Extending Peterson’s picture: a mirror approach to arbitrary Schubert varieties
This paper offers a new way to extend a classic idea in algebraic geometry and mirror symmetry. The authors generalize Dale Peterson’s construction — originally tied to whole flag varieties — to arbitrary Schubert varieties. Their main objects are a “generalised Peterson variety,” which can be a possibly non-reduced affine scheme sitting inside a Langlands-dual full flag variety, and a Lie-theoretic “superpotential,” a holomorphic function whose relative critical points recover that generalised Peterson variety.
To make this concrete the authors build several ingredients. They isolate two tori (the Picard torus and a divisor-class torus) attached to a given Schubert variety. They then define an open generalised Peterson variety inside a natural open Richardson cell by imposing vanishing conditions on certain weight-space projections of a fixed nilpotent element. They also define a Picard–Peterson (or Peterson–Toeplitz) variety and write down explicit holomorphic functions (their superpotentials). A key technical result in the paper is that the relative critical locus of the superpotential matches the generalised Peterson variety, and various maps between the schemes involved are shown to be isomorphisms in the factorial case.
Why this matters: one goal is to connect these geometric objects to quantum cohomology. Quantum cohomology is a version of the usual cohomology ring of a space that is deformed using counts of certain geometric curves (Gromov–Witten invariants). The authors make a central conjecture: when the Schubert variety is smooth and Fano (a class of varieties often considered in mirror symmetry), the coordinate ring of the generalised Peterson variety should recover the quantum cohomology ring after localizing the quantum parameters. In a further conjecture for a subclass of Fano cases (when the Picard number equals a certain rank), they produce a map to the cotangent bundle of a torus and a partial compactification that they expect to model the full, non-localized quantum cohomology. This is intended to mirror known presentations such as the Givental–Kim description via a degenerate leaf of the Kostant–Toda lattice.