Linear stability of projected canonical curves with applications to the slope of fibred surfaces (Q2478647): Difference between revisions

The paper deals with fibrations \(f:S\to B\), where \(S\) is a complex projective surface and \(B\) is a complex projective curve. \(f\) is assumed to be onto and \(S\) relatively minimal (there are no \(-1\) curves in fibers of \(f\)). The author focus on the case, when \(f\) is not locally trivial. In this case they define (after Xiao) the slope of \(f\) as \(s_f=\frac{K_f^2}{\chi_f}\). Here \(K_f=K_S-f^*K_B\) and \(\chi_f=\deg f_*\omega_{S/B}\). By Noether theorem \(s_f\) takes values between \(0\) and \(12\). The aim of this paper is to find a lower bound on \(s_f\). The authors state a conjecture that \(s_f\geq 4\frac{g-1}{g-q_f}\), where \(g\) is the genus of a generic fiber and \(q_f\) is the so-called relative irregularity of \(f\). Then they present instances when this conjecture holds and admit that there are some counterexamples known. In the subsequent parts of the paper the authors prove a weaker bound on \(s_f\), namely \(s_f\geq 4\frac{g-1}{g-\lfloor m/2\rfloor}\). Here \(m\) is the maximum of the irregularity \(q_f\) and the Clifford index \(\text{Cliff}(f)\). The most interesting fact about this bound is that \(q_f\) and \(\text{Cliff}(f)\) are apparently unrelated. The authors provide two different proofs of this bound: first one uses the Harder-Narashiman filtration of some subsheaf of \(f_*\omega_f\). This method was first introduced by Xiao. The other one is the application of the Cornalba--Harris method. The paper ends with some examples of mappings showing that \(q_f\) and \(\text{Cliff}(f)\) are in fact unrelated. Namely \(q_f\) may be large and \(\text{Cliff}(f)\) small and conversely.