Arxiv:1906.09414V2 [Math.CO]

Home , Hall subgroup, Rudvalis group

arXiv:1906.09414v2 [math.CO] 8 Oct 2020 hc seg-rmtv n fvlny20 u o 2-arc-transitive not of but group 2304 automorphism valency the of is and Ru edge-primitive group is Rudvalis which sporadic the [3], Atlas n[9], in 2,2]wl eeetv ol o st netgt h group-theor the investigate to us of for properties tools effective be will 25] [24, Aut detastv n h tblzr( stabilizer the and edge-transitive Γ iial,w a en the define may we Similarly, locle narc. an called also udmna eerhFnsfrteCnrlUniversities. Central the for Funds Research Fundamental and some ymti ru Sym( group symmetric Γ t etcshv qa valency ( equal have vertices its scle h aec of valency the called is finite. be to assumed Γ h set the lmnsin elements v h graph The . h group The . 0 Let Let upre yteNtoa aua cec onaino hn (1 China of Foundation Science Natural National the 20B25. by 05C25, Supported Classification. Subject Mathematics 2010 nti ae,algah r sue ob nt n ipe n l g all and simple, and finite pair be a is to graph assumed A are graphs all paper, this In if v , Γ g { 1 g v , . . . , u samxmlsubgroup. maximal a is Γ Γ Γ ∈ e f2ac.I hsppr epeetacasfiainfrthose for classification edge-stabilizers. a soluble Keywords acts present have we and group 2-arc-transitive paper, automorphism are this its which In if 2-arcs. 2-arc-transitive of and set set, edge the Abstract. rniiegop oul group. soluble group, transitive g Γ ∈ v , ( = sas r-rniie If arc-transitive. also is Aut ( ( = v g Aut ∈ } = ) ,E V, NEG-RMTV RPSWT SOLUBLE WITH GRAPHS EDGE-PRIMITIVE ON s V Γ fvrie with vertices of ) ,E V, Γ Γ Aut fti cini rniie hti,frec aro de hr ex there edges of pair each for is, that transitive, is action this If . and { Γ However, . E ea depiiiegaho aec ols hn3 hn sobse as Then, 3. than less no valency of graph edge-primitive an be ) u apn n det h te n,then one, other the to edge one mapping depiiiegah -r-rniiegah lotsml grou simple almost graph, 2-arc-transitive graph, Edge-primitive . eagah permutation A graph. a be ) Γ rp seg-rmtv fisatmrhs ru cspiiieyon primitively acts group automorphism its if edge-primitive is graph A scalled is o all for ∈ E Γ a aua cinon action natural a has V U A,HN ILA,ADZIPN LU PING ZAI AND LIAO, CI HONG HAN, HUA ( = r aldtevrie n de of edges and vertices the called are V { | v ,dntdby denoted ), esyta h graph the that say We . { ,E V, ,v u, ,v u, edge-primitive Γ vertex-transitivity ∈ } fannmt set nonempty a of ) ∈ } EDGE-STABILIZERS { d sntncsaiy2actastv;freape ythe by example, for 2-arc-transitive; necessarily not is v o ninteger an For . Γ i E v , 1. E Aut s2actastv hnPagrsrdcintheorems reduction Praeger’s then 2-arc-transitive is } i l uoopim of automorphisms All . +1 Introduction scle h egbrodof neighborhood the called is Γ Aut ∈ } ) { u,v Γ if E 1 E hc scle h uoopimgopof group automorphism the called is which , } Aut namely, , and fsm adhneeey edge every) hence (and some of , arc-transitivity Γ g s v V ≥ cspiiieyon primitively acts Γ i on 6= n set a and ,an 1, a valency has V v { i ,v u, +2 scle natmrhs of automorphism an called is Γ s o l possible all for } acin -arc Γ epciey For respectively. , g Γ depiiiegraphs edge-primitive E = 914,1710)adthe and 11731002) 1971248, scalled is and tcadgraph-theoretic and etic omasbru fthe of subgroup a form rniieyo the on transitively { f2sbesof 2-subsets of d u sn O’Nan-Scott Using . v g Γ s or v , -arc-transitivity in sa ( an is g Γ E } ak3graph, 3 rank a Γ edge-transitive htis, that , for is while , i -r is 1-arc A . ,2- p, d s { rglrif -regular +1)-tuple op are roups ,v u, { ,v u, V v | ∈ } The . Γ ∈ rved Γ } ( ists v V in of E is ) , | . 2 HAN, LIAO, AND LU

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

g Let G be a finite group and H,K ≤ G with |K : (H ∩ K)| = 2 and ∩g GH = 1. Let [G : H] = {Hx | x ∈ G}, and define a graph Cos(G,H,K) on [G : H]∈ such that 1 {Hx,Hy} is an edge if and only if yx− ∈ 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 . { }g Then K is maximal in G, and H∩K = Guv, which has index 2 in K. Noting that ∩g GH g g ∈ fixes V pointwise, ∩g GH = 1. Further, v 7→ 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 g (1) |K :(H ∩ K)| =2, ∩g GH =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 Γ (v) [1] v in Γ . LetGv be the transitive permutation group induced by Gv on Γ (v), and let Gv EDGE-PRIMITIVITY 3

G G u,v Gv d Remark 4{ } 3 PSL4(2).2 2 :S4 2 :SL3(2) 7 8 2 6 PSL5(2).2 [2 ]:S3.2 2 :(S3×SL3(2)) 7 22 2 20 F4(2).2 [2 ]:S3.2 [2 ].(S3×SL3(2)) 7 1+4 3 PSL4(3).2 3+ :(2S4×2) 3 :SL3(3) 13 2 1+4 Z2 3 Z PSL4(3).2 3+ :(2S4× 2) 3 :(SL3(3)× 2) 13 8 2 6 PSL5(3).2 [3 ]:(2S4) .2 3 .2S4.SL3(3) 13

Sp Zp:Zp 1 PSL2(p) p+1 p ∈{7, 11} 2 − M11 3 :Q8.2 M10 10 K11 J1 Z11:Z10 PSL2(11) 12 J3.2 Z19:Z18 PSL2(19) 20 O′N.2 Z31:Z30 PSL2(31) 32 B Z19:Z18×Z2 PGL2(19) 20 B Z23:Z11×Z2 PSL2(23) 24 M Z41:Z40 PSL2(41) 42 PSL2(19) D20 PSL2(5) 6 2 ∼ A6.2, A6.2 Z5:[4], Z10:Z4 PSL2(5), PGL2(5) 6 K6,6, G 6= S6 PGL2(11) D20 PSL2(5) 6 2 PSL3(r) 3 :Q8 PSL2(9) 10 r is a prime with 2 PSL3(r).2 3 :Q8.2 PGL2(9) r ≡ 4, 16, 31, 34(mod 45) 2 PSU3(r) 3 :Q8 PSL2(9) 10 r is a prime with 2 PSU3(r).2 3 :Q8.2 PGL2(9) r ≡ 11, 14, 29, 41(mod 45) 3 5 HS.2 [5 ]:[2 ] PSU3(5):2 126 3 5 Ru [5 ]:[2 ] PSU3(5):2 126 2 M10 Z8:Z2 3 :Q8 9 K10 2 PSL3(3).2 GL2(3):2 3 :GL2(3) 9 Z Z Z3 Z Z J1 7: 6 2: 7: 3 8 f Zf Z f K PSL2(p ).[o] D 2(pf −1) .[o] p : pf −1 .[o] p pf +1, o (2,p − 1)f (2,p−1) (2,p−1) f Zf Z f Sz(2 ).o D2(2f 1).o 2 : 2f 1.o 2 f is odd, o f − − Table 1. Graphs.

Γ (v) [1] [1] [1] [1] be the kernel of Gv acting on Γ (v). Then Gv ∼= Gv/Gv . Set Guv = Gu ∩ Gv . Then [1] Γ (u) [1] Gv induces a normal subgroup of (Gu )v with the kernel Guv. Since G is transitive on the arcs of Γ , there is some element in G interchanging u and v. This implies that

Γ (v) ∼ Γ (u) |G u,v :Guv| =2 and (Gv )u = (Gu )v. { } [1] Writing Gv and Gv in group extensions, the next lemma follows.

[1] [1] [1] Γ (u) [1] Γ (u) Γ (u) Γ (v) Lemma 2.2. (1) Gv = Guv.(Gv ) , (Gv ) ✂ (Gu )v =∼ (Gv )u. [1] [1] Γ (u) Γ (v) [1] [1] Γ (u) Γ (v) (2) G u,v = Guv.2, Guv =(Guv.(Gv ) ).(Gv )u, Gv =(Guv.(Gv ) ).Gv . { [1]} Γ (v) Γ (u) (3) If Guv =1 then Guv . (Gv )u × (Gu )v. 4 HAN, LIAO, AND LU

[1] By [32], s ≤ 7 , and if s ≥ 2 then Guv 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} ∈ Γ (v) E. Then G u,v is soluble if and only if (Gv )u is soluble, and Gv is soluble if and only Γ (v) { } if Gv is soluble.

Choose s maximal as possible, that is, Γ is (G,s)-arc-transitive but not (G,s+1)-arc- [1] transitive. In this case, Γ is said to be (G,s)-transitive. If further Guv =6 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] (1) If Guv =1 then s =2 or 3. [1] [1] ✂ Γ (v) qn 1 (2) If Guv =16 then Guv is a p-group for some prime p, PSLn(q) Gv , |Γ (v)| = q −1 and 6 =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) n ≥ 3, s ≤ 3 and Øp(Gv) is given as in Table 2, where Øp(Gv) is the maximal normal p-subgroup of Gv.

