M is a maximal subgroup of G\
K }
(6.1) _ _
(3) mp{CM(K)) > 2, where M = M/Op,(M)\ and
(4) If p is odd, then mp(M) > 4.
Under these assumptions we shall prove

T H E O R E M P U I . (Uniqueness grid, row 6) Assume the p-component preunique-

ness hypothesis. If M controls rank 1 fusion with respect to K, then one of the
following holds:
(i) M is strongly p-embedded in G; or
(ii) p is odd and M is almost strongly p-embedded in G.
T H E O R E M PU2. (Uniqueness grid, row 7) Assume the p-component preunique-
ness hypothesis. Further, if p = 2, assume that rri2(K) > 2. If M controls rank 2
fusion with respect to K, then M controls rank 1 fusion with respect to K (and so
the conclusion of Theorem PU\ holds).
T H E O R E M PU3. (Uniqueness grid, row 8) Assume the p-component preunique-
ness hypothesis. Further, if p = 2, assume that rri2(K) > 2. Then one of the
following holds:
(i) M is standard; or
(ii) M controls rank 2 fusion with respect to K.
Together, these three results have the following fundamental consequence.
THEOREM P S . Assume the p-component preuniqueness hypothesis. Further, if
p = 2, assume that rri2{K) > 2. Then one of the following holds:
(i) M is standard;
(ii) M is strongly p-embedded in G; or
(hi) p is odd and M is almost strongly p-embedded in G.
Finally, by combining Theorems PS and SZ and using the fact that in the
groups £2(2"), Sz(2n) and U${2n) all centralizers of involutions are solvable, we
obtain the basic Aschbacher-Gilman theorem [A4, Gil ] that terminal components
of 2-rank at least 2 are necessarily standard for p = 2.
T H E O R E M TS. (Uniqueness grid, row 9) Let x G 22(G) and let K be a com-
ponent of CG(X) with K terminal in G. If rrt2(K) > 2, then K is standard in
G.
7. The odd uniqueness theorem

The definition of strong p-uniqueness subgroup is limited to the case in
which G is of even type with o~(G) ^ 0 and p G o~(G), so that in particular it
follows then from the definition of cr(G) that mp(G) > 4. As the definition is
technical, requiring that a p-uniqueness subgroup satisfy some extra conditions,
we defer it until the next section and simply state the main theorem of the odd
uniqueness case.
T H E O R E M U ( < J ) . (Uniqueness grid, row 10) Assume that G is of even type with
cr(G) 7^ 0. If G contains a strong p-uniqueness subgroup for each p G cr(G), then
\M{S)\ = 1 for any S G Syl2(G).
Since a(G) is empty when G ^ L 2 (2 n ), 5z(2 n ), or U3(2n), Theorem M(5) then
implies that the hypothesis of Theorem U(cr) is never satisfied. Thus as a corollary
we obtain
COROLLARY U ( < J ). If G is of even type with a(G) ^ 0, then o~0(G) ^ 0. That
is, for some p G o~(G), G does not contain a strong p-uniqueness subgroup.
8. SOME TECHNICAL DEFINITIONS 93
Theorem U(cr) is analogous to Aschbacher's uniqueness theorem [A16] but

clearly there are significant differences in the hypotheses.
In the even type case of the classification proof, except in the quasithin case,
we work with primes p G CTQ(G) and attempt to construct strong p-uniqueness
subgroups to reach a contradiction. The preuniqueness results stated in the previous
section are useful for this purpose.
8. Some technical definitions

For completeness we include here the full definitions of the terms "almost
strongly p-embedded", "strong p-uniqueness subgroup", and "restricted even type".
This section can be omitted on a first reading.
As indicated earlier, of the three exceptional sets of conditions defining almost
strong p-embedding, two represent the best conclusions short of strong p-embedding
that we can prove about p-component preuniqueness subgroups in Theorem PUi;
the third represents a configuration reached in the analysis of G in the generic case
when G has a strongly closed subgroup of order p.
In the next four definitions, we assume that M is a p-component preuniqueness
subgroup with respect to K, and set M = M/Op'(M) and let Q G Sylp(CM{K))-
We let P G Sylp(M) with Q < P.
DEFINITION 8.1. M is wreathed if and only if the following conditions hold:
(1) p>5;
(2) mp(K) = 1;
(3) K* = (KM) has exactly p p-components;
(4) K* = F*(M);
(5) \M:K*\P = P]
(6) If we set R = P n K*, then TRA(G) < M; and
(7) r P f 2 (G) < M .
DEFINITION 8.2. M is almost p-constrained if and only if the following
conditions hold:
(1) p is odd;
(2) mp(K) = ll _
(3) F*(M) = X x Q ;
(4) P = T x Q, where T = PnK;
(5) CG{T) has a p-component J such that Q e Sylp(J) and J/Op>(J) =
L2(pn) {n > 3), U3(pn) (n > 2), or 2 G 2 (3 n ) (n > 3, p = 3); and
(6) r P | 2 ( G ) < M and TQA(G) < M.
DEFINITION 8.3. M is of strongly closed t y p e if and only if the following
conditions hold:
(1) p is odd;
(2) Keehev(2);
(3) mp(Q) = 1;
(4) Qi(Q) is strongly closed in P with respect to G;
(5) If X 3, then NG(X) < M; and
(6) r P , 2 (G ) < M.
It is a quick consequence of the above definitions and simple properties of DC-
groups that if M is of one of the three above types, then K is determined up to
conjugacy in M as the p-component of M of largest p-rank. Hence the terms need

not include the phrase "with respect to if".
Now we can define almost strong p-embedding.
DEFINITION 8.4. Let M be ap-component preuniqueness subgroup of G. Then
M is almost s t r o n g l y p - e m b e d d e d in G if and only if one of the following holds:
(1) M is strongly p-embedded in G;
(2) M is wreathed;
(3) M is almost p-constrained; or
(4) M is of strongly closed type.
Now we turn to the notion of s t r o n g p-uniquenes s s u b g r o u p , which incorpo-
rates those conclusions which we are able to prove at various points of our analysis
of G by signalizer functor theory and p-local analysis, but—in the case G is of even
type and p € CF{G)—before any elaborate analysis of 2-local subgroups of G.
First, when p is odd and &P(G) C Qp (this is part of the £C p -case, the condition
being an analogue for odd p of one of the conditions defining even type), a p-
uniqueness subgroup is obtained not as a p-component preuniqueness subgroup,
but purely from signalizer functors. We now formulate its key properties. First, we
define:
DEFINITION 8.5. For any group X, prime p, and positive integer n,
&*(X) = {E < X | E 9* Epn and mp(Cx(E)) = n}, and

S%Y) = U~ =1 SS(X).
Now we can define:

DEFINITION 8.6. Let M be a proper subgroup of G, p a prime, P e Sylp(M),
and assume that mp(P) > 4. Then M is an £ C p - u n i q u e n e s s s u b g r o u p of G if
and only if the following conditions hold:
(1) r P | 2 ( G ) < M;
(2) Opf(M) has even order, and if T e Syl2(Op'(M)), then CM{T) is ap'-group;
and
(3) For any A e S P (M), every ^-invariant p'-subgroup of M lies in Op>(M).
Finally, we can define the term "strong p-uniqueness subgroup".
DEFINITION 8.7. Let p be an odd prime and M a proper subgroup of G with
mp{G) > 4. Then M is a s t r o n g p-uniquenes s s u b g r o u p if and only if (1) and
(2) hold:
(1) Either M is a p-component preuniqueness subgroup which is almost strongly
p-embedded in G, or M is a £C p -uniqueness subgroup of G\ and
(2) Either (a) or (b) holds:
(a) Op\M) ± 1, but Op\X) = 1 for all X < G with M < X;
(b) M = NG{K) for some K < G with K G ehev(2), mp(CM{K)) < 1, and
M is almost strongly p-embedded in G.
In condition (2b), the almost strong p-embedding condition is actually redun-
dant, i.e., M cannot be a £C p -uniqueness group. For it follows easily from the
condition mp(M) > 4 that P C\ K ^ 1. If M were an £C p -uniqueness group, how-
ever, then [P fl K, Op>{M)\ would have even order by the definition, which would
9. GROUPS WITH A STRONGLY EMBEDDED SUBGROUP 95
contradict the quasisimplicity of K. A further observation in case (2b) is that if

^P(CM(^)) = 1> then M must be of strongly closed type.
Now we turn to the third technical definition, that of restricted even type. It
is used to record the fact that certain configurations in which G has even type
have already been ruled out by arguments from the special odd type case. For
example, as we shall see in section 12, C2 contains some Chevalley groups in odd
characteristic, namely L2(q) with q G 3\M9 (that is, q is a Fermat prime, Mersenne
prime or 9), whereas other .^(g), Q odd, lie in 62'. Thus, as might be expected, our
arguments in the case that G is of special odd type with some element of £>2(G) in
62' actually are strong enough to rule out the possibility, under certain conditions,
that G is of even type but &2(G) has an element isomorphic to L,2(q) for some
q G 9rM9. In particular a number of configurations involving 1^(5) and £2(9) can
be ruled out which would be awkward to handle from the point of view of a group of
even type. There are also some configurations involving £3(4) which can similarly
be ruled out. The definition of restricted even type incorporates the assumption
that these configurations have already been handled.
DEFINITION 8.8. We say that G is of restricted even typ e if and only if the
following conditions hold:
(1) G is of even type;
(2) Suppose that x G 22(G) and K is a component of CG{X) such that K =
L2(q) with q G 3TVC9. Then we have q = 5, 7, 9 or 17, and m2(G) > 4;
(3) Suppose that x G ^ ( G ) and K is a terminal component of CG{X) such that
K 2* L2(q) with q G 3 ^ 9 . Then
(a) m2(CG(x)) > 4;
(b) x is not 2-central in G; and
(c) Either q — 5 and NQ{K) has Sylow 2-subgroups isomorphic to EIQ, or
q = 9 and 777,2(0?(if)) = 1; and
(4) Suppose that x and y are commuting involutions of G, K is a terminal
component of CG{x), K/Z(K) = £3(4), Z(K) ** 1, Z2 or EA, and y induces
an automorphism of unitary type on K (that is, CK/Z{K){V) — %(2))- Then
\G:NG(K)\2>4.
9. Groups with a strongly embedded subgroup

At this point we have stated the principal results to be proved in the uniqueness
grid. We next discuss the stages of proof and explain the as yet undefined terms
appearing in the grid.
In row 1, the term "doubly transitive of Suzuki type" is also the definition of
G w G* for G* ^ L 2 (2 n ), Sz(2n) or U3(2n), n > 2. It incorporates the conclusion
of the first stage of the proof of Theorem SE (equivalently, the third stage of the
classification of groups with a strongly Z-embedded subgroup).
Let G then satisfy the assumptions of Theorem SE. Thus G is a 3C-proper simple
group containing a strongly embedded subgroup M with M a maximal subgroup
of G. In view of Theorem SZ, we may assume that G is doubly transitive on the
set Vt of G-conjugates of M.
DEFINITION 9.1. We say that G is doubly transitive of Suzuki type (on

Q.) if and only if a two-point stabilizer D fixing M (i.e., D < M) has a normal
complement X in M (i.e., M — DX with X < M and D n X = l) satisfying the

following conditions:
(1) X is nilpotent;
(2) X contains a Sylow 2-subgroup of M (and hence of G); and
(3) X acts regularly on Q - {M}.
Bender's theorem asserts the following:
THEOREM SE: STAGE 1. G is doubly transitive of Suzuki type on ft.
In particular, G is a split (B, iV)-pair of rank 1 (with B = M) and odd degree,
since the regularity condition implies that \Q\ = \X\ + 1.
This is the starting point for Suzuki's work on doubly transitive permutation
groups, which we have included as a Background Result with the proof as revised
by Peterfalvi. With this result, the proof of Theorem SE is completed, and with it
the classification of groups with a strongly ^-embedded subgroup.
THEOREM SE: STAGE 2. If G is doubly transitive of Suzuki type, then G =
L2{2n), Sz{2n) or U3(2n) for some n > 2.
10. Theorem M(S)

We next discuss the stages of the proof of Theorem M(5) (row 4 of the unique-
ness grid). In particular, we assume that G is of even type.
The theorem being a variation of the global C(G, 5)-theorem, its proof also
follows the Aschbacher %-block method. However, because of possible confusion
of terms between Aschbacher's blocks and Brauer's blocks of characters, we shall
adopt alternate terminology here: the term "near component" in place of "block",
and "linear or alternating near component" in place of "x-block".
DEFINITION 10.1. Let X be a group, set P = 02(X), and assume that
(a) CX(P) < P (equivalents F*(X) = P ) ;
(b) X/P is quasisimple or simple (including the case of prime order); and
(c) X = 0\X).
We call X a near component provided the following additional conditions hold:
(1) U = [P,X] < Oi(Z(P)); and
(2) X acts nontrivially and irreducibly o n F = U/Cu(X).
We can regard V as a faithful irreducible F2 [X/P]-module and refer to V as the
associated module of X. In addition, a subgroup X of a group Y is called a
near component of Y if X is a near component and X is subnormal in Y.
The justification for this terminology is that near components of a group behave
very much like components of a group; in particular, they satisfy an analogue [Fool,
Foo2] of Z/2>-balance, which is a basic property of components.
DEFINITION 10.2. A near component X with associated module V will be
called linear (respectively alternating) if and only if X/02(X) = L2(2n) with
n > 2 (respectively An with n = 3 or n > 5) and V is a natural 2-dimensional
F 2 n-module for X/02(X) 9* L 2 (2 n ), considered as an F2[X/02(X)]-module (re-
spectively the nontrivial irreducible constituent of the standard F2 permutation
module for X/02(X) ^ An).
[The notion of near component arises in the local C(G, 5)-theorem [ A l l , A12],
which asserts that a group Y such that F*(Y) = 02(Y) is the product of all its
10. THEOREM M(S) 97
linear and alternating near components and C(Y,S) = (Ny(R) 11 ^ R char 5),
where S G Syl2(Y). The notion of a near component can also be extended to
arbitrary primes p. However, for p = 3 and X/P = SX 2 (3 n ), -P = Oa(X), in order
for an analogue of the local C(G, 5)-theorem to be true, it is necessary to allow X
to have a structure approximating that of a maximal parabolic subgroup of G2(Sn).
In particular, [X, P]/$([X, P]) is the sum of two natural X/P-modules and X acts
nontrivially on Qi(Z(P)). Such "G2(3 n )-type" near components arise in the course
of the classification of groups with semidihedral or wreathed Sylow 2-subgroups.]
Assume now that G satisfies the assumptions of Theorem M(5), so that G is
of even type, |M(S)| = 1 for S G Syl2(G), and if M is the unique element of M(5),
then either (1) cr(G) = 0 or (2) mp(M) > 4 for some odd prime p, and for any such
p, M is a p-uniqueness subgroup. [This last assumption is not required for Stage
1.] The first stage of the proof of the theorem proceeds as follows.
By Theorem U(2), we may assume that G does not contain a 2-uniqueness
subgroup. Theorem SA therefore implies that S does not contain a nontrivial
strongly closed abelian subgroup. It follows that M does not control G-fusion in 5,
otherwise Z(02{M)) would be such a nontrivial strongly closed abelian subgroup
of S. (By definition of JVC(S), M is a 2-local subgroup, so 02(M) ^ 1.) This
observation constitutes the first step in the analysis.
As a consequence, one can apply the Alperin-Goldschmidt conjugation theorem
[Gol] to conclude that there is 1 ^ D < S such that H = NQ(D) has the following
properties:
(1) H £ M;
(10.1) (2)F*(iJ)-02(iJ);and
(3) D contains every involution of CG(D).
We consider the set N(M) of all maximal 2-local subgroups N of G such that
there exists a 2-subgroup D ^ 1 of N f) M such that if we set H = NN(D), then
D, H satisfy the conditions of (10.1). [By the discussion preceding (10.1), N(M) is
nonempty.]
We define a preordering >: on the set 3sf(M) as follows: if TV, JVi G N(M) and
T G Syl2(N fl M), Ti G Syl2{Nx n M), we write N^Nx if and only if m2(T) >
m2(Ti), with \T\ > |Ti| in case m2(T) = m2{Ti). We denote the set of maximal
elements of N(M) under ^ by 7f*(M). Finally, for N G JT(M) and T G Syl2(N),
set
C{N, T) = (NN(T0) 11 ^ T0 char T).
The goal of Stage 1 is the following.
THEOREM M(5): STAGE 1. Under the given assumptions on G, either G has
a 2-uniqueness subgroup, or N*(M) ^ 0 and the following conditions hold for any
N eW{M):
(i) F*(N) = 02{N);
(ii) M contains a Sylow 2-subgroup T of N;
(iii) C(N, T) < M; and
(iv) N contains a linear or alternating near component K with K j£ M.
We call M a near component 2-local uniqueness subgroup. Condition
(iv) follows directly from (i), (ii), (iii) and Aschbacher's local C(G,S) theorem, as
remarked above. The difficulty in establishing the critical condition (i) arises from
the presence of components in 2-local subgroups 2 . The following subsidiary result,

using Theorem TS, underlies the proof. Here N and T are as above.
T H E O R E M . Let x G X2(G), let L be a component of CQ(X), let P be a Sylow
2-subgroup ofCc(x), and assume
(a) L $£ M; and
(b) P<M and P±T.
Then L ^ L2(2n), Sz(2n) or U3(2n) for some n>2, and L is standard in G.
Next, let N G N* (M) and K be as in Theorem M(S) : stage 1. Our assumptions
on cr(G) are now used to restrict both the possibilities for K and the number r of
near components of N isomorphic to K: namely,
(1) If K/02(K) ^ Am, then m = 3, 5, or 9; and
(2) If K is nonsolvable, then r < 3.
[When K/02(K) = A m , Aschbacher's theorem implies that, in fact, m = 2 n + l
for some integer n, which explains why only the values m = 3, 5, or 9 occur here.]
As noted in Chapter 1, the analysis of this situation has been carried out by
Richard Foote. Underlying it is the following "standardness" theorem for the K
obtained in stage 1.
T H E O R E M M(5): STAGE 2. Assume the hypotheses and conclusions of Theo-
rem M(S): stage 1. Let Tx be a subgroup ofT with \T : Tx\ < 2 and 02(N) < T1}
and assume that if K is nonsolvable, then T\ normalizes K. If NQ is a maximal
2-local subgroup of G such that (K, T\) < iV0 and NQ >Z N, then K is a near
component of NQ.
We say that K is a standard near component. With this result Foote then
argues to a contradiction.
T H E O R E M M(5): STAGE 3. G cannot satisfy the assumptions and conclusions
of Theorem M(S): stage 1.
11. Theorem JJ(a)

Finally we describe the stages of the proof of Theorem U(cr). We assume
throughout that G is of restricted even type with o~(G) ^ 0 and that for each
p G cr(G), G contains a strong p-uniqueness subgroup Mp.
The goals of the first stage of the analysis are to reduce as much as possible
to the 2-local case F*(MP) = 02(Mp), and to strengthen both the p-uniqueness
and the 2-uniqueness properties of Mp. To describe the last of these, we need the
following terminology. Define Q(MP) to be the set of all p-subgroups Q of Mp such
that either mp(Q) > 3 or for some Epi subgroup A of Q, TA,I{G) < Mp.
DEFINITION 11.1. We say that Mp is a {2,p}-uniqueness subgroup of G if
and only if every 2-local subgroup of G containing an element of Q(MP) lies in Mp.
T H E O R E M U ( < J ): STAGE 1. The following conditions hold for each p G cr(G):
(i) Structure of Mp:
(a) 02'{MP) = 1; and
2
This difficulty arises from our assuming that G is of even type rather than of characteristic
2 type, as was assumed in the corresponding result in the original classification proof.
12. Cp-GROUPS 99
(b) Either F*(M P ) = 02(MP), or E(MP) e Ghev(2) with CMp(E(Mp))

having p-rank at most 1;
(ii) 2-uniqueness properties:
(a) Mp is a {2,p}-uniqueness subgroup of G; and
(b) M contains NG{S) for some S G Syl2(G);
(iii) Additional p-uniqueness properties (by definition Mp is a strong p-unique-
ness subgroup):
(a) Ifp = 3y then Mp is strongly p-embedded in G; and
(b) If p > 3, x G Ip(Mp), and CQ(X) % Mp, then LP>(CG(X)) = K satisfies
the conditions K/Op'(K) = L2(pn), n > 3, (x) is strongly closed in a
Sylow p-subgroup P of M with respect to G, P G Sylp(CG(x)), and
P = P0x(PDK)
with PQ cyclic.
In the original classification proof similar results were obtained by Aschbacher,
Gorenstein and Lyons [AGL1, GL1] and here our proof is similar. However we
achieve some simplification in the verification of the {2,p}-uniqueness property be-
cause of the definition of o~(G). The focus is on the elimination of 2-local subgroups
H such that HnMp contains both a Sylow 2-subgroup of H and an Ep2-subgroup.
For such a subgroup the p-uniqueness properties of M place severe restrictions on
the structure of H. The case mp(H fl Mp) > 4 is then relatively easy to rule out,
and once that has been done, the condition Mi (5) ^ 0 forces Mp itself to contain
a subgroup of index at most 2 in a Sylow 2-subgroup of G. This in turn eases the
analysis of the remaining possibility mp(H H Mp) < 3.
Once the preparatory uniqueness properties have been established, the proof
relies on the analysis of 2-amalgams and is the work of Gemot Stroth.
T H E O R EM U(<J): STAGES 2 AND 3. If G satisfies the hypotheses and conclu-
sions of Theorem U(a): stage 1, then |M(5)| = 1 for some S G Syl2(G).
If the conclusion of this theorem fails, Stroth makes a careful choice of p G or(G),
takes 5 G Syl2(Mp), and chooses a second element M G M(5). He then selects
certain subgroups N and Np of M and M p , respectively, containing S but with
02((N,NP)) = 1, and argues to a contradiction by analysis of the amalgam of Af
and Np. The {2,p}-uniqueness property is heavily used.
The conclusion of this theorem contradicts Theorem M(5), and so this com-
pletes the proof of Theorem U(cr).
C. T H E CLASSIFICATION G R I D :
GENERI C AND SPECIAL SIMPLE G R O U P S
12. Cp-groups
We now give an analogous description of the classification grid. We begin
with the definitions of the notions of the "generic" and "special" cases for G, which
underlie the basic three-part division of the grid. We then give the precise conditions
defining the seven rows of the grid and give the statements of Theorems 6*, 1 <
i<7.
Fundamental to the case division are partitions—one for each prime p—of the
set of quasisimple 3C-groups into three subsets. More precisely, for any p, recall
that Xp is the set of all quasisimple 3C-groups K such that Op>(K) = 1; we shall
define a partition
(12.1) Xp = epU7pUSp.
We begin with the definition of Cp. In view of Definition 1.3, this will complete
the precise definition of "even type." To specify which elements of Xp are to be in
Cp, we first do this for simple K and then consider the nonsimple case.
As remarked in the preceding chapter, Cp-groups are intended to include those
quasisimple X-groups considered to be "characteristic p-like." These include, of
course, all the simple groups in Qhev(p)—the groups of Lie type of characteristic p.
However, we are able to achieve certain economies by including some non-Qhev (p)
groups as well.
Unfortunately, the known quasisimple groups are too perverse to allow for a
neat conceptual definition that separates the elements of Cp (i.e., the Cp-groups)
from the other quasisimple groups. We shall therefore make the definition by sim-
ply listing the Cp-groups for each prime p. This will be followed by a conceptual
description which yields roughly the same set of groups.
DEFINITION 12.1. Let p be a prime. A simple X-group K is a C p -group if
and only if K G Ghev(p), or K = Ap, A2p, or A3p, or one of the following holds:
(1) p = 2 and either
(a) K = £2(3), Q £ 3^M9 (i.e., q a Fermat or Mersenne prime or 9);
(b) K & L 3 (3), L 4 (3), *74(3), or G 2 (3); or
(c) K = M n , M12, M 2 2 , M 2 3 , M 2 4 , J2, J3, J±, HS, Suz, Ru, Cox, Co2,
Fi22, Fi23, Fi'24,F3, F2, or Fx.
(a) K * U5(2), U6(2), Sp6(2), £>4(2), 3 ^ ( 2 ) , F 4 (2), 2 F 4 (2)', or P5p 4 (8);
or
(b) K ^ M n , J3, Cox, Co2, Co3, Mc, Ly, Suz, ON, Fi22, Fi23, Fi24, F 5 ,
F3, F2, o r F i .
(a) K ^ 2 F 4 (2)', 2 F 4 (32), or 5z(32); or
(b) K ^ J 2 , C01, Co2, Co3, HS, Mc, Ly, Ru, He, F5, F3, F2, or Fx.
(4) p = 7 and K ^ Cou He, ON, Fi24, F3, or F1.
(5) p=ll and if ^ J 4 .
Furthermore, a quasisimple DC-group K € Xp is a Cp-group if and only if K/Op(K)
is a Cp-group, with the following exceptions:
For p = 2: The groups SL2(q), q G 3 ^ 9 , 2A8, SX 4 (3), SU4(3), 5p 4 (3), and
[X]L 3 (4), X = 4, 4 x 2, or 4 x 4, are not included among the C 2-groups.
For p = 3: The groups 3^6 and 3CW are not included among the C3-groups.
Finally, a quasisimple DC-group K with Op>(K) = 1 is a C p >-group if K is not
a Cp-group; that is, Cp/ = Xp — C p.
In the above definition, we have not necessarily listed Chevalley groups of
ambiguous characteristic, since automatically all simple groups in Qhev(p) lie in
12. Cp-GROUPS 101
Cp. But we note that 62 contains 1/^(3), PSp 4 (3), and 2 G2(3)', since these are
isomorphic to £2(2)', 1/4(2), and £2(8), respectively, so lie in Ghev(2).
We shall not explicitly list the nonsimple Cp-groups at this time, but shall
introduce the notation to be used in connection with covering groups.
If K is a quasisimple group and Z < Z(K) with Z cyclic, we shall write
K 9* AK, where A = \Z\ and K = K/Z.
Thus, for example, SL 4 (3) = 2L 4(3) and 5^74(3) = 4t/ 4 (3) are not C2-groups, but
the group 2*74(3) 2* SU4(3)/Z2 is a C2-group.
On the other hand, if Z = Z\ x Z2 is noncyclic with Z\,Z2 cyclic, we shall
write
K = [Ax x A2]~K, where A* = \Z{\, i = 1,2.
Thus, for example, 2ftJ (3) = [2 x 2]PftJ(3); moreover, there exists a quasisimple
group of the form [4 x 4]L 3 (4).
It turns out, after the calculation of the Schur multiplier of each simple DC-group
K and the action of Aut(K) on it, that every quasisimple group K has one of the
forms
AK or \AX x A2]K
where K = K/Z(K), and this notation determines K up to isomorphism with the
following exceptions:
2Dn(q), q odd, n even, n > 4, 4L 3 (4), and 3I74(3),
for each of which there are exactly two isomorphism classes of groups.
We also use parentheses to indicate that we allow more than one possibility for
the central subgroup Z—that Z may be trivial or of the indicated order. Thus,
for example, K = (2)An means that either K = An or K = 2An. (The symbols
(S)Ln(q), (P)Qn(q), etc., are similarly ambiguous, representing a group which is
isomorphic either to SLn(q) or Ln(q), etc.) By contrast, a symbol inside a square
bracket denotes the actual structure of Z: K = [X]Ls(A), X = 4 or 4 x 2, means
that Z(K) ^ Z4 or Z 4 x Z 2 .
Finally, we give a roughly equivalent conceptual description of the notion of
Cp-group. We need some preliminary terminology. Fix a prime p.
First, an element x G 1P(H) is p-central (in H) if x is in the center of a Sylow
p-subgroup of H.
Next define
GJip = {K eXp\ F*(CK(x)) = Op(CK(x)) for every p-central x elp(K)}.
It is immediate that every K G CCKP is actually simple. Now given any set *KP
of known simple groups, each of order divisible by p, define
tip = {K G Xp I K/Z(K) G Up),
and let [K* be the set of all simple groups K G % such that either K G IKP,
or rrip(K) > 1 and LP(K) C J£p. Thus IK* consists of IKP plus certain additional
groups K, the components 3 of whose centralizers of elements of order p are required
to be covering groups of [Kp-groups. In particular, K G IK* if K is of characteristic
p-type, or more generally if HP(K) is empty and mp(K) > 1.
3
By the Bp property, th e p-components of such centralizers are actually components.
Finally we define
3o(p) = {J | J is simple and J G £hev(p) U {A p }};
di(p)=do{pTneKp; and
S2(p) = di{p)*neKp.
With this terminology, the set
eP = do(p)udi(p)ud2(p)
is a reasonable approximation 4 to Cp. Thus the Cp-groups consist of the groups in
Qhev(p) and Ap, together with certain of their pumpups and "double" pumpups.
An even closer approximation can be obtained by using the above definitions but
expanding G0ip to include all groups K G %p with the following property: for
every p-central x G TP(K), OP>(CK{X)) = 1 and each component of E(CK(X)) lies
in 3o(p) but not in any do{r) for any r ^ p.
The notion of a Cp-group has been introduced solely for expository reasons. It
is the Cp-groups with which we shall work.
In particular, the definition of "even type" (Definition 1.3) is now complete.
13. Sp-groups and T p -groups

