Study of parity sheaves arising from graded Lie algebras (Q6097858)

Let \(G\) be a complex, connected, reductive algebraic group and \( \chi\colon \mathbb{C}^{\times}\rightarrow G \) be a fixed cocharacter. Let \( G_{0} = \{ g\in G \mid g\chi ( t ) = \chi ( t )g \ \forall t\in\mathbb{C}^{\times} \} \) be the centralizer of \( \chi ( \mathbb{C}^{\times} ) \). If \( \mathfrak{g} \) is the Lie algebra of \(G\), let \( \mathfrak{g}_{n} \) be the subspace of \( \mathfrak{g} \) defined by \( \mathfrak{g}_{n} = \{ x\in \mathfrak{g} \mid \mathrm{Ad} ( \chi ( t ) )x = t^{n} x \ \forall t\in\mathbb{C}^{\times} \} \). This gives the decomposition \( \mathfrak{g} = \bigoplus_{n\in \mathbb{Z}}\mathfrak{g}_{n} \) which is a grading on \( \mathfrak{g} \). If \(H\) is a linear algebraic group, \( \Bbbk \) is a field of characteristic \(l\ge 0\) and \(X\) is an \(H\)-variety, then \( D_{H}^{b} ( X, \Bbbk ) \) denotes the derived category of \(H\)-equivariant constructive sheaves defined in [\textit{J. Bernstein} and \textit{V. Lunts}, Equivariant sheaves and functors. Berlin: Springer-Verlag (1994; Zbl 0808.14038)]. The paper under review has accomplished the detailed study of the category \( D_{G_{0}}^{b} ( \mathfrak{g}_{n}, \Bbbk) \) where \( \Bbbk \) is a field of positive characteristic. The simple perverse sheaves on \( \mathfrak{g}_{n} \) are indexed by pairs \( ( \mathcal{O}, \mathcal{L} )\), where \(\mathcal{O}\) is a \( G_{0} \)-orbit contained in \( \mathfrak{g}_{n} \) and \( \mathcal{L} \) is an irreducible \( G_{0} \)-equivariant \( \Bbbk \)-local system on \(\mathcal{O}\). This set of pairs is denoted by \( \mathcal{J} ( \mathfrak{g}_{n}, \Bbbk ) \). Next the subset of all cuspidal pairs, \( \mathcal{J} ( \mathfrak{g}_{n}, \Bbbk )^{\mathrm{cusp}} \subset \mathcal{J} ( \mathfrak{g}_{n}, \Bbbk ) \), is introduced, and the simple perverse sheaf associated to \( ( \mathcal{O}, \mathcal{L} )\) is denoted by \( \mathcal{I}\mathcal{C} ( \mathcal{O}, \mathcal{L} ) \). \textit{G. Lusztig} has proved in [Adv. Math. 112, No. 2, 147--217 (1995; Zbl 1001.17033)] that for \( \Bbbk = \mathbb{C} \), every simple perverse sheaf is a direct summand of the parabolic induction of some cuspidal pair. The topic of the paper under review stems primarily from the fact that this result becomes untrue for fields of non-zero characteristic. So, in the present paper, the author makes an attempt to develop Lusztig's theory for fields of non-zero characteristic. Towards this end, the author makes several assumptions concerning the characteristic of the field \(\Bbbk\), the size of this field and the group \(G\) itself. The author takes into account the idea that in the situation thus described, it is reasonable to replace the notion of semisimple complex by that of parity complex, and uses the notion of clean pair \( ( \mathcal{O}, \mathcal{L} )\in \mathcal{J} ( \mathfrak{g}_{n}, \Bbbk ) \) extensively in the course of the present investigation. In addition, the parity sheaf associated to the pair \( ( \mathcal{O}, \mathcal{L} )\) is denoted by \( \mathcal{E} ( \mathcal{O}, \mathcal{L} ) \). Within the framework of all these assumptions and conventions, the main results of the paper run as follows. \begin{itemize} \item[(1)] For any cuspidal pair \( ( \mathcal{O}, \mathcal{L} )\in \mathcal{J} ( \mathfrak{g}_{n}, \Bbbk )^{\mathrm{cusp}} \), \( \mathcal{I}\mathcal{C} ( \mathcal{O}, \mathcal{L} ) \) is clean and so \( \mathcal{I}\mathcal{C} ( \mathcal{O}, \mathcal{L} ) = \mathcal{E} ( \mathcal{O}, \mathcal{L} ) \). \item[(2)] Parabolic induction takes parity complexes to parity complexes. \item[(3)] For any pair \( ( \mathcal{O}, \mathcal{L} )\in \mathcal{J} ( \mathfrak{g}_{n}, \Bbbk ) \), \( \mathcal{E} ( \mathcal{O}, \mathcal{L} ) \) exists and is a direct summand of the parabolic induction of some cuspidal pair. \end{itemize}