Sheaves of maximal intersection and multiplicities of stable log maps (Q2046888)

Gromov-Witten invariants are a refined way of counting curves on a variety; similarly, logarithmic Gromov-Witten invariants are a refined way of counting (log) curves on a variety endowed with a logarithmic structure. The authors contribute to computing log GW invariants of a smooth projective surface \(X\) endowed with the log structure induced by an effective divisor \(D \subseteq X\). The most elegant special case is when \(X\) is a smooth del Pezzo surface, and \(D \subseteq X\) is a smooth anticanonical divisor. The basic study of this situation is given by the same authors in [\textit{J. Choi} et al., Trans. Am. Math. Soc. 374, No. 1, 687--732 (2021; Zbl 1465.14052)]; the article under review offers an interesting intermediate step excluded in the basic study, additional results, and generalizations to other situations, in particular blowups of toric surfaces, whose (relative) GW invariants have been studied using the tropical vertex group in [\textit{M. Gross} et al., Duke Math. J. 153, No. 2, 297--362 (2010; Zbl 1205.14069)]. To explain the special situation, let \((X,D)\) be a del Pezzo surface with a smooth anticanonical divisor. We also fix a curve class \(\beta \in H_2(X,\mathbb{Z})\), for which we assume that \(w := \beta . D > 0\). As an analog of the moduli of stable maps in classical GW invariants, we have the moduli stack \(\overline M_\beta(X,D)\) of basic stable log maps from \(1\)-marked log curves of genus \(0\) to \((X,D)\) which are of class \(\beta\) and have maximal tangency with \(D\) in the image of the marked point. Considered as an ordinary stack over the category of schemes, this is a proper DM stack and admits a virtual fundamental class in degree \(0\), i.e., we have a single GW invariant \(\mathcal{N}_\beta(X,D) \in \mathbb{Q}\). The image in \(D\) of the marked point of the curve is called the point of contact; the possible points of contact form a finite set \(D(\beta)\), and the moduli stack decomposes into a disjoint union of components \(\overline M^P_\beta(X,D)\) corresponding to the points \(P \in D(\beta)\). Each component gives rise to the finer invariant \(\mathcal{N}^P_\beta(X,D) \in \mathbb{Q}\). The component \(\overline M^P_\beta(X,D)\) further decomposes according to the underlying (classical) stable map, which is essentially a union of rational curves in \(X\) intersecting \(D\) only in \(P\). Moreover, only finitely many rational curves can occur at each \(P\). Often the part of \(\overline M^P_\beta(X,D)\) corresponding to a union of rational curves consists of finitely many isolated fat points, and the virtual fundamental class coincides with the actual fundamental class. In this case, the contribution to \(\mathcal{N}^P_\beta(X,D)\) is the sum of the lengths of the points, i.e., of their multiplicities mentioned in the title. The authors identify this contribution in two situations: \begin{itemize} \item for a single rational curve; \item for the union of two rational curves (in general position). \end{itemize} The identified contributions to \(\mathcal{N}^P_\beta(X,D)\) are positive integers, but the rational number \(\mathcal{N}^P_\beta(X,D)\) is, in general, not expected to be an integer. In the first main part of the article, the authors introduce sheaves of maximal intersection and study their moduli; the results of this are employed to determine the contribution of single rational curves. In the second main part of the article, the authors calculate the contribution of the union of two rational curves.