Unfortunately the existence of an element K in LP(G) with K a Cp/-group for
some prime p is not in itself sufficient to guarantee that the target group G* is a
"generic" simple group. On the contrary, there exist certain "in-between" situations
whose analysis requires special methods.
To accommodate this case separation, it is necessary to divide the set Cp/ into
two subsets, the larger called Sp, corresponding to the generic case, with the residual
"intermediate" set called 7pi 7 for "thin", as these groups all have low p-rank. We
first list the elements of the sets 7P for each prime p, and take Qp as the set of
Cp/-groups left over.
DEFINITION 13.1. Let p be a prime. Then 7P is the set of all (^/-groups K
such that one of the following holds:
(1) p = 2 and K ^ L2(q), q odd, q £ 3^M9, A7, SL2(q), q odd, 2An, 7 < n < 11,
or [X]L 3 (4), X = 4, 4 x 2, or 4 x 4;
(2) p = 3 and K/Z{K) ^ L|(g), e = ± 1 , q = e(mod 3), A7, M12, M22, J 2 or
G 2 (8), or K^3A6;
(3) p=5 and K^Fi22] or
(4) p is odd and mp(K) = 1.
Furthermore,
up — J^p ^->p Jp»
Since 72 by definition contains all groups in %2 with quaternion Sylow 2-

subgroups, it follows that 7p contains, for each prime p, all elements of %p of p-rank
1, with the exception of the Cp-groups L2(p) and Ap, and (for p = 5) Sz(32).
4
For example, to consider simple groups for illustration, the only simple elements of C p — C p
are £>4(3), He, F5, AQ and ^4io (for p = 2) and 1/3(4) (for p = 3). The only simple elements
of Qp - ep are A2p and A3p (for p > 5), Coi and F 3 (for p = 7), 2 F 4 (32), Sz(32) and tfe (for
p = 5), P S p 4 ( 8 ) , 6*2(8), £/ 5 (2), D 4 (2), C o i , Suz, O N and Ly (for p = 3), and Fi2z (for p = 2).
These elements fail to be in C p generally only because they contain a p-central element x such
t h a t E(CK(x)) ^ 1.
14. GROUPS OF SPECIAL ODD TYPE 103
Note also that for p = 2 the groups An, n = 9,10,11, are S2-groups according
to the above definition, while Ag(= 1/4(2)) is a C2-group and Aj is a T2-group; but
their 2-fold covers 2An (7 < n < 11) are all ^-groups.
We note also that if K is a sporadic Sp-group, then one of the following holds:
(1) P =• 2 and K ^ J i , Mc, Ly, ON, He, Co3, o r f 5 ;

(2) P =• 3 a n d K 9* M 2 3 , M 2 4 , J4, Ru, i J S \He, , or 3QN;
(3) P =• 5 and K 9* Suz, F i 2 3 , or F i 2 4 >
(4) p = = 7 and K= F2] or
(5) P =• 11 or 13, a n d K ^ F i .
In particular, if p is odd, then in every case mp(K) = 2 + mp(Z(K)).
The sets Cp, T p and %v have been constructed in such a way as to retain
a property of "closure" under taking p-components of centralizers of elements of
order p-indeed of any automorphisms of order p. Namely, it is verified without
difficulty that
If K G Cp (resp. T p ), then for any x G Aut(K) of order p,

all elements of LP(K) lie in Cp (resp. Cp U T p ).
Moreover,
If K G Qp (resp. 7P),
(13.2) then all nonidentity factor groups of K lie in Cp (resp. 6 p U T p ) ,
except lip = 2 and if = 2 ^ n , 9 < n < 11.
In some cases, these properties control whether pumpups are Cp, 7P, or Sp-
groups. For instance, suppose that K and L are p-components of CQ{X) and CG{V),
respectively, where a: and y are of order p and L is obtained from if by a chain of
pumpups. Because of (13.1) and (13.2), it follows that if K/Op>(K) G Sp, then also
LjOp\L) G Sp unless p = 2 and K/0(K) = An,9<n< 11.
14. Groups of special odd type

We can now define the major case divisions of the classification grid.
In the previous chapter we introduced the following refinements of 1p{G) and
DEFINITION 14.1. Let p be a prime. Then
T°(G) = {x G IP(G) I mp{CG(x)) > 4} if p is odd,

X°2{G) = {xe 12{G) I m2(CG(x)) > 3}, and
££((2) = {L I L is a component of CG(X)/OP>(CG(X)) for some x G Zp(G)}.
In this section we consider the cases where G is not generic for the prime 2.
These cases are incorporated into the following definition.
DEFINITION 14.2. We say that G is of ^ - s p e c i a l t y pe if and only if every
element of ^ ( G ) is either a C2-group or a T2-group (equivalently, no element of
&2(G) i s a S2-group). Moreover, G is of special odd type if and only if G is of
^2-special type but not of restricted even type.
It is clear from the definition that if ra2(G) = 2, then G is of special odd type.
Moreover, as noted in the previous chapter, if 7712(G) > 3, then £J%(G) — ^ ( G ) by
an argument using the Thompson transfer lemma.
The analysis of groups of special odd type itself splits into three cases, the first
of which is of course the solvability of groups of odd order, included as a Background
Result. This accounts for case 1, the first row of the classification grid.
When G is of even order, two cases are drawn according to the principal tech-
niques to be used in the analysis:
CASE 2. The Bender method together with 2-fusion analysis and character
theory.
CASE 3. The signalizer functor method together with 2-fusion analysis.
A brief discussion of the relevant methods appears in sections 28-31 and 35-36
below.
The two cases are distinguished by the nature of the elements of £,2(G). The
first arises primarily, but not exclusively, when some element of £<2(G) lies in the
set £2, whose definition we repeat.
DEFINITION 14.3. 3 2 consists of the groups SL2{q), q odd, q > 3; [X]L 3 (4),
X = 4, 4 x 2 or 4 x 4; and 2An, 7 < n < 11.
Thus 3 2 C T 2.
The following definition incorporates the configurations to be treated by the
Bender method. It includes the case in which some element of £<2(G) is a 232-group,
and some cases of small 2-rank.
DEFINITION 14.4. If G is of special odd type, we say that G is of £!B2-type if
and only if one of the following holds:
(1) Some element of &2(G) is a 232-group; or
(2) 1712(G) < 3.
The complementary set defines the cases to be treated by the signalizer functor
method.
DEFINITION 14.5. We say that G is of £ T 2 - t y p e if and only if G is of even
order, G is of special odd type, but G is not of £!B2-type.
Now we can state Theorems Ci, 62, and 63.
T H E O R E M CI . G is of even order.
T H E O R E M 62- If G is of £!B2-fa/pe? but contains no 2-uniqueness subgroup,
then G ^ G* for some G* G X&.
THEOREM C3. If G is of L>72-type, then G ^ G* for some G* e X^.
These three theorems together with Theorem U(2) show that if G is of special
odd type, then G *+G* for some K e %^ U X& u X^.
[The hypothesis in Theorem 62 that there is no 2-uniqueness subgroup could
be deleted, at the expense of having the groups £2(4), £^(4) and J\ added to the
conclusion. We have given the above statement of the theorem simply to preserve
the device of keeping the target sets X^ pairwise disjoint.]
15. GROUPS OF SPECIAL EVEN TYPE 105
15. G r o u p s of special even t y p e

For odd primes, generic and special type groups are denned only when G is
of restricted even type, so we make this assumption throughout. Moreover, when
<J(G) is nonempty, the definitions are limited to primes p G &(G) (Definition 1.5).
For such p, mp(G) > 4 by definition, and so T°(G) is nonempty.
DEFINITION 15.1. The group G of restricted even type is of £ p -special t y p e
for p G cr(G) if and only if
^(G)cepUTp,
or equivalently, no element of £J°(G) is a S p -group. Furthermore, we say that G
is of special even t y p e if and only if G is of restricted even type, and either
(TQ(G) — 0 or G is of £ p -special type for some p G cro(G).
The analysis of groups of special even type likewise divides into three major
cases, each again determined by the principal techniques used in their classification:
CASE 4. The Goldschmidt amalgam method
CASE 5. The Klinger-Mason method
CASE 6. The signalizer functor method
These methods will be briefly described in Sections 22, 23 and 29-34.

We now give the precise conditions defining each of these cases. The condition
for Case 4 is that G be quasithin (Definition 1.6). In this case, Theorem M(5)
from the uniqueness grid implies that |M(5)| > 2. It is this case whose analysis
will be based on the Goldschmidt amalgam method. When G is not of quasithin
type, then Theorem U(cr) from the uniqueness grid implies that o~0(G) ^ 0. This
case is further split into two subcases (cases 5 and 6) according to the nature of
the elements of &p(G) for p G cr0(G).
DEFINITION 15.2. Let p G <T0(G). Then G is of £C p -typ e if and only if G is
of restricted even type and &°(G) C Cp; and G is of £7p-typ e if and only if G is
of restricted even type, ££(G) C ep U 7P, but L°p(G) £ Cp.
Thus for p G (Jo(G), G is of /Cp-special type if and only if it is either of £C p -type
or of £Tp-type.
The analysis of groups of £C p -type depends on the Klinger-Mason method;
that of groups of £Tp-type on the signalizer functor method.
Now we can state Theorems C4, 65 and C6- Throughout, the hypotheses imply
in particular that G is assumed to be of restricted even type. In addition we assume,
as we may in view of Theorem U(2) of the uniqueness grid, that G does not contain
a 2-uniqueness subgroup.
THEORE M C4. IfG is of quasithin type but contains no 2-uniqueness subgroup,
then G9HG* for some C* G X^.
THEORE M C5. IfG is of L£p-type for some p G o-0(G), then p = 3 and G = G*
for some G* G3C(5).
THEOREM C6. G is not of Lip-type for any p e cr0(G).

In both Theorems 65 and 66 the proofs reach contradictions by the construction
of a strong p-uniqueness subgroup.
In view of Theorem U(cr), these three theorems completely classify G if G has
special even type. As with Theorem C2, we could delete the hypothesis about 2-
uniqueness subgroups from Theorem C4 at the expense of including Z/2(2n), Sz(2n)
and U3(2n) (n > 3) in the conclusion.
16. G r o u p s of generic t y p e
Finally we consider the generic situation.
DEFINITION 16.1. We say that G is of generic o d d t y p e if and only if some
element of -C^Ô *s a S2-group. We say that G is of generic even t y p e if and
only if G is of restricted even type, G is not of £ p -special type for any p G 0*0 (G),
but <TO(G) is nonempty. (Thus, taking p e GQ(G), it follows that some element of
&°{G) is a Sp-group.)
Finally, G is of generic t y p e if and only if it is either of generic odd type or
generic even type.
We remark that groups of Lie type of characteristic 2 and large Lie rank are
of generic even type, and those of odd characteristic are of generic odd type. The
primes p mentioned in Definition 16.1 are never the characteristic; thus p is even
when G is of odd generic type, and vice-versa.
In addition, the condition that G be of even type in the above definition forces
the odd and even generic cases to be mutually exclusive, since &2(G) C 62 for
groups of even type.
THEOREM C7. IfG is of generic type, then G = G* for some G* e 3C(7).
Together, Theorems 6^, 1 < i < 7, combined with results from the uniqueness
grid, establish the Classification Theorem.
For clarity we restate the main logic. It is immediate from Definitions 14.2,
15.1, and 16.1 that a OC-proper simple group G is of special odd type, special
even type or generic type. If G has a 2-uniqueness subgroup, then Theorem U(2)
applies. Otherwise, if G is of special odd type, then Definition 14.5 shows that one
of Theorems 61, 62 or C3 applies; if G is of special even type, then one of Theorems
64, C5 or 66 applies, as noted after Theorem QQ; and if G is of generic type, then
Theorem 67 applies. In every case, G G %.
D. T H E CLASSIFICATION G R I D :
T H E STAGES OF THE P R O O F
17. T h e o r e m e x
As with the uniqueness grid, we shall now describe the successive stages of the
proofs of Theorem (^, 1 < i < 7.
Although Theorem Ci is a Background Result, we are including a brief discus-
sion of it because of parallels with other cases, particularly Theorem U(cr).
17. THEOREM &i 107
Assume then that G has odd order. Since G is DC-proper, all proper simple
sections are of prime order and hence all proper subgroups of G are solvable. First,
define
o~*(G) = {p \p is a prime and mp(G) > 3}.
A simple transfer argument shows that <r*(G) ^ 0. The objective of the first
stage of the analysis is the following:
d : STAGE 1. For each p e <r*{G) and each subgroup E = Ep3 of
T H E O R EM
G, E lies in a unique maximal subgroup of G.
In the classification grid, we abbreviate the conclusion of this theorem by saying
that G has odd order uniqueness type. This is analogous in a broad sense to the
hypothesis of Theorem U(<r), and shows that the classification of groups of odd
order reduces in some sense to a uniqueness-type problem. However, this is only
partially true, for the subsequent stages of the analysis require precise information
about the structure and embedding of p-local subgroups of G for primes p for which
Tnp(G) < 2, so that there are also features of the analysis that have a resemblance
to the quasithin group problem.
These are included in the second stage of the analysis, the goal of which is
to pin down the exact structure and embedding of the maximal subgroups of G
to the extent possible by local group-theoretic methods. We shall not attempt to
describe the possibilities in detail here. An expository account of their structure
and embedding (as well as of the subsequent stages of the Feit-Thompson analysis)
appears in [G4, Chapter 1]. Roughly speaking, a maximal subgroup M of G is
shown to "resemble" a Frobenius group (in some cases, it is a Frobenius group) and
its "Frobenius kernel" is shown to be approximately disjoint from its conjugates
in G. Let us then abbreviate all these conditions, once the exact possibilities for
the structure and embedding of the maximal subgroups of G have been determined
to the extent possible by local group-theoretic methods, by saying that G is of
a* (G)-uniqueness type.
Thus one next proves
THEOREM Cii STAGE 2. G is of a* (G) -uniqueness type.
This is analogous to Theorem U(<r): stage 1.

The third stage of the analysis involves exceptional character theory. In this
stage, all but a single possibility for the configuration of maximal subgroups are
eliminated. This residual case is expressed in terms of the structure and embedding
of a maximal p-local subgroup of G for a suitable prime p. In particular, it is shown
that some maximal subgroup of G is not a Frobenius group.
The following definition incorporates the goal of this analysis. It represents
the set of conditions which Peterfalvi uses to derive a final contradiction in his
simplification of Feit and Thompson's original generator-relation analysis.
DEFINITION 17.1. The group G is of {p, g}-parabolie t y p e for the odd primes
p and q if and only if
(1) {{jfl - l)/p - 1), (p - 1)) = 1 and p > q;
(2) G contains a Frobenius group H = AP with kernel P = Epq and cyclic
complement A = Z&, where k = (pq — l)/(p — 1); and
(3) There exists a subgroup Po of P of order p and a Po-hivariant abelian q-
subgroup Q of G such that PQ normalizes A for some y G Q-
Thus Feit and Thompson next prove, using character theory:

T H E O R E M C I : STAGE 3. G is of {p,q}-parabolic type for suitable primes p, q.
Finally, this conclusion is shown to be impossible [P3].
T H E O R E M C I : STAGE 4. G is not of {p, q}-parabolic type for any primes p, q.
The last two stages are somewhat analogous to the last stages of Theorem C5
in their focus on the interplay between two primes, and also to Stroth's strategy
in Theorem U(cr): stage 2, namely, to construct a pair of subgroups and derive a
contradiction by amalgam-type analysis.
18. p - t e r m i n a l p - c o m p o n e n ts
The analysis of most of the remaining theorems involves the notion of a p-
terminal p-component in the centralizer of an element of TP(G), p a prime. The
notion extends that of a terminal component (Definition 1.2) in the presence of p'-
core obstruction, and was defined in section 7 of the preceding chapter. We briefly
recall the definition and basic properties.
First, by definition, the p-components of a group X are the minimal subnormal
subgroups K of X such that the image of K in X = X/Op> (X) is a component of
X. Each component of X is then the image of a unique p-component of X , and
the product of the p-components of X is called LP'(X), the p-layer.
Thus the set £JP(G) (resp. £J°(G)) consists of the groups K/Op>{K) as K ranges
over the set of p-components of CG(X) and x ranges over TP(G) (resp. T°(G)).
It follows quickly from the definition that a p-component K of X is a com-
ponent of X if and only if K centralizes O p /(X), or equivalently if and only if
K is quasisimple. Moreover, if K is a p-component of the group X, and we set
X = X/Op'(X), then for any y G Tp(Cx(K)), LP'(CK(y)) maps onto K. Here
Cx{K) is the subgroup of Nx(K) acting trivially on K by conjugation.
The fundamental definition is the following:
DEFINITION 18.1. Let K be a p-component of CG{x) for x G 1P(G). We say
that K, or (#, K), is p-terminal in G if and only if the following conditions hold:
(i) For any element y G Tp(CcG(x)(K/Op>{K))), Lp>(CK(y)) is contained in a
p-component J of Cc{y) such that J/Op'(J) = K/Op>{K)] and
(ii) For any Q e Sylp(CCG(x)(K/Op>(K))) and y e IP(Z(Q)), if J is determined
as in (i), then Q £ Sylp(CCG(y)(J/Op,{J))).
Note that if mp(CG{K/Op>(K))) = 1, then it is automatic that K is p-terminal
in C.
We shall abuse terminology slightly by calling such a K a p-terminal p-com-
p o n e n t (in G). Of course such a subgroup is a p-component not of G, but of CG{X)
for some x of order p. In addition, for brevity, if K is a p-terminal p-component in
G and K = K/Op>(K), we shall say that K itself is p-terminal in G.
Observe that if a p-component K of some CG(X) is quasisimple, then CG(K) =
CG{K/OP'(K)), by an elementary property of quasisimple groups. Thus K =
LP'(CK{y)) for each y £ XP(CG(K)). We therefore have the following criterion for
a p-terminal p-component to be terminal in G: If K is a p-component of GG(X) for
some x G lp(G), with K p-terminal in G, then K is terminal in G if and only if K
centralizes 0P'{CG{y)) for every y G XP(CG(K)). The condition includes the case
y = x.
19. CENTRALIZER OF INVOLUTION PATTERNS 109
As observed in the previous chapter, if we start with an element x G 1p{G)

and a p-component K of CG(X), it is possible by a sequence of pumpups to arrive
at a p-terminal p-component K* of CG(X*) for some x* G Xp(G). This process is
the first step of the analysis in several rows of the classification grid. Moreover,
x* and K* may be taken to inherit certain properties of x and K. For example,
because of Lp /-balance, the properties (13.1) and (13.2) imply that if K/Op'{K)
lies in Sp or Sp U 7P, then the same holds for K* unless p = 2 and KfO(K) = An,
9 < n < 11. Likewise it is important to be able to assert that if x G 2£(G), then
the same holds for #*. This is a triviality for p = 2, and is straightforward if p
is odd and K/Op>{K) G Sp5 since then K contains an Epi subgroup disjoint from
Op>p(K). With some more difficulty it can also be established for suitable K in the
case of groups of £T p -type, p odd.
Finally, we note that it is useful for us to extend the notion of p-terminality in
a natural way to so-called solvable p-components 5 for p = 2 and 3. We state the
definition of these objects here, but do not give the extension. A subgroup / of
CG{X) is called a solvable p-component if x has order p = 2 or 3, I > Op> (CG{X)),
and in CG(x) = CG(x)/Op>(CG{x)), 7 = SL2(S) o r SU3(2), respectively, and the
normal closure of / in CG(X) is the product of elementwise commuting 5L2(3)- or
5t/3(2)-subgroups of CG(X), one of which is / . / is a solvable component if and
only if the normal closure of I in CG(x) is the product of elementwise commuting
5Z/2(3)- or 5[/ 3 (2)-subgroups of CG{X), one of which is / .
19. Centralizer of involution patterns

