This paper investigates a specific class of finite groups with large p-subgroups, focusing on wreath product cases within the Local Structure Theorem, to enhance understanding of groups of Lie type and finite simple groups.
Contribution
It introduces a detailed analysis of wreath product cases in the Local Structure Theorem, providing new insights into the structure of groups with large p-subgroups.
Findings
01
Characterization of wreath product cases in the Local Structure Theorem
02
Enhanced understanding of groups of Lie type in characteristic p
03
Implications for the classification of finite simple groups
Abstract
Groups with a large p-subgroup, p a prime, include almost all of the groups of Lie type in characteristic p and so the study of such groups adds to our understanding of the finite simple groups. In this article we study a special class of such groups which appear as wreath product cases of the Local Structure Theorem.
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Full text
\frefformat
plainthmTheorem #1
\frefformatplaineq(#1)
\frefformatplainsecSection #1
\frefformatplainnotNotation #1
\frefformatplainhypHypothesis #1
\frefformatplainlemLemma #1
\frefformatplaindefDefinition #1
\frefformatplainpropProposition #1
\frefformatplaincorCorollary #1
\frefformatplaintabTable #1
\frefformatplaincorCorollary #1
\frefformatplainchpChapter #1
\frefformatplainclaimClaim #1
\frefformatplainclm#1
The Local Structure Theorem: The wreath product case
Groups with a large p-subgroup, p a prime, include almost all of the groups of Lie type in characteristic p and so the study of such groups adds to our understanding of the finite simple groups. In this article we study a special class of such groups which appear as wreath product cases of the Local Structure Theorem [MSS2].
Dedicated to the memory of Kay Magaard
1. Introduction
Throughout this article p is a prime and G is a finite group.
We say that L≤G has characteristic p if
[TABLE]
For T a non-trivial p-subgroup of G,
the subgroup NG(T) is called a * p-local* subgroup of G. By definition G has local characteristic p if all p-local subgroups of G have characteristic p and G has parabolic characteristic p if all p-local subgroups containing a Sylow p-subgroup of G have characteristic p.
A group K is called a K-group if all its composition factors are from the known finite simple groups. So, if K is a simple K-group, then K is a cyclic group of prime order, an alternating group, a simple group of Lie type or one of the 26 sporadic simple groups. A group G is a
Kp-group, provided all subgroups of all p-local subgroups of G are K-groups.
This paper is
part of a programme to investigate the structure of certain Kp-groups. See [MSS1, MSS2] for an overview of the project.
Of fundamental importance to the development of the programme are large subgroups of G: a
p-subgroup Q of G is large if
(i)
CG(Q)≤Q; and
(ii)
NG(U)≤NG(Q) for all 1=U≤CG(Q).
For example, if G is a simple group of Lie type defined in characteristic p, S∈Sylp(G) and Q=Op(CG(Z(S))),
then Q is a large subgroup of G unless there is some degeneracy in the Chevalley commutator relations which define G. This means that Q is a large subgroup of G unless G is one of Sp2n(2k), n≥2, F4(2k) or G2(3k).
If Q is a large subgroup of G, then it is easy to see that Op(NG(Q)) is also a large p-subgroup of G. Thus we also assume that
(iii)
Q=Op(NG(Q)).
One of the consequences of G having a large p-subgroup is that G has parabolic characteristic p. In fact any p-local subgroup of G containing Q is of characteristic p [MSS2, Lemma 1.5.5 (e)].
Further, if Q≤S∈Sylp(G), then Q is weakly closed in S with respect to G (Q is the unique G-conjugate of Q in S) [MSS2, Lemma 1.5.2 (e)].
A significant part of the programme described in [MSS1] aims to determine the groups which possess a large p-subgroup. This endeavour extends and generalizes earlier work of Timmesfeld and others in the original proof of the classification theorem where groups with a so-called large extraspecial 2-subgroup were investigated. The state of play at the moment is that the Local Structure Theorem has been completed and published [MSS2]. To describe this result we need some further notation.
For
a finite group L, YL denotes the unique maximal elementary abelian normal p-subgroup of L with
Op(L/CL(YL))=1. Such a subgroup exists [MSS1, Lemma 2.0.1(a)]. From now on assume that G is a finite Kp-group, S a
Sylow p-subgroup of G and Q a large p-subgroup of G with Q≤S and Q=Op(NG(Q)).
We define
[TABLE]
Under the assumption that S is contained in at least two maximal p-local subgroups, for L∈LG(S) with L≤NG(Q), the Local Structure Theorem provides information about L/CL(YL) and its action on YL. Given the Local Structure Theorem there are two cases to treat in order to fully understand groups with a large p-subgroup. Either there exists L∈LG(S) with YL≤Q or, for all L∈LG(S), YL≤Q. Research in the first case has just started and, for this situation, this paper addresses the wreath product scenario in the Local Structure Theorem [MSS2, Theorem A (3)]. This case is separated from the rest because of the special structure of L and YL. This structure allows us to use arguments measuring the size of certain subgroups to reduce to three exceptional configurations and has a distinct flavour from the remaining cases. For instance, the groups which are examples in the wreath product case typically have Q of class 3 whereas in the more typical cases it has class at most 2. The configurations in the Local Structure Theorem which are not in the wreath product case and have YL≤Q will be examined in a separate publication as there are methods which apply uniformly to cover many possibilities at once.
Contributions to the YL≤Q for all L∈LG(S) are the subject of [PPS].
For L∈LG(S) with Q not normal in L we set
[TABLE]
and use this notation throughout the paper.
Set q=pa. We recall from [MSS2, Remark A.25] the definition of a natural wreath SL2(q)-module for the group X with respect to K:
suppose that X is a group, V is a faithful X-module and K is a non-empty X-invariant set of subgroups of X. Then V is a natural SL2(q)-wreath product module for X with respect to K if and only if
[TABLE]
and, for each K∈K, K≅SL2(q) and [V,K] is the natural SL2(q)-module for K.
We now describe the wreath product case in [MSS2, Theorem A (3)]. For L∈LG(S) with L≤NG(Q), L is in the wreath product case provided
•
there exists a unique L-invariant set K of subgroups of L such that VL is a natural SL2(q)-wreath product module for L with respect to K.
•
L∘=Op(⟨K⟩)Q and Q acts transitively on K by conjugation.
•
YL=VL or p=2, ∣YL:VL∣=2, L∘≅SL2(4) or ΓSL2(4) and VL≤Q.
We say that L is properly wreathed if ∣K∣>1.
There are overlaps between the wreath product case and some other divisions in the Local Structure Theorem.
If L∘≅SL2(q) with VL=YL, then this situation can be inserted in the linear case of [MSS2, Theorem A (1)] by including n=2 is that case. Suppose that ∣K∣=2 and K≅SL2(2). If Q is a fours group, then, as Q conjugates K1 to K2,
[TABLE]
and YL is the tensor product module. This is an example in the tensor product case of [MSS2, Theorem A (6)].
We declare L to be in the unambiguous wreath product case if these two ambiguous configurations do not occur.
The ambiguous cases will be handled in a more general setting in a forthcoming paper mentioned earlier.
Main Theorem**.**
Suppose that p is a prime, G is a finite group, S a Sylow
p-subgroup of G and Q≤S is a large p-subgroup of G with Q=Op(NG(Q)).
If there exists L∈LG(S) with L in the unambiguous wreath product case and VL≤Q, then G≅Mat(22), Aut(Mat(22)), Sym(8), Sym(9) or Alt(10).
The proof of this theorem splits into four parts. First, in Section 3, we show that in the properly wreathed case we must have q=∣K∣=2 and, as L is unambiguous, S=Q≅Dih(8) and L∘≅O4+(2). If ∣K∣=1, we show that L∘≅ΓSL2(4) or SL2(4) and VL is the natural module with ∣YL:VL∣≤2, where, if L∘≅SL2(4), ∣YL:VL∣=2 holds. In the following three sections, we determine the groups corresponding to these three cases. Finally the Main Theorem follows by combining Propositions 3.5, 4.1, 5.1 and 6.2.
In [PPS] the authors proved that the unambiguous wreath product case does not lead to examples if for all L∈LG(S) we have YL≤Q, with the additional assumption that G is of local characteristic p. In this paper we do not make the assumption that G is of local characteristic p.
In the Local Structure Theorem there is also a possibility that L∈LG(S) is of weak wreath type. Any such group is contained in one, which is of unambiguous wreath type. A corollary of our theorem is
Corollary**.**
Suppose that p is a prime, G is a finite group, S a Sylow
p-subgroup of G and Q≤S is a large p-subgroup of G with Q=Op(NG(Q)). If L∈LG(S) is of weak wreath product type, then either G is as in the Main Theorem or VL≤Q.
