(S
3, S
6)-Amalgams V.
WOLFGANGLEMPKEN(*) - CHRISTOPHERPARKER(**) - PETERROWLEY(***)
Introduction.
In part IV of this present series the commuting case for (S3;S6)- amalgams was examined when a2O(S6) for (a;a0) a critical pair. This paper and the succeeding two parts are devoted to the commuting case when, for (a;a0) a critical pair,a2O(S3). In fact the bulk of our work is concerned with the situation h(Gb;Vb)1, where b2O(S6). Unlike Part IV, for this situation, there appear to be no subamalgams which can be exploited early on. Although a very precise description ofVbis obtained in Theorem 12.1 for our subsequent analysis we need to consider five sub- cases. This subdivision, given in Section 12, is done according to the size of coreGaVb(a2D(b)) and also whetherVb=Zbis an orthogonal module or not.
Of the five possibilities three, as it were, are the ``mainline'' cases - Cases 1, 3 and 4. We briefly discuss and compare each of these cases.
Case 3 is concerned with the smallest possibility for coreGaVb, namely coreGaVbZa. This case has the most complicated possibilities forVb=Zb, withVb=Zb being a quotient of
4 1
1. For this case the core argument (Lemma 9.9) is especially valuable as it often enables us to restrict the size of certain commutators. Our scrutiny of Case 3 takes place in Part VI - the end result being that this case cannot arise! For all the other cases we obtain bounds on the parameterbwhich are pursued further in [LPR2]. At the other end of the scale we have Case 1 with [Vb :coreGaVb]2 (and then Vb=Zb is a natural module). This, from the outset, is a very tight configuration (for example,Vbacts as a central transvection onVa0=Za0for
(*) Indirizzo dell'A.: Institute for Experimental Mathematics, University of Essen, Ellernstrasse 29, Essen, Germany.
(**) Indirizzo dell'A.: School of Mathematics and Statistics, University of Birmingham, Edgbaston, Birmingham B15 2TT, United Kingdom.
(***) Indirizzo dell'A.: School of Mathematics, The University of Manchester, P.O. Box 88, Manchester M60 1QD, United Kingdom.
(a;a0)2C), which nevertheless requires a delicate analysis. By contrast with Case 3 the core argument is no use here whatsoever. Case 4 (when [Vb:coreGaVb]22 and Vb=Zb is a natural module) lies somewhere be- tween Cases 1 and 3. The core argument is of little use and, initally, the configuration has a greater degree of freedom.
We remark that central transvections make their presence felt in a big way in Parts V, VI and VII. Indeed, without such transvections Cases 1, 2 and 3 would be virtually non-existent.
For ease of reference we continue the section numbering started in [LPR1]. We make the most use of material in Sections 1 and 2, and it is there we refer the reader to for notation and background results. Briefly the contents of this paper are as follows. Section 11 is, mostly, a warm-up for the work in Section 12 where the above mentioned structure ofVb is determined. Cases 1 and 2 are the subject of Section 13, the main con- clusions being given in Theorems 13.1 and 13.11.
11. Some preliminary results.
For this paper and Parts V and VI we assume the following hypothesis.
HYPOTHESIS11.0. If(a;a0)2C, then[Za;Za0]1anda2O(S3).
Some elementary observations on this hypothesis are gathered in our first result of this section.
LEMMA11.1. Letm2O(S3)andl2D(m).
(i) [Glm;Zm]ZlV1(Z(Glm))V1(Z(Gl)).
(ii) CZm(Gm)1, GGm(Zm)Qm and ZmZl1Zl2 whenever l1;l22D(m)withl16l2.
(iii) b1(2)and b>1.
(iv) If XGlmand X6Qm, then[X;Zm]ZlCZm(X).
(v) If XQlis such that[X;Vl]61, then[X;Vl]Zl. (vi) [Ql;Vl]Zl.
(vii) G is transitive on paths(l1;l2;l3)of length2wherel12O(S6).
PROOF. Let (a;a0)2C. Then [Za;Za0]1 implies Za0 Z(Ga0) and henceV1(Z(Ga0 1a0))Za0 Zj for allj2D(a0). Thereforeb1(2). Since Gab acts as an involution on Za, [Gab;Za]CZa(Gab)V1(Z(Gab))Zb. Likewise [Gaa 1;Za]Za 1. Hence Za ZbZa 1[Gab;Za]Za 1. By
Lemma 11.1(ii) CZa(Ga)1 and thus Zb\Za 11. Hence, by orders, Zb [Gab;Za]. SinceGacts edge-transitively uponG, we have verified (i) and (ii).