[1] Øp(Gv) Guv s n q Gv n(n 1)f (n 1)2f Zp − Zp − 3 SLn 1(q)×SLn(q) ✂ Gv/Øp(Gv) Znf Zf − p p 2 a.PSLn(q) ✂ Gv/Øp(Gv) with a q − 1 n(n−1)f (n−1)(n−2)f 2 2 Zp Zp 2 a.PSLn(q) ✂ Gv/Øp(Gv) with a q − 1 20 18 [q ] [q ] 3 3 even SL2(q)×SL3(q) ✂ Gv/Øp(Gv) Z6 Z4 Z6 3 3 2 3 3 3:SL3(3) Zn+1 Z2 Zn+1 2 2 2 2 2 :SLn(2) Z11 Z14 Z8 Z11 Z11 Z14 2 , 2 2, 2 2 4 2 2 :SL4(2), 2 :SL4(2) 30 26 30 [2 ] [2 ] 2 5 2 [2 ]:SL5(2) Table 2.

Lemma 2.5. Let Γ =(V, E) be a connected (G, 2)-arc-transitive graph, and {u, v} ∈ E. [1] If r is a prime divisor of |Γ (v)| then Ør(Gv )=1=Ør(Guv), and either Ør(Gv)=1, e ∼ Ze ∼ Γ (v) or |Γ (v)| = r and Ør(Gv) = r = soc(Gv ) for some integer e ≥ 1.

Γ (v) Proof. Since Γ is (G, 2)-arc-transitive, Gv is a 2-transitive group, and thus Guv is transitive on Γ (v) \{u}. Since Ør(Guv) ✂ Guv, all Ør(Guv)-orbits on Γ (v) \{u} have [1] the same size. Noting that r is coprime to |Γ (v) \{u}|, it follows that Ør(Guv) ≤ Gv . [1] [1] [1] Since Gv ✂ Guv, we have Ør(Gv ) ≤ Ør(Guv), and so Ør(Gv )=Ør(Guv). Similarly, [1] [1] considering the action of Guv on Γ (u)\{v}, we get Ør(Gu )=Ør(Guv). Then Ør(Gu )= [1] [1] [1] [1] Ør(Guv)=Ør(Gv ) ≤ Guv. By Theorem 2.4, either Guv =1 or Guv is a p-group, where [1] [1] p is a prime divisor of |Γ (v)| − 1. It follows that Ør(Gu )=Ør(Guv)=Ør(Gv ) = 1. EDGE-PRIMITIVITY 5

[1] [1] [1] [1] [1] Note that Ør(Gv)Gv /Gv ∼= Ør(Gv)/(Ør(Gv)∩Gv ). Since Ør(Gv)∩Gv ≤ Ør(Gv )= [1] [1] [1] Γ (v) 1, we have Ør(Gv) ∼= Ør(Gv)Gv /Gv ✂ Gv/Gv ∼= Gv . Thus Ør(Gv) is isomorphic to Γ (v) Γ (v) a normal r-subgroup of Gv . This implies that either Ør(Gv)=1,orGv 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. f ′ Thus f is a divisor of r − 1, and if r is a divisor of a − 1 for some f ′ ≥ 1 then f is a divisor of f ′. Thus we have the following lemma.

f Lemma 2.6. Let a ≥ 2, f ≥ 1 and f ′ ≥ 1 be integers. If a − 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 f divisor of f ′. If f ≥ 3 then a − 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 =16 , then P is also a Sylow subgroup of G.

Proof. Assume that P =6 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 ≤hNQ(P ),Hi ≤ NG(P ) =6 G. Since H is maximal in G, we have H = hNQ(P ),Hi 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]. { } 6 HAN, LIAO, AND LU

Γ (v) Γ (v) Γ (v) For the group X in Hypothesis 3.1, we have 1 =6 Xv ✂ Gv . Note that Gv is Γ (v) 2-transitive (on Γ (v)). Then Gv is aﬃne or almost simple, see [4, Theorem 4.1B] for Γ (v) Γ (v) example. It follows that soc(Gv ) = soc(Xv ).

Γ (v) 3.1. Assume that Gv is insoluble. Then Gv is an almost simple 2-transitive group Γ (v) Γ (v) (on Γ (v)). Recall that soc(Gv ) = soc(Xv ). Checking the point-stabilizers of almost Γ (v) Γ (v) simple 2-transitive groups, since (Gv )u is soluble, we conclude that either Xv is 2- Γ (v) ∼ transitive, or Gv = 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]. Γ (v) ∼ Lemma 3.2. Suppose that Hypothesis 3.1 holds. If d = 28 then Gv 6= PSL2(8).3.

Γ (v) ∼ [1] [1] Proof. Suppose that Gv = PSL2(8).3 and d = 28. Note that Xuv ≤ Guv = 1, see Γ (v) Γ (u) Theorem 2.4. Thus Xuv . (Xv )u×(Xu )v by Lemma 2.2. Γ (v) ∼ Γ (v) ∼ ∼ Assume that Xv = PSL2(8). Then (Xv )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), D q±1 ) possibly meet 2 (2,q−1) ∼ ± ∼ ∼ ± our requirements on X u,v , yielding X u,v = D2 q 1 . Then D36 = X u,v = D2 q 1 . { } { } (2,q−1) { } (2,q−1) Calculation shows that q = 37; however, PSL2(37) has no subgroup which has a quotient PSL2(8), a contradiction. Γ (v) Γ (v) ∼ Γ (v) ∼ Γ (u) ∼ Now let Xv = Gv = PSL2(8).3. Then (Xv )u = (Xu )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 { } [1] Γ (v) [1] [1] Γ (u) Γ (u) of order 9. Since X u,v = Xuv.2= Xv .(Xv )u.2 and Xv =∼ (Xv ) ✂ (Xu )v, we 2 {3 }2 4 2 5 2 6 3 5 3 6 have |X u,v | = 2 3 , 2 3 , 2 3 , 2 3 , 2 3 or 2 3 . Checking the Tables 14-20 given { } ∼ 6 [1] ∼ in [17], we conclude that X = G2(3).2, and X u,v = [3 ]:D8. In this case, Xv = ∼ { } 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.

Γ (v) Lemma 3.3. Suppose that Hypothesis 3.1 holds and Gv is almost simple. Then one of the following holds:

Γ (v) Γ (v) (s1) Gv = Xv = PSL3(2) or PSL3(3), and d =7 or 13, respectively; Γ (v) (s2) soc(Xv ) = PSL2(q) with q > 4, and d = q+1; [1] Γ (v) 3 (s3) Guv =1, soc(Xv ) = PSU3(q) with q > 2, and d = q +1; [1] Γ (v) 2n+1 2 (s4) Guv =1, soc(Xv ) = Sz(q) with q =2 > 2, and d = q +1; [1] Γ (v) 2n+1 3 (s5) Guv =1, soc(Xv ) = Ree(q) with q =3 > 3, and d = q +1.

In particular, Γ is (X, 2)-arc-transitive. EDGE-PRIMITIVITY 7

Recall that the Fitting subgroup Fit(H) of a ﬁnite group H is the direct product of Ør(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. f [1] Let q = p for some prime p. Assume that Xu,v = 1. Then Fit(Xuv)=Øp(Xuv), and either Fit(Xuv) = Fit(X u,v ) or Fit(X u,v ) = Fit(Xuv).2; in particular, |Fit(X u,v ): { } { } { } Øp(X u,v )| ≤ 2. { }

Proof. Let r be a prime divisor of |Xuv|. Then Ør(Xuv) is normal in Xuv. Since Γ is (X, 2)-arc-transitive, Xuv acts transitively on Γ (v) \{u}. Thus all Ør(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 Ør(Xuv) = 1. Then Fit(Xuv)=Øp(Xuv).

Note that Xuv is normal in X u,v as |X u,v : Xuv| = 2. Since Øp(Xuv) is a charac- { } { } teristic subgroup of Xuv, it follows that Øp(Xuv) is normal in X u,v , and so Øp(Xuv) ≤ { } Øp(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 Ør(X u,v ) ≤ Xuv, andso Ør(X u,v )=Ør(Xuv). It follows that Fit(X u,v )= { } { } { } Fit(Xuv)Ø2(X u,v )=Øp(Xuv)Ø2(X u,v ). In particular, Øp(Xuv)=Øp(X u,v ) if p =6 2. { } { } { } It is easily shown that Xuv ∩ Ø2(X u,v )= Ø2(Xuv). If Xuv ≥ Ø2(X u,v ) then { } { } p = 2, Fit(X u,v ) =Ø2(X u,v ) = Fit(Xuv), and the lemma holds. Assume that { } { } Ø2(X u,v ) 6≤ Xuv. Since |X u,v : Xuv| = 2, we have X u,v = XuvØ2(X u,v ). Then { } { } { } { } 2|Xuv| = |X u,v | = |Xuv||Ø2(X u,v ):(Xuv ∩Ø2(X u,v ))| = |Xuv||Ø2(X u,v ):Ø2(Xuv)|, { } { } { } { } yielding |Ø2(X u,v ) :Ø2(Xuv)| = 2. If p = 2 then Fit(X u,v ) = Ø2(X u,v ) and { } { } { } Fit(Xuv)=Ø2(Xuv), the lemma follows. If p =6 2 then Ø2(Xuv) = 1, |Ø2(X u,v )| = 2, { } and so Fit(X u,v )=Øp(Xuv) × Z2. This completes the proof. { } Γ (v) 3.2. Assume that Hypothesis 3.1 holds and Gv is soluble. Then Gv is an aﬃne 2- Γ (v) Zf f transitive group. Let soc(Gv ) = p . Then d = p . Recalling that d ≥ 6, we have [1] Γ (v) Γ (u) Guv = 1 by Theorem 2.4, and so Guv . (Gv )u×(Gu )v. If Guv is abelian then Γ Γ (v) is known by [20]. Thus we assume further that Guv is not abelian. Then (Gv )u is Γ (v) f Γ (v) nonabelian, and so (Gv )u 6≤ GL1(p ); in particular, f > 1. Since (Gv )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. Γ (v) Zf Let soc(Gv )= p , where p is a prime. Then f > 1, and one of the following holds.

