Elementary abelian groups of rank 5 are DCI-groups
Yan-Quan Feng
Y.-Q. Feng,
Department of Mathematics, Beijing Jiaotong University, Beijing 100044 PR China
[email protected]
and
István Kovács
I. Kovács,
IAM and FAMNIT, University of Primorska, Glagoljaška 8, 6000 Koper, Slovenia
[email protected]
Abstract.
In this paper, we show that the group Zp5 is a DCI-group for any odd prime p,
that is, two Cayley digraphs Cay(Zp5,S) and Cay(Zp5,T)
are isomorphic if and only if S=Tφ for some automorphism
φ of the group Zp5.
2010 Mathematics Subject Classification. 05C25, 05C60, 20B25.
Key words and phrases. Cayley graph, isomorphism, CI-group,
2-closed permutation group, Schur ring.
Y.-Q. Feng was supported in part by the National Natural Science Foundation of China (11571035, 11231008) and the 111 Project of China (B16002). I. Kovács was supported in part by
the Slovenian Research Agency (research program P1-0285, and
research projects N1-0032, N1-0038, J1-5433 and J1-6720) and is grateful
to Beijing Jiaotong University for hospitality.
1. Introduction
Let H be a finite group and S be a subset of H. The
Cayley digraph Cay(H,S) is the digraph that has vertex set H, and arc set {(x,sx):x∈H,s∈S}. It follows from the definition that Aut(Cay(H,S)) contains HR, the group of all right translations HR={hR:h∈H}, where
xhR=xh,x∈H. Also, Cay(H,S) is loopless if the identity element 1∈/S, and it is regarded as an undirected graph when S is an inverse-closed set, that is, S=S−1={x−1:x∈S}.
Two Cayley digraphs Cay(H,S) and Cay(H,T) are called Cayley isomorphic if T=Sφ for some automorphism φ∈Aut(H). It is trivial to show that Cayley isomorphic Cayley digraphs are isomorphic as digraphs. The converse, however, does not hold in general. There are examples of Cayley digraphs which are isomorphic but not Cayley isomorphic. A subset S⊆H is called a CI-subset if for any T⊆H, the isomorphism
Cay(H,T)≅Cay(H,S) implies that T=Sφ for some φ∈Aut(H). The group H is a DCI-group if each of its subsets are CI-subsets, and a CI-group if each of its inverse-closed subsets are CI-subsets. Motivated by a problem posed by Ádám in [1], Babai and Frankl [4] asked the following question: Which are the CI-groups?
Although the candidates of CI-groups have been reduced to a restricted list [9, 17], which was obtained by accumulating the work of several mathematicians, it is considered to be difficult to confirm that a particular group is a CI-group. We refer the reader to the survey paper [18] for most results on CI- and DCI-groups.
One of the crucial steps towards the classification of all CI-groups is to answer which elementary
abelian p-groups are CI-groups (see also [18, Question 8.3]). It is known that the group
Zpn is a CI-group in each of the following cases: n=1 [7, 11, 29];
n=2 [2, 12]; n=3 [2, 8]; n=4 and p=2 [5]; n=4 and p>2 [13]
(a proof for n=4 with no condition on p was given recently in [19]); n=5 and p=2 [5]; and n=5 and p=3 [27]. On the other hand, some examples of groups Zpn are also known which are not CI-groups,
and in each case the rank n≥6. Nowitz [23] found a non CI-subset of Z26, and more recently, Spiga [27] constructed a non-CI subset of Z38. Constructions of non-CI subsets of Zpn where n is expressed as a function in p were the subject of the papers [20, 25, 26]. The best bound is due to Somlai [25], which says that Zpn is not a CI-group if n≥2p+3. The question whether Zp5 is a CI-group for any odd prime p is mentioned in [18] as a crucial task for classifying CI-groups (see Section 8.4 and Problem 8.10). The goal of this paper is to complete this task by proving the following theorem:
Theorem 1.1**.**
The group Zp5 is a DCI-group for any odd prime p.
Our starting point is the following group theoretical criterion due to Babai [3]: A subset S⊆H is a CI-subset if and only if any two regular subgroups of Aut(Cay(H,S)) isomorphic to
H are conjugate in Aut(Cay(H,S)). Recall that, the group HR of right translations is always contained in Aut(Cay(H,S)). Motivated by this criterion, the following
definition was introduced by Hirasaka and Muzychuk [13]: A permutation group G≤Sym(Ω) containing a fixed subgroup F is F-transjugate if for each g∈Sym(Ω), the condition
g−1Fg≤G implies that g−1Fg and F and are conjugate in G. In this context,
Babai’s result can be rephrased as to say that a subset S⊆H is a CI-subset if and
only if the group Aut(Cay(H,S)) is HR-transjugate.
It is well-known that Aut(Cay(H,S)) is a 2-closed permutation group for any S⊆H
(for the definition of a 2-closed permutation group, see Section 2.1). Following [13], we say that H is a CI(2)-group if all 2-closed subgroups of Sym(H) containing HR are HR-transjugate. Clearly, if H is a CI*(2)*-group, then it is necessarily a DCI-group. In fact, instead of Theorem 1.1 we prove the following slightly more general theorem:
Theorem 1.2**.**
The group Zp5 is a CI(2)-group for any odd prime p.
We prove Theorem 1.2 following the so called S-ring approach (S-ring is the abbreviation of Schur ring, and
for a definition, see Section 2.2). Roughly speaking, S-rings are certain subalgebras of the group algebra QH which were introduced by Schur [24] in order to study permutation groups containing a regular subgroup isomorphic to H. The usage of S-rings in the investigation of CI-groups was proposed by Klin and Pöschel [14, 15].
For a concise survey on S-rings and their applications in combinatorics, we refer the reader to [21].
We finish the introduction with a brief outline of the paper:
Section 2 contains preliminary material, especially, a thorough introduction to S-ring theory.
We intend to keep our text as self-contained as possible.
In Sections 3, we turn to S-rings over elementary abelian p-groups of arbitrary rank. In particular, an equivalent condition will be derived for the group Zpn to be a CI*(2)*-group in terms of its S-rings (Proposition 3.4). We remark that, this condition is obtained by combining together several results
proved in [13, 19, 26]. Based on this equivalence, Theorem 1.2 will be reformulated in a statement
involving a particular class of S-rings over the group Zp5 (Theorem 3.5).
Then, in Section 4, we derive a property of S-rings over Zp4 which will be needed when dealing with S-rings over Zp5.
The proof of Theorem 3.5 will be divided into two parts depending on whether the S-rings in question are
decomposable or not (for a definition of a decomposable S-ring, see Section 2.2). The decomposable S-rings will be handled in Section 5, while the indecomposable ones in Section 6.
2. Preliminaries
All groups in this paper are finite.
In this section we collect all concepts and facts needed in this paper.
2.1. Permutation groups
Let G≤Sym(Ω) be a permutation group of a finite set
Ω. For ω∈Ω, we denote by Gω the stabilizer of ω in G, and by ωG the
G-orbit of ω.
For a subset Δ⊆Ω and permutation γ∈Sym(Ω),
we say that γ fixes Δ if Δγ=Δ, and
that γ fixes Δ pointwise if ωγ=ω for all ω∈Δ. The setwise stabilizer
and pointwise stabilizer of Δ in G will be denoted by G{Δ} and GΔ, resp.,
that is, G{Δ}={g∈G:Δg=Δ} and GΔ={g∈G:ωg=ω,ω∈Δ}.
The set of all G-orbits is denoted by Orb(G,Ω). Suppose that G is transitive on Ω. If δ={Δ1,…,Δn} is a block system for G, then we write Gδ for the kernel of the action of G on δ, and Gδ for the permutation group of δ induced by G.
Two permutation groups H,G≤Sym(Ω) are said to be 2-equivalent, denoted by H≈2G, if Orb(H,Ω2)=Orb(G,Ω2), see [31].
The equivalence class of G contains a largest subgroup, which is called the 2-closure of G, denoted by G(2).
The group G is called 2-closed if G(2)=G.
Proposition 2.1**.**
For any G≤Sym(Ω), Z(G)≤Z(G(2)).
Proof.
Let g1∈Z(G),γ∈G(2) and
ω∈Ω. Since Orb(G,Ω2)=Orb(G(2),Ω2), we have (w,wg1)G=(w,wg1)G(2), and so there exists g∈G, depending on w and wg1, such that (w,wg1)g=(w,wg1)γ.
Then ωg1γ=ωg1g=ωgg1=ωγg1. As ω was chosen arbitrarily
from Ω and γ from G(2),
it follows that g1∈Z(G(2)), hence
Z(G)≤Z(G(2)), and the assertion follows.
∎
A transitive permutation group G and its 2-closure G(2) have the same block systems (see [31, Theorem 4.11]).
Proposition 2.2**.**
([13, Proposition 2.1])*
Let G≤Sym(Ω) be a transitive permutation group,
and let δ be a block system for G. Then*
- (i)
(G(2))δ≤(Gδ)(2).
2. (ii)
If G is 2-closed and Fδ is 2-closed for some
F≤G, then FGδ is also 2-closed.
The following statement is given as Exercise 5.8 in [31].
It also appears as 5.1. Proposition in the preprint [22], where the authors give a
proof. Regarding the fact that [22] is a university preprint, we also present a proof.
Proposition 2.3**.**
(cf. [31])*
If G≤Sym(Ω) is a p-group, then
G(2) is also a p-group.*
Proof.
If G is intransitive, then G(2) is a subdirect product of the 2-closures
of the transitive components of G. Thus, it is sufficient to prove the theorem
for transitive groups G≤Sym(Ω).
Suppose to the contrary that G is a counterexample to the proposition
whose order is the smallest possible. If G is abelian, then it is regular on Ω.
By Proposition 2.1, G(2) centralizes G, and it follows that
G(2)=G. Thus, we may assume that G is non-abelian, and so
∣G∣≥p3.
The center Z(G) is nontrivial. Let P≤Z(G) such that ∣P∣=p, and let
g∈G(2) be an element of order q for some prime q=p. Then Orb(P,Ω) is a block system of G on Ω.
Let us consider the natural action of G on Orb(P,Ω).
To simplify notation, we write δ=Orb(P,Ω). The permutation group Gδ≤Sym(δ) induced by
G is transitive on δ and has order less than ∣G∣. By the minimality of G the group (Gδ)(2) is a p-group. Thus (G(2))δ is also
a p-group, see Proposition 2.2(i), which implies that g
acts on δ as the identity permutation.
Equivalently, g fixes any P-orbit Δ∈δ.
By Proposition 2.1, P≤Z(G(2)), hence g centralizes P. This implies that g is
semiregular on Δ. Since g has order q and ∣Δ∣=p, g fixes
pointwise Δ, and as this is true for any Δ∈δ, g is the identity permutation of Ω, a contradiction.
∎
Proposition 2.4**.**
( [13, Proposition 3.6(ii)] )*
Let H be an abelian p-group whose order ∣H∣≥p3, and
let G≤Sym(H) with G≥HR. If there exists a G1-orbit T such
that ∣T∣=p and ⟨T⟩=H, then ∣G∣=p⋅∣H∣.*
Finally, we recall a recent result of Morris [19].
Let H≅Zpn for an arbitrary prime p.
Assume that G≤Sym(H) is a p-group such that
[TABLE]
Let P be a Sylow p-subgroup of Sym(H) with G≤P.
Then P is permutation isomorphic to the iterated
wreath product Zp≀⋯≀Zp
(n copies of Zp), and this shows that P
admits block systems δ0,…,δn−1 such that
δi has blocks of size pi+1, and if 0≤i<j≤n−1, then
each class of δi is contained in a class of δj.
Since HR is abelian, the kernel (HR)δi has order
pi+1. In particular, there exist τ0,τ1∈HR
such that (HR)δ0=⟨τ0⟩ and (HR)δ1=⟨τ0,τ1⟩. Note that, we can write δ0=Orb(⟨τ0⟩,H) and δ1=Orb(⟨τ0,τ1⟩,H).
Proposition 2.5**.**
([19, Corollary 3.2])*
With the above notation, there exists ψ∈G(2)
such that ψ commutes with τ0, and
ψ−1π−1HRπψ contains τ1.*
Let us consider once more the above groups G=⟨HR,π−1HRπ⟩ and P, where P
is a Sylow p-subgroup of Sym(H) with G≤P.
Then Z(P)≤π−1HRπ because π−1HRπ is
abelian and regular on H. Similarly, Z(P)≤HR. On the other hand, Pδ0⊲P and τ0∈Pδ0, implying Pδ0∩Z(P)=1 because P is p-group. Then Pδ0∩Z(P)≤HR implies Pδ0∩Z(P)=⟨τ0⟩, hence τ0∈Pδ0∩Z(P)≤π−1HRπ. Proposition 2.5 together with the condition
that ψ centralizes τ0 shows that
τ0,τ1∈ψ−1π−1HRπψ, and hence
∣CHR(ψ−1π−1HRπψ)∣≥p2.
This inequality will be used later.
2.2. S-rings
In this subsection, we give the definition of an S-ring, and
review several basic properties.
Let H be a finite group with identity element 1, and let QH denote the group algebra of H over the rational number field. For a subset T⊆H, we define
the QH-element T as the formal sum T=∑h∈Hahh with ah=1 if h∈T, and ah=0 otherwise. We remark
that the QH-element T is traditionally called simple quantity, see [30].
By a Schur-ring over H (S-ring for short) we mean a subalgebra A⊆QH, which can be associated with a partition π of H satisfying the following conditions:
The set {1} belongs to π.
For every T∈π, the set T−1 belongs to π.
A is spanned by the QH-elements
T,T∈π.
The elements (classes) of π are also called the basic sets of A, and
from now on we will use the notation Bsets(A) for π.
The cardinality ∣Bsets(A)∣ is called the rank of A.
The concept of S-ring is due to Wielandt [30], which was motivated by the following result of Schur [24]:
Theorem 2.6**.**
(cf. [30, Theorem 24.1])*
Let H be a finite group, and let G≤Sym(H) with HR≤G.
Then the QH-elements T,T∈Orb(G1,H) span an S-ring
over H.*
The S-ring in the above theorem is also called the
transitvity module of G1, denoted by
V(H,G1). We note that, there exist S-rings which do not arise as transitivity
modules. Given an arbitrary S-ring A over a group H, we say that
A is Schurian if A=V(H,K1) for some permutation group K≤Sym(H) with HR≤K, and that A is non-Schurian otherwise.
Remark 2.7*.*
It should be noted that the pair (H,{Cay(H,T):T∈Bsets(A)})
forms a Cayley (association) scheme in the sense of [21].
Thus, S-ring theory can be regarded as a part of the theory of association schemes,
and several concepts defined for S-rings can be understood in this context.
Let A be any S-ring over a group H.
A subset S⊆H (subgroup K≤H, resp.) is called an A-subset (A-subgroup, resp.) if S∈A (K∈A, resp.). The radical of a subset S⊆H is the subgroup of H defined as
[TABLE]
In other words, rad(S) is the largest subgroup E≤H for
which S is equal to the union of some left E-cosets and also
some right E-cosets.