As indicated in the preceding chapter, the term G « G* when G* is a spo-
radic group or a low degree alternating group is expressed entirely in terms of the
structure of centralizers of involutions. We repeat the definition.
DEFINITION 19.1. The groups G and G* have the same involution fusion
pattern if and only if there is an isomorphism x i—• x* from a Sylow 2-subgroup S
of G onto a Sylow 2-subgroup of G* such that for any two involutions u, v G 5, u
and v are conjugate in G if and only u* and v* are conjugate in G*. Moreover, if this
is the case, we then say that G and G* have the same centralizer of involution
pattern if the isomorphism can be chosen so that in addition CG(U) = CG*(W*)
for every involution u G S.
DEFINITION 19.2. If G* is sporadic or G* = An, 7 < n < 12, n ^ 8, we write
G w G* if and only if G and G* have the same centralizer of involution pattern,
and in the cases G* = Aj and M\\ if and only if in addition \G\ = |G*|.
The additional condition in the M\\ case is required to distinguish it from the
group £3(3), which has the same centralizer of involution pattern as M\\. On the
other hand, in the A7 case the required order for G is in fact obtained in the process
of determining the structure of the centralizer of an involution. Finally, As = 1/4(2)
is more easily identified as the group £4(2) of Lie type.
Although the term G w G* differs from the above for the remaining known
simple groups, nevertheless much of the analysis of Theorem 62, to be described
in the next section, is concerned with showing that the centralizer of involution
5
This was done by Aschbacher for p = 2 [A5], and the case p = 3 is similar.
pattern in G approximates that of one of the target groups G*, this approximation
becoming sharper as one moves from stage to stage.
20. Theorem C2
The classification of groups of £23 2 -type (Definition 14.4) is the most elaborate
of any of the results to be derived in either grid. For one thing, this case involves
the most complex bifurcation pattern; for another, the conclusions to be reached at
various stages of the analysis have by far the most technical statements. For sim-
plicity, in this introductory chapter we shall give only an approximate description
of the precise results to be established. This will be entirely sufficient to convey the
nature of the overall proof. [A discussion of the SL2 (q) subcase is given in section
25 of the preceding chapter.]
First, it is necessary to enlarge the set X^ slightly. We set
^2)*=^2)u{L2(4),t/3(4),Ji}.
Our assumption that G is not of 2-uniqueness type eventually excludes these
three added groups, but this does not occur until a late stage of the analysis.
The first stage of the analysis uses almost exclusively the techniques of 2-fusion
and 2-transfer. The conclusions of this stage are incorporated into the following
definition.
DEFINITION 20.1. If G is of £!B 2 -type, we say that G is of 2-terminal G*-
type for G* G X^* if and only if the following conditions hold:
(i) G and G* have the same involution fusion pattern (in particular, G and G*
have isomorphic Sylow 2-subgroups); and
(ii) If z is a 2-central involution of G, then the structure of CG{Z)/Oy (GG(Z))
closely approximates that of CG* (z*) /Oy (CG* (z*)), where z* is a 2-central
involution of G*.
The precise meaning of condition (ii) will be given in Part IV. We confine our-
selves to the following remarks for now. Every G* e K (2) * has only one conjugacy
class of 2-central involutions, and most such G* have only one class of involutions
altogether. Furthermore, with the exception of the groups G* = Z/2(g), q odd,
173(4) and M12, CG*(Z*) contains a component or solvable component K* for each
G* G X^*. In all cases where such a K* exists, the precise meaning of (ii) includes
the assertion that CG(Z) possesses a 2-component or solvable 2-component K with
K/02f{K) isomorphic to K* or at least closely approximating K*. In these cases
we shall say that G is of 2-terminal G*-type (z,K).
T H E O R E M C2 : STAGE 1. IfGis ofX232-fr/pe, then G is of 2-terminal G*-type
for some G* e X&*.
The second stage of the analysis involves the application of the Bender method
to study 02'{CG{Z)). The sharpest results are achieved when rri2(G) > 3. In fact,
when G* is sporadic and rri2{G) > 3, namely when G* = M i 2 , Mc, Ly, ON or
J i , one already reaches the conditions defining G ~ G*. In particular, the case
G* = J\ is eliminated at this stage of the analysis since the conditions G w J\
immediately imply that G contains a 2-uniqueness subgroup.
The conclusions reached in the non-sporadic cases of 2-rank at least 3 are
incorporated in the following definition.
20. THEOREM C 2 111
DEFINITION 20.2. Assume that G is of 2-terminal G*-type (z, K) for G* e DC(2)

of Lie type with m2(G*) > 3. We say that G is of 2-central G*-type if and only
if the following conditions hold:
(1) L2,(CG(z)) = E(CG(z))-and
(2) Either 02>(CG(z)) = 1, or G* - LA{q) or UA(q) and 02>(CG(z)) is cyclic of
order dividing q ± 1, respectively.
We remark that (1) follows from (2), but it is included for emphasis. Again we
write 2-central G*-type (z, K) if we wish to specify z and K.
When G* has 2-rank 2, the corresponding definitions focus on a maximal sub-
group M of G containing CG(z), z G X2{G). By far the most elaborate analysis
occurs in the L^(q) case and here too the final conclusion is by far the sharpest.
DEFINITION 20.3. If G is of 2-terminal L 3 (g)-type, q = pn, p an odd prime,
we write G « Ls(q) if and only if the following conditions hold for some maximal
subgroup M of G containing CG(z):
(1) F*(M) = Op(M) ^ Eq2 with Op(M) inverted by z\
(2) CG(z) = M/Op(M) and M/Op(M) is isomorphic to a subgroup of TL2(q)
containing SL2{q)\ and
(3) M contains a Sylow p-subgroup of G.
Note that M is approximately isomorphic to a maximal parabolic subgroup of
G*. Prom this detailed information it is easy to reach the split (£?, JV)-pair structure
of G.
In the remaining cases, the information derived in this second stage places
rather weak restrictions on the structure of 02'(M). We emphasize this fact by
our choice of terminology. The precise definitions are technical and subdivided into
several cases. For simplicity we state only the principal conclusions. Recall that M
is a maximal subgroup of G containing CG{z).
DEFINITION 20.4. Assume that G is of 2-terminal G*-type with m2(G*) = 2
and G* ^ Ls(q). We say that G is of 2-maximal G*-type mod cores if and only
if the following conditions hold:
(1) 0 2 ( M ) ^ l ; a n d
(2) L2,(CG(z)) = E(CG{z)).
Condition (1) easily implies that M = CG(z) unless G* = L 2 (7), L2(9) or A7
and M/02>(M) = E 4 . For the cases G* = 1/3(3) or M n , condition (2) is replaced
by the condition that CG(z) contain a normal Qg-subgroup. Now at last we can
state the goal of stage 2.
THEOREM C2: STAGE 2. If G is of 2-terminal G*-type for G* e 3C(2)*, then
one of the following holds:
(i) G* = M12, Mc, Ly, ON or L3(q), q odd, and G w G*;
(ii) G* e Qhev, m2(G*) > 3, and G is of 2-central G* -type; or
(hi) m2(G*) = 2, G* ^ L^{q) for any q, and G is of 2-maximal G*-type mod
cores.
In conclusion (ii), the structure of CG(z) is essentially completely specified. On
the other hand, in conclusion (hi), we are given no information about the structure
and embedding in G of 02>(M) for the maximal subgroup M of G containing CG{z).
However, detailed information of this nature is required to prove that G « G* in
this case.
Achieving this objective involves a very long and difficult analysis involving
local methods, character theory and counting arguments. Moreover, although there
are many analogies and similarities among the arguments employed in the various
cases, the details are sufficiently different that it is best at this point to separate
the proof into cases according as (1) G* = L2(q) or A7; (2) G* = £/3(4); or (3)
G* = M n or U3(q), q odd. In all cases one determines the order of G and identifies
G as a doubly transitive permutation group with a specified point stabilizer. It is
in the latter phase of this analysis that the 1/3(4) case is eliminated by showing
that G contains a 2-uniqueness subgroup.
Note also that the doubly transitive target groups in X^ include the groups
2
G2(q) of 2-rank 3. Moreover, the passage from 2-central 2 G2(g)-type to double
transitivity again utilizes character-theoretic methods. Because of this, it is prefer-
able to treat the 2G2(#)-case along with the 2-rank 2 cases. Hence at this point
in the analysis, we split the groups in X^ of Lie type according as they have Lie
rank 1 or Lie rank at least 2.
This shows the complexity of the succeeding stages of the classification of groups
of £!B2-type. The precise conclusions one reaches from the local group-theoretic
analysis are extremely technical. We limit ourselves therefore to describing only
the one portion of its conclusions that is uniform for all G*, and to making a few
additional comments.
We assume then that G is of 2-maximal G*-type mod cores for some G* e X^*
of 2-rank 2 and with G* ^ L${q) for any q. With z and M as in the definition, set
Q = 02>{M), let t e I 2 ( M ) - 02(M) and put B = [Q,t].
Then one goal of the local analysis is to verify the following conditions:
(1) B is a Hall subgroup of F(Q);
(2) B is abelian and inverted by t;
(3) NG(B0) < M for all 1^B0<B.
[The goal of the subsequent character-theoretic and counting analysis is then
to force B = 1 except when the target group G* is A?, in which case it is to force
B * Z3.}
One of the principal alternative outcomes of the local analysis is that Q has
essentially the structure of a Probenius group with kernel B and complement A,
with additional information about the embedding in G of NC(AQ) for 1 ^ AQ < A.
A second alternative is that NG{X) ^ M for some 1 ^ X < F(Q), with specific
structural information concerning NG(X).
Further analysis proceeds by means of exceptional character theory, modular
character theory and counting arguments. We now describe the precise nature of
the double transitivity we reach in the third stage of the analysis when G* is of Lie
rank 1. It is incorporated into the following definition.
DEFINITION 20.5. Assume that G is either of 2-maximal G*-type mod cores
for G* = L2(pn), p odd, pn > 5, or C/3(pn), p odd, or of 2-central G*-type for
G* = 2 G 2 (p n ), p = 3, n > 1. We say that G is of doubly transitive G*-type and
write G « G* if and only if the following conditions hold:
(1) G acts doubly transitively on a set O;
(2) If Ga is the stabilizer of the point a G H, then Ga = NQ(P) for some
P G Sylp(G), and Ga has the following properties:
(a) \P\ = _pn, p3n or p 3 n , respectively;
20. T H E O R E M C2 113
(b) P acts regularly on Q — {a} (whence \Q\ = \P\ + 1);

(c) Ga — PY, where Y is cyclic of the respective order (pn — l ) / 2 , (p2n —
l)/d, d odd, d\(pn + 1), or pn — 1 with p = 3 and n odd, n > 1; and
(d) Correspondingly G a , Ga/§{P) or 02(Ga) is a Probenius group with
kernel P , P/$>(P) or P , respectively; and
(3) The following conditions hold:
(a) If G* = L2(pn), then Y is inverted by an involution t of G and CG{V) <
y(t) for every y E
(b) If G* = £/ 3 (p n ), then for any involution z of G, CG(z) 9* GU2{pn)/D,
where D is a cyclic normal subgroup of order d. Moreover, for a
suitable choice of z, 0 2 / ( G G ( Z ) ) < Y and CG{V) < CQ{Z) for every
l/G02,(CG(z))#;and
(c) If G* = 2 G 2 (p n ), then for any involution z of G, G G (z) ^ Z 2 x L2(pn).
Moreover, for a suitable choice of z, z G Y, Y is inverted by an involution
of CG(z), and CG(y) = Y for every y e 02>(Y)#.
For the groups of Lie rank > 2, the definition of G w G* is extremely detailed,
depending on the isomorphism type of G*. We have already defined G w L3 (#), so
we just consider the remaining cases. For all the cases, G ~ G* includes a condition
that G have subgroups H and N with H < N, resembling a Cartan subgroup and
monomial subgroup H* and N* of G*. For the cases G* = £3(9), P5p 4 (g), G2(<z)
and 3£>4(g), it includes a condition that G have a p-local subgroup M resembling
a certain maximal parabolic subgroup M* of G*, namely the normalizer of a long
root group. For the cases G* = Lf(q), it includes a condition that for 2-central
involutions z and z* of G and G*, respectively, G G ( Z ) and CG*(Z*) resemble each
other. The conditions are too technical to state here, but we note that for G* =
PSp±(q), for example, the critical requirements on M are that Op(M) is special of
order q3 with center of order q\ M/Op(M) = M*/Op(M*); the central involution
of M/Op(M) inverts Op{M)/Z\ and M contains a Sylow p-subgroup of G. We
summarize all these requirements by saying that G has subgroups H, N, M and
CG (Z) of the same complexion as the corresponding subgroups of G*.
DEFINITION 20.6. If G is of 2-central G*-type for G* e %{2) of Lie type of
Lie rank at least 2 but G* ^ Ls(q), q odd, we set G « G* if and only if G has
subgroups iJ and N of the same complexion as the corresponding subgroups of G*;
furthermore, according as G* = PSp^(q), G2(q), 3D±(q) or Lf(q), G has a subgroup
M, M, M or C<?(z), respectively, of the same complexion as the corresponding
subgroup of G*.
THEOREM C 2: STAGE 3. IfG is of H^-type, then G « G* /or some G* e 3C(2).
We complete the classification of groups of £232 -type by turning this internal
approximation into an isomorphism.
THEOREM C 2: STAGE 4. IfG is ofL^-type, with G « G* for some G* e 3C(2),
The sporadic and Lie rank 1 cases of the above theorem are Background Results.
In the Lie rank > 2 cases, one argues that the multiplication table of G is uniquely
determined either because G is a split (P, iV)-pair of rank 2 or by the Curtis-Tits
theorem or a variation of it.
21. Theorem C3
Next we discuss the classification of groups of £T2-type (Definition 14.5). In
the analysis of groups of £32-type, the existence of an involution z in G whose
centralizer contains a 2-component K with K 2-terminal in G was essentially a
direct consequence of the conditions defining this case. In contrast, the existence
of such an involution in the /COVcase is the entire objective of stage 1. The desired
conclusion is incorporated into the following definition.
DEFINITION 21.1. If G is of £T 2 -type, we say that G is of 2-terminal £T 2 -
type if and only if for some x G 22(G), CG{X) has a 2-component K with K
2-terminal in G and Kj02>{K) = L2(q), q odd, q > 5, or A7.
THEOREM C3: STAGE 1. IfG is of UJ2-type, then G is of 2-terminal L72-type.
In view of (13.1) and (13.2), as remarked in section 18, the difficulty in estab-
lishing this theorem centers around 2-components of centralizers of involutions that
are isomorphic mod 2'-cores to £ 2 (5), L2(l) or L2(9), since such 2-components are
also in C2 and hence may possess proper pumpups outside of the desired set, such
as L 2 (16), 1/3(4) and P5p4(4), respectively.
The aim after stage 1 is to force K to be terminal in G and hence standard
in view of Theorem TS. In the contrary case, we use signalizer functor theory to
embed CG{X) in a 2-component preuniqueness subgroup M and then invoke our
general results on such subgroups to show that M is a 2-uniqueness subgroup, a
contradiction in view of our uniqueness theorems. Once K is standard, it then
follows easily that the only possibilities for K are L 2 (5) (= A5), L2(9) (= AQ),
or A7. With some additional work, we pin down the exact isomorphism type of a
Sylow 2-subgroup of G and then argue that G has the same centralizer of involution
pattern as A9, A10, or A n , respectively, so that G & G* for some G* G K3, which
is the goal of stage 2.
THEOREM C3: STAGE 2. IfGis of 2-terminal L72-type, then G w G* for some
G* etf).
Finally, from the Thompson order formula (Theorem 35.1 below), it follows
directly that \G\ = |A n |, n = 9, 10 or 11, and we now have sufficient information
to complete the identification of G.
THEOREM C3: STAGE 3. IfG^G* for some G* e X ( 3 ) , then G^G*.
This completes the classification of groups of special odd type.
22. Theorem C4
We now present a similar outline for groups of special even type, beginning
with those of quasithin type. Because the analysis of groups of quasithin type has
not been completed as of this writing, the discussion in this section must be viewed
as only a tentative outline of their classification.
The goal of the first stage of the analysis is to focus on an amalgam in G, which
exists by virtue of Theorem M(5).
DEFINITION 22.1. Let G be of quasithin type. Then we say that G is of 2-
amalgam type if and only if there are S G Syl2(G) and 2-local subgroups Mi and
M2 of G such that the following conditions hold:
(i) 5 < Mi n M2 and NMl (S) = NM2 (5); and
22. T H E O R E M C 4 115
(ii) 0 2 ( ( M i , M 2 ) ) = l.
In this case we say that G is of 2-amalgam type (Mi, M 2 ; S).
THEOREM 64: STAGE 1. If G is of quasithin type, then G is of 2-amalgam
type.
The second stage has two goals. The first is to reduce the number of amalgam
problems that require resolution, that is, the number of possibilities for the sub-
groups Mi/02{Mi), perhaps by making a different choice of the Mi. The second is
to solve these amalgam problems, that is, use the amalgam method to analyze the
structure of 02(Mi). In the end the Mi should have a structure closely approxi-
mating that of maximal 2-local subgroups of some G* G ! K ^ .
The precise objectives for this stage will depend on the group G*. Let us say
that G is of 2-amalgam G*-type for G* G X^ if and only if G is of 2-amalgam
type ( M i , M 2 ; 5 ) , and there are maximal 2-local subgroups M*, M% in G* con-
taining a common Sylow 2-subgroup 5* of G* such that O2 (Mi) is approximately
isomorphic to O 2 (M*), i = 1 and 2; and O2 (MiflM 2 ) is approximately isomorphic
to O2' (Ml n M2). The notion of "approximately" will depend on G*.
At the minimum, this approximation asserts that chief factors of the groups
in question are isomorphic, but in some cases an exact isomorphism is achieved at
this stage. Moreover, when G* G X^ is alternating or sporadic, O2 (Mi) — Mi,
i = 1, 2, and O2'(MxnM2) = M1HM2. On the other hand, when G* G K (4) is of Lie
type, one must, in general, allow for automorphisms of odd order acting on O2 (Mi)
(so that, in effect, one is obtaining an approximation of G with some subgroup of
Aut(G*)), and it is for this reason that the conditions must be expressed in terms
of the groups O2'(Mi).
T H E O R E M 64: STAGE 2. If G is quasithin of 2-amalgam type, then G is of
2-amalgam G*-type for some G* G X^.
This condition defines the term G & G* when G* G X^ is of Lie type.
DEFINITION 22.2. If G* G X^ is of Lie type, then G « G* if and only if G is
of 2-amalgam G*-type.
When G* G X^ is alternating or sporadic, the goal is now to show that G and
G* have the same centralizer of involution pattern, whence G « G* in these cases
as well.
THEOREM 64: STAGE 3. If G is of 2-amalgam type for G* G XÂ\ then G «
G*.
It thus remains to turn these approximations into exact isomorphisms. In the
Lie type case, this is achieved either by showing that G is a split ( £ , AT)-pair,
with a uniquely determined multiplication table, or by establishing the Curtis-Tits
relations or a variant of them. [As usual, the argument depends on the fact that
G does not contain a 2-uniqueness subgroup, in the sense that one constructs a
subgroup Go = G*, and the nonexistence of a 2-uniqueness subgroup is used to
show that Go = G.) On the other hand, in the case G* = A12, one again has
\G\ = I-A12I by the Thompson order formula and one then identifies G via the
standard presentation for A\2 in terms of generating involutions. Furthermore, in
the X^ sporadic cases the desired isomorphism is a consequence of the Background
Results.
Thus the final stage is the following.

THEOREM C4: STAGE 4. IfG^G* for G* e %{A\ then G^G\
23. T h e o r e m C5
Next we assume that G is of £C p -type for some p G CFQ(G). Then according
to Definition 15.2, G is of restricted even type and every element of J^p(G) is a
Cp-group. Since p G ao(G), G does not contain a strong p-uniqueness subgroup,
and p G cr{G).
The analysis in this case has been carried out completely in most, but not all,
subcases. Thus the steps outlined here are strictly speaking provisional, though
they are accurate for all the subcases which have been completely analyzed.
To state the objective of the first stage of the analysis, we need a preliminary
definition.
DEFINITION 23.1. 23*(6?) is the set of elementary abelian p-subgroups B of G
lying in a 2-local subgroup of G and of maximal rank subject to this condition.
Since p G &(G), the definition of this set of primes implies that mp(B) > 4 for
any B G *B*(G). On the other hand, 2-locals containing a subgroup of index at
most 2 in a Sylow 2-subgroup of G do not necessarily contain an element of 23* (G).
Now we can define the key term.
DEFINITION 23.2. We say that G is of wide £ e p - t y p e if and only if it is of
.CCp-type and
mp(CG(B)) > mp(B) for every B G S£(G).
THEORE M C5: STAGE 1. If G is of£jQp-type, then G is of wide L£p-type.
In the contrary case, if one takes B G 23* (6?) with mp(CG(B)) = mp(B), then
as p is odd it follows rather quickly with the aid of the signalizer lemma of Bender
and Thompson [Bel; G L l , 1-20.1] that Q(CG(b)) = Op,(CG{b)) for b e B* defines
a nontrivial S-signalizer functor 0 on G whose closure is of even order. This implies
that TB,2{G) is contained in a proper subgroup M of G, and now using signalizer
functor theory, one argues that M is a strong p-uniqueness subgroup, contrary to
assumption.
This is the point from which Klinger and Mason [KMal] began their analysis
of groups G of both characteristic 2- and characteristic p-type. [Thus F* (N) is a
2-group for every 2-local subgroup N of G and a p-group for every p-local subgroup
N of G.) Our hypothesis that G is of £C p -type includes the possibility that G is
of characteristic p-type, and our analysis parallels and generalizes theirs. It was
shown in [GLl] that their arguments could be extended when the characteristic p-
type condition was relaxed to allow centralizers of p-elements to have components
in Ghev(p)] and now we further relax the hypotheses, expanding Ghev(p) to Qp
and allowing centralizers of involutions to have components in 62- These weaker
hypotheses lead us to some target groups, whereas Klinger and Mason simply reach
a contradiction, at least if some 2-local subgroup of G has p-rank at least 4, as is
the case here.
The conclusion of the first phase of their analysis is that a maximal £?-invariant
2-subgroup T of G is acted on faithfully by B, and is of symplectic t y p e : a
central product of an extra-special 2-group and a cyclic, quaternion, dihedral or
23. THEOREM C 5 117
semidihedral 2-group. Under the present weaker assumption that G is of even

type, there are additional possibilities, and so we make the following definition.
DEFINITION 23.3. A symplectic pair in G is a pair (B,T) such that
(i) B G B*(G7);
(ii) T is a maximal ^-invariant 2-subgroup of G; and
(hi) Either T is of symplectic type or T has order 2.
Furthermore, if T has order 2, we call the symplectic pair (B,T) trivial, and if
CB{T) = 1, we call the pair faithful.
In addition to having to cope with trivial symplectic pairs we unavoidably
encounter one other configuration which is mixed, and arises in the target group
G* = fi^"(3). We say that a symplectic pair (B,T) is of f ^ ( 3 ) - t y p e if and only if
p = 3, B ^ £ 3 4, BT/CB(BT) is the central product of two copies of 5L 2 (3), and
CG(Z(T)) has a component K 2* L 2 (9).
The following definition then incorporates the goal of Stage 2.
DEFINITION 23.4. If G is of wide £C p -type, then we say that G is of quasisym-
plectic type if and only if p = 3, G has symplectic pairs, and every symplectic
pair in G is trivial, faithful or of Sl^(3)-type.
THEOREM C5 : STAGE 2. If G is of wide LQp-type for p e o~(G), then G is of
quasisymplectic type.
The second phase of the Klinger-Mason analysis (Theorem D of [KMal]) shows
that there is no simple group of characteristic 2 type and characteristic p type which
is of quasisymplectic type. On the other hand, under the present weaker even type
and £C p -type hypothesis, there turn out to be twelve groups of quasisymplectic
£e p -type: six of Lie type (ft 7 (3), ^ s ( 3 ) , P&g(3), raio(2)? U7(2), 2E6(2)) and
six sporadic {Fi22, Fi23, Fi24, Go\, F 2 , F\).
For the six groups of Lie type, the next stage of the analysis consists in showing
that G satisfies conditions somewhat similar to those holding in £ p -generic groups
of Lie type, the latter expressed by the condition that G contain a "neighborhood"
from which the Steinberg relations for G can then be deduced. We shall use the
same term for the groups in X^ of Lie type (see section 25). As in the generic
case, we shall take the existence of such a neighborhood as definition of the term
G « G* for G* e %^ of Lie type.
In the case of the six sporadic groups of X^, the following definition incorpo-
rates the objective of the bulk of the third stage of the analysis. We have chosen
the term "neighborhood" here solely because of the suggestive parallel with the
generic situation.
DEFINITION 23.5. Let G* e X^5\ If G is of quasisymplectic type, we call the
configuration (C/, CG(Z), CG(V), Co{t)) a Z 6 x ^ - n e i g h b o r h o od of typ e G* if
and only if the following conditions hold:
(1) z, y and t have orders 2, 3 and 2, respectively, and U = (zy) x (t) = ZQ X Z2\
(2) There exist commuting elements 2*, y* and t* of G* and isomorphisms
CG(z) * C G *(z*), CG(y) = CG*(y*) and CG{t) <* CG.(t*), all carrying z to
z*, y to y* and t to £*;
(3) z is 2-central in G with F*(CG{Z)) a 2-group;
(4) Cc{y) has a component K with (y) G Sylz{CG{K))', and
(5) t G 02(CG(z)) and E(CG(t)) ± 1.
The existence of a Z 6 x ^-neighborhood of type G* G X^ with G* G Spor