Γ (v) (a1) f =2, and either SL2(3) ✂ (Gv )u ≤ GL2(p) and p ∈{3, 5, 7, 11, 23}, or p =3 Γ (v) and (Gv )u =Q8; 1+4 Z Γ (v) 1+4 Z Z 1+4 f 4 (a2) 2+ : 5 ≤ (Gv )u ≤ 2+ .( 5: 4) < 2+ .S5, and p =3 ; Γ (v) f Γ (v) f Γ (v) f (a3) (Gv )u 6≤ GL1(p ), (Gv )u ≤ ΓL1(p ) and |(Gv )u| is divisible by p − 1. Consider the case (a3) in Lemma 3.5. Write f pf 1 f 1 p ΓL1(p )= hτ, σ | τ − =1= σ , σ− τσ = τ i. Γ (v) m f Let hτi ∩ (Gv )u = hτ i, where m (p − 1). Then Γ (v) m ∼ Γ (v) (Gv )u/hτ i = hτi(Gv )u/hτi . hσi. 8 HAN, LIAO, AND LU

Γ (v) m e Set (Gv )u/hτ i ∼= hσ i for some divisor e of f. Then Γ (v) ∼ Z Z (Gv )u = pf −1 . f . m e l k Γ (v) Γ (v) m l k l k f m Choose τ σ ∈ (Gv )u such that (Gv )u = hτ ihτ σ i. Then (τ σ ) e ∈ hτ i but l k j m f k f k ie (τ σ ) 6∈ hτ i for 1 ≤ j < e . It follows that σ has order e . Then σ = σ for some i f k i′ e l k with (i, e ) = 1, and then (σ ) = σ for some i′. Thus, replacing τ σ by a power of it if necessary, we may let k = e. Then Γ (v) m l e (Gv )u = hτ ihτ σ i. Γ (v) m m i l e Further, (Gv )u = hτ ih(τ ) τ σ i 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

e (>) π(m) ⊆ π(p − 1), me 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.) [1] [1] Since Xuv ≤ Guv = 1, we have [1] Γ (u) ∼ Γ (v) Γ (u) Γ (v) Xv ✂ (Xu )v = (Xv )u, Xuv . (Xu )v×(Xv )u. Γ (v) This yields that (Xv )u is nonabelian. Then a limitation on π(|Xuv|) is given as follows.

Γ (v) Lemma 3.6. Assume that Lemma 3.5 (a3) holds and Xuv is nonabelian. Then (Xv )u ∼= Zm′ .Z f , where m′ and e′ satisfy e′ ∼ Γ (v) m f (i) Zm′ = (Xv )u ∩hτ i, mm′ p − 1, e e′ f; and f e′ (ii) m′ > 1, e′ < f, π(p − 1) \ π(p − 1) ⊆ π(m′) ⊆ π(|Xuv|).

Γ (v) Γ (v) m l e ∼ Z Z Proof. Recall that (Xv )u ✂ (Gv )u = hτ ihτ σ i = pf −1 . f . Then m e Γ (v) Γ (v) m ∼ Γ (v) m m Z (Xv )u/((Xv )u ∩hτ i) = (Xv )uhτ i/hτ i . f , e Γ (v) yielding (Xv )u ∼= Zm′ .Z f with m′ and e′ satisfying (i). Since Xuv is nonabelian, e′ Γ (v) (Xv )u is nonabelian, and so m′ > 1 and e′ < f. > f e′ m pf 1 By the above ( ), each r ∈ π(p − 1) \ π(p − 1) is a divisor of |hτ i| = m− . Let R m Γ (v) be the unique subgroup of order r of hτ i. Then, noting that R is normal in (Gv )u, Γ (v) Γ (v) Γ (v) either R ≤ (Xv )u or R(Xv )u = R×(Xv )u. Suppose that the latter case occurs. ′ n e′ Γ (v) m e′ pe Since e′ < f, we may let τ σ ∈ (Xv )u \hτ i. Then σ centralizes R. Thus x = x e′ Γ (v) m ∼ for x ∈ R, yielding r (p − 1), a contradiction. Then R ≤ (Xv )u ∩hτ i = 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 EDGE-PRIMITIVITY 9

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.

Lemma 4.1. Lemma 3.3 (s4) and (s5) do not occur.

[1] Proof. Suppose that Lemma 3.3 (s4) or (s5) holds. Then Xuv = 1 by Theorem 2.4. Thus [1] Γ (v) [1] [1] Γ (u) Γ (u) Γ (v) Γ (v) Γ (u) Xv = Xv .Xv , Xv ∼= (Xv ) ✂(Xu )v =∼ (Xv )u, and Xuv . (Xv )u×(Xu )v. f Γ (v) Γ (v) Set q = p with p a prime. Then the pair (Xv , (Xv )u) is given as follows:

Γ (v) Γ (v) Xv (Xv )u Sz(q).e pf+f :(q − 1).e e a divisor of f, p =2, odd f > 1 Ree(q).e pf+2f :(q − 1).e e a divisor of f, p =3, odd f > 1

In particular, Øp(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 Øp(X u,v ) is nonabelian, those pairs (G0,H0) with Øp(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. { } Γ (v) Case 1. Suppose that soc(Xv ) = Ree(q). Then p = 3, Ø3(X u,v ) is nonabelian 3f 4f 5f 6f 2 {2 } and of order 3 , 3 , 3 or 3 , |X u,v | is a divisor of 2(q − 1) f 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 t Z 6t Z2 Z t 3t Z conclude that (X,X u,v ) is one of (G2(3 ). 2l+1 , [3 ]: t . 2l+1 ) and (Ree(3 ), [3 ]: 3t 1), { } 3 1 − where 2l is the 2-part of t. If f > t then, by Zsigmondy’s− theorem, q − 1 has a prime t divisor which is not a divisor of 3 − 1, which contradicts that |X u,v | is divisible by Z ∼ {6 Z} 2 Z 2(q − 1). Thus f = t, and then X = G2(q). 2l+1 and X u,v = [q ]: q 1. 2l+1 . This { } − [1] [1] 3 implies that Xv =6 1, in fact, |Ø3(Xv )| = q . Thus Ø3(Xv) =6 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 f r is a prime. Since f > 1 is odd, 3 − 1 has an odd prime divisor, and so X u,v is not f { } a {2, 3}-group as |X u,v | is divisible by 3 − 1. Recall that Ø3(X u,v ) is nonabelian 3f 4f{ }5f 6f { } and of order 3 , 3 , 3 or 3 . Checking the groups H0 given in [17, Table 16-19], t t we conclude that soc(X) = PSLn(r ) or PSUn(r ), 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 t t t t in [1], we conclude that none of PSL3(r ), PSL4(r ), PSU3(r ) and PSU4(r ) has such maximal subgroups, a contradiction. 10 HAN, LIAO, AND LU

Γ (v) f 2f 3f Case 2. Suppose that soc(Xv ) = Sz(q). Then q = 2 , |Ø2(X u,v )| = 2 a, 2 a 4f { } or 2 a, where f > 1 is odd, and a = 1 or 2. Noting that X u,v has order divis- f { } ible by 2 − 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 2{ } 9 t Z 4t Z2 Z t 2t Z { } ( F4(2)′, [2 ]:5:4), (PSp4(2 ). 2l+1 , [2 ]: 2t 1. 2l+1 ) and (Sz(2 ), [2 ]: 2t 1), or soc(X) is t t − l − one of PSLn(r ) and PSUn(r ), where n ∈{3, 4},2 is the 2-part of t, and r is odd if n = 4. 3 The ﬁrst pair leads to q = 2 , and so |X u,v | is divisible by 7, a contradiction. Check- ing the maximal subgroups of soc(X) (refer{ } to [1, Tables 8.3-8.6, 8.8-8.14]), the groups t t t t PSL3(r ), PSU3(r ), PSL4(r ) and PSU4(r ) are excluded as they have no maximal sub- t Z 4t Z2 Z group with a simple section Sz(q). Thus (X,X u,v ) = (PSp4(2 ). 2l+1 , [2 ]: 2t 1. 2l+1 ) t 2t { } − or (Sz(2 ), [2 ]:Z2t 1). − f Since f > 1 is odd, 2 − 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. Z [1] ∼ 2 Z It follows that X = PSp4(q). 2l+1 , and Xv = [q ]: q 1. However, by [1, Table 8.14], Z 2−Z PSp4(q). 2l+1 has no maximal subgroup containing [q ]: q 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. { }

G X X u,v Xv s d 4{ } 3 X PSL4(2).2, S8 2 :S4 2 :SL3(2) 2 7 8 2 6 X PSL5(2).2 [2 ]:S3.2 2 :(S3×SL3(2)) 3 7 22 2 20 X F4(2).2 [2 ]:S3.2 [2 ].(S3×SL3(2)) 3 7 1+4 3 X,X.2 PSL4(3).2 3+ :(2S4×2) 3 :SL3(3) 2 13 8 2 6 X PSL5(3).2 [3 ]:(2S4) .2 3 .2S4.SL3(3) 3 13 Table 3.

