GW invariants and invariant quotients (Q1601062)

Gromov-Witten invariants have their origins in physics, namely in the topological sigma model coupled to gravity. There the \(A\)-model correlation function on a Calabi-Yau threefold is expressed in terms of certain invariants, which naively count the number of rational curves in the Calabi-Yau threefold that represent distinguished homology classes. Alas, in order to give a rigorous mathematical definition for those naively stipulated Gromov-Witten invariants, considerable geometric sophistication is required. There are several approaches to establishing a suitable ``Gromov-Witten theory'', using both algebraic geometry and symplectic geometry, and all of them have proved to be extremely useful in complex enumerative geometry by itself. However, it is usually quite difficult to explicitly compute these invariants, and therefore it is useful to better understand their behavior under certain modifications of the underlying manifold. In the paper under review, the author studies the behavior of the algebraically defined Gromov-Witten invariants (à la Li-Tian, Behrend-Fantechi, or Fulton-Pandharipande) with respect to group actions on the base variety. More precisely, the starting point of his investigations was the following problem: Given a projective variety acted on by a group, is there any fundamental relationship between the Gromov-Witten invariants of this variety and those of its quotient with respect to the groud action? As it looks rather hopeless, at this stage of the development, to tackle this problem in its full generality, the author turns to the special case of a polarized projective variety and a group action that is free on the semi-stable locus of the variety. Although this assumption makes the according geometric invariant theory trivial, it still remains natural and significant enough. In this context, the author's first main result provides a partial answer to the initial problem, in that it gives an explicit comparison formula for the genus-zero Gromov-Witten invariants of the variety \(X\) and its geometric quotient \(X//G\). The second part of the article is a contribution towards a deeper understanding, from an algebraic point of view of the so-called Hamilton invariants defined by \textit{K. Cieliebak}, \textit{A. R. Gaio} and \textit{D. Salamon} [``Curves, moment maps, and invariants of Hamiltonian group actions''], in the special case of torus actions. It is shown that those Hamiltonian invariants, which were constructed using real-analytic methods and symplectic techniques, fit perfectly well into the algebraic framework developed in the first part of the present paper. Actually, it is proved that the real-analytic and the algebro-geometric points of view are equivalent. Using this, the author proves a conjecture formulated by Cieliebak-Gaio-Salamon (cited above), which states that for certain choices of the parameters, these Hamiltonian invariants of \(X\) actually coincide with the (algebraic) Gromov-Witten invariants of the quotient \(X//G\). Alltogether, this work is very enlightening and sheds some more light on the relationship between differently defined invariants in quantum cohomology.