In addition to the notation already introduced, we will use the following
Notation**.**
For p a prime, G a group with a large p-subgroup Q=Op(NG(Q)) and L∈LG(S), we set QL=Op(L) and assume that VL≤Q. Define D=⟨VLNG(Q)⟩(L∩NG(Q))∈LG(S). Furthermore, set
[TABLE]
[TABLE]
and
[TABLE]
Notice that for L0=NL(S∩CL(YL)), we have L=CL(YL)L0 and CL(YL)≤D. Further
[TABLE]
by [MSS2, Lemma 1.2.4 (i)]. Since CL(YL) normalizes Q,
[TABLE]
Therefore, if L is in the unambiguous wreath product case, then so is L0. Hence we also assume that L=L0 and so
[TABLE]
2. Preliminaries
In this section we present some lemmas which will be used in the forthcoming sections.
Lemma 2.1**.**
Suppose that X is a group, E=O2(X) is elementary abelian of order 16 and X/E≅Alt(6) induces the non-trivial irreducible part of the 6-point permutation module on E. Then X splits over E.
Proof.
Choose R≤X such that R/E≅Sym(4) and Z(R)=1. Let T∈Syl3(R). As T acts fixed-point freely on O2(R), NR(T)≅Sym(3) and so there are involutions in X/E. Hence, as X/E has one conjugacy class of involutions, there are involutions in O2(R)∖E. Therefore O2(R)/Z(O2(R)) is elementary abelian of order 16. Now we consider O2(R). The fixed-point free action of T on O2(R)/Z(O2(R)) implies there is partition of this group into five T-invariant subgroups of order 4. As T acts fixed-point freely on O2(R) the preimages of all these fours groups are abelian.
As there are involutions in O2(R)∖E, there is a T-invariant fours group F∗≤O2(R)/Z(O2(R)) with F∗=E/Z(O2(R)) and such that the preimage F of F∗ is elementary abelian of order 16.
Now the action of X on E shows that for any involution i∈R∖E all involutions in the coset Ei are conjugate to i by an element of E. Hence all involutions in O2(R)∖E are in F. This shows that F is invariant under NR(T).
Again there is a partition of F into five groups of order four invariant under T. Let t be an involution in NR(T). Then ∣CF(t)∣=4, where ∣CE∩F(t)∣=2. Hence there is some fours group F1≤F, F1=E∩F and CF1(t)=1. This shows that F1 is normalized by t. Then F1⟨t⟩≅Dih(8) is a complement to E. Using a result of Gaschütz [GLS2, Theorem 9.26] , X splits over E.
∎
The next lemma is well-known.
Lemma 2.2**.**
Suppose that X≅Sym(5), F1 and F2 are fours groups of X with F1≤Alt(5) and V is a non-trivial irreducible GF(2)X-module. Then
(i)
V* is either the non-trivial irreducible part of the permutation module, which is the same as the natural O4−(2)-module, or V is the natural ΓL2(4)-module.*
2. (ii)
F1* acts quadratically on V if and only if V is the natural ΓL2(4)-module.*
3. (iii)
F2* acts quadratically on V if and only if V is the natural O4−(2)-module.*
Lemma 2.3**.**
Suppose that p is a prime, X is a group of characteristic p and U is a normal p-subgroup of X.
Let R be a normal subgroup of X with R≤CX(U/[U,Op(X)]). If [Op(X),Op(R)]≤U, then R≤Op(X).
Proof.
It suffices to prove that Op(R)=1. Suppose that n≥1 is such that [U,Op(R)]≤[U,Op(X);n]. Then
[TABLE]
and so
[TABLE]
We also have
[TABLE]
and thus the Three Subgroups Lemma implies
[TABLE]
This yields
[TABLE]
Since Op(X) is nilpotent, we deduce [U,Op(R)]=1. Hence
[TABLE]
As X has characteristic p, Op(R)=1 and so R≤Op(X) as claimed.
∎
Lemma 2.4**.**
Assume that X is a group, Y is a normal subgroup of X and xCX(Y)∈Z(X/CX(Y)). If [Y,x]≤Z(Y), then Y/CY(x)≅[Y,x] as X-groups.
Proof.
Define
[TABLE]
Then θ is independent of the choice of the coset representative in xCX(Y).
For y,z∈Y,
[TABLE]
and, for y∈Y and ℓ∈X, as [x,ℓ]∈CR(Y), xℓ=xc for some c∈CX(Y), and so
[TABLE]
Thus θ is an X-invariant homomorphism from Y to [Y,x]. As kerθ=CY(x), we have Y/CY(x)≅[Y,x] as X-groups.
∎
Lemma 2.5**.**
Assume that p is a prime, X is a group, Y is an abelian normal p-subgroup of X and R is a normal p-subgroup of X which contains Y. Suppose that Y=[Y,Op(X)], [R,Op(X)]≤CR(Y) and R acts quadratically or trivially on Y. Suppose that no non-central X-chief factor of Y/CY(R) is isomorphic to an X-chief factor of [Y,R]. Then Y≤Z(R).
Proof.
Assume that R>CR(Y). Using [R,Op(X)]≤CR(Y), we may select x∈R∖CR(Y) such that xCX(Y)∈Z(X/CX(Y))#. As Y is abelian, [Y,x]≤Z(Y) and so \freflem:abstracting 1.4 applies to give Y/CY(x)≅[Y,x] as X-groups. As R acts quadratically on Y,
[TABLE]
and so the hypothesis on non-central X-chief factors now gives Y/CY(x) and [Y,x] only have central X-chief factors. In particular, Y=[Y,Op(X)]≤CY(x) and this contradicts the initial choice of x∈R∖CR(Y). Hence Y≤Z(R).
∎
Lemma 2.6**.**
Suppose that p is a prime, X is a group, V≤U are normal p-subgroups of X, and Q is a large p-subgroup of X which is not normal in X.
Assume that V is a non-trivial irreducible GF(p)X-module and U/V is centralized by Op(X).Then
(i)
U* is elementary abelian; and*
2. (ii)
if U≤Ω1(Z(Op(X))), then Op(X)/COp(X)(U) contains a non-central chief factor isomorphic to V as a GF(p)X-module.
Proof.
Set ZX=Ω1(Z(Op(X))). We have [U,Op(X)]≤V≤ZX as V is irreducible. As Op(X) does not centralize U/Φ(U) by Burnside’s Lemma [GLS2, Proposition 11.1] and V is a non-trivial irreducible X-module, V≤Φ(U) and Φ(U) is centralized by Op(X). Therefore Φ(U)∩ZX is centralized by Op(X) and is normalized by Q. Since Q is large and Op(X)≤NX(Q), we deduce Φ(U)∩ZX=1. Thus Φ(U)=1 and so U is elementary abelian. Hence (i) holds.
Set Y=Op(X) and assume that U≤ZX. Select x∈U∖ZX such that [X,x]≤U∩ZX≤Z(Y).
Then xCX(Y)∈Z(X/CX(Y)). Thus \freflem:abstracting 1.4 implies
Y/CY(x)≅[Y,x]≤U∩ZX and this isomorphism is as X-groups.
Since [Y,x] is normalized by Q, [Y,x]=1 and Q is large, Op(X) does not centralize [Y,x]. Thus [Y,x]≥V as [U,Op(X)]≤V. This proves (ii).
∎
Lemma 2.7**.**
Assume that p is a prime, X is a group, U is an elementary abelian normal subgroup of X, U=[U,Op(X)] and Op(X) acts quadratically and non-trivially on U. Set R=Op(X), W=R/CR(U), and Z=[U,R]. Then W, U/Z and Z are X/R-modules and W is isomorphic to an X/R-submodule of Hom(U/Z,Z). In particular, if Z is centralized by X, then the set of X-chief factors of W can be identified with a subset of the GF(p)-duals of the X-chief factors of U/Z.
Proof.
Since R acts quadratically on U, W is elementary abelian. Furthermore, R centralizes W, U/Z and Z. Hence all of these groups can be regarded as GF(p)X/R-modules. For w∈R, define
[TABLE]
The calculation in the proof of \freflem:abstracting 1.4 shows that the commutator [u,w] defines a homomorphism from U to Z and, as w centralizes Z, θw is a well-defined homomorphism from U/Z to Z. Thus θ is a well-defined map. Consider w1,w2∈R, uZ∈U/Z and ℓ∈X. Then
[TABLE]
which means θw1w2=θw1θw2 and so θ is a group homomorphism. We show that θ is an X-module homomorphism. So let ℓ∈X, uZ∈U/Z and w∈R. Then
(wℓ)θ=θwℓ and
[TABLE]
Since kerθ=CR(U), this completes the proof of the main claim.
If Z is centralized by X, then
[TABLE]
where n is such that ∣Z∣=pn. This completes the proof of the lemma.