Assume b1 holds. Since, by part (ii), [Za;Qb]Zb this then gives h(Gb;Qb)0, contrary to the hypothesisCGb(Qb)Qb. Thereforeb>1.
For part (iv) we haveGlmXQmand so (iv) follows easily from (i).
Because [X;Vl]61 there exists j2D(l) such that [X;Zj]61. Thus X6Qjby (ii) and now (iv) implies (v). Part (vi) is an easy consequence of (v). While (vii) follows from G being edge-transitive and, as l22O(S3), Gl1l2being transitive onD(l2)nfl1g.
Our next lemma will be put to use after we have established Theo- rem 12.1.
LEMMA 11.2. If Vb=Zb contains either a natural or orthogonal S6- module, then QaQb Gab and[Vb;Qb]Zb.
PROOF. LetVbXZbbe such thatX=Zbis a natural or orthogonal S6-module, and supposeQaQb6Gab. Then we haveQaQband soQjQb for allj2D(b) by Lemma 1.1(i), and henceVbZ(Qb). If [Vb;Qb]6Zb, then, by Lemma 11.1(v), we also haveVbZ(Qb). So, if the lemma is false, we may view Vb as aGF(2) (Gb=Qb)-module. Appealing to Lemma 2.2(iv) givesXYCX(Gb) withY4 or
4 1
. In the former caseCX(Gb)Zb and in the latter [Zb:CX(Gb)]2. In either case (using Proposition 2.5(i) forY
4 1
) we obtainV1(Z(Gab))>Zb, against Lemma 11.1(i). There- foreQaQbGab and [Vb;Qb]Zb, as required.
LEMMA 11.3. Let(a;a0)2C and put CCVb(O2(Gb)). Then Za\CZb. PROOF. Since CZa(Ga)1,Za is a direct sum of 2-dimensional irre- ducible Ga=Qa-modules. IfZa\C>Zb were to hold, then Za\C would contain a non-zeroGa=Qa-submodule ofZa, against Lemma 1.1(ii).
LEMMA 11.4. Let (d;d0)2C. If h(Gd1;Vd1)>1, then [Vd1\
\Qd0;Vd0]61.
PROOF. Suppose the result is false and, without loss of generality, we take ad and a0d0. So h(Gb;Vb)>1 and [Vb\Qa0;Va0]1. Since [Vb:Vb\Qa0]23, [Vb :CVb(Va0)]23. If [Va0 :Va0\Qb]22, then h(Gb;Vb)1. So [Va0 :Va0\Qb]2. Moreover,Va0\QbQaimplies that
[Va0 :CVa0(Za)]2, contrary to h(Ga0;Va0)>1. Therefore Va0\Qb6Qa
and, similarly,Va06Qb.
jZa 1j 2 (11:4:1)
By Lemma 11.1(iv) CZa(Va0\Qb)Zb. Hence Lemma 11.1(ii) implies CZa1(Va0\Qb)1. So
1Za 1\Vb\Qa0 Za 1\Qa0:
Since [Va0 :Va0\Qb]2, we have [Va0:CVa0(Za 1)]22. Therefore, as h(Ga0;Va0)>1, we getjZa 1Qa0=Qa0j 2, which yields (11.4.1).
Forx2Vb; [Va0 :CVa0(x)]22: (11:4:2)
Let x2Vb. From Va0\Qb6Qa and Lemma 11.1(v) [Va0\Qb;Vb]Zb. Thus j[Va0\Qb;x]j 2 by (11.4.1) from which, as [Va0 :Va0\Qb]2, (11.4.2) follows.
If jVbQa0=Qa0j 22, then Vb contains an element y, y62Qa0, which is not a transvection on some non central Ga0-chief factor in Va0 whence h(Ga0;Va0)1 by (11.4.2). Thus [Vb:Vb\Qa0]2 and so [Vb:CVb(Va0)]2. But then since Va0 6Qb, this gives h(Gb;Vb)1, a contradiction. So the lemma holds.
