This paper classifies edge-primitive, 2-arc-transitive graphs with soluble edge-stabilizers, advancing understanding of their symmetry properties and automorphism groups.
Contribution
It provides a complete classification of a specific class of highly symmetric graphs with soluble edge-stabilizers.
Findings
01
Classification of edge-primitive, 2-arc-transitive graphs with soluble edge-stabilizers
02
Identification of automorphism group actions on these graphs
03
Characterization of their symmetry and structural properties
Abstract
A graph is edge-primitive if its automorphism group acts primitively on the edge set, and 2-arc-transitive if its automorphism group acts transitively on the set of 2-arcs. In this paper, we present a classification for those edge-primitive graphs which are 2-arc-transitive and have soluble edge-stabilizers.
\begin{array}[]{l|l|l}X_{v}^{{\it\Gamma}(v)}&(X_{v}^{{\it\Gamma}(v)})_{u}&\\
\hline\cr{\rm Sz}(q).e&p^{f{+}f}{:}(q-1).e&e\mbox{ a divisor of }f,\,p=2,\,\mbox{odd }f>1\\
\hline\cr{\rm Ree}(q).e&p^{f{+}2f}{:}(q-1).e&e\mbox{ a divisor of }f,\,p=3,\,\mbox{odd }f>1\end{array}
\begin{array}[]{l|l|l}X_{v}^{{\it\Gamma}(v)}&(X_{v}^{{\it\Gamma}(v)})_{u}&\\
\hline\cr{\rm Sz}(q).e&p^{f{+}f}{:}(q-1).e&e\mbox{ a divisor of }f,\,p=2,\,\mbox{odd }f>1\\
\hline\cr{\rm Ree}(q).e&p^{f{+}2f}{:}(q-1).e&e\mbox{ a divisor of }f,\,p=3,\,\mbox{odd }f>1\end{array}
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Full text
On edge-primitive graphs with soluble edge-stabilizers
Hua Han
H. Han
School of Science, Tianjin University of Technology
A graph is edge-primitive if its automorphism group acts primitively on the edge set, and 2-arc-transitive if its automorphism group acts transitively on the set of 2-arcs. In this paper, we present a classification for those edge-primitive graphs which are 2-arc-transitive and have soluble edge-stabilizers.
Supported by the National Natural Science Foundation of China (11971248, 11731002) and the Fundamental Research Funds for the Central Universities.
1. Introduction
In this paper, all graphs are assumed to be finite and simple,
and all groups are assumed to be finite.
A graph is a pair Γ=(V,E)
of a nonempty set V and a set E of 2-subsets of V.
The elements in V and E are called the vertices and edges of Γ, respectively.
For v∈V, the set Γ(v)={u∈V∣{u,v}∈E} is called
the neighborhood of v in Γ, while ∣Γ(v)∣ is called the valency
of v. We say that the graph Γ has valency d or Γ is d-regular
if its vertices have equal valency d.
For an integer s≥1, an s-arc in Γ is
an (s+1)-tuple (v0,v1,…,vs) of vertices with
{vi,vi+1}∈E and vi=vi+2 for all possible i.
A 1-arc is also called an arc.
Let Γ=(V,E) be a graph. A permutation g on V is called an automorphism
of Γ if {ug,vg}∈E for all {u,v}∈E. All automorphisms of Γ
form a subgroup of the symmetric group Sym(V), denoted by AutΓ, which
is called the automorphism group of Γ. The group AutΓ has a natural
action on E, namely, {u,v}g={ug,vg} for {u,v}∈E
and g∈AutΓ. If this action is transitive, that is, for each
pair of edges there
exists some g∈AutΓ mapping one edge to the other one, then
Γ is called edge-transitive.
Similarly, we may define the vertex-transitivity, arc-transitivity
and s-arc-transitivity** of Γ.
The graph Γ is called edge-primitive if
AutΓ acts primitively on E, that is, Γ is edge-transitive and
the stabilizer (AutΓ){u,v} of some (and hence every) edge {u,v}
in AutΓ is a maximal subgroup.**
Let Γ=(V,E) be an edge-primitive graph of valency no less than 3.
Then, as observed in ****[9]****, Γ is also arc-transitive.
If Γ is 2-arc-transitive then Praeger’s reduction theorems
****[24, 25]**** will be effective tools for us to
investigate the group-theoretic and graph-theoretic properties of Γ.
However, Γ is not necessarily 2-arc-transitive; for example,
by the Atlas ****[3]****,
the sporadic Rudvalis group Ru
is the automorphism group of a rank 3 graph, which is edge-primitive
and of valency 2304 but not 2-arc-transitive.
Using O’Nan-Scott Theorem for (quasi)primitive groups ****[24]****,
Giudici and Li ****[9]**** gave a reduction theorem on
the automorphism group of Γ. They proved that, as a primitive group on E,
only four of the eight O’Nan-Scott types for primitive groups may occur
for AutΓ, say SD, CD, PA and AS.
They also considered the possible O’Nan-Scott types for AutΓ acting on V,
and presented constructions or examples to verify the existence of
corresponding graphs. Then what will happen if we assume further that
Γ is 2-arc-transitive? The third author of this paper showed that either
AutΓ is almost simple or Γ is a complete bipartite graph if Γ is 2-arc-transitive, see ****[20]****.
This stimulate our interest in classifying those edge-primitive graphs
which are 2-arc-transitive.
In this paper, we present a classification result stated as follows.
Theorem 1.1**.**
Let Γ=(V,E) be a graph of valency d≥6, and let G≤AutΓ such that
G acts primitively on the edge set and transitively on the 2-arc set of Γ.
Assume further that G is almost simple and, for {u,v}∈E, the edge-stabilizer G{u,v} is soluble.
Then either Γ is (G,4)-arc-transitive, or
G, G{u,v}, Gv and d are listed as in Table 1.
Remark.
If Γ is edge-primitive and either 4-arc-transitive or of valency less
than 6, then the edge-stabilizers must be soluble. The reader may find a
complete list of such graphs in ****[10, 11, 17, 29**]****.
For each triple (G,Gv,G{u,v}) listed in
Table 1, the coset graph Cos(G,Gv,G{u,v}),
see Section 2 for the definition,
is both (G,2)-arc-transitive and G-edge-primitive.
**□
2. Preliminaries
Let G be a finite group and H,K≤G with ∣K:(H∩K)∣=2 and
∩g∈GHg=1. Let [G:H]={Hx∣x∈G}, and define a
graph Cos(G,H,K) on [G:H] such that
{Hx,Hy} is an edge if and only if yx−1∈HKH∖H.
The group G can be viewed as a subgroup of AutCos(G,H,K),
where G acts on [G:H] by right multiplication.
Then Cos(G,H,K) is G-arc-transitive and,
for x∈K∖H, the edge {H,Hx} has
stabilizer K in G. Thus Cos(G,H,K)
is G-edge-primitive if and only if K is maximal in G.
Assume that Γ=(V,E) is a G-edge-primitive graph of valency d≥3.
Then Γ is G-arc-transitive by ****[9, Lemma 3.4]****.
Take an edge {u,v}∈E, let H=Gv and K=G{u,v}.
Then K is maximal in G, and H∩K=Guv, which has index 2 in K.
Noting that ∩g∈GHg fixes V pointwise, ∩g∈GHg=1.
Further, vg↦Gvg,∀g∈G gives an isomorphism from Γ to
Cos(G,H,K). Then, by ****[5, Theorem 2.1]****, the following lemma holds.
Lemma 2.1**.**
Let Γ=(V,E) be a connected graph of valency d≥3, and G≤AutΓ.
Then Γ is both (G,2)-arc-transitive and G-edge-primitive if and only if
Γ≅Cos(G,H,K) for some subgroups H and K of G satisfying
(1)
∣K:(H∩K)∣=2, ∩g∈GHg=1 and K is maximal in G;
(2)
H* acts 2-transitively on [H:(H∩K)] by right multiplication.*
Let Γ=(V,E) be a connected graph of valency no less than 3,
{u,v}∈E and G≤AutΓ. Assume that
Γ is (G,s)-arc-transitive for some s≥1, that is,
G acts transitively on the s-arc set of Γ.
Then Gv acts transitively on the neighborhood Γ(v) of v in Γ.
Let GvΓ(v) be the transitive permutation group
induced by Gv on Γ(v), and
let Gv[1] be the kernel of Gv acting on Γ(v).
Then GvΓ(v)≅Gv/Gv[1].
Set Guv[1]=Gu[1]∩Gv[1].
Then Gv[1] induces a normal subgroup of
(GuΓ(u))v with the kernel Guv[1].
Since G is transitive on the arcs of Γ,
there is some element in G interchanging u and v.
This implies that
[TABLE]
Writing Gv[1] and Gv in group extensions, the next lemma follows.
By ****[32]****, s≤7 , and if s≥2 then Guv[1]
is a p-group for some prime p, refer to ****[7]****.
Thus Lemma 2.2 yields a fact as follows.
Corollary 2.3**.**
Let Γ=(V,E) be a connected (G,2)-arc-transitive graph,
and {u,v}∈E. Then G{u,v} is soluble if and only if
(GvΓ(v))u is soluble, and Gv is soluble if and only if
GvΓ(v) is soluble.
Choose s maximal as possible, that is, Γ is (G,s)-arc-transitive but not (G,s+1)-arc-transitive. In this case, Γ is said to be
(G,s)-transitive. If further Guv[1]=1,
then one can read out the vertex-stabilizer Gv from
****[8, 31]**** for s≥4 and
from ****[28]**** for 2≤s≤3.
In particular, we have the following result.
Theorem 2.4**.**
Let Γ=(V,E) be a connected (G,s)-transitive graph of valency on less 3,
and {u,v}∈E.
Assume that s≥2.
(1)
If Guv[1]=1 then s=2 or 3.