eventually yields that G and G* have the same centralizer of involution pattern and
hence that G « G*.
Thus the full objective of stage 3 is the following.
T H E O R EM C5: STAGE 3. IfGis of quasisymplectic type, then G m G* for some
G* G 3 C ( 5 ) .
The final stage recognizes G by means of the Steinberg relations or the Back-
ground Results.
THEOREM C5: STAGE 4. IfG^G* for some G* G X&\ then G ^ G*.
24. T h e o r e m C6
There is a considerable parallel between the proofs of Theorems C3 and ^6.
The principal distinction is that as G is now of even type, the analysis leads to a
contradiction: if p G CTQ(G) and G is of £T p -type, it is shown that G has a strong
^-uniqueness subgroup, contrary to the definition of 00(G). We let or(G) be the
set of primes p G 00(G) such that G is of £T p -type. Thus for any p G cr<j{G) and
any x G 2£(G), every component of CG(X)/Ov>(Cc{x)) is in Cp or T p ; and there is
some x for which some such component is in T p . (The list of T p-groups is given in
Definition 13.1.)
DEFINITION 24.1. We say that G is of p - t e r m i n al £Tp-type if and only if

there is x G 1p(G) and a p-component K of CG{X) such that K is j^-terminal in G
and K/Op>(K) G T p .
T H E O R E M 65: STAGE 1. If G is of restricted even type and p G o~<y(G), then
G is of p-terminal UJv-type.
This result has been mentioned already in section 18. If Tp(G) were equal to
T°(G), it would be a consequence of general properties of p-components and the
fact that pumpups of elements of T p cannot lie in Cp. The difficulty is due to the
setJp(G)-J°(G).
Next, using signalizer functor theory and the fact that groups in 7P have low
p-rank, we are able to apply our p-component preuniqueness theorems to prove
T H E O R E M QQ: STAGE 2. IfG is of restricted even type, then for anyp G cr<j(G),
G contains a strong p-uniqueness subgroup.
However, by definition cr<j(G) C <To(G), that is, such uniqueness subgroups do
not exist for any p G cr^(G). Therefore o~<y(G) = 0, proving Theorem Cg.
25. Vertical n e i g h b o r h o o d s
To describe the goal of the analysis in the generic case, we begin by stating a
property of the groups of Lie type in X^7\
First, if K is any quasisimple group of Lie type, it will be convenient to call
inner-diagonal-graph automorphisms of K algebraic a u t o m o r p h i s m s . The term
is natural since these are the automorphisms arising from automorphisms of the
algebraic group overlying K.
25. V E R T I C A L N E I G H B O R H O O D S 119
PROPOSITION 25.1. Let G* e X^ with G* = dL(q) a simple group of Lie

type. Then for some prime p dividing q±l, and specifically for p = 2 if q is odd,
G* contains an Epi -subgroup U* with the following properties:
(1) For some x* € U*#, CG*(X*) contains a component K* with K* a %p-group
and K* terminal in G*;
(2) U* leaves K* invariant and elements of U* — (x*) induce algebraic auto-
morphisms on K*;
(3) CK*(U*) contains a component I* such that the image I of I* in K*U* —
K*U*/CK*u*(K*) is terminal in lC*U*;
(4) For each u* £ U*#, I* is contained in a component Iu* of CQ*(U*); and
(5) For some u* e U* - (x*),I* < Iu*.
Thus Ix* = K*. We call such a configuration (U*,I*,{IU* \u* G (*7*)#})

a vertical neighborhood in G*. By condition (4), each pumpup Iu* of I* is
vertical. The subgroup U* is called the base of the neighborhood. Also each of the
groups Iu* is called a neighbor of (#*, K*).
[The base of a vertical neighborhood consists of semisimple elements, and usu-
ally but not always lies in a torus in the algebraic group overlying G. Nor is it true
conversely that a subgroup isomorphic to Ep2 with p\(q ± 1), and lying in a torus,
necessarily is the base of some vertical neighborhood; the terminality condition may
easily fail.]
We emphasize that we require p = 2 if q is odd. Thus according to the definition,
only groups of Lie type of characteristic 2 possess vertical neighborhoods for odd
p. In fact, groups of Lie type over Fq, q odd, in general also contain Ep2-subgroups
U* satisfying the above conditions for suitable odd p dividing q±l. However, as it
is our intent to identify each group in X^ only once in the course of the analysis,
we enforce this limitation on p.
Furthermore, G* may possess more than one conjugacy class of vertical neigh-
borhoods and the Steinberg relations for G* may be more easily constructed from
one such vertical neighborhood than from another. It is in stage 3 below that we
find a preferred neighborhood.
A slight modification of the above definition allows us to include the six groups
of Lie type in X^ as groups of Lie type possessing a vertical neighborhood. Indeed,
if G* is one of these groups, we let p = 2 or 3 according as the characteristic of G*
is 3 or 2, respectively. Then G* contains a subgroup U* = Ep2 satisfying all the
above conditions except that K* £ S p ; indeed the components Iu* for u* G (J7*)#
are Cp-groups. Namely, for G* = ^7(3), PQg~(3) or ft^(3), two Iu* are 2£74(3) and
the third is P5p 4 (3), 2J74(3) or L 4 (3), respectively; for G* = 2D5(2) or 2E6{2), half
the Iu* are 224(2) and the rest are 1/4(2) or UQ(2), respectively; and for G* = U?(2),
half the Iu* are SUQ(2) and the rest are U$(2).
An entirely analogous result holds for the alternating groups in X^. If G* =
An, n > 13, taking a four subgroup U* generated by two disjoint products of two
transpositions, we obtain three components Iu* isomorphic to A n _ 4 , A n _ 4 , and
An-g. Again we call the corresponding configuration a vertical neighborhood.
Likewise the subgroup of order 4 generated by a pair of disjoint transpositions yields
a vertical neighborhood in the symmetric groups E n , n > 9.
The definition of a vertical neighborhood can be transferred to G in the natural
way. Let G* be a group of Lie type in X^ \JX^7\ or An, n > 13, or E n , n > 11. Let
(£/*, /*, {Iu* | u* G (U*)#}) be a vertical neighborhood in G*. Let U be a subgroup

of G such that the following conditions hold:
(1) There is an isomorphism a : U —* U*;
(2) CQ{U) contains a component / such that mp(CG(I)) = 2;
(3) For each u G U&, I is contained in a component Iu of CQ{U)\ and
(4) There are isomorphisms 7TU : IUU —• Ia(uÛ*, for each -U G E/#, which
coincide on / and extend a.
Under these circumstances, we say that the configuration (U,I,Iy, y G U^) is
a vertical neighborhood (of type G?*).
In practice, it is often the case that the relations determined by the Iu* (and
their intersections) are already determined by a proper subset of the Iu*. Indeed,
this is obviously true if one of the Iu* is I* itself. In such cases, we may suppress
superfluous Iu* in describing the given vertical neighborhood.
26. Theorem C7
In sections 22, 26 and 27 of the preceding chapter, we have described the prin-
cipal moves in the classification of generic groups. Here we formalize the successive
stages of the analysis.
Thus we assume that G is of generic odd type or generic even type. Moreover,
in the latter case if we let <?s{G) be the set of all primes p G cr(G) such that some
element of ££(G) is a 9 p -group, then v$(G) i=- 0 and G contains a g-uniqueness
subgroup for every q G c(G) — <jg(G), i.e., ao(G) C CTQ(G).
If G has generic even type we fix a prime p G cr0(G), so that p G <rg(G) and
G does not contain a strong p-uniqueness subgroup. If G is of generic odd type we
put p = 2, so that by Theorem U(2), G does not contain a 2-uniqueness subgroup.
Choosing a natural ordering on the elements of &P(G) that are SP -groups, one
easily produces an element x G Tp{G) such that CG(X) has a p-component K with
K/Op'(K) a Sp-group and K p-terminal in G. [The exceptions An, 9 < n < 11,
noted after (13.2) could lead to terminal 2An components, 9 < n < 11, but such
configurations have already been analyzed as a side problem in conjunction with
the £!B2-type case, a more natural context.]
The case in which p = 2 and K/02>{K) ^ 2An (n > 12 as Kj02\K) is a
Sp-group) is exceptional and is eliminated by a fusion analysis. In all the other
cases, we use signalizer functor theory to eliminate p'-core obstruction, which is the
aim of the first two stages of the analysis.
Indeed, in the first stage we verify that G is 3/2-balanced with respect to a
suitable Epz-subgroup A of CG(X) suitably embedded in K(x) (see section 29 for
the definition). Since G does not contain a (strong) p-uniqueness subgroup, it
follows with the aid of our uniqueness results that CG(X) cannot be embedded in a
p-component preuniqueness subgroup. This is shown in the second stage to imply
that the associated functor @3/2 on Evi-subgroups of A is trivial (see section 29),
which in turn leads to the elimination of p'-core obstruction.
DEFINITION 26.1. We say that G is of 3/2-balanced type with respect to
x and K if and only if
(1) x G Zp(G) and K is a p-component of CG{X)\
(2) If p = 2, then K/02>{K) ^2An,n> 12; and
(3) G is 3/2-balanced with respect to a suitably chosen Eps-subgroup A oiK(x).
26. THEOREM C 7 121
T H E O R E M C7: STAGE 1. Let G be of generic type. Then G is of3/'2-balanced

type with respect to a suitable x G T^{G) and p-component K of CQ(X) such that
K is p-terminal in G and KjOv>(K) is a %v-group.
The next definition describes the aim of stage 2. For uniformity, in the case
that p — 2 and KjOi' (K) = An we shall call any involution of E n algebraic. Note
that n > 9 as K/02>{K) is a S P -group.
DEFINITION 26.2. We say that G is of semisimple typ e (with respect to
re, y and K) if and only if the following conditions hold:
(1) x G Zp{G) and K is a component of CG{X)\
(2) K is terminal in G;
(3) y G TP(CG{X)), y induces a nontrivial algebraic automorphism on K and
CK{V) contains a component / (for p = 2 or 3, possibly a solvable 51,2(3) or
SUs(2) component) such that the image i" of / in K(y) = K(y)/Cjc{y)(K)
is terminal in K(y); moreover, for every u G {x,y)#, I < E(CG{U)) (with
an appropriate modification in the 52^(3) and SUs(2) cases).
T H E O R E M C7: STAGE 2. If G is of3/2-balanced type with respect to x and K,
then for a suitable y, G is of semisimple type with respect to x, y and K.
As described in Section 26 of the preceding chapter, Stage 3 has a number of
conclusions, involving the isomorphism type of K as well as properties of a sub-
terminal (#, If )-pair (2/, / ) and its associated neighborhood 3Sf = ((#, y), I, {Iu \ u G
(x, y)#}), where Iu denotes the pumpup of i" in CG{U). These properties are ex-
pressed in terms of the definition of vertical neighborhood and, when K is of Lie
type, of the definitions of both a level neighborhood and a splitting prime. More-
over, when G is of even type (whence necessarily K is of Lie type of characteristic
2), a choice of the prime p G cr$(G) is necessary to achieve splitting.
We shall not repeat the rather detailed conclusions of Stage 3, but for brevity
shall refer to them by saying that G is of proper semisimple type (with respect
to N). Thus in Stage 3, we prove
T H E O R E M C7: STAGE 3. There exists a prime p G o~(G) and a p-terminal S P -
pair (#, K) possessing a subterminal pair (y, / ) and associated neighborhood DSf such
that G is of proper semisimple type with respect to 3ST .
We define the span of N to be the subgroup
G0{N) = {Iu\ue{x,y)#).
Likewise as described in the preceding chapter, the goal of Stage 4 is to prove

T H E O R E M 67: STAGE 4. If G is of proper semisimple type with respect to the
neighborhood 3sT, then G0(N) ^ G* for some G* G ft(7).
We take the conclusion of Stage 4 as the definition of the term G w G*.
Finally, using 2-uniqueness results, we prove
THEOREM C7: STAGE 5. IfGzzG* for some G* G X^\ then G ^ G*.
E. P R I N C I P A L T E C H N I Q U E S O F T H E P R O O F
27. Fusion
We briefly describe the principal methods underlying the proof of the Classifi-
cation Theorem. We make no attempt to be comprehensive, but rather single out
some of the key techniques that are basic for the analysis. Since the purpose here
is solely to illuminate the proof, we shall omit the references needed to justify our
assertions. Throughout, as always, G is a 3C-proper simple group.
We begin with fusion. Fusion analysis is most critical for the prime 2 in the
study of groups of special odd type, but certain general consequences of the Alperin-
Goldschmidt conjugation theorem are critical for other primes and in other cases.
The first two results concern 2-fusion. Fix S G SyfaiG).
THEOREM 27.1. (Glauberman's Z*-theorem) If z e T2{S), then z9 e S - (z)
for some g G G.
T H E O R E M 27.2. (Extremal conjugation) Suppose T < S with S/T cyclic. If
y G X2(S — T), then the following conditions hold for some g G G:
(i) y9 G T;
(ii) Cs(y9) G Syl2(CG{y9)); and
(hi) Cs(y)9<Cs(y9).
This is a version of the Thompson transfer lemma. Involutions x of S for which
Cs(x) € SU12{CG(X)) a r e s a id to be e x t r e m a l in S (with respect to G).
Next we state the Alperin-Goldschmidt conjugation theorem.
THEOREM 27.3. Let P G Sylp(G), and let A be a subset of P. If g G G and
A9 < P, then there exist subgroups Di of P and elements gi G Ni = NG(Di),
1 (Ni) xDi,l<i< n;
(hi) NiHPe Sylp{Ni), l<i<n;
(iv) i C D i and A9192"'9* C Dj+1, l<j<n-l; and
(v) A9 = A9l92"'9n, indeed g = cgi • • • gn for some c G GG(A).
The theorem asserts that all G-fusion in P can be "factored" as a product of
conjugations within p-local subgroups of restricted types. Furthermore, as Z(P) <
Ca{Di), (i) and (ii) imply that Z(P) < Di for each i.
It is this last fact that enables one to establish the following corollary of the
theorem.
THEORE M 27.4. Let M be a subgroup of G containing P and z an element of
order p in Z(P). If CQ{Z) < M and M controls the G-fusion of z in P, then the
following conditions hold:
(i) If Q is a p-subgroup of G containing z, then NQ{Q) < M; and
(ii) M controls G-fusion in P.
[The definitions of control of fusion appear in Section 4.]
28. THE BENDER METHOD 123
28. The Bender method

Underlying the Bender method is the Bender uniqueness theorem, a basic result
which is critical in several parts of the classification proof, including the analysis of
groups of £!B2-type and the proofs of the solvable signalizer functor theorem and the
Odd Order Theorem. Recall that for any group X, the generalized Fitting subgroup
F*(X) is the product of the layer E(X) and the Fitting subgroup F(X) of X, that
is, the product of all components and normal p-subgroups of X. Furthermore, for
any set of primes 7r, On(X) is the unique largest normal 7r-subgroup of X, and n'
is the set of all primes not in 7r. We again limit the discussion to G.
DEFINITION 28.1. Let M and TV be maximal subgroups of G. Then M —> N
if and only if there is 1 ^ Q < F(M) such that NF*(M)(Q) < N.
T H E O R E M 28.2. (Bender) Let M and N be maximal subgroups of G such that
M ~^> N, and let n be the set of primes dividing \F(M)\. Then
(i) / / |TT| > 1, then Ov(F(N)) < M ;
(ii) 0^{F{N)) fl M = 1; and
(iii) / / also N ^ M, then either M = N or both F* (M) and F* (AT) are p-groups
for the same prime p.
To illustrate the Bender method, we consider the analysis of the £332-case. We
fix an involution z and consider maximal subgroups M of G containing CG(Z). The
analysis divides naturally into three cases:
(1) For some such M, F*(M) has even order;
(2) For any such M, F*(M) has odd order but not prime power order;
(3) For some such M, F*(M) is a p-group for some odd prime p.
In case (1), the fact that G is DC-proper together with the Bender-Suzuki the-
orem yields in almost all cases that M = CQ(Z). This in turn leads easily to the
£ 2 -property for CG(z), i.e., L2,(CG{z)) = E(GG(z)).
The specific analysis in case (2) depends on the nature of a Sylow 2-subgroup
S of G; for simplicity let us assume here that S is dihedral, a case analyzed by
Bender. An examination of DC-groups shows that if N is any maximal subgroup of G
containing z, then N has the following critical property: for every CJV(Z)-invariant
subgroup W of N of odd order, [z, W] < 0(N). Bender chooses a suitable M and,
using this property, makes a sequence of applications of the uniqueness theorem.
Eventually it applies to M and M9, where g G G is chosen so that s = z9 and sz
both centralize z and are (7-conjugate but not M-conjugate to z. (The existence
of g follows from fusion analysis and the Bender-Suzuki theorem.) The uniqueness
theorem yields that M — M9 so that g G M, a contradiction since z and s are not
M-conjugate.
Thus, as is typical of the method, we are reduced to the case that a suitably
chosen maximal subgroup M of G has the property that F*(M) = Op(M) for some
odd prime p. Unfortunately, in the presence of quadratic modules, this case has
been amenable to successful analysis only under very restricted assumptions. The
definition of a group of £232-type in fact incorporates the extent to which we have
been able to push through this final case, whose elaborate resolution accounts in
part for the length of the analysis of groups of £232-type. It is from this analysis
that the parabolic structure of G emerges leading to the case G w L 3 (p n ), p odd.
Fortunately, the signalizer functor method for the prime 2 is applicable when G is
not of £232-type.
29. Signalizer functors and fc-balanced groups

The signalizer functor method is built on two theories. One consists of the
signalizer functor theorems, from the basic solvable signalizer functor theorem to
McBride's extension to the nonsolvable case (Theorem 29.2 below). The other,
developed by Gorenstein and Walter, concerns the existence of fc-balanced signalizer
functors for various values of k. The signalizer functor method is utilized for all
primes p in both the generic and special cases.
We fix a prime p and a positive integer k. First denote by £P{X) the set of
elementary abelian p-subgroups of the group X and by Epk{X) the subset of EP(X)
consisting of those elements of rank > k.
DEFINITION 29.1. If A e £P(GQ, an A-signalizer functor on G is a corre-
spondence G associating to each a G A& an A-invariant p'-subgroup S(CG(a)) of
CG(a) such that
0(CG(a)) n CG(a') = e(CG(a')) n CG(a) for all a, a' e A*.
Furthermore G is said to be solvable if each 0(CG(a)) is solvable and to be
trivial if each S(CG(a)) = 1. In addition, the closure G(G; A) of G in G is defined
to be
G(G',A) = (G(CG(a))\aeA#).
The signalizer functor G is said to be closed if and only if its closure is a p'-group;
it is said to be complete if and only if it is closed and in addition, S(CG(a)) =
C&(G-A)(a) f° r all a G A#.
Note that for p = 2, each G(Cc»(a)), being a 2'-group, is of odd order and so is
solvable. Hence in this case all A-signalizer functors on G are necessarily solvable.
T H E O R E M 29.2. (Signalizer functor theorem) If A e 8%(G) and G is an A-
signalizer functor on G, then G(G;A) is a p'-group, and moreover® is complete.
R E M A R K . In the solvable case, the theorem in this generality is due to Glauber-
man. In the general case, the only known proof uses the hypothesis that G is a
3C-proper group in case G is not solvable, the case dealt with by McBride. The
reason is that the proof requires certain properties for automorphisms of order p
of simple sections of G of order prime to p; in fact these necessary properties are
enjoyed by all simple X-groups of order prime to p.
Now we turn to the second pillar of the method, the /c-balanced signalizer
functors. For any group X and any B G £ P (X), define Ax(B) by
AX(B)= f) 0„(Cx(b)).
beB*
DEFINITION 29.3. If A e £fc+i(G), we say that G is ^-balanced with respect

to A if and only if for each B e E^ (A) and each a € A#
AG(B)nCG(a)<Op,(CG(a));
and we say that G is weakly /c-balanced with respect t o A if and only if for
each such B and a
[AG(B)nCG(a),A}<Op,(CG(a)).
29. S I G N A L I Z E R F U N C T O R S A N D K - B A L A N C E D G R O U P S 125
The following result shows that the existence of ^4's with either of these prop-
erties is sufficient for most applications to yield signalizer functors on G.
THEOREM 29.4. Let A G £^+2 (*-*)> ^ a P°s^ve integer, and assume that G is
either k-balanced or weakly k-balanced with respect to A. Correspondingly, for each
a G A&, define S(CG(a)) as follows:
e(CG(a)) = (AG(B) H CG(a) \ B G E*(A)); or
e(CG(a)) = (AG(B) n CG(a) \ B G £pk(A))Op,(CG(A)).
Then B is an A-signalizer functor on G.
We refer to B as t h e ^-balanced or weakly ^-balanced A-signalizer func-
tor on G.
Thus in either case the signalizer functor theorem implies that B is complete,
and in particular that the closure B(G; A) of A is a //-group.
In the /c-balanced case, one can easily establish that
e(G;A) = (AG(B)\BeZk(A)).
Similarly, for any E G £fc +i(A),
(29.1) Q(G; A) = {AG(B) \ B e £*(£)) = 6(G; E).
However, NG(E) obviously normalizes the right side of this equation, so we
have
T H E O R E M 29.5. Let A G £^+2^), k a positive integer, and assume that G is
k-balanced with respect to A. Let B be the k-balanced A-signalizer functor on G.
Then
(29.2) TA^1(G)<NG(0(G;A)).
On the other hand, because of the presence of the factor Op'{CG{A)) in each
G(CG(a)) in the weakly /c-balanced case, it is not possible to verify (29.1) in that
case and thus obtain an analogue of Theorem 29.5. However, there exists an effec-
tive substitute for Theorem 29.5 under slightly stronger conditions, developed by
Goldschmidt and Aschbacher. The conditions are incorporated into the following
definition.
D E F I N I T I ON 29.6. If A e ££ + 1 (G), we say that G is (k + ^-balanced with
respect to A if and only if G is both weakly ^-balanced and (k + l)-balanced with
respect to A.
Moreover, suppose that A G £^ + 2(^0 a n c ^ G is (k + ^-balanced with respect
to A. For each E G £pk+1(A), set
(29.3) ekH(G;E) = <AG(£7), [AG(D),E]\De EP(E)).
We refer to B f c + i as the associated (k + |)-balanced functor on G.