∎
Lemma 2.8**.**
Suppose that V is a p-group and X is a group which acts faithfully on V with Op(X)=1. Assume A≤X is an elementary abelian p-subgroup of order at least p2
which has the property CV(A)=CV(a) for all a∈A#. If L is a non-trivial subgroup of X and L=[L,A], then A acts faithfully on L.
In particular, A centralizes every p′-subgroup which it normalizes,
[A,F(X)]=1, E(X)=1 and, if L is a component of X which is normalized but not centralized by A, then A acts faithfully on L.
Proof.
Suppose that L=[L,A] is a non-trivial subgroup of X. Assume that there is b∈A# with [L,b]=1. Then L normalizes CV(b) and so, as CV(b)=CV(A), L=[L,A] centralizes CV(b). Since L=[L,A], L=Op(L) and the Thompson A×B-Lemma implies [L,V]=1, a contradiction. Hence A acts faithfully on L.
Let F be a p′-subgroup of X which is normalized by A. Then F=⟨CF(a)∣a∈A#⟩. If A does not centralizes F, then there exists a∈A# such that 1=[CF(a),A]=[CF(a),A,A]. Hence, taking L=[CF(a),A], we have L=[L,A] and a∈CA(L), a contradiction. Hence [F,A]=1. Now A centralizes F(X) and therefore E(X)=1.
If L is a component of X which is normalized by A, then either [L,A]=L or [L,A]=1. If [L,A]=1, then we have A acts faithfully on L.
∎
Lemma 2.9**.**
Let X be a group, N a normal subgroup of G and T∈Sylp(X). Assume that X=NT, CT(N)=1, q=pa and
[TABLE]
where Ni≅SL2(q) for 1≤i≤s. Then the p-rank of G is sa.
Proof.
Assume first that q=2. Then T acts faithfully on O3(N). As the 2-rank of GLs(3) is s, we are done. Similarly, if q=3, then T acts faithfully on O2(N)/Z(N), which is elementary abelian of order 22s we are done as GL2s(2) has 3-rank s.
Thus we may assume that q>3. In particular, the subgroups Ni are quasisimple and T permutes the set {Ni∣1≤i≤s}.
Assume that p is odd. Let A be an elementary abelian subgroup in T of maximal rank and assume that A≤N. Then by Thompson replacement [GLS2, Theorem 25.2] we may assume that A acts quadratically on T∩N. This shows that A has to normalize each Ni. As non-trivial field automorphisms are not quadratic on T∩Ni, we get that A centralizes T∩N and so A≤T∩N, the assertion.
Assume that q=2a with a≥2. Let B=NN(T∩N). We have that T normalizes B and T/(T∩N) acts faithfully on B/(T∩N). Now the Thompson dihedral Lemma [GLS2, Lemma 24.1] says that for any elementary abelian subgroup A of T we have a B-conjugate Ag such that U=⟨A,Ag⟩(T∩N)/(T∩N) is a direct product of r dihedral groups where 2r=∣A/(A∩N)∣≤2s and A(T∩N)/(T∩N) is a Sylow 2-subgroup of U. Set T1=[O2′(U),T∩N]. As U is generated by two conjugates of A we see that ∣T1∣=∣CT1(A/A∩N)∣2. This now shows that ∣A∣≤∣T∩N∣, the assertion again. This proves the lemma.
∎
In the next two lemmas we use the notation presented in the introduction though we do not assume that L is unambiguous.
Lemma 2.10**.**
Suppose that L∈LG(S), L≤NG(Q) and VL=[YL,L∘]. Then
(i)
CYL(L∘)=1.
(ii)
Ω1(Z(S))≤VL.
(iii)
If VL is an irreducible L-module, VL≤Q and Ω1(Z(QL))<QL, then VL≤QL′≤Φ(QL).
Proof.
As CYL(L∘)≤CG(Q) is normalized by L, (i) is a consequence of Q being large.
By [MSS2, Lemma 1.24 (g)], Ω1(Z(S))≤YL now Gaschütz Theorem [GLS2, Theorem 9.26] and (i) give (ii).
Assume that N is a non-trivial normal p-subgroup of L. Then Ω1(Z(S))∩N=1. Since VL is irreducible as a L-module, (ii) gives VL≤N. In particular, as VL≤Q, N≤Q.
Suppose that QL is abelian. Then, as Q=Op(NG(Q)) and [Q,QL,QL]≤QL′=1, QL is quadratic on Q, and hence QLQ/Q is elementary abelian and so Φ(QL)≤Q. By the remark earlier taking N=Φ(QL) we obtain Φ(QL)=1, contrary to Ω1(Z(QL))<QL. Hence QL is non-abelian. Thus QL′=1 and so, as VL is irreducible, VL≤QL′≤Φ(QL). This proves (iii).
∎
Lemma 2.11**.**
Suppose that L∈LG(S), L≤NG(Q) and VL=[YL,L∘]. Assume that YL=Ω1(Z(QL)), m∈L and Op(L)QL≤KQL, where K=⟨W,Wm⟩. Then Op(L)≤K and the following hold
(i)
[Op(L),QL]≤[W,QL][Wm,QL]≤(W∩QL)(Wm∩QL)=UL.**
2. (ii)
If [W,W]≤VL, then W acts quadratically on the non-central chief factors of QL/VL.
Assume, in addition, that VL is irreducible as a K-module, [VL,W,W]=1,
and [W,W]≤VL. Then
(iii)
W∩Wm∩QL≤YL;
2. (iv)
UL/YL* is elementary abelian or trivial;
and*
3. (v)
either QL=YL or UL′≥VL.
Proof.
Since W and Wm are normalized by QL, K=⟨W,Wm⟩ is normalized by QLK and so Op(L)≤K.
Since W, Wm, [QL,W] and [QL,Wm] are normalized by QL, we have
[TABLE]
In particular, A=(W∩QL)(Wm∩QL) is normalized by Op(L). Since (W∩QL)L=(W∩QL)SOp(L)=(W∩QL)Op(L), we have A=UL.
Thus (i) holds.
By the additional hypothesis,
[TABLE]
and so W acts quadratically on all the non-central L-chief factors in QL/VL, which is (ii).
Notice that part (ii), VL irreducible as a K-module and [VL,W,W]=1 together imply that the non-central K-chief factors in QL/VL are not isomorphic to VL.
Set I=W∩Wm∩QL. Then I≤W∩Wm and so
[TABLE]
and
[TABLE]
Hence IVL/VL is centralized by ⟨W,Wm⟩=K.
As W acts quadratically on all the non-central chief factors of K in QL/VL by (ii) and by assumption, W does not act quadratically on VL, \freflem: U¿YL abelian implies that I≤Ω1(Z(QL))=YL. This proves (iii).
Since W is generated by elements of order p, W/[W,W] is elementary abelian and therefore, as [W,W]≤VL, WVL/VL is also elementary abelian. Since W∩QL and QL∩Wm normalize each other parts (i) and (iii) give (iv).
If VL≤UL′ and QL=YL, then, as UL/YL is elementary abelian by (iv), \freflem:VL sub QL’ (ii) implies UL is elementary abelian. Select E with QL≥E>VL of minimal order such that E=[E,Op(L)] and E/VL has a non-central K-chief factor. Then
[TABLE]
Furthermore, VL[E,QL]<E and so [[E,QL],Op(L)]≤VL. Therefore \freflem: U¿YL abelian implies that [E,QL]≤YL and so QL acts quadratically on E.
Hence \freflem:isochieffactors implies that E≤YL, a contradiction. Hence UL′ is non-trivial and it follows that VL≤UL′.
∎
3. The reduction
We use the notation presented in the introduction. For the rest of this article we have L∈LG(S) with Q not normal in L and L is in the unambiguous wreath product case.
This means that YL=VL unless we are in the special case that L∘≅SL2(4) or ΓSL2(4), ∣YL:VL∣=2 and
[TABLE]
We start with a general result which just requires VL≤Q.
Lemma 3.1**.**
The following hold.
(i)
⟨VLD⟩* is not a p-group;*
2. (ii)
[Q,⟨VLD⟩]≤W; and
3. (iii)
W≤CG(VL).
Proof.
Let C~=NG(Q) and K=⟨VLC~⟩. As D=KNL(Q) and NL(Q) acts on VL we have ⟨VLD⟩=⟨VLK⟩ is subnormal in H. If ⟨VLD⟩ is a p-group, we obtain VL≤Op(NG(Q))=Q which is a contradiction. This proves (i).
We have [Q,VL]≤Q∩VL≤W. As W and Q are normalized by D, (ii) holds.
