Dimension of zero weight space: an algebro-geometric approach (Q2253019)

Let \(G\) be a connected, adjoint, simple algebraic group over the complex numbers, with Borel subgroup \(B\). Let \(\Lambda\) denote the (integral) weight lattice and \(\Lambda^+\) denote the subset of dominant weights (which parameterizes the finite dimensional irreducible \(G\)-modules). For any \(\lambda \in \Lambda^+\), let \(V(\lambda)\) denote the associated irreducible \(G\)-module with highest weight \(\lambda\). A question of fundamental representation-theoretic interest is the dimension of the zero weight space of \(V(\lambda)\). In this paper, the authors take a global approach to this problem. Define a function \(\mu_0 : \Lambda^+ \to {\mathbb Z}\) by \(\mu_0(\lambda) = \dim V(\lambda)_0\). The main result of the paper is that the function \(\mu_0\) is essentially piecewise polynomial. By ``polynomial'' is meant that \(\mu_0\) is polynomial in the coefficients when \(\lambda\) is expressed as a linear combination of fundamental weights. The proof, as well as the precise statement, makes use of Geometric Invariant Theory (GIT). Set \(\Lambda({\mathbb R}) := \Lambda\otimes_{\mathbb Z}{\mathbb R}\). Let \(\Lambda^+({\mathbb R})\) denote the cone in \(\Lambda({\mathbb R})\) generated by \(\Lambda^+\) and \(\Lambda^{++}({\mathbb R})\) denote its relative interior. Consider the GIT classes of maximal dimension in \(\Lambda^{++}({\mathbb R})\). The authors identify these classes (or chambers) as the connected components of \(\Lambda^{++}({\mathbb R})\) modulo a certain set of hyperplanes (parameterized by the associated Weyl group and simple roots). Another key object involved is a certain sublattice \(\Gamma\) of the root lattice which depends upon the underlying root system. It is defined so that the natural \(G\)-equivariant line bundle on the flag variety \(G/B\) associated to a weight \(\lambda\) satisfies a certain descent property. Since \(G\) is of adjoint type, the weight lattice may be identified with the root lattice. As such, each weight \(\lambda \in \Lambda^+\) lies in some coset of \(\Gamma\). A more precise statement of the main result is that for each GIT class (mentioned above) and each coset of \(\Gamma\), there is a polynomial function \(f : \Lambda({\mathbb R}) \to {\mathbb R}\) with rational coefficients so that \(\mu_0(\lambda) = f(\lambda)\) for \(\lambda\) in the intersection of the coset and the GIT class. The proof uses in part the Riemann-Roch Theorem. The authors illustrate their result by explicitly computing \(\mu_0\) for all rank two groups (based on previously known information) as well as for type \(A_3\).