There is no signalizer functor B f c + i , but we use the notation B f c + i(G;i£) be-
cause of the following analogy with (29.1), which is proved by a direct generalization
of an argument made originally by Aschbacher in the case p — 2, k — 1.
(29.4) e f c + i ( G ; A) = Gk+,(G;E) for any E € Epk+1(A)-

However, it is immediate from the definition of O fc+ 1 (G; E) that it is invariant

under NQ{E). Thus we have
T H E O R E M 29.7. If A e £fc+2(G0; & a positive integer, andG is (fc+|) -balanced
with respect to A, then
TA,k+1(G)<NG(Gk+i(G;A)).
Though B fc+ 1 (G; A) does not arise from an A-signalizer functor on G, we have
by its definition and from weak /c-balance the condition
(29.4) ek+h(G;A)<G(G;A),
where O denotes, as above, the weakly ^-balanced A-signalizer functor on G. As

G is an A-signalizer functor on G, it follows from Theorem 29.2 and (29.4) that
©fc+i(G; A) is a p'-group, which is the critical use made of the signalizer functor
theorem. Thus Theorem 29.7 is indeed an effective substitute for Theorem 29.5,
allowing one to extend the signalizer functor method to the case of (h+ ^)-balance.
The value of this is that for the applications, (k + ^)-balance is a significantly
weaker condition to verify than fc-balance, for a given value of k.
It is thus critical to be able to verify either fc-balance or (k + |)-balance with
respect to A. As shown by Gorenstein and Walter, this verification can be reduced
to checking local properties of certain elements of £ P (G), namely the components
of CG{Q)/OP>(CG(CL)) as a ranges over A&. For convenience, denote this subset of
LP(G) by HP(G; A).
DEFINITION 29.8. Let K be a quasisimple 3C-group. We say that K is locally
A>balanced or weakly locally fc-balanced if and only if for each H such that
Inn(K) < H < Aut(K), each E e £pk(H) and each x e XP(CH{E)), we have
respectively
AH(E) = 1 or
[AH(E),x] = l
Furthermore, we say that K is locally {k + |)-balanced if K is both locally
(k + l)-balanced and weakly locally /c-balanced.
The following result shows the connection between local balance and balance.
THEOREM 29.9. Let A G £/C + 2(£0> & a positive integer. If each element of
Jdp(G',A) is locally k-balanced, then G is k-balanced with respect to A. If each
element ofXP(G; A) is locally (fc-h ^-balanced and A-invariant, then G is (k-\- \)-
balanced with respect to A.
REMARKS . A more practical version of the theorem holds, with the hypothesis
that the elements of LP{G\A) satisfy only a weaker form of local fc-balance or
(k + |)-balance, relativized to A. The theorem also indicates that verification of
(k + ^)-balance requires additional analysis in the presence of components not left
invariant by A.
30. IMBALANCE, THE £ p -PROPERTY AND PUMPUPS 127
30. Lp> -balance, the £? p -property and pumpups

These topics were introduced in sections 6 and 7 of the previous chapter. We
make some further remarks here, following the work of Gorenstein and Walter.
Throughout this section, p is a fixed prime. Combining the fact that G is JC-proper
with L p /-balance and the Z?p-property, we obtain:
THEORE M 30.1. LetK be ap-component ofCG(x), x eTp(G), y eIp(CG(x)),
L — (K(yS)), and let I be a p-component6 of CL(V). Then the following conditions
hold:
(i) I < Lp,(CG(y));
(ii) Set CG(V) = Cc(y)/Op'(CG(y)) and let J be the normal closure of I in
Lp>(CG(y)). Then I is a component of Cj(x), and one of the following
holds:
(a) 1 = 1;
(b) J is a single x-invariant p-component of Cc(y) with I < J; or
(c) J is the product ofp isomorphicp-components J i , . . . , Jp ofCc(y) cycled
by x, with I/Op>(I) a homomorphic image of each Ji/Op>(Ji).
In (ii) one can say "component" rather than "p-component" by virtue of the Bp-
property. The pumpup J of / is trivial, vertical and proper, or diagonal according
as (a), (b), or (c) holds. When I covers K/Op,{K), i.e., y e CCG{x){K/Opf(K)),
then J is also called a pumpup of K itself.
The crucial significance of this result is that one is able to determine the possible
isomorphism types of J from that of I. Namely, if J is quasisimple, then one may
use the known tables of components of the centralizers of automorphisms of order
p of quasisimple JC-groups; if J is not quasisimple, then each component of J is
isomorphic to a covering group of i". This fact underlies the passage to p-terminal
components described in section 18 as well as innumerable specific calculations.
Of course, if J < E(Cc{y)), then J itself is a product of components of Cc(y)
and the J5p-property implies that I is quasisimple and a component of Cj(x).
There exists an important related result, known as L*,-balance, that describes
the embedding of OP'(CG{X)) PI Cc(y) in Ca{y)- To state it, we need some further
notation.
First, observe that if X is any group with Op'(X) = l , p a prime, then clearly
F*(X) = E{X)Op(X) and E(X) = Lp,{X).
30.2. Let X be a group with Op>(X) = 1, and let P be a Sylow
DEFINITION
p-subgroup oiE{X). We define L*,{X) to be E(X)Op>(Cx(P)). More generally, if
Op>(X) ^ 1, we define L*p/(X) to be the full preimage in X of L$,(X/Op>(X)).
Note that it is immediate from the definition that L*, (X) normalizes each p-
component of X. Moreover, by a Prattini argument, the definition is independent
of the choice of P.
THEOREM 30.3. (L*,-balance) Let X be a %-group and N a p-local subgroup
ofX. ThenL*p,(N)<L;,(X).
6
Thus either L = K (when y leaves K invariant) or L is the product of p isomorphic p-
components of CQ(X) cycled by y\ and in the latter case, J is a "diagonal" of L, so that I/Op/p(I)
projects isomorphically onto each component of L/Opfp(L), and I/Op/(I) is isomorphic to a
homomorphic image of K/Op/(K).
Like Lpf-balance, this has consequences for pairs of commuting elements of G

of order p, an important example of which is the following.
T H E O R E M 30.4. If x and y are commuting elements of G of order p and we
set D = OP'(CG{X)) H Cc{y) and Cc{y) = Co(y)/'Op>{Cc{y)), then the following
conditions hold:
(i) D<L;,(CG(y));
(ii) D normalizes every p-component ofCc{y);
(iii) D centralizes every component of Co(y) that is either centralized by x or
not normalized by x; and
(iv) D centralizes Op(CG{y))-
If D 7^ 1, then by the F*-theorem there is a component K of Cc{y) not
centralized by D. By the theorem TV = KD(x) is a group with K < N and x not
centralizing ~K. Hence if we set iV* = ~N/Cjf(K), then K* is simple, iV* < Aut(K*),
D* 7^ 1, and x* ^ 1. Moreover, as D < Op>(CcG(y){x)), one can show that
D* < Opf (CM* (#*)) and consequently according to the definition K* is not locally
1-balaneed with respect to {x*). Thus K* is limited to the simple 3C-groups that
are not locally 1-balanced for the prime p, and the structure of D* is limited by
that of Op/(Cjv* (#*))> which is always small in some sense of the word, because K
is a DC-group.
31. The signalizer functor method

We now explain the objectives of the signalizer functor method as it relates
to the X-proper simple group G. In brief, trivial signalizer functors lead to the
elimination of p'-core obstruction, while nontrivial signalizer functors lead to the
existence of p-uniqueness subgroups.
This principle is most evident when mp(G) > 3 and every element of &P(G) is
locally balanced (i.e., locally 1-balanced) for the prime p—for instance, when &P(G)
is empty. In that case G is balanced with respect to every A € £>3(G). Using the
signalizer functor theorem, one obtains the following result rather easily:
T H E O R E M 31.1. Let p be a prime such that mp(G) > 3. / / every element of
LP(G) is locally balanced for p and A is any element of E^(G)f then according as
the 1-balanced A-signalizer functor G on G is trivial or nontrivial, we have:
(i) G(G; A) = 1 and Op,(CG{x)) = 1 for all x e 1P(G) with mp{CG(x)) > 3; or
(ii) 0(G; A) ^ 1, M = NG(@(G; A)) is a proper subgroup of G, and for any
P e Sylp(G) containing A, T°P2{G) < M.
We remark once again that when p = 2 and rri2(G) > 3, the Thompson transfer
lemma implies directly that rri2(CG(x)) > 3 for every x G ^(G), so that if (i) holds,
then 0(CG(X)) = 1 for every involution x of G. In particular, under the hypotheses
of the theorem, we conclude that either every 2-component of CG (X) is quasisimple
for every x G %2(G) or M is a 2-uniqueness subgroup. For p odd the conclusions
are still very strong as well.
Application of the signalizer functor method is more elaborate in the presence of
elements of &P(G) that are not locally balanced for p. In the first place, the analysis
becomes "relativized", centered around a suitably chosen element x G 1p{G) and
p-component K of CG(X) with K p-terminal in G. The pair x and K are found via
the pumping up process, in practice K being chosen so that K/Op>(K) is maximal
32. N E A R C O M P O N E N T S , P U S H I N G UP, A N D F A I L U RE O F F A C T O R I Z A T I O N 129
in a suitable ordering of the elements of £ p ( G ) . Usually this ordering yields in

particular that K/Op>(K) is "well-generated" with respect to elementary abelian
p-subgroups of CG{X) acting on it.
One then argues that there is A £ EP(CG(X)) with mp(A) > k + 2 for some
positive integer A: for which G is either ft-balanced or (fc+ |)-balanced with respect
to A. Moreover, one chooses A so that k is as small as possible (the best one can
hope for in general is k = 1 with G being |-balanced with respect to A). Then
(31.1) TAik+1(G)<NG(Q{G;A)),
where <£> is either the fc-balanced A-signalizer functor G or the associated (k + \)-
balanced functor Ok+i on G, respectively.
The goal of the analysis is entirely similar to that in the 1-balanced case above,
relativized to x and K: If $(G;A) = 1, the aim is to show that K and certain
of its neighbors are semisimple, while if <&(G\A) ^ 1, the aim is to show that
M = NG{®{G\A)) is a p-component preuniqueness subgroup containing CQ(X).
The value of k is important here, for a crucial step in verifying that CQ{X) < M
is the inclusion K < M , which will follow from (31.1) provided K/0P'{K) is well
enough generated with respect to A, namely,
K<rA,k+1(G).
The smaller the value of &, the fewer "non-generated" counterexamples K to this
desired condition there will be. On the other hand, the smaller the value of k, the
less likely it is for /c-balance or k + ^-balance to hold; the choice of k is therefore
a compromise dictated by the specific nature of the elements of LP{G\A). The
problem with many of the groups in T p is that there is no satisfactory value of k.
32. Near components, pushing up, and failure of factorization

In the next three sections we describe some general techniques that are central
to the 2-local analysis of groups of even type. The results of this and the next
section are crucial in the study of the quasithin case as well as in certain uniqueness
theorems: Theorem U(cr) and Theorem JA(S).
We refer the reader to section 10 for the definitions connected with near com-
ponents. We first describe the general context that leads to the existence of a near
component. It is the existence of a subgroup M of G (again we limit the discussion
to our OC-proper simple group G) which "controls" many 2-local subgroups of G
and a second subgroup X of G which represents an obstruction to the control of
further 2-local data by M. The conditions on X are the following:
(1) X £ M;
(2)F*(X ) = 0 2 ( X ) ;
(3) There is T € Syl2{X) with T < M ;
(4) NG(T0) < M for all 1 + TQ char T; and
(5) C(X,T)<X.
Here C(X, T) is defined by C(X, T) = (NX(T0) \ 1 ^ T0 char T), and condition
(5) is an obvious consequence of the other conditions.
Aschbacher has established the following general result [ A l l , A12], known as
the local C(X, T)-theorem, about groups X satisfying conditions (2) and (5). For
us, as G is 3C-proper, his result is only required under the additional assumption
that X is a 9C-group.
T H E O R E M 32.1. Assume that X is a %-group with F*(X) = 02{X). Then
X is the product of C(X, T) and all the linear and alternating near components K
of X such that K/02(K) = L2(2n), n > 2, or A2™+i, n > 1. Consequently, if
C(X, T) < X, then X possesses such a near component.
The proof of Theorem 32.1 involves two techniques that are fundamental for
the analysis of groups of even type. The first step of the proof concerns failure
of Thompson factorization and quadratic modules, introduced in section 8 of the
previous chapter.
Let
A(T) = {A | A < T, A 2* E2n, n = m2(T)} and
J(T) = (A\AeA(T)).
J(T) is called the Thompson subgroup of T. Also set ZX(T) = f&i(Z(T)). Then
Z\(T) and J(T) are each characteristic in T and so by our assumption on C(X,T),
certainly
(32.1) CX(Z1(T))NX(J(T))<X.
If we set V = (ZX(T)X), then the condition F*(X) = 02(X) implies that V is
an elementary abelian normal 2-subgroup of X. Furthermore, by (32.1), if we set
C = Cx(V), it follows by a Prattini argument that J(T) ^ C. Indeed if we put
X = X/C, the following conditions quickly follow for some A G A(T):
(1) V is a faithful X-module;
(2) 02(X) = 1;
(3) A 7^ 1 and A is elementary abelian; and
(4) \A\\CV(A)\ > |n
with (2) following from the definition of V, and (4) following because A G A(T)
has the consequences \A\ > \(Ar\C)V\ and VC\ A = CV(A) = CV(A). Therefore V
is a failure of factorization X-module, by definition. Furthermore, an application
of the Thompson Replacement Theorem shows that for a suitable choice of A, the
following condition also holds:
[V,A,A] = 1,
so that V is quadratic with respect to A. This is the situation that is first analyzed
in the proof of Theorem 32.1.
When F*(X) is solvable, Thompson's original analysis, refined by Glauberman,
shows that SL2(2) = £3 is involved in X and describes precisely how it is involved.
When E(X) ^ 1, the following basic result originally obtained by Aschbacher is
used for a first reduction in the structure of X. [The proof can now take advantage
of the recent analysis of quadratic modules for DC-groups by Meierfrankenfeld and
Stroth [MeStl].]
T H E O R E M 32.2. Let Y be a group with F*(Y) e X. IfY possesses a failure of
factorization module W, then F*(Y) e £hev(2) UAlt.
Failure of factorization modules as well as quadratic modules have been ex-
tensively studied, and the possibilities for W have been determined for various
33. T H E A M A L G A M M E T H O D 131
choices of the X-group F*(Y). Such information is needed in the analysis of all the
theorems mentioned at the beginning of this section.
When E(X) ^ 1, one can arrange to apply Theorem 32.2 to suitable com-
ponents of X = X/Cx(V). We consider here only the case that there is such a
component in Ghev(2). Then one can use the generation of the groups in Ghev(2) of
Lie rank at least 2 by their parabolic subgroups and a further reduction argument
of Baumann to reduce the proof of Theorem 32.1 in the Lie type characteristic 2
case to the following minimal case:
X/02(X)^L2(2n), n>2.
In this case, Nx(T) is, in fact, the unique maximal subgroup of X containing
T, whence the condition C(X,T) < X is equivalent to the assertion:
No nontrivial characteristic subgroup of T is normal in X.
The Baumann-Glauberman-Niles theorem covers this special case of the local
C(X,T) theorem, namely:
T H E O R E M 32.3. Let X be a group such that F*(X) = 02(X) and X/02(X) =
L2(2n), n > 1. If no nontrivial characteristic subgroup of a Sylow 2-subgroup T of
X is normal in X, then 02{X) is a near component.
Theorem 32.3 is an example of a pushing up situation, characterized in general
by replacing the condition C(X,T) < X by the condition that for some subgroup
A of Aut(T), no nontrivial ^-invariant subgroup of T is normal in X. In practice,
A arises within NQ{T). Such pushing up problems with other possibilities for
F*{X/02(X)) besides L2(2n) have also been extensively studied.
33. The amalgam method

The amalgam method, introduced by Goldschmidt [Go6], is a general technique
for studying groups generated by a pair of subgroups without a common normal
subgroup. The most important applications occur in the following situation:
(1) X is generated by two subgroups Mi and M2;
(2) No nonidentity subgroup of Mi D M2 is normal in both Mi and M 2 ;
(3) Mi and M2 share a Sylow p-subgroup S for some prime p; and
(4) F*(Mi) = Op(Mi),i = 1,2.
The amalgam method involves a combination of local group theory and geom-
etry, with large portions of the argument not requiring X to be finite. The purpose
is to understand the amalgam consisting of Mi, M2, and the inclusion mappings
of Mi fl M2 into these subgroups, rather than X, which is a "completion" of this
amalgam. Indeed, one typically replaces X by the free amalgamated product of
Mi and M2 amalgamated over the subgroup Mi n M2, and the main goal of the
method is to force the structure of the "parabolic" subgroups Mi and M2, given
the structure of the factor groups Mi/Op(Mi). Short of this, one aims at least to
bound the number of chief factors of each Mi contained within Op(Mi), which in
particular will bound \Op(Mi) |, i = 1,2. In practice, this method applies with p = 2
in the analysis both of quasithin groups and of Theorem U(cr). In both cases, the
g-strueture of Mi/Op(Mi) for odd primes q is restricted severely by the hypotheses
of the respective cases, viz., cr(G) = 0 and CFQ(G) = 0.
One associates a graph T with the given configuration: the vertices of T are the
right cosets of Mi and M 2 in X, with two distinct vertices defined to be adjacent
if they have a nontrivial intersection.
Obviously distinct cosets of Mi have trivial intersection for each i — 1, 2, so if
two distinct vertices MjX, M^y for x,y e G are adjacent, then necessarily j ^ k, so
by definition T is bipartite. Furthermore, X acts by right multiplication on T, and
this action is faithful because of (2) above, so that X < Aut(T).
One easily establishes the following elementary properties of this set-up.
PROPOSITION 33.1. The following conditions hold:
(i) r is connected;
(ii) X operates edge-transitively but not vertex-transitively on T;
(iii) The stabilizer in X of a vertex ofT is an X-conjugate of Mi or M2;
(iv) The stabilizer in X of an edge ofT is an X-conjugate of M\ D M2; and
(v) If X is the free amalgamated product of Mi and M2 over Mi D M2, then T
is a tree.
For simplicity, one denotes the vertices of T by letters a, /3, 7, etc., with Xa,
Xp, Xy, etc., their respective stabilizers in X. For a vertex a of T, let T(a) be the
set of vertices adjacent to a. Then the following facts are also part of the set-up.
PROPOSITION 33.2. Xa is transitive on T(a) for any a e T.
PROPOSITION 33.3. Let a, (3 be adjacent vertices in T and let T < Xa H Xp .
If (ATG(T))7 is transitive on Tfr) for both 7 = a and 7 = ft, then T = 1.
In the ensuing discussion, we shall assume that p = 2 and shall describe briefly
the key ingredients and objectives of the amalgam method. For any a G T, let
Qa be the kernel of the action of Xa on T(a). By (33.1)(3) and Proposition 33.2,
Qa ^ 02(Xa), and for simplicity we shall also assume that Qa = C>2(Xa) for each
a.
Also, let Sa e Syl2{Xa) and set Aa = Q1(Z(Sa)). Since CXa(Qac) < Qa by
our hypothesis (33.1)(4) on Mi and M2, we have Aa < Qi(Z(Qa)). Hence if we
set
then Za < î(Z(Q a ))_and Za < Xa.

Finally, if we set Xa = Xa/Cxa(Za), it quickly follows that
02(Xa) = 1, and
Za is a faithful X a -module.
The analysis focuses on the action of the subgroups Za on the graph T, more
specifically on connected paths
a = ai ot2 &b-i otb = P
o o ••• o o
of minimal length b with the following property: Za fixes each vertex of r(a»)
for all 1 < i < 6, but does not fix all the vertices of r(a^) = r(/3).
Group-theoretically, this is equivalent to the assertion that
Za < Qoa for all 1 < i < 6, Za < Xp, but Za £ Qp.
33. T H E A M A L G A M M E T H O D 133
The goal of the analysis is to force the integer b to be small, and use this to
bound the number of chief factors of X contained within Qa.
Now since Za < Xp, Za acts on Zp. Furthermore, because of the minimal
choice of b, it follows that likewise Zp < Xa and so Zp acts on Za. Therefore
[Za, Zp] < Za n Zp, so [Za, Zp, Zp] — [Zp, Za, Za] = 1. Thus the action of Zp on
Za is quadratic as is the action of Za on Zp.
The analysis divides into two major cases according as
[Za, Zp) ^ I or [Za,Zp} = I.
In the first case, Zp ^ Qa and so there is symmetry between a and /3. Defining
Ca — Cza(Zp) and Cp = C ^ ( Z Q ) , we can therefore choose notation so that
\za\/\ca\ > \z0\/\c0\.

However, this immediately yields that
\Za\\Cz,{Za)\ > \Z0\,
and hence that Zp is a failure of factorization module for Xp .

Thus the analysis in the first case leads directly to the study of failure of factor-
ization modules for the group Xp in its action on Zp . It is here that the possible
structures of the groups Mi/O^Mi) enter critically. Because of the restrictions on
these groups in the applications, as noted above, the possible structures of Xp are
severely restricted.
The analysis in the second case is similar to the first, but somewhat more
complicated. We mention here only one subcase, related to the local C(X,T)-
theorem. It involves a and an adjacent vertex 7. Then Qa < X7. The assumption
in this subcase is that
(33.2) Qa e Syl2(H7), where H7 = (Q«y)-
Assume also that
(33.3) 02((Xa,N)) = l for all N < X1 such that N £ Xa.
This condition can frequently be arranged by an appropriate initial choice of

Mi and M2, and holds for instance if Xa is a maximal 2-local subgroup of G.
Now (33.3) implies that for any 1 ^ Qo char Qa, we have Nx (Qo) ^ Xa fl X1.
Moreover, by (33.2) and a Frattini argument, Nxy(Qa) and hence Xa D Xy cover
X1jK1. But Xy ^ Xa and so H7 n Xa < H1. Thus
C(tf 7 ; Qa) = (NH^(Q0) \1^Q0 char Qa) <H7nXa< ff7.
In view of (33.1)(4) and the local C(X, T)-theorem, H1 therefore necessarily

contains a near component.
34. Local analysis at two primes

To the extent possible, our analysis deals with the p-local structure of G for
a fixed prime p. The choice of p depends on the case being analyzed. However,
at several critical places, there is an interplay between p- and g-local structures
for distinct primes p and q. Such interplay, for example, is intrinsic to the Bender
method and central to the proof of the Feit-Thompson theorem.
We mention here three important examples in the analysis of groups of even
type where p-local and 2-local structures interact. First, in the generic case, once
the existence of a terminal component K has been established in CQ(X) for some x G
T°(G), p G o~o{G), the proof that K G &hev(2) relies on an analysis of centralizers
of involutions y e K, using the fact that CK(V) < CcG(y)(x) and the restrictions
imposed on Cdy) by the condition that G have even type.
Second, in the analysis of groups of £C p -type, the argument in the case where G
is of wide £C p -type moves freely between p-local and 2-local subgroups, and indeed
the goal of the argument, that G has a Z§ x ^-neighborhood, clearly involves both
2- and 3-local structure.
Third, the initial stage of the proof of Theorem U(a) establishes 2-uniqueness
properties of a subgroup Mv which was constructed from p-local analysis as having
certain p-uniqueness properties.
Unfortunately, there is no direct analogue of L^-balance (Theorem 30.2) to
apply to a pair of commuting elements x, y G G of distinct prime orders p and
q. Indeed, if N is a g-local subgroup of the DC-group X, then LP>(N) need not lie
in Lp'(X).7 Nevertheless the obstruction to such an inclusion can sometimes be
pinpointed, as the following elementary proposition shows.
PROPOSITION 34.1. Letp and q be distinct primes, and let X be a %-group such
that Op'(X) = 1. Let Q be a q-subgroup of X and K a component of E(Cx(Q))-
Then one of the following holds:
(i) K < E{X), and Q permutes transitively the set of components of X not
centralized by K; or
(ii) There exists a QK-invariant p-subgroup R of X such that [K,R] / 1.
This proposition, like Lp>-balance, can be applied when x and y are commuting
elements of orders p and q, respectively, OP>(CG{X)) = 1, and K is a component
°f CE{cG(y)){x))i to give the embedding of K in X = CG{X). When the first
alternative holds, we have the analogue of L p /-balance: K < E(CG(X)), and indeed
K is a component of E(CcG(x)(y))i s o if t n e structure of K is known, then the
structure of E{CG(X)) is limited.
On the other hand, the second alternative can sometimes be avoided by choos-
ing x and y appropriately. For example, the Thompson "dihedral lemma", originally
used in the iV~group paper for a similar purpose, is used in this way in the analysis
of groups of wide £C p -type. The lemma is the following.
PROPOSITION 34.2. If the elementary abelian 2-group A of rank n acts faith-

fully on the p-group R, p odd, then the semidirect product AR contains the direct
product of n dihedral groups of order 2p. Moreover, rri2,p(AR) > n — 1.
7
For example, take X to be the holomorph of E = Epn and let q be a prime divisor of p — 1.
Then a complement H to E in X is a g-local subgroup of X , and Lp/(H) = Hf, but Lp/(X) = 1.
35. CHARACTER THEORY AND GROUP ORDER FORMULAS 135
Here the 2-local p-rank of a group X is defined by

m2,P(X) = max{mp(AT) \N<X, 02(N) ^ 1}.
We then can draw the following useful conclusion.
PROPOSITION 34.3. Suppose thatp is an odd prime, x G 1p{G), Op,(CG{x)) —
I, y is an involution of CQ{X), and K is a component of E(CcG(y)(x)). Suppose
that m2(K) - m2(Z(K)) > m2,P(G) + 1. Then K < E(CG(x)).
This proposition is one reason why it is advantageous to study elements x of
order p lying in an element B G 32(G), since by definition mp(B) = m2^p(G).
When G is of even type and E(CG{y)) has a component L, then L G G2 and so
typically the 2-rank of L is much larger than its p-rank. As long as mp(CB(K)) is
not too large, Proposition 34.3 finds application.
We conclude by mentioning another elementary result which is at the heart of
the analysis of groups of wide £C p -type and gives rise to 2-groups of symplectic
type in G.
PROPOSITION 34.4. Let p be an odd prime and let B G S£(G), so that B =
Epn, n = m2)P{G), and B normalizes a nontrivial 2-subgroup T of G. Suppose that
n>2 and that for every hyperplane A of B, F*(CG(A)) is a p-group. Further, if
p = 3, assume that CT(B) ^ 1. Then every characteristic abelian subgroup ofT is
cyclic, i.e., T is of symplectic type or cyclic.
To illustrate this, suppose that G is of £C p -type and OP'(CG(X)) = 1 for all
x G T°(G). Then the components of CG(X) for x G B# lie in Gp. In the simplest
case, these components actually lie in Chev(p) and each one admits a nontrivial
inner automorphism induced by some element of a hyperplane A of B. By i m -
balance and the fact that groups in Qhev(p) are of characteristic p-type, this leads to
the desired condition that F*(CG(A)) is a p-group. The condition that CT{B) ^ 1
can usually be achieved by a suitable choice of B and T.
35. Character theory and group order formulas

