The paper explores connections between different ways of counting geometric objects called curves and sheaves (which are algebraic structures encoding how functions behave on spaces) in the context of a special type of high-dimensional geometric space called a Calabi-Yau fivefold. These counting problems, known as enumerative invariants, arise in mathematics and physics, particularly in string theory and its conjectured eleven-dimensional cousin called M-theory. The central idea is that several seemingly different counting problems should secretly be related to one another, and the paper investigates these relationships both through rigorous proof and numerical experimentation.
The key players are three types of invariants: Gromov-Witten invariants (which count maps of surfaces called membranes into the fivefold), Pandharipande-Thomas invariants (which count curve-like structures defined using sheaves on a lower-dimensional threefold sitting inside the fivefold), and so-called membrane indices, which act as an intermediary between these two and generalize a well-known concept from string theory called Gopakumar-Vafa invariants. The authors prove the equivalence between these counting theories in certain tractable cases called strip geometries, and provide strong numerical evidence for the correspondence in more complicated settings including local surfaces and local projective spaces, for which they also propose new conjectural formulas.
In the final part of the paper, the authors prove a result called geometric engineering for a specific class of fivefolds built as products of simple spaces. This connects the sheaf-counting invariants to instanton partition functions, which are quantities originally arising in quantum field theory and computed using very different methods. Establishing this bridge between geometric counting problems and physics-inspired partition functions is a significant step toward placing M-theory on a more rigorous mathematical footing, and the paper's mix of proofs and numerical tests lays out a detailed roadmap for future work.