Linear groups of alternating type containing non-scalar elements with all but one eigenvalues of multiplicity 1 1 Alexandre Zalesski Abstract Our main goal is to determine the irreducible representations of alternating and symmetric groups and their universal central extensions that contain a non-scalar element with all but one eigenvalues of multiplicity 1. 1 1 1 2000 Mathematics subject classification: 20B15, 20H30 2 2 2 keywords: alternating groups, irreducible representations, almost cyclic elements 1 Introduction This paper completes a long-standing project of describing finite primitive irreducible linear groups G containing an element g of prime power order whose all but one eigenvalues have multiplicity 1. The ground ring is assumed to be a field, usually algebraically closed, and of arbitrary characteristic \ell coprime to |g| . Earlier contributions to this problem are in works [ 5 , 6 , 7 , 8 , 20 , 21 , 28 , 29 ] ; more detailed will be given below. An element with this property is called almost cyclic . If all eigenvalues are of multiplicity 1 then g is called cyclic , the term goes back to module theory terminology: a module with one generator is called cyclic. It is known that a module over a cyclic group \langle g\rangle is cyclic if and only if the minimal polynomial of g coincides with the characteristic polynomial, and if g is diagonalizable then this is equaivalent to saying that all eigenvalues of g are of multiplicity 1. The bulk of the project is the case where G is a quasi-simple group, or more precisely those whose derived group is quasi-simple. (A quasi-simple group G is, by definition, a finite group such that G=G^{\prime} and G^{\prime}/Z(G^{\prime}) is a non-abelian simple group.) The case where the simple factor of G is an alternating group is the only one that was not considered yet. The main goal of this paper is to prove the following theorem. The permutational representation of A_{n} of degree n is called natural , and any non-trivial composition factor of it is called subnatural . (If n\neq 6 then all permutational representations of A_{n} of degree n are equivalent, whereas those of A_{6} partition in two equivalence classes; we call natural the one sending a 3-cycle to a 3-cycle). For g\in G the order of g modulo Z(G) is denoted by o(g) . Theorem 1.1 . Let G=c.A_{n} , n\geq 5 , be the universal central extension of A_{n} , and \phi:G\rightarrow GL_{m}(F) be a non-trivial irreducible representation of G. Suppose that F is algebraically closed of characteristic \ell\neq p and \phi(g) is almost cyclic for some non-central p-element of G. Then m\leq n+1 , and either \phi(Z(G))=\mathop{\rm Id}\nolimits and \phi is subnatural, or one of the cases in Tables 1.1 and 2.1 holds, in the latter case m\leq 8 and o(g)\leq 9 . In fact we settle a slightly more general case of groups G where G^{\prime}/Z(G^{\prime})\cong A_{n} , see Lemmas 4.2 , 4.3 and 3.8 below. The assumption that g\in G is of prime power order is not strictly necessary but obtaining a similar result for arbitrary elements of G requires a significally larger work. Elements of prime power, in particular, of prime order are of interest for some applications, see [ 12 , 1 , 11 , 18 ] . In [ 12 ] the authors stated the problem of describing the irreducible subgroups G of GL(V) (for V being a vector space of finite dimension over a finite field) that contain