2. (2)
If Guv[1]=1 then
Guv[1] is a p-group for some prime p,
PSLn(q)⊴GvΓ(v), ∣Γ(v)∣=q−1qn−1 and 6=s≤7, where n≥2 and q=pf for some integer f≥1; moreover, either
(2.1)
n=2* and s≥4; or*
2. (2.2)
n≥3, s≤3 and Op(Gv) is given as in Table 2, where Op(Gv) is the maximal normal p-subgroup of Gv.
Lemma 2.5**.**
Let Γ=(V,E) be a connected (G,2)-arc-transitive graph, and {u,v}∈E.
If r is a prime divisor of ∣Γ(v)∣ then Or(Gv[1])=1=Or(Guv), and either
Or(Gv)=1, or
∣Γ(v)∣=re and
Or(Gv)≅Zre≅soc(GvΓ(v)) for some integer e≥1.
*Proof. *Since Γ is (G,2)-arc-transitive,
GvΓ(v) is a 2-transitive group, and thus Guv is transitive on Γ(v)∖{u}.
Since Or(Guv)⊴Guv, all Or(Guv)-orbits on Γ(v)∖{u} have the same size.
Noting that r is coprime to ∣Γ(v)∖{u}∣, it follows that
Or(Guv)≤Gv[1]. Since Gv[1]⊴Guv, we have Or(Gv[1])≤Or(Guv), and so Or(Gv[1])=Or(Guv).
Similarly, considering the action of Guv on Γ(u)∖{v}, we get
Or(Gu[1])=Or(Guv). Then Or(Gu[1])=Or(Guv)=Or(Gv[1])≤Guv[1].
By Theorem 2.4, either Guv[1]=1 or Guv[1] is a p-group, where p is a prime divisor of ∣Γ(v)∣−1. It follows that Or(Gu[1])=Or(Guv)=Or(Gv[1])=1.
**Note that Or(Gv)Gv[1]/Gv[1]≅Or(Gv)/(Or(Gv)∩Gv[1]).
Since Or(Gv)∩Gv[1]≤Or(Gv[1])=1, we have Or(Gv)≅Or(Gv)Gv[1]/Gv[1]⊴Gv/Gv[1]≅GvΓ(v).
Thus Or(Gv) is isomorphic to a normal r-subgroup of GvΓ(v).
This implies that either Or(Gv)=1, or GvΓ(v) is an affine 2-transitive group of degree re for some e. Thus the lemma follows.
**□
Let a≥2 and f≥1 be integers. A prime divisor r of af−1 is primitive if r is not a divisor of ae−1 for all 1≤e<f. By Zsigmondy’s theorem ****[35]****, if f>1 and af−1 has no primitive prime divisor then
af=26, or f=2 and a=2t−1 for some prime t.
Assume that af−1 has a primitive prime divisor r. Then a has order f modulo r. Thus f is a divisor of r−1, and if r is a divisor of af′−1 for some f′≥1 then f is a divisor of f′.
Thus we have the following lemma.
Lemma 2.6**.**
Let a≥2, f≥1 and f′≥1 be integers. If
af−1 has a primitive prime divisor r then f is a divisor of r−1, and
r is a divisor of af′−1 if and only if f is a divisor of f′.
If f≥3 then af−1 has a prime divisor no less than 5.
We end this section with
a fact on finite primitive groups.
Lemma 2.7**.**
Assume that G is a finite primitive group with a point-stabilizer H.
If H has a normal Sylow subgroup P=1, then P is also a Sylow subgroup of G.
*Proof. ***Assume that P=1 is a normal Sylow subgroup of H. Clearly, P is not normal in G. Take a Sylow subgroup Q of G with P≤Q. Then H≤⟨NQ(P),H⟩≤NG(P)=G. Since H is maximal in G, we have
H=⟨NQ(P),H⟩ and so NQ(P)≤H. It follows that NQ(P)=P, and hence P=Q. Then the lemma follows.
**□
3. Some restrictions on stabilizers
In Sections 4 and 5, we shall prove Theorem 1.1 using the result given in ****[17]**** which classifies finite primitive groups with soluble point-stabilizers.
Let Γ=(V,E) be a graph of valency d≥6,
{u,v}∈E and G≤AutΓ. Assume
that G is almost simple, G{u,v} is soluble, Γ is G-edge-primitive and (G,2)-arc-transitive.
Clearly, each nontrivial normal subgroup of G acts transitively on the edge set E.
Choose a minimal X among the normal subgroups of G which act primitively
on E. By the choice of X, we have soc(X)=soc(G), X{u,v}=X∩G{u,v}, G=XG{u,v} and
G/X=XG{u,v}/X≅G{u,v}/X{u,v}.
Then, considering the restrictions on both X{u,v} and Xv caused by the
2-arc-transitivity of Γ,
we may work out the pair (X,X{u,v}) from ****[17, Theorem 1.1]****, and then determine the group G and the graph Γ.
Thus we make the following assumptions.
Hypothesis 3.1**.**
Let Γ=(V,E) be a G-edge-primitive graph of valency d≥6, and {u,v}∈E, where G is an almost simple group with socle T. Assume that
(i)
Γ* is (G,2)-arc-transitive, and the edge-stabilizer G{u,v} is soluble;*
(ii)
G* has a normal subgroup X such that soc(X)=T, X{u,v} is maximal in X, and (X,X{u,v}) is one of the pairs (G0,H0) listed in
[17, Tables 14-20].*
For the group X in Hypothesis 3.1, we have 1=XvΓ(v)⊴GvΓ(v). Note that GvΓ(v) is 2-transitive (on Γ(v)).
Then GvΓ(v) is affine or almost simple, see ****[4, Theorem 4.1B]**** for example. It follows that soc(GvΓ(v))=soc(XvΓ(v)).
3.1.
Assume that Gv is insoluble. Then GvΓ(v) is an almost simple 2-transitive group (on Γ(v)). Recall that soc(GvΓ(v))=soc(XvΓ(v)). Checking the point-stabilizers of almost simple 2-transitive groups, since (GvΓ(v))u is soluble, we conclude that either XvΓ(v) is 2-transitive, or GvΓ(v)≅PSL2(8).3 and d=28.
For a complete list of finite 2-transitive groups, the reader may refer to
****[2, Tables 7.3 and 7.4]****.
Lemma 3.2**.**
Suppose that Hypothesis 3.1 holds. If d=28 then GvΓ(v)≅PSL2(8).3.
*Proof. *Suppose that GvΓ(v)≅PSL2(8).3 and d=28.
Note that Xuv[1]≤Guv[1]=1, see Theorem 2.4.
Thus Xuv≲(XvΓ(v))u×(XuΓ(u))v by Lemma 2.2.
Assume that XvΓ(v)≅PSL2(8).
Then (XvΓ(v))u≅D18, and Xuv≅D18, (Z3×Z9):Z2, (Z9×Z9):Z2 or D18×D18. In particular, the unique
Sylow 3-subgroup of X{u,v}=Xuv.2 is isomorphic to Zm×Z9, where m=1, 3 or 9. Checking the primitive groups listed in ****[17, Tables 14-20]****, we know that only the pairs (PSL2(q),D2(2,q−1)q±1) possibly meet our requirements on X{u,v}, yielding X{u,v}≅D2(2,q−1)q±1. Then D36≅X{u,v}≅D2(2,q−1)q±1. Calculation shows that q=37; however, PSL2(37) has no subgroup which has a quotient PSL2(8), a contradiction.
Now let XvΓ(v)=GvΓ(v)≅PSL2(8).3. Then (XvΓ(v))u≅(XuΓ(u))v≅Z9:Z6 and Xuv≲Z9:Z6×Z9:Z6. In particular, a Sylow 2-subgroup of X{u,v}=Xuv.2
is not a cyclic group of order 8, and the unique Sylow 3-subgroup of
X{u,v} is nonabelian and contains elements of order 9. Since X{u,v}=Xuv.2=Xv[1].(XvΓ(v))u.2 and Xv[1]≅(Xv[1])Γ(u)⊴(XuΓ(u))v,
we have ∣X{u,v}∣=2233, 2234, 2235, 2236, 2335 or 2336.
Checking the Tables 14-20 given in ****[17]****, we conclude that X=G2(3).2, and X{u,v}≅[36]:D8. In this case, Xv[1]≅Z9:Z6 and
Xv≅Z9:Z6.PSL2(8).3; however, X has no such subgroup by the Atlas
****[3**]****, a contradiction.
This completes the proof.
**□
By Lemma 3.2, combining with Theorem 2.4, the next lemma
follows from checking the point-stabilizers of finite almost simple 2-transitive groups.
Lemma 3.3**.**
Suppose that Hypothesis 3.1 holds and GvΓ(v) is almost simple. Then one of the following holds:
(s1)
GvΓ(v)=XvΓ(v)=PSL3(2)* or PSL3(3), and d=7 or 13, respectively;*
2. (s2)
soc(XvΓ(v))=PSL2(q)* with q>4, and d=q+1;*
3. (s3)
Guv[1]=1, soc(XvΓ(v))=PSU3(q) with q>2, and d=q3+1;
4. (s4)
Guv[1]=1, soc(XvΓ(v))=Sz(q) with q=22n+1>2, and d=q2+1;
5. (s5)
Guv[1]=1, soc(XvΓ(v))=Ree(q) with q=32n+1>3, and d=q3+1.
In particular, Γ is (X,2)-arc-transitive.
Recall that the Fitting subgroup Fit(H) of a finite group H is the direct product of Or(H), where r runs over the set of prime divisors of ∣H∣.
Lemma 3.4**.**
Suppose that Hypothesis 3.1 holds and one of Lemma 3.3 (s2)-(s5)* occurs. Let q=pf for some prime p.
Assume that Xu,v[1]=1. Then Fit(Xuv)=Op(Xuv), and either
Fit(Xuv)=Fit(X{u,v}) or Fit(X{u,v})=Fit(Xuv).2;
in particular,
∣Fit(X{u,v}):Op(X{u,v})∣≤2.*
*Proof. *Let r be a prime divisor of ∣Xuv∣. Then Or(Xuv) is normal in Xuv. Since Γ is (X,2)-arc-transitive,
Xuv acts transitively on Γ(v)∖{u}. Thus all Or(Xuv)-orbits (on Γ(v)∖{u}) have equal size, which is a power of r and a divisor of ∣Γ(v)∖{u}∣. Note that ∣Γ(v)∖{u}∣=d−1, which
is a power of p. It follows that either r=p or Or(Xuv)=1. Then Fit(Xuv)=Op(Xuv).