COROLLARY11.5. Let(d;d0)2C and assumeh(Gd1; Vd1)>1. Then (i) [Vd1\Qd0;Vd0]Zd0,[Vd0\Qd1;Vd1]Zd1;and
(ii) Vd0 6Qd1.
PROOF. Lemmas 11.1(v) and 11.4 yield [Vd1\Qd0;Vd0]Zd0. Suppose Vd0 Qd1. Using Lemma 11.1(vi) gives
Zd0[Vd1\Qd0;Vd0][Vd1;Vd0]Zd1
and thus [Vd1;Vd0]Zd0 by orders. Then h(Gd0;Vd0)0 which is im- possible. Therefore Vd0 6Qd1. Hence we may find r2D(a0) such that (r;d1)2C , and Lemmas 11.1(v) and 11.4 give the remaining part of (i).
LEMMA 11.6. Let (a;a0)2C and assume that h(Gb;Vb)>1. If jVa0Qb=Qbj 22, thenjVbQa0=Qa0j 2 jVa0Qb=Qbj.
PROOF. By Corollary 11.5 we may find l2D(b) such that [Zl;Va0\Qb]61. Lemma 11.1(iv) then gives CZl(Va0\Qb)Zb and so CZl\Qa0(Va0\Qb)Zb. Since [Va0 :CVa0(Zl)]2[Va0:Va0\Qb]23 and
h(Ga0;Va0)>1, we infer that [Zl:Zl\Qa0]2. Now Va0\Qb acts as an involution onZl\Qa0and hence
j[Zl\Qa0;Va0\Qb]j [Zl\Qa0 :Zb] jZbj=2:
Because [Zl\Qa0;Va0\Qb]Za0\Zbwe see that [Zb:Za0\Zb]2. Let x2Vb and put Va0 Va0=Za0. Then j[Va0\Qb;x]j jZbj 2 (note that Zb Va0). Therefore [Va0 :CVa0(x)]2[Va0 :Va0\Qb]23 which, as h(Ga0;Va0)>1, forces jVbQa0=Qa0j 2. From Corollary 11.5 there exists r2D(a0) such that (r;b)2C and so, asjVbQa0=Qa0j 2, we may repeat the above argument to obtainjVa0Qb=Qbj 2. This proves the lemma.
LEMMA 11.7. Suppose that b>3,(a;a0)2C andh(Gb;Vb)>1. Then UaGa0.
PROOF. If Ua6Qa0 2, then there exists a 22D[2](a) such that (a 2;a0 2)2C. Applying Corollary 11.5 to (a 2;a0 2) gives
Za 1[Va0 2\Qa 1;Va 1]Va0 2Qa0;
asb>3. ThenZaZa 1ZbQa0, a contradiction. ThereforeUaQa0 2. Now, by Corollary 11.5, Za0 [Vb\Qa0;Va0]Vb and so [Ua;Za0]1.
Therefore
UaCGa0 1(Za0)Ga0 1a0; as required.
LEMMA11.8. Let(a;a0)2C and suppose b>3andh(Gb;Vb)>1. Then [Ua;Va0]6Ua.
PROOF. Supposing [Ua;Va0]Ua we seek to uncover a contradiction.
SettingPb hGab;Va0iandVbbVb=CVb(O2(Gb)) we first show that (11.8.1) (i) Pb=QbS4Z2
(ii) h(Gb;Vb)2 and Vbb is isomorphic to a direct sum of two isomorphic naturalS6-modules
(iii) jZaj 24 andjZbj jZa0j 22. (iv) jZaQa0=Qa0j 2 andj[Za;Va0=Za0]j 22
(v) CZa(Va0\Gab)Zb
From [Ua;Va0]Uawe have thatPbnormalizesUaand hencePb6Gb by Lemma 1.1 (ii). By the parabolic argument (Lemma 3.10)
[Va0 :Va0\Gab]2, and so [Va0 :Va0\Qa]22. Thus [Va0 :CVa0(Za0)]22. Thereforeh(Ga0;Va0)2 withjZaQa0=Qa0j 2, [Va0:CVa0(Za)]22and both non central chief factors ofVa0 are isomorphic naturalS6-modules. Clearly Va0 6Gab and so we conclude that Pb=QbS4Z2. Combining Lemmas 2.16 and 11.1(i), (vi) gives (ii) forVba0, so proving part (ii). Because Vbb hZbGabi it follows from part (ii) that jZbaj 22. Hence jZaj 24 and jZbj 22 by Lemmas 11.1(ii) and 11.3, and we have (iii). Also, by Lemma 11.1(vi), Za acts as an involution upon Va0=Za0. By Lemma 11.1(i) h(Ga0;Va0=Za0)2 and so [Va0=Za0 :CVa0=Za0(Za)]22. Hence j[Za;Va0=Za0]j 22, so establishing (iv). Finally we prove (v). If Va0\Gab Qa, then [Va0:CVa0(Za)]2 which is impossible. So Va0\Gab 6Qa and thusZbCZa(Va0\Gab).
