Stability of branching laws for highest weight modules. (Q2257456): Difference between revisions

Let \(G\) be a reductive algebraic group over \(\mathbb C\), and \(B\) its Borel subgroup. Let \(X\) be a \(G\)-variety which is spherical, i.e., \(B\) has a dense open orbit [cf. \textit{M. Brion}, Prog. Math. 80, 11-26 (1989; Zbl 0724.14034); \textit{M. Brion} et al., Invent. Math. 84, 617-632 (1986; Zbl 0604.14047)]. Spherical varieties arise as one of key ingredients in representation theory, and there are so many good properties for spherical varieties to hold. In this paper, the author investigates a property of \((\mathbb C[X],G)\)-module \(M\) over \(X\). Let \(V_\lambda^G\) be an irreducible finite dimensional representation of \(G\) with the highest weight \(\lambda\), and denote by \(m_M(\lambda)\) its multiplicity in \(M\). Put \(\Lambda^+(M):=\{\lambda\mid m_M(\lambda)>0\} \) and \(\Lambda^+(X):=\Lambda^+(\mathbb C[X])\) for brevity. We also choose a point \(x_0\in X\) which generates a dense open \(B\)-orbit in \(X\). Stabilizer of the orbit \(B\cdot x_0\) is a parabolic subgroup \(P\) in \(G\), and \(L=P_{x_0}\) is a reductive subgroup, where \(H_{x_0}\) denotes the fixed point subgroup of \(x_0\) in \(H\). The main theorem (Theorem 4.5) can be stated as follows: Suppose that \(X\) is quasi-affine and \(M\) is a finitely generated \(\mathbb C[X]\)-module without zero divisors. Then there exists a \(\lambda_0\in\Lambda^+(X)\) for which \(m_M^G(\lambda+\lambda_0)=m_{M_{x_0}}^L(\lambda|_{B_{x_0}})\) holds for any \(\lambda\in\Lambda^+(M)\), where \(M_{x_0}=M/\mathfrak m(x_0)M\) (\(\mathfrak m(x_0)\) denoting a maximal ideal of \(\mathbb C[X]\) corresponding to \(x_0\)). Since \(\mu|_{B_{x_0}}=0\) (\(\mu\in\Lambda^+(X)\)), it is immediate to conclude \(m_M^G(\lambda+\lambda_0)=m_M^G(\lambda+\mu+\lambda_0)\) for any \(\mu\in\Lambda^+(X)\), i.e., the multiplicity becomes ``stable'' along \(\Lambda^+(X)\) for sufficiently ``large'' weights (cf.~Theorem 4.1). This phenomenon was first noticed by \textit{Fumihiro Sato} and called \textit{stability of branching coefficients} by himself [Comment. Math. Univ. St. Pauli 42, No. 2, 189-207 (1993; Zbl 0824.22013)]. As another interpretation, \(M_{x_0}\) can be regarded as an isotropy representation of associated variety of a Harish-Chandra module. Thus the theorem roughly states that the multiplicity of \(K\)-types of a ``good'' Harish-Chandra module can be obtained from the branching of isotropy representation (if the associated variety is the closure of a spherical nilpotent orbit). The assumption that \(M\) has no zero divisors sounds very strong, however, there are many useful applications. Among others, the author discusses about 1) multiplicities in induced modules when \(X\) is \(G\)-homogeneous (Theorem 5.2); 2) global sections of vector bundles over projective varieties, which is \(H\)-spherical for a certain reductive subgroup \(H\subset G\) (Theorem 5.3); 3) stability theorem for unitary highest weight modules (Theorem 5.9); and 4) branching laws of holomorphic discrete series to a symmetric subgroup of Hermitian type (Theorem 5.13). Those results are quite interesting individually, but the nicer is that they can be derived from beautiful universal theorems in the framework of spherical varieties.