If S is an A-subset, then both groups rad(S) and ⟨S⟩ are A-subgroups
(see [30, Propositions 23.5 and 23.6]).
If K,L≤H are two A-subgroups, then it can be easily checked that both
K∩L and ⟨K∪L⟩ are A-subgroups.
The thin radical of A is defined as
[TABLE]
The following simple, but useful property is a simple observation:
[TABLE]
Here eT={et:t∈T} and Te={te:t∈T}. It follows that
the thin radical Oθ(A) is an A-subgroup.
Let K≤H be an A-subgroup. Then, for any basic set T∈Bsets(A) there exist positive integers k and k′ such that
[TABLE]
These can be verified by considering the products
K⋅T and T⋅K.
Here and in what follows the symbol ⋅ denotes the
multiplication of QH. Since both products belong to A, these
can be expressed as a linear combination of the simple quantities
T′,T′∈Bsets(A). The coefficient by
T is equal to k in the case of K⋅T, and
it is equal to k′ in the case of T⋅K.
The subalgebra QK∩A is an S-ring over an A-subgroup K, denoted by
AK, which is called the S-subring of A induced by K.
Assume, in addition, that K⊴H for an A-subgroup K.
For a subset S⊆H, we let S/K={Kh,h∈S}.
Let T1,T2∈Bsets(A) such that KT1∩KT2=∅, or equivalently,
k1t1=k2t2 holds for some ki∈K and ti∈Ti (i=1,2).
This shows that, the coefficient aT1>0 by the linear combination
K⋅T2=∑T∈Bsets(A)aTT. This implies
that T1⊆KT2, and hence KT1⊆KT2.
Similarly, bT2>0 by K⋅T1=∑T∈Bsets(A)bTT,
hence T2⊆KT1, and so KT2⊆KT1 also holds.
We conclude that KT1=KT2, and thus the sets KT,T∈Bsets(A) form a
partition of H, and consequently, the sets T/K,T∈Bsets(A) form a partition of
H/K. The corresponding QH/K-elements T/K span an S-ring over H/K, which is called the quotient of A by K, denoted by AH/K.
Assume that H=E×F is the internal direct product of its subgroups E and F,
and that A is an S-ring over H such that both
E and F are A-subgroups. Since E∩F={1},
it follows that XY=X⋅Y for any subsets X⊆E and Y⊆F. A straightforward computation yields that the simple quantities
RS,R∈Bsets(AE),S∈Bsets(AF) span an S-ring over H.
The latter S-ring is called the tensor product of AE with AF,
denoted by AE⊗AF. Clearly, AE⊗AF=AF⊗AE,
and AE⊗AF⊆A. The following lemma can be
easily shown using Eq. (1).
Lemma 2.8**.**
Let A be an S-ring of the internal direct product H=E×F such that
both E and F are A-subgroups. If AE=QE or AF=QF,
then A=AE⊗AF.
The following result is also known as Schur’s first theorem on multipliers (see [21]).
Theorem 2.9**.**
(cf. [30, Theorem 23.9(a)])*
Let A be an S-ring over an abelian group H, T∈Bsets(A) be any basic set, and suppose that k is an integer coprime to ∣H∣.
Then the set T(k):={hk:h∈T} is a basic set of A.*
Finally, we recall the concept of E/F-wreath product after [21].
This was defined in [16] under the name wedge product, and independently in [10] under the name generalized wreath product.
Let A be an S-ring over a group H. If there exist A-subgroups E and F such that
[TABLE]
then we say that A is an E/F-wreath product and
write A=AE≀E/FAH/F. Note that, the S-ring A can be reconstructed uniquely from the S-rings AE and AH/F. In the particular case when E=F,
we use term wreath product, and write
AE≀AH/E for AE≀E/EAH/E.
In what follows we say that A is decomposable if it can be decomposed as A=AE≀E/FAH/F where E=H and F={1}, and that A is
indecomposable otherwise.
2.3. Automorphisms of S-rings
Let A⊆QH be an S-ring over a group H.
By an automorphism of A we mean a permutation of H that is an automorphism of all Cayley graphs Cay(H,T),T∈Bsets(A).
This definition is due to Klin and Pöschel [14] (see also [21]). The group of all automorphisms of A will be denoted by Aut(A), that is,
[TABLE]
In what follows, we write Aut(A)1 for the stabilizer (Aut(A))1.
Note that, as a permutation group of H, Aut(A) is 2-closed.
Moreover, if A=V(H,G1) for some G≤Sym(H) with G≥HR, then
Aut(V(H,G1))=G(2). If K≤G is a subgroup with HR≤K, then
V(H,K1)⊇V(H,G1). Also, given two S-rings A and B of the same group H, the inequality B⊆A implies that
Aut(B)≥Aut(A).
For two arbitrary S-rings A,B⊆QH, their intersection A∩B
is also an S-ring over H (cf. [13, 21]). Therefore, given any subset
S⊂H, it is possible to define the S-ring
⟨⟨S⟩⟩:=∩A∗A∗, where A∗ runs over the set of all S-rings over
H that contain S. Then, the following identity holds:
[TABLE]
Indeed, let G=Aut(Cay(H,S)) and A=V(H,G1).
The fact that G≥Aut(⟨⟨S⟩⟩) follows if we observe
that S can be expressed as S=∪i=1kTi for some basic sets
Ti∈Bsets(⟨⟨S⟩⟩), and thus
Aut(⟨⟨S⟩⟩)≤⋂i=1kAut(Cay(H,Ti))≤G.
On the other, since G is 2-closed, G=Aut(A). Also, as any element of
G1 maps S to itself, it follows that S∈A. This implies in turn that
⟨⟨S⟩⟩⊆A, and so Aut(⟨⟨S⟩⟩)≥Aut(A)=G,
and Eq. (3) follows.
Suppose that K≤H is an A-subgroup and write G=Aut(A).
Then any element of the stabilizer G1 maps K to itself. This implies that
the setwise stabilizer G{K} factorizes as
G{K}=G1KR. In particular, G1≤G{K}, and hence the
G{K}-orbit of 1 is a block for G (see [6, Theorem 1.5A]). The latter orbit is K, and we conclude that the induced block system δ={Kg:g∈G}
is equal to the set H/K of all right cosets of K in H.
Finally, we point out a relation between Aut(A) and the thin radical Oθ(A). For h∈H, the left translation hL∈Sym(H) is the permutation acting as xhL=h−1x,x∈H.
If A is a Schurian S-ring over H, then its thin radical Oθ(A) satisfies the following:
[TABLE]
Indeed, if h∈Oθ(A) then every
g∈Aut(A) acts as an automorphism of
Cay(H,{h}). It is straightforward to check that this implies that g and
(h−1)L commute, and so
hL∈CSym(H)(Aut(A)). On the other hand, if hL∈CSym(H)(Aut(A)) and
g∈Aut(A)1, then (h−1)g=1hLg=1ghL=h−1. Therefore, the orbit of h−1 under Aut(A)1 is equal to the set {h−1}. Now, since A is Schurian, its basic sets are the
Aut(A)1-orbits on H, in particular, h−1∈Oθ(A), and this implies that
h∈Oθ(A) as well.
2.4. Isomorphisms of S-rings
Let A be an S-ring over a group H and B be an S-ring over a group K. A bijection f:H→K is called an (combinatorial) isomorphism
between A and B if
[TABLE]
Here Cay(H,T)f is the image of the digraph Cay(H,T) under f,
that is, it has vertex set K and arc set
{(xf,yf):x,y∈H and yx−1∈T}.
It follows from the definition that f induces a bijection
f∗:Bsets(A)→Bsets(B) defined by Tf∗=S for T∈Bsets(A)
exactly when Cay(H,T)f=Cay(K,S).
We say that f is normalized if f maps the identity element
1H to the identity element 1K. In the special case when f is
an isomorphism from H to K, we call f a Cayley isomorphism.
Notice that, when f is a normalized isomorphism
of A, then Tf∗=Tf holds for all T∈Bsets(A).
It is well-known that the linear map defined by T↦Tf∗ is an algebra isomorphism between the Q-algebras A and B (cf. also [13, 21]). Using this fact, it is not hard to show that (ST)f=SfTf holds for any normalized isomorphism f of A and any basic sets S,T∈Bsets(A). Some properties are listed below.
Proposition 2.10**.**
([13, Proposition 2.7])*
Let f:A→B be a normalized isomorphism from an
S-ring A over a group H to an S-ring B over a group K,
and let E≤H be an A-subgroup. Then,*
- (i)
the image Ef is a B-subgroup of K. Moreover, the restriction
fE:E→Ef is an isomorphism between AE and AEf.
2. (ii)
For each h∈H, (Eh)f=Efhf.
3. (iii)
If E⊴H and Ef⊴K,
then the mapping fH/E:H/E→K/Ef,
defined by (Eh)fH/E:=Efhf is a normalized isomorphism between
AH/E and BK/Ef.
In this paper, we will be interested exclusively in isomorphisms
between S-rings over the same group H. We adopt the notation used in [13], and
denote by Iso(A) the set of all isomorphisms from A to S-rings over H, that is,
[TABLE]
and let Iso1(A)={f∈Iso(A):1f=1}.
Note that, Iso(A)⊆Sym(H), but it is not necessarily
a subgroup. It follows from the definition that for any γ∈Aut(A) and
ψ∈Aut(H), their product γψ is an isomorphism from A to an S-ring over H.
Therefore, Aut(A)Aut(H)⊆Iso(A). Now, we say that
A is a CI-S-ring or simply that A is CI, if
Iso(A)=Aut(A)Aut(H).
This definition was given by Hirasaka and Muzychuk in [13], where the
following proposition was proved:
Proposition 2.11**.**
([13, Theorem 2.6])*
Let G≤Sym(H) be a 2-closed group with HR≤G, and
let A=V(H,G1). Then the following conditions are equivalent:*
- (i)
G* is HR-transjugate.*
2. (ii)
Iso(A)=Aut(A)Aut(H).
3. (iii)
Iso1(A)=Aut(A)1Aut(H).
Thus, the CI*(2)*-property for a group H is equivalent to the CI-property
for all Schurian S-rings over H.
In the last lemma of this subsection we collect further properties of S-ring isomorphisms.
Lemma 2.12**.**
Let A be an S-ring over a group H, and let f∈Iso1(A).
- (i)
If H is abelian and Tf=T for some T∈Bsets(A),
then (T(k))f=T(k) for any integer k coprime to ∣H∣.
2. (ii)
Let E≤H be an A-subgroup such
that E≤rad(T) for some T∈Bsets(A). If (TE)f=TE
then Tf=T.
Proof.
(i): Since Tf=T, f∈Aut(Cay(H,T)). Let us consider the
S-ring ⟨⟨T⟩⟩. By Eq. (3), f∈Aut(⟨⟨T⟩⟩). On the other hand, by Theorem 2.9, T(k)∈Bsets(⟨⟨T⟩⟩), and (i) follows.
(ii): This follows from (TE)f=TE and ET=TE=T as E≤rad(T).
∎
2.5. p-S-rings
We say that an S-ring A over a group
H is a p-S-ring if H is a p-group, and all basic sets T∈Bsets(A)
have p-power size, see [13]. The following proposition follows from results about
p-schemes proved in [32] (see [13, Theorem 3.3]). For sake of
easier reading, we give a proof using only the definition of an S-ring.
Proposition 2.13**.**
Let A be a p-S-ring over a p-group H. Then
- (i)
the thin radical Oθ(A) is non-trivial;
2. (ii)
there exists a chain of A-subgroups
[TABLE]
such that ∣Hi+1:Hi∣=p for all i∈{0,…,r−1}.
Proof.
By definition, ∑T∈Bsets(A),T={1}∣T∣=∣H∣−1. Since all cardinality ∣T∣ as
well as ∣H∣ are p-powers, (i) follows.
We prove (ii) by induction on ∣H∣. The statement is trivial for ∣H∣=p. For
the rest of the proof it is assumed that ∣H∣>p.
Choose a maximal non-trivial A-subgroup K<H, that is, H is the only
A-subgroup which contains properly K. Let ∣K∣=pm.
Let us consider the sets KTK,T∈Bsets(A).
These sets form a partition of H because T1⊆KT2K
or T1∩KT2K=∅ for any basic sets T1,T2∈Bsets(A) by Eq. (2), and therefore, KT1K=KT2K
or KT1K∩KT2K=∅.
Note that, pm divides ∣KTK∣ for all T,
and KTK=K for all basic sets T⊂K. Thus,
∣H∣=∑T∈Bsets(A),T⊆K∣KTK∣+∣K∣,
and so there exists a basic set
T1 such that T1⊆K and ∣KT1K∣=pm. Then,
Kt⊆KT1K and tK⊆KT1K for all t∈T1.
This together with ∣Kt∣=∣tK∣=∣KT1K∣=pm shows that any t∈T1 normalizes
K. Thus, K⊴⟨K,T1⟩=H, where the latter equality follows by the maximality of K. Now, (ii) follows by applying the induction hypothesis to
the S-rings AK and AH/K.
∎
Proposition 2.14**.**
([13, Proposition 3.4(i)])*
Let A be a p-S-ring over an abelian p-group H.
If there exists a basic set T∈Bsets(A) with ∣T∣=∣H∣/p, then
A decomposes to the wreath product A=AK≀AH/K, where
K≤H is an A-subgroup with index ∣H:K∣=p.*
The next result is the classification of all p-S-rings over Zp3.
Theorem 2.15**.**
([13, 28])*
Every p-S-ring over the group Zp3 for an odd prime
p is Cayley isomorphic to one of the S-rings given in Table 1.*
Remark 2.16*.*
Hirasaka and Muzcyhuk [13] classified the Schurian S-rings, and it was proved later by Spiga and Wang [28]
that all p-S-rings over Zp3 are Schurian (see [28, Theorem 1]).
Let H be a group isomorphic to Zp3 for an odd prime p. An S-ring A
over H is called exceptional if it is Cayley isomorphic to the S-ring in the 6th row of Table 1, see [13]. Exceptional S-rings will play an important role in later sections.
Lemma 2.17**.**
Let A be an exceptional S-ring over a group H,H≅Zp3. Then ∣Aut(A)∣=p4 and Iso1(A)=Aut(H).
Proof.
Consider the S-ring in the 6th-row of Table 1. Denote by T its basic
set containing the element (1,0,0)∈Zp3. Since p>2, it
follows quickly that
∣T∣=p and ⟨T⟩=Zp3. This implies that A has also a basic set
T′ such that ∣T′∣=p and ⟨T′⟩=H. The S-ring A is Schurian.
Thus, T′ is equal to an Aut(A)1-orbit, and by
Proposition 2.4, ∣Aut(A)∣=p4. Thus, HR⊴Aut(A), and so
Aut(A)1≤Aut(H). Since H is a CI*(2)*-group, A is a CI-S-ring,
see Proposition 2.11, and we can write
Iso1(A)=Aut(A)1Aut(H)=Aut(H).
