ABSTRACT. We prove the following result, which is analogous to two theorems, one due to Kodaira and Krasnov and another one due to Jouanolou and Ghys. Let M be a compact complex manifold and f a dominant endomorphism of M. If there exist k totally invariant irreducible hypersurfaces W i ⊂ M, with k > dim(M) + h 1,1 (M) then f preserves a nontrivial meromorphic fibration. We then study the case where f is a meromorphic map.1. Introduction 1.1. Let M be a connected compact complex manifold of dimension n, and f : M → M be a surjective endomorphism of M. Let W ⊂ M be a hypersurface. By definition, W is totally invariant if f −1 (W ) = W. This property implies, and is stronger than, the forward invariance f (W ) = W. These notions coincide when f is an automorphism. The topological degree of f is then equal to the product of the multiplicity m and the topological degree of f |W : W → W ; the divisor defined by the jacobian determinant of f contains W with multiplicity m − 1.Theorem A. Let M be a compact complex manifold, and f be a dominant endomorphism of M. If there are k totally invariant hypersurfaces W i ⊂ M, with, then there is a non constant meromorphic function Φ and a non zero complex number α suchThis result has already been used, and proved, by the author and by Kawaguchi when M is a surface (see [6,4,14]), and by Zhang when f is an automorphism of a compact Kähler manifold with positive entropy (see [18]). We shall extend (a weak form of) Theorem A to the case of meromorphic transformations in section 3, Theorem B.
Remarks.Received by the editors May 28, 2010. M is the sheaf of holomorphic 1-forms, and H i (M, Ω 1 M ) denotes the i-th Cech cohomology group of this sheaf. By Hodge theory, these cohomology groups are finite dimensional complex vector spaces (see [12], chapter 0, section 6, page 100).
1.2.2.The proof of Theorem A provides a slightly stronger statement: The number k in inequality ( * ) can be replaced by the total number of irreducible components of the W i .
1.2.3.If g is an endomorphism of the complex projective space P n (C) with two totally invariant hypersurfaces, then there is a constant α = 0 and a non constant meromorphic function Ψ such that ψ • g = αΨ deg(g) . This conclusion is different from the invariance propertyThe coordinate axis of the plane P 2 (C) are totally invariant under the action of the endomorphism g[x : y : z] = [x 2 : y 2 : z 2 ]. In this case,P 2 (C) )) = 2 + 1 = 3 is equal to the number of totally invariant lines, and all meromorphic functions Φ such that Φ • g = αΦ for some α ∈ C are indeed constant. The same example, but in dimension n, shows that Theorem A is sharp for M = P n (C).
1.2.4.Theorem A is analogous, in a dynamical setting, to the following statement: If a compact complex manifold M has k irreducible hypersurfaces, withthen there is a non constant meromorphic function on M (see [1], section IV.6, page 129, or [10,15] and [3,16]). Another similar statement has been obtained by Jouanolou and Ghys for foliated manifolds: If a codimension one (singular...