The thin Stefan problem is a mathematical model for how a thin sheet of ice melts when it sits on top of water. The interesting mathematical challenge is tracking the "free boundary," which is the moving edge that separates the part of the thin sheet that is still frozen from the part that has already melted. Points on this boundary come in two flavors: regular points, where the melting front behaves smoothly and predictably, and singular points, where the geometry becomes more complicated and the standard tools break down. Understanding singular points is both technically difficult and physically important, because they represent places where the melting process behaves in an unexpected or degenerate way.
The authors prove several results that together give a detailed picture of how bad the singular set can be. They first show that the melting front can be described locally by a single function of position, and that singular points are exactly the places where this function has a flat spot (a critical point, in the sense of calculus). They then bound the size of the singular set using a geometric measure called parabolic Hausdorff dimension, showing it is no larger than what you would expect from an n-dimensional surface evolving in time. Crucially, they prove "generic regularity" in low dimensions, meaning that if you freeze a snapshot in time, almost every moment you look at will show a completely smooth melting front, with singular points appearing only at rare instants.
To reach these conclusions, the authors combine techniques from the well-studied classical Stefan problem (melting of a three-dimensional block of ice) with several new tools tailored specifically to the thin, lower-dimensional setting. These include frequency formulas that measure how rapidly a solution oscillates near a singular point, geometric inequalities that control how quickly solutions can approach a singular configuration, and a careful hierarchical decomposition of the singular set by how degenerate each point is. Together, these ingredients allow the authors to show that outside a very small exceptional set, the singular points all lie on a smooth surface, giving a nearly complete geometric description of where and how the melting front can fail to be regular.