An LLT-type algorithm for computing higher-level canonical bases (Q989073): Difference between revisions

From the introduction: Let \(e > 2\) be an integer. In this paper the author considers the integrable representation theory of the quantised enveloping algebra \(\mathcal U = U_q(\widehat{\mathfrak{sl}}_e)\). For any dominant integral weight \(\Lambda\) for \(\mathcal U\), the irreducible highest-weight module \(V(\Lambda)\) for \(\mathcal U\) can be constructed as a submodule \(M^s\) of a Fock space \(\mathcal F^s\) (which depends not just on \(\Lambda\) but on an ordering of the fundamental weights involved in \(\Lambda\)). Using the standard basis of the Fock space, one can define a canonical basis (in the sense of Lusztig/Kashiwara) for \(M^s\). There is considerable interest in computing this canonical basis (that is, computing the transition coefficients from the canonical basis to the standard basis) because of Ariki's theorem, which says that these coefficients, evaluated at \(q = 1\), yield decomposition numbers for certain cyclotomic Hecke algebras. In the case where \(\Lambda\) is of level 1, there is a fast algorithm due to \textit{A. Lascoux, B. Leclerc} and \textit{J.-Y. Thibon} [Commun. Math. Phys. 181, 205--263 (1996; Zbl 0874.17009)] for computing the canonical basis. The purpose of this paper is to give a generalisation of this algorithm to higher levels. \textit{B. Leclerc} and \textit{J.-Y. Thibon} [Int. Math. Res. Not. 1996, No. 9, 447--456 (1996; Zbl 0863.17013)] showed how the canonical basis could be extended to a basis for the whole of the Fock space in the level 1 case. This was generalised to higher levels by \textit{D. Uglov} [Prog. Math. 191, 249--299 (2000; Zbl 0963.17012)] , but using a `twisted' Fock space (which is not obviously isomorphic to a tensor product of level 1 Fock spaces). By using Uglov's construction and taking a limit, one can define a canonical basis for the whole of the (untwisted) Fock space, and this in principle gives an algorithm for computing the canonical basis of \(M^s\). However, in practice this algorithm is extremely slow. We give a much faster algorithm here; the way we do this is to compute the canonical basis for an intermediate module \(M^s\), which is defined to be the tensor product of level 1 highest-weight irreducibles. It is then straightforward to discard unwanted vectors to get the canonical basis for \(M^s\). The author remarks that \textit{N. Jacon} [J. Algebra 292, No. 1, 100--109 (2005; Zbl 1104.20017)] and \textit{X. Yvonne} [J. Algebra 309, No. 2, 760--785 (2007; Zbl 1172.17300)] have also given algorithms for computing higher level canonical bases. However, Yvonne's algorithm is very slow, since it computes the canonical basis for the whole of the Fock space, while Jacon's algorithm works in a particular type of twisted Fock space, whereas the author's algorithm remains in the more natural setting of the untwisted Fock space; although these Fock spaces are isomorphic, so that in principle one canonical basis determines the other, in practice it is very difficult to give an explicit isomorphism.