(11.8.2)Pb=Qbis the parabolic with restriction 2
2
on the non-central chief factors inVbb.
Suppose that (11.8.2) is false. Then, by (11.8.1)(i),(ii), Pb=Qb is the parabolic with restriction 1
21
!
on all the (isomorphic) naturalS6-modules inVbb. From Lemma 11.1(i)bZaCbVb(Gab). Letdbe an element of order 3 inPb. SincedcentralizesCVbb(Gab), we infer thatZaCVb(d). Thus
Za/hGab;di Pb:
In particular,Va0normalizesZa0and so [Va0;Za]Za. From (11.8.1)(iv),(v) jZa\Qa0j 23 and CZa(Va0\Gab)Zb and consequently [Za\Qa0;Va0\
\Gab]61. So [Za\Qa0;Va0]61 and an appeal to Lemma 11.1(ii) gives [Za\Qa0;Va0]Za0. Combining this with j[Za;Va0=Za0]j 22 (since jZa0j jZbj) yieldsj[Za;Va0]j 24. But then
Za[Za;Va0]Va0 Qa0;
contrary to (a;a0)2C. With this contradiction we have proved (11.8.2).
jVbQa0=Qa0j 22: (11:8:3)
If jVbQa0=Qa0j 622, then VbQa0 ZaQa0 whence [Vb;Va0=Za0]
[Za;Va0=Za0]. SinceZa0 [Vb;Va0] by Corollary 11.5, using (11.8.1)(iii),(iv) we see thatj[Vb;Va0]j 24. Hence, asjZbj 22,j[Va0;Vb=Zb]j 22. Conse- quentlyVa0acts as a transvection upon each of the non-central chief factors in Vbb. Together (11.8.2) and Proposition 2.5(viii) imply thatO2(Pb)=Qb is the unique quadratically acting (upon each of the non-central chief factors inVbb) E(23)-subgroup of each Sylow 2-subgroup ofPb=Qb. Then Proposition 2.5(iii)
forcesVa0 O2(Pb)Gab, contradicting (11.8.1)(i). Thus we conclude that jVbQa0=Qa0j 22.
(11.8.4) A contradiction
By Corollary 11.5 there exists r2D(a0) such that (r;b)2C. If jVa0Qb=Qbj 22, then applying Lemma 11.6 to (r;b) yieldsjVbQa0=Qa0j 2, against (11.8.3). Therefore, asVa0 is elementary abelian,jVa0Qb=Qbj 23, and, of course, Va0 acts quadratically on Vb. Again, using (11.8.2) and Proposition 2.5(viii), we deduce the untenableVa0 O2(Pb)Gab. This is the desired contradiction which completes the proof of Lemma 11.8.
12. The structure ofVb/Zb.
THEOREM12.1. Let(a;a0)2C. Then
(i) h(Gb;Vb)1 with Vb=Zb isomorphic to a quotient of 4
1
1;
and
(ii) jZaj 22,jZbj 2.
PROOF. First we suppose thatb>3 andh(Gb;Vb)>1 and argue for a contradiction. So Lemmas 11.7 and 11.8 are available to giveUaGa0and [Ua;Va0]6Ua. Also, from Corollary 11.5, [Vb;Va0]Za0. So if UaQa0
VbQa0, we then haveUaVb(Ua\Qa0) whence, using Lemma 11.1(vi), [Ua;Va0][Vb;Va0]VbUa, a contradiction. ThereforeUaQa0 6VbQa0. SinceUa is elementary abelian we infer thatjVbQa0=Qa0j 22. Now Lem- ma 11.6 and Corollary 11.5(ii) give
jVbQa0=Qa0j jVa0Qb=Qbj 2:
(12:1:1)
PutVb Vb=Zb.
