Colength sequences for matrices. (Q1763748): Difference between revisions

Let \(\chi_n=\sum m_\lambda\chi^\lambda\), \(\lambda\vdash n\), be the \(n\)-th cocharacter of the polynomial identities of \(k\times k\) matrices over a field of characteristic 0, decomposed as a sum of irreducible \(S_n\)-characters. The determination of the multiplicities \(m_\lambda\) remains one of the basic unsolved problems in the study of polynomial identities and invariants of matrices. Approaching this problem, one tries to describe the asymptotics of \(m_\lambda\). In particular, a lot of information can be obtained from the asymptotics of the colength \(l(n)=\sum m_\lambda\). The paper under review contains the important result that \(l(n)\) behaves asymptotically as a polynomial of degree \(k^2\choose 2\). The author obtains similar results also for the colength of the pure trace cocharacter sequence, the mixed trace cocharacter sequence, and for the cocharacter sequence of the centre of the generic matrix algebra. Again, the colength behaves as a polynomial, for some (possibly different) leading coefficients. The paper is very well written and uses several different techniques from the theory of PI-algebras, matrix invariants, combinatorics of symmetric functions and representation theory of \(GL_m\) and \(S_n\), and even evaluation of multiple integrals. Finally the author gives several conjectures about the colength, involving partitions \(\lambda\) with fixed number of parts, as well as \(*\)-polynomial identities of matrix algebras with involution.