On large theta-characteristics with prescribed vanishing (Q2420261)

David Mumford and Joe Harris begun the study of various algebro-geometric aspects of theta characteristics. Then, the field was established as an active field of study by various authors. The paper under review, being a prolongation to one of the authors' previous studies (e.g. [the third author, Math. Ann. 363, No. 1--2, 385--392 (2015; Zbl 1329.14059); ``Components of moduli spaces of spin curves with the expected codimension II'', Preprint, \url{arXiv:1502.05262}]); adds considerable and interesting ingredients to the literature of the field. Upon their main basic results, the moduli of curves \([C]\) admitting theta characteristics \(L\) with prescribed vanishing and multiplicities and having specific number of linearly independent global sections, if non empty, is of codimension at most \(g-1+\frac{r(r-1)}{2}\), where \(h^0(L)\geq r+1\) and \(h^0(L)\equiv r+1\pmod 2\). As their second main result, they prove that for an explicitly determined integer number, depending only on \(r\), the pointed locus inside the moduli space of pointed curves of genus \(g\), is non-empty with an irreducible component of expected dimension. The first result involves a smaller of the paper. They apply a relative study of symmetric product of curves with large theta characteristics together with basic facts about the dimension of the spaces involved in a morphism. Then, they apply Harris' bound on \(\mathcal{M}^r_g\) which is the locus of curves admitting theta characteristics \(L\) with \(h^0(L)\geq r+1\) and \(h^0(L)\equiv r+1\pmod 2\). As for their second main result; they prove: Theorem 1.2. For any genus \(g\geq g(r ),\) and for any partition \(\underline{k} = (k_1, \dots , k_n)\) of \(g-1\), the locus \(\mathcal{G}^r_g (\underline{k})\) is non-empty, and there exists an irreducible component \(\mathcal{Z}\subset \mathcal{G}^r_g (\underline{k})\) having expected dimension. In particular, at a general point \([C, p_1,\dots,p_n] \in \mathcal{Z}\), the large theta characteristic \(\mathcal{O}_C(\sum^n_{i=1} k_i p_i)\) possesses exactly \(r + 1\) independent global sections and, apart from the cases \((r , g) = (0, 2)\) and \((1, 3)\), the curve \(C\) is non-hyperelliptic. In the small genus cases \(g=2, 3\), they use simple combinatorics of their locus \(\mathcal{G}^r_g(\underline{k})\) together with an ad-hoc argument to conclude. For other values of \(g\), they first reduce the assertion to the case of \(\mathcal{G}^r_g(g-1)\). In order to do so, for a fixed genus \(g\), they use the partial order on the set of partitions of \(g-1\) to prove that if the subcanonical locus \(\mathcal{G}^r_g(g-1)\) admits an irreducible component of expected dimension with general point \([C, p_1, \cdots , p_m]\) consisting a non-hyper elliptic curve \(C\) with \(h^0(C, \mathcal{O}_C(\sum h_ip_i))=r+1\), then \(\mathcal{G}^r_g(\underline{k})\) would have the same properties for any partition \(\underline{k}\) of \(g-1\). If for an integer \(g(r)\) the locus \(\mathcal{G}^r_{g(r)}\) contains an irreducible component \(\mathcal{Z}_{g(r)}\) having expected dimension with general \([C, p]\) satisfying \(h^0(C, \mathcal{O}_C(g(r)-1)p)=r+1\), then for any \(g\geq g(r)\) the locus \(\mathcal{G}^r_g\) has the same properties, as well. Therefore, summarizing; what the authors have obtained along section \(2\) is that, in order to establish their Theorem 1. 2, it suffices to reduce to the case \(\underline{k}=(g-1)\). Section \(3\) of the paper is dedicated to establish Theorem 1. 2 in the case \(\underline{k}=(g-1)\). The strategy is to find a locus inside \(\mathrm{Hilb}^r_{g, g-1}\), the Hilbert scheme of reduced curves of arithmetic genus \(g\) and degree \(g-1\) with at most nodes as singularities. So, they need to extend the notion of sub-canonical points to reduced curves, upon which they introduce the ``limit sub-canonical points''. The authors then construct reducible curves admitting limit sub-canonical points deforming to smooth curves with sub-canonical points. Through Remark 3.5 and Remark 3.6, they explain interesting properties of their constructed curves. Theorem 3.10 almost completes the proof. In order to establish this via the pattern created inside \(\mathrm{Hilb}^r_{g, g-1}\), it would suffice to prove that for \(r\geq 2\) and for any hyperplane \(M\), there exists an irreducible component \(Z_{g(r)}\subset Q^r_{g(r), sm}(M)\) such that \(\dim Z_{g(r )} = 2g(r )-2-\frac{r(r-1)}{2}+(r + 1)^2-r \), and the general point \([C] \in Z_{g(r )}\) parameterizes a non-degenerate linearly normal curve \(C \subset \mathbb{P}^r\) with \(h^1(C, N_C/\mathbb{P}^r) = 0\), where \(Q^r_{g(r), sm}(M)\) is the locus of smooth curves \(C\subset \mathbb{P}^r\) such that \(M\) cuts a sub-canonical point \(p\) on \(C\). For \(r=2\), the authors use an ad-hoc argument to establish this assertion and Lemma 3.8 furnishes an inductive setting to prove the statement for arbitrary \(r\). They, then, map the locus \(Z_{g(r )}\) to \(\mathcal{G}^r_{g(r)}\), where they obtain an irreducible component \(\mathcal{Z}^r_{g(r)}\subset \mathcal{G}^r_{g(r)}\) having the desired properties. The paper is accompanied by several interesting examples to explain the authors ideas, specifically for small \(g\) and \(r\).