Note that Xuv is normal in X{u,v} as ∣X{u,v}:Xuv∣=2.
Since Op(Xuv) is a characteristic subgroup of Xuv, it follows that Op(Xuv) is normal in X{u,v}, and so Op(Xuv)≤Op(X{u,v})≤Fit(X{u,v}).
For each odd prime divisor r of ∣X{u,v}∣, since ∣X{u,v}:Xuv∣=2,
we have Or(X{u,v})≤Xuv, and so Or(X{u,v})=Or(Xuv).
It follows that Fit(X{u,v})=Fit(Xuv)O2(X{u,v})=Op(Xuv)O2(X{u,v}).
In particular, Op(Xuv)=Op(X{u,v}) if p=2.
**It is easily shown that Xuv∩O2(X{u,v})=O2(Xuv).
If Xuv≥O2(X{u,v}) then p=2, Fit(X{u,v})=O2(X{u,v})=Fit(Xuv), and the lemma holds.
Assume that O2(X{u,v})≤Xuv.
Since ∣X{u,v}:Xuv∣=2, we have X{u,v}=XuvO2(X{u,v}).
Then 2∣Xuv∣=∣X{u,v}∣=∣Xuv∣∣O2(X{u,v}):(Xuv∩O2(X{u,v}))∣=∣Xuv∣∣O2(X{u,v}):O2(Xuv)∣, yielding
∣O2(X{u,v}):O2(Xuv)∣=2. If p=2 then Fit(X{u,v})=O2(X{u,v}) and Fit(Xuv)=O2(Xuv), the lemma follows. If p=2 then O2(Xuv)=1,
∣O2(X{u,v})∣=2, and so Fit(X{u,v})=Op(Xuv)×Z2.
This completes the proof.
**□
3.2.
Assume that Hypothesis 3.1 holds and Gv is soluble.
Then GvΓ(v) is an affine 2-transitive group. Let soc(GvΓ(v))=Zpf. Then d=pf.
Recalling that d≥6, we have Guv[1]=1 by Theorem 2.4, and so Guv≲(GvΓ(v))u×(GuΓ(u))v.
If Guv is abelian then Γ is known by ****[20]****.
Thus we assume further that Guv is not abelian.
Then (GvΓ(v))u is nonabelian,
and so (GvΓ(v))u≤GL1(pf); in particular, f>1. Since (GvΓ(v))u is soluble, by ****[2, Table 7.3]****,
we have the following lemma.
Lemma 3.5**.**
Suppose that Hypothesis 3.1 holds, Gv is soluble and Guv is not abelian. Let soc(GvΓ(v))=Zpf, where p is a prime. Then f>1, and one of the following holds.
(a1)
f=2, and either SL2(3)⊴(GvΓ(v))u≤GL2(p) and p∈{3,5,7,11,23}, or p=3 and (GvΓ(v))u=Q8;
2. (a2)
2+1+4:Z5≤(GvΓ(v))u≤2+1+4.(Z5:Z4)<2+1+4.S5,
and pf=34;
3. (a3)
(GvΓ(v))u≤GL1(pf), (GvΓ(v))u≤ΓL1(pf) and ∣(GvΓ(v))u∣ is divisible by pf−1.
Consider the case (a3) in Lemma 3.5.
Write
[TABLE]
Let ⟨τ⟩∩(GvΓ(v))u=⟨τm⟩, where m\,\big{|}\,(p^{f}-1).
Then
[TABLE]
Set (GvΓ(v))u/⟨τm⟩≅⟨σe⟩ for some divisor e of f. Then
[TABLE]
Choose τlσk∈(GvΓ(v))u such that (GvΓ(v))u=⟨τm⟩⟨τlσk⟩. Then (τlσk)ef∈⟨τm⟩ but (τlσk)j∈⟨τm⟩ for 1≤j<ef. It follows that σk has order ef.
Then σk=σie for some i with (i,ef)=1, and then (σk)i′=σe for some i′. Thus, replacing τlσk by a power of it if necessary, we may let k=e. Then
[TABLE]
Further, (GvΓ(v))u=⟨τm⟩⟨(τm)iτlσe⟩ for an arbitrary integer i, thus we may assume further 0≤l<m. By ****[6, Proposition 15.3]****, letting π(n) be the set of prime divisors of a positive integer n, we have
(⋇)
π(m)⊆π(pe−1), me\,\big{|}\,f and (m,l)=1; in particular, m=1 if l=0.
Suppose that Xuv is nonabelian. (The case where Xuv is abelian is left in Section 5.)
Since Xuv[1]≤Guv[1]=1, we have
[TABLE]
This yields that (XvΓ(v))u is nonabelian. Then a limitation on π(∣Xuv∣) is given as follows.
Lemma 3.6**.**
Assume that Lemma 3.5 (a3)* holds and Xuv is nonabelian. Then (XvΓ(v))u≅Zm′.Ze′f, where m′ and e′ satisfy*
(i)
Zm′≅(XvΓ(v))u∩⟨τm⟩, mm^{\prime}\,\big{|}\,p^{f}-1, e\,\big{|}\,e^{\prime}\,\big{|}\,f; and
2. (ii)
m′>1, e′<f, π(pf−1)∖π(pe′−1)⊆π(m′)⊆π(∣Xuv∣).
*Proof. *Recall that (XvΓ(v))u⊴(GvΓ(v))u=⟨τm⟩⟨τlσe⟩≅Zmpf−1.Zef. Then
[TABLE]
yielding (XvΓ(v))u≅Zm′.Ze′f with m′ and e′ satisfying (i). Since Xuv is nonabelian, (XvΓ(v))u is nonabelian, and so m′>1 and e′<f.
**By the above (⋇), each r∈π(pf−1)∖π(pe′−1) is a divisor of ∣⟨τm⟩∣=mpf−1. Let R be the unique subgroup of order r of ⟨τm⟩. Then, noting that R is normal in (GvΓ(v))u, either R≤(XvΓ(v))u or R(XvΓ(v))u=R×(XvΓ(v))u. Suppose that the latter case occurs. Since e′<f, we may let τnσe′∈(XvΓ(v))u∖⟨τm⟩.
Then σe′ centralizes R.
Thus xpe′=x for x∈R, yielding r\,\big{|}\,(p^{e^{\prime}}-1), a contradiction.
Then R≤(XvΓ(v))u∩⟨τm⟩≅Zm′.
Noting that m′ is a divisor of ∣Xuv∣, the result follows.
**□
4. Graphs with insoluble vertex-stabilizers
In this and next sections, we prove Theorem 1.1. Thus, we let G, T, X and Γ=(V,E) be as in Hypothesis 3.1.
Our task is to determine which pair (G0,H0) listed in ****[17, Tables 14-20]**** is a possible candidate for (X,X{u,v}), and determine whether or not the
resulting triple (G,Gv,G{u,v}) meets the conditions (1) and (2) in Lemma 2.1.
In this section, we deal with the case where Gv is insoluble, that is,
Xv is described as in Lemma 3.3. First, by the following lemma,
Lemma 3.3 (s4) and (s5) are excluded.
*Proof. *Suppose that Lemma 3.3 (s4) or (s5) holds.
Then Xuv[1]=1 by Theorem 2.4. Thus
Xv=Xv[1].XvΓ(v), Xv[1]≅(Xv[1])Γ(u)⊴(XuΓ(u))v≅(XvΓ(v))u,
and
Xuv≲(XvΓ(v))u×(XuΓ(u))v. Set q=pf with p a prime. Then the pair (XvΓ(v),(XvΓ(v))u) is given as follows:
[TABLE]
In particular, Op(X{u,v}) is not abelian.
We next show that none of the pairs (G0,H0) in ****[17, Tables 14-20]**** gives a desired pair (X,X{u,v}).
Since Op(X{u,v}) is nonabelian, those pairs (G0,H0) with Op(H0) abelian are not in our consideration. In particular, soc(X) is not isomorphic to an alternating group. Also, noting that X{u,v} has a subgroup of index 2, those H0 having no subgroup of index 2 are excluded.
Case 1.
Suppose that soc(XvΓ(v))=Ree(q). Then p=3,
O3(X{u,v}) is nonabelian and of order 33f, 34f, 35f or 36f, ∣X{u,v}∣ is a divisor of 2(q−1)2f2 and divisible by 2(q−1). Checking the orders of those H0 given in ****[17, Tables 15]****, we conclude that soc(X) is not a sporadic simple group.
Suppose that soc(X) is a simple exceptional group of Lie type.
By ****[17, Table 20]****, we conclude that (X,X{u,v}) is one of (G2(3t).Z2l+1,[36t]:Z3t−12.Z2l+1) and (Ree(3t),[33t]:Z3t−1), where 2l is the 2-part of t. If f>t then, by Zsigmondy’s theorem, q−1 has a prime divisor which is
not a divisor of 3t−1, which contradicts that ∣X{u,v}∣ is divisible by 2(q−1). Thus f=t, and then X=G2(q).Z2l+1 and X{u,v}≅[q6]:Zq−12.Z2l+1.
This implies that Xv[1]=1, in fact, ∣O3(Xv[1])∣=q3.
Thus O3(Xv)=1 and Xv has a quotient Ree(q).e.
Checking the maximal subgroups of G2(q).Z2l+1, refer to ****[15, Theorems A and B]****, we conclude that G2(q).Z2l+1 has no maximal subgroup containing such Xv as a subgroup, a contradiction.
Suppose that soc(X) is a simple classical group
over a finite field of order rt, where r is a prime.
Since f>1 is odd, 3f−1 has an odd prime divisor, and so X{u,v} is not a
{2,3}-group as ∣X{u,v}∣ is divisible by 3f−1.
