Projective modules for the symmetric group and Young's seminormal form. (Q2515624)

This paper concerns the problem of computing the decomposition numbers for the symmetric group in characteristic \(p\). This area recently suffered a huge shock when Williamson disproved James's Conjecture, which predicted that the decomposition numbers for \(\mathfrak S_n\) in characteristic \(p>\sqrt n\) (or, more generally, for any block of \(\mathfrak S_n\) of defect less than \(p\)) could be computed via the LLT algorithm. However, known counterexamples are extremely large, and James's Conjecture appears to be true a great deal of the time. The paper under review can be seen as a contribution to the on-going effort to determine what the range of validity of the James Conjecture is. The results in the paper are presented in slightly strange way, in that the central statement is a conjecture (Conjecture 1) which is immediately disproved by showing that it implies the James Conjecture. But the point is that Conjecture 1 holds for certain pairs \((n,p)\), and therefore implies the James Conjecture for these pairs. Moreover, the author gives a method for determining the validity of Conjecture 1 for a given pair \((n,p)\), and carries out some computations which verify James's Conjecture in small cases. However, these cases were already known. The main idea centres on the LLT algorithm. Translated to the setting of the Hecke algebra at a \(p\)th root of unity in \(\mathbb C\), this algorithm computes, for each simple module, the corresponding column of the decomposition matrix. This is done by first computing an `auxiliary vector' which equals the desired column plus some linear combination of other columns (smaller in a suitable ordering); a triangular reduction (using graded decomposition numbers) is then used to compute the desired column. The approach taken in the present paper is that the auxiliary vector corresponds to a (typically decomposable) projective module for the symmetric group, which can in theory be constructed. The author gives a construction of an idempotent in the group algebra, and Conjecture 1 asserts that the projective module it defines is the projective corresponding to the LLT auxiliary vector. The paper is on the whole very well written; in particular, the background material is laid out clearly and in detail. This paper therefore serves as a useful introduction to this problem and the methods used. Whether these methods will actually yield strong results not easily obtainable by other means is less clear. Unfortunately, as is increasingly common these days, this paper remains untroubled by the attentions of a copy editor; the references are not even in alphabetical order. What do we pay publishers for?