Assume W≤CG(VL). Then [W,VL]=1 and so [W,⟨VLD⟩]=1. Hence X=Op(⟨VLD⟩) centralizes Q by (ii). Since CG(Q)≤Q, we have X≤Q. Thus X=1 and ⟨VLD⟩ is a p-group, which contradicts (i). Hence W≤CG(VL).
∎
We adopt the following notation.
Let B≥CL(VL) be such that B=⟨K⟩ and let S0=S∩B.
We write B=K1…Ks where Ki≥CL(VL), Ki∈K, Ki≅SL2(q) and, for 1≤i≤s, put
[TABLE]
[TABLE]
[TABLE]
and
[TABLE]
We begin by showing that W is not contained in the base group B.
Lemma 3.2**.**
Suppose that L is either properly wreathed, or q=pa (where p divides a) and some element of L∘ induces a non-trivial field automorphism on Op(L∘)≅SL2(q). Then W is not contained in S0. In particular, if L is properly wreathed with q=s=2, then Q is not cyclic of order 4.
Proof.
Set F=⋂g∈DCQ(VL)g.
Suppose that W is contained in S0. As Q normalizes W and acts transitively on K when L is properly wreathed and, as VL is the natural SL2(q)-module when s=1, and field automorphisms are present, the structure of VL yields
[TABLE]
Suppose that g∈D. Then using \freflem:w1(ii) and (VL)g=VLg yields
(3.2.1) **[Z0,[VLg,Q]]≤[Z0,W]=1.
We also remark that as W≤Q, Z0≤[VL,Q]≤W=Wg≤S0g and Z0≤Z(W). In particular, as S0g
normalizes every element of Kg, so does
Z0. Therefore, for 1≤i≤s, Z0 also normalizes each Kig and so also [YLg,Kig]=(VLi)g.
If s=1 and we have field automorphisms in L∘, then [VL,Q]>Z0 and so \frefclm:notbase1 provides Z0≤CQ([VLg,Q])=CQ(VLg). Thus
[TABLE]
in this case.
We will show that the same holds in the properly wreathed case.
Because Q acts transitively on Kg,
[TABLE]
As [Z0,[VLg,Q]]=1 by \frefclm:notbase1,
[TABLE]
Hence Z0≤CQ(VLg) and this implies that
[TABLE]
in the properly wreathed case too.
Therefore,
[TABLE]
Hence VL stabilizes the normal series Q≥W≥W∩F≥1 in D and so VL≤Op(D). But then ⟨VLD⟩ is a p-group contrary to \freflem:w1 (i). We conclude that W≤S0 as claimed.
If q=s=2 and Q is cyclic of order four, then, as W is generated by involutions, W=Q∩S0, a contradiction. Thus Q is not cyclic of order 4 in this case.
∎
We now reduce the properly wreathed case to one specific configuration which will be handled in detail in \frefsec:o4.
Proposition 3.3**.**
Assume that L is properly wreathed and unambiguous. Then ∣K∣=2, q=2, and W permutes K transitively by conjugation. Furthermore, Q=S≅Dih(8), L∘≅O4+(2) and YL=VL is the natural O4+(2)-module.
Proof.
Since Q permutes K transitively by conjugation and S0 normalizes Q, we have
As W=⟨VLg∩Q∣g∈D⟩, \freflem:not base implies there exists g∈D such that VLg∩Q≤S0. We fix this g.
(3.3.2) ** We have VLg∩Q∩S0=1.
Suppose that VLg∩Q∩S0=1. Then, as Q∩S0 and VLg∩Q normalize each other, VLg∩Q centralizes Q∩S0.
If VLg∩Q normalizes some Ki∈K, then, as Q acts transitively on K and normalizes VLg∩Q, VLg∩Q normalizes every member of K.
As VLg∩Q centralizes [Q,S0], 3 (iii) implies that
[TABLE]
Since Q acts transitively on K, this is true for each Ki∈K.
Thus
[TABLE]
which contradicts the choice of g∈D.
Hence VLg∩Q does not normalize any member of K. As B is a direct product we calculate that CS0(VLg∩Q) has index at least qp−1 in S0. However \frefclm:index [Q,S_0] (ii) states that Q∩S0 has index at most q in S0 and, as this subgroup is centralized by VLg∩Q, we deduce that
[TABLE]
Furthermore, as VLg∩Q does not normalize any member of K, if s>2, we have CS0(VLg∩Q) has index at least q2 in S0, and so we must have
[TABLE]
Since VLg∩Q centralizes [S0,Q] by \frefclm:index Q,S_0, no element in VLg∩Q can act as a non-trivial field automorphism on K1 and so we infer from VLg∩Q∩S0=1, that ∣VLg∩Q∣=2. In particular, ∣CVL(VLg∩Q)∣=q2 as VLg∩Q exchanges VL1 and VL2.
We know that ∣VLg∣=q4. As ∣[VLg,Q]∣≥q3, we have
[TABLE]
and we have just determined that
[TABLE]
Hence VLg∩Q∩CG(VL) has order at least 23a−1, where q=2a.
Assume that a=1. Then, as VLg1 has order q2,
[TABLE]
It follows that VL∩Q normalizes both K1g and K2g. As (VL∩Q)CLg(VLg)/CLg(VLg) is normalized by Q and Q permutes {K1g,K2g} transitively, (VL∩Q)CLg(VLg)/CLg(VLg) does not centralize Kig/CLg(VLg) for i=1,2. Thus ∣CVLgi(VL∩Q)∣≤q for i=1,2. But then
[TABLE]
which contradicts a=1. We conclude that q=s=2 and ∣VLg∩Q∣=2. Furthermore, VLg∩Q is centralized by Q and so Q is elementary abelian of order 4. It follows that L∘≅Ω4+(2) and VL is the natural module. Hence L is ambiguous and we conclude that VLg∩Q∩S0=1.\hfill■
(3.3.3) ** We have
∣CVL(VLg∩Q)∣≤qs/p.
We know VLg∩Q≤S0 and VLg∩Q∩S0=1 by \frefclm:meet base. As VLg∩Q is normalized by Q, VLg∩Q∩S0=1 implies that
[TABLE]
If some element d∈VLg∩Q induces a non-trivial field automorphism on Ki for some Ki∈K, then
CVLi(VLg∩Q)≤CZi(d) has order q1/p and the result follows by transitivity of Q on K. On the other hand, if d∈VLg∩Q has an orbit of length p on K, then C⟨(VL1)⟨d⟩⟩(VLg∩Q)≤C⟨Z1⟨d⟩⟩(d) which has order q. Using the transitivity of Q on K, we deduce ∣CVL(VLg∩Q)∣≤qs/p. This proves the result.
\hfill■
As Q acts transitively on the {Vi∣1≤i≤s}, we have VL=[VL,Q]V1. By \frefclm:meet base Q∩S0=1 and so ∣[V1,Q]∣≥q. In particular
[TABLE]
Since VL∩Q∩CLg(VLg)≤CVL(VLg∩Q), \frefclm:cent bound and ∣VL∣=q2s together give
[TABLE]
On the other hand, by \freflem:prank the p-rank of L is as where q=pa. Hence
[TABLE]
and so
[TABLE]
In particular, \freflem:prank implies
(3.3.4) ** ∣(VL∩Q)CLg(VLg)/CLg(VLg)∣=q2=22a.
Assume that q>2. Since Sg/S0g has 2-rank 2 and VL∩Q is elementary abelian, (VL∩Q∩S0g)CLg(VLg)/CLg(VLg) has rank at least 2a−2=1.
Since VL∩Q∩S0g is normalized by Q and Q permutes {K1g,K2g} transitively, VL∩Q∩S0g contains an element which projects non-trivially on to both S1gCLg(VLg)/CLg(VLg) and S2gCLg(VLg)/CLg(VLg).
Thus VL≥[VL∩Q,[VLg,Q]]≥Z0g. But then, using \frefclm:cent bound yields the contradiction
[TABLE]
Thus q=s=2.
It follows from \freflem:not base that W is transitive on K and Q≅Dih(8) or Q is elementary abelian of order 4. The second possibility gives L∘≅Ω4+(2), which is ambiguous. This proves \frefprop:notwreath.
∎
Next we deal with the case s=1.
Proposition 3.4**.**
Suppose that Op(L∘)≅SL2(q) where q=pa=rp, VL=YL is the natural Op(L∘)-module and that some element of L∘ induces a non-trivial field automorphism on Op(L∘). Then p=2=r.
Proof.
We may assume that rp>4. By \freflem:not base we have that W≤S0 and, as W is generated by elements of order p, we have that ∣S0W:S0∣=p. As Q is normal in S, 1=Q∩S0, so Z0≤Q∩YL. Furthermore, as Q contains elements which act as field automorphisms on Op(L∘),
[TABLE]
by assumption. Thus no element in S∖QL centralizes a subgroup of index p in VL∩Q.