The only general result that we use that depends on character theory is Glaub-
erman's Z*-theorem, which generalizes the Brauer-Suzuki theorem on groups with
quaternion Sylow 2-subgroups. Beyond this, character theory is used in the classi-
fication proof only in connection with groups of special odd type (including groups
of odd order) and in connection with certain properties of the sporadic groups and
a few other DC-groups.
The primary use of ordinary character theory is through the Brauer-Suzuki
theory of exceptional characters and related isometries and coherence between the
character ring of G and character rings of suitable subgroups of G. Such character
calculations play a key role in the analysis both of groups of odd order and of groups
with dihedral, rank 2 homocyclic abelian or type ^3(4) Sylow 2-subgroups.
The theory of blocks of characters, in particular Brauer's first and second main
theorems, is essential for the analysis of groups with Sylow 2-subgroups which are
semidihedral, wreathed, or elementary abelian of order 8, but it is also used in the
rank 2 homocyclic abelian and ^3(4) type cases.
One of the principal uses of character theory is to derive formulas for the order
of the group G under investigation in terms of the structure of the centralizer
of an involution. We give such an example below. However, when G has more

than one conjugacy class of involutions, Thompson has derived such a formula by
entirely elementary, non character-theoretic calculations. Frequent application of
this formula is made in the course of the classification of groups of special type, so
we shall state it here.
Assume then that G has more than one conjugacy class of involutions and fix
involutions u,v of G with u not conjugate to v in G. For any involution w of
G, define r(iz, v,w) to be the number of ordered pairs (ui,vi) such that U\ is G-
conjugate to u, V\ is G-conjugate to v, and w G (uiVi). (Since u\ and v\ are not
conjugate, their product must have even order.)
With this notation, Thompson counts the ordered pairs consisting of a conju-
gate of u and a conjugate of v in two different ways to arrive at his formula:
THEOREM 35.1. (Thompson Order Formula) The order of G is given by the
following expression:
r(u,v,w)
\G\ = \CG(u)\\CG(v)\J2 \CG(W)\
the sum taken over a system of representatives w of the conjugacy classes of invo-
lutions ofG.
In the particular case in which there are just two conjugacy classes of involutions
in G, the formula becomes
|G| =r(u,v,v)\CG(u)\ +r(u,v,u)\CG(v)\.
It is for the case in which G has only one conjugacy class of involutions that
character theory is needed to derive analogous formulas for \G\. We shall state
Brauer's result in the semidihedral/wreathed case (under the assumption that G is
what Brauer calls "regular"). His formula depends on the prior determination of
the degrees of the characters in the principal 2-block of G in terms of the structure
of the centralizer of an involution.
Assume then that G has a semidihedral or wreathed Sylow 2-subgroup S and
let z be the unique involution of Z(S). An elementary fusion analysis implies that
CG{Z) does not have a normal 2-complement. But now as G is DC-proper, it follows
that CG{Z) = CG(Z)/02'(CG(Z)) contains a normal subgroup L = SL2(q) for some
odd prime power q. Since G has only one class of involutions, q depends only on
G\ we refer to it as the characteristic power of G, and it is a key entry in the
order formula for G. We write q — pn, p an odd prime.
Note that the target groups here are the groups L^(q), Us(q), q odd, and M n ;
correspondingly their characteristic powers are in fact q, q and 3.
Now let U be a four-subgroup of S with U < S if S is wreathed. Since
m2(S) = 2, z e U, so CG{U) < CG{z).
DEFINITION 35.2. We say that G is regular if and only if the p-part of |G| is
at least as large as the p-part of \CG(Z)\3/\CG(U)\2.
Brauer has shown that G is necessarily regular when S is wreathed. However,
when S is semidihedral, G need not be regular. Indeed, if G = M n , then CG(Z) =
GL 2 (3), whence the 3-part of \CG(z)\3/\CG(U)\2 is 27, but \MU\ - 8 • 9 • 10 • 11 is
not divisible by 27.
Here then is Brauer's formula.
36. IDENTIFICATION OF THE GROUPS OF LIE TYPE 137
T H E O R E M 35.3. Assume that G has a semidihedral or wreathed Sylow 2-sub-

group and is regular of characteristic power q. Then for a suitable value of e = ±1,
we have
|r , \CG{Z)? q2-eq + l
1
' \CG(U)\*' {q-eY •
The two possibilities for e correspond to the two target groups G* = L^(q) or
36. Identification of t h e g r o u ps of Lie t y p e

In sections 11, 26 and 27 of the preceding chapter we have described the Stein-
berg presentation for the groups of Lie type (of Lie rank > 2) and have discussed
several alternative ways of verifying these relations—via Tits's classification of
(J5, iV)-pairs, the Curtis-Tits theorem (Chapter 1, Theorem 27.3) and variations
of it, or the Gilman-Griess theorem for the groups of characteristic 2 and Lie rank
at least 4 (Chapter 1, Theorem 27.6). The last two approaches are most effective for
groups of Lie rank at least 3. [Gilman and Griess have established a modification
of Theorem 27.6, applicable in the Lie rank 3 cases.]
The Curtis-Tits and Gilman-Griess theorems do not apply to groups of Lie rank
at most 2, on the other hand. Variants of the Curtis-Tits theorem can be used if
the untwisted Lie rank is at least 3, for instance to identify Un(q) for n = 4 or 5 by
means of subgroups stabilizing nonsingular subspaces on the natural n-dimensional
module over Fq2. Other groups of Lie rank at most 2 are best identified from their
split (J5, iV)-pair structure, which we now describe in some detail. We begin with
the Lie rank 2 case, where the required conditions are more easily described. As
always, G is a X-proper simple group.
Let G(q) = G* be a simple target group of Lie rank 2. Thus G* - £*iV*£* with
H* = B* H AT* < TV*, # * an abelian p'-group, £* = H*U*, where U* e Sylp{G*),
U* is the product of iI*-root subgroups, the Weyl group W* — N*/H* is a dihedral
group generated by two fundamental reflections, and the Bruhat lemma (Definition
ll.l(iv)) describes the products of double cosets B*'x*B* • B*y*B*, for x*, y* € N*.
Again this description of G* determines G* up to isomorphism. Indeed if G is a
simple split (£?, iV)-pair of rank 2, and we set H = B fl N and let U be a nilpotent
normal complement to H in J5, then one can define i/-root subgroups in U and the
following result holds [Ril,Ti5]:
T H E O R E M 36.1. In the situation just described, if there are isomorphisms W —>
W* and B —> B* carrying H to H* and H-root subgroups of U to corresponding
H* -root subgroups of U*, then G = G*.
Indeed, Tits shows that under these hypotheses, the corresponding buildings
for G and (2* are isomorphic. We also remark that an analogous result holds for
split (JE?, Af)-pairs of rank > 3; however, this is not in general the most efficient
approach for establishing the Steinberg relations.
The condition in the above result that B = B* can in fact be weakened to
conditions which are more appropriate to the results of amalgam analysis (corre-
sponding to a target group G(q), q even) or to the results of analysis of central-
izers of involutions (for the case q odd). If we let P*, i = 1, 2, be the maximal
parabolic subgroups of G(q) containing 5 *, then the P* have Levi decompositions
P* = K*Ul with [ / * n K * = l , [ / * a p-group and K\ the product of # * and a
normal rank 1 subgroup of Lie type. If one assumes that W = W* and G has
subgroups Pi, i = 1, 2, such that Pi > B, P{ = KiUi with K{ ^ X*, \U{\ = \U?\,
and such that certain sections of Ui are isomorphic to corresponding sections of U*,
even as /^-modules, then it turns out that the condition B = B* can be derived
without repeating the complete classification of split (J5, iV)-pairs of rank 2 of Fong
and Seitz (see [GLSl, FoSel]).
The identification problem in the Lie rank 1 case is much more difficult. In this
case, it is necessary to delve into portions of the multiplication table of G outside
proper subgroups to show how the table is determined from the given rank 1 split
(J3, iV)-pair data.
We shall explain this briefly. In this case W = N/H = Z2, and hence if we fix
w e N - H, then
(36.1) G = BNB = BU BwB,
whence G is a doubly transitive subgroup of £ Q , where ft is the set of right cosets
of B. Now the prior analysis reduces one to the case that B = HU with U < B
and U G Sylp(G). In particular, it follows from (36.1) that
(36.2) G = BU BwU,
and the (£?, iV)-pair axioms imply that each element y G BwU has a unique repre-
sentation of the form
(36.3) y = huwv with h G H and u,v E U.
Hence for each t G U#, the element wtw can be uniquely expressed in the form
(36.4) wtw = h(t)g(t)wf(t),
where h(t) G H and /(£), g(t) G U.
It is clear from (36.3) and (36.4) that the multiplication table of G is completely
determined by the structures of B and N = H(w) and the functions / , g and h. But
much less is sufficient. Namely, since G is simple, it is easy to see that G is generated
by U and w. Moreover, because of (36.2) and (36.3), Q, can be parametrized by U
and a single additional symbol. The action of U on Q is then determined by the
structure of U, and the action of w on ft is completely determined by the single
function f(t).
Thus one obtains:
PROPOSITION 36.2. The multiplication table of G is completely determined by
the structure of U and the function f from U# to U^.
Peterfalvi uses this last result in his revision and extension of O'Nan's earlier
characterization of Us(q), q odd, as a split (J5, iV)-pair of rank 1.
37. Properties of X-groups

As the preceding discussion makes amply clear, almost all of the above tech-
niques depend for their validity and usefulness on the fact that the critical subgroups
of G to which they are to be applied are %-groups. Verification of the necessary
%-group properties reduces ultimately — frequently by Bender's F*-theorem (The-
orem 4.8(i) of Chapter 1) — to a combination of specific properties of simple and
almost simple X-groups, and properties of solvable groups.
37. PROPERTIES OF DC-GROUPS 139
We have also indicated the nature of many of these simple %-group properties.
We shall not present any of these results here, but instead shall simply list the
various categories into which they fall:
1. Automorphisms
2. Schur multipliers
3. Structure of local subgroups, especially the p-components of centralizers of
elements of order p, but also other local subgroups
4. Generation, particularly by p-local subgroups
5. Local balance
6. Sylow p-structure and fusion
7. p-rank and embedding of elementary abelian p-subgroups
8. Other subgroup structures
9. jPp-representations for the groups in Qhev(p)
It seems that the smaller the X-groups one encounters in an analysis, the longer
and more detailed is the list of their properties which one needs. Thus the analysis
of simple groups of special type seems to require digging deeply into the minutiae
of their subgroup structures. In contrast, the generic case analysis utilizes only a
shorter list of comparatively general properties of simple 3C-groups, and even within
the set of generic groups, the ones of larger Lie rank are amenable to more general
arguments.
F . NOTATIONAL CONVENTIONS
In general, our notation will be standard. However, our notation for the simple
groups is that of Table I of Chapter 1; this differs in some respects from the Atlas
notation [ C P N N W 1 ] .
For the sake of compactness, we shall adopt some abbreviated conventions,
such as the "bar convention" for homomorphic images: if X, X,X*, etc. is a
homomorphic image of the group X , then the image of any subgroup or subset Y
of X will be denoted by Y,Y,Y*, etc., respectively.
For any subset Y of the ambient X-proper simple group G under investigation,
we shall write Cy and Ny for CG(Y) and NG(Y), respectively. [This notation will
be used only relative to G itself.]
In addition, as the prime p will be fixed in each chapter, we shall write for
any group X, m ( X ) , L(X), J ( X ) , £ fc (X), L(X), etc., for mp(X), L p /(X), XP(X),
£^(X), £JP(X), etc., respectively.
However, this abbreviated notation will be used only in the body of the text,
the introduction to each chapter being written with the standard, fuller notation.
Finally, at the end of each volume, we shall include a glossary of terms and
symbols introduced in that volume.
Background References
Reference [P4] has been added in the Second Printing. It can be used as a substitute Background
Reference for [FTl].
[Al] M. Aschbacher, Finite Group Theory, Cambridge University Press, Cambridge, 1986.
[A2] , Sporadic Groups, Cambridge University Press, Cambridge, 1994.
[BeGll] H. Bender and G. Glauberman, Local Analysis for the Odd Order Theorem, London
Math. Soc. Lecture Notes Series # 1 8 8 , Cambridge University Press, Cambridge, 1994.
[Brl] R. Brauer, Some applications of the theory of blocks of characters of finite groups. Ill,
J. Algebra 3 (1966), 225-255.
[Br2] , Some applications of the theory of blocks of characters of finite groups. IV, J.
Algebra 17 (1971), 489-521.
[Br3] , Character theory of finite groups with wreathed Sylow 2-subgroups, J. Algebra
19 (1971), 547-592.
[Cal] R. W. Carter, Simple Groups of Lie Type, Wiley and Sons, London, 1972.
[Dl] J. Dieudonne, La Geometrie des Groupes Classiques, Springer-Verlag, Berlin, 1955.
[El] M. Enguehard, Obstructions et p-groupes de classe 3. Caracterisation des groupes de
Ree, Asterisque 1 4 2 - 1 4 3 (1986), 3-139.
[Fl] W. Feit, The Representation Theory of Finite Groups, North Holland, Amsterdam, 1982.
[FTl] W. Feit and J. G. Thompson, Solvability of groups of odd order (Chapter V, and the
supporting material from Chapters II and III only), Pacific J. Math. 13 (1963), 775 -
1029.
[Gil] G. Glauberman, On solvable signalizer functors in finite groups, Proc. London Math.
Soc. (3) 3 3 (1976), 1-27.
[Gl] D. Gorenstein, Finite Groups, 2nd Edition, Chelsea, New York, 1980.
[GLl] D. Gorenstein and R. Lyons, The local structure of finite groups of characteristic 2 type
(Part I only), Memoirs Amer. Math. Soc. 276 (1983).
[GL2] , On Aschbacher's local C(G,T)-theorem, Israel J. Math. 8 2 (1993), 227-279.
[Hunl] D. C. Hunt, A check of the Ree group conjecture for small fields, Invent. Math. 58 (1980),
99.
[Hul] B. Huppert, Endliche Gruppen I, Springer-Verlag, Berlin, 1967.
[HuBl] B. Huppert and N. Blackburn, Finite Groups III, Springer-Verlag, Berlin, 1985.
[Isl] I. M. Isaacs, Character Theory of Finite Groups, Academic Press, New York, 1976.
[McBl] P. P. McBride, Near solvable signalizer functors on finite groups, J. Algebra 7 8 (1982),
181-214.
[McB2] , Nonsolvable signalizer functors on finite groups, J. Algebra 78 (1982), 215-238.
[Odl] A. M. Odlyzko, The numerical verification of Thompson's identity, Invent. Math. 58
(1980), 97-98.
[PI] T. Peterfalvi, Le Theoreme de Bender-Suzuki (Part II only), Asterisque 1 4 2 - 1 4 3 (1986),
141-296.
[P2] , Sur la caracterisation des groupes Uz(q), pour q impair, J. Algebra 1 4 1 (1991),
253-264.
[Stl] R. Steinberg, Lectures on Chevalley Groups, Notes by J. Faulkner and R. Wilson, Mimeo-
graphed notes, Yale University Mathematics Department (1968).
[St2] , Endomorphisms of linear algebraic groups, Memoirs Amer. Math. Soc. 80 (1968).
[Stel] B. Stellmacher, Pushing up, Arch. Math. 46 (1986), 8-17.
[Sul] M. Suzuki, Group Theory I, II, Springer-Verlag, Berlin, 1982; 1986.
[P4] T. Peterfalvi, Character Theory for the Odd Order Theorem (Part I, Character Theory
for the Odd Order Theorem), London Math. Soc. Lecture Notes Series # 2 7 2 , Cambridge
University Press, Cambridge, 2000.
140
Expository References
[All] J. L. Alperin, Sylow intersections and fusion, J. Algebra 6 (1967), 222-241.

[ABG1] J. L. Alperin, R. Brauer, and D. Gorenstein, Finite groups with quasi-dihedral and
wreathed Sylow 2-subgroups, Trans. Amer. Math. Soc. 151 (1970), 1-261.
[ABG2] , Finite simple groups of 2-rank two, Scripta Math. 29 (1973), 191-214.
[Anl] S. Andrilli, On the uniqueness of O'Nan's sporadic simple group, Ph.D. Thesis, Rut-
gers University (1980).
[Arl] E. Artin, The orders of the classical simple groups, Comm. Pure and Appl. Math. 8
(1955), 455-472.
[Ar2] , Geometric Algebra, Wiley-Interscience, New York, 1957.
[Al] M. Aschbacher, Finite Group Theory, Cambridge University Press, Cambridge, 1986.
[A2] , Sporadic Groups, Cambridge University Press, Cambridge, 1994.
[A3] , Finite groups with a proper 2-generated core, Trans. Amer. Math. Soc. 197
(1974), 87-112.
[A4] , On finite groups of component type, Illinois J. Math. 19 (1975), 78-115.
[A5] , 2-components infinite groups, Comm. Algebra 3 (1975), 901-911.
[A6] , Tightly embedded subgroups of finite groups, J. Algebra 4 2 (1976), 85-101.
[A7] , On finite groups in which the generalized Fitting group of the centralizer of
some involution is symplectic but not extraspecial, Comm. Algebra 4 (1976), 595-616.
[A8] , On finite groups in which the generalized Fitting group of the centralizer of
some involution is extraspecial, Illinois J. Math. 2 1 (1977), 347-364.
[A9] , A characterization of Chevalley groups over fields of odd order. I, II, Ann.
of Math. 106 (1977), 353-468; Correction, Ann. of Math. I l l (1980), 411-414.
[A10] , Thin finite simple groups, J. Algebra 4 5 (1978), 50-152.
[All] , On the failure of Thompson factorization in 2-constrained groups, Proc.
London Math. Soc. (3) 4 3 (1981), 425-449.
[A 12] , A factorization theorem for 2-constrained groups, Proc. London Math. Soc.
(3) 4 3 (1981), 450-477.
[A 13] , Finite groups of rank 3. I, II, Invent. Math. 6 3 (1981), 357-402; Invent.
Math. 71 (1983), 51-163.
[A 14] , Weak closure in finite groups of even characteristic, J. Algebra 70 (1981),
561-627.
[A15] , GF(2) -representations of finite groups, Amer. J. Math. 104 (1982), 683-771.
[A16] , The uniqueness case for finite groups. I, II, Ann. of Math. 117 (1983), 3 8 3 -
551.
[A 17] , On the maximal subgroups of the finite classical groups, Invent. Math. 76
(1984), 469-514.
[A 18] , Quasithin groups post-Mason, unpublished typescript.
[AGL1] M. Aschbacher, D. Gorenstein, and R. Lyons, The embedding of 2-locals in finite
groups of characteristic 2 type. 1, II, Ann. of Math. 114 (1981), 335-456.
[ASel] M. Aschbacher and G. M. Seitz, On groups with a standard component of known
type. 1,11, Osaka J. Math. 1 3 (1976), 439-482; Osaka J. Math. 18 (1981), 703-723.
[Bal] B. Baumann, Uber endliche Gruppen mit einer zu Z/2(2 n) isomorphen Faktorgruppe,
Proc. Amer. Math. Soc. 74 (1979), 215-222.
[Bel] H. Bender, Uber den grossten p'-Normalteiler in p-auflosbaren Gruppen, Arch. Math.
18 (1967), 15-16.
[Be2] , On groups with Abelian Sylow 2-subgroups, Math. Zeit. 117 (1970), 164-176.
141
142 EXPOSITORY REFERENCES
[Be3] , Transitive Gruppen gerader Ordnung, in dene jede Involution genau einen
Punkt festldfit, J. Algebra 1 7 (1971), 527-554.
[Be4] , Goldschmidt's 2-signalizer functor theorem, Israel J. Math. 22 (1975), 208-
213.
[Be5] , Finite groups with dihedral Sylow 2-subgroups, J. Algebra 70 (1981), 216-228.
[BeGll] H. Bender and G. Glauberman, Local Analysis for the Odd Order Theorem, London
Math. Soc. Lecture Notes Series # 1 8 8 , Cambridge University Press, Cambridge, 1994.
[BeG12] , Characters of finite groups with dihedral Sylow 2-subgroups, J. Algebra 70
(1981), 200-215.
[Bol] E. Bombieri, Thompson's problem (a2 = 3), Appendices by A. M. Odlyzko and D.
C. Hunt, Invent. Math. 58 (1980), 77-100.
[Brl] R. Brauer, Some applications of the theory of blocks of characters of finite groups.
Ill, J. Algebra 3 (1966), 225-255.
[Br2] , Some applications of the theory of blocks of characters of finite groups. IV,
J. Algebra 17 (1971), 489-521.
[Br3] , Character theory of finite groups with wreathed Sylow 2-subgroups, J. Algebra
19 (1971), 547-592.
[BrSal] R. Brauer and C.-H. Sah (eds.), Finite Groups, A Symposium, Benjamin, New York,
1969.
[BrSul] R. Brauer and M. Suzuki, On finite groups of even order whose 2-Sylow subgroup is
a quaternion group, Proc. Nat. Acad. Sci. U.S.A. 4 5 (1959), 1757-1759.
[BuWil] N. Burgoyne and C. Williamson, On a theorem of Borel and Tits for finite Chevalley
groups, Arch. Math. 2 7 (1976), 489-491.
[CPSS1] P. J. Cameron, C. E. Praeger, J. Saxl, and G. M. Seitz, On the Sims conjecture and
distance transitive graphs, Bull. London Math. Soc. 15 (1983), 499-506.
[Cal] R. W. Carter, Simple Groups of Lie Type, Wiley and Sons, London, 1972.
[Ca2] , Finite Groups of Lie Type: Conjugacy Classes and Complex Characters,
Wiley and Sons, London, 1985.
[Chi] C. Chevalley, The algebraic theory of spinors, Columbia University Press, New York,
1951.
[Ch2] , Sur certains groupes simples, Tohoku Math. J. 7 (1955), 14-66.
[CI] M. J. Collins (ed.), Finite Simple Groups II, Academic Press, London, 1980.
[Col] J. H. Conway, A group of order 8,315,553,613,086,720,000, Bull. London Math. Soc.
1 (1969), 79-88.
[CCPNWl] J. H. Conway, R. T. Curtis, S. P. Norton, R. A. Parker, and R. A. Wilson, Atlas of
Finite Groups, Clarendon Press, Oxford, 1985.
[CoWal] J. H. Conway and D. B. Wales, Construction of the Rudvalis group of order 145,926,-
144,000, J. Algebra 2 7 (1973), 538-548.
[Cul] C. W. Curtis, Central extensions of groups of Lie type, J. Reine Angew. Math. 22 0
(1965), 174-185.
[Dal] K. M. Das, Homotopy and homology of p-subgroup complexes, Ph.D. Thesis, Califor-
nia Institute of Technology (1994).
[Dil] L. E. Dickson, Linear Groups, with an Exposition of the Galois Field Theory, Dover
Publ., New York, 1958.
[Dl] J. Dieudonne, La Geometrie des Groupes Classiques, Springer-Verlag, Berlin, 1955.
[El] M. Enguehard, Obstructions et p-groupes de classe 3. Caracterisation des groupes de
Ree, Asterisque 1 4 2 - 1 4 3 (1986), 3-139.
[Fl] W. Feit, The Representation Theory of Finite Groups, North Holland, Amsterdam,
1982.
[FT1] W. Feit and J. G. Thompson, Solvability of groups of odd order, Pacific J. Math. 13
(1963), 775-1029.
[FinFrl] L. Finkelstein and D. Frohardt, Simple groups with a standard 3-component of type
An(2), with n > 5, Proc. London Math. Soc. (3) 4 3 (1981), 385-424.
[FinSl] L. Finkelstein and R. Solomon, A presentation of the symplectic and orthogonal
groups, J. Algebra 60 (1979), 423-438.
[FinS2] , Finite simple groups with a standard 3-component of type 5p2n(2), n > 4,
J. Algebra 59 (1979), 466-480.
EXPOSITORY REFERENCES 143
[FinS3] , Finite groups with a standard 3-component isomorphic to Q±(2m, 2), m > 5,
F 4 (2) or En{2), n = 6, 7, 8, J. Algebra 7 3 (1981), 70-138.
[Fil] B. Fischer, Groups generated by 3-transpositions, Preprint, University of Warwick
(1969).
[Fi2] , Finite groups generated by 3-transpositions, Invent. Math. 13 (1971), 232-
246.
[FoSel] P. Fong and G. M. Seitz, Groups with a (B,N)-pair of rank 2. I, II, Invent. Math.
21 (1973), 1-57; Invent. Math. 24 (1974), 191-239.
[Fool] R. Foote, Aschbacher blocks, Proc. Symp. Pure Math. 3 7 (1980), 37-42.
[Foo2] , Component type theorems for finite groups in characteristic 2, Illinois J.
Math. 26 (1982), 62-111.
[Gil] R. H. Gilman, Components of finite groups, Comm. Alg. 4 (1976), 1133-1198.
[GiGrl] R. H. Gilman and R. L. Griess, Finite groups with standard components of Lie type
over fields of characteristic two, J. Algebra 80 (1983), 383-516.
[Gil] G. Glauberman, On solvable signalizer functors infinite groups, Proc. London Math.
Soc. (3) 33 (1976), 1-27.
[G12] , Central elements in core-free groups, J. Algebra 4 (1966), 403-420.
[G13] , On the automorphism group of a finite group having no nonidentity normal
subgroups of odd order, Math. Zeit. 9 3 (1966), 154-160.
[G14] , On groups with quaternion Sylow 2-subgroups, 111. J. Math. 18 (1974), 60-65.
[GINil] G. Glauberman and R. Niles, A pair of characteristic subgroups for pushing-up finite
groups, Proc. London Math. Soc. (3) 4 6 (1983), 411-453.
[Gol] , A conjugation family for finite groups, J. Algebra 16 (1970), 138-142.
[Go2] D. M. Goldschmidt, Solvable signalizer functors on finite groups, J. Algebra 21
(1972), 137-148.
[Go3] , 2-signalizer functors on finite groups, J. Algebra 2 1 (1972), 321-340.
[Go4] , Weakly embedded 2-local subgroups of finite groups, J. Algebra 2 1 (1972),
341-351.
[Go5] , 2-fusion in finite groups, Ann. of Math. 99 (1974), 70-117.
[Go6] , Automorphisms of trivalent graphs, Ann. of Math. I l l (1980), 377-406.
[Gl] D. Gorenstein, Finite Groups, Second Edition, Chelsea, New York, 1980.
[G2] , On the centralizers of involutions in finite groups, J. Algebra 11 (1969),
243-277.
[G3] , Finite Simple Groups: An Introduction to their Classification, Plenum Press,
New York, 1982.
[G4] , The Classification of Finite Simple Groups. I, Plenum Press, New York,
1983.
[GH1] D. Gorenstein and K. Harada, Finite groups whose 2-subgroups are generated by at
most 4 elements, Memoirs Amer. Math. Soc. 147 (1974).
[GL1] D. Gorenstein and R. Lyons, The local structure of finite groups of characteristic 2
type, Memoirs Amer. Math. Soc. 27 6 (1983).
[GL2] , On Aschbacher's local C(G,T)-theorem, Israel J. Math. 82 (1993), 227-279.
[GL3] , Nonsolvable signalizer functors on finite groups, Proc. London Math. Soc.
(3) 35 (1977), 1-33.
[GL4] , Nonsolvable signalizer functors revisited, J. Algebra 133 (1990), 446-466.
[GL5] , A local characterization of large sporadic groups, Preprint (1992).
[GLS1] D. Gorenstein, R. Lyons, and R. M. Solomon, On finite split BN-pairs of rank two,
J. Algebra 157 (1993), 224-270.
[GW1] D. Gorenstein and J. H. Walter, The characterization of finite groups with dihedral
Sylow 2-subgroups, J. Algebra 2 (1964), 85-151, 218-270, 354-393.
[GW2] , The -K-layer of a finite group, Illinois J. Math. 15 (1971), 555-565.
[GW3] , Centralizers of involutions in balanced groups, J. Algebra 20 (1972), 284-319.
[GW4] , Balance and generation in finite groups, J. Algebra 3 3 (1975), 224-287.
[Grl] R. L. Griess, Schur multipliers of finite simple groups of Lie type, Trans. Amer. Math.
Soc. 183 (1973), 355-421.
[Gr2] , Schur multipliers of finite simple groups of Lie type II, Proc. Symp. Pure
Math. 37 (1980), 279-282.
[Gr3] , The friendly giant, Invent. Math. 69 (1982), 1-102.
[GrMaSel] R. L. Griess, D. Mason, and G. M. Seitz, Bender groups as standard components,

