Abstract. We want to associate to an n-vector on a manifold of dimension n a cohomology which generalizes the Poisson cohomology of a 2-dimensional Poisson manifold. Two possibilities are given here. One of them, the Nambu-Poisson cohomology, seems to be the most pertinent. We study these two cohomologies locally, in the case of germs of n-vectors on K n (K = R or C).2000 Mathematics Subject Classification. 53D17.
Introduction.A way to study a geometrical object is to associate to it a cohomology. In this paper, we focus on the n-vectors on an n-dimensional manifold M.If n = 2, the 2-vectors on M are the Poisson structures thus, we can consider the Poisson cohomology. In dimension 2, this cohomology has three spaces. The first one, H 0 , is the space of functions whose Hamiltonian vector field is zero (Casimir functions). The second one, H 1 , is the quotient of the space of infinitesimal automorphisms (or Poisson vector fields) by the subspace of Hamiltonian vector fields. The last one, H 2 , describes the deformations of the Poisson structure. In a previous paper (see [9]) we have computed the cohomology of germs at 0 of Poisson structures on K 2 (K = R or C).In order to generalize this cohomology to the n-dimensional case (n ≥ 3), we can follow the same reasoning. These spaces are not necessarily of finite dimension and it is not always easy to describe them precisely.Recently, a team of Spanish researchers has defined a cohomology, called NambuPoisson cohomology, for the Nambu-Poisson structures (see [6]). In this paper, we adapt their construction to our particular case. We will see that this cohomology generalizes in a certain sense the Poisson cohomology in dimension 2. Then we compute locally this cohomology for germs at 0 of n-vectors, with the assumption that f is a quasihomogeneous polynomial of finite codimension ("most of" the germs of n-vectors have this form). This computation is based on a preliminary result that we have shown, in the formal case and in the analytical case (so, the Ꮿ ∞ case is not entirely solved). The techniques we use in this paper are quite the same as in [9].