Set W1=⟨Z0D⟩. As Z0 centralizes W∩S0, every element of Z0 centralizes a subgroup of index at most p in W. As W1 is generated by conjugates of Z0, and these conjugates all contain elements which centralize a subgroup of index at most p in W, W1 is generated by elements which centralize a subgroup of index at most p in VL∩Q. As no element in S∖QL has this property, we conclude that W1≤QL. Hence [VL,W1]=1. In particular [VL∩Q,W1]=1 and so also [W,Z0]=[W,W1]=1. This shows W≤S0 and contradicts \freflem:not base.
∎
We collect the results of this section in the following proposition:
Proposition 3.5**.**
Suppose that L∈LG(S), L≤NG(Q), VL≤Q and L is in the unambiguous wreath product case. Then one of the following holds:
(i)
L∘≅O4+(2)* , Q=S≅Dih(8) and YL=VL is the natural module.*
(ii)
L∘≅ΓSL2(4), VL is the natural SL2(4)-module and ∣YL:VL∣≤2.
(iii)
L∘≅SL2(4), VL is the natural module and ∣YL:VL∣=2.
Proof.
If ∣K∣>1, then (i) holds by \frefprop:notwreath, so we may assume that ∣K∣=1. As L is unambiguous, either YL=VL or L∘≅SL2(q). If YL=VL, then by definition of the wreath product case, (ii) or (iii) holds. So we may assume YL=VL and L∘≅SL2(q). Now (ii) holds by \frefprop:fieldauto.
∎
4. L∘≅O4+(2)
In this section we analyse the configuration from \frefprop:unambiguous(i). We prove
Proposition 4.1**.**
Suppose that L∈LG(S), L≤NG(Q) and L in the unambiguous wreath product case. If YL≤Q and L∘≅O4+(2), then G≅Sym(8), Sym(9) or Alt(10).
Proof.
By \frefprop:unambiguous we have Q≅Dih(8). Since YL is the natural O4+(2)-module for L/CL(YL) and VL is also the wreath product module for L/CL(YL) with respect to {K1,K2}, we have the following well known facts.
(4.1.1) **
(i)
∣[YL,Q]∣=23, ∣[YL,Q,Q]∣=22 and CYL(Q)=[YL,Q,Q,Q] has order 2.
2. (ii)
[YL,S0]=CYL(S0) has order 22;
3. (iii)
∣[YL,Q′]∣=22;
4. (iv)
CL([YL,Q])≤CL(YL).
Our first aim is to prove
(4.1.2) ** W is elementary abelian of order 22, [YL,W]=[YL,Q]=YL∩Q and [YL,W,W]=CYL(W)=CYL(Q)=Z.
Applying \freflem:w1, we consider x∈D such that YLx∩Q≤CL(YL). Then YLx∩Q is normalized by Q and so
[TABLE]
In particular, \frefclm:YLprops(iii) gives
[TABLE]
As YL is elementary abelian, YLx∩Q is elementary abelian.
Suppose that [YL,YLx∩Q,YLx∩Q]=1. Then
[TABLE]
by \frefclm:YLprops (iv). Hence [YL,YLx∩Q,YLx]=1. Then as ∣[YL,YLx∩Q]∣=22 and ∣YL∩Q∣=23, we conclude that (YL∩Q)CLx(YLx)/CLx(YLx) has order 2. Thus [YLx,YL∩Q,YL∩Q]=1. Now the argument just presented implies that ∣YLx∩Q∣=2 and so, as Q normalizes YLx∩Q, YLx∩Q=Z(Q).
In particular, as [YL,S0,S0]=1, we have proved that
if YLx∩Q≤S0, then YLx∩Q=Z(Q).
For a moment let Q1 be the fours subgroups of Q not equal to S0.
Then as Φ(YLx∩Q)=1 the displayed line implies that W≤Q1 and \freflem:not base and Q′≤YLx∩Q imply W=Q1. The remaining statements in \frefclm:W4 now follow from the action of L on YL.\hfill■
We have that Z(Q) centralizes [YL,Q] and so Z(Q)≤S∩CL(YL)=QL. Hence
using \frefclm:W4 we obtain
[TABLE]
(4.1.3) **
We have QL=YL.
Suppose that QL>YL. Let m∈L be such that K≅SL2(2)×SL2(2), where K=⟨W,Wm⟩. Recall that by the choice of L in the Notation at the end of the introduction, we have YL=Ω1(Z(QL)) and by \frefprop:unambiguous and \frefclm:W4, K acts irreducibly on YL=VL. Hence we may apply
\freflem:[WO_p][W^mO_p] (iii), (iv) and (v) which combined yield UL/YL is elementary abelian and
[TABLE]
Since [QL,W,W]≤[W,W]=Z≤YL, we have W acts quadratically on every chief factor of L in QL/YL. In particular, no non-central L-chief factor of QL/YL is isomorphic to YL.
Let E be the preimage of CUL/YL(K). Then E is normal in L and application of \freflem: U¿YL abelian implies that E=YL. Let X∈Syl3(K). By \freflem:[WO_p]W^mO_p, [K,CL(YL)]≤UL, so XUL is normal in L. As L is solvable, CL(YL)=CX(YL)QL and either CX(YL)=1 or X≅3+1+2. The latter case is impossible as W is quadratic on UL/YL. Hence UL=[UL,O2(L)] and UL/YL contains no central L-chief factors. We know that every L-chief factor in UL/YL is a wreath product module for SL2(2)≀2 with W acting quadratically. In particular, for every non-central chief factor F of L in UL/YL we have [F,W]=[F,Z(Q)]. Set W1=[W,D]. Then
[TABLE]
Hence [F,W]=[F,W1] for every non-central chief factor F of L in UL/YL. Set L=L/YL and let z∈Q with Z(Q)=⟨z⟩. As CF(Z(Q))=[F,Z(Q)] for each F, we have CUL(z)=[UL,z]; then as W acts quadratically on UL, we have
[W,UL]=CUL(W). Thus [UL,W]YL=[UL,W1]YL.
In particular,
[TABLE]
and so UL acts quadratically on W/W1. Therefore ULCD(W/W1)/CD(W/W1) is elementary abelian. Hence
[TABLE]
Set R=⟨YLD⟩. Then, as YL≤O2(D) by \freflem:w1 (i), YL∩O2(D)=YL∩Q≤W and so R centralizes O2(D)/W and W/W1.
\freflem: cent W/[W,Q] yields YL≤O2(D) and this contradicts \freflem:w1 (i). We have shown QL=YL. \hfill■
(4.1.4) ** ∣S∣=27 and NG(Q)/Q≅Sym(3).
Since QL=YL=VL and Q≅Dih(8), ∣S∣=27 and ∣Q∣=26. Then NG(Q)=SX, where X is a Hall 2′-subgroup of NG(Q) and QX is normal in NG(Q). Furthermore W is extraspecial of order 25. As W/Z=J(Q/Z), we have W is normal in NG(Q). Hence X acts faithfully on W and embeds in O4+(2). As [W,Q]=Z(Q), S/W is faithful on W/Z, so NG(Q)/W embeds into O4+(2). Because O4+(2)≅Sym(3)≀2, and O2(NG(Q)/W)=1, we get the claim. \hfill■
Taking T∈Syl3(L), we have NL(T) is a complement to QL and so L=QLNL(T) is a split extension of QL by O4+(2). In particular, the isomorphism type of S is uniquely determined. As Sym(8) has a subgroup isomorphic to L and Sym(8) has odd index in Alt(10), we have S is isomorphic to a Sylow 2-subgroup of Alt(10).
Let z∈CYL(Q)#, then as YL is a +-type space for L, there is a fours group A of YL which has all non-trivial elements L-conjugate to z. Since CG(z) has characteristic 2, CO(G)(z)=1 and so by coprime action
[TABLE]
Assume that G has no subgroup of index two. Then S is isomorphic to a Sylow 2-subgroup of Alt(10). Therefore [Mas, Theorem 3.15] implies that F∗(G)≅Alt(10), Alt(11), PSL4(r), r≡3(mod4), or PSU4(r), r≡1(mod4). Notice that Z(Q)=CYL(Q)=⟨z⟩ and so CG(z)=NG(Q) has characteristic 2. In Alt(11), z corresponds to (12)(34)(56)(78) and so CG(z)≤(Alt(8)×Z3):2, which implies that CG(z) is not of characteristic 2. In the linear and unitary groups CG(z) has a normal subgroup isomorphic to SL2(r)∘SL2(r), and this contradicts \frefclm:NGQ. Hence G≅Alt(10).