Recall that O3(X{u,v}) is nonabelian and of order 33f, 34f, 35f or 36f. Checking the groups H0 given in ****[17, Table 16-19]****,
we conclude that soc(X)=PSLn(rt) or PSUn(rt), where n∈{3,4}.
Take a maximal subgroup M of X such that Xv≤M. Then M has a simple section (i.e., a quotient of some subgroup) Ree(q). Recall that q>3. Checking Tables 8.3-8.6 and 8.8-8.11 given in ****[1]****, we conclude that none of PSL3(rt), PSL4(rt), PSU3(rt) and PSU4(rt) has such maximal subgroups, a contradiction.
Case 2.
Suppose that soc(XvΓ(v))=Sz(q). Then q=2f,
∣O2(X{u,v})∣=22fa, 23fa or 24fa, where f>1 is odd, and a=1 or 2.
Noting that X{u,v} has order divisible by 2f−1, by Lemma 2.6, we conclude that X{u,v} is not a {2,3}-group. Since X{u,v}
is nonabelian, it follows from ****[17, Table 15-20]**** that either
(X,X{u,v}) is one of (2F4(2)′,[29]:5:4), (PSp4(2t).Z2l+1,[24t]:Z2t−12.Z2l+1) and (Sz(2t),[22t]:Z2t−1), or
soc(X) is one of PSLn(rt) and PSUn(rt), where n∈{3,4}, 2l is the 2-part of t, and r is odd if n=4.
The first pair leads to q=23, and so ∣X{u,v}∣ is divisible by 7, a contradiction. Checking the maximal subgroups of soc(X) (refer to ****[1, Tables 8.3-8.6, 8.8-8.14]****), the groups PSL3(rt), PSU3(rt), PSL4(rt) and PSU4(rt) are excluded as they have no maximal subgroup with a simple section Sz(q). Thus (X,X{u,v})=(PSp4(2t).Z2l+1,[24t]:Z2t−12.Z2l+1) or (Sz(2t),[22t]:Z2t−1).
Since f>1 is odd, 2f−1 has a primitive prime divisor, say s.
Recalling that ∣X{u,v}∣ is divisible by 2f−1, it follows that
s is a divisor of 2t−1, and so t≥f. Then t=f. It follows that X=PSp4(q).Z2l+1, and
Xv[1]≅[q2]:Zq−1. However, by ****[1**, Table 8.14]****, PSp4(q).Z2l+1 has no maximal subgroup containing [q2]:Zq−1.Sz(q), a contradiction.
**□
Lemma 4.2**.**
Assume that (s1) of Lemma 3.3 occurs. Then G, X, X{u,v} and Xv are listed as in Table 3.
*Proof. *Assume first that Xuv[1]=1. Then,
Xv=Xv[1].XvΓ(v), Xv[1]≅(Xv[1])Γ(u)⊴(XuΓ(u))v≅(XvΓ(v))u,
and
Xuv≲(XvΓ(v))u×(XuΓ(u))v.
Suppose that XvΓ(v)=PSL3(2). Then (XvΓ(v))u≅S4, and thus Xv[1] and X{u,v} are given as follows:
[TABLE]
In particular, 22≤∣O2(X{u,v})∣≤25. Check all possible pairs (X,X{u,v}) in ****[17, Tables 14-20]****. Noting that
A8≅PSL4(2) and PSU4(2)≅PSp4(3), we conclude that
X≅A8, X{u,v}≅24:S32 and Xv[1]≅A4; or
X=M12 with X{u,v}≅2+1+4:S3; or
X≅PSU4(2) with X{u,v}≅2A42.2. The group A8 is excluded as it has no subgroup of the form of
Xv[1].PSL3(2). The groups M12 and PSU4(2)
are excluded as their orders are not divisible by d=7.
Suppose that XvΓ(v)=PSL3(3). Then (XvΓ(v))u≅32:2S4.
Thus Xv[1] and X{u,v} are given as follows:
[TABLE]
Note that O3(X{u,v})≅32 or 34. Checking the possible pairs (X,X{u,v}), we have X{u,v}≅34:23.S4 and
X=A12 or PΩ8+(2); in this case, d=13 is not a divisor of
∣X∣, a contradiction.
Now let Xuv[1] be a nontrivial p-group. Then, by Theorem 2.4,
Xv and X{u,v} are given as follows:
[TABLE]
Suppose that p=2. Then ∣X{u,v}∣ is divisible by 9 if and only if ∣O2(X{u,v})∣≥8, and O2(X{u,v}) contains no elements of order 8 unless ∣O2(X{u,v})∣≥222. Check the pairs (G0,H0) given in ****[17, Tables 14-20]**** by estimating ∣H0∣ and ∣O2(H0)∣. We conclude that
one of the following holds:
(i)
X=PSL4(2).2≅S8** and X{u,v}=24:S4;**
2. (ii)
X=PSL5(2).2** and X{u,v}=[28].S32.2;**
3. (iii)
X=F4(2).2** and X{u,v}=[222].S32.2;**
4. (iv)
soc(X)=PSL3(4)** and ∣O2(X{u,v})∣=26;**
5. (v)
X=PSU4(3).23** and ∣O2(X{u,v})∣=27;**
6. (vi)
X=He.2** and X{u,v}=[28]:S32.2.**
Case (iv) yields that Xv≅23:SL3(2) or 24:SL3(2); however, X
has no such subgroup by the Atlas ****[3]****. Similarly, cases (v) and (vi) are excluded.
For (i),
G=X and Γ is (isomorphic to) the
point-plane incidence graph of the projective geometry PG(3,2).
For (ii),
G=X and Γ is (isomorphic to) the
line-plane incidence graph of the projective geometry PG(4,2).
If (iii) holds then G=X and Γ is the line-plane incidence graph of the metasymplectic space associated with F4(2), see ****[30]****.
Now let p=3. Then ∣O3(X{u,v})∣=35 or 38, and X{u,v} has no normal Sylow subgroups. Checking all possible pairs (X,X{u,v}) in
****[17, Tables 14-20]****,
we know that (X,X{u,v}) is one of the following pairs:
[TABLE]
**Note that O3(Xv)≤O3(X{u,v}).
Then, for the first pair,
O3(X{u,v})≅Z94 has no subgroup isomorphic to Z36, which is impossible. For the second pair, G=X and Γ is (isomorphic to) the
line-plane incidence graph of the projective geometry PG(4,3).
The last pair implies that X≅PGL4(3), G=X or X.2, and Γ is (isomorphic to) the
line-plane incidence graph of the projective geometry PG(3,3).
**□
Lemma 4.3**.**
Assume that Lemma 3.3 (s2)* holds.
Then d=q+1, and either Γ is (X,4)-arc-transitive, or G, X, X{u,v} and Xv are listed in Table 4.*
*Proof. *Let XvΓ(v)=PSL2(q).[o], and q=pf>4, where p is a prime and o\,\big{|}\,(2,q-1)f.
Note that Γ is (X,2)-arc-transitive, see Lemma 3.3.
By Theorem 2.4, if Xuv[1]=1 then Γ is (X,4)-arc-transitive. Thus we assume next that Xuv[1]=1, and then Lemma 3.4 works.
Note that
Xv=Xv[1].XvΓ(v), Xv[1]≅(Xv[1])Γ(u)⊴(XuΓ(u))v≅(XvΓ(v))u=pf:(2,q−1)q−1.[o],
and
Xuv≲(XvΓ(v))u×(XuΓ(u))v. We have
Op(X{u,v})=Zpif.a, where i∈{1,2} and a is a divisor of (2,p).
It is easily shown that i=2 if and only if Op(Xv[1])=Zpf.
Combining with Lemma 3.4, we need only consider those pairs (G0,H0) in
****[17, Tables 14-20]**** which satisfy
(a)
Op(H0)=Zpif.a, where i∈{1,2} and a is a divisor of (2,p);
∣Fit(H0):Op(H0)∣≤2; G0 has a subgroup, say M0, such that ∣M0:(M0∩H0)∣=q+1,
∣H0:(M0∩H0)∣=2, and M0 has a simple section PSL2(q);
2. (b)
∣H0:Op(H0)∣** is a divisor of 2(q−1)2f2 and divisible by q−1; if i=1 then ∣H0:Op(H0)∣ is a divisor of 2(q−1)f.**
Case 1. Assume that soc(X) is an alternating group.
Using ****[17, Table 14]****,
we have G=X=Sp and X{u,v}≅Zp:Zp−1, where p∈{7,11,17,23}. Then Xv=PSL2(p) and d=p+1. For p=17 or 23, the group PSL2(p) has no transitive permutation representation of degree p, and thus it cannot occur as a subgroup of Sp. Therefore, p=7 or 11, and G, X and X{u,v} are listed in Table 4.
In fact, Xuv and X{u,v} are the normalizers of some Syolw p-subgroup
in PSL2(p) and Sp, respectively.
(Note that, for p=7, the resulting graph is the point-plane non-incidence graph of PG(3,2).)
Case 2. Assume that soc(X) is a simple sporadic group. By ****[17, Table 15]****, with the restrictions (a) and (b), the only pairs (G0,H0) are listed as follows:
(M11,32:Q8.2),
(J1,Z11:Z10), (J1,Z7:Z6), (J3.2,Z19:Z18), (J4,Z29:Z28),
(O′N.2,Z31:Z30), (B,Z19:Z18×Z2), (B,Z23:Z11×Z2), (M,Z41:Z40) and (M,Z47:Z23×Z2). In particular, Op(H0) is a Sylow p-subgroup of G0. This yields that Xv[1]=1, and so soc(Xv)=PSL2(pf).
Suppose that (X,X{u,v}) is one of (J1,Z7:Z6), (J4,Z29:Z28) and (M,Z47:Z23×Z2).
Then Xv=PSL2(p) for p=7, 29 and 47, respectively; however,
by the Altas ****[3]**** and ****[34, Tables 5.6 and 5.11]****, X has no subgroup PSL2(p), a contradiction.
Thus G, X and X{u,v} are listed in Table 4.
(Note that the Monster M has a maximal subgroup PSL2(41) by ****[22]****.)