Trans. Amer. Math. Soc. 235 (1978), 179-211.
[GroScl] F. Gross and W. R. Scott (eds.), Proceedings of the Park City Conference on Finite
Groups, Academic Press, New York, 1976.
[HaWal] M. Hall and D. B. Wales, The simple group of order 604,800, J. Algebra 9 (1968),
417-450.
[HaHil] P. Hall and G. Higman, The p-length of a p-soluble group and reduction theorems for
Burnside's problem, Proc. London Math. Soc. 7 (1956), 1-41.
[HI] K. Harada, On the simple group of order 2 1 4 3 6 5 6 7 • 11 • 19, 119-195 in [GroScl].
[H2] , Groups with nonconnected Sylow 2-subgroups revisited, J. Algebra 70 (1981),
339-349.
[Hel] D. Held, The simple groups related to M 2 4, J. Algebra 13 (1969), 253-296.
[He2] , The simple groups related to M 2 4, J. Austral. Math. Soc. 16 (1973), 24-28.
[HeKaSel] C. Hering, W. Kantor, and G. M. Seitz, Finite groups with a split BN-pair of rank
1, J. Algebra 20 (1972), 435-475.
[HiMcKl] G. Higman and J. McKay, On Janko's simple group of order 50,232,960, Bull. London
Math. Soc. 1 (1969), 89-94.
[HiSil] D. G. Higman and C. C. Sims, A simple group of order 44,352,000, Math. Zeit. 105
(1968), 110-113.
[Hoi] D. Holt, Transitive permutation groups in which a 2-central involution fixes a unique
point, Proc. London Math. Soc. (3) 3 7 (1978), 165-192.
[Hunl] D. C. Hunt, A check of the Ree group conjecture for small fields, Appendix to [Bol],
Invent. Math. 58 (1980), 99.
[Hul] B. Huppert, Endliche Gruppen I, Springer-Verlag, Berlin, 1967.
[HuBl] B. Huppert and N. Blackburn, Finite Groups III, Springer-Verlag, Berlin, 1985.
[Isl] I. M. Isaacs, Character Theory of Finite Groups, Academic Press, New York, 1976.
[Iwl] N. Iwahori, ed., Finite Groups, Japan Society for the Promotion of Science, Tokyo,
1976.
[Jl] Z. Janko, A new finite simple group with abelian 2-Sylow subgroups and its charac-
terization, J. Algebra 3 (1966), 147-186.
[J2] , Some new simple groups of finite order. I, 1st. Naz. Alta Math. Symposia
Math. 1 (1968), 25-65.
[J3] , A new finite simple group of order 86,775,571,046,077,562,880 which pos-
sesses M24 and the full cover of M22 as subgroups, J. Algebra 4 2 (1976), 564-596.
[JT1] Z. Janko and J. G. Thompson, On a class of finite simple groups of Ree, J. Algebra
4 (1966), 274-292.
[KLST1] W. Kimmerle, R. Lyons, R. Sandling, and D. N. Teague, Composition factors from
the group ring and Artin's theorem on orders of simple groups, Proc. London Math.
Soc. (3) 60 (1990), 89-122.
[KlLil] P. Kleidman and M. Liebeck, The Subgroup Structure of the Classical Groups, Cam-
bridge University Press, Cambridge, 1990.
[KMal] K. Klinger and G. Mason, Centralizers of p-subgroups in groups of characteristic 2,
p type, J. Algebra 3 7 (1975), 362-375.
[LeSil] J. Leon and C. C. Sims, The existence and uniqueness of a siple group generated by
^^-transpositions, Bull. Amer. Math. Soc. 8 3 (1977), 1039-1040.
[LI] R. Lyons, A characterization ofU3(4), Trans. Amer. Math. Soc. 164 (1972), 371-387.
[L2] , Evidence for a new finite simple group, J. Algebra 20 (1972), 540-569; Er-
rata, J. Algebra 34 (1975), 188-189.
[L3] , Generation of finite Chevalley groups of odd characteristic, Comm. Algebra
18 (1990), 1433-1444.
[MacWl] A. MacWilliams, On 2-groups with no normal abelian subgroups of rank 3 and their
occurrence as Sylow 2-subgroups of finite simple groups, Trans. Amer. Math. Soc.
150 (1970), 345-408.
[Mai] G. Mason, Quasithin finite groups, unpublished typescript.
[Ml] E. Mathieu, Memoire sur le nombre de valeurs que peut acquerir une fonction quand
on y permut ses variables de toutes les manieres possibles, J. Math. Pures et Ap-
pliquees 5 (1860), 9-42.
[M2] , Memoire sur Vetude des fonctions de plusieurs quantites, sur la maniere
des formes et sur les substitutions qui les laissent invariables, J. Math. Pures et
Appliquees 6 (1861), 241-323.
[M3] , Sur la fonction cinq fois transitive des 24 quantites, J. Math. Pures et Ap-
pliquees 18 (1873), 25-46.
[McBl] P. P. McBride, Near solvable signalizer functors on finite groups, J. Algebra 78
(1982), 181-214.
[McB2] , Nonsolvable signalizer functors on finite groups, J. Algebra 78 (1982), 215-
238.
[McLl] J. McLaughlin, A simple group of order 898,128,000, 109-111 in [BrSal].
[MeStl] U. Meierfrankenfeld and G. Stroth, On quadratic GF{2)-modules for Chevalley groups
over fields of odd order, Arch. Math. 5 5 (1990), 105-110.
[Nil] R. Niles, Pushing-up in finite groups, J. Algebra 5 7 (1979), 26-63.
[Nol] S. Norton, The construction of J 4 , Proc. Symp. Pure Math. 37 (1980), 271-278.
[Odl] A. M. Odlyzko, The numerical verification of Thompson's identity, Appendix to
[Bol], Invent. Math. 58 (1980), 97-98.
[ONI] M. O'Nan, Automorphisms of unitary block designs, J. Algebra 20 (1972), 495-511.
[ON2] , A characterization ofUz{q), J. Algebra 22 (1972), 254-296.
[ON3] , Some evidence for the existence of a new finite simple group, Proc. London
Math. Soc. (3) 32 (1976), 421-479.
[PI] T. Peterfalvi, Le Theoreme de Bender-Suzuki. I, II, Asterisque 1 4 2 - 1 4 3 (1986),
141-296.
[P2] , Sur la caracterisation des groupes U$(q), pour q impair, J. Algebra 141
(1991), 253-264.
[P3] , Simplification du chapitre VI de Varticle de Feit et Thompson sur les groupes
d'ordre impair, C. R. Acad. Sci. Paris 299 (1984), 531-534.
[Phi] K.-W. Phan, On groups generated by three-dimensional special unitary groups I, J.
Austral. Math. Soc. A 2 3 (1977), 67-77.
[Ph2] , On groups generated by three-dimensional special unitary groups II, J. Aus-
tral. Math. Soc. A 23 (1977), 129-146.
[PoThl] M. Powell and G. Thwaites, On the nonexistence of certain types of subgroups in
simple groups, Quart. J. Math. Oxford Series (2) 26 (1975), 243-256.
[Rl] R. Ree, A family of simple groups associated with the simple Lie algebra of type (F4),
Amer. J. Math. 8 3 (1961), 401-420.
[R2] , A family of simple groups associated with the simple Lie algebra of type (G2),
Amer. J. Math. 8 3 (1961), 432-462.
[Ril] F. Richen, Modular representations of split BN pairs, Trans. Amer. Math. Soc. 140
(1969), 435-460.
[Rul] A. Rudvalis, A new simple group of order 2 1 4 • 3 3 • 5 3 • 7 • 13 • 29, Notices Amer. Math.
Soc. 20 (1973), A-95.
[Schl] I. Schur, Uber die Darstellungen der symmetrischen und alternierenden Gruppen
durch gebrochene lineare Substitutionen, J. Reine Angew. Math. 139 (1911), 155-
250.
[Scl] L. L. Scott, Representations in characteristic p, Proc. Symp. Pure Math. 3 7 (1980),
319-331.
[Sel] G. M. Seitz, Chevalley groups as standard subgroups. I, II, III, Illinois J. Math. 2 3
(1979), 36-57, 516-553, 554-578.
[Se2] , Standard subgroups in finite groups, 41-62 in [CI].
[Se3] , Generation of finite groups of Lie type, Trans. Amer. Math. Soc. 271 (1982),
351-407.
[Shi] E. Shult, On a class of doubly transitive groups, Illinois J. Math. 16 (1972), 434-455.
[SGH1] E. Shult, T. Gagen and M. Hale (eds.), Finite Groups '72, North Holland, Amster-
dam, 1973.
[Sil] C. C. Sims, The existence and uniqueness of Lyons' group, 138-141 in [SGH1].
[Smil] F. Smith, On finite groups with large extraspecial 2-subgroups, J. Algebra 4 4 (1977),
477-487.
[Sml] S. D. Smith, Large extraspecial subgroups of widths 4 a r ^ 6, J. Algebra 58 (1979),
251-280.
[Sm2] , A characterization of finite Chevalley and twisted groups of type E over

GF(2), J. Algebra 62 (1980), 101-117.
[Sm3] , The classification of finite groups with large extraspecial subgroups, Proc.
Symp. Pure Math. 3 7 (1980), 111-120.
[SI] R. M. Solomon, Maximal 2-components in finite groups, Comm. Algebra 4 (1976),
561-594.
[Stl] R. Steinberg, Lectures on Chevalley Groups, Notes by J. Faulkner and R. Wilson,
Mimeographed notes, Yale University Mathematics Department (1968).
[St2] , Endomorphisms of linear algebraic groups, Memoirs Amer. Math. Soc. 80
(1968).
[St3] , Variations on a theme of Chevalley, Pacific J. Math. 9 (1959), 875-891.
[St4] , Generateurs, relations, et revetements de groupes algebriques, Colloque sur
la theorie des groupes algebriques, Bruxelles (1962), 113-127.
[St5] , Generators, relations, and coverings of algebraic groups II, J. Algebra 71
(1981), 527-543.
[Stel] B. Stellmacher, Pushing up, Arch. Math. 4 6 (1986), 8-17.
[Ste2] , Finite groups all of whose maximal parabolic subgroups are thin and of char-
acteristic 2-type, unpublished typescript.
[Sul] M. Suzuki, Group Theory I, II, Springer-Verlag, Berlin, 1982; 1986.
[Su2] , A new type of simple groups of finite order, Proc. Nat. Acad. Sci. U.S.A. 4 6
(1960), 868-870.
[Su3] , On a class of doubly transitive groups, Ann. of Math. 7 5 (1962), 105-145.
[Su4] , On a class of doubly transitive groups. II, Ann. of Math. 79 (1964), 514-589.
[Su5] , A simple group of order 448,345,497,600, 113-119 in [BrSal].
[Su6] , Finite groups with a split (B,N)-pair of rank one, Proc. Symp. Pure Math.
3 7 (1980), 139-147.
[Tl] J. G. Thompson, Toward a characterization of E£(q). I, II, III, J. Algebra 7 (1967),
406-414; J. Algebra 20 (1972), 610-621; J. Algebra 4 9 (1977), 163-166.
[T2] , Nonsolvable finite groups all of whose local subgroups are solvable. I, II, III,
IV, V, VI, Bull. Amer. Math. Soc. 74 (1968), 383-437; Pacific J. Math. 3 3 (1970),
451-536; Pacific J. Math. 3 9 (1971), 483-534; Pacific J. Math. 4 8 (1973), 511-592;
Pacific J. Math. 5 0 (1974), 215-297; Pacific J. Math. 51 (1974), 573-630.
[T3] , A simple subgroup of E8(S), 113-116 in [Iwl].
[Timl] F . G. Timmesfeld, A characterization of Chevalley and Steinberg groups over F2,
Geom. Ded. 1 (1973), 269-321.
[Tim2] , Groups generated by root involutions. I, II, J. Algebra 3 3 (1975), 75-134; J.
Algebra 3 5 (1975), 367-441.
[Tim3] , Finite simple groups in which the generalized Fitting group of the centralizer
of some involution is extraspecial, Ann. of Math. 107 (1978), 297-369.
[Til] J. Tits, Groupes simples et geometries associees, Proc. Int. Congr. Math. Stockholm
(1962), 197-221.
[Ti2] , Groupes semi-simples isotropes, Colloque sur la theorie des groupes alge-
briques, Bruxelles (1962), 137-146.
[Ti3] , Algebraic and abstract simple groups, Ann. of Math. 80 (1964), 313-329.
[Ti4] , Buildings of spherical type and finite BN-pairs, Lecture Notes in Mathematics
# 3 8 6, Springer-Verlag, Berlin, 1974.
[Ti5] , Classification of buildings of spherical type and Moufang polygons: A survey,
Att. Conv. Lincei 1 7 (1976), 229-246.
[Wal] J. H. Walter, The characterization of finite groups with abelian Sylow 2-subgroups,
Ann. of Math. 8 9 (1969), 405-514.
[Wa2] , The B-conjecture: 2-components in finite simple groups, Proc. Symp. Pure
Math. 3 7 (1980), 57-66.
[Wa3] , The B-conjecture; characterization of Chevalley groups, Memoirs Amer.
Math. Soc. 3 4 5 (1986).
[Warl] H. Ward, On Ree's series of simple groups, Trans. Amer. Math. Soc. 121 (1966),
62-89.
[Wol] W. Wong, Generators and relations for classical groups, J. Algebra 32 (1974), 529 -
553.
[Yl] T. Yoshida, Character-theoretic transfer, J. Algebra 52 (1978), 1-38.

Added in Second Printing:
[P4] T. Peterfalvi, Character Theory for the Odd Order Theorem (Part I: Character The-
ory for the Odd Order Theorem), London Math. Soc. Lecture Notes Series #272 ,
Cambridge University Press, Cambridge, 2000.
[P5] T. Peterfalvi, Character Theory for the Odd Order Theorem (Part II: A Theorem of
Suzuki), London Math. Soc. Lecture Notes Series # 2 7 2 , Cambridge University Press,
Cambridge, 2000.
GLOSSARY
PAGE SYMBOL
90 rvj
97 y
4-5 [Ii], [I2], [IA], M, [HA], [HG], [Up], [Us]
60, 101 Âni [2)An
101 2Dn(q)
101 3C/4(3)
101 4L 3 (4) _
101 [Ai x A2]K (central extension)
101 AK (central extension)
81 Alt
6, 8 An
7,8 An{q), 2An{q), A+(q), A~{q)
130 A{T)
60, 104 £ 2 , "B2-group
7, 8, 10 Bn{q), 2B2{2n)
116 n{G)
24 BP,{X)
7,8 Cn{q)
9, 11 Coi, C02, Cos
100 C p , Sp-group
102
100 S P ', Cp'-group
81 Qhev, ehev(p)
90 C{G,S)
101 Q'Kp
22 C{K, x)
63 ^X
22 CX{A/B)
14 CX{V)
139 CY
64, 124 &x{B)
7, 8, 10 Dn{q), 2Dn{q), 3 D 4 (q), D+{q), D~(q)
7, 8, 10 En{q),2E6{q),E+(q),Ee{q)
14 EpTt
124 £*(X), Zpk(X)
139 £(X), £ fc (X)
17 E(X)
148
GLOSSARY 149
18 *(*)
9, 11 Fl, i<2, F3, F5
7, 8, 10 FM, 2FA(2n), 2 F 4 (2)'
9, 11 Fi22, Fi23, Fi24
95 9^9
6 Fq
16 F(X)
17 F*(X)
66 7(G)
132 T, T(a)
82 TJ-aW
82 Tp,fc(X)
27 G*
109-121 G^G*
66, 121 Go(X)
7, 8, 10 G2(q),*G2(3n)
71 Ga> G*a
6 GLn(q)
102 %
32-33 G(q), G{q)
7 GUn(q)
9, 11 He
101
9, 11 HS
55, 81 Zp(G)
55, 103 I°p(G), J£(G)
13 Inn(X)
14 Int, Int(x)
139 X(X)
102 3O(P), 3I(P), 32(P)
9, 11 <A> J2, J3, JA
26, 130 J(P)
90 \K]
53, 81 %
110 %(2)*
86 3CW, i = 0,...,7
53, 81 Jvp
12 DC-group
12 DC-proper
6, 8 Ln(q) = L+(q)
8 Ln(Q)
55, 103 Z°P(G)
53, 81 LP{G)
126 HP(G;A)
20 LP,(X)
21 LP,{X)
127 L;,{X)
7 £(?)
150 GLOSSARY
139 JL(X)
139 L(X)
9, 11 Ly
135 m2,p(G)
9, 11 M n , M12, M 2 2 , M 2 3 , M 2 4
9, 11 Mc
123 M ~> N
30, 81 rrip(A)
58, 60, 82 M(5),Mi(5)
139 m(X)
65 3sr
97 N(M), N*(M)
19 Nx(Q)
139 NY
26 «i(B)
117 ^8(3)-type
9, 11 O'N
24 Ov,p{X)
19 0*{X),0*\X\O"(X)
29 Out(X)
19 0(X),Op(X),Op,(X),On(X)
19 7T, 7 r '
32 n
7, 8, 101 Pfi±(«), Ptln(q), (P)Cln(q)
6, 8 PSLn(q)
7, 8 PSpn(q)
7 ,8 PSUn(q)
98 Q(Afp)
9,11 Ru
37, 82 <r{G)
107 a*(G)
58, 83 <ro{G)
120 <r9(G)
118 <T7(G)
32 E, E+, E~
6 ^-*n
94 §P(X), §r{X)
29 5(G)
6 S£«(«)
101 (S)Ln(q)
81 Spor
7 SUn(q)
9, 11 Suz
8, 10 Sz(2n)
30, 124-125 6 , 0 ( C G ( a ) ), 6 ( G ; A ) , 9 f e + i , e f e + i ( G ; A )
102
8 ^n(«)
139 X, X, X*
GLOSSARY 151
101 [X]L3(4)
26 Z(J(P))
6, 8 Zp
88 z,zx
115 2-amalgam G*-type
114-115 2-amalgam type, 2-amalgam type (Mi, M<i\ S)
111 2-central G*-type
88 2-central involution
135 2-local p-rank
111 2-maximal G*-type mod cores
98 {2, p}-uniqueness subgroup
110 2-terminal G*-type
114 2-terminal -CJVtype
82 2-uniqueness subgroup
64 3/2-balance
120 3/2-balanced type
13 ^-composition factor, length, series
118, 121 algebraic automorphism
93 almost p-constrained p-component preuniqueness subgroup
18 almost simple group
94 almost strongly p-embedded subgroup
131-132 amalgam method: Aa, Qa, Sa,Xa, Za
125 associated (k + |)-balanced functor
96 associated module of a near component
44-50 Background Results, Background References
119 base of a neighborhood
34 (£,iV)-pair, split (£,iV)-pair
34 Borel subgroup
24 £?p-property
33, 34 Bruhat decomposition: £?, iJ, iV, R, U, V, Xa, W, ha{t), na(t)
33 Cartan subgroup
46, 109 centralizer of involution pattern
90 characteristic 2-core
136 characteristic power
16 characteristic subgroup
25 characteristic p-type
32-33 Chevalley commutator formula
7 Chevalley group
13 chief factor, series
6 classical group
17, 81 component
13 composition factor, length, series
87 control of (strong) G-fusion (in T)
91 control of rank 1 (or rank 2) fusion
16 covering group
112 doubly transitive (7*-type
95-96 doubly transitive of Suzuki type
55, 81 even type
122 extremal conjugation
152 GLOSSAR Y
26 failure of factorization module

19 Fitting length, series
16 Fitting subgroup
39 four-group
18 Frattini subgroup
17 generalized Fitting subgroup
58, 106 generic, generic type
106 generic even type
106 generic odd type
25 Hall cr-subgroup
46, 109 involution fusion pattern
124 /c-balance, weak /c-balance
125 /c-balanced signalizer functor, weakly /c-balanced signalizer functor
125 (fc+ |)-balance
17 layer
104 ££ 2 -type
105 £C p -type
94 £C p -uniqueness subgroup
32 Lie rank (twisted or untwisted)
128 local balance
126 local (k + ^)-balance
126 local /c-balance, weak local /c-balance
19 local subgroup
21 L p /-balance
127 L*/-balance
57-58 •Cp-generic type
58, 103 £ p -special type, ^-special type
104 £T 2 -type
105 £T p -type
34 monomial subgroup
96, 98 near component (of Y): linear, alternating, standard
97 near component 2-local uniqueness subgroup
64, 119 neighbor, (y, 7)-neighbor
65, 119, 120 neighborhood: level, split, vertical
8, 9 notation for simple groups
58-59, 81 odd type
101 p-central element of order p
20 p-component
91 p-component preuniqueness hypothesis
90 p-component preuniqueness subgroup
20 p-constrained group
19 p'-core
20 p-layer
19 p-local subgroup
107 {p, g}-parabolic type
89 proper 2-generated core
121 proper semisimple type
24 p-solvable group
64 p-source
GLOSSARY 153
23, 108 p-terminal p-component

63 p-terminal Sp-pair
118 p-terminal £T p -type
22, 127 pumpup: vertical, trivial, diagonal, proper
82 p-uniqueness subgroup
26 quadratic chief factor
25 quadratic i^X-module
16, 81 quasisimple group
117 quasisymplectic type
60, 82 quasithin, quasithin type, quasithin case
34 reduced monomial subgroup
10 Ree group
136 regular
95 restricted even type
32 root subgroup, root system
25 cr-subgroup
21 Schreier property
18 Schur multiplier
12 section
16 semisimple group
64 semisimple neighbor
121 semisimple type
124 signalizer functor: A-signalizer functor, solvable, closed, complete
13 simple group
109 solvable component, solvable p-component
13 solvable group
66, 121 span of 3Sf
58 special, special type
105 special even type
103 special odd type
53 standard component
91 standard preuniqueness subgroup
32-33 Steinberg presentation, relations
10 Steinberg variation
43, 89 strongly closed subgroup
93 strongly closed type p-component preuniqueness subgroup
30-31, 82 strongly embedded subgroup
52, 82 strongly p-embedded subgroup
88 strongly Z-embedded subgroup
52, 94 strong p-uniqueness subgroup
64 subterminal (x, if)-pair
10 Suzuki group
117 symplectic pair: faithful, trivial
116 symplectic type
23, 81 terminal component
26, 130 Thompson subgroup
34 Tits system
7 twisted group, untwisted group
17 universal covering group
154 GLOSSARY
33 universal version
89 weakly ^-embedded subgroup
33 Weyl group
116 wide £C p -type
93 wreathed p-component preuniqueness subgroup
71 Y-compatible
65 (y, 7)-neighborhood
117 Z6 x ^-neighborhood
INDEX
2-central involution 88
2-local p-rank 135
3/2-balanced functor 43, 64-65
algebraic automorphism 118, 121

almost strongly p-embedded subgroup, see uniqueness subgroups
Alperin, J. 39, 41
alternating group An 6, 32, 36
as C2, T2 , or S 2 -group 103
amalgam method 5-6, 26, 39, 41, 43, 60-61, 105, 131-133
associated graph 132-133
Artin, E. 11
Aschbacher, M. 18, 30, 37, 39-43, 45-48, 50, 53, 89, 99, 125, 129, 130
Aschbacher x-block 39, 41, 53
associated (k + ^-balanced functor 125
Atlas of Finite Groups 45, 50, 139
Background References 47-50, 140