[1] [1] Γ (v) [1] [1] Γ (u) Γ (u) Proof. Assume ﬁrst that Xuv = 1. Then, Xv = Xv .Xv , Xv ∼= (Xv ) ✂(Xu )v =∼ Γ (v) Γ (v) Γ (u) (Xv )u, and Xuv . (Xv )u×(Xu )v. Γ (v) Γ (v) ∼ [1] Suppose that Xv = PSL3(2). Then (Xv )u = S4, and thus Xv and X u,v are given as follows: { }

[1] 2 Xv 1 2 A4 S4 2 4 4 2 4 2 X u,v 2 :S3.2 2 .S3.2 2 :3 .[4] 2 :S3.2 { } 2 5 In particular, 2 ≤|Ø2(X u,v )| ≤ 2 . 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, ∼ 4 2 [1] ∼ ∼ 1+4 ∼ X u,v = 2 :S3 and Xv = A4; or X = M12 with X u,v = 2+ :S3; or X = PSU4(2) { } ∼ 2 { } with X u,v = 2A4.2. The group A8 is excluded as it has no subgroup of the form of [1] { } Xv .PSL3(2). The groups M12 and PSU4(2) are excluded as their orders are not divisible by d = 7. EDGE-PRIMITIVITY 11

Γ (v) Γ (v) ∼ 2 [1] Suppose that Xv = PSL3(3). Then (Xv )u = 3 :2S4. Thus Xv and X u,v are given as follows: { } [1] 2 2 2 2 2 Xv 1 3 3 :2 3 Q8 3 :2A4 3 :2S4 2 4 4 4 2 4 2 4 2 X u,v 3 :2S4.2 3 :2S4.2 3 :([4].S4).2 3 :Q8.S3.2 3 :(2A4) .[4] 3 :(2S4) .2 { } ∼ 2 4 Note that Ø3(X u,v ) = 3 or 3 . Checking the possible pairs (X,X u,v ), we have ∼ 4 3 { } + { } X u,v = 3 :2 .S4 and X = A12 or PΩ8 (2); in this case, d = 13 is not a divisor of |X|, a contradiction.{ } [1] Now let Xuv be a nontrivial p-group. Then, by Theorem 2.4, Xv and X u,v are given as follows: { } Xv X u,v s d p 6 8 { 2} 2 .(S3×SL3(2)) [2 ].S3.2 3 7 2 20 22 2 [2 ].(S3×SL3(2)) [2 ].S3.2 3 7 2 3 5 2 .SL3(2) [2 ].S3.2 2 7 2 4 6 2 :SL3(2) [2 ].S3.2 2 7 2 6 8 3 .(2A4×SL3(3)) [3 ].(2A4×2S4).2 3 13 3 6 8 2 3 .(2S4×SL3(3)) [3 ].(2S4) .2 3 13 3 3 5 3 .SL3(3) [3 ].2S4.2 2 13 3 3 5 3 .(2×SL3(3)) [3 ].(2×2S4).2 2 13 3 6 8 3 :SL3(3) [3 ].2S4.2 2 13 3

Suppose that p = 2. Then |X u,v | is divisible by 9 if and only if |Ø2(X u,v )| ≥ 8, and { } 22 { } Ø2(X u,v ) contains no elements of order 8 unless |Ø2(X u,v )| ≥ 2 . Check the pairs { } { } (G0,H0) given in [17, Tables 14-20] by estimating |H0| and |Ø2(H0)|. We conclude that one of the following holds: ∼ 4 (i) X = PSL4(2).2 = S8 and X u,v =2 :S4; { }8 2 (ii) X = PSL5(2).2 and X u,v = [2 ].S3.2; { } 22 2 (iii) X =F4(2).2 and X u,v = [2 ].S3.2; { } 6 (iv) soc(X) = PSL3(4) and |Ø2(X u,v )| =2 ; { } 7 (v) X = PSU4(3).23 and |Ø2(X u,v )| =2 ; 8 { 2 } (vi) X = He.2 and X u,v = [2 ]:S3.2. { } ∼ 3 4 Case (iv) yields that Xv = 2 :SL3(2) or 2 :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]. 5 8 Now let p = 3. Then |Ø3(X u,v )| = 3 or 3 , 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: { } 4 1+4 2 8 2 1+4 (F4(8).2, 9 .(2+ :S3).2), (PSL5(3).2, [3 ]:(2S4) .2), (PSL4(3).2, 3+ :(2×2S4)). ∼ Z4 Note that Ø3(Xv) ≤ Ø3(X u,v ). Then, for the ﬁrst pair, Ø3(X u,v ) = 9 has no Z6 { } { } subgroup isomorphic to 3, 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). 12 HAN, LIAO, AND LU ∼ The last pair implies that X 6= 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. { } Γ (v) f Proof. Let Xv = PSL2(q).[o], and q = p > 4, where p is a prime and o (2, q − 1)f. [1] Note that Γ is (X, 2)-arc-transitive, see Lemma 3.3. By Theorem 2.4, if Xuv =6 1 then Γ [1] is (X, 4)-arc-transitive. Thus we assume next that Xuv = 1, and then Lemma 3.4 works. [1] Γ (v) [1] ∼ [1] Γ (u) ✂ Γ (u) ∼ Γ (v) f q 1 Note that Xv = Xv .Xv , Xv = (Xv ) (Xu )v = (Xv )u = p : (2,q− 1) .[o], − Γ (v) Γ (u) Zif and Xuv . (Xv )u×(Xu )v. We have Øp(X u,v )= p .a, where i ∈{1, 2} and a is a { } [1] Zf divisor of (2,p). It is easily shown that i = 2 if and only if Øp(Xv ) = p . Combining with Lemma 3.4, we need only consider those pairs (G0,H0) in [17, Tables 14-20] which satisfy Zif (a) Øp(H0)= p .a, where i ∈{1, 2} and a is a divisor of (2,p); |Fit(H0):Øp(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 2 (b) |H0:Øp(H0)| is a divisor of 2(q − 1) f and divisible by q − 1; if i = 1 then |H0:Øp(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 2 the restrictions (a) and (b), the only pairs (G0,H0) are listed as follows: (M11, 3 :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, Øp(H0) is a Sylow p- [1] f subgroup of G0. This yields that Xv = 1, and so soc(Xv) = PSL2(p ).

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 =6 p.

Suppose that r = p. Then, by (a), either Øp(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 f 16-20] such that Øp(H0) is abelian. Thus we have r = p = 2. Recalling that p > 4 and f |H0/Øp(H0)| is divisible by 2 − 1, it follows from Lemma 2.6 that H0/Øp(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: EDGE-PRIMITIVITY 13

G X X u,v Xv d Remark { } Sp Sp Zp:Zp 1 PSL2(p) p+1 p ∈{7, 11}, Γ bipartite 2 − M11 M11 3 :Q8.2 M10 10 K11 J1 J1 Z11:Z10 PSL2(11) 12 J3.2 J3.2 Z19:Z18 PSL2(19) 20 Γ bipartite O′N.2 O′N.2 Z31:Z30 PSL2(31) 32 Γ bipartite B B Z19:Z18×Z2 PGL2(19) 20 Xv < Th < B Z23:Z11×Z2 PSL2(23) 24 Xv < Fi23 < B M M Z41:Z40 PSL2(41) 42 see [22] for Xv PSL2(19) PSL2(19) D20 PSL2(5) 6 X,X.2 PGL2(9) D20 PSL2(5) 6 K6,6 X,X.2 M10 Z5:Z4 PSL2(5) 6 K6,6 PGL2(11) PGL2(11) D20 PSL2(5) 6 Γ bipartite 2 X,X.2 PSL3(r) 3 :Q8 PSL2(9) 10 r prime, [1, Tables 8.3, 8.4] r ≡ 4, 16, 31, 34(mod 45) 2 X,X.2 PSU3(r) 3 :Q8 PSL2(9) 10 r prime, [1, Tables 8.5, 8.6] r ≡ 11, 14, 29, 41(mod 45) Table 4.

t t t 3t (2t 1)2 t 3t Z Z t − Z 2t− (PSL2(2 ), 2: 2 1), (PSL3(2 ), [2 ]:[ (3,2t 1) ].2); (PSU3(2 ), [2 ]: 2 1 ); − − (3,2t+1) t Z 4t Z2 Z l (PSp4(2 ). 2l+1 , [2 ]: 2t 1. 2l+1 ), where 2 is the 2-part of t; t 2t 3 − 11 2 9 (Sz(2 ), [2 ]:Z2t 1); ( D4(2), [2 ]:(Z7 × S3)); ( F4(2)′, [2 ]:5:4). − t 2t t First, the pair (Sz(2 ), [2 ]:Z2t 1) is excluded as Sz(2 ) has no subgroup with a section f − PSL2(2 ). For the last two pairs, we have f = 5 and 4 respectively, which yields that f 2 − 1 is not a divisor of |H0|, a contradiction. For the three pairs after the ﬁrst one, we f have t < f, thus G0 has no maximal subgroup with a section PSL2(2 ), a contradiction. t Zt Z Suppose ﬁnally that (X,X u,v ) = (PSL2(2 ), 2: 2t 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 2f [1] f t =2f. Then Ø2(X u,v )=2 , yielding |Ø2(Xv )| =2 . Thus Ø2(Xv) =6 1 and Xv has { } f 2f a simple section PSL2(2 ). Checking the subgroups of PSL2(2 ), refer to [12, II.8.27], 2f we conclude that PSL2(2 ) has no subgroup isomorphic to Xv, a contradiction. Assume that r =6 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 =6 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 t (G0,H0) =(E8(q′), Zq′8 q′7 q′5 q′4 q′3 q′+1.Z30), where q′ = r . Suppose that (G0,H0)= ± ∓ − ∓ ± 8 7 5 4 3 (E8(q′), Zq′8 q′7 q′5 q′4 q′3 q′+1.Z30). Then q′ ±q′ ∓q′ −q′ ∓q′ ±q′+1 is divisible by some ± ∓ − ∓ ± 15 30 primitive prime divisor s of q′ − 1 or of q′ − 1. Noting that s ≥ 17, we know that H0 8 7 5 has normal cyclic Sylow s-subgroup. It follows from (a) that 17 ≤ p = s = q′ ±q′ ∓q′ − 4 3 q′ ∓q′ ±q′+1. In particular, Øp(H0)= Zp and f = 1. By (b), |H0| is divisible by p − 1, 8 7 5 4 3 and then 30 is divisible by p−1. This implies that 30 = p−1= q′ ±q′ ∓q′ −q′ ∓q′ ±q′, which is impossible. Therefore, H0/Fit(H0) is a {2, 3}-group. 14 HAN, LIAO, AND LU

