The paper addresses a system called "morphological calculus," which is a way of describing the shape and structure of geometric objects using polynomial-like expressions. Think of it as a shorthand for encoding how a space is built from simpler pieces, like edges, faces, and higher-dimensional cells. The problem is that this shorthand is too lossy: different geometric objects can end up with the same expression, and the notation throws away crucial information about how the pieces connect to each other. The authors fix this by replacing the polynomial notation with a richer mathematical structure called a chain complex, which keeps track of not just how many pieces there are at each dimension, but also how they fit together at their boundaries.
With this improved framework, the authors can cleanly distinguish between different ways geometric spaces can be assembled. For example, a product of two spaces (like a cylinder formed from a circle times an interval) looks algebraically similar to a fiber bundle (like a twisted version of that cylinder), but the two behave differently in ways the old calculus could not capture. The authors introduce tools to measure this difference precisely, including something called a monodromy defect, which quantifies how much a bundle "twists" compared to a plain product. They also develop a diagrammatic version of the construction that handles spaces built by gluing several pieces together in complicated ways.
As a concrete application, the authors revisit some specific calculations that had unresolved discrepancies in the original morphological calculus work. These involve spaces called bivector spheres, which arise naturally in four and five dimensions from the geometry of rotations. In four dimensions, the relevant space turns out to be a join of two ordinary spheres, and the discrepancy in the old calculation is explained by an extra piece inserted at the midpoint of that join. In five dimensions, a symmetry group acts on the space and breaks it into orbits of different types; carefully accounting for all these orbits using the new chain-level framework resolves the remaining numerical mismatches, with seven specific pairs of cells identified as the source of the previous errors.