Representation theory via cohomology of line bundles (Q6063027)

For a connected reductive algebraic group \(G\) and a Borel subgroup \(B\) defined over an algebraically closed field \(k\), this expository article explores results of the author and others, on the relationship between the cohomology of line bundles on a certain flag variety \(G/B\) and the representation theory of \(G\). The base of this relationship is that a \(B\)-module \(E\) gives rise to \(G\)-modules \(H^{i}(E)\), either via the sheaf cohomology of an associated vector bundle \(\mathcal{L}(E)\) on \(G/B\), or via right-derived functor of induction \(R^{i}\operatorname{Ind}_{B}^{G}(E)\). Particularly important is the case that \(E\) is \(1\)-dimensional, so that \(E = \lambda\) for a weight \(\lambda\) of a fixed maximal torus \(T\) of \(B\); in this case we just write \(H^{i}(\lambda)\). This article focuses on two key problems: Describe when \(H^{i}(\lambda)\) vanishes and describe its \(G\)-module structure when it does not. This article gives an accessible, historical account of work on these problems over the past 70 years. The article is kept concise with a focus on the logical narrative and state of the art; terminology and technical details are brought in only to highlight key ideas. Section 2 begins by describing how the theory is at its most straightforward in characteristic \(0\), largely due to the Borel-Weil and Borel-Weil-Bott theorems showing that \(H^{0}(\lambda)\) is irreducible for dominant \(\lambda\), and in general \(H^{i}(\lambda)\) vanishes except possibly in one degree, where it is isomorphic to an induced module \(H^{0}(\lambda^{+})\) for an appropriate dominant \(\lambda^{+}\). In positive characteristic, all of this fails in generality, motivating both of the key problems. Section 3 discusses Kempf's vanishing theorem for dominant weights and its fundamental consequences. Sections 4 and 5 then directly address work on the vanishing problem and on the relevant \(G\)-module structures. This includes statements and applications of such central results as the Frobenius-Steinberg theorem, the strong linkage principle and structure results such as the Jantzen filtration and sum formula for Weyl modules (and more besides). The article closes with a discussion of modules \(H^{i}(\mu)\) for so-called generic weights \(\mu\), the study of which is much more tractable than the general case.