(12.1.2) (i) (Va0\Qb)Qa Gab.
(ii) h(Gb;Vb)2 and there exists Xb/Gb with Zb XbVb such thatVb XbCVb(Gb) andXbis isomorphic to a quotient of
4 1
4 1
. Moreover, the two 4's inXbare isomorphic naturalS6-modules.
(iii) jZaj 24,jZbj jZa0j 22andh(Ga;Za)2.
By (12.1.1) [Va0 :Va0\Qb]2 and so, ash(Ga0;Va0)>1,Va0\Qb6Qa. Thus (i) holds, and [Va0 :CVa0(Za)]22 with h(Ga0;Va0)2. Applying
Lemmas 2.16 and 2.2(iv) toVa0Va0=Za0yields (ii) forVa0, so proving (ii).
Using Lemmas 11.1(ii), 11.3, part (ii) and the fact thatZ(Ga)1 gives (iii).
SetD(a) fb;a 1;lg.
[Va0 :Va0\Qa 1]24: (12:1:3)
From (12.1.1) we have ZaQa0VbQa0 and henceVa 1Qa0 VbQa0. By (12.1.2), UaQa0Va 1Qa0. Since UaQa0 6VbQa0, jVa 1Qa0=Qa0j 22 and hence, by (12.1.2)(ii) and Proposition 2.5(iii), [Xa0 :CXa0(Va0 1)]24. If [Xa0\Qa 1;Va 1]61, then by Lemma 11.1(v)
Za 1[Xa0\Qa 1;Va 1]Va0 Qa0
whence ZaZa 1Zb Qa0. Hence [Xa0\Qa 1;Va 1]1 from which (12.1.3) follows.
Va0 2Qa 1 and [Ua;Va0 2]1:
(12:1:4)
Suppose Va0 26Qa 1. Then there exists r2D(a0 2) such that (r;a 1)2C. Applying Corollary 11.5 to (r;a 1) gives Za 1 [Va0 2;Va 1]Va0 2Qa0, since b>3. This is against (a;a0)2C , and therefore Va0 2Qa 1. If [Va0 2;Va 1]61, then Lemma 11.1(v) gives Za 1[Va0 2;Va 1] which again contradicts (a;a0)2C. Thus [Va0 2;Va 1]1 and likewise [Va0 2;Vl]1, so we have (12.1.4).
[Va0 :Va0\Va0 2]24: (12:1:5)
This follows from (12.1.3) and (12.1.4).
(12:1:6)j[Ua;Va0\Qb]j24[Va0:Va0\Va0 2] and (hence)j[Ua;Va0\Qb]j28: In view of (12.1.3)Va0\Qb\Qa must act as at least a fours group on Va 1=Za 1, and thusj[Va0\Qb\Qa;Va 1]j 24 by (12.1.2)(ii) and Propo- sition 2.5(iii). Further, we observe, as [Va0\Qb\Qa;Va 1]Va0 and Va0\Qb interchangeslanda 1, that
[Va0\Qb\Qa;Va 1][Va0\Qb\Qa;Vl]Va 1\Vl:
Let t2Va0\Qb be such that Va0\Qb(Va0\Qb\Qa)hti. Then, as t normalizes Va 1Vl and Va 1\Vl, we have j[Va 1Vl=Va 1\Vl;t]j
[Va 1:Va 1\Vl]. Now
[Va0\Qb;Va 1Vl][Va0\Qb\Qa;Va 1][t;Va 1Vl] from which we deduce that
j[Va0\Qb;Va 1Vl]j 24[Va 1:Va 1\Vl]:
So, by Lemma 11.1(vii),
j[Va0\Qb;Ua]j 24[Va0 :Va0\Va0 2]:
This, together with (12.1.5), yieldsj[Va0\Qb;Ua]j 28. Ua doesn0t act quadratically uponXa0=Za0: (12:1:7)
Suppose (12.1.7) is false. Then, combining (12.1.2)(ii), (iii) and Propo- sition 2.5(ii), we have j[Ua;Xa0]j 28. Noting that [Ua;Xa0] [Ua;Va0], (12.1.6) forces
[Ua;Va0][Ua;Va0\Qb]Ua; contradicting Lemma 11.8. Thus (12.1.7) holds.
