A proof of Wahl's conjecture in the symplectic case (Q1949288)

Let \(G\) be a connected semisimple algebraic group over an algebraically closed field \(k\), and let \(B\) be a Borel subgroup and \(P\supset B\) a parabolic subgroup. Let \(X_P= G/P\) and \(X= G/B\). Wahl conjectured (in \(\text{char\,}k= 0\)) that the Gaussian map \[ H^0(X_P\times X_P, I_\Delta\otimes({\mathcal L}_1\boxtimes{\mathcal L}_2))\to H^0(X_P, \Omega^1_{X_P}\otimes{\mathcal L}_1\otimes{\mathcal L}_2), \] is surjective for any ample line bundles \({\mathcal L}_1\), \({\mathcal L}_2\) on \(X_P\), where \(I_\Delta\) is the ideal sheaf of the diagonal \(\Delta\) and \(\Omega^1_{X_P}\) is the sheaf of 1-forms on \(X_P\). This conjecture was proved (in \(\text{char\,}k= 0\)) by the reviewer by using the combination of geometric and representation theoretic techniques. The same question when \(\text{char\,}k= p> 0\) has been considered by Frobenius splitting methods. For any smooth projective variety \(Y\) over \(\text{char\,}k= p> 0\), any Frobenius splitting of \(Y\) canonically gives rise to a section \(\sigma\) of the line bundle \(\omega^{\otimes 1-p}_Y\), where \(\omega_Y\) is the canonical line bundle of \(Y\). Moreover, if an irreducible smooth closed subscheme \(Z\) of \(Y\) of codimension \(d\) is compatibly split, then \(\sigma\) vanishes with multiplicity at most \((p- 1)d\) along \(Z\). If \(\sigma\) vanishes with multiplicity exactly equal to \((p- 1)d\) along \(Z\), we call the splitting of \(Y\) to maximally compatibly split along \(Z\). Such a Frobenius splitting of \(Y\) lifts to a Frobenius splitting of the blow-up \(B_Z(Y)\) of \(Y\) along \(Z\) compatibly splitting the exceptional divisor (and conversely). \textit{V. Lakshmibai}, \textit{V. B. Mehta} and \textit{A. J. Parameswaran} [J. Algebra 208, No. 1, 101--128 (1998; Zbl 0955.14006)] showed that if the blow-up \(B_\Delta(X_P\times X_P)\) is Frobenius split compatibly splitting the exceptional divisor, then the analogue of Wahl's conjecture for \(X_P\) is true over \(k\) of \(\text{char\,} p> 0\). Moreover, they conjectured that \(X_P\times X_P\) admits a Frobenius splitting maximally compatibly splitting the diagonal \(\Delta\). We will refer to this conjecture as the LMP conjecture for \(X_P\). The LMP conjecture was proved for Grassmannians by \textit{V. B. Mehta} and \textit{A. J. Parameswaran} [Int. J. Math. 8, No. 4, 495--498 (1997; Zbl 0914.14021)] for symplectic and orthogonal Grassmannians for any odd \(p\) by \textit{V. Lakshmibai}, \textit{K. N. Raghavan} and \textit{P. Sankaran} [Cent. Eur. J. Math. 7, No. 2, 214--223 (2009; Zbl 1200.14100)]; for any minuscule \(G/P\) by \textit{J. Brown} and \textit{V. Lakshmibai} [Proc. Indian Acad. Sci., Math. Sci. 119, No. 5, 571--592 (2009; Zbl 1192.14036)]; and for \(X= SL_N/B\) (and hence for \(X_P= SL_N/P\) for any parabolic subgroup \(P\)) by \textit{N. Lauritzen} and the author [Invent. Math. 128, No. 3, 437--442 (1997; Zbl 0871.20035)]. In the paper under review, the author realizes the splitting of \(SL_N/B\times SL_N/B\) (maximally splitting the diagonal) given by \textit{Lauritzen} and the author [loc. cit.] in a different way and shows that the splitting is a canonical splitting. This new realization of the splitting generalizes easily to give a proof of the LMP conjecture for any flag variety \(X_P= SP_{2N}/P\) of \(SP_{2N}\). This is the main result of the paper.