By (a), Fit(H0) a {2,p}-group. Then |H0| has no prime divisor other than 2, 3 f and p. Since p − 1 is a divisor of |H0|, by Lemma 2.6, we have f < 3. Recall that Γ (u) ∼ Γ (v) f q 1 Γ (v) Γ (u) (Xu )v = (Xv )u = p : (2,q− 1) .[o], and Xuv . (Xv )u×(Xu )v, where o is a divisor − of (2, q − 1)f. It follows that Xuv/Øp(Xuv) has an abelian Hall 2′-subgroup. Note that XuvØp(X u,v )/Øp(X u,v ) ∼= Xuv/(Øp(X u,v ∩ Xuv) = Xuv/Øp(Xuv), and |X u,v : { } { } { } { } XuvØp(X u,v )| ≤ 2. It follows that X u,v /Øp(X u,v ) has an abelian Hall 2′-subgroup. { } { } { } Thus, as a possible candidate for X u,v , the quotient of H0 over Øp(H0) has abelian { } Hall 2′-subgroups. In particular, H0/Øp(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:

t t (rt 1)2 t (rt+1)2 Z t± Z − (PSL2(r ), r 1 : 2), (PSL3(r ), [ (3,rt 1) ].S3), (PSU3(r ), [ (3,rt+1) ].S3); (2,rt−1) − t Z Z2 l+4 t Z Z l+3 (PSp4(2 ). 2l+1 , 2t 1.[2 ]), (PSp4(2 ). 2l+1 , 22t+1.[2 ]), where t ≥ 3; t Z Z ± t Z Z (Sz(2 ), 2t 1: 2), (Sz(2 ), 2t √2t+1+1: 4), where t ≥ 3;; t Z− Z ± t Z Z (Ree(3 ), 3t √3t+1+1. 6), (Ree(3 ), 3t+1. 6), where t ≥ 3;; t Z ±Z2 l+3 t Z Z l+2 (G2(3 ). 2l+1 , 3t 1.[3 · 2 ]), (G2(3 ). 2l+1 , 32t 3t+1.[3 · 2 ]), where t ≥ 2; 3 t Z ± Z 2 t Z ± Z ( D4(r ), r4t r2t+1: 4); ( F4(2 ), 22t √23t+1+2t √2t+1+1. 12), where t ≥ 3; t − l+3 ± ± (F4(2 ).Z2l+1 , Z24t 22t+1.[3 · 2 ]), where t ≥ 2; − l where the power 2 appeared means the 2-part of t. Recall that |Fit(H0):Øp(H0)| ≤ 2 f f and |H0 : Øp(H0)| is divisible by p − 1. This allows us determine the values of p t and r . As an example, we only deal with the second pair. Suppose that (G0,H0) = t (rt 1)2 t − (PSL3(r ), [ (3,rt 1) ].S3). Considering the structures of Fit(H0) and Øp(H0), either (3,r − 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 −1−1or(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, 2 we conclude that (G0,H0,p,f) is one of (PSL2(19), D20, 5, 1), (PSL3(8), 7 :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/Øp(H0) is a {2, 3}-group. Recall that r =6 p. Suppose that p = 2. Recalling that q = 2f > 4, we have f ≥ 3. In particular, f since |H0| is divisible by 2 − 1, H0 is not a {2, 3}-group by Lemma 2.6. Then the 2 9 9 only possibility is that G0 = F4(2)′ and H0 = [2 ]:5:4. Thus |Ø2(H0)| = 2 , it follows 4 9 from (a) that f = 4 or 9, and then G0 has a section PSL2(2 ) or PSL2(2 ), which is 2 impossible by checking the (maximal) subgroups of F4(2)′. Thus p> 2, and H0/Øp(H0) Zif is a {2, 3}-group; in particular, by (a), Øp(H0)= p 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 Øp(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 =6 p> 2 and Øp(H0)= EDGE-PRIMITIVITY 15

Zif Z2 Z2 p , it follows that either Øp(H0)= 3 or (G0,H0)=(F4(2).4, 7:(3 × SL2(3)).4). Since f 2 3 is a divisor of (q − 1)f, we get G0 =F4(2).4 and q = p =7or7 . By (b), for q =7 2 or 7 , 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. Inspect- ing carefully the remaining pairs, the possible candidates for (X,X u,v ) are as follows: { }

(PGL2(9), D20), (M10, Z5:Z4), (PGL2(11), D20); 2 (PSL3(r), 3 :Q8), where r ≡ 4, 7(mod 9); Z Z 2 5 (PSp4(4).4, 17: 16), (PSp4(4).4, 5 :[2 ]); 2 (PSU3(r), 3 :Q8), where 5

For the ﬁrst three pairs, G, X and X u,v are easily determined and given as in Table Z Z { } 4. The pair (PSp4(4).4, 17: 16) is excluded as PSp4(4).4 has no subgroup PSL2(17), 2 2 and the pair ( F4(2), Z13:Z12) is excluded as F4(2) has no subgroup PSL2(13). Suppose ∼ t ∼ 2 ∼ that X = PSU3(2 ) and X u,v = 3 :Q8. Then we have Xv = PSL2(9); however, by t{ } [1, Tables 8.3, 8.4], PSU3(2 ) has no subgroup PSL2(9), a contradiction. Suppose that 2 5 (X,X u,v ) = (PSp4(4).4, 5 :[2 ]). 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 ∼ 2 [3], we conclude that Xv ≤ M = (A5 × A5):2 < PSp4(4).2 < PSp4(4).4. Note that 2 4 Xuv =5 :[2 ], 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 ∼ 2 that Xv = M = (A5 × A5):2 , yielding d = |Xv : Xuv| = 36 =6 q + 1, a contradiction. 2 ∼ 2 Let (X,X u,v ) = (PSL3(r), 3 :Q8). Then Xuv = 3 :4. It is easily shown that p = 3 ∼ { } and Xv = PSL2(9). Since r ≡ 4, 7(mod 9), we know that PSL3(r) has a Sylow 3- Z2 subgroup 3. By [1, Tables 8.3, 8.4], PSL3(r) has a subgroup PSL2(9) if and only if r ≡ 1, 4(mod 15). Thus, in this case, we have r ≡ 11, 14, 29, 41(mod 45). For a subgroup PSL2(9) of PSL3(r), taking a Sylow 3-subgroup Q of PSL2(9), the normalizers 2 2 of Q in PSL2(9) and PSL3(r) are (isomorphic to) 3 :4 and 3 :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 ∼ ∼ 2 G = X or X.2, and if G = X.2 then Gv = Xv.2 = PGL2(9) and G u,v = 3 :Q8.2. The 2 { } pair (PSU3(r), 3 :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. { }

G X X u,v Xv d Remark 3{ }5 HS.2 HS.2 [5 ]:[2 ] PSU3(5):2 126 Γ bipartite 3 5 Ru Ru [5 ]:[2 ] PSU3(5):2 126 Table 5. 16 HAN, LIAO, AND LU

Γ (v) f Proof. Let Xv = PSU3(q).[o] and q = p > 2, where p is a prime and o 2(3, q+1)f. Γ (v) f+2f q2 1 [1] Then (Xv )u = p : − .[o], and Xuv = 1 by Theorem 2.4. Thus |Øp(X u,v )| = (3,q+1) { } 3f 4f 5f 6f p .a, p .a, p .a or p .a, where a is a divisor of (2,p). Moreover, Øp(X u,v ) is non- { } abelian, and X u,v /Øp(X u,v ) has a subgroup Z (q2−1) . We next determine which pair { } { } (3,q+1) (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 Øp(X u,v ) is nonabelian, we conclude that (X, Øp(X u,v )) is one of the following pairs: { } { }

t 3t t 3t (PSL3(p ).2, [p ]), (PGL3(p ).2, [p ]) (with p = 2), t 3t t Z 4t (PSU3(p ), [p ]), (PSp4(p ). 2l+1 , [p ]) (with p = 2), t 2t t 3t t 6t (Sz(p ), [p ]), (Ree(p ), [p ]) and (G2(p ).Z2l+1 , [p ]),

l t Z t where 2 is the 2-part of t. Check the maximal subgroups of PSp4(p ). 2l+1 , Sz(p ) and Ree(pt), refer to [1, Table 8.14], [26, Theorem 9] and [15, Theorem C], respectively. We f Z t t conclude that none of PSp4(t ). 2l+1 , Sz(p ) and Ree(p ) has maximal subgroups with a simple section PSU3(q), and they are excluded. For the ﬁrst three and the last pairs, t 2 3t 6t |X/Øp(X u,v )| is a divisor of 2(p − 1) , and Øp(X u,v ) = [p ] or [p ]. Clearly, t ≤ 2f. { } { } 2 2 6 Suppose that t = 2f. Then soc(X) = PSL3(q ) or PSU3(q ), and Øp(X u,v ) = [q ]. { } [1] 3 It follows that Øp(Xv ) = [q ]. Thus Øp(Xv) =6 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 t 6 (X, Øp(X u,v )) = (G2(p ).Z2l+1 , [q ]), and we get a similar contradiction by checking { } t the maximal subgroups of G2(p ).Z2l+1 .

Suppose that f =6 t < 2f. Then f > 1. Recalling that X u,v /Øp(X u,v ) has a 2f { }t 2 { } 2f subgroup Z (q2−1) , we know that p − 1 is a divisor of 2(3, q + 1)(p − 1) . If p − 1 (3,q+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 2f in Case 1. Assume that X u,v /Øp(X u,v ) is a {2, 3}-group. Then p − 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 /Øp(X u,v ) is a 2-group. { } { } These observations yield that either q = p = 3 and X u,v /Øp(X u,v ) is a 2-group, or { } { } X u,v is not a {2, 3}-group. { } if Recalling that X u,v /Øp(X u,v ) has a subgroup Z (q2−1) and |Øp(X u,v )| = p .a for { } { } (3,q+1) { } 3 ≤ i ≤ 6, it follows that (X,X u,v ) is one of the following pairs: { } 3 5 3 5 3 3 (HS.2, [5 ]:[2 ]), (Ru, [5 ]:[2 ]), (McL, [5 ]:3:8), (Co2, [5 ]:4S4), 3 3 (Th, [5 ]:4S4), (J4, [11 ]:(5 × 2S4)). EDGE-PRIMITIVITY 17

[1] Then q = p ∈ {5, 11} and Xv = 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 ≤ 3 5 3 5 { } NXv (P ). For the pairs (HS.2, [5 ]:[2 ]) and (Ru, [5 ]:[2 ]), by the Atlas [3], X u,v is a { } normalizer of some Sylow 5-subgroup, which intersects a maximal subgroup PSU3(5):2 3 of soc(X) at [5 ]: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 3 PSU3(5), refer to [34, Table 5.8]. For the pair (McL, [5 ]:3:8), by the Atlas [3], we have 3 Xv = PSU3(5), and so Xuv ≤ NPSU3(5)(P ) = [5 ]:8, which contradicts that |X u,v : 3 { } Xuv| = 2. For the pair (J4, [11 ]:(5 × 2S4)), by [34, Table 5.8], Xv = PSU3(11).2, 3 yielding Xuv ≤ NXv (P ) = [11 ]:(5 × 8:2), we get a similar contradiction. For the 3 pair (X,X u,v ) = (Co2, [5 ]: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 3 3 5 Xuv ≤ NXv (P ) = [5 ]:8 or [5 ]:[2 ], and then |X u,v : Xuv|= 6 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 Γ (v) if Gv 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 2m+1 ∼ ∼ Z2m+1 Z Aut Aut T = Sz(2 ), T u,v = D2(22m+1 1), Tv = 2 : 22m+1 1, Γ = (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: ∼ ∼ 2 ∼ (1) G = X or X.2, X = M10, X u,v = Z8:Z2, Xv = 3 :Q8 and Γ = K10. ∼{ } ∼ 2 (2) G = X = PSL3(3).2, X u,v = GL2(3):2, Xv = 3 :GL2(3), and Γ is the point-line non-incidence graph of {PG(2} , 3).