elements g\in G of prime order p\not|\,q such that g acts irreducibly on V/C_{V}(g) (equivalently, on (\mathop{\rm Id}\nolimits-g)V ). They obtained a full description of such groups assuming that \dim C_{V}(g) <\dim V/2 . Such element g\in G becomes almost cyclic under a suitable field extension. The results of [ 12 ] were refined in [ 1 ] and further extended in [ 9 ] , where for a p -element g the assumption was relaxed to |g|>\dim V/3 . In [ 11 ] some more details on the groups listed in [ 9 ] is given, with focus on the case where \dim C_{V}(g)=\dim V/2 and identifying the minimal groups containing elements in question. Our results together with [ 5 , 6 , 7 , 8 , 21 , 29 ] potentially allows to obtain similar results for subgroups of GL(V) with no restriction to \dim C_{V}(g) . Note that the case of characterisitic 0 ground field is not excluded. Observe that the irreducible linear quasi-simple finite groups over the complex numbers containing a cyclic element have been determined in [ 16 , Theorem 3.1.5] and a special case of alternating groups has been earlier settled in [ 15 , Theorem 6.2] . In linear group theory a significant role is played by finite irreducible groups generated by reflections and quasi-reflections. A quasi-reflection is a matrix g\in GL_{n}(F) such that g-\alpha.\mathop{\rm Id}\nolimits is of rank 1 for some \alpha\in F , if \alpha=1 then g is called a reflection. These were classified nearly 50 years ago, see [ 27 , 25 ] . Note that a quasi-reflection is a special case of an almost cyclic matrix. In general, almost cyclic matrices can be defined as follows [ 6 , Definition 2.1] : Definition 1.2 . Let M be an (m\times m) -matrix over a field F . We say that M is almost cyclic if there exists \alpha\in F such that M is similar to \mathop{\rm diag}\nolimits(\alpha\cdot\mathop{\rm Id}\nolimits_{k},M_{1}) where M_{1} is cyclic and 0\leq k4 is a central extension of A_{n} with center of order c in which every proper normal subgroup lies in its center. GL_{n}(F) is the group of non-singular (n\times n) -matrices over F , and we often write GL_{n}(F)=GL(V) in order to say that V is the underlying vector space for GL_{n}(F) . If g\in GL_{n}(F)=GL(V) then \deg g denotes the degree of the minimal polynomial of g . We say that g is fixed point free if 1 is not an eigenvalue of g . The eigenvalue 1 eigenspace of g is often introduced as C_{V}(g) . The notion of a cyclic and almost cyclic matrix is defined in the introduction. We write \mathop{\rm diag}\nolimits(g_{1},\ldots,g_{k}) for a block-diagonal matrix with diagonal blocks g_{1},\ldots,g_{k} (not necessarily of the same size). Let \rho:G\rightarrow GL_{n}(F) be a representation. If H is a subgroup of G then the restriction of \rho to H is denoted by \rho|_{H} . 2 Preliminaries If an almost cyclic matrix M acts in a vector space W and U is an M -stable subspace of W then the restrictions of M to U and to W/U are almost cyclic matrices. Lemma 2.1 . Let g\in G\subset GL_{n}(F),n>1 be a finite irreducible subgroup generated by k conjugates of g . Suppose that g is almost cyclic, and d=\dim(g-\lambda\cdot\mathop{\rm Id}\nolimits)V for some \lambda\in F . Then n\leq dk and n\leq(o(g)-1)k . Proof. Let g_{1},\ldots,g_{k} be conjugates of g . Then G=\langle g_{1},\ldots,g_{k}\rangle stabilizes the subspace W:= (g_{1}-\lambda\cdot\mathop{\rm Id}\nolimits)V+(g_{2}-\lambda\cdot\mathop{\rm