Massey products and Fujita decompositions on fibrations of curves (Q2301243): Difference between revisions

Consider a fibration \(f:S\rightarrow B\) from a smooth complex projective surface to a smooth curve. The second Fujita's decomposition of the Hodge bundle is \( f_*\omega_f= \mathcal{ U} \oplus \mathcal{ A}\) where \(A\) is ample and \(\mathcal{ U} \) is a unitary flat bundle, see [\textit{F. Catanese} and \textit{M. Dettweiler}, C. R., Math., Acad. Sci. Paris 352, No. 3, 241--244 (2014; Zbl 1310.14017)]. Those authors proved that \(\mathcal{ U} \) is quasiample iff the monodromy of the associated local system is finite. They provided examples when monodromy is infinite. The paper under review introduces and studies certain flat subbundles \(\mathcal {M} \) of \(\mathcal{ U}\), one main result is the finiteness of their monodromy. The bundles in question are built by means of a condition that roughly requires that some Massey-like products all vanish, the unfamiliar reader can consult the article's section \(3\), where the construction of families of Massey-like products is motivated and explained. The non vanishing of a Massey product implies that the infinitesimal invariant of the normal function of the Ceresa cycle is non trivial. Now if the monodromy of \(\mathcal{ U} \) is not finite then \(\mathcal{ U} \) cannot be of Massey type, so that there must be a nonvanishing Massey product, hence the normal function associated with the Ceresa cycle on the general fibre is not trivial. This is then the case for the examples of Catanese and Dettweiler. Conversely when the normal function is trivial then \(\mathcal{ U}\) is of Massey type, hence of finite monodromy. This is the case for families of hyperelliptic curves. The heart of the paper is the construction of the Massey bundles and the investigation of their monodromy. The proof of finiteness rests on a remarkable circle of ideas, one such being the use of a new non-classical version of the theorem of Castelnuovo-de Franchis.