∎
We finish the subsection with further properties.
Lemma 2.18**.**
Let A be a p-S-ring over a group H, K≤H be an
A-subgroup with index ∣H:K∣=p, and let T∈Bsets(A). Then the following hold:
- (i)
T* is contained in a K-coset. In particular, rad(T)≤K.*
2. (ii)
Let L≤H be an A-subgroup of order p such that
L⊴H and L≤rad(T). Then for any h∈T,
hL∩T=Lh∩T={h}.
3. (iii)
If H is an abelian group and ∣Oθ(A)∩K∣∣T∣>∣H∣/p, then Oθ(A)∩rad(T)={1}.
Proof.
(i): Let us consider the quotient S-ring AH/K.
By Eq. (2), AH/K is a p-S-ring.
Since H/K≅Zp, its only p-S-ring is QH/K, and so
AH/K=QH/K. In particular, ∣T/K∣=1, hence
T⊆Kh for a coset Kh and (i) follows.
(ii) By Eq. (2), there is a positive integer k such that
∣hL∩T∣=∣Lh∩T∣=k for all h∈T.
As ∣T∣ is a p-power, k=1 or p. If k=p,
then we find LT=TL=T, so L≤rad(T), which is excluded by one of the assumptions. Thus, k=1 and (ii) follows.
(iii): Assume that ∣Oθ(A)∩K∣∣T∣>∣H∣/p.
Let us consider the sets eT,e∈Oθ(A)∩K.
By Eq. (1) and (i), these are all basic sets contained
in a coset Kh. If these are pairwise distinct, then
∣Oθ(A)∩K∣∣T∣=∑e∈Oθ(A)∩K∣eT∣≤∣Kh∣=∣H∣/p, a contradiction.
Thus, eT=e′T for distinct e,e′∈Oθ(A)∩K.
Using this and that H is abelian, we find e−1e′T=Te−1e′=T, and so
e−1e′∈Oθ(A)∩rad(T), by which (iii) follows.
∎
3. On CI-S-rings over Zpn
In this section we give three propositions about CI properties of S-rings over the groups Zpn. The first one is a necessary condition for an S-ring to be non-CI. It is essentially contained in the proof of [13, Proposition 3.9].
Proposition 3.1**.**
Suppose that A=V(H,P1) is a non-CI-S-ring, where
H≅Zpn and HR≤P≤Sym(H) is a p-group.
Then
the normalizer NAut(A)(HR) contains a subgroup
K for which the following hold:
- (i)
K≅Zpn,* it is regular on H, and K=HR.*
2. (ii)
The stabilizer (KHR)1 is elementary abelian, and
[TABLE]
Proof.
Let G=Aut(A) and N=NAut(A)(HR).
Note that, since G=P(2) and P is a
p-group, G is a p-group as well, see Proposition 2.3.
We first show the existence of a subgroup K≤N that has
all properties given in (i). If G=N, then the existence
of the required subgroup K follows
from the condition that A is a non-CI-S-ring. Now, suppose
that N<G.
Then, since G is a p-group, the normalizer
NG(N)=N, hence we may choose
some g∈NG(N)∖N. We let K=(HR)g. It
is straightforward to see that K has all properties given in (i).
Now, we turn to part (ii). Consider the group Q=KHR.
Clearly, Q=Q1K=Q1HR and Q1≅Q/HR≅K/(K∩HR). Thus, Q1 is an elementary abelian group. Also,
as both K and HR are abelian, K∩HR≤Z(Q), implying
that ∣CHR(Q1)∣≥∣K∩HR∣. This completes the proof of part (ii).
∎
The second proposition will be a sufficient condition for an S-ring over Zpn to be CI.
Definition 3.2**.**
We say that an S-ring A of a group H is ≈2-minimal if
[TABLE]
For example, the full group algebra QH is a ≈2-minimal S-ring.
The obvious examples for non-≈2-minimal S-rings are the S-rings of rank 2 (the two basic sets are {1} and H∖{1}). Clearly, Aut(A)=Sym(H), and
thus Aut(A)≈2X whenever HR≤X≤Sym(H) and X is 2-transitive on H.
Proposition 3.3**.**
Let A be a Schurian p-S-ring over a group H≅Zpn, and K≤H be an A-subgroup of order p such that AH/K is a ≈2-minimal CI-S-ring. Then A is a CI-S-ring.
Proof.
Let G=Aut(A) and choose L≤G such
that L is regular on H and L≅H. Because of Proposition 2.11 it is enough to show that L and HR are conjugate in G.
We write Gˉ=GH/K,HˉR=(HR)H/K and
Lˉ=LH/K. Note that HˉR=(H/K)R.
The group Lˉ is abelian acting transitively on
H/K. It follows that it is regular, and Lˉ≅Zpn−1.
Since AH/K is a CI-S-ring,
Lˉ=(HˉR)x for some x∈Aut(AH/K).
For sake of simplicity we denote by 1ˉ the identity of H/K.
We claim that Aut(AH/K)=Gˉ. To settle this it is
sufficient to show that Aut(AH/K)≈2Gˉ, and
use the assumption that AH/K is ≈2-minimal.
We have to show that AH/K=V(H/K,Gˉ1ˉ).
Here we copy the proof of [13, Proposition 2.8(ii)].
Since K is an A-subgroup, Kg=K for
any g∈G1, and thus by Proposition 2.10(ii), the coset Kh is mapped by gH/K to (Kh)gH/K=Khg. A basic set of AH/K is in the form T/K where
T∈Bsets(A). Since A=V(H,G1),
we find that T=hG1 for some h∈H. Observe that G{K}=KRG1, and for any g∈G,
gH/K∈Gˉ1ˉ if and only if g fixes setwise the subgroup K. This implies that g=kRg′ for some k∈K and g′∈G1. Now, we can express the
Gˉ1ˉ-orbit of Kh as
[TABLE]
We conclude that T/K is an orbit of Gˉ1ˉ,
and the claim follows.
Recall that Lˉ=(HˉR)x for some x∈Aut(AH/K)=Gˉ.
Choose g∈G such that gH/K=x−1. Then Lg≤GH/KHR, where GH/K denotes the kernel of G acting on H/K. Write M=GH/KHR.
Since both G and (HR)H/K are
2-closed groups, it follows by Proposition 2.2(ii) that
M is also 2-closed. We are done if we show that
M is HR-transjugate, because then
(Lg)g′=HR for some g′∈M, showing that L and HR are indeed conjugate in G.
Again, because of Proposition 2.11 we are done if we show that
the S-ring B=V(H,M1) is CI. Then A⊆B, and thus K is also a B-subgroup. Note that GH/K∩HR=KR and M=M1HR. Then ∣M1∣=∣M∣/∣HR∣=∣GH/KHR∣/∣HR∣=∣GH/K∣/∣KR∣ and ∣GH/K∣=∣(GH/K)1∣∣KR∣. It follows ∣M1∣=∣(GH/K)1∣ and hence M1=(GH/K)1. Since (GH/K)1⊲G1 and all orbits of GH/K are the cosets of K in H which have order p, we have K≤Oθ(B) and
all basic sets of B=V(H,M1) not contained in Oθ(B) are K-cosets.
Let f∈Iso1(B). In order to prove that B is a CI-S-ring, we have to find an automorphism φ∈Aut(H) such that
[TABLE]
Choose a minimal generating set {h1,…,hn} of H such that {h1,…,hℓ}, ℓ≤n, is a generating set of Oθ(B) with h1∈K. By Proposition 2.10, Kf≤H and (Khi)f=Kfhif. Since 1f=1,
Tf∈Bsets(Bf) for every basic set T∈Bsets(A).
Using this and that each Khi is a B-subset, we find that each Kfhif is a Bf-subset, and so
⟨Kfh1f,…,Kfhnf⟩≤H is a Bf-subgroup.
By Proposition 2.10(i),
∣⟨Kfh1f,…,Kfhnf⟩∣=∣⟨Kfh1f,…,Kfhnf⟩f−1∣.
Thus, H=⟨Kfh1f,…,Kfhnf⟩.
Since h1∈K, it follows that Kfh1f=Kf, and
{h1f,⋯,hnf} is also a minimal generating set of H. Define φ as the induced automorphism of H by
φ:hif↦hi for 1≤i≤n. Then hifφ=hi. To finish the proof, it suffices to show that Eq. (5) holds.
Set f1=fφ. Clearly, f1∈Iso1(B). Recall that for any S,T∈Bsets(B), (ST)f1=Sf1Tf1 (see the paragraph preceding Proposition 2.10). Then f1 fixes each element in Oθ(B) because f1 fixes a generating set of Oθ(B).
In particular, Kf1=K as K≤Oθ(B), and
Eq. (5) holds whenever T⊂Oθ(B).
Now, suppose that T⊂Oθ(B).
Let us consider the isomorphism f1H/K of BH/K induced by f1
(for the definition of f1H/K, see Proposition 2.10(iii)).
The quotient S-ring BH/K=QH/K. Since
QH/K is a CI-S-ring and Aut(QH/K)=(H/K)R, it follows that Iso1(BH/K)=Aut(QH/K)1Aut(H/K)=Aut(H/K). Also, f1H/K∈Iso1(BH/K), because Kf1=K, and so
f1H/K∈Aut(H/K).
On the other hand, as f1 fixes all generators hi,
f1H/K fixes a generating set of H/K, and so f1H/K is the identity mapping.
Since T⊂Oθ(B),
T=Kh for some h∈H∖K,
and we can write
Tf1=(Kh)f1=(Kh)f1H/K=Kh=T.
∎
Proposition 3.3 will be especially useful in conjunction with the fact that
all indecomposable Schurian p-S-rings over Zp4 are ≈2-minimal.
We prove the latter fact in Section 4.
Recall that, the CI*(2)*-property for a group H is equivalent to the CI-property
for all Schurian S-rings over H (see Proposition 2.11).
The third proposition is the following refinement:
Proposition 3.4**.**
Let H be a group isomorphic to Zpn for an
odd prime p.
Then the following conditions are equivalent:
- (i)
H* is a CI**(2)**-group.*
2. (ii)
All S-rings V(H,A) are CI-S-rings where
A<Aut(H) is a p-group with ∣CHR(A)∣≥p2.
Proof.
Notice that, the implication (i) ⇒ (ii) is a
direct consequence of Proposition 2.11.
Now, we turn to the implication (ii) ⇒ (i).
Let G≤Sym(H) be a 2-closed subgroup with HR≤G, and let K≤G be a regular subgroup such that K≅H. We have to show that K and HR are conjugate in G.
Now, choose a Sylow p-subgroup P of G such that HR≤P. Since G
is 2-closed, by Proposition 2.3, P is 2-closed, that is, P(2)=P.
By Sylow Theorem, Kx≤P for some x∈G, hence we may
assume that K≤P. According to Proposition 2.5 there exists some y∈⟨HR,K⟩(2)≤P(2)=P such that
∣CHR(Ky)∣≥p2. Let Q=⟨HR,Ky⟩.
Then Q(2)≤P(2)=P, and Q(2) is also a p-group.
It is sufficient to show that Q(2) is HR-transjugate.
Let us consider the normalizer N=NQ(2)(HR). Since ∣CHR(Ky)∣≥p2, it follows that ∣CHR(Q)∣≥p2. By Proposition 2.1, CHR(Q)=HR∩Z(Q)≤HR∩Z(Q(2))=CHR(Q(2)).
Therefore, ∣CHR(Q(2))∣≥p2, and as
N(2)≤Q(2), it follows that
∣CHR(N1(2))∣≥p2.
By the hypothesis in (ii), the S-ring V(H,N1(2)) is a CI-S-ring. Equivalently,
N(2) is HR-transjugate.
We finish the proof by showing that
N(2)=Q(2). In doing this we use the same idea as in
the proof of [27, Proposition 1]. Assume to the contrary
that N(2)<Q(2). Since Q(2) is a p-group,
we can choose an element z∈NQ(2)(N(2))∖N(2). Then (HR)z≤N(2) . Since N(2) is HR-transjugate, (HR)z=(HR)z′ for some z′∈N(2), and so we find z′z−1∈NQ(2)(HR)=N, from which
z∈N(2), a contradiction. Therefore, Q(2)=N(2),
showing that Q(2) is HR-transjugate, as required.
∎
In fact, we are going to derive Theorem 1.2 by showing that the condition in case (ii) of Proposition 3.4 holds when H≅Zp5.
Theorem 3.5**.**
Let H≅Zp5 for an odd prime p. Then all S-rings
V(H,A) are CI-S-rings where A<Aut(H) is a p-group with
∣CHR(A)∣≥p2.
The proof of Theorem 3.5 will be given in Sections 5 and 6.
4. Indecomposable Schurian p-S-rings over Zpn are
≈2-minimal for n≤4
We set some notation that will be used throughout the rest of the paper:
Notation. From now on p will stand for an odd prime, and
H will denote a group isomorphic to Zpn. The group H will be regarded as the additive group of an n-dimensional vector space over the field GF(p).
The elements of H will be denoted by lower case letters u,v, etc., while the subgroups of H by upper case letters U,V, etc. As usual, the identity element will be denoted by 0, and
the inverse of an element u∈H by −u.
For an integer k and a subset T⊆H we write T(k)=kT={ku:h∈T}, where ku=u+⋯+u, with ∣k∣ summands if k>0, and ku=−(u+⋯+u) otherwise.
It turns out that all indecomposable p-S-rings over the group
Zpn are ≈2-minimal for any odd prime p and n≤3. This is not hard to see for the groups Zp and Zp2. The full group algebra QZp is the only p-S-ring over Zp; and up to Cayley isomorphisms, there are two p-S-rings over
Zp2: QZp2 and QZp≀QZp, and the latter one is decomposable. Theorem 2.15 shows that, up to Cayley isomorphisms, there are two indecomposable p-S-rings over Zp3: QZp3 and
the exceptional p-S-ring given in the 6th row of Table 1.
By Lemma 2.17, the automorphism group of an exceptional p-S-ring has order p4, hence it is ≈2-minimal. In this section, we extend this result to the Schurian indecomposable p-S-rings over Zp4.
Theorem 4.1**.**
All indecomposable Schurian p-S-rings over the group
Zp4 are ≈2-minimal for any odd prime p.
The proof of the theorem will be given in the end of the section following three preparatory lemmas.
Recall that, if A is an S-ring over H and W≤H is an A-subgroup, then the W-cosets in H form a block system for Aut(H).
Lemma 4.2**.**
Let A be an indecomposable S-ring over a group
H≅Zpn, and let W be an A-subgroup with
∣W∣=p. Then the kernel
[TABLE]
where δ denotes the block system H/W.
Proof.
Let K=Aut(A)δ. It is clear that
WR≤K, and thus it is enough to prove that the stabilizer
K0 is trivial. By Lemma 2.18(ii), for every basic set T∈Bsets(A),
[TABLE]
We define recursively a finite sequence T1,…,Tr of basic sets of A as follows. Let T1={w} where w is an arbitrary nonzero element in W.
