Proof of Stembridge's conjecture on stability of Kronecker coefficients (Q5964804)

If \(\lambda\) is a partition of \(n\), let \({\mathcal M}_{\lambda}\) denote the associated irreducible complex representation of the symmetric group \(S_n\). The Kronecker coefficients \(g_{\lambda,\mu,\nu}\) are the multiplicities which occur in the decomposition of the tensor product \({\mathcal M}_{\mu}\otimes{\mathcal M}_{\nu}\cong{\mathcal M}_{\lambda}^{\oplus g_{\lambda,\mu,\nu}}\). \textit{J. R. Stembridge} [``Generalized stability of Kronecker coefficients'', \url{http://www.math.lsa.umich.edu/~jrs/papers/kron.pdf}] made the following definition: A triple of partitions \((\alpha,\beta,\gamma)\) with \(|\alpha|=|\beta|=|\gamma|\) and \(g_{\alpha,\beta,\gamma}>0\) is stable if, for any other triple of partitions \((\lambda,\mu,\nu)\) with \(|\lambda|=|\mu|=|\nu|\), the Kronecker coefficient \(g_{d\alpha+\lambda,d\beta+\mu,d\gamma+\nu}\) is constant for \(d\gg 0\). He then conjectured: A triple \((\alpha,\beta,\gamma)\) is stable if and only if \(g_{d\alpha,d\beta,d\gamma}=1\) for \(d>0\). (Murnaghan's theorem, which states that \(g_{(d)+\lambda,(d)+\mu,(d)+\nu}\) is constant for \(d\gg 0\), is then an easy corollary.) J. R. Stembridge [loc. cit.] proved the sufficiency of the stability condition (as does the author in Sect. 4.1.); the paper under review proves the necessity. For partitions \(\lambda,\mu,\nu\) of \(n\), let \({\mathcal K}_{\lambda,\mu,\nu}\) be the \({\mathcal M}_{\lambda}\)-multiplicity space in \({\mathcal M}_{\mu}\otimes{\mathcal M}_{\nu}\), i.e. \({\mathcal M}_{\mu}\otimes{\mathcal M}_{\nu}=\bigoplus_{\lambda}{\mathcal K}_{\lambda,\mu,\nu}\otimes{\mathcal M}_{\lambda}\). (\({\mathcal K}_{\lambda,\mu,\nu}\) is a vector space whose dimension is \(g_{\lambda,\mu,\nu}\).) Then define \(B_{\alpha,\beta,\gamma}=\bigoplus_{d\geq 0}{\mathcal K}_{d\alpha,d\beta,d\gamma}\) and \(N_{\alpha,\beta,\gamma}^{\lambda,\mu,\nu}=\bigoplus_{d\geq 0}{\mathcal K}_{d\alpha+\lambda,d\beta+\mu,d\gamma+\nu}\). The main idea is that Hilbert functions on these spaces are identified with sequences of the relevant Kronecker coefficients, and the structure of these spaces (given in the main result, Theorem 1.3) then gives information about the Kronecker coefficients. Stembridge's conjecture then follows easily using the structure theorem for finitely generated \(\mathbb C[t]\)-modules (Corollary 1.4). Theorem 1.3 asserts that the space \(B_{\alpha,\beta,\gamma}\) has the structure of a finitely generated graded integral domain, and \(N_{\alpha,\beta,\gamma}^{\lambda,\mu,\nu}\) has the structure of a finitely generated torsion-free graded \(B_{\alpha,\beta,\gamma}\)-module. The proof links \(B_{\alpha,\beta,\gamma}\) and \(N_{\alpha,\beta,\gamma}^{\lambda,\mu,\nu}\) with the spaces \(A_{\alpha}\) and \(M_{\alpha,\lambda}\) (which are sums of Schur functors) using the Segre product and Schur-Weyl duality. The result follows from an analogous result for \(A_{\alpha}\) and \(M_{\alpha,\lambda}\) (Proposition 3.1), which is proved using invariant theory and the Borel-Weil theorem. The last section of the paper discusses related results, including Littlewood-Richardson and plethysm versions of Theorem 1.3, as well as twisted commutative algebras (of which the rings \(A_{\alpha}\) are an example) which led to the proof in this paper.