Quadratic differentials and equivariant deformation theory of curves (Q714922)

Let \(X\) be a non-singular complete curve of genus \(g_X \geq 2\) defined over an algebraic closed field \(k\) of characteristic \(p,\) and let \(G\) be a subgroup of the automorphism group of \(X.\) The dimension of the tangent space of the associated equivariant deformation functor (in the sense of Schlesinger) roughly measures in ``how many ways'' \(X\) can be deformed to another curve which also admits \(G\) as a group of automorphisms. It was shown by \textit{J. Bertin} and \textit{A. Mézard} [Invent. Math. 141, No. 1, 195--238 (2000; Zbl 0993.14014)] that this tangent space is isomorphic to \(V_1 := H^1 (G, {\mathfrak T}_X)\), \({\mathfrak T}_X\) being the tangent sheaf of \(X\). If \(p \not= 0\) and \(G\) is a \(p\)-group, \(V_1\) has the same dimension as \(V_0 := H^0 (X, \Omega_X^{\otimes 2})_G\), the co-invariants of \(G\) acting on the space of global holomorphic quadratic differentials on \(X\) [\textit{A. Kontogeorgis}, J. Pure Appl. Algebra 210, No. 2, 551--558 (2007; Zbl 1120.14020)]. When the action of \(G\) is tame, it is known that dim \(V_1 = 3(g_{Y}-1) + r\), where \(g_Y\) is the genus of the quotient curve \(Y = X/G\) and \(r\) the cardinality of the branch locus \(Y_{\mathrm{ram}}\). When the action of \(G\) is wild, the problem is of course much more difficult. Here, assuming that \(p \not= 0\) and \(G\) is a \(p\)-group, the authors aim to compute \(\dim V_0\) in particular cases by using known results on the Galois module structure of Riemann-Roch spaces \(H^0 (X, {\mathcal O}_X (D))\) for certain \(G\)-invariant divisors on \(X\) : -- if \(G\) is cyclic, \textit{N. Borne} [Adv. Math. 201, No. 2, 454--515 (2006; Zbl 1105.14063)] has explicitly determined the Galois module structure of \(H^0 (X, {\mathcal O}_X (D))\) for any \(G\)-invariant divisor \(D\) on \(X\) of degree greater that \(2 g_X - 2\). Using this result, the authors compute the total number of direct summands in the decomposition of \(H^0 (X, {\mathcal O}_X (D))\) into indecomposable \(k[G]\)-modules, from which they derive a formula giving dim \(V_0\) in the same (but more complicated) terms as in the tame case. They note that this could also be obtained from core results in Bertin-Mézard [loc. cit.]. -- if the action of \(G\) is weakly ramified, i.e. if the second ramification group \(G_2 (P)\) vanishes for all \(P \in X,\) they introduce the space \(W\) of global meromorphic quadratic differentials on \(X\) which may have a pole of order at most 3 at each ramified point and which are holomorphic everywhere else. It is known [\textit{B. Köck}, Am. J. Math. 126, No. 5, 1085--1107 (2004; Zbl 1095.14027)] that \(W\) is a free \(k[G]\)-module of rank \(3 (g_Y - 1 + r)\), and the authors can express the difference \(\dim W_G - \dim V_0\) in terms of group homology of \(G\) acting on the space of global sections of a certain skyscraper sheaf, hence dim \(V_0\) using a spectral sequence argument. In a concluding section, assuming that \(p > 3\) and \(X \to Y\) is not unramified, the authors show that there exists a \(G\)-invariant effective canonical divisor \(D\) on \(X,\) which allows them to split the \(k[G]\)-module \(H^0(X, \Omega_X^{\otimes 2}) \simeq H^0 (X, \Omega_X (D))\) into its semi-simple and nilpotent parts with respect to the corresponding Cartier operator: \(H^0 (X, \Omega_X (D)) \simeq H^s_D \oplus H^n_D.\) They show that the \(k[G]\)-module \(H^s_D\) is free and compute its rank and hence \(\dim (H^s_D)_G\) provided \(D\) satisfies further conditions. In cases where the \(k [G]\)-module structure of \(H^0(X, \Omega_X^{\otimes 2})\) is known, this gives information about the mysterious part \(H^n_D\).