The log product formula (Q6101526)

The paper under review is an important contribution to the subject of logarithmic Gromov-Witten theory. The goal in this subject is to study curves in a pair \((X,D)\) consisting of a smooth algebraic variety and a simple normal crossings divisor. The idealized object in logarithmic Gromov-Witten theory can be imagined of as a non-compact curve \(C^\circ\) in the complement \(X\setminus D\), with prescribed ``asymptotics''. Precisely, one studies maps of pairs from pointed curves \((C,p_1,\ldots,p_n)\to (X,D)\) where the tangency orders at the different \(p_i\) along the different components of \(D\) are fixed. There is a significant increase in complexity in theory when going from the case where \(D\) is smooth to the general case when \(D\) is simple normal crossings. Consequently, there is practical and theoretical value in studying pairs of ``product'' type: \(X = Y\times Z\), where \(D\) is the union of a pullback of a smooth divisor in each of \(Y\) and \(Z\). The main result of this paper proves that in this product setting, the appropriate moduli spaces of logarithmic stable maps to \(X\), together with its virtual class, can be recovered from maps to the components \(Y\) and \(Z\). It should be noted that the most naive statement would be that \(M(X)\) is well-approximated by \(M(Y)\times_M M(Z)\), where \(M\) is the moduli space of curves and the rest are logarithmic mapping spaces. However, this statement is \emph{false} both at the level of spaces and at the level of their virtual fundamental classes. The failure comes from the logarithmic structure. If \(Y\) and \(Z\) have the trivial logarithmic structure, the statement becomes true and this was established in Behrend in the late 1990s. In order to rememdy this, the author develops a new type of intersection theory for logarithmic schemes, based on the notion of a logarithmic normal cone and uses this to equip the spaces in question with enhancements of standard intersection theory operations. The author then shows that the naive statement can be salvaged, provided one works in the context of logarithmic intersection theory. A significant part of the paper is dedicated to the development of logarithmic intersection theory, normal cones, Gysin maps, and so on. The results have potential to be useful beyond Gromov-Witten theory.