Id}\nolimits)V+\cdots+(g_{k}-\lambda\cdot\mathop{\rm Id}\nolimits)V . As G is irreducible, we have W=V . So the first inequality follows. For the second see [ 8 , Lemma 2.11] . ∎ Lemma 2.2 . [ 28 , Lemma 3.2] Let M=M_{1}\otimes M_{2} be a Kronecker product of diagonal non-scalar matrices M_{1},M_{2} of sizes m\leq n , respectively. (1) The eigenvalue multiplicities of M do not exceed mt , where t is the maximal eigenvalue multiplicity of M_{2} . (2) If M_{1},M_{2} are cyclic then the eigenvalue multiplicities of M do not exceed m . (3) Suppose that M is almost cyclic. Then M_{1} and M_{2} are cyclic. (4) Suppose that M is almost cyclic and M_{i} is similar to M_{i}^{-1} for i=1,2 . Then the eigenvalue multiplicities of M do not exceed 2 . In addition, if e is an eigenvalue of M of multiplicity 2 then e\in\{\pm 1\} . 3 Representations of A_{n} of small degrees Let G denote A_{n} or S_{n} , n>4 . We say that an irreducible representation \phi of S_{n} is subnatural if the irreducible constituents of \phi|_{A_{n}} are subnatural representations of A_{n} . This extends to S_{n} the notion of a subnatural representation of A_{n} given in the introduction. In fact, any subnatural representation of S_{n} is irreducible on A_{n} . In this section we denote by P_{n} the permutation FS_{n} -module of dimension n arising from a natural representation of S_{n} . It is well known that P_{n} has a unique non-trivial composition factor, and one or two trivial factors depending of whether \ell\nmid n or \ell|n , respectively. We denote the non-trivial factor by W_{n} , and denote by W_{n}^{\prime} the restriction of W_{n} to A_{n} . Then W_{n}^{\prime} remains irreducible. If \ell\neq 2 then we set P_{n}^{-}=P_{n}\otimes\alpha_{n} , where \alpha_{n} is a non-trivial one-dimensional F -representation of S_{n} , and W_{n}^{-}=W_{n}\otimes\alpha . We call W_{n},W_{n}^{-} and W_{n}^{\prime} standard FG -modules (some authors call them deleted). The representations of G afforded by standard FG -modules are subnatural. It is well known that \dim W_{n}=\dim W_{n}^{\prime}=n-1 if \ell\nmid n , and n-2 otherwise. If n>8 then the converse holds, in the sense that every FG -module of dimension d with 17 and let \phi be a non-trivial irreducible representation of A_{n} such that all composition factors of \phi|_{A_{n-1}} are either trivial or subnatural representations of A_{n-1} . Then \phi is a subnatural representations of A_{n} . Using induction on n with n=8 the induction base, we have: Corollary 3.2 . Let 66 , isomorphic to A_{n-1} are conjugate. (By [ 10 , §5.2] A_{n-1} is intransitive, and, by an order reason, a subgroup X\cong A_{n-1}\subset A_{n} fixes a point and even coincides with the stabilizer of a point; as A_{n} is transitive, the point stabilizers are conjugate in A_{n} .) Lemma 3.3 . (1) The alternating group A_{n} with n odd is generated by two cycles of order n . (2) For n>6 even the group A_{n} is generated by two elements of order n/2 of double cycle structure, and S_{n} is generated by two cycles of order n . Proof. (1) are well known. For a proof, observe that every g\in A_{n} is a product of two n -cycles, see for instance [ 26 ] . For n=3 the statement is trivial, for n=5 this follows by inspection of the subgroups of A_{5} in [ 3 ] . Let n>5 . By the Bertrand postulate, if n-2>3 then there exists a prime p>2 such that (n-2)/2w_{2}>...