The quadratic type of the 2-principal indecomposable modules of the double covers of alternating groups (Q1655797)

Let \(K\) be an algebraically closed field of characteristic \(2\) and let \(G\) be a finite group. An indecomposable module \(V\) for the group algebra \(KG\) is self-dual if and only if it has a non-degenerate \(K\)-valued \(G\)-invariant bilinear form \(\beta\). The form \(\beta\) may be assumed to be symmetric; if it is in addition symplectic, so \(\beta(v,v) = 0\) for all \(v \in V\), then it may in some cases be obtained as the polarisation \(q(v+w)-q(v)-q(w)\) of a \(G\)-invariant quadratic form on \(V\). Say that such modules have \emph{quadratic type}. \textit{R. Gow} and the author [Proc. Am. Math. Soc. 147, No. 7, 2783--2796 (2019; Zbl 07073434)] proved that the number of isomorphism classes of quadratic type indecomposable projective \(KG\)-modules is equal to the number of strongly real \(2\)-regular conjugacy classes of \(G\). For the alternating group \(A_n\), every real \(2\)-regular conjugacy class is strongly real: that is, if \(g\), \(g^{-1} \in A_n\) are in the same \(2\)-regular class then \(g^x = g^{-1}\) for some involution \(x \in A_n\). The double cover \(2.A_n\) is the unique group up to isomorphism such that \[ \langle z \rangle \hookrightarrow 2.A_n \twoheadrightarrow A_n \] is a non-split central extension of \(A_n\) by \(\langle z \rangle \cong C_2\). (In fact, by a theorem of Schur, \(2.A_n\) is the universal perfect central extension of \(A_n\) for each \(n \not= 6, 7\).) Writing \(\overline{g}\) for the image of \(g \in 2.A_n\) under the quotient map \(2.A_n \twoheadrightarrow A_n\), it may be that \(\overline{g}\) is strongly real while \(g\) is not: for example, since the only involution in \(2.A_6\) is the central involution \(z\), \emph{no} \(2\)-regular conjugacy class in \(2.A_6\) is strongly real. It is therefore an interesting problem, solved in this paper, to determine which indecomposable projective \(K(2.A_n)\)-modules are of quadratic type. To state the author's main result we need some further notation. Let \(S_n\) denote the symmetric group of degree \(n\) and let \(D^\mu\) denote the simple \(KS_n\)-module canonically labelled by the partition \(\mu\) of \(n\) with distinct parts. By a theorem of Benson, quoted as Lemma 5 in the paper under review, \(D^\mu \downarrow_{A_n}\) is reducible if and only if \(\mu_{2j-1} - \mu_{2j} \in \{1,2\}\) and \(\mu_{2j-1} + \mu_{2j} \not\equiv 2\) mod \(4\) for each \(j\); when reducible, the two summands are each self-dual if and only if \(\mu_2 + \mu_4 + \cdots \) is even. Let \(D^\mu_A\) denote an indecomposable direct summand of \(D^\mu \downarrow_{A_n}\). Since the centre \(\langle z \rangle\) of \(2.A_n\) acts trivially on every simple \(2.A_n\)-module, the simple \(2.A_n\) modules are, up to isomorphism, the \(D^\mu_A\) for \(\mu\) a partition of \(n\) with distinct parts. Let \(P^\mu\) denote the projective cover of \(D^\mu_A\) \emph{as a module for \(2.A_n\)}. Let \(|\mu|_a = \mu_1 - \mu_2 + \mu_3 - \cdots \) be the alternating sum of the parts of the partition \(\mu\) and let \(\ell_o(\mu)\) be the number of odd parts: note that \(|\mu|_a \ge \ell_o(\mu)\). The main result of this paper is that \(P^\mu\) has quadratic type if and only if \[ \frac{n-|\mu|_a}{2} \le 4m \le \frac{n-\ell_o(\mu)}{2} \] for some integer \(m\). The proof uses the construction of \(D^\mu\) as the quotient of the Specht module \(S^\mu\) by the radical of the canonical bilinear form on \(S^\mu\), and combinatorial arguments with tableaux. The paper ends with the helpful example of \(2.A_{13}\), which shows for instance that while the basic spin module \(D^{(7,6)}_A\) of dimension \(32\) is self-dual, its projective cover does not have quadratic type. Finally, it is worth mentioning in connection with this paper a conjecture of Kleshchev, that every simple module for \(KA_n\) has even dimension. This is known for the self-dual simple modules, since by a theorem of Gow, they have symplectic type. Thus the real problem is to prove the conjecture for the summands of \(D^\lambda \downarrow_{A_n}\) when the restriction splits as \(U \oplus V\) with \(U^\star \cong V\); in this case the conjecture is open and appears to be difficult.