Case 3. Assume that soc(X) is a simple group of Lie type over a finite field of order rt, where r is a prime. We first show r=p.
Suppose that r=p. Then, by (a), either Op(H0) is abelian or r=p=2. For r=p>2, noting that ∣H0∣ has a divisor q−1, there does not exist H0 in ****[17, Tables 16-20]**** such that Op(H0) is abelian.
Thus we have r=p=2. Recalling that pf>4 and ∣H0/Op(H0)∣ is divisible by 2f−1, it follows from Lemma 2.6 that H0/Op(H0) is not a {2,3}-group. Checking those H0 given in ****[17, Tables 16-20]****, we conclude that (G0,H0) is one of the following pairs:
First, the pair (Sz(2t),[22t]:Z2t−1) is excluded as Sz(2t) has no subgroup with a section PSL2(2f).
For the last two pairs, we have f=5 and 4 respectively, which yields that
2f−1 is not a divisor of ∣H0∣, a contradiction.
For the three pairs after the first one, we have t<f, thus G0 has no maximal subgroup with a section PSL2(2f), a contradiction.
Suppose finally that (X,X{u,v})=(PSL2(2t),Z2t:Z2t−1). Then 3≤f<t≤2f+1.
Noting that 2f−1 is a divisor of 2t−1, it follows that f is a divisor of t, and so t=2f. Then O2(X{u,v})=22f, yielding ∣O2(Xv[1])∣=2f.
Thus O2(Xv)=1 and Xv has a simple section
PSL2(2f). Checking the subgroups of PSL2(22f), refer to ****[12, II.8.27]****, we conclude that PSL2(22f) has no subgroup isomorphic to Xv, a contradiction.
Assume that r=p in the following.
Subcase 3.1**. We first deal with those pairs (G0,H0) such that H0 is included in some infinite families in [17, Table 16-20].
Note that r=p, and we consider only those H0 having subgroups of index 2.
It follows that either H0/Fit(H0) is a {2,3}-group or
(G0,H0)=(E8(q′),Zq′8±q′7∓q′5−q′4∓q′3±q′+1.Z30), where q′=rt.
Suppose that (G0,H0)=(E8(q′),Zq′8±q′7∓q′5−q′4∓q′3±q′+1.Z30). Then q′8±q′7∓q′5−q′4∓q′3±q′+1 is divisible by some primitive prime divisor s of q′15−1 or of q′30−1. Noting that s≥17, we know that H0 has normal cyclic Sylow s-subgroup.
It follows from (a) that 17≤p=s=q′8±q′7∓q′5−q′4∓q′3±q′+1.
In particular, Op(H0)=Zp and f=1.
By (b), ∣H0∣ is divisible by p−1, and then 30 is divisible by p−1.
This implies that 30=p−1=q′8±q′7∓q′5−q′4∓q′3±q′, which is impossible. Therefore, H0/Fit(H0) is a {2,3}-group.**
By (a), Fit(H0) a {2,p}-group. Then ∣H0∣ has no prime divisor other than 2, 3 and p. Since pf−1 is a divisor of ∣H0∣, by Lemma 2.6,
we have f<3.
Recall that (XuΓ(u))v≅(XvΓ(v))u=pf:(2,q−1)q−1.[o],
and
Xuv≲(XvΓ(v))u×(XuΓ(u))v, where o is a divisor of (2,q−1)f. It follows that Xuv/Op(Xuv) has an abelian Hall 2′-subgroup.
Note that XuvOp(X{u,v})/Op(X{u,v})≅Xuv/(Op(X{u,v}∩Xuv)=Xuv/Op(Xuv), and ∣X{u,v}:XuvOp(X{u,v})∣≤2.
It follows that X{u,v}/Op(X{u,v}) has an abelian Hall 2′-subgroup.
Thus, as a possible candidate for X{u,v}, the quotient of H0 over Op(H0) has abelian Hall 2′-subgroups. In particular, H0/Op(H0) has no section A4.
Considering the restrictions on H0, r and f, we conclude that
(G0,H0) can only be one of the following pairs:
(PSp4(2t).Z2l+1,Z2t±12.[2l+4]), (PSp4(2t).Z2l+1,Z22t+1.[2l+3]), where t≥3;
(Sz(2t),Z2t−1:Z2)**,
(Sz(2t),Z2t±2t+1+1:Z4), where t≥3;; **
(Ree(3t),Z3t±3t+1+1.Z6)**,
(Ree(3t),Z3t+1.Z6), where t≥3;; **
(G2(3t).Z2l+1,Z3t±12.[3⋅2l+3]),
(G2(3t).Z2l+1,Z32t±3t+1.[3⋅2l+2]), where t≥2;
(3D4(rt),Zr4t−r2t+1:Z4)**;
(2F4(2t),Z22t±23t+1+2t±2t+1+1.Z12), where t≥3;
**
(F4(2t).Z2l+1,Z24t−22t+1.[3⋅2l+3]), where t≥2;
where the power 2l appeared means the 2-part of t.
Recall that ∣Fit(H0):Op(H0)∣≤2 and ∣H0:Op(H0)∣ is divisible by pf−1. This allows us determine the values of pf and rt. As an example, we only deal with the second pair. Suppose that (G0,H0)=(PSL3(rt),[(3,rt−1)(rt−1)2].S3). Considering the structures of Fit(H0) and Op(H0), either (3,rt−1)=1, p=rt−1 and f∈{1,2}, or f=1 and p=rt−1=3. The latter implies that PSL2(q) is soluble, which is not the case. Assume that the former case holds.
Then ∣S3∣ is divisible by rt−1−1 or (rt−1)2−1. Then the only possibility is that (pf,rt)=(7,8).
The other pairs can be fixed out in a similar way, the details is omitted here.
Eventually, we conclude that (G0,H0,p,f) is one of
(PSL2(19),D20,5,1), (PSL3(8),72:S3,7,1) and
(Sz(8),Z5:Z4,5,1).
By the Atlas ****[3]****, neither PSL3(8) nor Sz(8) has subgroup
with a section PSL2(p).
Thus, in this case, G, X and X{u,v} are given as in Table 4.
Subcase 3.2**. For the pairs (G0,H0) not appearing in Subcase 3.1, we check the finite number of H0 one by one. We observed that
either p=2, or H0/Op(H0) is a {2,3}-group.
Recall that r=p.**
Suppose that p=2. Recalling that q=2f>4, we have f≥3.
In particular, since ∣H0∣ is divisible by 2f−1,
H0 is not a {2,3}-group
by Lemma 2.6. Then the only possibility is that
G0=2F4(2)′ and H0=[29]:5:4. Thus ∣O2(H0)∣=29, it follows from (a) that f=4 or 9, and then G0 has a section PSL2(24) or PSL2(29), which is impossible by checking the (maximal) subgroups of 2F4(2)′. Thus p>2, and H0/Op(H0) is a {2,3}-group; in particular, by (a),
Op(H0)=Zpif for some i∈{1,2}.
Suppose that H0 has a section A4.
Then H0 has no normal Sylow 3-subgroup. Further, H0 has no quotient A4 as H0 has a subgroup of index 2.
If (3,(q−1)f)=1 then, by (b), we conclude that p=3 and Op(H0) is the unique Sylow 3-subgroup of H0, a contradiction. Thus 3 is a divisor of (q−1)f.
Check those H0 in ****[17, Table 16-20]**** which have a section A4 and
do not appear in Subcase 3.1.
Recalling r=p>2 and Op(H0)=Zpif, it follows that either Op(H0)=Z32 or (G0,H0)=(F4(2).4,Z72:(3×SL2(3)).4). Since 3 is a divisor of (q−1)f, we get G0=F4(2).4
and q=pf=7 or 72.
By (b), for q=7 or 72, the order of H0 should be a divisor of
72 or 192 respectively, which is impossible.
The above argument allows us ignore many cases without further inspection.
Inspecting carefully the remaining pairs, the possible candidates for (X,X{u,v}) are as follows:
(PSU3(2t),32:Q8), where t is a prime no less than 5;
(2F4(2),Z13:Z12).
For the first three pairs, G, X and X{u,v} are easily determined and given as in Table 4.
The pair (PSp4(4).4,Z17:Z16) is excluded as PSp4(4).4 has no subgroup PSL2(17), and the pair (2F4(2),Z13:Z12) is excluded as 2F4(2) has no subgroup PSL2(13). Suppose that X≅PSU3(2t) and X{u,v}≅32:Q8. Then we have Xv≅PSL2(9); however,
by ****[1, Tables 8.3, 8.4]****, PSU3(2t) has no subgroup PSL2(9), a contradiction.
Suppose that (X,X{u,v})=(PSp4(4).4,52:[25]).
Then Xv contains a Sylow 5-subgroup P of X and has a section PSL2(5) or
PSL2(25). By the information for PSp4(4).4 given in the Atlas ****[3]****,
we conclude that Xv≤M≅(A5×A5):22<PSp4(4).2<PSp4(4).4.
Note that Xuv=52:[24], which should be the normalizer of P in Xv.
Using GAP ****[27]****, computation shows that ∣NL(P)∣≤200 for any maximal subgroup L of M with P≤L. It follows that Xv=M≅(A5×A5):22,
yielding d=∣Xv:Xuv∣=36=q+1, a contradiction.
Let (X,X{u,v})=(PSL3(r),32:Q8). Then Xuv≅32:4.
It is easily shown that p=3 and Xv≅PSL2(9).
Since r≡4,7(mod9), we know that PSL3(r) has a Sylow 3-subgroup Z32.
By ****[1, Tables 8.3, 8.4]****, PSL3(r) has a subgroup PSL2(9)
if and only if r≡1,4(mod15). Thus, in this case, we have
r≡11,14,29,41(mod45). For a subgroup PSL2(9) of PSL3(r), taking a Sylow 3-subgroup Q of PSL2(9), the normalizers of Q in PSL2(9) and PSL3(r) are (isomorphic to) 32:4 and 32:Q8, respectively.
Then these two normalizers of Q can serve as the roles of Xuv and X{u,v}, respectively.