Γ (v) Proof. Case 1. Assume that Lemma 3.5 (a1) holds. Suppose that (Xv )u =Q8. Then 7 4 Xuv . Q8×Q8. This implies that |X u,v | is a divisor of 2 and divisible by 2 . Checking ∼{ } ∼ the Tables 14-20 in [17], we have X = PSL2(9).2 = M10 and X u,v = Z8:Z2; in this case, ∼ 2 { } Xv = 3 :Q8, and d = 9. Then part (1) of this lemma follows. 18 HAN, LIAO, AND LU

Γ (v) Γ (v) Γ (v) Suppose that (Xv )u =6 Q8. If p = 3 and (Gv )u = Q8, then (Xv )u is abelian, Γ (v) it follows that Xuv is abelian, a contradiction. Thus we have SL2(3) ✂ (Gv )u ≤ Γ (v) Z GL2(p), and p ∈ {3, 5, 7, 11, 23}. Then (Gv )u ≤ NGL2(p)(SL2(3)) = p 1 ◦ GL2(3). Γ (v) Γ (v) Γ (v) − Since (Xv )u is nonabelian and normal in (Gv )u, we have Q8 ✂ (Xv )u, and hence Γ (v) 7 2 2 4 SL2(3) ✂ (Xv )u. Moreover, |X u,v | is a divisor of 2 3 (p − 1) and divisible by 2 . Let { } M be an arbitrary normal abelian subgroup of X u,v . Then M ∩ Xuv has index at most [1] [1] { } Γ (v) 2 in M, and (M ∩ Xuv)Xv /Xv is isomorphic to a normal subgroup of (Xv )u. Thus [1] [1] [1] [1] [1] (M ∩ Xuv)Xv /Xv . Zp 1. Since M ∩ Xv ✂ Xv and Xv is isomorphic to a normal Γ (v) − [1] [1] [1] ∼ subgroup of (Xv )u, we have M ∩ Xv . Zp 1. Noting that (M ∩ Xuv)Xv /Xv = − [1] 2 M ∩ Xuv/(M ∩ Xv ), it follows that |M ∩ Xuv| is a divisor of (p − 1) . 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:

7 2 2 4 (c1) |H0| is a divisor of 2 3 (p − 1) and divisible by 2 ; H0 has a factor (a quotient 2 2 2 of some subnormal subgroup) Q8; and H0 has no element of order 3 , 5 or 11 . 2 (c2) If M is a normal abelian subgroup of H0 then |M| is a divisor of 2(p − 1) ; if (p 1)2 p ∈{7, 11, 23}, the order of Ø p−1 (H0) is a divisor of − . 2 4

Checking the those H0 which satisfy conditions (c1) and (c2), we conclude that the possible pairs (X,X u,v ) are listed as follows: { } 2 5 2 6 (M11, 3 :Q8.2), (M11, 2S4), (M12, [2 ].S3), (M12, 3 :2S4), (J2, [2 ]:(3×S3)), 6 9 2 8 2 6 2 (J3, [2 ]:(3×S3)), (Co3, [2 ].3 .S3)), (He.2, [2 ]:3 .D8)), (McL.2, [2 ]:S3), 2 2+4 (PSL3(3), 3 :2S4), (PSL3(3).2, 2S4:2), (PSL3(4).2, 2 .3.2), 6 2 8 2 (PGL3(4).2, [2 ].3.S3), (PSL4(3).2, 2.S4.2), (PSL5(2).2, [2 ].S3.2), 8 2 5 7 2 (PSp4(4).4, [2 ]:3.12), (PSp4(4).4, 5 :[2 ]), (PSp6(2), [2 ]:S3), 8 3 2 2 (PSp6(3), [2 ]:3 .S3), (PSU3(3), 4.S4), (PSU4(2), 2.A4.2), (PSU4(3), 2.A4.4), 5 + 2 (PSU4(3).2, [2 ].S4), (PΩ8 (3).A4, 10 :4A4), 2 2 (G2(2)′, 4.S4), (G2(3), SL2(3) ◦ SL2(3):2), ( F4(2)′, 5 :4A4).

Note these groups X are included in the Atlas [3]. Inspecting the subgroups of X, only ∼ 2 the pair (PSL3(3).2, 2S4:2) gives a desired Xv = 3 :GL2(3). Then part (2) of this lemma follows. 1+4 Z Γ (v) 1+4 Z Z 1+4 Γ (v) Case 2. Let2+ : 5 ≤ (Gv )u ≤ 2+ .( 5: 4). Then 2+ ✂(Xv )u, and so, |X u,v | is a divisor of 21552 and divisible by 26. Further, if M is a normal abelian subgroup{ of} 5 X u,v then a similar argument as in Case 1 yields that |M| is a divisor of 2 . It is { } easily shown that Ø2(Xuv) =6 1, and hence Ø2(X u,v ) =6 1. Checking the pairs (G0,H0) { } in [17, Tables 14-20], either Ø2(H0)=1or |H0| has an odd prime divisor other than 5. Thus, in this case, no desired pair (X,X u,v ) exists. { } Γ (v) f Γ (v) We assume next that Lemma 3.5 (a3) occurs. Thus (Gv )u 6≤ GL1(p )and(Gv )u ≤ f Γ (v) Γ (u) Γ (v) ΓL1(p ). Then f > 1 and (Gv )u . Zpf 1:Zf . Recalling Xuv . (Xu )v×(Xv )u ≤ Γ (u) Γ (v) − (Gu )v×(Gv )u, we have the following simple fact. EDGE-PRIMITIVITY 19

Z3 Z5 Z6 Lemma 5.4. If (a3) of Lemma 3.5 occurs then X u,v has no section t , r or 2, 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 6.

f 6 6 Proof. Suppose that p =2 . Then X has order divisible by 2 , Xuv . Z63:Z6×Z63:Z6, 4 and thus X u,v has a normal Hall 2′-subgroup and |X u,v | is indivisible by 2 . Checking { } { } Tables 14-20 given in [17], (X,X u,v ) is one of the following pairs: { } Z Z 1+2 6 3 (S7, 7: 6), (M12.2, 3+ :D8), (PSL2(2 ), D126), (PSL2(5 ), D126), 2 6 (PSL2(7937), D7938), (PSL3(8), 7 :S3), (Sz(8), D14), (G2(3).2, [3 ]:D8). 6 ∼ 6 The pair (PSL2(2 ), D126) yields that Xv = 2 :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, ∼ ∼ 2 ∼ X u,v = Z8:Z2, Xv = 3 :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: { }

(i) (PGL2(7), D16), (PSL3(2).2, D16), (PGL2(9), D16), (M10, Z8:Z2); (A5, D10), 2 2 (A6, 3 :Z4), (M11, 3 :Q8.2), (J1, D6×D10), (PGL2(7), D12), (PGL2(9), D20), Z Z a a Z Z (M10, 5: 4), (PGL2(11), D20), (PSL2(t ), D 2(t ±1) ); (PSp4(4).4, 17: 16); (2,t−1) a Za Z a a (ii) (PSL2(t ), t : t −1 ), t is a prime a ≤ 4 and t − 1 a power of 2; 2 Z2 mod (PSL3(t), 3:Q8), t is a prime with t ≡ 4, 7( 9); Z2 mod (PSU3(t), 3:Q8), t is a prime with t ≡ 2, 5( 9); a Z2 (PSU3(2 ), 3:Q8) with prime a> 3; a Z 2 Z a Z Z Z b (PSp4(2 ). 2b+1 , D2(q 1):2. 2b+1 ), (PSp4(2 ). 2b+1 , 22a+1.4. 2b+1 ), where 2 k a; 2a+1 ± 2a+1 (Sz(2 ), D2(22a+1 1)), (Sz(2 ), Z22a+1 2a+1+1:Z4); 3 a − ± ( D4(t ), Zt4a t2a+1:Z4), t is a prime. − ∼ 2 The pair (M10, Z8:Z2) yields that Xv = 3 :Q8 and d = 9. The third pair in (i) implies ∼ Z2 Z a a that Xv = 3: 8; however, Xuv is abelian, which is not the case. For (PSL2(t ), D 2(t ±1) ), (2,ta−1) a a f ∼ Zf Z checking the subgroups of PSL2(t ), we have t = p and Xv = p : pf −1 , and then Xuv (2,p−1) 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 Ør(X u,v ). Then Ør(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 f { } |Xv:Xuv| = d = p 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: 20 HAN, LIAO, AND LU

2a+1 ∼ Z2a+1 Z (iii) X = Sz(2 ) and Xv = 2 : 22a+1 1; a Z 2a− Z (iv) X = PSp4(2 ). 2b+1 and Xv . Sp2(2 ):2. 2b ; a Z a Z (v) X = PSp4(2 ). 2b+1 and Xv . Sp2(2 ) ≀ S2. 2b .