>w_{s} . Also let F be any field with characteristic \neq 2 . Then: (1) the degree of any faithful representation of 2.S_{n} over F is divisible by 2^{\lfloor(n-s)/2\rfloor}; (2) the degree of any faithful representation of 2.A_{n} over F is divisible by 2^{\lfloor(n-s-1)/2\rfloor}. Lemma 3.6 . In notation of Theorem 3.5 we have 2^{\lfloor(n-s-1)/2\rfloor}>2(n-1) for n>13 . Proof. Suppose the contrary, that 2^{\lfloor(n-s-1)/2\rfloor}\leq 2(n-1) . Then 2^{(n-s-3)/2}\leq 2(n-1) and 2^{n-s-3}\leq 4(n-1)^{2} , 2^{n-3}\leq(n-1)^{2}2^{s} . Note that 1+2+\cdots+2^{s-1}=2^{s}-1\leq n , so 2^{n-3}\leq(n-1)^{2}(n+1) and n\leq 14 . One checks that 2^{\lfloor(n-s-1)/2\rfloor}>2(n-1) for n=14 . ∎ In Lemma 3.7 we write [k,l] for a permutation with two cycles of size k,l . Lemma 3.7 . Let G=A_{n} or S_{n} , n>4 , and, for a prime p, let 1\neq g\in G be a p-element, |g|=p^{a} . Suppose that p\neq\ell and g is almost cyclic on W_{n} . Then either g=[1^{n-p^{a}},p^{a}] , or p=2 , n=2^{a}+2 , g=[2,n-2] . In addition, \deg g\geq|g|-1 , and the equality holds if and only if |g|=n and g is cyclic, or p=2 , n=2^{a}+2 , g=[2,n-2] and the eigenvalue -1 multiplicity equals 2 . Proof. Observe that P_{n} is a direct sum of P_{c} , where c runs over the sizes of the cycles in the cycle decomposition of g . The characterrstic polynomial of a c -cycle element in P_{c} is x^{c}-1 , so the eigenvalues of this element in P_{c} are all c -roots of unity, each occurs with multiplicity 1. Let [c_{1}^{m_{1}},\ldots,c_{k}^{m_{k}}] be a cycle decomposition of g , where we can assume that c_{1}<\cdots1 . Recall that the composition factors of P_{n} are W_{n} and 1_{G} , the latter appears with multiplicity 1 if n is not a multiple of \ell , and with multiplicity 2 otherwise. Clearly, the multiplicities of all eigenvalues \lambda\neq 1 of g\in G in P_{n} and W_{n} are the same. If \lambda\neq\pm 1 then this occurs with multiplicity 1 in P_{n} . Therefore, g has at most one cycle of size greater than 2. If g has no cycle of size 2 then g=[1^{m_{1}},c_{2}] ; this case is recorded in the lemma conclusion. Suppose that g has a cycle of size 2 and g\neq[1^{m_{1}},2] . Then p=2 and -1 eigenvalues of g on W_{n} of multiplicity at least 2. Therefore, g=[2,c_{2}] (otherwise each 1 and -1 are eigenvalues of g on W_{n} of multiplicity at least 2). Whence the result. ∎ Lemma 3.8 . For 43 , in one of these two representations \phi(g) is cyclic and 1 is not an eigenvalue of \phi(g) , in the other representation \phi(g) is almost cyclic (not cyclic), \deg\phi(g)=7 and 1 is an eigenvalue of \phi(g) of multiplicity 2. Let n=8 and \ell=2 . Recall that A_{8}\cong SL_{4}(2) and |g|\in\{3,5,7\} . There are two non-equivalent irreducible representations of G of degree 4. One observes that \phi(g) is cyclic for |g|=5,7 . There are two classes of elements g of order 3, and \phi(g) is almost cyclic with \deg\phi(g)=3 exactly for one of two classes. Let n=7 and G=6.A_{7} . Then G has two irreducible representations \phi of degree 4 for \ell\neq 7 , |\phi(Z(G))|\leq 2 , and |\phi(Z(G))|=1 if and only if \ell=2 . These extend to 2.S_{7} if and only if \ell=7 . The group A_{7} has two conjugacy classes of elements of order 3. If g corresponds to a 3-cycle element in A_{7} then \phi(g) is almost cyclic whenever \dim\phi=4 . If g corresponds to a double 3-cycle element in A_{7} then \phi(g) is not almost cyclic. Let