Generic Cartan invariants for Frobenius kernels and Chevalley groups (Q1121386)

Let G be a semisimple, simply connected algebraic group over an algebraically closed field K of characteristic \(p>0\). For \(n>0\), let \(G_ n\) be the schemetheoretic kernel of the n-th power of the Frobenius morphism and let \(G_ nT\) be the related group scheme for a fixed maximal torus T of G. For each \(\lambda\) in the character group X(T), L(n,\(\lambda)\) (resp. L(\(\lambda)\)) denotes the simple \(G_ nT\)-module (resp. G-module) of highest weight \(\lambda\) and Q(n,\(\lambda)\) the injective hull \((=projective\) cover) of L(n,\(\lambda)\). Denote by \(X_ n(T)\) the set of dominant weights with coordinates between 0 and \(p^ n- 1\). Each \(\lambda\in X(T)\) can be written uniquely as \(\lambda^ 0+p^ n\lambda^ 1\) with \(\lambda^ 0\in X_ n(T)\). When \(p\geq 2h-2\) (h being the Coxeter number of G), it is known that there is a G-module, for each \(\lambda \in X_ n(T)\), restricting to the \(G_ nT\)-module Q(n,\(\lambda)\), this G-module is also called Q(n,\(\lambda)\). The group of rational points of G over a field of \(p^ n\) elements is denoted by G(n). The simple KG(n)-modules are the restrictions of the L(\(\lambda)\) for \(\lambda \in X_ n(T)\). When \(p\geq 2h-2\), the G-module Q(n,\(\lambda)\) is injective and projective for KG(n) with the injective hull \((=\) projective cover) U(n,\(\lambda)\) of L(\(\lambda)\) occurring once as a summand. Given \(\lambda \in X_ n(T)\), the Cartan invariant \(\hat c_ n(\lambda,\mu)\) is the multiplicity of L(n,\(\mu)\) as a factor of the \(G_ nT\)-module Q(n,\(\lambda)\). Define \(c_ n(\lambda,\mu)\) to be the multiplicity of the G(n)-module L(\(\mu)\) as a factor of U(n,\(\lambda)\). Let \(W_ p\) be the affine Weyl group associated with the Weyl group of G and the corresponding alcoves in X(T)\(\otimes {\mathbb{R}}\). \(W_ p\) acts on weights via the dot action: \(w\cdot \lambda =w(\lambda +\rho)-\rho\), where \(\rho\) is the sum of fundamental dominant weights. In the paper under review, the author regards (A)-(D) below as the conditions on Cartan invariants for the KG(n)-modules U(n,\(\lambda)\) to be generic. (A) The weight \(\lambda\) lies in the interior of some alcove for \(W_ p.\) (B) As a G(n)-module, Q(n,\(\lambda)\) remains indecomposable and is therefore isomorphic to U(n,\(\lambda)\). (C) Let \(L(\mu^ 0)\otimes L(\mu^ 1)^{[n]}\) be any G composition factor of Q(n,\(\lambda)\), where [n] denotes twisting by the n-th power of the Frobenius map. If A is the alcove containing \(\mu^ 0\), then \(\mu^ 0\) lies at a distance of at least \(r(\mu^ 1)\) from all walls of A, where \(r(\mu^ 1)\) denotes half the diameter of the weight diagram of \(L(\mu^ 1).\) (D) Let both L(\(\mu)\) and L(\(\nu)\) be G composition factors of Q(n,\(\lambda)\). If \(\mu^ 0\) and \(\nu^ 0\) lie in the same alcove but are distinct, then \(\mu^ 0+\pi \neq \nu^ 0+\sigma\) for all weights \(\pi\), \(\sigma\) of \(L(\mu^ 1)\), \(L(\nu^ 1)\), respectively. Then the main result of this paper is stated as follows. Suppose \(\lambda \in X_ n(T)\) satisfies conditions (A)-(D). Then there is a natural 1-1 correspondence between the distinct \(G_ nT\) composition factors of Q(n,\(\lambda)\) and the distinct G(n) composition factors of \(U(n,\lambda)=Q(n,\lambda)\). The correspondence sends \(L(n,\mu^ 0+p^ n\pi)\) to \(L(\mu^ 0+\pi)\). Moreover, the composition factor multiplicities agree, i.e. the respective Cartan invariants are equal: \(\hat c_ n(\lambda,\mu^ 0+p^ n\pi)=c_ n(\lambda,\mu^ 0+\pi).\) Note that L. Chastkofsky and Ye Jia-Chen have described the generic Cartan invariants for finite groups of Lie type in two different ways. The present paper offers a more conceptual treatment on these Cartan invariants which avoids the messy explicit calculations.