The paper studies certain geometric objects called torus-orbit closures, which arise when a high-dimensional torus (think of a multi-dimensional version of a circle) acts on a space called the complete flag variety. The flag variety is a fundamental object in algebra and geometry that encodes how nested sequences of vector spaces can sit inside one another. The specific orbit closures the authors focus on come from a structure called a cluster fan, which is a combinatorial object that organizes certain algebraic data into cone-shaped regions. Understanding the geometry of these orbit closures, particularly their sizes and shapes as measured by something called equivariant homology, is the central goal.
To compute these geometric measurements, the authors develop a set of combinatorial tools called long-range divided difference operators, which are algebraic operations that can be read off from certain tree-like diagrams called noncrossing alternating forests. These forests provide a systematic way to break complicated geometric calculations into smaller, manageable steps. The key achievement is expressing the homology classes of these orbit closures as combinations of simpler building blocks called Schubert cycles, where all the coefficients are guaranteed to be positive in a precise technical sense called Graham-positivity. Positivity results like this are highly desirable because they suggest there is a deeper combinatorial or geometric reason behind the formulas, rather than them being accidents of calculation.
As concrete payoffs, the authors recover and generalize several known results in the field. They give new positive formulas for all torus-invariant curves in the flag variety, for generic orbit closures, and for a family of geometric objects called Richardson varieties associated to particular intervals in a combinatorial ordering called the Bruhat order. They also translate their results to a simpler related space called the Grassmannian, where their orbit closures correspond to objects from matroid theory, specifically lattice path matroids on rearranged ground sets. The paper thus bridges algebraic geometry, combinatorics, and matroid theory through a unified framework.