Now, suppose that the sets T1,…,Ti are already defined for i≥1.
If H=⟨T1∪⋯∪Ti⟩, then finish the procedure and let r=i. Otherwise,
choose Ti+1 to be a basic set in
H∖⟨T1∪⋯∪Ti⟩ such that
W≤rad(Ti+1). Notice that, such Ti+1 does exist because A is indecomposable.
Let S=T1∪⋯∪Tr and consider the Cayley graph
Cay(H,S). Note that, Aut(A)≤Aut(Cay(H,S)).
It is clear from the construction that ⟨S⟩=H, hence
Cay(H,S) is connected. We claim that S has the property that, whenever a W-coset intersects S, it does intersect it at exactly one element. Suppose to the contrary that there exist u1,u2∈S such that
u1=u2 and u1−u2∈W.
Then u1∈Ti and u2∈Tj for some i,j∈{1,…,r}. It follows from the construction of S and Eq. (6)
that i=j, and we may assume w. l. o. g. that i<j. Thus, u2∈⟨Ti,W⟩≤⟨T1,…,Ti⟩, and so u2∈⟨T1∪⋯∪Tj−1⟩∩Tj, a contradiction.
Now, using that Aut(A)≤Aut(Cay(V,S)) and the above property of S, we find that every element in K0 fixes
all neighbors of [math] in Cay(H,S). This and the connectedness of Cay(H,S) yield that ∣K0∣=1.
∎
For x∈Aut(H), we define CH(x)={u∈H:uR∈CHR(x)}.
Lemma 4.3**.**
Let A be an indecomposable S-ring over a group
H≅Zp4 and let x∈NAut(A)(HR)
such that x=idH and 0x=0. Then ∣CH(x)∣≤p2.
Proof.
Since x=idH, it follows that ∣CH(x)∣≤p3.
Assume to the contrary that ∣CH(x)∣=p3.
Let U=CH(x), and for a fixed v1∈H∖U, let W=⟨v1x−v1⟩. Then ∣W∣=p, and it follows that the orbit v⟨x⟩=W+v for all v∈H∖U. Observe that, W is not an A-subgroup. For otherwise,
x belongs to the kernel of Aut(A) acting on H/W, which is impossible
by Lemma 4.2.
Let U′ be an A-subgroup of order p3. If U=U′, then define
V=⋂T∈Bsets(A),T⊆H∖Urad(T).
Clearly, V is an A-subgroup such that W≤V≤U, and
A is an U/V-wreath product, a contradiction.
Hence, U′=U, and in particular, U is not an A-subgroup.
Let T1∈Bsets(A) such that T1⊂U′ and
T⊂U. Then ∣T1∣≥p, and since W is not an
A-subgroup, it follows that
∣T1∣=p2. By Proposition 2.14, T1 is equal to a U′′-coset for some A-subgroup U′′ such that ∣U′′∣=p2 and
W<U′′. We find
U′′=W+W′ for an A-subgroup W′ of order p.
Since W′≤Oθ(A), W′<U, and it follows that U′′=U′∩U.
Now, choose T∈Bsets(A) such that T⊂U∪U′ and
T∩U=∅.
Notice that such T exists because U is not an A-subgroup.
It follows that ∣T∣>p. Fix an element v∈T∩U.
Then T⊆U′+v, see Lemma 2.18(i).
Since A is indecomposable, it follows from
Proposition 2.14 that ∣T∣=p2. Let v′∈T∖U.
If W′≤rad(T), then we find U′′+v′=W′+(W+v′)⊆W′+T=T. Thus, T=U′′+v′, contradicting that
T∩U=∅. Thus, W′≤rad(T), and by
Eq. (2), every W′-coset intersects T in at most 1 element . Consequently, any U′′-coset intersects T in at most p elements. The U′′-cosets contained in U′+v can be listed as U′′+ku+v, where
k∈{0,1,…,p−1} and u is any fixed element in U′∖U.
Let Ti=T∩(U′′+iu+v),i∈{0,1,…,p−1}.
It follows that the sets Ti form a partition of T, Ti is
a W-coset for all i>0, and ∣T0∣=p.
Let us consider the product T⋅(−T) in QH. Writing
T⋅(−T)=∑u∈Hauu, it follows quickly from the above
description of the sets Ti that
∑u∈U′′∖{0}au=p3−p2.
On the other hand, T⋅(−T)∈A, and it can be expressed as
the linear combination T⋅(−T)=∑T′∈Bsets(A)bT′T′.
Let w∈W,w=0. Since W is not an A-subgroup, it
follows that the coset
W′+w is a basic set of A. Let us denote the latter basic set
by T(w). It also follows
from description of the sets Ti that aw≥p2−p. Thus,
bT(w)≥p2−p as well, and as w was chosen arbitrarily from
W∖{0}, we arrive at a contradiction as follows:
[TABLE]
∎
Let A be a p-S-ring over H. In what follows, we call an ordered n-tuple (v1,…,vn) of generators of H an A-basis if all
subgroups in the chain below are A-subgroups
[TABLE]
Notice that, if x∈Aut(A) normalizes HR and 0x=0, then
x∈Aut(H), and it can be written in an A-basis as an upper triangular matrix having 1’s in the diagonal.
Lemma 4.4**.**
Let A be an indecomposable p-S-ring over a group H≅Zp4. Then
- (i)
∣NAut(A)(HR∣≤p6;**
2. (ii)
HR* is normal in Aut(A).*
Proof.
Let G=Aut(A) and N=NAut(A)(HR).
(i): Assume to the contrary that ∣N∣>p6, that is, for
the stabilizer N0 we have ∣N0∣>p2. Let us fix an A-basis (v1,v2,v3,v4). This means that ⟨v4⟩ is an A-subgroup, and we can consider the action of N0 on H/⟨v4⟩. By Lemma 4.2, the latter action is faithful, and hence N0 is isomorphic to a subgroup
of the group of all upper triangular 3×3 matrices
with each diagonal element equal to 1.
Therefore, ∣N0∣=p3, and we can choose
x∈Z(N0) that can be written in the basis (v1,v2,v3,v4) as
[TABLE]
Then, the orbit v1N0 has size at most p2. This
follows from Proposition 2.14 and the fact that
A is indecomposable. Therefore, there exists y∈N0
such that y=idH and y fixes v1. Using also that [x,y]=1, we find that (v1x)y=(v1y)x=v1x,
hence v1x=v1+v3+av4∈CH(y). It follows
that each of v1,v3 and v4 is in CH(y).
This, however, contradicts Lemma 4.3.
(ii): We have to show that G=N. Assume to the contrary that
G>N. Then NG(N)>N. Choose g∈NG(N)∖N and let P=(HR)gHR. Since
P0≤N0, ∣P0∣≤p2.
If ∣P0∣=p, then ∣P∣=∣HR∣⋅∣P0∣=p5, hence
∣(HR)g∩HR∣=p3.
Then every x∈P0 satisfies ∣CHR(x)∣≥∣(HR)g∩HR∣=p3, a contradiction to Lemma 4.3.
Therefore, ∣P0∣=p2, P0=N0 and each z∈N0 satisfies
∣CHR(z)∣≥∣(HR)g∩HR∣=p2.
By Lemma 4.3, CHR(z)=(HR)g∩HR whenever z=idH.
Therefore, letting U={u∈H:uR∈HRg∩HR}, we can write
[TABLE]
Let us consider the S-ring B=V(H,N0).
Clearly, U≤Oθ(B). Fix a B-subgroup V such that V has order p3 and
U<V. Let v∈H∖V and T∈Bsets(B) be a basic set such that
v∈T. By Lemma 2.18(i), vz−v∈V.
Suppose that vz−v∈U for all z∈N0. This implies that
T⊆U+v, and thus either T=U+v, or ∣T∣≤p.
In the latter case, however, it follows that N0 contains a nontrivial element z fixing some v∈T, and hence CH(z)≥⟨U,v⟩>U, which contradicts
Eq. (7). Observe that, if T=U+v, then it is also a
basic set of A. For otherwise, A would have a basic set of
size p3, contradicting that A is indecomposable (see Proposition 2.14). Now, since A is not a V/U-wreath product, there exists v1∈H∖V and x∈N0 for which
v1x−v1∈/U.
Now, define the elements v2=v1x−v1, v3=v2x−v2, and let v4∈U be an element such that U=⟨v3,v4⟩. It follows that (v1,v2,v3,v4) is a B-basis, in which
[TABLE]
Let y∈N0 such that N0=⟨x,y⟩.
By Eq. (7), CH(y)=U=⟨v3,v4⟩, and thus y can be written in the basis (v1,v2,v3,v4) in the form
[TABLE]
Using also that [x,y]=1, we find d=a and e=0. Thus,
v2y=v2+dv3. On the other hand, v2(xd)=v2+dv3 also holds, and we find v2∈CH(xdy−1), where xdy−1=idH.
This contradicts Eq. (7).
∎
All Schurian p-S-rings over Zp4 are CI [13, Theorem 3.1]. Combining this with Lemma 4.4 yields the following
corollary (see also the proof of Lemma 2.17):
Corollary 4.5**.**
Let A be an indecomposable Schurian p-S-ring over a group
H,H≅Zp4 for an odd prime p. Then
Iso0(A)=Aut(H).
Everything is prepared to prove the main result of this section.
Proof of Theorem 4.1.
Assume to the contrary that A is a Schurian indecomposable p-S-ring over
H, which is not ≈2-minimal. Let G=Aut(A). By Lemma 4.4,
HR⊴G and ∣G∣≤p6.
As A is not ≈2-minimal, ∣G∣=p6, and there exists
x∈G0 such that x has order p, and A=V(H,⟨x⟩).
In other words, G≈2K where K=⟨HR,x⟩. Note that
G=K(2). Let u∈CH(x). Then uR∈Z(K), and by
Proposition 2.1, uR∈Z(K(2))=Z(G), implying
that u∈CH(y) for any y∈G0. We obtain that
CH(x)≤CH(y) for all y∈G0.
Also, as A=V(H,⟨x⟩), every basic set of A has size at most p.
Thus, if ∣CH(x)∣=p2, then one can find y∈G0 such that y=idH and ∣CH(y)∣=p3, which is impossible by
Lemma 4.3. It remains to consider the case
when ∣CH(x)∣=p. Equivalently, rank(x−I)=3, and this implies
that, in a suitable basis, denoted by (v1,v2,v3,v4), x has the following Jordan normal form:
[TABLE]
Since x has order p, it follows that p>3.
Let T be the orbit of v1 under ⟨x⟩ (hence under G0). It is not hard to check that ∣T∣=p and ⟨T⟩=H. Then, by Proposition 2.4,
∣G0∣=p, a contradiction. This completes the proof of the theorem.
∎
We finish the section with a corollary of
Proposition 3.3 and Theorem 4.1, which will be used several times in the next two sections.
Corollary 4.6**.**
Let A be a Schurian p-S-ring over the group
H≅Zp5, and let W be an A-subgroup of order p. If A is a non-CI-S-ring, then the S-ring AH/W is decomposable.
5. Proof of Theorem 3.5 I: The decomposable S-rings
We record all assumptions of Theorem 3.5 in the following hypothesis:
Hypothesis 5.1**.**
A=V(H,A) is an S-ring over a group H≅Zp5 for
some odd prime p, and for some subgroup
A≤Aut(H) with ∣CH(A)∣≥p2.
Our eventual goal is to show that, assuming Hypothesis 5.1,
the S-ring A is CI. In this section, we deal with the particular case when
A is decomposable.
Theorem 5.2**.**
Assuming Hypothesis 5.1,
suppose that A is decomposable.
Then A is a CI-S-ring.
The theorem will be proved in the end of the section following four preparatory lemmas.
In the next three lemmas we study the S-ring A described in Theorem 5.2 which satisfies
additional conditions.
Lemma 5.3**.**
Assuming Hypothesis 5.1, suppose that there exist A-subgroups U1,U2,W1 and W2 with
∣U1∣=∣U2∣=p4,∣W1∣=∣W2∣=p,U1=U2,W1=W2 and W1+W2<U1∩U2, and also that the following hold:
- (1)
A* is both a U1/W2- and a U2/W1-wreath product.*
2. (2)
Both AU1/W1 and AU2/W2 are indecomposable.
3. (3)
∣Av∣=1* for some v∈H∖U1∪U2.*
Then one of the following possibilities holds:
- (i)
∣NAut(A)(HR)∣=p8,* and there exists an A-subgroup
U3 such that ∣U3∣=p4,U3=Ui for i∈{1,2}, and
every basic set of A not contained in U1∪U2∪U3 is equal to a (U1∩U2)-coset.*
2. (ii)
∣NAut(A)(HR)∣=p9,* and every basic set of A not contained in U1∪U2 is equal to a (U1∩U2)-coset.*
Proof.
We let N=NAut(A)(HR),W=W1+W2 and U=U1∩U2. Note that A=V(H,N0), where N0 denotes the stabilizer of
[math] in N. Also notice that, wγ=w for all w∈W and
γ∈N0.
Fix a non-trivial element x∈Av, and some u1∈U1∖U which is not fixed by x.
Then there exists an integer k such that ku1+v∈U2. Since v∈/U1, it follows that
ku1+v∈/U. We define the elements v1=ku1,v2=ku1+v and v3=v1x−v1.
We find that v2x=v2+v3. For i∈{1,2},
let Ti=viA, the A-orbit containing vi (in other words, Ti is the basic set of
A that contains vi). Note that, since AUi/Wi is indecomposable, it follows that rad(Ti)=Wi for both i=1,2.
Since v3=vix−vi, it follows that v3∈Ti−Ti, hence v3∈Ui.
We conclude that v3∈U1∩U2. Suppose that v3∈W. Then both v1 and v1x are in
Ti∩⟨v3⟩+v1, and this together with Eq. (2) shows that ∣Ti∩⟨v3⟩+v′∣=p for all v′∈Ti, and hence ⟨v3⟩≤rad(Ti).
It follows that ⟨v3⟩≤rad(T1)∩rad(T2)=W1∩W2.
Thus, v3=0, that is, v1x=v1. This contradicts
that v1=ku1 and u1 was chosen so that it is not fixed by x. We conclude that
v3∈U∖W.
Next, assume for the moment that v3x−v3∈W1. Let us consider the automorphism
xU1/W1. First, as v1x−v1=v3∈/W1, we see that xU1/W1 is not the identity mapping.
On the other hand, xU1/W1 fixes W/W1 pointwise and the element W1+v3 (here W1+v3 is regarded as
an element of the group U1/W1). By all these we find ∣CU1/W1(xU1/W1)∣=p2, which implies that
AU1/W1 is a nontrivial generalized wreath product, a contradiction to the assumption given in (2).
We conclude that v3x−v3∈/W1.
Notice that, there is a symmetry between the conditions satisfied by the pairs (U1,W1) and (U2,W2). Therefore, any statement, which involves the subgroups U1,U2,W1 and W2, and which can be derived from these conditions, gives always rise to yet another statement that is obtained by replacing U1 with U2 and W1 with W2. In what follows, we will refer to the new statement as the dual counterpart.
