Note on the linearisation of finite abelian groups

Abstract

If $K$ is a field with enough roots of unity and $V$ an abelian group, the $K$-algebra $K[V]$ of the group $V$ is split semisimple, so that the canonical morphism $K[V]\to K^{V^\sharp}$, where $V^\sharp$ denotes the dual group of $V$ (which may be seen as Hom$(V,K^\times)$), is an isomorphism of $K$-algebras. If one removes the assumption that $K$ has enough roots of unity, one can easily deduce from it (by using a base change and Krull-Schmidt) that it remains a $K$-linear isomorphism $K[V]\xrightarrow{\simeq} K^{V^\sharp}$ natural in the group $V$ if one restricts to finite groups $V$ canceled by a fixed nonzero integer. The question of whether such an isomorphism, natural in the abelian group $V$, still exists without any other restriction than $V$ is finite and its order is invertible in $K$, is less obvious; we solve it positively, in a somewhat more general setting ($K$ being any commutative ring), by using Gauss sums. We also explore some related functorial questions.