Background Results 59, 63, 79, 87, 104, 118
listed, 44-50
balance, /c-balance, see group; also see signalizer functor
bar convention 18, 139
Baumann, B. 39, 131
Bender, H. 16-17, 48-50, 123
Bender method 30, 38, 43, 60, 62, 104, 110, 123, 134
Blackburn, N. 46, 47
(£,iV)-pair 34
split 34
recognition of rank 1 36-37, 39, 49, 50, 63, 113, 138
recognition of rank 2 37, 63, 111, 113, 115, 137-138
Bombieri, E. 49
Borel, A. 25
£ p -property 24, 28, 40-41, 62, 123, 127
partial 30, 36, 42, 64
Brauer, R. 51
theory of blocks 38, 46, 50, 62
defect groups of 2-rank at most 3 50
Brauer-Suzuki theory of exceptional characters 38, 135
Burnside, W. 29-30
building 34, 73, 138
155
156 INDEX
Carter, R. 45, 47
centralizer
of element of odd prime order p 35, 41, 42, 51, 54-56
of element of prime order p 108
of involution 11, 27ff., 35, 39, 41-43, 51, 52, 54, 61-62
of semisimple element 51, 54-56
centralizer of involution pattern 46, 59, 77, 109-110
character theory 31, 46, 50, 60, 62, 104, 108, 135-137
ordinary vs. modular 50
Chevalley, C. 3
Chevalley groups, see groups of Lie type
chief factor, series 13-14
classical groups 6ff., see also groups of Lie type
Classification Grid 79, 83, 85, 99-121
Classification Theorem, see Theorems
Theorems Ci-Cy 104-106
component (see also p-component) 17, 51, 81
standard 53, 91-92
terminal 23, 42, 53, 81, 90-92, 108
solvable 51, 67, 109
composition factor, length, series 12
A-composition factor, length, series 13
computer 35, 45, 68
control
of2-locals 129
of fusion 87, 122
of rank 1 or 2 fusion 91-92
Conway, J. 11
core 20; see also p'-core
elimination 40, 43, 60, 110-111
covering group 16
universal 17, 33
notation for 101
e p -groups 54, 57, 81, 99-101
as pumpups 101-102
Cp'-groups 95, 100
Curtis, C. 35
Das, K. M. 35, 71
Delgado, A. 37
Dickson, L. 7
Dieudonne, J. 47
double transitivity of Suzuki type 95-96
Enguehard, M. 49-50
Expository References 47, 141-146
Feit, W. 46, 47, 48, 107-108
Finkelstein, L. 35
Fischer, B. 11, 39-40
transpositions 11, 39
INDEX
Fitting length, series 19

Fitting subgroup 16
generalized Fitting subgroup 17, 123
Foote, R. 5, 38, 53, 98
Frattini subgroup 18
Frobenius, G. 29
Frobenius group 107
Frohardt, D. 35
fusion 29, 60, 62, 63, 104, 120, 122
extremal conjugation 122
G « G* 59, 62-63, 67-68, 76-77, 83

G*=An
n < 12 109, 114
n > 13 121
G* of Lie type
of large Lie rank 116, 121
of small Lie rank
of characteristic 2 95-96, 115
of characteristic 2 and Lie rank 1 95-96
of odd characteristic 110, 112, 113
G* sporadic 109
general local group theory 45-48
generalized Fitting subgroup 17, 123
geometry associated with a finite group 35, 73-74
Gilman, R. 35, 39, 41
Glauberman, G. 21, 38, 39, 48-50, 124, 130
Goldschmidt, D. 39, 43, 49, 125
Gomi, K. 37
Gorenstein, D. 29, 38, 39, 41, 46, 47-50, 99, 124, 126, 127
Sp-groups 57-58, 63, 103
Griess, R. L. 17, 35, 41, 45
group
almost simple 18
alternating, see alternating group
covering 16
universal 17, 33
notation for 101
A;-balanced 124-125, 129
(k+ |)-balanced 125, 129
locally balanced 128
locally /c-balanced, k -f ^-balanced 126
weakly /c-balanced, weakly locally /c-balanced 124, 126
X-proper 12, 21
nilpotent 15-16
of Lie type, see groups of Lie type
p-constrained 20
perfect 16
p-solvable 24
158 INDEX
quasisimple 16, 53, 81

semisimple 16
simple, table of 8-10
solvable 13, 16, 73-74
sporadic, see sporadic group
group order formulas 50, 135-137
groups in Cp, see Cp-groups
groups in Cp/, see Cp/-groups
groups in Sp, see S p-groups
groups in T p , see T p-groups
groups of 3/2-balanced type 120-121
groups of characteristic 2-type 55, 98, 116-117
groups of characteristic p-type 25, 116-117
groups of even type 36, 53-55, 57-59, 81, 87, 98
groups of restricted even type 58-59, 82, 90, 95, 99, 102, 106
groups of G*-type' (G* a target group)
groups of doubly transitive G*-type, G* = L3(pn), U3(pn), 2 G 2 (3 n ) 112
groups of 2-amalgam G*-type, G* € 0C(4) 115
groups of 2-central G*-type, G* e 0C^ of Lie type 111
groups of 2-maximal G*-type mod cores, G* £ X^* 111
groups of 2-terminal G*-type, G* G 3C(2)* 110
groups of generic type 35-36, 55-59, 63-68, 79, 106, 118-121
groups of generic even type 106
groups of generic odd type 106
groups of £ p -generic type 57-59, 61
groups of semisimple type 121
groups of proper semisimple type 121
groups of GF(2)-type 40-43
groups of large sporadic type 37-38, 61
groups of Lie type 6ff., 32ff., 45-49
as (S, iV)-pair 34
Borel subgroup of 33
Bruhat decomposition of 34
Cartan subgroup of 33
Chevalley group 3
Chevalley commutator formula 32
Dynkin diagram of 32, 71
generation of 48, 65-66
Lie rank of
twisted 32
untwisted 32
monomial subgroup of 34
reduced 34
parabolic subgroup of 26, 62
rank 1 subgroup of 33, 35, 71
Ree group 3, 10, 49, 50
root subgroup of 32
root system 32
fundamental system 32
INDEX 159
Schur multiplier of 45, 48

semisimple element of 51
Steinberg variation 7
Steinberg presentation of, relations 33
Suzuki group 3, 10
universal version of 33
and universal covering group 33
untwisted, see Chevalley group
Weyl group of 33, 35
groups of odd order
groups of odd order uniqueness type 107
groups of {p, g}-parabolic type 107-108
groups of cr*(G)-uniqueness type 107
groups of odd type 58-59, 81, 86
groups of quasithin type 82, 105, 114-116
groups of 2-amalgam type 114-115
thin subcase 37, 41
groups of special type 36-38, 58-61, 79, 103-106, 106-108, 110-118
groups of ^2-special type 58-60, 103-104
groups of £3 2 -typ e 61-63, 83, 104, 110-113, 123
51/2(<z)-subcase 61-63
groups of £T2-type 104,114
groups of 2-terminal £T2-type 114
groups of £ p -special type 58-60, 105-106
groups of £C p -type 105,116-118
groups of quasisymplectic £C p -type 117
groups of wide £C p -type 116, 133-135
groups of-CTp-type 106,118
groups of p-terminal £T P -type 118
groups of odd order uniqueness type 107
groups of {p, #}-parabolic type 107-108
groups of cr*(G?)-uniqueness type 107
groups of special even type 105, 106, 114-118
groups of special odd type 103-104, 106-108, 110-114
groups with specified 2-structure
groups of 2-rank at most 2 36, 135
groups of 2-rank at most 3 50
groups with semidihedral or wreathed Sylow 2-subgroups 76, 136-137
characteristic power 136-137
Brauer group order formula for regular groups 137
Hall, M. 11
Hall, P. 16, 25
Hall cr-subgroup
Harada, K. 11, 39, 89
Hayashi, M. 37
Holt, D. 89
Hunt, D. 49
Huppert, B. 46, 47
160 INDEX
identification of simple groups see recognition of simple groups

involution fusion pattern 46, 60, 62, 109, 110
Isaacs, I. M. 46, 47
isomorphism question 11
Janko, Z. 11, 49
JC-groups 5, 12
theory of 12, 45, 48, 75-76, 138-139
3C-proper group 12, 75-76, 79
Klinger, K. 38, 116-117
Klinger-Mason method 61, 105, 116-118
layer, see also p-layer 17
Leech lattice 11, 35
linear group 18
local subgroup, see also p-local subgroup 19
general structure of 27-28
Imbalance 21, 127-128
analogue for near components 96
analogue for two primes 134
Imbalance 127-128
Lyons, R. 41, 48-50, 99
Mason, G. 37-38, 41, 116-117
Mathieu, E. 11
maximal subgroup 60, 62-63, 123
McBride, P. 30, 49, 124
Meierfrankenfeld, U. 130
modules
failure of factorization 26, 130, 133
quadratic 25-26, 123, 130
near component 50, 96-97, 129-131, 133
alternating 96-97, 130
associated module of 96
linear 96-97, 130, 131
standard 98
typeG 2 (3") 97
neighbor 42, 64-65, 71, 119
semisimple 64-65
neighborhood 55-58, 59, 63-68, 118-120
base of 119
example of 56-57, 68-70
level 66, 121
span of 66-67, 121
vertical 65, 76-77, 118-120
ZQ x Z2- 61, 117
TV-group 38-39, 73
Niles, R. 39, 50
notation 139
INDEX
Odlyzko, A. 49
O'Nan, M. 18, 45, 49, 138
outer automorphisms 18
overall strategy of proof 35-38, 42-43
Parts of the series 4-5, 59, 77-78, 80
Part II 38, 52-53
Part III 36
Part IV 37
Part V 37, 38
p-central p-element 101
p-component 20
p-terminal 22ff., 63, 108-109
pumping up to 23, 108-109
solvable 64, 109
p-component preuniqueness hypothesis 91-92
p-component preuniqueness subgroup, see uniqueness subgroups
p-component uniqueness theorems 30-31, 38, 53, 65, 90-92, 118
p'-core 19; see also core
elimination 23-24, 37, 61, 120
embedding of p'-core of p-local subgroups 21, 127-128
p-layer 20
p-local subgroup 19
embedding of p'-core of 20-21, 127-128
embedding of p-layer of 21-24, 127
perfect central extension, see covering group
permutation group 18, 35
doubly transitive 74, 112-113, 138
of Suzuki type 95-96
split (£,iV)-pair of rank 1 95-96, 112-113, 138
highly transitive 11, 35
of rank 3 11
representation as 31
Peterfalvi, T. 48-49, 96, 107, 138
Phan, K.-W. 35, 71
presentation 31-35
of classical groups 35
of symmetric groups 32, 36, 68-70
Steinberg presentation of groups of Lie type 32, 33, 36, 51, 118
a la Curtis-Tits 34, 67, 70
a la Gilman-Griess 35, 67, 71
p-source 64
pumpup 22, 127
as Cp, T p , or Sp-group 103, 114
diagonal, proper, trivial, or vertical 22, 127
quasisimple group 16, 53, 81
quasithin 5, 37, 41, 43, 60-61, 82, 105, 114-116
recognition of simple groups 31-35
groups of Lie type 32-35, 42, 49, 137-138
162 INDEX
alternating groups 32
sporadic groups 35
recognition of symmetric groups 68-70
Ree, R. 10
Ree groups 3, 10, 49, 50
Rowley, P. 37
Schreier, O. 21
Schreier property 21, 24, 29
Schur, I. 17
Schur multiplier 18, 45, 48, 139
Scott, L. 18
section 12
Seitz, G. 48
a(G) 37-38, 52-53, 55, 58-59, 61, 82, 86, 92-93, 105, 131
a0(G) 58-61, 83, 86, 105-106, 116, 118, 131
Sibley, D. 48
signalizer functor 29-30, 124ff.
closure 30, 124
closed 124
complete 124
^-balanced, weakly ^-balanced 124-126
solvable 124
trivial 124, 128
signalizer functor method 29-30, 36-38, 41, 60, 64-65, 104-105, 116, 118, 120,
128-129
simple groups, table of 8-10
simplicity criteria 29-31
Sims, C. 39, 45
conjecture 74
Smith, F. 41
Smith, S. 41
Solomon, R. 35, 39, 53
sporadic groups 3, 6, 9, 11, 87, 100
as target group 77, 87
as Sp-group 103
background properties of 44-48, 50
existence and uniqueness of 46-47, 50, 72
individual groups:
Baby Monster F2 = BM 11, 38, 87, 100
Conway groups Cox = -1, Co2 = -2, Co3 = -3 11, 38, 42, 72, 87, 100
Fischer 3-transposition groups Fi22, Fi23, Fi'2A 11, 38, 87, 100
Fischer-Griess Monster Fx = M 3, 11, 38, 42, 47, 61, 72, 87, 100
Harada-Norton F 5 = HN 11, 67-68, 87, 100
Held He = HUM 11, 87, 100
Higman-Sims HS 11, 87, 100
Janko J i , J 2 = HJ, J 3 = HJM, J 4 3, 11, 87, 100
Lyons-Sims Ly 11, 45, 87, 100
Mathieu M n , M12, M 2 2 , M 2 3 , M 2 4 11, 42, 46, 72, 87, 100
INDEX 163
McLaughlin Mc 11, 87, 100

O'Nan-Sims O'N 11, 87, 100
Rudvalis Ru 11, 87, 100
Suzuki Suz 11, 87, 100
Thompson F 3 = Th 11, 87, 100
standard form problems 38
Steinberg, R. 3, 7, 10, 17, 45, 47, 48
Stellmacher, B. 37, 50
strongly closed subgroup 61
strongly embedded subgroup, see uniqueness subgroups
strongly p-embedded subgroup, see uniqueness subgroups
strongly ^-embedded subgroup, see uniqueness subgroups
Stroth, G. 5, 37, 38, 43, 53, 98-99, 107, 130
subterminal pair 64-66
Suzuki, M. 10, 38, 46, 47, 49
Suzuki groups Sz(2n) 3, 10
symmetric group E n 6
symplectic type 2-group 116
symplectic pair 117
faithful, of Sl^(3)-type, trivial 117
Tanaka, Y. 37
target group (G*) 27, 54-55, 86-87
the 8 families 3CW (0 13 63-68, 76-77
E n 68-70
i n i ^ ° ) 89
of Lie type 76-77
Lf(q), PSPi{q), G2(q), *DA{q), Lf(q), q odd 61-63
of large Lie rank 63-68, 71, 76
of Lie rank 1 77, 110-113
of small Lie rank 77, 110-116
quasithin 77
sporadic 77, 87
large sporadic 116-117
Theorems
Alperin-Brauer-Gorenstein-Walter classification of groups of 2-rank 2 39, 41
Alperin-Goldschmidt conjugation theorem 97, 122
Aschbacher classical involution theorem 41-43
Aschbacher proper 2-generated core theorem 89
Aschbacher uniqueness case theorem 43, 92
Aschbacher-Bender-Suzuki strongly Z-embedded subgroup theorem 88
Aschbacher-Gilman component theorem 39, 40, 52, 75
generalized to odd primes 75-76
Baumann-Glauberman-Niles theorem 50, 130-131
revision of 50
Bender F*-theorem 17, 138
Bender-Suzuki strongly embedded subgroup theorem 30-31, 33, 36, 38, 52
164 INDEX
for DC-proper groups 75

Bender-Thompson signalizer lemma 116
Bender uniqueness theorem 123
Brauer-Suzuki quaternion theorem 52
Burnside p a g 6 -theorem 30
Classification Theorem 6, 79
comparison of new and old proofs 41-44, 55, 63, 72, 74-76, 98
four-part division 58-59
main logic 106
stages of the proof 61, 106-121
Theorems 61-67 104-106
Curtis-Tits theorem 35, 63, 67, 113, 115
variations for classical groups 35, 63, 113, 115
Feit-Thompson theorem, see Odd Order Theorem
Fong-Seitz classification of split (B, iV)-pairs of rank 2 63, 138
Gilman-Griess recognition theorem for groups of Lie type 35, 67, 71
Glauberman ZJ-theorem 26, 38
Glauberman Z*-theorem 30, 31, 36, 38, 43, 48, 79, 122, 135
global C{G, S)-theorem 38, 53
Goldschmidt strongly closed abelian 2-subgroup theorem 43, 89
Gorenstein-Harada sectional 2-rank at most 4 theorem 39, 42, 43
Gorenstein-Walter dihedral Sylow 2-subgroup classification 50, 74
revision by Bender and Glauberman 50, 74
Hall-Higman Theorem B 26
Holt's Theorem 89
Jordan-Holder Theorem 12fF.
local C(X, T)-theorem 50, 96-97, 129-131
Mason Quasithin Theorem 37, 41
Odd Order Theorem 30, 31, 36, 38-39, 48, 79, 104, 134, 135
revision of 48, 74, 123
p-component uniqueness theorems 30-31, 38, 53, 65, 90-92, 118
Schur-Zassenhaus theorem 24
signalizer functor theorems 49, 124
Solomon maximal 2-component theorem 39
Theorem Q1 104
stages 1-4 106-108
Theorem C2 104
stages 1-4 110-113
Theorem C3 104
stages 1-3 113
Theorem C4 105
stages 1-4 114-116
Theorem C5 105
stages 1-4 116-118
Theorem G6 106
stages 1-2 118
Theorem C7 106
stages 1-5 118-121
Theorem M(5) 5, 90, 99, 129
INDEX 165
stages 1-3 97-98

Theorem PS 92
Theorems PUi, PU 2 , PU 3 92
Theorem SA 89, 97
Theorem SE 89-90
stages 1-2 95-96
Theorem SF 89
Theorem SZ 88-89
Theorem TS 92, 114
Theorem U(<r) 5, 86, 92, 106-108, 129, 131
stages 1-3 98-99
Corollary U(a) 92
Theorem U(2) 89, 97, 106
Thompson A x B-lemma 21
Thompson dihedral lemma 134
Thompson factorization theorem 26, 130
Thompson iV-group classification theorem 38-40, 43
Thompson order formula 114, 136
Thompson replacement theorem 26, 130
Thompson transfer lemma 55, 104, 122, 128
Three subgroup lemma 17
Timmesfeld root involution theorem 42-43
Tits classification of spherical buildings of rank at least 3 73
Walter classification of groups with abelian Sylow 2-subgroups 74
revision by Bender 74
Thompson, J. G. 37, 39-40, 43, 48, 49, 107-108, 130
Thompson subgroup 26, 130
tightly embedded subgroup 39, 43
Timmesfeld, F. G. 37, 39, 40
Tits, J. 10, 25, 34, 137
Tits system, see ( £ , iV)-pair
Tp-groups 57, 102-103, 129
as pumpups 101-103
transfer 29, 31, 74, 107
triality 10
uniqueness grid 79, 83, 84, 87-89

uniqueness case 5, 38, 43, 53
uniqueness subgroups
{2,p}-uniqueness subgroup 53, 98-99
2-uniqueness subgroup 82, 87-89
p-component preuniqueness subgroup 90-93
controlling rank 1 or 2 fusion 91-92
of strongly closed type 93
standard 91-92
wreathed 93
p-uniqueness subgroup, p odd 52-53, 82
strong p-uniqueness subgroup 52-53, 58, 63, 68, 82-83, 92-96, 99, 106
£C p -uniqueness subgroup 94
166 INDEX
near component 2-local uniqueness subgroup 97

strongly embedded subgroup 30-31, 75, 82, 87-89, 95-96
strongly p-embedded subgroup 31, 82, 91-92
almost strongly p-embedded subgroup 91-94
almost p-constrained 93
of strongly closed type 93
wreathed 93
strongly Z-embedded subgroup 87-89, 96
weakly Z-embedded subgroup 89
Walter, J. H. 29, 38-39, 124, 126, 127
Ward, H. 49
weak closure method 43
weakly Z-embedded subgroup, see uniqueness subgroups
Wong, W. 35
Yoshida, T. 29

(40, No 1) (Mathematical Surveys and Monographs, 40, No 1) Daniel Gorenstein, Richard Lyons, Ronald Solomon - The Classification of The Finite Simple Groups-Amer Mathematical Society (1994)

Uploaded by

Copyright:

Available Formats

(40, No 1) (Mathematical Surveys and Monographs, 40, No 1) Daniel Gorenstein, Richard Lyons, Ronald Solomon - The Classification of The Finite Simple Groups-Amer Mathematical Society (1994)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(40, No 1) (Mathematical Surveys and Monographs, 40, No 1) Daniel Gorenstein, Richard Lyons, Ronald Solomon - The Classification of The Finite Simple Groups-Amer Mathematical Society (1994)

Uploaded by

Copyright:

Available Formats

http://dx.doi.org/10.1090/surv/040.

Selected Titles in This Series

2000 Mathematics Subject Classification. P r i m a r y 20D05, 20E32.

C o p y i n g an d reprinting. Individual readers of this publication, and nonprofit libraries

© 1994 by the American Mathematical Society. All rights reserved.

37. Properties of DC-groups 138

Though elated at the successful completion of the classification of finite simple

Richard Lyons and Ronald Solomon

Richard Lyons and Ronald Solomon

INTRODUCTION TO THE SERIES

Finally we note that we are intentionally adopting a fuller style of exposition

This first monograph is preliminary, presenting the conceptual framework un-

This series of monographs is divided into five major parts:

The proof of the Classification Theorem proceeds by induction, focusing on a

It is appropriate here to comment on the status, at the moment of this writing,

group PSLn(q) is simple except for the two cases n = 2, q = 2, and n = 2, g = 3,

TABLE I — T H E FINITE SIMPLE G R O U P S

Group ( n o t e s ) Other names Order

4 , n >5 A.ltn \n\

PSLn+1(q) = Ln+1(q) n+1

#„(<?), n > 2 (2) P&2n+\{q) = ^2n+l(q)

C»(«), n > 2 W PSp2n{q)

£>«(«), n > 3 Pntn(q) = D+{q)

Fi2s M(23) 2is . 313 . 52 . 7 . n . 13 . i 7 . 23

Notes: (1) Ai(2), Ai(3), 2 A 2 ( 2 ), and 2B2(2) are solvable.

of period 3, it is not possible to use this automorphism to construct a twisted form

sequence of finite simple groups, known as the composition factors of X, which

Furthermore, in a chief series, it follows from the maximality of Xi in Xi+i

4. The generalized Fitting subgroup and quasisimple groups

PROPOSITION 4.1. IfV* is the subgroup of X generated by all minimal normal

Xi+1/Xi < Z(X/Xi) for all f, 0 < i < n - 1.

In particular, each Xi+i/Xi must then be abelian. In fact, as Xi+i/Xi is a

group X, and acts trivially on X / Z ( X ) , then it acts trivially on X, This is a simple

5. p'-cores and p-components

F(X) = Fi(X ) < F2{X) < < Fh(X) = X.

LEMMA 5.3. For any set of primes n, we have

6. The embedding of p'-cores and p-components

7. Terminal and p-terminal p-components

Furthermore, we see that / is in fact a p-component of Lp>(Cx{x)) H Cx(y)- In

V € Cx{K) of order p. Indeed, let Q and R be Sylow p-subgroups of C(K,x)

8. p-constrained and p-solvable groups

(i) Y possesses a complement in X (i.e., there is A < X such that X = YA

9. Internal analysis: targeting local structure

C* <* GLk(q) x GLn„k(q).

In particular, we see that C* has at most 2 nonsolvable composition factors, which

10. Internal analysis: passing from global t o local information

(1) G has no proper abelian quotient, i.e., G = [G, G];

G W 3 , GW4]. It can be attempted whenever G contains an elementary abelian

0(CG(a)) H CG(b) = 0(CG{b)) n CG{a) for all a, b e A*.

the subgroup H of G is strongly embedded in G if and only if H < G, H has even

11. Identifying simple groups

A. By presentations (i.e., by generators and relations).

For example, the symmetric group E n is generated by the transpositions x\ =

(11.2) [xa,xp] = Y[X7

12. A capsule summary of this series

is "p-generic". This means that L is neither a group of Lie type in characteristic

We consider the cases

13. The existing classification proof

A four-group is a group isomorphic to Z2 x Zi.