h(Ga;Ua)4:
(12:1:8)
Set Ua(i)[Ua;Qa;i] and Vb(i)[Vb;Qa;i]. From (12.1.2)(ii) and Lemma 11.2,GabQaQb. HenceVb(3)61 by Proposition 2.5(i). Our aim is to show that U(1)a 6Vb(1) andUa(2)6Vb(2). As a consequences (see Lemma 1.2(v))Ua=Ua(1),Ua(1)=Ua(2),Ua(2)=Ua(3)all contain at least one non-central chief factor for Ga. Because Gab.Vb(3)61, Vb(3)\Zb61 and so h(Ga;Ua(3))60 by Lemma 1.1(ii), we then obtain (12.1.8).
Suppose Ua(1)Vb(1). Then [Va0;Qa0 1][Va0 2;Qa0 1] and so, by (12.1.4), Ua centralizes [Va0;Qa0 1][Xa0;Qa0 1], contrary to (12.1.7). If Ua(2)Vb(2), then likewise Ua centralizes [Xa0;Qa0 1;Qa0 1] which again contradicts (12.1.7). ThereforeUa(1)6Vb(1)andUa(2)6Vb(2), as desired.
(12.1.9) UaQa0=Qa0 is the non-quadratic E(23)-subgroup of Ga0 1a0=Qa0
acting onXa0=Za0.
Since [Za:Za\Qa0]2, (12.1.2)(i), (iii) imply that [Za\Qa0;Va0\
\Qb]61. So there exists r2D(a0) for which [Za\Qa0;Zr]61. Thus Za\Qa0 6Qr. Note that [Zr:Zr\Qb]2 by (12.1.1). Hence Zr\Qb 6Qa for Zr\QbQa would imply [Zr:CZr(Za\Qa0)]2 and thenceh(Gr;Zr)1, contrary to (12.1.2)(iii).
Now
[Ua:CUa(Zr\Qb)][Ua:Ua\Qr]2[Ua:Ua\Qa0]
and therefore (12.1.8) forces [Ua:Ua\Qa0]23. Thus, by (12.1.7), we have established (12.1.9).
We are now in a position to deduce the desired contradiction.
Combining (12.1.9), (12.1.2)(ii) with Proposition 2.5(viii) we get
[Xa0:CXa0(Ua)]26. Therefore [Va0 :Va0\Va0 2]26 by (12.1.4). Hence (12.1.6) forces j[Ua;Va0\Qb]j 210. Since jZa0j 22 and Za0[Ua;Va0], (12.1.2)(ii) impliesj[Ua;Va0]j 210. Consequently
[Ua;Va0][Ua;Va0\Qb]Ua;
again contradicting Lemma 11.8. This completes the proof that h(Gb;Vb)1 when b>3. Now we examine the situation h(Gb;Vb)>1 whenb3.
(12.1.10) (i) Za2[Vb;Va0];
(ii) either jVbQa0=Qa0j jVa0Qb=Qbj 2 or jVbQa0=Qa0j
jVa0Qb=Qbj 23;
(iii) h(Gb;Vb)2 and, for i1;2, there exists Xb(i)/Gb with ZbXb(i) such that VbXb(1)Xb(2) and Xb(i)=Zb4 or
4 1
. Further the 4's inX(1)b andX(2)b are isomorphic.
Part (i) follows from Corollary 11.5(i) while Corollary 11.5(ii) and Lemma 11.6 imply (ii). For (iii) we first note that (i) gives [Vb;Gb]Vb. If jVbQa0=Qa0j jVa0Qb=Qbj 2, then [Va0 :CVa0(Za)]22 and so h(Ga0;Va0)2 with both non-central chief factors being isomorphic nat- ural modules. Using Lemma 2.16 and [Va0;Ga0]Va0 yields the desired structure of Vb in this case. Turning to the latter possibility given in (ii) we deduce, just as in Lemma 11.6, that for all x2Vb [Va0 :CVa0(x)]24. Hence h(Ga0;Va0)2 and, as Vb acts as a quadraticE(23)-group on Va0, both non central chief factors are isomorphic natural modules. Appealing to Lemma 2.4 and Proposition 2.7(ii) completes the proof of (iii).
