The incompressible Euler equations describe idealized fluid flow, but finding solutions is notoriously hard. A common strategy involves constructing approximate solutions called "subsolutions," which satisfy a relaxed version of the equations where an extra term, the Reynolds stress, absorbs the error. Researchers want to understand what happens in the limit as these approximations get closer and closer to true solutions. The trouble is that the collection of all such limit objects does not form a smooth geometric space in any traditional sense, which makes it difficult to talk about small perturbations or directions of variation near a given limit point. This paper builds a rigorous geometric framework, using a flexible approach called diffeology, to handle exactly this situation. The authors treat the fluid velocity, its quadratic flux, and the Reynolds stress as independent variables and study the space formed by taking weak limits of smooth approximate solutions.
A central contribution is making precise what it means to "perturb" one of these limit objects infinitesimally. The authors construct tangent spaces intrinsic to the limit space and show that these tangent directions satisfy the linearized version of the governing equations in a generalized sense called distributions, even though the surrounding space lacks the usual smooth structure required for calculus. They also identify which tangent directions are consistent with being a true Euler solution rather than just an approximation, and they separate perturbations that can actually be detected by measuring velocity or flux from "hidden" directions that only affect the Reynolds stress without any observable consequence.
Finally, the paper introduces a concrete model built from finite mixtures, meaning states formed by combining a small number of simpler fluid configurations with different weights. The derivative of such a mixture construction produces explicit tangent vectors and connects the Reynolds stress to what happens when a single fluid phase begins to split into multiple distinct phases. Under a mild generality assumption, the authors show that every possible stress configuration can be realized this way to first order. This yields counting results about how many phases are needed and identifies a minimal hierarchy of observable quantities, giving a practical geometric picture of how the complexity of turbulent-like fluid behavior emerges from these approximate solution spaces.