A simple approach to matrix realizations for Littlewood-Richardson sequences (Q1300842)

Littlewood-Richardson sequences are sequences of partitions used to index various objects in combinatorics, representation theory, and module theory. The interest of the author in this paper, is to discuss their use to code isomorphism types for torsion modules, submodules and their quotients over a discrete valuation ring of characteristic zero (e.g. finite abelian \(p\)-groups). For more details on Littlewood-Richardson sequences, see \textit{I. G. Macdonald} [Symmetric functions and Hall polynomials (Oxford University Press, London)(1979; Zbl 0487.20007)]. Let \(S=(\lambda^{(0)},\lambda^{(1)},\ldots ,\lambda^{(r)})\) be a sequence of partitions. The sequence \(S\) is a Littlewood-Richardson sequence of type \((\mu,\nu;\lambda)\) if \(\lambda^{(0)}=\mu\), \(\lambda^{(r)}=\lambda\), and \(\lambda^{(i-1)}\subseteq \lambda^{(i)}\) for all \(i\), the skew diagram \(\lambda^{(i)}-\lambda^{(i-1)}\) is a horizontal strip containing \(\nu_i\) boxes, for all \(i\), and \(\sum_{j=1}^k(\lambda_j^{(i)}-\lambda_j^{(i-1)})\geq \sum_{j=1}^{k+1}(\lambda_j^{(i+1)}-\lambda_j^{(i)})\) for \(i\geq 1\) and \(k\geq 0\). This last condition can be interpreted as the existence of a filling of \(\lambda-\mu\) satisfying certain properties. If \(\lambda\) is a partition, let \(\lambda'\) denote the conjugate partition. Let \(c_{\mu\nu}^{\lambda}\) denote the number of Littlewood-Richardson sequences of type \((\mu,\nu;\lambda)\), and let \(R\) denote an arbitrary discrete valuation ring of characteristic zero. It can be shown (see Macdonald, reference as above, \textit{T. Klein} [J. Lond. Math. Soc. 43, 280-284 (1968; Zbl 0188.09504)], and \textit{J. A. Green} [Les polynomes de Hall et les caractères des groupes \(GL(n,q)\), in: Colloque d'Algèbre supérieure, C. B. R. M., Louvain, 207-215 (1957; Zbl 0086.02501)]) that \(c_{\mu\nu}^{\lambda}>0\) if and only if there exist finitely generated torsion modules \(M\), \(N\) and \(L\) such that \(N\subseteq L\), \(M\cong L/N\), \(M\cong (R/{t^{\mu_1}R})\oplus (R/{t^{\mu_2}R})\oplus\cdots\), \(N\cong (R/{t^{\nu_1}R})\oplus (R/{t^{\nu_2}R})\oplus\cdots\), and \(L\cong (R/{t^{\lambda_1}R})\oplus (R/{t^{\lambda_2}R})\oplus\cdots\), where \(t\) is a uniformizing parameter for the discrete valuation ring \(R\). The author recasts the above result in a language which removes the need for conjugate partitions. This is important in this context, as it makes the connection with the modules simpler and more explicit, as well as allowing them to rewrite it in terms of matrix realizations of the modules over \(R\) (called matrix realizations of the original Littlewood-Richardson sequence). He is then enabled to show that any matrix realization of a Littlewood-Richardson sequence may be `factored' in a certain sense, in a way compatible with the rows of \(\nu\) (this has a close relationship with the Littlewood-Richardson filling described above). The author also describes this factorization explicitly. The proof of this result involves a combinatorial analysis of matrix realizations associated to the Littlewood-Richardson sequence.