For instance, the statement v3x−v3∈/W1 has dual counterpart: v3x−v3∈/W2.
Now, as v3x−v3∈/W1∪W2, we can choose elements v4∈W1
and v5∈W2 such that v3x=v3+v4+v5.
Now, it follows from the above construction
that (v1,v2,v3,v4,v5) is an A-basis.
In this basis, the automorphism x is represented by the matrix as shown in Eq. (8).
[TABLE]
Furthermore, it is straightforward to check that each of y1 and y2, defined in Eq. (8), acts on H as an automorphism of A, and therefore, it belongs
to N0. Let M=⟨x,y1,y2⟩. Clearly, M≤N0, and for i∈{1,2}, the basic set Ti is equal to the orbit viM.
Now, let z∈N0∩Nv1. For i∈{1,2}, let us consider the automorphism zUi/Wi∈Aut(AUi/Wi)0. The latter group is generated by the element xUi/Wi,
and we find z^{U_{i}/W_{i}}\in\big{\langle}x^{U_{i}/W_{i}}\big{\rangle}. Moreover, as
v1z=v1, it follows that zU1/W1 is the identity mapping. All these yield that
z can be written in the following form:
[TABLE]
This shows that ∣N0∩Nv1∣≤p2, and therefore,
∣N∣=p8 or p9.
Fix an element v′=kv1+v2 for some k∈{1,…,p−1},
and let T be the basic set of A that contains v′.
Since A=V(H,N0) we can write T=(v′)N0.
It follows from Eqs. (8) and (9) that
W+v′⊆T⊆U+v′. This together with
Lemma 2.14 yields that T=W+v′ or T=U+v′. By Eq. (8), (v′)x=kv1+kv3+v2+v3=v′+(k+1)v3.
Assume at first that k=p−1. Then (v′)x∈/v′+W, and hence T=U+v′. The latter condition together with Theorem 2.9 yields that every basic set of A contained in ⟨U,v′⟩∖U is equal to a U-coset.
If ∣N∣=p8, then
v3N0=v3+⟨v4+v5⟩, and this with the previous paragraph implies that (i) holds with U3=⟨U,−v1+v2⟩.
Finally, suppose that ∣N∣=p9. Then for z with a=1 and
b=0 in Eq. (9), we find
(−v1+v2)z=−v1+v2+v3, thus (−v1+v2)N0=U+(−v1+v2), and (ii) follows.
∎
Lemma 5.4**.**
With the notation of Lemma 5.3, the S-ring A is CI.
Proof.
We keep all notations from the previous proof, and let, in addition,
W3=⟨v4+v5⟩.
Let f∈Iso0(A) such that f fixes all elements vi with i=2, and also −v2. Recall that, (v1,…,v5) is the A-basis defined in the proof of Lemma 5.3.
We settle the lemma by showing that Tf=T for all basic sets T∈Bsets(A). This will be done in five steps.
Claim (a). Tf=T* for all T∈Bsets(A),T⊂U1∪U2.*
This is trivial for the basic sets {v4} and {v5}, and hence
Claim (a) follows for all basic sets
T⊂W=⟨v4,v5⟩.
Let N=NAut(A)(HR).
It follows from the proof of Lemma 5.3 that
any basic set T⊂U∖W is in the form
T=k(W3+v3+w) for some w∈W and some k∈{1,…,p−1} if ∣N∣=p8; whereas in the form T=k(W+v3) for some k∈{1,…,p−1} if ∣N∣=p9.
As both W3+v3 and {w} are
basic sets, we can write (W3+v3+w)f=(W3+v3)f+wf=W3f+v3f+wf=W3+v3+w, where the second equality follows by Lemma 2.10(ii). Now, this together with Lemma 2.12(i) yields
Tf=T in the case when ∣N∣=p8. Similarly,
(W+v3)f=W+v3, and this with Lemma 2.12(i) yields
Tf=T if ∣N∣=p9.
Next, let us consider the isomorphism fU1/W1. Since AU1/W1 is indecomposable, fU1/W1∈Aut(U1/W1), see Lemma 2.17. This together with the fact that f fixes
pointwise a generating set of U1 shows that fU1/W1 is the identity mapping.
Using this and that AU1 is a U1/W1-wreath product, we deduce that Tf=T for all basic sets
T⊂U1∖U. Observe that, the latter statement has dual counterpart: Tf=T
for all basic sets T⊂U2∖U. This completes the proof of Claim (a).
Claim (b). There exist an integer k and a function F2:U2→{0,1,…,p−1} such that
[TABLE]
The S-ring AH/U=QH/U. Form this and the fact that f fixes a basis of H, we deduce that
fH/U is the identity mapping, and therefore, f maps any
U-coset to itself. In particular, it fixes the coset U2+v1.
Let f~ denote the permutation of U2+v1 induced by the action of f on U2+v1.
Choose an arbitrary basic set T∈Bsets(AU2), and let Σ be the
subdigraph of Cay(H,T) induced by the set
U2+v1. By Claim (a), Tf=T, and this in turn implies that
f∈Aut(Cay(H,T)), and f~∈Aut(Σ).
It is straightforward to check that
(v1)R is an isomorphism between Cay(U2,T) and Σ. Therefore, the permutation g∈Sym(U2), defined by g=(v1)Rf~(−v1)R, belongs to
Aut(Cay(U2,T)). As T∈Bsets(AU2) was chosen arbitrarily,
by definition, g∈Aut(AU2). Furthermore, 0g=0(v1)Rf~(−v1)R=0. We have
already shown that \operatorname{Aut}(\mathcal{A}_{U_{2}/W_{2}})_{0}=\big{\langle}x^{U_{2}/W_{2}}\big{\rangle}, and thus we can write
gU2/W2=(xk)U2/W2 for some integer k. This allows us to define the function
F2:U2→{0,1,…,p−1}, by letting F2(u)v5=ug−uxk for each u∈U2.
Then,
uxk+F2(u)v5=ug=u(v1)Rf~(−v1)R=(u+v1)f−v1, and Claim (b) follows.
Recall that N=NAut(A)(HR).
Claim (c). Suppose that ∣N∣=p8. Then for any
u,u′∈U2 with u−u′∈U, F2(u)=F2(u′).
Suppose that u,u′∈U2 such that u−u′∈U.
Recall that g∈Aut(AU2) and g fixes any U-coset. Let us consider the automorphism
gU2/W3. It can be easily seen from the proof of
Lemma 5.3(i) that AU/W3=QU/W3. This implies that gU2/W3 acts on the coset (U+u)/W3=(U/W3)+(W3+u) as a translation by
some element from U/W3. This implies that, ug−u+u′−(u′)g∈W3. It follows from Eq. (8) that
uxk−u+u′−(u′)xk∈W3 also holds, and by the definition of F2 we can write
[TABLE]
On the other hand, ⟨v5⟩∩W3={0}, and this yields
Claim (c).
Notice that, the symmetry between the pairs (U1,W1) and (U2,W2) extends to the triples
(v1,T1,v4) and (−v2,−T2,v5). Here −T2={−v:v∈T2}, which is also a basic set of
A (see the definition of an S-ring). As a consequence, the statements in Claims (b) and (c) have the following dual counterparts:
Claim (d). Suppose that ∣N∣=p8.
Then there exist an integer l and a function F1:U1→{0,1,…,p−1} such that
[TABLE]
Moreover, if u,u′∈U1 with u−u′∈U, then F1(u)=F1(u′).
We are ready to handle the remaining basic sets of A.
Claim (e). Tf=T* for all T∈Bsets(A).*
In view of Claim (a), we can assume that T⊂U1∪U2. If ∣N∣=p9, then T is equal to a U-coset, see
Lemma 5.3(ii). Using this
and that f maps any U-coset to itself, Claim (e) follows at once
if N∣=p9.
In the rest of the proof it will be assumed that ∣N∣=p8.
Let us consider the coefficient of v3 in the linear combination of
(v1−v2)f. By Eq. (10), this follows to be equal to −k. Then, applying Eq. (11), we find that this is equal to l. We conclude that l=−k.
Next, let us consider the element w=(v1−v2+v3)f−(v1−v2)f.
Using Eq. (10) and Claim (c), we find
[TABLE]
On the other hand, using Eq. (11), the fact that l=−k and Claim (d), we find
[TABLE]
We conclude that k=0.
Let U3=⟨U,v1−v2⟩.
We have shown in the proof of Lemma 5.3(i) that the
all basic sets of A not contained in U1∪U2∪U3 are
U-cosets, and all basic sets contained in U3∖U are
W-cosets. Combining this with the fact that f maps any
U-coset to itself, we get that Claim (e) holds whenever T⊂U1∪U2∪U3. It remains to check whether the basic sets contained in U3∖U are fixed by f.
By Theorem 2.9, any such a basic set can be written in the form
T(k)=kT={ku:u∈T}, where k∈{1,…,p−1} and T⊂U3∩(U2+v1).
Then, By Lemma 2.12(i),
we are done if we show that Tf=T. Since T⊂U2+v1, there exists some u∈U2 such that
T=W+v1+u. Now, applying Eq. (10) and using that k=0, we finally get
[TABLE]
This completes the proof of Claim (e), and thus the proof of the lemma as well.
∎
Lemma 5.5**.**
Assuming Hypothesis 5.1, suppose that A is an U/W-wreath product, where ∣U∣=p4,∣W∣=p and AU is indecomposable.
Then the S-ring A is CI.
Proof.
If ∣A∣=p, then each basic set has size at most
p, and therefore, each basic set T⊂H∖U is equal to a W-coset. This implies that AH/W=QH/W.
In particular, AH/W is indecomposable, and we can apply Corollary 4.6 to get that A is a CI-S-ring. For the rest of the proof we assume that ∣A∣≥p2.
Let AU denote the group of automorphisms of U induced by
restricting A to U. We show next that ∣AU∣≤p.
Assume to the contrary that ∣AU∣>p. Since AU is indecomposable, it follows by
Lemma 4.3 that ∣Oθ(A)∣=p2. Fix u1,u2∈Oθ(A) such that
Oθ(A)=⟨u1,u2⟩. Now, let V<U be an A-subgroup such that
∣V∣=p3 and Oθ(A)<V, and fix an element u3∈V∖Oθ(A).
If AU is not semiregular on the orbit u3A, then it follows that
∣CU(z)∣=p3 for any nontrivial z∈(AU)u3. This contradicts Lemma 4.3, and thus u3A=Oθ(A)+u3 and
∣AU∣=p2. There exist unique elements x,y∈A that satisfy
u3x=u1+u3 and u3y=u2+u3. Now, let u be an arbitrary element from
U∖V. Then both ux−u and uy−u are in V. Therefore,
there are integer numbers k,l,m,k′,l′,m′∈{0,1,…,p−1} for which
[TABLE]
Since ∣AU∣=p2, the group AU is abelian, and
uxy=uyx. The coefficients of u1 and u2, resp.,
have to be the same in both sides, and this results in the equalities:
k+k′=k+k′+m′ and l+l′+m=l+l′, resp., which gives that
m=m′=0. This implies that the orbit u⟨x,y⟩=Oθ(A)+u, and
therefore, rad(T)≥Oθ(A) for the basic set of A that contains u. As u was chosen arbitrarily from U∖V, we obtain finally that AU is a
V/Oθ(A)-wreath product. This contradicts the assumption that AU is indecomposable, and by this we have proved that ∣AU∣≤p.
Let us fix T1∈Bsets(A) such that T1⊂H∖U and
∣rad(T1)∣ is the the smallest among all
∣rad(T)∣ where T runs over the set of all basic sets
T∈Bsets(A),T⊂H∖U.
Now, let (v1…,v5) be an A-basis such that
vi∈U for all i<5 and v5∈T1.
Let f∈Iso0(A) such that f fixes vi for all i∈{1,…,5}.
We settle the lemma by showing that Tf=T for all basic sets
T∈Bsets(A).
Since f fixes all vi and U=⟨v1,…,v4⟩, the A-subgroup U is mapped to itself by f.
Then fU is a normalized isomorphism of AU. Since AU is indecomposable, by Corollary 4.5, fU∈Aut(U), which implies that fU is the identity mapping. In particular,
Tf=T for all basic sets T⊂U.
Now, we turn to the basic sets contained in H∖U.
Let
[TABLE]
Let us consider the S-ring AH/V.
Since V≤U and f fixes U pointwise, we get Vf=V,
and thus fH/V∈Iso0(AH/V) (see Proposition 2.10(iii)).
We finish the proof by showing that
fH/V=idH/V. Indeed, then any basic set T⊂H∖U satisfies (T+V)f=T+V, and since V≤rad(T), it follows by Lemma 2.12(ii) that Tf=T.
Case 1. ∣A∣=p2.
Let V1=rad(T1). Clearly, V≤V1.
We show at first that we may choose v5 such that
[TABLE]
Since ∣AU∣≤p, there exists x∈A such that CH(x)=U. Then the ⟨x⟩-orbits not contained in U
coincide with the cosets of a subgroup W′<H of order p, in particular, A is a
U/W′-wreath product. We may assume w. l. o. g. that W′=W. Let A=⟨x,y⟩.
Let us consider the S-ring AH/W. It follows that
AH/W=V(H/W,⟨yH/W⟩), where yH/W is
the automorphism of H/W induced by the action of
y on H/W. In view of Corollary 4.6 we may assume
that AH/W is decomposable. This implies that any
⟨yH/W⟩-orbit is contained in a coset of a fixed subgroup of
H/W of order p. It is not hard to show that this implies
that ∣CH/W(yH/W)∣=p3. Let U′ be the unique subgroup of H that contains W and for which
U′/W=CH/W(yH/W).
Suppose at first that U′=U. Then, any ⟨y⟩-orbit in U is contained
in a W-coset, and hence ∣CU(y)∣≥p3. On the other hand,
AU indecomposable, and it follows
from Lemma 4.3 that y acts as the identity on U. Hence, CH(y)=U, and so CH(A)=U. This shows that any basic set T⊂H∖U is equal to a V-coset, in particular, Eq. (12) follows.
Next, suppose that U′=U, and choose an element
v∈U′∖U. Then the basic set T(v) containing v
is equal to the coset W+v. Thus, ∣rad(T(v))∣=p, which is
clearly minimal among all orders ∣rad(T)∣,T⊂H∖U. Then choosing
v5 to be v, we get T1=T(v)=W+v5 and also W=V1=V, that is, Eq. (12) holds also in this case.
Let U′′=⟨V,v5⟩.
The group H/V decomposes to the internal direct sum H/V=Uˉ+Uˉ′′, where both factors Uˉ=U/V and Uˉ′′=U′′/V are AH/V-subgroups.