Item (iii) yields that Xuv is abelian, which is not the case. Item (iv) gives Xuv = Xv, ∼ Za Z 2 Z a contradiction. Suppose that (v) occurs, we have Xv = ( 2: 2a 1) :2. 2b . Then 1 =6 − Ø2(Xv) ≤ Ø2(Gv), and hence d = |Ø2(Gv)| by Lemma 2.5. Since Xv is transitive on Γ (v), f 2a a 2 b+1 it follows that p = d =2 . Thus |Xuv| = (2 − 1) 2 , 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 ∼ ∼ 2 ∼ then G = X or X.2, X = M10, X u,v = Z8:Z2, Xv = 3 :Q8 and Γ = K10. { } Γ (v) Proof. Let f = 2. Then (Xv )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. − −

f f Let π0(p − 1) be the set of primitive primes of p − 1. By Zsigmondy’s theorem, if f f 6 t π0(p − 1) = ∅ and f > 1 then p = 2 , or f = 2 and p = 2 − 1, where t is a prime. f Thus, in view of Lemma 5.5 and Corollary 5.7, we assume next that π0(p − 1) =6 ∅. f Lemma 5.8. Assume that π: = π0(p − 1) =6 ∅, Xuv is nonabelian and (a3) of Lemma 3.5 occurs. Then f ≥ 3, and

(1) π =6 π(|X u,v |) \{2}, min(π) ≥ max{5, f+1}; { } (2) p 6≡ ±1(mod r) and Ør(X u,v ) =16 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.

Γ (v) Proof. By the assumption and Lemma 3.6, we have (Xv )u =∼ Zm′ .Z f , and ∅= 6 π = e′ f r 1 π0(p − 1) ⊆ π(m′). For r ∈ π, since p − ≡ 1(mod r), we have f ≤ r − 1, and so r ≥ f+1. In particular, r ≥ 5 and p 6≡ ±1(mod r). Recall that X u,v = Xuv.2 and { } Xuv . Zm′ .Z f ×Zm′ .Z f . It follows that Ør(X u,v ) =6 1, and Ør(X u,v ) is the unique e′ e′ { } { } Sylow r-subgroup of X u,v . Clearly, Ør(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 =6 1 and, by Lemma 5.6, X u,v has { } no normal abelian Hall 2′-subgroup. Then π =6 π(|X u,v |) \{2}, the lemma follows. { } Z6 Z5 Recall that X u,v has no section 2 or 3, see Lemma 5.4. Combining with Lemma { } 5.8, we next check the pairs (G0,H0) listed in [17, Tables 14-20]. f Lemma 5.9. Assume that π0(p − 1) =6 ∅, Xuv is nonabelian and (a3) of Lemma 3.5 occurs. Then T = soc(X) is not a simple group of Lie type.

a Proof. Suppose that T is a simple group of Lie type over a ﬁnite ﬁeld of order q′ = t , where t is a prime. Since T ✂ G, we know that T is transitive on the edge set of Γ (v) Γ (v) Γ (v) Γ (v) Γ (v) Γ . Then Tv =6 1. Noting that Tv ✂ Gv , we have soc(Gv ) ≤ Tv . In f particular, Tv is transitive on Γ (v), and so |Tv| = p |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 ). { } { } EDGE-PRIMITIVITY 21

By Lemmas 5.6 and 5.8, X u,v is not a {2, 3}-group and has no normal abelian Hall { } f 2′-subgroup. Assume that t ∈ π0(p − 1). By Lemmas 5.4 and 5.8, t ≥ 5, X u,v has Z3 { } no section t and Øt(X u,v ) =6 1 is abelian. Checking the pairs (G0,H0) listed in [17, { } 2 ∼ Z2 Z Tables 16-20], we have X = PSL2(t ) and X u,v = t : t2−1 . For this case, checking the { } 2 2 f subgroups of PSL2(t ), no desired Xv arises, a contradiction. Therefore, t 6∈ π0(p − 1). f By Lemma 5.8, Ør(X u,v ) =6 1 for each r ∈ π0(p − 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 Øt(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 6= Z30, by Lemma 3.6, f Z3 f π0(p − 1) ⊆ π(|N|), and thus, by Lemma 5.4, N has no subgroup r for r ∈ π0(p − 1). With these restrictions, only one of the following (i)-(iv) occurs. ∼ 1 Z2 (i) Either X = PSL3(q′) and X u,v = ′ q′ 1.S3 with q′ =26 , 4, or X = PSU3(q′) { } (3,q 1) ∼ 1 Z2 − 3 − f 2 and X u,v = ′ q′+1.S3. Then |Xv| = ′ p (q′∓1) . Checking Tables 8.3-8.6 { } (3,q +1) (3,q 1) 3 ∓ 1 2 Z ′ given in [1], we have X = PSL3(q′) and Xv . [q′ ]: (3,q′ 1) q 1. It follows that p = t = 3, f+1 − − and |Ø3(Xv)| =3 =3d, which contradicts Lemma 2.5. + ∼ 2 2 (ii) T = soc(X) = PΩ8 (q′) and T u,v = D 2(q′2+1) .[2 ]. In this case, noting that { } (2,q′−1) ′ ′ 4 f (q 2+1)2 3 f (q 2+1)2 |T u,v :Tuv| ≤ 2, we have |Tv| = 2 p ′ 2 or 2 p ′ 2 . Let M be a maximal { } (2,q 1) (2,q 1) − − 2 2 subgroup of T with Tv ≤ M. By [14], since |M| is divisible by (q′ +1) , we have ∼ 2 2 2 2 2 2 M = PSL2(q′ ) .2 . It is easily shown that PSL2(q′ ) .2 does not have subgroups of ′ ′ 4 f (q 2+1)2 3 f (q 2+1)2 order 2 p (2,q′ 1)2 or 2 p (2,q′ 1)2 , a contradiction. − − 2 2 2 (iii) (X,X u,v ) is one of ( F4(2)′, 5 :4A4) and ( F4(2), 13:12). For the ﬁrst pair, we f { } f 2 f 4 have π0(p − 1) = {5} and, since p is a divisor of | F4(2)′|, we conclude that p = 2 4 f f 12 3 or 3 . The second pair implies that π0(p − 1) = {13}, and then p =2 or 3 . 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: { }

T N |T u,v :N| Remark a { } Ree(3 ) Z a+1 6 odd a ≥ 3 3a 3 2 +1 ± Z2×Z 3a+1 6 & X = T 2 a Z2 G2(3 ) 3a 1 12 odd a ≥ 2 ± Z32a 3a+1 6 Z2 ± 2a+1 48 odd a ≥ 3 2 a Z2 F4(2 ) a+1 96 & X = T 2a 2 2 +1 ± a a+1 Z 3a+1 a+1 12 &2 ±2 2 +1 > 5 22a 2 2 +2a 2 2 +1 Z2 ± ± 22a 2a+1 72 a Z2 ± F4(2 ) 22a+1 96 a ≥ 2 Z24a 22a+1 12 Z2 − E8(q′) q′4 q′2+1 288 X = T − Zq′8 q′7 q′5 q′4 q′3 q′+1 30 ± ∓ − ∓ ± 22 HAN, LIAO, AND LU

f Let M be a maximal subgroup of T with Tv ≤ M. Then |M| is divisible by p |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 f 2 a no maximal subgroups of order divisible by p |N|. Similarly, by [21], the group F4(2 ) is excluded. We next deal with the remaining cases. a 2a a (1) Let T = G2(3 ). Suppose that |N| = 3 ±3 +1. By [15, Theorems A and B], 2a a ∼ a a since |M| is divisible by 3 ±3 +1, we have M = SL3(3 ):2 or SU3(3 ):2. By [1, Tables 8.3-8.6], we conclude that Tv . Z32a 3a+1:[6], which is impossible. ± a 2 a a a Similarly, for |N| = (3 ±1) , we have Tv . (SL2(3 ) ◦ SL2(3 )).2, SL3(3 ):2 or a 1 f f a 2 SU3(3 ):2. Since |Tv| is divisible by 2 |T u,v |p = 6p (3 ±1) , checking the maximal a a {a } ba Z2 subgroups of SL2(3 ), SL3(3 ) and SU3(3 ), we have p = 3 and Tv . [3 ]: 3a 1.2 for − b = 2 or 3. Since Tuv has order divisible by 3, it follows that Ø3(Tuv) =6 1, which contradicts Lemma 2.5. a f (3) Let T = F4(2 ). By [18, 19], noting that |M| is divisible by p |N|, we conclude ∼ a + a 2a 2 ∼ a 2 that M = Sp8(2 ) or PΩ8 (2 ).S3 with |N| = (2 +1) , or M = c.PSL3(2 ) .c.2 or a 2 2a a 2 a c.PSU3(2 ) .c.2 with |N| = (2 ±2 +1) , where c = (3, 2 ± 1). Then a contradic- a + a a tion follows from checking the maximal subgroups of Sp8(2 ), PΩ8 (2 ), PSL3(2 ) and a PSU3(2 ), refer to [1, Tables 8.3-8.6, 8.48-8.50]. 4 2 2 ∼ 2 2 (4) Let T = E8(q′). Then |N| =(q′ − q′ +1) and M = PSU3(q′ ) .8. For this case, 2 checking the maximal subgroups PSU3(q′ ), we get a contradiction.

f Lemma 5.10. Assume that π0(p − 1) =6 ∅, Xuv is nonabelian and (a3) of Lemma 3.5 ∼ Z Z ∼ Z3 Z Z occurs. Then G = X = J1, X u,v = 7: 6, Xv = 2: 7: 3 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:Z r−1 for r 6∈ {7, 11, 17, 23}, or X = Sr and X u,v = Zr:Zr 1 for { } 2 { } − 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. − f Assume that T is a sporadic simple group, and let r ∈ π0(p − 1). Then (X,X u,v ,r) is one of the following triples: { }

