(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)
(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)
(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)
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.
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
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
Glossary 148
Index 155
This page intentionally left blank
Preface
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.
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.
PRELIMINARIES
This page intentionally left blank
http://dx.doi.org/10.1090/surv/040.1/01
PART I, CHAPTER 1
OVERVIEW
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
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
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.
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
2 PSUn+1(q) = Un+1(q)
An(q), n >2 «
= L-+1(q) = A-(q)
i=2
2
252(9) (D.(3) Sz{q) = B2(^q) ? 2 ( ? - i ) ( « 2 + i)
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
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
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
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<i x PSL2(q), q odd. The general solution is Ree's twisted group 2G?2(#)> 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 ] .
12 PART I, CHAPTER 1: OVERVIEW
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
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
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 O^iX) 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.
20 PART I, CHAPTER 1: OVERVIEW
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.
22 PART I, CHAPTER 1: OVERVIEW
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
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
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
[ 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
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.
34 PART I, CHAPTER 1: OVERVIEW
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.
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 }.
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.
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
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 ] .
44 PART I, CHAPTER 1: OVERVIEW
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
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]).
46 PART I, CHAPTER 1: OVERVIEW
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.
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.
20
I n Part II, Peterfalvi makes occasional reference to technical lemmas from Part I and we
include these implicitly in the above reference.
50 PART I, CHAPTER 1: OVERVIEW
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.
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
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.
/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^SL^iq);
(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^J^An-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
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.
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.
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
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.
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.
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
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)#).
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*.
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.
We hope that this rather detailed outline provides a useful summary of the
generic case analysis.
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
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.
<* 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
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
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.
74 PART I, CHAPTER 1: OVERVIEW
v(1)
;M is solvable; or
(29.1)
V ;
(2) E(M/02>(M)) ^ L2(q) for some q = 3 mod 4
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
76 PART I, CHAPTER 1: OVERVIEW
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.
PART I, CHAPTER 2
OUTLINE OF PROOF
INTRODUCTION
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
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.
82 PART I, CHAPTER 2: OUTLINE OF PROOF
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
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
UNIQUENESS GRID
Hypothesis stage 1 stage 2 stage 3 stage 4
p=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
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)*)
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 )
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
86 PART I, CHAPTER 2: OUTLINE OF PROOF
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 < i < 7 of target groups:
3C(°> = {£ 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
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
88 PART I, CHAPTER 2: OUTLINE OF PROOF
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) G^A9.
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.
90 PART I, CHAPTER 2: OUTLINE OF PROOF
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.
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
98 PART I, CHAPTER 2: OUTLINE OF PROOF
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
100 PART I, CHAPTER 2: OUTLINE OF PROOF
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.
(2) p = 3 and either
(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 .
(3) p = 5 and either
(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.
102 PART I, CHAPTER 2: OUTLINE OF PROOF
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.
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:
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.
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.
104 PART I, CHAPTER 2: OUTLINE OF PROOF
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.
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
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:
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^^O *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.
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
5
This was done by Aschbacher for p = 2 [A5], and the case p = 3 is similar.
110 PART I, CHAPTER 2: OUTLINE OF PROOF
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
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
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.
114 PART I, CHAPTER 2: OUTLINE OF PROOF
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 X^A\ 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.
116 PART I, CHAPTER 2: OUTLINE OF PROOF
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
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.)
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
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
G0{N) = {Iu\ue{x,y)#).
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 < i < n, for some positive integer n, satisfying
(i) CG{Di) < Op,p(Ni), l<i<n;
(ii) Oplp(Ni) = Op>(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
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*
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
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),
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
(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.
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.
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
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
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
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.
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
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
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.
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
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.
144 EXPOSITORY REFERENCES
[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.
146 EXPOSITORY REFERENCES
PAGE SYMBOL
90 rvj
97 y
4-5 [Ii], [I2], [IA], M, [HA], [HG], [Up], [Us]
60, 101 ^Ani [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
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
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
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
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
Tanaka, Y. 37
target group (G*) 27, 54-55, 86-87
the 8 families 3CW (0 < i < 7) 86
An, n < 12 77, 111-112, 114-116
An, n > 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