By (12.1.10)(iii) Vb=Zb is a quotient of 4
1
4
1
. Put Yb
CVb(O2(Gb)) and, fori1;2, letYb(i)be such thatYb Yb(i),Yb(i)/Gband Yb(i)=Yb4 with Vb=YbYb(1)=YbYb(2)=Yb. Let Yb(3) denote the inverse image in Vb of the diagonal submodule. Clearly VbYb(i)Yb(j) for i6j2 f1;2;3g so, without loss of generality, we may assume that Ya(1)0 6Qb andYa(2)0 6Qb.
Assume for the moment thatjVbQa0=Qa0j jVa0Qb=Qbj 2. SoYa(1)0 acts as a transvection upon each of the non-central chief factors withinVb. If [Ya(1)0 \Qb;Vb]61, then E(22)Zb[Ya(1)0 \Qb;Vb] whence [Ya(1)0 ;Vb]
E(24). However consideringVb acting uponYa(1)0 yields, asVb acts as a transvection uponYa(1)0 =Za0, that [Vb;Ya(1)0 ]E(23). Thus we conclude that [Ya(1)0 \Qb;Vb]1 and, similarly, [Ya(2)0 \Qb;Vb]1. So Ya(i)0 \Qb
CY(i)
a0(Vb) for i1;2 and, in particular, Ya(i)0 \Qb Yb. Since
[Ya(i)0 :Ya(i)0 \Qb]2[Va0 :Va0\Qb] and, for i1;2, Ya(i)0 6Qb, we see thatVa0\Qb(Ya(1)0 \Qb)(Ya(2)0 \Qb)Ya(3)0 . In particularYa(3)0 Qband, since h(Ga0;Ya(3)0 )1, Ya(3)0 6Qa. Therefore Zb [Za;Ya(3)0 ]Ya(3)0 . But then Va0Ya(3)0 which is impossible. So now we only need consider the possibility jVbQa0=Qa0j jVa0Qb=Qbj 23. Recalling that Za0[Vb;Va0], Propositio- n 2.5(ii) implies that
[Va0;Vb]Za0
Xa(1)0 ;Ga0a0 1;2
Xa(2)0 ;Ga0a0 1;2 : Since [Va0;Vb]CVa0(Vb) we see that
[Va0;Vb]Za0
Xa(1)0 ;Ga0a0 1;2
Xa(2)0 ;Ga0a0 1;2 :
Because [Va0 :Va0\Qb]23 we have Va0\Qb hxi[Va0;Vb] where xx1x2withxi2Xa(i)0. Since [Va0\Qb;Vb]ZbZa2, [x;Vb]Za2.
Set Va0Va0=Ya0. Then Va0Xa(1)0 Xa(2)0 44 and by Lemma 11.3 Za2(Za2\Xa(1)0 )(Za2\X(2)a0 )E(22). SinceX(i)a0 /Ga0, we infer that [xi;Vb]Za2\Xa(i)0 (i1;2). Without loss of generality we may suppose x162[Xa(1)0 ;Ga0a0 1;2] (becauseVa0\Qb6[Va0;Vb]). Sohx1i[X(1)a 1;Ga0a0 1;2]
E(23) with
Vb;hx1i
Xa(1)0 ;Ga0a0 1;2
Za2\Xa(1)0 :
But Proposition 2.5(ix) shows this to be impossible and so we have ruled out the possibility h(Gb;Vb)>1 and b3. Therefore we have shown that h(Gb;Vb)1. The remainder of the theorem follows from Proposition 2.9(i) and Lemmas 2.2(iv), 11.3.
LEMMA12.2. Let(a;a0)2C andD(a) fl1;l2;l3g.
(i) V1(Z(Gab))ZbZ2 andV1(Z(Qa))ZaZliZlj E(22), 1i<j3.
(ii) Gab QaQb and[Vb;Qb]Zb. (iii) coreGaVbVli\Vlj, i6j.
(iv) Zb V1(Z(Qb)) 1
1
.
PROOF. (i) is mostly a restatement of Theorem 12.1; V1(Z(Qa))Za
follows sinceZa is a projectiveGa=Qa-module and V1(Z(Qab))Zb Za. Part (ii) follows from Theorem 12.1 and Lemma 11.2.
Iffi;j;kg f1;2;3g, then clearlyGalknormalizesVli\Vlj. Also, [Vli\Vlj;Qli][Vli;Qli]ZliZaVli\Vlj:
Using part (ii) we now see thatVli\Vlj/Ga, so proving (iii).
