Support varieties of line bundle cohomology groups for \(SL_3(k)\). (Q903936)

Let \(G\) be a semisimple simply connected algebraic group over an algebraically closed field \(k\) of prime characteristic \(p\). Let \(B\) be a Borel subgroup and \(G/B\) denote the associated flag variety. A weight \(\lambda\) defines a one-dimensional \(B\)-module and an associated line bundle \(\mathcal L(\lambda)\) on \(G/B\). For any \(i\geq 0\), the sheaf cohomology groups \(H^i(\lambda):=H^i(G/B,\mathcal L(\lambda))\) are rational \(G\)-modules. The module \(H^0(\lambda)\) is nonzero if and only if \(\lambda\) is a dominant weight. In that case, it is dual to a Weyl module. Conditions are also known for which \(H^1(\lambda)\neq 0\), but the vanishing of \(H^i(\lambda)\) for higher \(i\) remains an open problem. The goal of this paper is to study the cohomological support varieties \(V_{G_1}(H^i(\lambda))\) of \(H^i(\lambda)\) over the first Frobenius kernel \(G_1\) of \(G\). For good primes, \textit{D. K. Nakano, B. J. Parshall} and \textit{D. C. Vella} [J. Reine Angew. Math. 547, 15-49 (2002; Zbl 1009.17013)] computed \(V_{G_1}(H^0(\lambda))\) (for a dominant \(\lambda\)), identifying it with a certain \(G\)-orbit of nilpotent elements in the Lie algebra of \(G\). The author conjectures that, for good primes, if \(H^i(\lambda)\neq 0\), then \(V_{G_1}(H^i(\lambda))=V_{G_1}(H^0(w\cdot\lambda))\), where \(w\) is a Weyl group element such that \(w\cdot\lambda\) is a dominant weight. Using the notion of ``generic dimension'' of a module, the author obtains a general result that if \(H^i(\lambda)\neq 0\) for at least one \(i\geq 0\), then there exists some \(i\) for which \(V_{G_1}(H^i(\lambda))\) may be identified with a certain \(G\)-orbit of nilpotent elements. For a fixed \(\lambda\), if there is precisely one \(i\) for which \(H^i(\lambda)\) is nonzero, the conjecture would follow. This occurs for example when \(G=SL_2(k)\), but fails in general. The main focus of the paper is verifying the conjecture in the case \(G=SL_3(k)\). The proof is split into a variety of cases. For \(p\)-regular weights, the author makes use of corrected recursive character formulas originally found by \textit{S. Donkin} [J. Algebra 307, No. 2, 570-613 (2007; Zbl 1115.14008)]. For weights that are not \(p\)-regular, the author investigates the projectivity of \(H^i(\lambda)\) over \(G_1\), with the \(p=2\) case being dealt with separately.