To simplify notation, we write vˉ for the coset
V+v, where v is any element in H,
and Sˉ for the set S/V⊆H/V, where
S⊆H. Notice that, fH/V fixes pointwise both Uˉ and
Uˉ′′. Let vˉ∈H/V be an arbitrary element. Then, Uˉ+vˉ=Uˉ+vˉ1 for some vˉ1∈Uˉ′′, and we can write
[TABLE]
Similarly, Uˉ′′+vˉ=Uˉ′′+vˉ2 for
some vˉ2∈Uˉ, and hence
[TABLE]
Therefore,
vˉfH/V−vˉ∈Uˉ∩Uˉ′′={0ˉ}, and fH/V=idH/V, as claimed.
Case 2. ∣A∣≥p3.
Let K denote the kernel of A acting on the set U.
Since ∣AU∣≤p, ∣K∣≥p2. Since K≤Aut(H) and each
element of K fixes U pointwise, there exists a subgroup V′≤U such that ∣V′∣=∣K∣≥p2, and the K-orbits not contained in U are equal to the V′-cosets not contained in U.
Note that, V≥V′, and thus ∣V1∣≥∣V∣≥p2, where V1=rad(T1).
Suppose at first that T1=V1+v5.
By Proposition 2.4, ∣T1∣=p3, which implies that
∣V1∣=p2, and hence V=V1. Also, the S-ring AH/V is an exceptional S-ring over the group H/V, and it follows that fH/V∈Aut(H/V). Since f fixes all generators vi,
it follows that fH/V=idH/V.
Now, suppose that T1=V1+v5. It is sufficient to
show that V=V1. Then the conditions in Eq. (12) hold, and fH/V=idH/V follows as in Case 1.
Since V≤V1,∣V∣≥p2 and ∣V1∣≤p3, the equality V=V1 follows if ∣V1∣=p2. We are left with the case when ∣V1∣=p3. Then, for each basis set T⊂H∖U, ∣rad(T)∣=p3,
implying that T is a coset of a subgroup that contains V.
Assume to the contrary that V=V1. Then ∣H/V∣=p3, and AH/V has two basic sets which are cosets of distinct subgroups of order p. Since ∣H/V∣=p3, the S-ring AH/V is Cayley isomorphic to one of the S-rings given in Table 1. A quick look at the table shows that none of these S-rings has two basic sets which are cosets
of distinct subgroups of order p. Therefore, V=V1, and this completes the proof of the lemma.
∎
Before we prove Theorem 5.2, one more technical lemma is needed.
Lemma 5.6**.**
Assuming Hypothesis 5.1, suppose that A is a U/W-wreath product, where ∣U∣=p4 and ∣W∣=p.
Furthermore, let V be an
A-subgroup such that W<V<U,∣V∣=p3 and ∣Oθ(A)∩V∣≥p2, and let f∈Iso0(A).
Then there exists φ∈Aut(H) such that fφ maps each of U and W to itself, and the following conditions hold:
- (i)
Tfφ/W=T/W* for all T∈Bsets(A).*
2. (ii)
Tfφ=T* for all T∈Bsets(A) with
T⊂V∪(H∖U).*
3. (iii)
If either V⊆Oθ(A), or each
basic set of A contained in V∖Oθ(A) is equal to
a W-coset, then fφ fixes pointwise V.
Proof.
Let us fix five elements of H as follows:
[TABLE]
Clearly, (v1,…,v5) is an A-basis.
Let φ1∈Aut(H) be the automorphism defined
by φ:vif↦vi,i∈{1,…,5}, and
let f1=fφ. It follows that f1∈Iso0(A), and
f1 fixes all vi. In particular, Wf1=W, and thus
f1H/W∈Iso0(AH/W).
The S-ring AH/W is a CI-S-ring. Therefore, there exists some ϕ∈Aut(H/W) such that Tf1/W=(T/W)ϕ for all T∈Bsets(A). Let ϕ1∈Aut(H) such that ϕ1 fixes W pointwise, and
ϕ1H/W=ϕ. Then we have
W+v4ϕ1=(W+v4)ϕ1H/W=(W+v4)ϕ=W+v4f1=W+v4.
Thus, v4ϕ1=v4+w for some w∈W. Now, let ϕ2∈Aut(H) be defined by
[TABLE]
and let φ2=ϕ1ϕ2.
It follows that φ2 satisfies the following:
- (a)
uφ2=u for all u∈⟨v4,v5⟩,
2. (b)
φ2H/W=ϕ.
Let T1∈Bsets(A) such that v3∈T1.
Then T1=X+v3 for some subgroup X≤⟨v4,v5⟩, and
we find T1f1=Xf1+v3f1=X+v3=T1. Thus, T1φ2/W=(T1/W)φ2H/W=(T1/W)ϕ=T1f1/W=T1/W. This gives T1φ2+W=T1+W, and thus
T1φ2−1 is contained in the coset T1+W=X+W+v3.
By (a), Xφ2=X, implying that
T1φ2−1=X+v3φ2−1. This with the previous observation yields that T1φ2−1=T1+w1 for some w1∈W. Now, we define φ3∈Aut(H) by letting
[TABLE]
Let φ=φ1φ2−1φ3−1.
Then fφ=f1φ2−1φ3−1.
It is easily seen that f1 maps each of
U and W to itself. The condition that φ2 maps W to itself
follows from (a) and the fact that W=⟨v5⟩. Then, W≤Uφ2, and for
Uφ2=U it is enough to show that
Uφ2/W=U/W. This follows along the line:
Uφ2/W=(U/W)φ2H/W=(U/W)ϕ=Uf1/W=U/W.
The definition of φ3 shows that it also maps U and W to itself, and therefore,
we obtain that fφ maps each of U and W to itself.
We finish the proof by showing that all conditions (i)-(iii) hold for fφ.
(i): Recall that f1=fφ1.
It follows from (b) that Tf1φ2−1/W=T/W for all
T∈Bsets(A). We claim that
(W+vi)φ3=W+vi for al i∈{1,…,5}.
This is obvious if i=3, or i=3
and X=W. Let i=3 and X=W. Then
T1=W+v3, and since T1f1=T1, it follows that
T1φ2/W=T1f1/W=T1/W, implying that
W+v3φ2−1=W+v3.
Therefore, (W+v3)φ3=Wφ3+v3φ3=W+v3φ2−1=W+v3.
Since H=⟨v1,…,v5⟩, it follows that
(W+x)φ3=W+x for all x∈H. Consequently,
Tfφ/W=Tf1φ2−1φ3−1/W=Tφ3−1/W=T/W for all
T∈Bsets(A), and (i) follows.
(ii): Let T be an arbitrary basic set of A. Note that, if
W≤rad(T), then using that fφ maps W to itself,
we can write W+Tfφ=(W+T)fφ=Tfφ.
Combining this with (i) yields Tfφ=Tfφ+W=T+W=T.
In particular, (ii) holds whenever T⊂H∖U or
T=T1 and W=X.
Also, by (a) and the definition of φ3, fφ fixes pointwise ⟨v4,v5⟩, and this gives that (ii) also
holds when T⊂⟨v4,v4⟩. It remains to consider the
case when T⊂V∖⟨v4,v5⟩. Observe that, T can be written in the form T=kT1+v for some
k∈{1,…,p−1} and some v∈⟨v4,v5⟩. In view of
Eq. (1) and Theorem 2.9,
in order to prove Tfφ=T it is sufficient to show that
T1fφ=T1. We have already shown above that this holds if
X=W. If X=W, then the statement follows along the following line:
[TABLE]
(iii): Since V=⟨v3,v4,v5⟩, we are done if we show that
vifφ=vi for all i∈{3,4,5}. Using that fφ=f1φ2−1φ3−1 and all vi are fixed by f1, it reduces to show
that viφ2−1φ3−1=vi for all i∈{3,4,5}.
This follows immediately from (a) and the definition of φ3 if i=4 or i=5.
The condition V⊂Oθ(A) is equivalent to T1={v3}, and
therefore, T1={v3} or T1=W+v3.
Now, recall that T1φ2−1=T1+w1.
This shows that, if T1={v3}, then v3φ2−1=v3+w1, and so v3φ2−1φ3−1=(v3+w1)φ3−1=v3. Finally, if T1=W+v3, then
v3φ3=v3φ2−1, that is,
v3φ2−1φ3−1=v3.
∎
Everything is prepared to settle the main result of the section.
Proof of Theorem 5.2.
Since A is decomposable, it is a U/W-wreath product where
W<U,∣W∣=p and ∣U∣=p4. Furthermore,
we may assume because of Lemma 5.5 that
AU is a V/X-wreath product where X<V<U,∣X∣=p
and ∣V∣=p3.
Let f∈Iso0(A). We have to show that
there exists some φ∈Aut(H) such that
Tfφ=T for all basic sets T∈Bsets(A).
By Lemma 5.6(i)-(ii), there exists some
φ1∈Aut(H) such that f1=fφ1
satisfies
[TABLE]
and
[TABLE]
Now, if W≤rad(T) for any basic set T⊂V∖U,
then Eq. (13) and Lemma 2.12(ii) imply that
Tf1=T also holds, and hence we are done by letting
φ=φ1.
For the rest of the proof it will be assumed that there exists
some basic set T1⊂U∖V such that W≤rad(T1). Note that, this implies that
∣T1∣=p or ∣T1∣=p2.
We consider below the two cases separately.
Case 1. ∣T1∣=p.
In this case T1=X+u for some u∈U∖V.
Since W≤rad(T1), it follows that X=W. Since X<V, by Eq. (14), Xf1=X, and thus T1f1=Xf1+uf1=X+uf1. Using this and Eq. (13), we conclude that
T1f1+W=T1+W, implying that T1f1 is contained in the coset X+W+u. Thus, T1f1=T1+w1 for some
w1∈W. Choose an automorphism φ2∈Aut(H) satisfying the following:
[TABLE]
Let f2=f1φ2. It is not hard to show that f2 satisfies both Eqs. (13) and (14).
We show below that Tf2=T also holds for each T∈Bsets(A) with T⊆U∖V,
and therefore, will be done by letting φ=φ1φ2.
First, T1f2=(T1+w1)φ2=(X+u+w1)φ2=T1. Let us consider the S-ring AU/X and its isomorphism f2U/X. Note that, the latter isomorphism belongs to
Iso0(AU/X) because f2 maps each of U and X to itself.
The group U/X can be written as the internal direct sum
V/X+⟨X,u⟩/X where both subgroups V/X and ⟨X,u⟩/X are
AU/X-subgroups. By Lemma 2.8,
AU/X=AV/X⊗A⟨X,u⟩/X.
Also, f2U/X fixes any basic
set T′∈Bsets(AU/X) if T′⊂V/X or T′⊂⟨X,u⟩/X. This together with AU/X=AV/X⊗A⟨X,u⟩/X yields that f2U/X fixes all basic sets of AU/X.
Then, using also that AU is a V/X-wreath product, we conclude f2 fixes all basic sets of A contained in
U∖V, as required.
Case 2. ∣T1∣=p2.
Assume for the moment that T1 is equal to a coset X′+u
for some subgroup X′<V,∣X′∣=p2, and ∣W∩X′∣=1.
It follows from Eq. (13) that T1f1=T1+w1 for some w1∈W. Repeating the argument used in Case 1, we find φ2∈Aut(H) such that f2=f1φ2 satisfies
both Eqs. (13) and (14), and also T1f2=T1. Now, if T∈Bsets(A) is an arbitrary basic set such that T⊂U∖V, then it can be written in the form T=k(T1+w2) for some k∈{1,…,p−1} and some
w2∈W. In view of Lemma 2.12(i), Tf2=T follows
because (T1+w2)f2=T1f2+w2f2=T1+w2.
All these show that we are done by choosing φ=φ1φ2.
For the rest of the proof it will be assumed that T1 is not a coset.
Equivalently, AU/X is an exceptional S-ring.
This implies that any basic set in U∖V generates
U, and therefore, V is the only A-subgroup
of order p3 contained in U. This fact will be used later.
By Corollary 4.6, the S-ring AH/W must be decomposable. Equivalently, there exist
A-subgroups Y and U′ such that
W<Y<U′,∣Y∣=p2 and ∣U′∣=p4, and
AH/W is a (U′/W)/(Y/W)-wreath product.
Case 2.1. U′=U.
Assume that Y=W+X. In this case A is a U/X-wreath
product, and we can finish this case by replacing first
W with X, and then apply the argument used right after
Eq. (14).
Now, suppose that Y=W+X. Since W<Y, this gives X∩Y={0} and ∣X+Y∣=p3.
Then Y<U′=U. We obtain that X+Y is an A-subgroup of
order p3 contained in U. As it was noted above, this forces that X+Y=V, in particular,
Y<V. It follows that, either V⊆Oθ(A), or
any basic set contained in V∖(W+X) is equal to
a W-coset. By Lemma 5.6(iii), f1 fixes pointwise V.
Let us consider f1U/X, the isomorphism of AU/X
induced by f1 acting on U/X. Then,
f1U/X∈Iso0(AU/X), and since AU/X is exceptional, f1U/X∈Aut(U/X), see Lemma 2.17.
Also, as f1 fixes pointwise V,
f1U/X centralizes V/X. This implies that
f1U/X acts on U/X∖V/X as a translation by some element X+u1 where u1∈V. In particular, T1f1/X=(T1+u1)/X, and
thus T1f1+X=T1+u1+X. Since X≤rad(T1), it
follows that X=Xf1≤rad(T1f1), and we can
write T1f1=T1f1+X=T1+u1+X=T1+u1.
Using also that ∣X∩Y∣=1, we can further write
T1+u1=T1+u2 for some u2∈Y.
Fix an element u∈U∖V, and then choose an
automorphism φ2∈Aut(H) such that
[TABLE]
Let f2=f1φ2. Using that u2∈Y and Y≤rad(T)
for all T∈Bsets(A) with T⊂H∖U,
it is not hard to show that f2 satisfies Eq. (14). Also, T1f2=(T1+u2)φ2=T1.
Finally, let T∈Bsets(A) be an arbitrary basic set such that
T⊂U∖V. Then, T can be written in the form
T=T1+w2,w2∈W, and thus we can write Tf2=T1f2+w2f2=T1+w2=T.
All these show that we are done by choosing φ=φ1φ2.
Case 2.2. U′=U.
Recall that, V is the only
A-subgroup of order p3 contained in U. This implies
that U∩U′=V. The radical rad(T1/W)=(X+W)/W.
On the other hand, Y/W≤rad(T1/W), and thus Y=X+W.
This shows that A is a U′/X-wreath product.
The S-ring AU′ is a U′/W-wreath product, and we
may assume that AU′/W is indecomposable. Observe that,
letting U1=U,U2=U′,W1=X and W2=W,
conditions (1) and (2) of Lemma 5.3 hold. Therefore, in view of Lemma 5.4,
we may assume that condition (3) does not hold, that is,
A acts regularly on any of its orbits not contained in
U∪U′. Now, fix an A-basis (v1,…,v5) as follows:
[TABLE]
Then, let ψ be the automorphism of H defined by
ψ:vif↦vi for all i∈{1,…,5}, and let
f3=fψ. We finish the proof by showing that Tf3=T for all basic sets T∈Bsets(A). One can settle the equality Tf3=T for T⊂U∪U′ by copying the argument used in the proof of
Claim (a) in the proof of Lemma 5.4.