Turning to (iv), we setMV1(Z(Qb)). Observing thatM is aGb=Qb- module with soc(M)Zb we haveM,!P(1). SincejV1(Z(Gab))j 2, Pro- position 2.5(i) implies that M doesn't contain submodules of type
4 1
. HenceM
11
by Lemma 2.4.
We will make frequent use of Lemma 12.2, often without specific re- ference.
LEMMA 12.3. Let (a;a0)2C and suppose that coreGaVb>Za0. Then Vb=Zbis isomorphic to4or
4 1
.
PROOF. Since coreGaVb>Za, [coreGaVb;Qa]61 by Lemma 12.2(i).
Clearly [coreGaVb;Qa]/Ga and so Lemma 12.2(i) implies that Za [coreGaVb;Qa]. Hence
Za[coreGaVb;Qa][Vb;Gb]
and therefore Vb [Vb;Gb]. Now the lemma is a consequence of Theo- rem 12.1.
LEMMA12.4. Let(a;a0)2C. Then
(i) [Qb;O2(Gb)]6Qa; in particular Qa\Qa6/Gb. (ii) GbO2(Gb)Qa.
PROOF. (i) PutRb:[Qb;O2(Gb)] and assume thatRbQa. Clearly, Rb Q:Qa\Qb and so Q/Gb with QCQb(Za)CQb(Vb). Since [Qb:Q]2 we get 22 jV1(Z(Qb))j jCVb(Qb)j
jVbj
p and hence
jVbj 24, a contradiction. This proves (i).
(ii) If Lb :O2(Gb)Qa6Gb, then [Gb:Lb]2 with Qa2SyI2Lb; so Qa\QbO2(Lb)/Gb, but this contradicts (i).
Let (a;a0)2C and put YbCVb(O2(Gb)). Assume that jYbj 22. We now define
Fa hYljl2D(a)i DYbGaE and
Hb hFmjm2D(b)i FaGb
:
ClearlyFa/GaandHb/Gb. Some less obvious properties of these groups are given in the next lemma.
LEMMA12.5.
(i) h(Ga;Fa)2; and (ii) h(Gb;Hb)2.
PROOF. (i) First we observe that ZaFa and, as jYbj 22, that h(Ga;Fa)2. Ifh(Ga;Fa)1, thenFaYbZaand hence
[Qa;Fa][Qa;YbZa][Qa;Yb]Zb <Za:
Since [Qa;Fa]/Ga, we then getFaV1(Z(Qa))Za, which is impossible.
Thush(Ga;Fa)2.
(ii) SinceZa Fa, clearlyVb Hb. So if (ii) is false, thenHbFaVb. Hence [FaVb;Qb]/Gb and
[FaVb;Qb][Fa;Qb][Vb;Qb][Fa;Qb]Zb:
By GabQaQb and (i) j[Fa;Qb]j 22 and so [FaVb;Qb][Fa;Qb].
Because Fa=Za2 or 21, we see that j[Fa;Qb]j 23 and therefore [Fa;Qb] is centralized by O2(Gb). Consequently Za6[Fa;Qb] and so Za\[Fa;Qb]Zb. Next we consider
[Fa;Qa\Qb][Fa;Qa]\[Fa;Qb]Za\[Fa;Qb]Zb:
Therefore [VbFa;Qa\Qb]Zb. So Qa\QbCQb(VbFa=Zb). Recalling that [FaVb;Qb][Fa;Qb] has order at least 22 we deduce thatQa\Qb
CQb(VbFa=Zb). But thenQa\Qb/Gb, contrary to Lemma 12.4(i). This completes the verification of (ii).
The groupsFaandHb will be important in later arguments in, for ex- ample, Section 13 and Part VI. Now we return to examine further the si- tuation in Lemma 12.3 in our next lemma, where Fa makes a brief ap- pearance.
LEMMA 12.6. Let (a;a0)2C, and assume that coreGaVbVa 1\
\Vb>Za.
(i) If Vb=Zb4, then Va 1\Vb[Vb;Gab] or Va 1\Vb
[Vb;Gab;Gab].
(ii) Assume that Vb=Zb 4
1
, and set Yb CVb(O2(Gb)). Then Yb6Va 1\Vband Yb(Va 1\Vb)6Vb. Hence one of the following holds: