The Coxeter element and the branching law for the finite subgroups of SU(2)

Abstract

Let $\Gamma$ be a finite subgroup of SU(2) and let $\widetilde {\Gamma} = \{\gamma_i\mid i\in J\}$ be the unitary dual of $\Gamma$. The unitary dual of SU(2) may be written $\{\pi_n\mid n\in \Bbb Z_+\}$ where $dim \pi_n = n+1$. For $n\in \Bbb Z_+$ and $j\in J$ let $m_{n,j}$ be the multiplicity of $\gamma_j$ in $\pi_n|\Gamma$. Then we collect this branching data in the formal power series, $m(t)_j = \sum_{n=0}^{\infty}m_{n,j} t^n$. One shows that there exists a polynomial $z(t)_j$ and known positive integers $a,b$ (independent of $j$) such that $m(t)_j = {z(t)_j \over (1-t^a)(1-t^b)}$. The problem is the determination of the polynomial $z(t)_j$. If $o\in J$ is such that $\gamma_o$ is the trivial representation, then it is classical that $z(t)_o = 1 +t^h$ for a known integer $h$. The problem reduces to case where $\gamma_j$ is nontrivial. The McKay correspondence associates to $\Gamma$ a complex simple Lie algebra $\g$ of type A-D-E. We explicitly determine $z(t)_j$ for $j\in J-\{o\}$ using the orbits of a Coxeter element on the set of roots of $\frak{g}$. Mysteriously the polynomial $z(t)_j$ has arisen in a completely different context in some papers of Lusztig. Also Rossmann has recently shown that the polynomial $z(t)_j$ yields the character of $\gamma_j$.