n=6 . If \dim\phi=4 and o(g)=3 then, as in the case with n=7 , there are two conjugacy classes of elements of order 3 in A_{6} , and \phi(g) is almost cyclic in one of them and not almost cyclic in the other (for \ell\neq 3 ). Note that one of these representations is subnatural, and if \phi is not subnatural then \phi(g) is almost cyclic when g corresponds to a double 3-cycle element in A_{6} . Let \dim\phi=5 . Then Z(G)=1 , \ell\neq 2 and G has two irreducible representation of degree 5, one of them is subnatural. (These are obtained from each other by a twist with an outer automorphism of A_{6}). If o(g)=3 then there are two conjugacy classes of elements of order 3, elements of one of them are almost cyclic and \deg\phi(g)=3 , those in the other class are not almost cyclic. Note that {\rm Out}\,(A_{n}) has order 2 if 6\neq n>4 and of order 4 (and of exponent 2) for n=6 . If G^{\prime}/Z(G^{\prime})\cong A_{n} and g\in G is of odd prime power order then g\in G^{\prime}Z(G) . For 42 be a prime, p\neq\ell and let g\in G be a p-element. Suppose that \phi(g) is almost cyclic. Then one of the cases of Table 1.2 holds. The entries of Table 1.2 are extracted from [ 14 ] . Note that {\rm Aut}\,A_{6} have no irreducible projective representation that is reducible on every normal proper subgroup. (Otherwise, by Clifford’s theorem, the group c.A_{6} , for some fixed c and \ell , would have 4 distinct irreducible Brauer characters of the same degree that are conjugate in {\rm Aut}\,A_{6} . This is not the case by [ 14 ] .) Let p=2 and \ell\neq 2 . Table 2.1 is obtained by inspection of [ 14 ] . In this table G/Z(G)=A_{n} and we differ the groups in question by indicating G^{\prime} in the 1st column. Table 2.2 lists the cases where G=c.A_{n}.2 , a non-split semidirect product of G=c.A_{n} and a group of order 2. As we shall see, this is sufficient for using in the proof of Theorem 1.1 and describe details on maximum eigenvalue multiplicity for g\in G . Recall that G/Z(G)\cong S_{n} if n\neq 6,n>4 , whereas the group A_{6} has 3 outer automorphisms of order 2, and hence 3 non-isomorphic groups A_{n}.2 . Following [ 14 ] we denote the projection of g into G/Z(G) by 2_{i} , i=1,2,3 , where i=2 if G/Z(G)\cong S_{n} , i=1 if G/Z(G)\cong PGL_{2}(9) , and 2_{3}=2_{1}\cdot 2_{2} . For every irreducible representation \phi of G one can consider \phi\otimes\alpha , where \alpha is a non-trivial one-dimensional representation of G . As G=\langle g,G^{\prime}\rangle , we have \alpha(g)=-1 . If e is a unique eigenvalue of \phi(g) of multiplicity m>1 then -e is an eigenvalue of (\phi\otimes\alpha)(g) of multiplicity m>1 too. In particular, \phi and \phi\otimes\alpha are not equivalent. In order to avoid repetitions of these very similar cases we write \pm e in the column headed e . The only case in Table 2.2 where \phi|_{G^{\prime}} is reducible is the one of degree 4 written at the 3rd row. 4 The main result Lemma 4.1 . Let G be a group such that G/Z(G)\cong A_{n} or S_{n} , n>9 , and let g\in G be a p-element for a prime p. Let \ell\neq p and let \phi:G\rightarrow GL(m,F) be a faithful irreducible representation such that \phi(g) is almost cyclic. Suppose that n=o(g) . Then \phi|_{G^{\prime}} has at most two irreducible constituents, both of them are subnatural. Proof. We identify G/Z(G) with a subgroup of S_{n} . Let h be the projection of g into G/Z(G) . Then h