Meromorphic fibrations on certain manifolds with trivial canonical bundle (Q930820)

The aim of this article is to study a number of questions related to the existence of rational, non-birational, dominant endomorphisms \(f: X \dashrightarrow X\) of projective manifolds with trivial canonical divisor. If \(X\) is a \(K3\) surface that admits an elliptic fibration it is fairly easy to construct such endomorphisms, but if \(X\) is a general \(K3\) surface with Picard number one, we expect that all the endomorphisms are birational [cf. \textit{T. Dedieu}, Severi varieties and self rational maps of \(K3\) surfaces, \url{arXiv:0704.3163} (2007) for recent progress on this question]. The main strategy of the paper is to relate the existence of non-birational endomorphisms to the existence of certain rational fibrations. More precisely the authors show that if \(f: X \dashrightarrow X\) is a dominant endomorphism, there exists a \(f\)-invariant fibration \(g: X \dashrightarrow B\) such that the fibre \(X_b\) over a very general point \(b \in B\) is the Zariski closure of the orbit of a very general point in \(X_b\). The authors then apply this tool to the case of varieties with trivial canonical divisor: they show that if \(X\) has trivial canonical bundle and Picard number one, the general fibre of any rational, non-trivial fibration \(g: X \dashrightarrow B\) is a variety of general type. Since a variety of general type does not admit endomorphisms of degree greater than one, an endomorphism \(f: X \dashrightarrow X\) of degree greater than one does not preserve \(g\). In particular the orbit of a very general point \(x \in X\) is Zariski dense in \(X\). As an application one obtains that if \(X\) is defined over an uncountable field \(K\) the variety is potentially dense, that is there exists a finite field extension \(K \subset L\) such that the \(L\)-rational points are Zariski dense in \(X\). The authors also study the relationship between an endomorphism \(f: X \dashrightarrow X\) and the fibrations naturally defined on \(X\), like the rational quotient map, the Iitaka fibration or the core fibration [cf. \textit{F. Campana}, Ann. Inst. Fourier 54, No. 3, 499--630 (2004; Zbl 1062.14014)]. \textit{C. Voisin} [in: The Fano conference. Papers of the conference organized to commemorate the 50th anniversary of the death of Gino Fano (1871--1952), Torino, Italy, September 29--October 5, 2002. Torino: Università di Torino, Dipartimento di Matematica. 761--792 (2004; Zbl 1177.14040)] has shown that the Fano surface of lines \(X\) of a smooth cubic fourfold admits an endomorphism of degree 16. This variety has trivial canonical bundle and generically the Picard number is one. \textit{E. Amerik} and \textit{C. Voisin} have shown in their recent preprint [Potential density of rational points on the variety of lines of a cubic fourfold, \url{arXiv:0707.3948} (2007)] that if \(X\) is defined over a number field and sufficiently general, it is potentially dense.