Assume now that G has a subgroup of index two. As VL≤G′ we also have W≤G′. Therefore (G′∩L)/YL≅Ω4+(2) and so G′ has Sylow 2-subgroups isomorphic to those of Alt(8). Applying [GH, Corollary A*] we have F∗(G)≅Alt(8), Alt(9) or PSp4(3). Again in G′≅PSp4(3), we have that G′ contains a subgroup of shape SL2(3)∘SL2(3). This contradicts \frefclm:NGQ and proves the proposition.
∎
5. L∘≅ΓSL2(4)
In this section we attend to the case from \frefprop:unambiguous(ii). Hence we have p=2, L∘≅ΓSL2(4), VL is the natural SL2(4)-module and either YL=VL or ∣YL/VL∣=2. Notice that as L≤NG(Q) and L centralizes YL/VL, if YL>VL, YL does not split over VL and CYL(Q)=CVL(Q) has order 2. Furthermore, CS([YL,Q])=QL.
Our aim is to prove
Proposition 5.1**.**
Suppose L∈LG(S) and L≤NG(Q) with L in the unambiguous wreath product case. If YL≤Q and L∘≅ΓSL2(4), then G≅Mat(22) or Aut(Mat(22)).
Notice that as QL∈Syl2(CL(YL)), CL(YL)/QL is centralized by L∘, and so CL∘(YL)=QL∩L∘ as the Schur multiplier of SL2(4) has order 2. We also have ∣Q∣≥4 and ∣Z(Q)∩VL∣=2.
Lemma 5.2**.**
For N=NG(QL) we have (Z(Q)∩VL)N∩YL⊆VL. In particular, N normalizes VL.
Proof.
If VL=YL, there is nothing to prove. Assume that ∣YL:VL∣=2. Choose g∈N, put U=(Z(Q)∩VL)g and assume that U≤VL. Recall that YL=Ω1(Z(QL)) and so U≤YL and YL is normalized by N. Then CL(U)CN(YL)/CN(YL)≅5:4 or 2×Sym(3). As CN(Ug−1) normalizes Q∩YL, CN(Ug−1) is not irreducible on YL/Ug−1. This excludes the possibility CL(U)CN(YL)/CN(YL)≅5:4 which is irreducible on YL/U. Hence we see that Z(Q)∩VL has exactly 15+10=25 conjugates under N, but 25 does not divide the order of SL5(2)=Aut(YL). This contradiction proves the lemma.
∎
Lemma 5.3**.**
We have QL=YL and either
(i)
∣S∣=27, L/QL≅ΓSL2(4), NG(Q)/Q≅SL2(2), there exists a subgroup E≤S of order 24 which is normalized by NG(Q) such that NG(E)/E≅Alt(6) and NL(E) has index 5 in L. Furthermore all the involutions in ⟨NG(E),L⟩ are conjugate.
2. (ii)
G* has a subgroup of index 2 which satisfies the conditions in *(i).
Proof.
We have S≅Dih(8) and Q≤S0 as L∘≅ΓSL2(4). \freflem:not base implies that W≤S0. By assumption, we either have YL=VL or ∣YL:VL∣=2. In particular, 24≤∣YL∣≤25. Since Q is normal in S and contains W we know
(5.3.1) ** Either
Q is elementary abelian of order 4 or Q=S
As VL is a natural SL2(4)-module and L≤NG(Q), we have CYL(Q)=CYL(S) has order 2 and [YL,Q]=[VL,Q] has order 8. Furthermore, as W is normal in S and is not contained in S0, we have [YL,Q,W]=Z where Z=CVL(S) has order 2. Thus, arguing exactly as before \frefclm:qlyl and in the proof of \frefclm:W4 we obtain
(5.3.2) ** ∣W∣=4, [W,W]=Z and [QL,W,W]≤YL.
(5.3.3) **
Assume that QL>YL. Then [QL,O2(L)]≤YL.
Suppose that [QL,O2(L)]≤YL. Then VL≤Φ(QL) by Burnside’s Lemma [GLS2, Proposition 11.1], which contradicts \freflem:VL sub QL’(iii). This proves the claim\hfill■
(5.3.4) ** If VL<YL, then Q=S.
If Q has order 4, then Q=W by \frefclm:S5-W quad, so Q normalizes a Sylow 3-subgroup T of L and so Q normalizes CYL(T) which has order 2 and complements VL. Hence CYL(T)≤Z(Q), so T≤NG(Q) and therefore L=⟨T,S⟩≤NG(Q), a contradiction. Thus Q=S has order 8.\hfill■
(5.3.5) ** We have QL=YL.
Suppose false. By \frefclm:S5-W quad W acts quadratically on QL/YL and ∣W∣=4. Also W≤S0, so \freflem:Sym5-modules implies that the non-central L-chief factors in QL/YL are orthogonal modules for L≅O4−(2). In particular, as L-modules, the non-central L-chief factors of QL/YL are not isomorphic to VL.
Choose E≤QL normal in L and minimal so that E/YL contains a non-central L-chief factor and let F be the preimage of CE/YL(O2(L)). Then [F,O2(L)]≤YL and \freflem: U¿YL abelian applies to yield F≤YL. In particular, [E,E]≤YL.
We claim E′≤VL. This is obviously the case if VL=YL. So suppose that ∣YL:VL∣=2. If E′≤VL. Then the minimal choice of E and E′VL=YL implies that E/VL is extraspecial of order 25. Notice that [E,W]≤W and W/Z is elementary abelian as [W,W]=Z by \frefclm:S5-W quad. Hence, as [E,W]YL/VL has order 23, we infer that E/VL has +-type contrary to L≅ΓSL2(4). Hence E/VL is elementary abelian. If [QL,E]=VL, then E/VL has order 24 by \freflem:Sym5-modules and so QL/CQL(E) embeds into
[TABLE]
by \freflem:dualchieffactors. Since QL/CQL(E) involves only trivial and orthogonal modules this contradicts [Pr, Lemma 2.2].
Thus [E,QL]=YL>VL.
By \frefclm:S5-V_LnotY_L
Q=S has order 8.
In summary we now know ∣W∣=4 and [W,Q]=[W,S]=Z(S).
We calculate using Z is normal in D by \frefclm:S5-W quad that
[TABLE]
Therefore
[TABLE]
As ∣[Z(S),E/YL]∣=4 and Q=S, this implies that ∣CE/YL(S)∣=4.
As E/YL is the orthogonal O4−(2)-module for L, this is impossible. We have proved the claim.
\hfill■
(5.3.6) **
Suppose that YL=VL. Then L is a maximal 2-local subgroup of G, NG(Q)/Q≅SL2(2), there exists a subgroup E≤S of order 24 which is normalized by NG(Q) such that NG(E)/E≅Alt(6) and NL(E) has index 5 in L.
By \frefclm:Q_L=Y_L we have ∣S∣=27, and ∣W∣=22. Also ∣[W,YL]∣=8 and YL≤Q, so Q∩YL=[W,YL]=W∩YL, Therefore ∣W∣=25. Set C=CQ(W). Then C centralizes [YL,Q] which has order 23 and so C≤CL([YL,Q])=YL. Thus C≤CYL(W) which has order 2. Then, by \frefclm:S5-W quad, W′=Z=C and, as W is generated by involutions, we have W is extraspecial. Since [YL,Q]≤W, W has +-type.
Observe W/Z=J(Q/Z), so W is normal in NG(Q) and NG(Q)/Z embeds into Aut(W)≅24:O4+(2).
Assume that YLQ/Q normalizes a subgroup T of O3(NG(Q))/Q which has fixed points on W/Z. Then W=[W,T]CW(T) and [W,T]≅CW(T)≅\mboxQ8 and these subgroups are normalized by YL. But then
[TABLE]
Since [W,YL] is elementary abelian and Ω1(P)=Z(P) for P≅\mboxQ8, we conclude that
[TABLE]
and then [W,YL] has order 2 which is nonsense as YL is the natural module. Therefore YL normalizes no such subgroup.
Let F=O2,3(NG(Q)). Assume that ∣F/Q∣=9. Then the previous argument implies that CF/Q(YL)=1. Let T1 be the preimage of this subgroup. Then [YL,Q] is normalized by T1. Hence YL=CYLQ([YL,Q]) is normalized by T1. Using the fact that Q is weakly closed in any 2-group which contains it, for w∈YL#, we let Qw be the unique conjugate of Q in O2(CG(w)). Then T1 permutes the elements of YL and so T1 normalizes L∘=⟨Qw∣w∈YL#⟩. Since L=L∘YL, we have that T1 normalizes L.
On the other hand,
WYL is normalized by T1 and, as T1 acts fixed-point freely on W/Z, T1 acts transitively on WYL/YL≅W/[YL,Q]≅22 and this is impossible as W∩O2(L) is a maximal subgroup of W and is normalized by T1.