Thus X and X{u,v} are given as in Table 4.
Noting that G=XG{u,v}, we have G{u,v}/X{u,v}≅G/X≲Out(PSL3(r))≅S3, and so
G=X.[m] and G{u,v}=X{u,v}.[m], where m is a divisor of 6.
Thus ∣Guv:Xuv∣=m, since ∣Gv:Guv∣=10=∣Xv:Xuv∣, we have
∣Gv:Xv∣=m. By ****[1**, Table 8.4]****, NAutPSL3(r)(Xv)=Xv.2.
Since Xv⊴Gv, it follows that m≤2. Thus G=X or X.2, and if G=X.2 then Gv=Xv.2≅PGL2(9) and G{u,v}≅32:Q8.2.
The pair (PSU3(r),32:Q8) is similarly dealt with.
**□
Lemma 4.4**.**
If Lemma 3.3 (s3)* holds then G, X, X{u,v} and Xv are listed in Table 5.*
Proof. Let XvΓ(v)=PSU3(q).[o] and q=pf>2, where p is a prime and o\,\big{|}\,2(3,q{+}1)f. Then
(XvΓ(v))u=pf+2f:(3,q+1)q2−1.[o], and Xuv[1]=1 by Theorem 2.4. Thus
∣Op(X{u,v})∣=p3f.a, p4f.a, p5f.a or p6f.a, where a is a divisor of (2,p).
Moreover, Op(X{u,v}) is nonabelian, and
X{u,v}/Op(X{u,v}) has a
subgroup Z(3,q+1)(q2−1). We next determine which pair (G0,H0) in ****[17**, Tables 14-20]** is a possible candidate for (X,X{u,v}).
Note that
we may ignore those H0 which either has no subgroup of index 2 or has abelian maximal normal p-subgroup. In particular, soc(X) is not an alternating group.**
Case 1. Let (G0,H0) be a pair with H0 included in some infinite families given in ****[17, Table 16-20]****. Since
Op(X{u,v}) is nonabelian, we conclude that (X,Op(X{u,v})) is one of the following pairs:
(Sz(pt),[p2t]), (Ree(pt),[p3t]) and (G2(pt).Z2l+1,[p6t]),
where 2l is the 2-part of t.
Check the maximal subgroups of PSp4(pt).Z2l+1, Sz(pt) and Ree(pt), refer to ****[1, Table 8.14]****, ****[26, Theorem 9]**** and ****[15, Theorem C]****, respectively.
We conclude that none of PSp4(tf).Z2l+1, Sz(pt) and Ree(pt) has maximal subgroups with a simple section PSU3(q), and they are excluded.
For the first three and the last pairs, ∣X/Op(X{u,v})∣ is a divisor of 2(pt−1)2, and Op(X{u,v})=[p3t] or [p6t]. Clearly, t≤2f.
Suppose that t=2f. Then soc(X)=PSL3(q2) or PSU3(q2), and Op(X{u,v})=[q6]. It follows that Op(Xv[1])=[q3].
Thus Op(Xv)=1 and Xv has an almost simple quotient PSU3(q).[o].
Checking Tables 8.3 and 8.5 given in ****[1]****, we conclude that X has no maximal subgroup
containing Xv, a contradiction.
If t=f then we have (X,Op(X{u,v}))=(G2(pt).Z2l+1,[q6]), and
we get a similar contradiction by checking the maximal subgroups of G2(pt).Z2l+1.
Suppose that f=t<2f. Then f>1.
Recalling that X{u,v}/Op(X{u,v}) has a
subgroup Z(3,q+1)(q2−1), we know that p2f−1 is a divisor of 2(3,q+1)(pt−1)2.
If p2f−1 has a primitive prime divisor say s, then s≥2f+1≥5, and s
is not a divisor of 2(3,q+1)(pt−1)2, a contradiction.
It follows from Zsigmondy’s theorem that 2f=6 and p=2, and so t=1 or 2.
Then 7 is a divisor of p2f−1 but not a divisor of 2(3,q+1)(pt−1)2, a contradiction.
Case 2.
Let (G0,H0) be one of the pairs in ****[17, Table 15-20]**** which is not considered in Case 1.
Assume that X{u,v}/Op(X{u,v}) is a {2,3}-group.
Then p2f−1 has no prime divisor other than 2 and 3.
It follows that f=1, and so p=q>2. Calculation shows that p∈{3,5,7}.
For q=p=3, it is easily shown that X{u,v}/Op(X{u,v}) is a 2-group.
These observations yield that
either q=p=3 and X{u,v}/Op(X{u,v}) is a 2-group,
or X{u,v} is not a {2,3}-group.
Recalling that X{u,v}/Op(X{u,v})
has a subgroup Z(3,q+1)(q2−1) and
∣Op(X{u,v})∣=pif.a for 3≤i≤6,
it follows that (X,X{u,v}) is one of the following pairs:
Then q=p∈{5,11} and Xv[1]=1.
In particular, soc(Xv)=PSU3(p), and X{u,v} is the normalizer
NX(P) of some Sylow p-subgroup P of X.
Thus Xuv=Xv∩X{u,v}≤NXv(P).
For the pairs (HS.2,[53]:[25]) and (Ru,[53]:[25]),
by the Atlas ****[3]****,
X{u,v} is a normalizer of some Sylow 5-subgroup,
which intersects a maximal subgroup PSU3(5):2 of soc(X) at [53]:8:2,
thus G, X and X{u,v} are listed in Table 5.
The other pairs are excluded as follows.
First, the group Th is excluded as it has no maximal subgroup with
a simple section PSU3(5), refer to ****[34, Table 5.8]****.
For the pair (McL,[53]:3:8), by the Atlas ****[3]****,
we have Xv=PSU3(5), and so Xuv≤NPSU3(5)(P)=[53]:8, which
contradicts that ∣X{u,v}:Xuv∣=2.
For the pair (J4,[113]:(5×2S4)), by ****[34, Table 5.8]****,
Xv=PSU3(11).2, yielding Xuv≤NXv(P)=[113]:(5×8:2),
we get a similar contradiction.
For the pair (X,X{u,v})=(Co2,[53]:4S4), by the Atlas ****[3**]****, Xv<HS.2<Co2. Checking the maximal subgroups
of HS.2, we have Xv=PSU3(5) or Xv=PSU3(5):2. It follows that
Xuv≤NXv(P)=[53]:8 or [53]:[25], and then ∣X{u,v}:Xuv∣=2, a contradiction.
**□
5. Graphs with soluble vertex-stabilizers
Let G, T, X and Γ=(V,E) be as in Hypothesis 3.1.
We now deal with the case that Gv is soluble. First, the following lemma says that Γ is not a complete bipartite graph if GvΓ(v) is soluble.
Lemma 5.1**.**
Assume that Γ≅Kd,d. Then T≅A6, d=6, Tv=PSL2(5) and Tuv≅D10. In particular, Xuv is nonabelian.
Proof. Let G+ be the subgroup of G fixing the bipartition of Γ.
Then Gv≤G+, and Gv is 2-transitive on the partite set
which does not contain v. Thus G+ acts 2-transitively on
each partite set, and these two actions are not equivalent.
Check the almost simple 2-transitive groups, refer to
****[2**, Table 7.4]**. We conclude that T≅A6 or M12,
and Tv≅A5 or M11, respectively. Since Tuv is soluble,
the lemma follows.
**□
By Lemma 5.1 and ****[20]****, we have the following result.
Lemma 5.2**.**
Assume that Xuv is abelian. Then either T≅PSL2(q) and Γ≅Kq+1, or T=Sz(22m+1), T{u,v}≅D2(22m+1−1), Tv≅Z22m+1:Z22m+1−1, AutΓ=Aut(Sz(q)) and Γ is (T,2)-arc-transitive.
Lemma 5.3**.**
Assume that (a1) or (a2) of Lemma 3.5 holds,
and Xuv is nonabelian. Then one of the following holds:
(1)
G=X* or X.2, X=M10, X{u,v}≅Z8:Z2,
Xv≅32:Q8 and Γ≅K10.*
2. (2)
G=X=PSL3(3).2, X{u,v}≅GL2(3):2, Xv≅32:GL2(3), and Γ is the point-line non-incidence graph of PG(2,3).
Proof. Case 1. Assume that Lemma 3.5 (a1) holds.
Suppose that (XvΓ(v))u=Q8. Then Xuv≲Q8×Q8.
This implies that ∣X{u,v}∣ is a divisor of
27 and divisible by 24. Checking the Tables 14-20 in ****[17**],
we have X≅PSL2(9).2=M10 and X{u,v}≅Z8:Z2;
in this case, Xv≅32:Q8, and d=9.
Then part (1) of this lemma follows.
Suppose that (XvΓ(v))u=Q8. If
p=3 and (GvΓ(v))u=Q8, then (XvΓ(v))u is abelian,
it follows that Xuv is abelian, a contradiction. Thus we have
SL2(3)⊴(GvΓ(v))u≤GL2(p), and p∈{3,5,7,11,23}. Then (GvΓ(v))u≤NGL2(p)(SL2(3))=Zp−1∘GL2(3).
Since (XvΓ(v))u is nonabelian and normal in (GvΓ(v))u, we have Q8⊴(XvΓ(v))u, and hence
SL2(3)⊴(XvΓ(v))u. Moreover, ∣X{u,v}∣ is a divisor of
2732(p−1)2 and divisible by 24.
Let M be an arbitrary normal abelian subgroup of X{u,v}. Then M∩Xuv has index at most 2 in M,
and (M∩Xuv)Xv[1]/Xv[1] is isomorphic to a normal subgroup of
(XvΓ(v))u. Thus (M∩Xuv)Xv[1]/Xv[1]≲Zp−1. Since M∩Xv[1]⊴Xv[1] and Xv[1] is isomorphic to a normal subgroup of
(XvΓ(v))u, we have M∩Xv[1]≲Zp−1.