Now, suppose that T∈Bsets(A) with T⊂H∖(U∪U′). By Lemma 2.18, T is contained in both a
U-coset and a U′-coset, hence it is contained in a V-coset, recall that V=U∩U′. We claim that T is, in fact, equal to a V-coset.
Assume to the contrary that T is properly contained in a V-coset.
In particular, we have ∣T∣≤p2.
Let G=HRA, N=NAut(A)(HR), and N0 be the stabilizer of [math] in N. Then, Aut(A)=G(2) and CHR(A)≤Z(G). By Proposition 2.1, CHR(A)≤Z(Aut(A)), and
hence CHR(A)≤CHR(N0). The S-ring
A can be expressed as A=V(H,N0) such that
p2≤∣CH(N0)∣. Now, we can replace A with N0, and
assume, in addition, that N0 is regular on T. Therefore,
∣N0∣=∣T∣≤p2. On the other hand, N0 contains two
elements x and x′ such that CH(x)=U and
x acts on the elements v∈H∖U as
the right translation vx=v+w for a fixed nonzero
w∈W, and CH(x′)=U′ and
x′ acts on the elements v∈H∖U′ as
the right translation vx=v+w′ for a fixed nonzero
w′∈X. It is easily seen that
V(H,⟨x,x′⟩)=A=V(H,N0), implying that
⟨x,x′⟩<N0. This, however, contradicts the previously
obtained bound ∣N0∣≤p2, and by this we have proved that
T is indeed equal to a V-coset.
Finally, let us consider the S-ring AH/V, and the induced
isomorphism f3H/V∈Iso0(AH/V). It is easily seen
that AH/V=Q(H/V),
and therefore, f3H/V∈Aut(H/V). Using this and that
f3 fixes all vi,i∈{1,…,5}, we find
f3H/V=idH/V. Thus, (T+V)f3=T+V, and combining
this with V≤rad(T), Lemma 2.12(ii) gives that Tf3=T. This completes the proof of Case 2.2.
∎
6. Proof of Theorem 3.5 II: The indecomposable S-rings
In this section, we turn to the indecomposable S-rings in
Theorem 3.5 and prove the following theorem:
Theorem 6.1**.**
Assuming Hypothesis 5.1, suppose that A is indecomposable.
Then A is a CI-S-ring.
Our main result Theorem 3.5 follows then as the consequence of Theorems 5.2 and 6.1.
Theorem 6.1 will be proved in the end of the section after six preparatory lemmas.
In the first two lemmas we derive some properties of indecomposable p-S-rings
with thin radical of order at least p2.
Lemma 6.2**.**
Let B be an indecomposable p-S-ring
over a group H≅Zp5 such that ∣Oθ(B)∣≥p2,
and let x∈NAut(B)(HR) such that x=idH and
0x=0. Then ∣CH(x)∣≤p3.
Proof.
Since x=idH, it follows that CH(x)≤p4.
Assume to the contrary that ∣CH(x)∣=p4.
Let U=CH(x), and for a fixed v1∈H∖U, let
W=⟨v1x−v1⟩. Then ∣W∣=p, and it
follows that the orbit v⟨x⟩=W+v for all v∈H∖U.
It can be shown in the same way as in the proof of
Lemma 4.3 that neither U nor W are B-subgroups.
Let V1≤Oθ(B) such that ∣V1∣=p2 and let
V2=V1+W. Then V1<U,∣V1∩W∣=1, and
therefore ∣V2∣=p3.
Let U′ be a B-subgroup of order p4 such that
V1<U′. Then U=U′, and thus there exists
u∈U′∖U. Let T∈Bsets(B) such that u∈T.
Then W+u⊆T⊆U′, implying that
W≤U′. We conclude that U∩U′=V1+W=V2.
Let xU′ denote the restriction of x to the B-subgroup
U′. Clearly, CU′(xU′)=V2, and thus by Lemma 4.3, BU′ is a V′/W′-wreath product for
some B-subgroups 0<W′<V′<U′, where ∣W′∣=p and
∣V′∣=p3. Let T∈Bsets(B) such that
T⊆U′∖V′ and T⊂U.
Then using that T contains a W-coset and W′≤rad(T), it
follows that T contains a (W+W′)-coset (recall that W′=W as
W is not an A-subgroup). This together with Proposition 2.18(ii)
yields that W≤rad(T)≤V′, and thus W<V′. On the other hand it is clear that V1<V′, and this with the previous
observation yields V′=W+V1=V2. In particular, V2 is
a B-subgroup. Since U is not a B-subgroup, the S-ring
BH/V2≅QZp≀QZp, and this implies that
U′ is the only B-subgroup which has order p4 and
contains V2. This property will be used later. Note also that,
[TABLE]
Next, let us consider the S-ring BH/V1.
Let T∈Bsets(B) such that T∩U=∅ and
T⊂U. The latter condition implies that T contains
some W-coset, and thus ∣T/V1∣≥p2. It follows by
Proposition 2.14, that T/V1 is equal to a U′/V1 coset.
It is easy to see that AU′/V1≅QZp≀QZp;
and we conclude that
BH/V1≅QZp≀QZp≀QZp.
Now, fix a coset U′+v1 distinct from U′, and let
T1∈Bsets(B) such that T1⊆U′+v1.
Let T∈Bsets(B) be another basic set contained in U′+v1.
Since BH/V1≅QZp≀QZp≀QZp,
it follows that both T1∩V1+v1 and T∩V1+v1 are nonempty. Choose u1∈T1∩V1+v1 and u2∈T∩V1+v1. Then, u2−u1∈V1≤Oθ(B), hence by Eq. (1), T1+u2−u1=T. Thus,
rad(T)=rad(T1), and combining this with Theorem 2.9, we conclude that rad(T′)=rad(T1) for any basic set
T′⊂U′. Since B is indecomposable, it follows
that rad(T1) is trivial.
This together with Lemma 2.18(ii) yields that
[TABLE]
Indeed, if u1,u2∈T1∩V1+v, then for W=⟨u1−u2⟩,
∣T1∩W+u2∣≥2, and so W≤rad(T1), a contradiction.
Thus, ∣T1∩V1+v∣≤1, and since ∣T1∩V1+v∣≥1 also holds, see the above paragraph, Eq. (16) follows.
By Eq. (15), we can choose a subgroup W1<V1 which satisfies the following property: for all T∈Bsets(B),T⊂U′∖V2, either
∣T∣=p3 or W1<rad(T). Notice that, this property
implies that BU′/W1≅QZp≀QZp≀QZp.
Let us consider xH/W1, the automorphism of H/W1
induced by x acting on H/W1. By Lemma 4.2,
x cannot be in the kernel of Aut(B) acting on H/W1,
and so xH/W1=idH/W1. Then,
CH/W1(xH/W1)=U/W1, and thus by Lemma 4.3, BH/W1 is an (X/W1)/(Y/W1)-wreath product for some B-subgroups X and Y, where
∣X∣=p4, ∣Y∣=p2 and W1<Y<X.
Notice that, X∩U′ is a B-subgroup of order p3.
On the other hand, since BU′/W1≅QZp≀QZp≀QZp, V2 is the only B-subgroup in U′ that has order p3 and contains W1.
Using this and that W1<X∩U′, it follows that
V2=X∩U′. We have shown above that
U′ is the only B-subgroup containing V2, hence X=U′.
Thus, W1<Y<U′, and using again that
BU′/W1≅QZp≀QZp≀QZp, we can see
that V1 is the only B-subgroup in U′ that
has order p2 and contains W1, and so Y=V1.
To sum up, BH/W1 is an
(U′/W1)/(V1/W1)-wreath product.
Recall that, T1 is a basic set contained in the coset
U′+v1. Then, T1 satisfies
V1/W1<rad(T1/W1). In particular, for any u∈T1,
∣T1∩(V1+u)∣≥∣T1/W1∩(V1+u)/W1∣=p, which,
however, contradicts Eq. (16).
This completes the proof of the lemma.
∎
Lemma 6.3**.**
Let B be an indecomposable non-CI p-S-ring
over a group H≅Zp5 such that ∣Oθ(B)∣≥p2.
Then ∣NAut(B)(HR)∣≥p7.
Proof.
Let N=NAut(B)(HR). Assume to
the contrary that ∣N∣<p6. Let
K≤N be the subgroup given
in Proposition 3.1.
The stabilizer (KHR)0≤N0,
hence we can write
[TABLE]
Since K=HR, we can choose a non-identity element
x∈(KHR)0. Then the above inequality yields
∣CH(x)∣≥∣K∩HR∣≥∣K∣/p=p4.
This, however, contradicts Lemma 6.2.
∎
The key step in proving Theorem 6.1 will be to show that, if
A is non-CI, then Aut(A)∩Aut(H) contains a subgroup
L such that ∣L∣=p2 and ∣CH(L)∣=p3.
The next three lemmas are devoted to the arising S-ring V(L,H).
Lemma 6.4**.**
Assuming Hypothesis 5.1,
suppose that A is indecomposable,
and let L≤Aut(A)∩Aut(H) such that ∣L∣=p2
and ∣CH(L)∣=p3. Then the following conditions hold:
- (i)
The S-ring V(H,L) is indecomposable.
2. (ii)
Let T be a basic set of V(H,L) such that ∣T∣>1.
Then T is equal to an X-coset for some subgroup
X<CH(L) of order ∣X∣=p2.
3. (iii)
Let T and T′ be two basic sets of V(H,L) of size p2 for which ⟨T,CH(L)⟩=⟨T′,CH(L)⟩. Then
rad(T)=rad(T′).
Proof.
Let N=NAut(A)(HR) and N0 be the stabilizer of
[math] in N. Note that, L≤N0 and A=V(H,N0).
Furthermore, we let B=V(H,L) and U=CH(L).
(i): Assume to the contrary that B is a X/Y-wreath product, where X>Y,∣X∣=p4 and ∣Y∣=p.
Let x∈Aut(H) be defined
by vx=v for all v∈X, and vx=v+v1 for all
v∈H∖X, where v1∈Y is a fixed nonzero element. Then CH(x)=X and x∈Aut(B). Also, as L≤N0,
B=V(H,L)⊇V(H,N0)=A, and thus
Aut(B)≤Aut(A) . In particular, x∈Aut(A), which
contradicts Lemma 6.2.
(ii): As ∣L∣=p2, ∣T∣≤p2, and if ∣T∣<p2, then Lv is nontrivial, where v∈T, and Lv is the stabilizer of v in L.
Then for x∈Lv, v∈CV(x), and thus
∣CV(x)∣≥∣⟨U,v⟩∣=p4, which contradicts Lemma 6.2, recall that B is indecomposable. We deduce that
∣T∣=p2. Note that, since there is no basic set of size p,
it follows that every B-subgroup of order p2 must be contained
in U. This fact will be used later.
Let V<H be a B-subgroup such that U<V and ∣V∣=p4.
If T⊂V, then it easy to see that T is equal to a coset of a subgroup of U of order p2, and (ii) follows.
Now, suppose that
T⊂U.
By Lemma 2.18(iii), U∩rad(T)={0}, and thus we can choose W<U such that ∣W∣=p and W≤rad(T).
Let us consider the S-ring BH/W.
Then T/W is a basic set of BH/W of size ∣T/W∣=p.
Denote by LH/W the subgroup of Aut(BH/W) induced
by L acting on H/W.
By Lemma 4.2, ∣LH/W∣=p2. It follows from
this and Proposition 2.4 that
T/W cannot generate H/W. This together with the fact that
W≤rad(T) shows that ⟨T⟩=H, and thus ∣⟨T⟩∣=p3
or ∣⟨T⟩∣=p4. If ∣⟨T⟩∣=p3, then it is easily seen that T is equal to a coset of a subgroup of U, and so (ii) follows.
Assume that ⟨T⟩=p4, and let V′=⟨T⟩.
We show below that this case cannot occur.
If U<V′, then it is easy to see that T is equal to a coset of a
subgroup of U, contradicting that ⟨T⟩=p4.
Thus, ∣U∩V′∣=p2, and H can be expressed as the internal
direct sum H=V′+X for some subgroup X<U,∣X∣=p.
Note that, as X≤Oθ(B), it follows from
Lemma 2.8 that B=BV′⊗BX.
Let Y<V′ be a B-subgroup of
order p3 such that U∩V′<Y. Since ⟨T⟩=V′,
T⊂Y. The radical rad(T)=U∩V′, for otherwise,
T cannot generate V′. It follows that the basic sets of B contained in V′∖Y are in the form
k(T+u) for some k∈{1,…,p−1} and some
u∈U∩V′. Since W≤rad(T), we obtain that
BV′ is a Y/W-wreath product. This implies that
B=BV′⊗BX is a (Y+X)/W-wreath product, which contradicts (i).
(iii): Assume to the contrary that rad(T)=rad(T′) and
⟨U,T⟩=⟨U,T′⟩. Let X=rad(T), and let us consider the
S-ring BH/X.
By (ii), both T and T′ are X-cosets. Since ⟨U,T⟩=⟨U,T′⟩, we find that the elements T/X and T′/X generate a subgroup of H/X of order
p2, and this subgroup intersects U/X trivially.
We conclude that BH/X≅QZp3. Consequently, every basic set of B is contained in some X-coset.
This together with (ii) shows that B is an U/X-wreath product,
which contradicts (i).
∎
Lemma 6.5**.**
With the notation of Lemma 6.4, ∣Aut(V(H,L))∣=p8 .
Proof.
As in the previous lemma, we let B=V(H,L) and U=CH(L).
We start with fixing a suitable B-basis.
Fix an element v1∈H∖U, and another
v2∈H∖⟨U,v1⟩. For i=1,2, Let Ti∈Bsets(B) such that vi∈Ti. By Lemma 6.4(ii)-(iii),
Ti−vi is a subgroup of U of order p2.
It will be convenient to denote these subgroups by U∞ and
U0, namely, we let T1=U∞+v1 and T2=U0+v2.
Then ∣U∞∩U0∣=p. Let v4∈U∞∩U0,v4=0. Then, there exist x,y∈L satisfying v2x−v2=v1y−v1=v4. Let v3=v1x−v1 and v5=v2y−v2. We prove next that ⟨v1,v2,v3,v4,v5⟩=H and ⟨x,y⟩=L.
For the first part it is enough to show that ⟨v3,v4,v5⟩=U.
Now, v1x∈T1=U∞+v1, hence v3∈U∞.
Suppose for the moment that v3=kv4 for some integer
k. Then
(kv2)x−kv2=kv4=v3=v1x−v1,
implying that kv2−v1 is fixed by x, and hence
kv2−v1∈CH(x)=U, which is impossible. We conclude that
U∞=⟨v3,v4⟩.
We obtain by a similar argument that U0=⟨v4,v5⟩, and therefore, ⟨v3,v4,v5⟩=U∞+U0=U.
For the second part, if y=xm for some integer m, then
we can write v2+v5=v2y=v2xm=v2+mv4,
contradicting that ⟨v4,v5⟩=U0 has order p2.