is a cycle of length n (as n=o(g) ). By Lemma 3.3 , if p=2 then G/Z(G)\cong S_{n} is generated by two conjugates of h ; if p>2 then A_{n} is generated by two conjugates of h . Let h^{\prime} be a conjugate of h such that \langle h,h^{\prime}\rangle\cong S_{n} , resp., A_{n} if p=2 , respectively, p>2 . Let g^{\prime}\in G be an element such that h^{\prime}=gZ(G) . Then \langle g,g^{\prime}\rangle projects onto \langle h,h^{\prime}\rangle . Set G_{1}=\langle Z(G),g,g^{\prime}\rangle . Then G_{1}=G if p=2 or p>2 and G/Z(G)=A_{n} , otherwise |G:G_{1}|=2 and G_{1}/Z(G_{1})\cong A_{n} . In the latter case \phi|_{G_{1}} has at most 2 irreducible constituents. Let \phi_{1} be one of them; then \dim\phi=a\dim\phi_{1} with a\leq 2 . As \phi is irreducible, the dimension of \phi , resp., of \phi_{1} does not exceed 2o(g)-2 by Lemma 2.1 . In each case, the dimension of an irreducible constituent \tau of \phi|_{G^{\prime}} do not exceed 2n-2 . Suppose first that \phi(Z(G^{\prime}))\neq 1 . Then \ell\neq 2 and \tau(G^{\prime})\cong 2.A_{n} . By Lemma 3.5 , \dim\tau\geq 2^{\lfloor(n-s-1)/2\rfloor} , where s is the number of non-zero terms in the 2-adic expansion of n . So \dim\tau\leq 2n-2 , which implies n\leq 13 by Lemma 3.6 . Let n=13 . Let \rho be a faithful irreducible representation of G^{\prime}=2.A_{13} . We show that d\geq 32 . Indeed, let X\cong 2.A_{12}\subset G^{\prime} . Then the irreducible constituents of \rho|_{X} are faithful representations of X . By [ 14 ] , the degree d^{\prime} , say, of a faithful irreducible representation \rho^{\prime} of X is not less that 32, unless \ell=3 where d^{\prime}\geq 16 . Moreover, if d^{\prime}>16 then d^{\prime}\geq 144 for \ell=3 . So we are left with \ell=3 , d^{\prime}=16 and \rho|_{X} is irreducible. Let x\in X be of order 11. By [ 14 ] , the Brauer character of \rho^{\prime} , and hence of \rho , takes an irrational value at x . This is a contradiction as x is conjugate to every x^{i} for 014 and \sigma is an irreducible modular representation of S_{n} then either all irreducible constituents of \sigma|_{G^{\prime}} are subnatural or \dim\sigma>n(n-5)/2 . If n>14 and \tau is not subnatural then \dim\tau\leq 2n-2 , so \dim\phi\leq 4n-4 , whence n\leq 12 . (Indeed, we have n(n-5)/2\leq 4n-4 implies n(n-5)\leq 8n-8 , n^{2}-13n+8\leq 0 .) As n is a prime power, n\leq 11 or n=13 . If \phi|_{G_{1}} is irreducible then \dim\phi\leq 2n-2 , whence n(n-5)/2\leq 2n-2 , a contradiction for n=13 . If \phi|_{G_{1}} is reducible then \dim\tau=\dim\phi_{1}\leq 2n-2 , which is 24 for n=13 . If \tau is not subnatural then \dim\tau>12 , and then \dim\tau\geq 32 by [ 2 ] , a contradiction. So n\leq 11 . As n is a p -power, n\neq 10 ; for n=11 the lemma is true by Tables 1.1 and 1.2 . ∎ Lemma 4.2 . Let p=2 and let G be a group such that G^{\prime}\cong c.A_{n} , n>11 , and G=\langle g,G^{\prime}\rangle , where g\in G is a 2 -element. Let \phi:G\rightarrow GL(m,F) be a faithful irreducible representation of G such that \phi(g) is almost cyclic. Then G^{\prime}\cong A_{n} and \phi|_{G^{\prime}} is a subnatural. Proof. Note that G/Z(G)\cong A_{n} or S_{n} and c=|Z(G^{\prime})|\leq 2 . Let h be the projection of g into G/Z(G) , and let k be the largest cycle size in the cycle decomposition of h . Note that k=|h|=o(g) . Suppose first that o(g)\geq 8 . If k=n then the result is contained in Lemma 4.1 . So 8\leq k11 .) Then h