Hence ∣F/Q∣=3, NG(Q)=FS and NG(Q)/Q≅SL2(2). In particular, ∣Q∣=26, S=YLQ, and FYL/W≅2×SL2(2). It follows that
[TABLE]
et E=CS([W,Q]). As W is normal in NG(Q), so is E. As ∣S∣=27 and ∣GL3(2)∣2=23, we have ∣E∣≥24. Since F acts fixed-point freely on W/Z (being normalized by YL), we have E≤Q and then E is normal in NG(Q). Since E∩W=[W,Q], we find ∣E∣=24. Let S≤L1≤L be such that L1/QL≅Sym(4) has index 5 in L. Notice that O2(L1)=S0. Then E≤CL([YL,Q,Q])=YLS0. Also YL≤S0, so S0=YLS0. Therefore E≤S0. Now EYL/YL acts as a Sylow 2-subgroup of SL2(4) on the natural module. In particular for any involution e∈E∖YL we have that CYL(e)=E∩YL. This implies that all involutions in EYL are contained in YL∪E and therefore E and YL are the only elementary abelian subgroups of S0 of order 24. In particular, L1 normalizes E. Now NG(E)≥⟨L1,NG(Q)⟩∈LG(S). Notice that L1 has orbits of lengths 3, and 12 on E and that NG(Q) does not preserve these orbits. Hence NG(E) acts transitively on E#. As NG(Q)=CG(Z), we now have that ∣NG(E)∣=15∣NG(Q)∣=27⋅32⋅5. We have that X=NG(E)/E is isomorphic to a subgroup of GL4(2)≅Alt(8) of order 23⋅32⋅5. We consider the action of X on a set of size 8. As Alt(8) has no subgroups of order 45, X is not transitive. Hence X is isomorphic to a subgroup of Alt(7), Sym(6) or X≅(Alt(5)×3):2. Suppose that X≅(Alt(5)×3):2. As NG(Q)/Q≅Sym(4), we see that EQ/E≤Alt(5). Since E is the natural SL2(4)-module, we get that ∣Z(Q)∣=4. But, by \frefclm:S5-W quad, ∣Z(Q)∣=2. Hence we have one of the first two possibilities and then obviously X=NG(E)/E≅Alt(6).
We just have to show that L is a maximal 2-local subgroup. Let M be a 2-local subgroup with L≤M. As Q≤M, we have that M is of characteristic 2. Then YL=YM and CG(YL)=YL. As ∣NG(Q):S∣=3 and YL is not normal in NG(Q), we have NM(Q)=S=NL(Q). As L acts transitively on YL#, we conclude M=NM(Q)L=NL(Q)L=L.
\hfill■
(5.3.7) ** If YL=VL, then G has just one conjugacy class of involutions.
By \frefclm:structure NG(E)/E≅Alt(6). As YL≤E, there is an involution y∈YL∖E. Now y inverts an element of order 5 in NG(E) and so ∣[E,y]∣=∣CE(y)∣=4. This shows that all involutions in Ey are conjugate. As all involutions in S/E are conjugate in Alt(6) and all the involutions in YL are L-conjugate, this proves the claim.\hfill■
We have now proved that (i) holds when YL=VL.
(5.3.8) **
Suppose that YL>VL. Then G has a subgroup of index 2.
We have that ∣S∣=28. By \frefclm:S5-V_LnotY_L and \frefclm:Q_L=Y_L, S=QYL. We are going to show that J(S)=YL.
For this let A≤S be elementary abelian of maximal order and assume that A=YL. Then ∣AYL/YL∣≤4. As there are no transvections on VL, we get ∣AYL/YL∣=4 and we may assume that A acts quadratically on YL by [GLS2, Theorem 25.2]. As W≤S0 by \freflem:not base and ∣W∣=4 by \frefclm:S5-W quad, W does not act quadratically on YL, AYL/YL≤S0/YL and S0=AYL. Now A∩YL has order 8 and so ∣CYL(S0)∣=8. But (L∘)′ is generated by two conjugates of S0, which gives CYL(L∘)=1 a contradiction to \freflem:VL sub QL’(i).
Thus YL=J(S) is the Thompson subgroup of S. In particular, NG(YL) controls G-fusion of elements in YL.
As S∈Syl2(G) and CS(YL)=QL, QL∈Syl2(CG(YL)) and we have NG(YL)=CG(YL)NNG(YL)(QL). By \freflem:fusioncontrol
[TABLE]
Suppose that O2(L)≥YL. Then O2(L)/VL≅SL2(5) has quaternion Sylow 2-subgroups and ∣L:O2(L)∣=2. On the other hand,
there exists g∈NG(Q)∖NG(YL) with WYL≥(YLg∩Q)YL=YL and (YLg∩Q)VL/VL is elementary abelian, which is a contradiction. Therefore O2(L)/VL≅SL2(4) and, as W does not act quadratically on YL, we see that ∣W:W∩O2(L)∣=2 and thus O2(L)W/VL≅ΓSL2(4). Hence L has a subgroup L0=O2(L)W of index 2 with YL∩L0=VL.
Let T∈Syl2(L0) and w∈YL∖T. Suppose that for some x∈G, wx∈T and ∣CS(wx)∣≥∣CS(w)∣. As L∘ has orbits of length 6 and 10 on YL∖VL, we may assume ∣CS(wx)∣≥∣S∣/2. But then as VL is the natural module, it does not admit transvections and so wx∈VL. As NG(YL)=NG(VL) and NG(YL) controls fusion in YL, this is not possible. Hence the supposed condition cannot hold. Application of [GLS2, Proposition 15.15], shows that G has a subgroup of index 2. This proves \frefclm:56.\hfill■
Let G0 be a subgroup of G of index 2, and set Q0=Q∩G0. We have VL≤L∘≤G0. Hence W=⟨[VL,Q]D⟩≤G0. In particular, W≤Q0 and so Z(Q0)=Z and Q0 is large in G0. Set L0=O2(L)Q0=O2(L)W. Then L0∘/VL≅ΓSL2(4) and YL0=VL0=VL≤Q0. Thus (G0,L0) satisfies the hypotheses of (i). This proves (ii) holds if VL=YL.
∎
Proof of \frefprop:S5-1:.
By \freflem:S5-2 we just have to examine the structure in \freflem:S5-2(i), so we may assume that \freflem:S5-2(i) holds.
By \freflem:splitA6
[TABLE]
As NG(Q)≤NG(E), for a 2-central involution z we have that CG(z) is a split extension of E by Sym(4). As O(CG(z))=1 coprime action yields O(G)=⟨CO(G)(e)∣e∈E#⟩=1. In particular F(G)=1 and E(G)=1. Suppose that J∗ is a non-trivial subnormal subgroup of G normalized by ⟨L,NG(E)⟩. Then S∩J∗=1.
Since 1=J∗∩NG(E) is normal in NG(E) and 1=J∗∩L is normal in L, it follows that J∗∩NG(E)≥J∗∩S≥EYL. Hence J∗≥⟨YLNG(E)⟩=NG(E)≥S and J∗≥⟨SL⟩=L. Therefore there is a unique non-trivial subnormal subgroup of G of minimal order normalized by ⟨L,NG(E)⟩. It follows that ⟨L,NG(E)⟩ is contained in a component J of G. Since O(G)=1 and S≤J, J=E(G).
As J has just one conjugacy class of involutions by \freflem:S5-2(i) and, for z∈E#, CG(z)≤NG(E), it follows that G=J is simple. Using G has just one conjugacy class of involutions and applying [J, Theorem] yields G≅Mat(22). This proves the proposition when \freflem:S5-2(i) holds. If \freflem:S5-2(ii) holds, then G≅Aut(Mat(22)).
∎
6. L∘≅SL2(4)
In this section we investigate the configuration in \frefprop:unambiguous(iii). Thus L∘≅SL2(4), ∣YL:VL∣=2 and VL is the natural SL2(4)-module.
As Q≤L∘, CVL(S0)=CVL(Q)≤Z(Q), so Q is normal in NL∘(CVL(S0)) and hence Q=S0 is a Sylow 2-subgroup of L∘. In particular Z(Q)∩YL=Z(Q)∩VL is of order 4.
Lemma 6.1**.**
The subgroup Q is elementary abelian. In particular, Q∩YL=Q∩VL=CYL(Q)=Z, ∣YLQ/Q∣=23 and ∣VLQ/Q∣=22.
Proof.
We know that [Q,VL]=CVL(Q)=Q∩VL and, as Q is elementary abelian, Φ(Q)≤QL. If Φ(Q)=1, then, since Z(S)∩Φ(Q)=1, we deduce Φ(Q)∩VL=1. As NL(QQL) normalizes Q and is irreducible on [VL,Q], [VL,Q]≤Φ(Q). But then VL centralizes Q/Φ(Q), so VL≤Op(NG(Q))=Q, a contradiction. This shows Q is elementary abelian and then also YL∩Q=VL∩Q=CYL(Q).
