The paper tackles two related problems in algebraic geometry, specifically about how functions and geometric objects behave near singular points, but working over fields where arithmetic wraps around after a prime number p (called positive characteristic). The first problem concerns "determinacy": when can a complicated function near a singular point be fully captured by just its first few Taylor series terms? The authors show that for matrix-valued functions, a geometric condition called finite codimension is exactly equivalent to being able to truncate the Taylor series at some finite level and lose no essential information. This works over essentially any field, with careful handling needed for finite fields. As a bonus, their methods reveal useful structural properties of certain algebraic objects called determinantal ideals.
The second problem concerns "modality": how many continuous parameters are needed to describe all nearby singular points of a similar type? In standard calculus over the real or complex numbers, there are two natural notions of modality depending on which transformations you allow, and it was known they can differ. The authors introduce a new algebraic tool called a Frobenius algebra, which exploits a special feature unique to positive characteristic arithmetic (the Frobenius map, which raises elements to the p-th power). Using this algebra and a construction called a finite flat critical cover, they prove that two competing notions of modality actually agree in positive characteristic. This has a satisfying consequence: if singularities have bounded modality, their complexity can always be tamed by a finite family of algebraic models, independent of the ambient dimension.
Running through both parts of the paper are these Frobenius-based techniques, which turn out to be surprisingly powerful tools for measuring singularities. The same methods pin down finer numerical invariants called F-jumping numbers and the F-pure threshold, which are positive-characteristic analogs of invariants used to measure how bad a singularity is in classical geometry. A particularly clean result is that the F-pure threshold, which in principle requires checking infinitely many levels of Frobenius, is actually already determined at a single level. Together, the results show that working in positive characteristic, far from being an obstacle, provides extra algebraic structure that makes several classical problems more tractable.