is contained in a subgroup Y of G isomorphic to A_{k}\times A_{n-k} or A_{k}\times S_{n-k} if h\in A_{n} and h\notin A_{n} , respectively. So h\in X_{1}\times X_{2} , where X_{1}\cong A_{k} and X_{2} is isomorphic to A_{n-k} or S_{n-k} . Let G_{1} be the preimage of X_{1} in G and G_{3}=\langle g,G_{1}\rangle . Then G_{3}^{\prime}=c.A_{k} with k\in\{8,10\} , and \tau(g) is almost cyclic for every irreducible constituent of \phi|_{G_{3}} . As in this case k\leq 10 , we can use Lemma 3.8 (instead of Lemma 4.1 used above). So either \tau is almost cyclic or \tau is as indicated in Table 2.1 or 2.2. However, none of the table entries for k=8,10 satisfies o(g)\leq 4 . This rules out the option o(g)\leq 4 , and completes the proof. ∎ Lemma 4.3 . For n>11 let G be a group such that G/Z(G)\cong A_{n} or S_{n} . Let g\in G\setminus Z(G) be a p-element for p>2 , and let \phi:G\rightarrow GL(m,F) be a faithful irreducible representation such that \phi(g) is almost cyclic. Then \phi(Z(G))=\mathop{\rm Id}\nolimits and \phi|_{G^{\prime}} are subnatural. Proof. As \phi is faithful, Z(G) is a cyclic group, and then we can assume that G=G^{\prime} . (Indeed, g\in G^{\prime}Z(G) as p>2 , and if g=g_{1}z for z\in Z(G),g_{1}\in G^{\prime} then \phi(g) is almost cyclic if and only if so is \phi(g_{1}) .) So we can assume that G=c.A_{n} for c\leq 2 , and then G=G^{\prime} . Let \Omega be the natural permutation set for A_{n} and let h be the projection of g into G/Z(G)=A_{n} . Let k=o(g) . By Lemma 4.1 , we can assume that h stabilizes some subset of \Omega_{k}\subset\Omega of k points, 1\leq k9 . The we claim that c=1 and the irreducible constutuents of \phi|_{G_{3}^{\prime}} are trivial or subnatural. Suppose the contrary, and let \tau be an irreducible constituent of \phi|_{G_{3}} with \dim\tau>1 . Then \tau(g) is almost cyclic. As \Omega_{k} is a maximal h -orbit on \Omega , we observe that k=o(g) is the minimal integer m such that g^{m}\in Z(G_{3}) . By Lemma 4.1 , c=1 and \tau|_{G_{3}^{\prime}} is subnatural. Whence the claim. As c=1 , we have G\cong A_{n} . It follows that every irreducible constituent of \phi|{G_{3}^{\prime}} is either trivial or subnatural. By Corollary 3.2 , \phi is subnatural. So the result follows if o(g)>9 . Suppose that o(g)=|g|\leq 9 . We first fix some special cases. Recall that n>11 . For n=12 the result follows by Lemma 3.8 . If n=13 then h fixes a point on \Omega , so the result follows from the one with n=12 . Let n>13 . Suppose the contrary, that \phi is not almost cyclic. If n=14 then h fixes a point on \Omega , unless h is a double cycle of order 7. In the latter case, by Lemmas 3.3 and 2.1 , we have \dim\phi\leq 12 . If c=1 then, arguing as in the proof of Lemma 4.1 , using [ 13 ] we observe that \dim\phi\geq n(n-5)/2 , whence n(n-5)/2\leq 4o(g)-4 ; for n=14 this yields 63\leq 24 , a contradiction. If c=2 then, by Lemma 3.5 , \dim\phi\geq 2^{5}=32 , which contradicts the above inequality \dim\phi\leq 24 . The argument for h a double cycle in A_{14} works similarly for h a double cycle of order 9 in A_{18} . So in this case \phi is almost cyclic. Note in addition, that, by Lemma 3.7 , \phi(g) is not almost cyclic if g is a double cycle in A_{2n} with n odd. In general, let n>14 . Then we revise the above partition \Omega=\Omega_{k}\cup\Omega_{n-k} , by choosing k=9,10,14 for o(g)=3,5,7 , respectively, and k=12,18 if o(g)=9 and