Automorphisms of an irregular surface of general type acting trivially in cohomology (Q1947349)

The automorphisms of a complex smooth manifold which act trivially in cohomology have been classified for several classes of surfaces, but the problem is still open for the surfaces of general type (unless some additional hypotesis is considered). The author of this note had previously shown [Tohoku Math. J. (2) 64, No. 4, 593--605 (2012; Zbl 1259.14045)] that, if the Euler characteristic of a surface of general type \(S\) is big enough, then the group \(G\) of the automorphisms which act trivially in cohomology has order at most 4. In this note the author classifies the boundary case under the additional assumption that the irregularity is at least 3. More precisely, he shows that, if \(G\) has order 4, the irregularity is at least 3 and the geometric genus is at least 35, then the surface belongs to the family constructed in the first section of the paper. These are surfaces with irregularity 3 isogenous to a product of two curves: more precisely the quotients of the product of any hyperelliptic curve with a very special curve of genus 5 having three pairwise commuting involutions, by an involution acting separately on the two factors, and as the hyperelliptic involution on the first factor, and as one of the three given involution (without fixed points) on the second factor. The two further involutions on the second factor induce the required cohomologically trivial action of a (noncyclic) group of order 4. About the assumptions: the hypothesis on the geometric genus is technical and not expected to be sharp; the assumption on the irregularity is necessary, by some constructions in smaller irregularity in the Ph.D. thesis of Wenfei Liu. The strategy of the proof is the following. The author first uses some Hodge theory to show that under his assumptions, the canonical map of \(S\) factors through the quotient map onto \(S/G\). Then, by an argument of \textit{A. Beauville} [Manuscr. Math. 110, No. 3, 343--349 (2003; Zbl 1016.14016)], he shows that the canonical map has order at most 9, and therefore 4 or 8. He then excludes the case 4 by using the Castelnuovo inequality. So the canonical map has degree 8 and by a result of \textit{G. Xiao} [Compos. Math. 56, 251--257 (1985; Zbl 0594.14029)] the canonical image is ruled. The author concludes by showing that the pull-back of the ruling is an isotrivial fibration of genus 5, and then by using it to prove that the surface is one of surfaces described above.