Ah! si loin des carquois, des torches et des flèches, On se sauvait un peu vers des choses... plus fraîches ! E. Rostand, Cyrano de Bergerac, Acte III, Scène 7 1. Kac polynomials for quivers and curves 1.1. Quivers. Let Q be a locally finite quiver with vertex set I and edge set Ω. For any finite field F q and any dimension vector d ∈ N I , let A Q,d (F q ) be the number of absolutely indecomposable representations of Q over F q , of dimension d. Kac proved the following beautiful result :Here , is the Euler form, see Section 2.1. The fact that the number of (absolutely) indecomposable F q -representations behaves polynomially in q is very remarkable : the absolutely indecomposable representations only form a constructible substack of the stack M Q,d of representations of Q of dimension d and we count them here up to isomorphism, i.e. without the usual orbifold measure. Let us briefly sketch the idea of a proof. Standard Galois cohomology arguments ensure that it is enough to prove that the number of indecomposable F q -representations is given by a polynomial I Q,d in q (this polynomial is not as well behaved as or as interesting as A Q,d ). Next, by the Krull-Schmidt theorem, it is enough to prove that the number of all F q -representations of dimension d is itself given by a polynomial in F q . This amounts to computing the (orbifold) volume of the inertia stack IM Q,d of M Q,d ; performing a unipotent reduction in this context, we are left to computing the volume of the stack N il Q,d parametrizing pairs (M, φ) with M a representation of dimension d and φ ∈ End(M ) being nilpotent. Finally, we use a Jordan stratification of N il Q,d and easily compute the volume of each strata in terms of the volumes of the stacks M Q,d ′ for all d ′ . This actually yields an explicit formula for A Q,d (t) (or I Q,d (t)), see [Hu]. We stress that beyond the case of a few quivers (i.e. those of finite or affine Dynkin type) it is unimaginable to classify and construct all indecomposable representations; nevertheless, the above theorem says that we can count them.The positivity of A Q,d (t) was only recently 1 proved : 1 the special case of an indivisible dimension vector was proved earlier by Crawley-Boevey and Van den Bergh,
see [CBVdB]1