Lectures on Frobenius splittings and \(B\)-modules (Q1316606)

In these lecture notes the author proves some conjectures of Donkin, Joseph and Polo. The key idea behind his proofs comes from \textit{O. Mathieu} [Ann. Sci. Éc. Norm Supér., IV. Sér. 23, 625-644 (1990; Zbl 0748.20026)] but he has also used arguments that go back to earlier papers, in particular to the work of \textit{P. Polo} [e.g. Astérisque 173-174, 281-311 (1989; Zbl 0733.20021)] and his own contribution [Math. Z. 201, 19-31 (1989; Zbl 0642.20037)]. One of the nice features of his treatment is the cohomological criteria he gives for a module to possess various filtrations, see below. Moreover, he has done a considerable effort in explaining and streamlining the arguments and in his presentation he has put the main emphasis on \(B\)-modules although the geometry relating to the Frobenius splitting techniques (introduced by \textit{V. Mehta} and \textit{A. Ramanathan} [Ann. Math., II. Ser. 122, 27-40 (1985; Zbl 0601.14043)] and recalled in an appendix in this book) certainly still plays a crucial role. The reader who would like to have even more purely representation theoretic proofs may want to use the reviewer's paper [Invent. Math. 79, 611-618 (1985; Zbl 0591.14036)] which contains most of those basic results (vanishing theorems, normality of Schubert varieties and Demazure's character formula) which the author here derives from Frobenius splitting. Let \(G\) be a semi-simple algebraic group with a Borel subgroup \(B\). Then the Schubert varieties \(X_ w\) in the flag variety \(G/B\) are indexed by elements \(w\) of the Weyl group \(W\) for \(G\). After some quick preliminaries on the basics about reductive groups, about the Demazure desingularization of \(G/B\) and about \(B\)-modules the author presents Polo's theorem. In his terminology this is the statement that the dual Joseph modules are injective in certain weight limited subcategories of the category of \(B\)-modules. By a dual Joseph module he understands a \(B\)-module of the form \(H^ 0(X_ w,{\mathcal L})\) where \(\mathcal L\) is an effective line bundle on \(G/B\). Following Polo we say that a \(B\)-module has an excellent filtration (in short is excellent) if it has a filtration whose quotients are dual Joseph modules. This is a generalization of Donkin's concept of good filtrations: if \(w\) is the longest element in \(W\) then \(X_ w = G/B\) and the corresponding dual Joseph module is a dual Weyl module (in particular is a \(G\)-module). A \(G\)-module has a good filtration (in short is good) if it has a filtration whose quotients are dual Weyl modules. Donkin's conjecture says that the tensor product of two good modules is again good and that goodness is preserved upon restriction to parabolic subgroups. This is the next theorem which the author presents. Here he uses as a crucial ingredient Mathieu's clever application of Frobenius splitting (see loc. cit.). Recall that \textit{S. Donkin} [Rational representations of algebraic groups (Lect. Notes Math. 1140, 1985; Zbl 0586.20017)] and \textit{J.-P. Wang} [J. Algebra 77, 162-185 (1982; Zbl 0493.20023)] (on the first part) had proved the result in most cases but that Mathieu's proof not only works in general but also avoids all case by case considerations. See \textit{J. Paradowski}'s paper [Proc. Symp. Pure Math. 56, Part 2, 93- 108 (1994)] for an alternative approach relying on Lusztig's canonical basis. Joseph's (resp. Polo's) conjecture says that the tensor product of a good (resp. excellent) module with a one dimensional \(B\)-module which induces an effective line bundle on \(G/B\) is excellent. These two conjectures are proved next (in a somewhat different fashion from the way Mathieu obtains them). Finally it should be mentioned that the author throughout the text gives many variations on the theme (e.g. Schubert filtrations, passage to characteristic zero, etc.) and one also finds many interesting remarks and questions scattered around in this book.