The paper investigates how smooth the solutions to a particular mathematical equation can be. The equation in question connects the geometry of a curved surface (specifically, something called the determinant of the second derivative matrix, which measures how a surface curves in multiple directions simultaneously) to a function of the solution itself. The authors focus on two-dimensional settings and work with solutions that are convex, meaning they curve upward like a bowl. Without assuming anything special about the right-hand side function, they prove that solutions are as smooth as "C^1,1," which roughly means the solution and its first derivatives exist and are well-behaved, though the second derivatives might just barely fail to be continuous. They also show that near points where the gradient vanishes (called critical points), the level curves of the solution look increasingly like ellipses as you zoom in.
A key technical innovation is the discovery of a new quantity that decreases monotonically along a certain geometric flow. This flow tracks how the "support functions" of the level sets of the solution evolve, and the fact that this quantity only ever decreases gives the authors powerful control over the solution's behavior. This kind of monotonicity tool is valuable because it converts a complicated nonlinear problem into something more tractable, allowing the authors to extract precise geometric and analytic information about the solution.
Beyond the main regularity result, the work has two notable applications. First, it yields generalized Liouville theorems, which are results that classify all globally defined solutions of a given equation. These extend a classical 1954 result by Jorgens and also fully classify solutions to a related obstacle problem previously studied by other researchers. Second, the methods apply to meteorology: the semigeostrophic equations, which model large-scale atmospheric flow, admit special solutions called rotating waves, and the authors establish both how smooth these solutions are and how rigid their structure is, meaning the solutions cannot deviate much from a standard form.