Noting that (M∩Xuv)Xv[1]/Xv[1]≅M∩Xuv/(M∩Xv[1]), it follows that
∣M∩Xuv∣ is a divisor of (p−1)2. Thus ∣M∣ is a divisor of 2(p−1)2.
The above observations allow us to consider only the pairs
(G0,H0) in ****[17, Tables 14-20]**** which satisfy the following conditions:
(c1)
∣H0∣** is a divisor of
2732(p−1)2 and divisible by 24; H0 has a factor (a quotient of some subnormal subgroup) Q8; and H0 has no element of order 32, 52 or 112.**
2. (c2)
If M is a normal abelian subgroup of H0 then ∣M∣ is a divisor of 2(p−1)2;
if p∈{7,11,23}, the order of O2p−1(H0) is a divisor of 4(p−1)2.
Checking the those H0 which satisfy conditions (c1) and (c2), we conclude that the possible pairs (X,X{u,v}) are listed as follows:
Note these groups X are included in the Atlas ****[3]****. Inspecting the subgroups of X, only the pair (PSL3(3).2,2S4:2) gives a desired Xv≅32:GL2(3). Then part (2) of this lemma follows.
Case 2. Let 2+1+4:Z5≤(GvΓ(v))u≤2+1+4.(Z5:Z4). Then 2+1+4⊴(XvΓ(v))u, and so, ∣X{u,v}∣ is a divisor of
21552 and divisible by 26. Further, if M is a normal abelian subgroup of
X{u,v} then a similar argument as in Case 1 yields that ∣M∣ is a divisor of 25. It is easily shown that O2(Xuv)=1, and hence O2(X{u,v})=1. Checking the pairs (G0,H0) in ****[17**, Tables 14-20]****, either O2(H0)=1 or ∣H0∣ has an odd prime divisor other than 5. Thus, in this case, no desired pair (X,X{u,v}) exists.
**□
We assume next that Lemma 3.5 (a3) occurs. Thus (GvΓ(v))u≤GL1(pf) and (GvΓ(v))u≤ΓL1(pf). Then f>1 and (GvΓ(v))u≲Zpf−1:Zf. Recalling Xuv≲(XuΓ(u))v×(XvΓ(v))u≤(GuΓ(u))v×(GvΓ(v))u, we have the following simple fact.
Lemma 5.4**.**
If (a3) of Lemma 3.5 occurs then X{u,v} has no section Zt3, Zr5 or Z26, where t is a primitive prime divisor of pf−1 and r is an arbitrary odd prime.
Lemma 5.5**.**
Assume that Xuv is nonabelian and (a3) of Lemma 3.5 occurs.
Then pf=26.
Proof. Suppose that pf=26. Then X has order divisible by 26, Xuv≲Z63:Z6×Z63:Z6, and thus X{u,v} has a normal Hall 2′-subgroup and
∣X{u,v}∣ is indivisible by 24. Checking Tables 14-20 given in ****[17**], (X,X{u,v}) is one of the following pairs:
**The pair (PSL2(26),D126) yields that Xv≅26:Z63, and thus
Xuv is abelian, this is not the case. The other pairs are easily excluded as none of them gives a desired Xv.
**□
Lemma 5.6**.**
Assume that Xuv is nonabelian and (a3) of Lemma 3.5 occurs.
Suppose that Xuv has a normal abelian Hall 2′-subgroup.
Then G=X or X.2, X=M10, X{u,v}≅Z8:Z2,
Xv≅32:Q8 and Γ≅K10.
Proof. Note that X{u,v}=Xuv.2.
The unique Hall 2′-subgroup of Xuv is also the Hall 2′-subgroup of X{u,v}.
Checking Tables 14-20 given in ****[17**], we know that (X,X{u,v}) is one of the following pairs:
The pair (M10,Z8:Z2) yields that Xv≅32:Q8 and d=9.
The third pair in (i) implies that Xv≅Z32:Z8; however, Xuv is abelian, which is not the case. For (PSL2(ta),D(2,ta−1)2(ta±1)), checking the subgroups of PSL2(ta),
we have ta=pf and Xv≅Zpf:Z(2,p−1)pf−1, and then Xuv is abelian, a contradction.
The other pairs in (i) are also excluded as ∣X∣ is indivisible by pf. (Note that f>1.)
Now we deal with the pairs in (ii).
Note that, for an odd prime r, the edge-stabilizer X{u,v} has a unique Sylow r-subgroup Or(X{u,v}). Then Or(X{u,v}) is a Sylow subgroup of X by Lemma 2.7. This implies that the unique Hall 2′-subgroup of X{u,v}, say K, is a Hall subgroup of X. Since X{u,v}=Xuv.2, we have K≤Xuv. Note that ∣Xv:Xuv∣=d=pf and Xv is contained in a maximal subgroup of X.
We now check the maximal subgroups of X which contain K, refer to ****[12, II.8.27]****, ****[1, Tables 8.3-8.6, 8.14, 8.15]**** and ****[14, 26]****. Then one of the following occurs:
(iii)
X=Sz(22a+1)** and Xv≅Z22a+1:Z22a+1−1;**
2. (iv)
X=PSp4(2a).Z2b+1** and Xv≲Sp2(22a):2.Z2b;**
3. (v)
X=PSp4(2a).Z2b+1** and Xv≲Sp2(2a)≀S2.Z2b.**
**Item (iii) yields that Xuv is abelian, which is not the case.
Item (iv) gives Xuv=Xv, a contradiction.
Suppose that (v) occurs, we have Xv≅(Z2a:Z2a−1)2:2.Z2b.
Then 1=O2(Xv)≤O2(Gv), and hence
d=∣O2(Gv)∣ by Lemma 2.5. Since Xv is transitive on Γ(v),
it follows that pf=d=22a. Thus ∣Xuv∣=(2a−1)22b+1, and
so ∣X{u,v}:Xuv∣=8>2, a contradiction.
**□
Corollary 5.7**.**
Assume that Xuv is nonabelian and (a3) of Lemma 3.5 occurs.
If f=2 then G=X or X.2, X=M10, X{u,v}≅Z8:Z2,
Xv≅32:Q8 and Γ≅K10.
*Proof. ***Let f=2. Then (XvΓ(v))u≲Zp2−1.Z2. Note that X{u,v}=Xuv.2 and Xuv≲Zp2−1.Z2×Zp2−1.Z2. Then
Lemma 5.6 is applicable, and the result follows.
**□
Let π0(pf−1) be the set of primitive
primes of pf−1.
By Zsigmondy’s theorem, if π0(pf−1)=∅ and f>1 then
pf=26, or f=2 and p=2t−1, where t is a prime.
Thus, in view of Lemma 5.5 and Corollary 5.7, we assume next that π0(pf−1)=∅.
Lemma 5.8**.**
Assume that π:=π0(pf−1)=∅, Xuv is nonabelian and (a3) of Lemma 3.5 occurs.
Then f≥3, and
(1)
π=π(∣X{u,v}∣)∖{2}, min(π)≥max{5,f+1};
(2)
p≡±1(modr)* and Or(X{u,v})=1 for each r∈π;*
(3)
X{u,v}* has a unique *(nontrivial) Hall π-subgroup, which is either cyclic or a direct product of two cyclic subgroups.
*Proof. ***By the assumption and Lemma 3.6, we have
(XvΓ(v))u≅Zm′.Ze′f, and ∅=π=π0(pf−1)⊆π(m′). For r∈π, since pr−1≡1(modr), we have f≤r−1, and so r≥f+1. In particular, r≥5 and
p≡±1(modr). Recall that X{u,v}=Xuv.2 and Xuv≲Zm′.Ze′f×Zm′.Ze′f. It follows that
Or(X{u,v})=1, and Or(X{u,v}) is the unique Sylow r-subgroup of X{u,v}. Clearly, Or(X{u,v}) is either cyclic or a direct product of two cyclic subgroups.
Then X{u,v} has a unique Hall π-subgroup F, which is either cyclic or a direct product of two cyclic subgroups. Clearly, F=1 and, by Lemma 5.6, X{u,v} has no normal abelian Hall 2′-subgroup.
Then π=π(∣X{u,v}∣)∖{2}, the lemma follows.
**□
Recall that X{u,v} has no section Z26 or Z35, see Lemma 5.4. Combining with Lemma 5.8, we next check the pairs (G0,H0) listed in ****[17, Tables 14-20]****.
Lemma 5.9**.**
Assume that π0(pf−1)=∅, Xuv is nonabelian and (a3) of Lemma 3.5 occurs.
Then T=soc(X) is not a simple group of Lie type.
*Proof. *Suppose that T is a simple group of Lie type over a finite field of order q′=ta, where t is a prime.
Since T⊴G, we know that T is transitive on the edge set of Γ. Then TvΓ(v)=1. Noting that TvΓ(v)⊴GvΓ(v), we have soc(GvΓ(v))≤TvΓ(v). In particular, Tv is transitive on Γ(v), and so ∣Tv∣=pf∣Tuv∣. In view of this, noting that Tv=T∩Xv=T∩Gv and T{u,v}=T∩X{u,v}=T∩G{u,v}, we sometimes work on the triple (T,Tv,T{u,v}) instead of (X,Xv,X{u,v}).
By Lemmas 5.6 and 5.8, X{u,v} is not a {2,3}-group and has no normal abelian Hall 2′-subgroup.
Assume that t∈π0(pf−1). By Lemmas 5.4 and 5.8, t≥5, X{u,v} has no section Zt3 and Ot(X{u,v})=1 is abelian. Checking the pairs (G0,H0) listed in ****[17, Tables 16-20]****, we have X=PSL2(t2) and X{u,v}≅Zt2:Z2t2−1. For this case, checking the subgroups of PSL2(t2), no desired Xv arises, a contradiction. Therefore, t∈π0(pf−1).
By Lemma 5.8, Or(X{u,v})=1 for each r∈π0(pf−1). Recall that
X{u,v} is not a {2,3}-group and has a subgroup of index 2. Checking the pairs (G0,H0) listed in ****[17, Tables 16-20]****, we conclude that Ot(X{u,v})=1. Further, we observe that a desired X{u,v} if exists has the form of
N.K, where N is an abelian subgroup of T and either K is a {2,3}-group
or (X,K)=(E8(q′),Z30).
