The Kodaira dimension of the moduli space of curves of genus 15 (Q1105653)

The purpose of this paper is to prove that the moduli space \({\mathcal M}_{15}\) of curves of genus 15 has Kodaira dimension \(\kappa =-\infty\), i.e. that \(H^ 0({\mathcal M}_{15},nK_{{\mathcal M}_{15}})=0\) for all \(n>0\). Our proof is based on an analysis of a particular divisor \(D\subset \bar {\mathcal M}_{15}\), namely \(D=\) some component of the locus of curves carrying a \(g^ 3_{14}\). We show that D is unirational. Moreover, for some rational curve \(F\subset D\) which is a member of a family of rational curves ``filling up'' \(D\), we show that the intersection numbers \((*)\;D\cdot F>0,\) while \(F\cdot K_{{\mathcal M}_{15}}<0.\) As is easily seen, this implies \(\kappa({\mathcal M}_{15})=-\infty\). Roughly speaking, our proof goes as follows. We start with a rather common and explicit sort of pencil \(F_ 0\) of curves in \({\mathcal H}\), which itself, like - it would seem - any pencil one can construct ``explicitly'', could not possibly have properties (*) above \(({\mathcal H}\) is the closure of some component of the Hilbert scheme of nonsingular curves of genus 15 and degree \(14\) in \({\mathbb{P}}^ 3)\). We then lift \(F_ 0\) to a pencil of matrices in \({\mathcal N} = space\) of special \(4\times 9\) matrices. Since \({\mathcal N}\) is just a linear space, the latter pencil can be moved around freely, and we then map it over to \({\mathcal M}_{15}\), which yields a rational curve F as above. Using formulas of \textit{D. Eisenbud} and \textit{J. Harris} [Invent. Math. 90, 359- 387 (1987; Zbl 0631.14023)] and \textit{D. Mumford} [Enseign. Math., II. Sér. 23, 39-110 (1977; Zbl 0363.14003)] for the classes of D and \(K_{{\mathcal M}_{15}}\) in Pic(\({\mathcal M}_{15})\), respectively, we establish (*) above. We emphasize that F is not the image of \(F_ 0.\) Note that, since our methods apply only to curves Y in \({\mathbb{P}}^ 3\) with \(H^ 1({\mathcal J}_ Y(5))=0\), it follows that genus 15 is the only genus \(g\geq 14\) for which curves of this type fill up at least a divisor in \({\mathcal M}_{15}\). - We note finally that the curve F above comes very ``close'' to making \({\mathcal M}_{15}\) uniruled, but does not quite make it. On the other hand, it is apparently conjectured in general that \(\kappa =-\infty\) implies uniruledness. Thus \({\mathcal M}_{15}\) seems like a good test case for this conjecture.