a cycle of order 9 occurs once or twice, respectively, in the cycle decomposition of h . We mimic the above reasoning to deduce that c=1 and \phi is subnatural. Let \Omega_{k},G_{1},G_{3}=\langle g,G_{1}\rangle be as above for k just specified. By the above, we are left with the cases where the cycles of h on \Omega_{1} are of maximal possible sizes. So (o(g),k)\in\{(3,9),(5,10),(7,14),(9,12) , (9,18)\} . Recall that G_{3}/Z(G_{3})\cong A_{k} and G_{3}^{\prime}\cong c.A_{k} for c\leq 2 . Let \tau be an irreducible constituent of \phi|_{G_{3}} with \dim\tau>1 . Then \tau(g) is almost cyclic. We claim that \tau|_{G_{3}^{\prime}} is subnatural. If k\leq 12 then the claim follows from Lemma 3.8 and Table 1, if k=14,18 then this is proved in the previous paragraph. So c=1 , G\cong A_{n}, G_{3}^{\prime}\cong A_{k} and \tau|_{G_{3}^{\prime}} is irreducible. Therefore, every irreducible constituent of \phi|_{G_{3}^{\prime}} is either trivial or subnatural. So the result follows from Corollary 3.2 .∎ Corollary 4.4 . Let G=2.A_{n} , n>4 , and let g\in G be a p-element, p>2 . Let \phi:G\rightarrow GL_{m}(F) be a non-trivial irreducible representation of G. Suppose that \ell\neq p and \phi(g) is almost cyclic. If \deg\phi(g)2 and \deg\phi(g)4 , and let g\in G be a 2 -element. Let \phi:G\rightarrow GL_{m}(F) be a non-trivial irreducible representation of G. Suppose that \ell\neq 2 and \phi(g) is almost cyclic. If \deg\phi(g)2 , then we detect two more cases: G=3.A_{6}.2_{2} , \ell\neq 2 , o(g)=8 , m=\deg\phi(g)=3 and G=3.A_{6}.2_{3} , o(g)=8 , \ell\neq 2,3 , m=\deg\phi(g)=6. Proof of Theorem 1.1 . If n\leq 11 , the result follows by inspection of Tables 1,1 and 2.1. If n>11 then Lemma 4.2 for p=2 and Lemma 4.3 for p>2 imply the result. ∎ Department of Mathematics, University of Brasilia, Brasilia-DF 70910-900, Brazil e-mail: alexandre.zalesski@gmail.com Table 1.1. The non-trivial non-natural irreducible \ell -modular representations \phi of G=c.A_{n} , n\leq 10 , and p -elements g\in G , p>2 , with \phi(g) almost cyclic n \ell \dim\phi c o(g) \deg\phi(g) e m 5 2 2 1 3,5 2 - 1 5 \neq 2,p 2 2 3,5 2 - 1 5 \neq 2,p 3 1 3,5 3 - 1 5 \neq 2,3 4 2 3 4 1 2 5 \neq 2,5 4 2 5 4 - 1 5 \neq 2,3,5 5 1 5 5 - 1 5 \neq 2,5 6 2 5 4 1 2 6 3 2 2 5 2 - 1 6 \neq 3,p 3 3 3,5 3 - 1 6 3 3 1 5 3 - 1 6 2 4 1 3A 3 1 2 6 2 4 1 5 4 - 1 6 \neq 2,3 4 2 3B 4 1 2 6 \neq 2,5 4 2 5 4 - 1 6 \neq 2,3,5 5 1 5 5 - 1 6 \neq 2,3 5 1 3B 3 1 3 6 3 6 2 5 5 1 2 6 >5 6 3,6 5 5 1 2 7 5 3 3 3,7 3 - 1 7 2 4 1 3B 3 1 2 7 2 4 1 5,7 4 - 1 7 \neq 2,3 4 2 3B 3 1 2 7 \neq 2,p 4 2 5,7 4 - 1 7 \neq 2,3,p 6 3,6 5 5 1 2 7 \neq 2,3,7 6 3,6 7 6 - 1 7 2 6 3 5 5 1 2 7 3 6 2 7 6 - 1 8 2 4 1 3B 3 1 2 8 2 4 1 5,7 4 - 1 8 \neq 2,7 8 2 7 7 1 2 9 2 8 1 7 7 1 2 9 \neq 2,7 8 2 7 7 1 2 9 2 8 1 9 8 - 1 9 2 8 1 9 7 1 2 9 \neq 2,3 8 2 9A 7 1 2 9 \neq 2,3 8 2 9B 8 - 1 10 5 8 2 7 7 1 2 10 5 8 2 9A 7 1 2 10 5 8 2 9B 8 -1 2 In the tables e stands for the eigenvalue of \phi(g) of multiplicity greater than 1, if it exists, and m is the maximal eigenvalue multiplicity of \phi(g) . See comments after the proof of Lemma 0000000 on the last two lines in Table 2.1. In column o(g) we use notation of [ 14 ] to specify the conjugacy class of a certain order elements if this is necessary. Table 1.2 . Almost cyclic elements g\in G^{\prime} , |g| a p -power, p>2 , in non-trivial non-subnatural irreducible representations of G with G/Z(G)=S_{n} , 4