Thus, ⟨x,y⟩=L, as required. It is clear that
(v1,…,v5) is a B-basis.
Let V=⟨v1,v2⟩.
Then H can be written as the internal direct sum H=V+U.
For w∈H, let wV and wU denote the projection of
w into V and U, resp.
Furthermore, let I=GF(p)∪{∞}; and for
i∈I, define the elements v^i∈V, and subgroups
Ui≤U as follows
[TABLE]
Let G=Aut(B) and Gw be the stabilizer of an element w in
G. Observe that, the lemma is equivalent to show that
∣G0∣=p3. We are going to derive this in six steps.
Claim (a). The basic sets of B not contained in
U are in the form
[TABLE]
By definition, the basic sets in question are equal to the L-orbits
wL,w∈H∖U, where L=⟨x,y⟩. Now,
w=jv^i+u for some j∈GF(p)∖{0},i∈I and
u∈U. A direct computation yields that the L-orbit
v^iL=Ui+v^i. This together with the fact that
u∈CH(L) yields Claim (a).
Let Fun0(V,U) denote the set of all functions F:V→U such that F(0)=0. For F∈Fun0(V,U), we define the permutation gF∈Sym(H) as follows:
[TABLE]
where wV denotes the projection of w to V (recall that,
we have H=V+U).
Claim (b). For every g∈G0,
g=gF for some F∈Fun0(V,U).
Since U is a B-subgroup, the set H/U form a
block system for G. Let us consider gH/U, the permutation of H/U induced by g acting on H/U.
Then, gH/U∈Aut(B/U). It is easy to see that
B/U=QH/U, and thus we get that gH/U=idH/U.
Equivalently, g fixes setwise every U-coset.
On the other hand, by Eq. (4), g centralizes uR
for all u∈U. These two facts imply Claim (b).
Claim (c). For every F∈Fun0(V,U),
gF∈G0 if and only if the following conditions hold:
[TABLE]
Let F∈Fun0(V,U). By definition, gF∈G0 if and only if
gF∈Aut(Cay(H,T)) for any basic set T∈Bsets(B).
It is clear that gF centralizes uR for all u∈U.
Hence, gF∈Aut(Cay(H,T)) whenever T⊂U.
Now, suppose that T⊂U. By Claim (a),
T=Ui+jv^i+u for some j∈GF(p)∖{0},i∈I and u∈U.
Therefore, gF∈Aut(Cay(H,T)) if and only if
[TABLE]
By Eq. (18), this reduces to
[TABLE]
Equivalently, F(v+jv^i)−F(v)∈Ui for all v∈V.
Since,
[TABLE]
it follows that
F(v+jv^i)−F(v)∈Ui for all v∈V if and only if
F(v+v^i)−F(v)∈Ui for all v∈V,
and Eq. (19) follows.
Claim (d). *If i,j,k∈I are pairwise distinct,
and u1,u2,u3∈U are arbitrary elements, then
∣Ui+u1∩Uj+u2∩Uk+u3∣=1.
*
It is not hard to show that Claim (d) follows from
Ui∩Uj∩Uk={0}.
Let αv3+βv4+γv5∈Ui∩Uj∩Uk.
Suppose at first that none of i,j and k is equal to ∞.
Then, using the definition of the subgroups Ui,Uj and Uk,
we find α1,α2 and α3 in GF(p) such
that
[TABLE]
Using also that i,j and k are pairwise distinct, we deduce that
β=γ(i+j)=γ(i+k)=γ(j+k), and hence
α=β=γ=0, and so Ui∩Uj∩Uk={0}.
Now, suppose that k=∞. Since U∞=⟨v3,v4⟩,
γ=0, and
αv3+βv4=α1(iv3+v4)=α2(jv3+v4).
Since i=j, it follows that
α1=α2=α=β=0, and Ui∩Uj∩Uk={0}, as required.
Claim (e). ∣G0∩Gv1∣≤p.
Let g∈G0∩Gv1. By Claim (b), g=gF for some
F∈Fun0(V,U). Notice that, F(0)=F(v1)=0.
Let us consider the image F(2v1).
Then, we can express 2v1 as 2v1=v1+v^∞, hence
by Eq. (19), F(2v1)−F(v1)∈U∞. Also,
v1+v2=2v1+v^−1 and v2=2v1+v^−2, and using again Eq. (19),
we find F(v1+v2)−F(2v1)∈U−1 and F(v2)−F(2v1)∈U−2.
All these yield
[TABLE]
On the other hand, F(v2)∈U0∩U−1=⟨−v4+v5⟩
and F(v1+v2)∈U0∩U1=⟨v4+v5⟩.
Using also that F(v1+v2)−F(v2)∈U∞=⟨v3,v4⟩,
we find F(v2)=α(−v4+v5) and F(v1+v2)=α(v4+v5) for some α∈GF(p).
Substitute these in Eq. (20). After a
direct computation we find F(2v1)=2αv3.
This shows that the orbit of 2v1 under the group G0∩Gv1 has size at most p. Therefore,
∣G0∩Gv1∣≤∣G0∩Gv1∩G2v1∣⋅p,
and to derive Claim (e) it is enough to show that
∣G0∩Gv1∩G2v1∣=1.
Now, choose g∈G0∩Gv1∩G2v1.
Then, F(0)=F(v1)=F(2v1)=0, thus applying Eq. (19) to F(iv1+v2),i∈GF(p), and using Claim (d),
we find F(iv1+v2)∈Ui∩Ui−1∪Ui−2={0}.
Therefore, F(iv1+v2)=0 for all i∈GF(p).
In particular, F(v2)=F(v1+v2)=F(2v1+v2)=0, and we
can repeat the same argument to have F(iv1+2v2)=0 for all
i∈GF(p). Since V=⟨v1,v2⟩,
the process can be continued to cover all v∈V, and this leads to that F(v)=0 for all v∈V, that is, g=idH, as required.
Claim (f). ∣G0∣=p3.
The G0-orbit of v1 is the basic set U∞+v1.
This together with Claim (e) shows that ∣G0∣≤p2⋅∣G0∩Gv1∣≤p3. To settle Claim (f) it is enough to
find a non-trivial automorphism g∈G0∩Gv1.
We claim that gF is such an automorphism where F is defined
as follows:
[TABLE]
Then, F(0)=0, and by Claim (c), we have gF∈G0 if
F satisfies the conditions in Eq. (19).
This can be verified directly. After letting
v=iv1+jv2 and using the above definition of
F, we compute for k∈GF(p),
[TABLE]
These show that the conditions in Eq. (19) hold, and
gF∈G0. Also, F(v1)=0 and F(2v1)=2v3, and hence
gF is a non-trivial element in G0∩Gv1. This completes
the proof of the lemma.
∎
Lemma 6.6**.**
With the notation of Lemma 6.4, V(H,L) is a CI-S-ring.
Proof.
We keep all notations from the previous proof, that is,
[TABLE]
In addition, let N=NG(HR).
In view of Lemma 2.11, it is enough to show that all
regular subgroups of G isomorphic to H are conjugate in G.
First, the number of subgroups of G that are conjugate to HR is equal to
the index ∣G:N∣. By Lemma 6.5, ∣G∣=p8, and since
HRL≤N, it follows that ∣N∣≥p7. If G=N, then for every non-trivial element
z∈G0∩Gv1, CH(z)=⟨v1,v3,v4,v5⟩, contradicting Lemma 6.2.
Thus, ∣N∣=p7, and there are exactly
p subgroups of G that are conjugate to HR.
Therefore, to finish the proof it is sufficient to show that there are exactly p regular subgroups of G isomorphic to H.
Note that, we have L=N0.
Let K≤G be any regular subgroup isomorphic to HR
such that K=HR. Let M=⟨K,HR⟩. Since K=HR, ∣M∣≥p6. This implies that ∣M∩L∣>1.
Indeed, if ∣M∣=p6, then HR⊴M, and hence
M0=1 and M0≤N0=L. If ∣M∣>p6, then
∣M∩L∣>1 follows because ∣L∣=p2 and ∣LM∣≤∣G∣=p8. Using also that
K∩HR≤Z(M) and Lemma 6.2, we deduce that
∣K∩HR∣≤p3. On the other hand, since K is regular and abelian, it follows that
Z(G)≤K, and hence K∩HR=UR. Note that, we have proved that every regular subgroup isomorphic to H intersects HR at UR, unless it is equal to HR.
This fact will be used in the next paragraph.
We claim that K≤N. Suppose to the contrary that
there exists some g∈K∖N.
Then g=z1vR for some z1∈G0∖L and
v∈H∖U. On the other hand,
∣N∩K∣≥p4, and
and thus K contains an element in the form
z2wR,z2∈L and w∈H∖U. Clearly, since UR≤K, the element
w cannot be in U. Then,
z1z2(vR)z2wR=z1vRz2wR=z2wRz1vR=z2z1(wR)z1vR. By Lemma 6.5, G0 is abelian, and
we get (wR)z1=(vR)z2wR(vR)−1=(vz2+w−v)R.
Thus, (wR)z1∈HR, and since w∈/U, (wR)z1∈/UR, and
∣HRz1∩HR∣≥p4. Now, it follows by the previous
paragraph that HRz1=HR, and hence z1∈L, a contradiction.
Now, there exist z1,z2∈L such that
[TABLE]
Recall that, the L-orbit of v1 is in the form
v1L=U∞+v1, and the L-orbit of v2 is in the form
v2L=U0+v2. Let u=v1z2−v1. Clearly, u∈U∞.
Then, since z1(v1)R and
z2(v2)R commute, 0z1(v1)Rz2(v2)R=v1z2+v2
and 0z2(v2)Rz1(v1)R=v2z1+v1, it follows that u=v2z1−v2.
This shows that u∈U0 also holds, and hence u∈U∞∩U0=⟨v4⟩.
Now, as L is regular on both orbits v1L and v2L, the automorphisms z1 and z2
are uniquely determined by u, and thus K is determined as well.
This yields that there are exactly p regular subgroups of G
isomorphic to H. This completes the proof of the lemma.
∎
Lemma 6.7**.**
Assuming Hypothesis 5.1, suppose that A is indecomposable,
and let U and W be
A-subgroups such that W<Oθ(A)<U,∣W∣=p
and ∣U∣=p4. Then A has a basic set T such that
[TABLE]
Proof.
Since A is indecomposable, there exists a basic set
T1⊂H∖U such that W≤rad(T1).
It is clear that ∣T1∣>1.
We have to show that ∣T1∣≤p2. To the contrary
assume that ∣T1∣≥p3. If ∣T1∣=p4, then A is
decomposable, see Proposition 2.14, thus ∣T1∣=p3.
This together with ∣Oθ(A)∩U∣=∣Oθ(A)∣≥p2 gives
∣Oθ(A)∩U∣⋅∣T1∣>p4=∣H∣/p, and
we can apply Lemma 2.18(iii) to obtain that
Oθ(A)∩rad(T1)={0}. Let W′≤Oθ(A)∩rad(T1) such that ∣W′∣=p. Since W≤rad(T1), we get, using
Eq. (1), pairwise distinct
basic sets in the form T1+w,w∈W. As the union of the
latter basic sets is equal to the coset U+v1, it follows that W′≤rad(T) for all T∈Bsets(A) with T⊂U+v1. This together with Theorem 2.9 yields that W′≤rad(T) for all T∈Bsets(A) with T⊂U, that is, A is a U/W′-wreath product, a contradiction.
∎
Everything is prepared to settle the main result of the section.
Proof of Theorem 6.1.
Let U=Oθ(A), N=NAut(A)(HR) and N0 be the stabilizer
of [math] in N. Assume to the contrary that A is a non-CI-S-ring.
We prove first the following:
[TABLE]
If ∣U∣≥p3, then we are done by choosing L
to be any subgroup of N0 of order p2 (see also
Lemma 6.3). Thus we assume for the moment that
∣U∣=p2. Let K≤N be the subgroup given in Proposition 3.1, and let M=(KHR)0. If ∣M∣≤p2, then
[TABLE]
By Lemma 6.2, M must have order p2, and therefore, we are done by choosing
L to be M.
Let ∣M∣≥p3. It follows from Lemma 6.7 that there exists a basic set T such that ∣T∣≤p2 and rad(T)=U.
Let v∈T, and Mv be the stabilizer of v in M.
Then CH(Mv)≥⟨U,v⟩. If ∣T∣=p or the orbit vM=T,
then it follows that ∣Mv∣≥p2. Using this and that
∣⟨U,v⟩∣=p3, we can choose
L to be any subgroup of Mv of order p2.
Now, suppose that ∣T∣=p2, and the orbit vM=T.
Choose a non-identity element x0∈Mv, and let u∈T be
an arbitrary element. Then u=vx for some x∈M, and since M is abelian, we can write ux0=vxx0=vx0x=vx=u. As a corollary we find CH(x0)≥⟨U,T⟩.
Clearly, ∣⟨U,T⟩∣≥p3; in fact, ⟨U,T⟩∣=p3 must hold by Lemma 6.2. It follows that T is equal to a
U-coset, that is, rad(T)=U. This is a contradiction, and
Eq. (22) follows.
By Lemma 6.6, the S-ring V(H,L) is a CI-S-ring. Therefore,
A=V(H,L), hence A>L, in particular, ∣A∣≥p3.
For sake of simplicity we let V=Oθ(V(H,L)). Clearly, U≤V and
∣V∣=p3. Fix W1<U,∣W1∣=p. By Lemma 6.7, there exists a basic set T1∈Bsets(A) such that 1<∣T1∣≤p2 and W1<rad(T1). Since every basic set of V(H,L) outside V is a coset of a subgroup of V of order p2, we find that either
T1 is contained in H∖V,
∣T1∣=p2 and rad(T1)<V, or T1≤V.
In the former case V=rad(T1)+W1, whereas in the latter case
V=⟨T1,W1⟩ because W1<rad(T1). We conclude that V is an A-subgroup. This shows that we may
choose the above T1 such that
T1⊂H∖V.
Fix some v1∈T1.
Since ∣N0∣≥p3, there exists a non-identity element
x∈N0 such that v1x=v1, and thus
CH(x)≥⟨U,v1⟩, and so ∣CH(x)∣≥p⋅∣U∣.
This together with Lemma 6.2
shows that ∣U∣=p2, in particular, U<V.
Now, using also that W1<U and W1<rad(T1),
we infer in turn that ∣U∩rad(T1)∣=p,
Arad(T1)=QCp≀QCp, and finally that,
AV=(QCp≀QCp)⊗QCp.
Let W2=U∩rad(T1).
It is easy to see that every A-subgroup of order p2 contained in V contains W2.
(In fact, such an A-subgroup intersects rad(T1) at exactly W2.)
Now, apply Lemma 6.7 with W=W2.
We obtain that, there exists a basic set T2 of A such that
T2⊂V,∣T2∣=p2 and W2<rad(T2). As before, rad(T2) is an A-subgroup of order p2 contained V, contradicting our earlier observation that such an A-subgroup must contain W2. This completes the proof of the theorem.
∎