∎
Proposition 6.2**.**
Suppose L∈LG(S) and L≤NG(Q) with L in the unambiguous wreath product case. If YL≤Q, L∘≅SL2(4) and ∣YL:VL∣=2, then G is Aut(Mat(22)).
Proof.
We start by observing that the action of L on YL gives
(6.2.1) **
(i)
∣VLQ/Q∣=∣Q:CQ(VL)∣=22;
2. (ii)
for all v∈VL∖Q, CQ(v)=CQ(VL); and
3. (iii)
for all w∈Q∖QL, [w,VL]=[Q,VL].
Let B=NL(QQL). Then B contains an element β of order 3 which acts fixed-point freely on VL and irreducibly on [VL,Q]=CYL(Q).
Using \frefclm:1 (ii) and \freflem:soldual yields [VL,F(NG(Q)/Q)]=1.
Let K≥Q be the preimage of
[TABLE]
Then K is non-trivial, normalized by B and \freflem:soldual implies VLQ/Q acts faithfully on K/Q.
The three involutions of QQL/QL each centralize a subgroup of YL of order 23 and by \freflem:VL sub QL’(i), there are three elements of YLQ/Q which act on Q as GF(2)-transvections, they generate YLQ/Q and are permuted transitively by B/Q. As B normalizes K and as VLQ/Q acts faithfully on K/Q, at least one and hence all of the transvections in YLQ/Q act faithfully on K/Q.
If CQ(K)=1, then CCQ(K)(S)=1. As Ω1(Z(S))=CVL(S) by \freflem:VL sub QL’ (ii), and CQ(K) is normalized by B, we have [Q,VL]≤CQ(K). But then K=⟨VLK⟩Q centralizes Q/CQ(K) contrary to CK(Q)=Q. Hence CQ(K)=1.
Let V be a non-trivial minimal KYL-invariant subgroup of Q.
Then KYL acts irreducibly on V. Moreover, as YL does not centralize V, V≤QL and, as VL is the natural L∘-module we have [YL,V]=[YL,Q]=YL∩Q≤V. It follows that K centralizes Q/V and so K/Q acts faithfully on V=[Q,K] which is normalized by B. Hence CYL(V)=YL∩V=YL∩Q and YLQ/Q acts faithfully on V. Recall that YLQ/Q is generated by elements which operate as transvections on Q and hence on V. Therefore [McL, Theorem] applies to give KYL/Q≅SLm(2) with m≥3, Sp2m(2) with m≥2, O2m±(2) with m≥2, or Sym(m) with m≥7. Furthermore, V=[Q,K] is the natural module in each case.
Since CYLQ/Q(S/Q) contains a transvection and has order 22, KYL/Q≅SLm(2) with m≥3 or O2m±(2) with m≥2. Suppose that KYL/Q≅Sym(m) with m≥7. Then, as YLQ/Q is generated by three transvections, we see that YLQ/Q is generated by three commuting transpositions in KYL/Q. Let t be the product of these transpositions. Then, as m≥7, ∣[V,t]∣=23. However, ∣[V,YL]∣=22, and so we have a contradiction. We have demonstrated
(6.2.2) ** KYL/Q≅Sp2m(2), m≥2 and [Q,K]=[Q,KYL] is the natural module.
Since [Q,K] is the natural KYL/Q-module and [VL,Q]≤[Q,K] has order 22, we have [[VL,Q],S]=1. In particular, QQL/QL<S/QL≅Dih(8) and SQ/Q∩K/Q acts non-trivially on [Q,VL].
Consider Q∗=O2(KS). Since Q∗ centralizes [Q,K], Q∗ centralizes [VL,Q] and so Q∗QL=QQL. Hence Φ(Q∗)≤QL. If Φ(Q∗)=1, then
[TABLE]
and so also [Q∗,K]=[Q∗,K,K]≤[Q,K]≤Φ(Q∗) which is impossible. Hence Q∗ is elementary abelian and it follows that Q≤Q∗=CQ∗(Q)≤Q. Since KS acts on [Q,K] and KYL/Q≅Sp2m(2), we now deduce S≤KYL from the structure of \mboxOut(K/Q). Hence B=⟨SB⟩≤KYL as B normalizes KVL. It follows that B/Q is the minimal parabolic subgroup P of K/Q irreducible on [YL,V] and with O2(P) centralizing [YL,V]⊥/[YL,V]=CYL(V)/[YL,V]. Therefore there is β∈K of order three such that ⟨β⟩ is transitive on the transvections in YLQ/Q and normalizes QLQ/Q which has index 2 in S/Q. In particular, from the structure of the natural Sp2m(2)-module β centralizes
[TABLE]
As V is abelian, V acts quadratically on QL/VL. By \freflem:Sym5-modules, QL/VL involves only natural SL2(4)-modules and trivial modules as L-chief factors. We know β acts fixed-point freely on the natural module and so, as β centralizes [QL,V]YL/YL, all the L-chief factors of QL/VL are centralized by L. In particular, VL is the unique non-central L-chief factor in Q and so YL∩Φ(QL)=1. As Ω1(Z(S))≤VL by \freflem:VL sub QL’ (ii), Φ(QL)=1, so QL=Ω1(Z(QL))=YL, which together with S/QL≅Dih(8) implies
We next show that G has a subgroup of index two. In NG(Q) we have a subgroup U of index 2 of shape 24.Alt(6). Furthermore YL≤U and VL≤U. Since [v,Q]=CQ(v) for v∈VL∖Q and U/Q has one conjugacy class of involutions, all the involutions in U∖Q are U-conjugate. Since L acts transitively on VL and U is transitive on Q#, we have that all the involutions in U are G-conjugate. As Q is large, we have CG(z)≤NG(Q) for z∈Q#. Hence all the involutions in U have centralizer which is a {2,3}-group. There is an involution t in YL∖VL, which is not in U and centralized by an element of order 5 in L. Hence t is not conjugate to any involution of U. Application of [GLS2, Proposition 15.15] gives a subgroup G1 of index two in G. We have NG1(Q)/Q≅Alt(6). By \freflem:splitA6 this extension splits and we have that the centralizer of a 2-central involution z∈G1 is a split extension of an elementary abelian group of order 16 by Sym(4). In particular O(CG(z))=1 and so by coprime action O(G)=⟨CO(G)(e)∣e∈Q#⟩=1. As YL≤Q, there is an involution y∈NG1(Q)∖Q. Since all involutions in Qy and in NG1(Q)/Q are conjugate, G1 has just one conjugacy
class of involutions. In particular F∗(G1) is simple. Application of [J, Theorem] gives that F∗(G1)≅Mat(22) and so G≅Aut(Mat(22)).
∎
Acknowledgment
We thank the referee for numerous comments which have improved the readability and clarity of our work.
The second author was partially supported by the DFG.
Bibliography9
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[GH] D. Gorenstein, K. Harada, On finite groups with Sylow 2 2 2 -subgroups of type A n subscript 𝐴 𝑛 A_{n} , n = 8 𝑛 8 n=8 , 9 9 9 , 10 10 10 , 11 11 11 , Math. Z., 117 (1970), 207–238.
2[GLS 2] D. Gorenstein, R. Lyons, R. Solomon, The classification of the finite simple groups, Amer. Math. Soc. Surveys and Monographs 40(2), (1996).
3[J] Z. Janko, A characterization of the Mathieu simple groups, I, J. Algebra 9, 1968, 1–19.
4[Mas] D. Mason, Finite simple groups with Sylow 2 2 2 -subgroup dihedral wreath Z 2 subscript 𝑍 2 Z_{2} , J. Algebra 26, (1973), 10–68.
5[MSS 1] U. Meierfrankenfeld, B. Stellmacher, G. Stroth, Finite groups of local characteristic p: an overview in Groups, combinatorics and geometry , Durham 2001 (eds. A. Ivanov, M. Liebeck, J. Saxl), Cambridge Univ. Press, 155–191.
6[MSS 2] U. Meierfrankenfeld, B. Stellmacher and G. Stroth, The local structure theorem for finite groups with a large p 𝑝 p -subgroup , Mem. Amer. Math. Soc. 242, Nr. 1147 (2016).
7[Mc L] J. Mc Laughlin, Some subgroups of SL n ( F 2 ) subscript SL 𝑛 subscript 𝐹 2 {\operatorname{SL}}_{n}(F_{2}) , Illinois J. Math 13, 1969, 105-115.
8[PPS] Chr. Parker, G. Parmeggiani, B. Stellmacher, The P 𝑃 P !-Theorem, Journal of Algebra 263 (2003), 17–58.