Z Z Z Z Z Z Z2 (J1, 7: 6, 7), (J1, 11: 10, 11), (J1, 19: 6, 19), (J2, 5:D12, 5), (J3.2, Z19:Z18, 19), (J4, Z29:Z28, 29), (J4, Z37:Z12, 37), (J4, Z43:Z14, 43), Z Z Z2 Z2 (O′N.2, 31: 30, 31), (He, 5:4A4, 5), (Co1, 7:(3×2A4), 7), Z Z Z Z Z Z (Ly, 37: 18, 37), (Ly, 67: 22, 67), (Fi24′ , 29: 14, 29), (B, Z13:Z12×S4, 13), (B, Z19:Z18×Z2, 19), (B, Z23:Z11×2, 23), (M, Z23:Z11×S4, 23), (M, (Z29:Z14×3).2, 29), (M, Z31:Z15×S3, 31), (M, Z41:Z40, 41), (M, Z47:Z23×2, 47). EDGE-PRIMITIVITY 23

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:

X J1 J2 J4 Co1 O′N.2 He B 2 2 3 2 2 4 2 5 4 |X u,v | 2·3·7 2 ·3·5 2·7·43 2 ·3 ·7 2·3·5·31 2 ·3·5 2 ·3 ·13 r { } 7 5 43 7 31 5 13 pf 23 24 214 23, 36 25 24 33, 54, 212 f p − 1 |Guv| X X × X, × X X X, X, × X B B M M M M M 2 2 3 2 2 3 |X u,v | 2 ·3 ·19 2·11·23 2 ·3·11·23 2 ·3·7·29 2·3 ·5·31 2 ·5·41 2·23·47 r { } 19 23 23 29 31 41 47 pf 218 211, 311 211, 311 228 25, 53 220, 38 223 f p − 1 |Guv| × ×, × ×, × × X, × ×, × ×

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 f f (p − 1) is a divisor of |Guv| = |Xuv||G:X|. It follows that (X,X u,v ,r,p ) is one of { } Z Z 3 Z2 4 Z2 3 (J1, 7: 6, 7, 2 ), (J2, 5:D12, 5, 2 ), (Co1, 7:(3×2A4), 7, 2 ), Z Z 5 Z2 4 Z Z 3 (O′N.2, 31: 30, 31, 2 ), (He, 5:4A4, 5, 2 ), (B, 13: 12×S4, 13, 3 ), 4 5 (B, Z13:Z12×S4, 13, 5 ), (M, Z31:Z15×S3, 31, 2 ). Z2 3 3 3 For (Co1, 7:(3×2A4), 7, 2 ), we have Xuv . ΓL1(2 )×ΓL1(2 ), yielding that Xuv has 3 odd order, a contradiction. Similarly, for (B, Z13:Z12×S4, 13, 3 ), the order of Xuv is 4 5 indivisible by 2 , a contradiction; for (M, Z31:Z15×S3, 31, 2 ), the order of Xuv is indi- Z2 4 visible by 3, also a contradiction. For (He, 5:4A4, 5, 2 ), the order of Xuv is divisible 3 2 f 4 7 2 by 2 · 3 · 5 and, since p = 2 , the order of Xu is divisible by 2 · 3 · 5 ; however, He has no soluble subgroups of order divisible by 27 · 3 · 52, a contradiction. Similarly, the 5 quadruple (O′N.2, Z31:Z30, 31, 2 ) is excluded as O′N.2 has no soluble subgroups with 5 order divisible by 2 · 31. (Note that Gv is soluble.) By the Altas [3], J2 has no sub- 4 2 Z2 4 groups with order divisible by 2 · 5 , and then (J2, 5:D12, 5, 2 ) is excluded. By the Altas [3] and [33, Theorem 2.1], B has no subgroups with order divisible by 32 · 54 · 13, 4 3 and then (B, Z13:Z12×S4, 13, 5 ) is excluded. Then (J1, Z7:Z6, 7, 2 ) is left, which gives ∼ Z3 Z Z f Xv = 2: 7: 3, d = p = 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. { } Γ (v) Now let G and Γ = (V, E) satisfy the assumptions in Theorem 1.1. If Gv is an almost simple 2-transitive group then, by Lemmas 3.3, 4.1-4.3, the triple (G,Gv,G u,v ) { } Γ (v) f is listed in Table 1. Assume that Gv is a soluble 2-transitive group of degree d = p , Γ (v) f Γ (v) where p is a prime. Then either Gv ≤ ΓL1(p ), or Gv has a normal subgroup SL2(3) 1+4 or 2+ . For the latter case, the triple (G,Gv,G u,v ) is known by Lemmas 5.2 and 5.3. { } Γ (v) f f f Let Gv ≤ ΓL1(p ) and consider the primitive prime divisors of p − 1. If p − 1 has no primitive prime divisor then, by Lemmas 5.2, 5.5 and Corollary 5.7, (G,Gv,G u,v ) { } 24 HAN, LIAO, AND LU

f is listed in Table 1. If p − 1 has primitive prime divisors, then (G,Gv,G u,v ) is known by Lemmas 5.2 and 5.10. This completes the proof. { }

References [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] P.J. Cameron, Permutation Groups, Cambridge University Press, Cambridge, 1999. [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] J.D. Dixon and B. Mortimer, Permutation Groups, Springer-Verlag, New York, 1996. [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] D.A. Foulser, The flag-transitive collineation groups of the Desarguesian affine planes, Canad. J. Math. 16 (1964), 443-472. [7] A. Gardiner, Arc transitivity in graphs, Quart. J. Math. Oxford 24(1973), 399-407. [8] A. Gardiner, Doubly primitive vertex stabilizers in graphs, Math. Z. 135 (1974), 157-166. [9] M. Giudici and C.H. Li, On finite edge-primitive and edge-quasiprimitive graphs, J. Combin. Theory Ser. B 100 (2010), 275-298. [10] S.T. Guo, Y.Q. Feng, C.H. Li, The finite edge-primitive pentavalent graphs, J. Algebraic Combin. 38 (2013) 491-497. [11] S.T. Guo, Y.Q. Feng, C.H. Li, Edge-primitive tetravalent graphs, J. Combin. Theory Ser. B 112 (2015) 124-137. [12] B. Huppert, Endliche Gruppen I, Springer-Verlag, Berlin, 1967. 3 [13] P.B. Kleidman, The maximal subgroups of the Steinberg triality groups D4(q) and of their automorphism groups, J. Algebra 115 (1988), 182-199. + [14] P.B. Kleidman, The maximal subgroups of the finite 8-dimensional orthogonal group PΩ8 (q) and of their automorphism groups, J. Algebra 110 (1987), 173-242. [15] P.B. Kleidman, The maximal subgroups of the Chevalley groups G2(q) and with q odd, the Ree 2 group G2(q), and their automorphism groups, J. Algebra 117 (1988), 30-71. [16] P.B. Kleidman and M.W. Liebeck, The Subgroup Structure of the Finite Classical Groups, Cam- bridge University Press, 1990. [17] C.H. Li and H. Zhang, The finite primitive groups with soluble stabilizers, and the edge-primitive s-arc transitive graphs, Proc. Lond. Math. Soc. 103 (2011) 441-472. [18] M.W. Liebeck, J. Saxl and G.M. Seitz, Subgroups of maximal rank in finite exceptional groups of Lie type, Proc. London Math. Soc. 65 (1992), 297-325. [19] M.W.Liebeck and G.M. Seitz, A survey of maximal subgroups of exceptional groups of Lie type, Groups, Combinatorics and Geometry, pp. 139-146 (2003). [20] Z.P. Lu, On edge-primitive 2-arc-transitive graphs, J. Combin. Theory Ser. A 171 (2020), 105172, https://doi.org/10.1016/j.jcta.2019.105172. 2 2 [21] G. Malle, The maximal subgroups of F4(q ), J. Algebra 139 (1991), 52-69. [22] S. P. Norton and R.A Wilson, A correction to the 41-structure of the Monster, a construction of a new maximal subgroupL 2(41) and a new Moonshine phenomenon, J. Lond. Math. Soc. (2) 87 (2013), 943-962. [23] J.M. Pan, C.X. Wu and F.G. Yin, Finite edge-primitive graphs of prime valency, European J Combin. 73 (2018), 61-71. [24] C.E. Praeger, An O’Nan-Scott theorem for finite quasiprimitive permutation groups and an appli- cation to 2-arc transitive graphs, J. London Math. Soc. 47 (1992), 227-239. [25] C.E. Praeger, On a reduction theorem for finite, bipartite 2-arc-transitive graphs, Austral. J. Combin. 7 (1993), 21-36. [26] M. Suzuki, On a class of doubly transitive groups, Ann. Math. 75 (1962), 105-145. EDGE-PRIMITIVITY 25

[27] The GAP Group, GAP-Groups, Algorithms, and Programming, Version 4.8.6, 2017. http://www.gap-system.org [28] V.I. Troﬁmov, Vertex stabilizers of locally projective groups of automorphisms of graphs: a sum- mary, In: Groups, Combinatorics and Geometry, pp.313-326, World Sci. Publ. Co., Durham, 2001. [29] R. Weiss, Kantenprimitive Graphen vom Grad drei, J. Combin. Theory Ser. B 15 (1973), 269-288. [30] R. Weiss, Symmetric graphs with projective subconstituents, Pro. Amer. Math. Soc. 72 (1978), 213-217. [31] R. Weiss, s-transitive graphs, Algebraic methods in graph theory, Colloq. Soc. Janos Bolyai 25(1981), 827-847. [32] R. Weiss, The nonexistence of 8-transitive graphs, Combinatorica 1 (1981), 309-311. [33] R.A. Wilson, The maximal subgroups of the Baby Monster, I, J. Algebra 211 (1999), 1-14. [34] R.A. Wilson, The Finite Simple Groups, Springer-Verlag, 2009. [35] K. Zsigmondy, Zur Theorie der Potenzreste, Monatsch. Math. Phys. 3 (1892), 265-284.

H. Han, School of Science, Tianjin University of Technology, Tianjin 300384, P. R. China Email address: [email protected]

H. C. Liao, Center for Combinatorics, LPMC, Nankai University, Tianjin 300071, P. R. China Email address: [email protected]

Zaiping Lu, Center for Combinatorics, LPMC, Nankai University, Tianjin 300071, China Email address: [email protected]