For the case where K≅Z30, by Lemma 3.6, π0(pf−1)⊆π(∣N∣), and thus, by Lemma 5.4, N has no subgroup Zr3 for r∈π0(pf−1). With these restrictions, only one of the following (i)-(iv) occurs.
(i) Either X=PSL3(q′) and X{u,v}≅(3,q′−1)1Zq′−12.S3 with q′=2,4, or X=PSU3(q′) and X{u,v}≅(3,q′+1)1Zq′+12.S3. Then ∣Xv∣=(3,q′∓1)3pf(q′∓1)2. Checking Tables 8.3-8.6 given in ****[1]****, we have X=PSL3(q′) and Xv≲[q′3]:(3,q′−1)1Zq′−12. It follows that p=t=3, and ∣O3(Xv)∣=3f+1=3d, which contradicts Lemma 2.5.
(ii) T=soc(X)=PΩ8+(q′) and T{u,v}≅D(2,q′−1)2(q′2+1)2.[22]. In this case, noting that ∣T{u,v}:Tuv∣≤2, we have ∣Tv∣=24pf(2,q′−1)2(q′2+1)2 or 23pf(2,q′−1)2(q′2+1)2. Let M be a maximal subgroup of T with Tv≤M.
By ****[14]****, since ∣M∣ is divisible by (q′2+1)2,
we have M≅PSL2(q′2)2.22. It is easily shown that PSL2(q′2)2.22
does not have subgroups of order 24pf(2,q′−1)2(q′2+1)2 or 23pf(2,q′−1)2(q′2+1)2, a contradiction.
(iii) (X,X{u,v}) is one of
(2F4(2)′,52:4A4) and (2F4(2),13:12).
For the first pair, we have π0(pf−1)={5} and, since pf is a divisor of
∣2F4(2)′∣, we conclude that
pf=24 or 34. The second pair implies that
π0(pf−1)={13}, and then pf=212 or 33. By the Atlas ****[3]****, X has no maximal subgroup containing Xuv as a subgroup of index divisible by pf, a contradiction.
(iv) T{u,v} has a normal abelian subgroup N listed as follows:
[TABLE]
Let M be a maximal subgroup of T with Tv≤M. Then ∣M∣ is divisible by pf∣N∣. Check the maximal subgroups of T of order divisible by ∣N∣, refer to
****[15, 19, 21]****. Then we may deduce a contradiction.
First, by ****[15, Theorem C]****, we conclude that Ree(3a) has no maximal subgroups of order divisible by pf∣N∣. Similarly, by ****[21]****, the group 2F4(2a) is excluded. We next deal with the remaining cases.
(1) Let T=G2(3a).
Suppose that ∣N∣=32a±3a+1.
By ****[15, Theorems A and B]****, since ∣M∣ is divisible by 32a±3a+1, we have M≅SL3(3a):2 or SU3(3a):2.
By ****[1, Tables 8.3-8.6]****, we conclude that Tv≲Z32a±3a+1:[6], which is impossible.
Similarly, for ∣N∣=(3a±1)2, we have
Tv≲(SL2(3a)∘SL2(3a)).2, SL3(3a):2 or SU3(3a):2.
Since ∣Tv∣ is divisible by 21∣T{u,v}∣pf=6pf(3a±1)2,
checking the maximal subgroups of SL2(3a), SL3(3a) and SU3(3a), we have p=3 and Tv≲[3ba]:Z3a−12.2 for b=2 or 3.
Since Tuv has order divisible by 3, it follows that O3(Tuv)=1, which contradicts Lemma 2.5.
(3) Let T=F4(2a). By ****[18, 19]****, noting that ∣M∣ is divisible by pf∣N∣, we conclude that M≅Sp8(2a) or PΩ8+(2a).S3 with ∣N∣=(22a+1)2, or M≅c.PSL3(2a)2.c.2 or c.PSU3(2a)2.c.2 with ∣N∣=(22a±2a+1)2, where c=(3,2a±1). Then a contradiction follows from checking the maximal subgroups of Sp8(2a), PΩ8+(2a), PSL3(2a) and PSU3(2a), refer to ****[1, Tables 8.3-8.6, 8.48-8.50]****.
**(4) Let T=E8(q′). Then ∣N∣=(q′4−q′2+1)2
and M≅PSU3(q′2)2.8. For this case, checking the maximal subgroups PSU3(q′2), we get a contradiction.
**□
Lemma 5.10**.**
Assume that π0(pf−1)=∅, Xuv is nonabelian and (a3) of Lemma 3.5 occurs.
Then G=X=J1, X{u,v}≅Z7:Z6, Xv≅Z23:Z7:Z3 and d=8.
*Proof. *By Lemma 5.9, T=soc(X) is either an alternating group or a sporadic simple group. Note that X{u,v} is not a {2,3}-group and has no normal abelian Hall 2′-subgroup.
Assume that T is an alternating group. Then, by ****[17, Table 14]****, either X=Ar and X{u,v}≅Zr:Z2r−1 for r∈{7,11,17,23}, or X=Sr and X{u,v}≅Zr:Zr−1 for r∈{7,11,17,23}. For these two cases, Xv is a transitive subgroup of
Sr in the natural action of Sr. Then either Xv is almost simple or
Xv≲Zr:Zr−1 (refer to ****[4, page 99, Corollary 3.5B]****),
a contradiction.
Assume that T is a sporadic simple group, and let r∈π0(pf−1). Then (X,X{u,v},r) is one of the following triples:
Recall that pf is a divisor of ∣X∣ and r is a primitive prime divisor of pf−1. Searching all possible pairs (pf,r), we get the following table:
[TABLE]
Recalling that G{u,v}=X{u,v}.(G/X), we have
2∣Guv∣=∣G{u,v}∣=∣X{u,v}∣∣G:X∣=2∣Xuv∣∣G:X∣, and so
∣Guv∣=∣Xuv∣∣G:X∣. Since Gv is 2-transitive on Γ(v), we know that (pf−1) is a divisor of ∣Guv∣=∣Xuv∣∣G:X∣. It follows that (X,X{u,v},r,pf) is one of
For (Co1,Z72:(3×2A4),7,23), we have Xuv≲ΓL1(23)×ΓL1(23), yielding that Xuv has odd order, a contradiction. Similarly, for (B,Z13:Z12×S4,13,33), the order of Xuv is indivisible by 24, a contradiction; for (M,Z31:Z15×S3,31,25), the order of Xuv is indivisible by 3, also a contradiction.
For (He,Z52:4A4,5,24), the order of Xuv is divisible by
23⋅3⋅52 and, since pf=24, the order of Xu is divisible by
27⋅3⋅52; however, He has no soluble subgroups of order divisible by 27⋅3⋅52, a contradiction.
Similarly, the quadruple (O′N.2,Z31:Z30,31,25) is excluded as O′N.2 has no soluble subgroups with order divisible by 25⋅31. (Note that Gv is soluble.) By the Altas ****[3]****, J2 has no subgroups with order divisible by 24⋅52, and then (J2,Z52:D12,5,24) is excluded.
By the Altas ****[3]**** and ****[33**, Theorem 2.1]****, B has no subgroups
with order divisible by 32⋅54⋅13, and then (B,Z13:Z12×S4,13,54) is excluded. Then (J1,Z7:Z6,7,23) is left, which gives Xv≅Z23:Z7:Z3, d=pf=8 and G=X.
**□
Finally, we summarize the argument for proving Theorem 1.1 as follows.
Proof of Theorem 1.1.
Clearly, each (G,Gv,G{u,v}) in Table 1
gives a G-edge-primitive graph Cos(G,Gv,G{u,v}).
It is not difficult to check the 2-arc-transitivity of G on Cos(G,Gv,G{u,v}), we omit the details.
**Now let G and Γ=(V,E) satisfy the assumptions in Theorem 1.1.
If GvΓ(v) is an almost simple 2-transitive group then, by Lemmas 3.3, 4.1-4.3, the triple (G,Gv,G{u,v}) is listed in Table 1. Assume that GvΓ(v) is a soluble 2-transitive group
of degree d=pf, where p is a prime. Then either GvΓ(v)≤ΓL1(pf), or GvΓ(v) has a normal subgroup SL2(3) or 2+1+4.
For the latter case, the triple (G,Gv,G{u,v}) is known by Lemmas 5.2 and 5.3. Let GvΓ(v)≤ΓL1(pf)
and consider the primitive prime divisors of pf−1.
If pf−1 has no primitive prime divisor then, by Lemmas 5.2, 5.5 and Corollary 5.7, (G,Gv,G{u,v}) is listed in Table 1. If pf−1 has primitive prime divisors, then (G,Gv,G{u,v}) is known by Lemmas 5.2 and 5.10.
This completes the proof.
**□
Bibliography35
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] J.N. Bray, D.F. Holt and C.M. Roney-Dougal, The Maximal Subgroups of the Low-Dimensional Finite Classical Groups , Cambridge University Press, New York, 2013.
2[2] P.J. Cameron, Permutation Groups , Cambridge University Press, Cambridge, 1999.
3[3] J.H. Conway, R.T. Curtis, S.P. Noton, R.A. Parker and R.A. Wilson, Atlas of Finite Groups , Clarendon Press, Oxford, 1985.
4[4] J.D. Dixon and B. Mortimer, Permutation Groups , Springer-Verlag, New York, 1996.
5[5] X.G. Fang and C.E. Praeger, Finite two-arc transitive graphs admitting a Suzuki simple group, Comm. Algebra 27 (1999), 3727-3754.
6[6] D.A. Foulser, The flag-transitive collineation groups of the Desarguesian affine planes, Canad. J. Math. 16 (1964), 443-472.
7[7] A. Gardiner, Arc transitivity in graphs, Quart. J. Math. Oxford 24(1973), 399-407.
8[8] A. Gardiner, Doubly primitive vertex stabilizers in graphs, Math. Z. 135 (1974), 157-166.