Generic regularity in time for solutions of the Stefan problem in 4+1 dimensions
Giacomo Colombo · 2025
Read Generic regularity in time for solutions of the Stefan problem in 4+1 dimensions by Giacomo Colombo (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-11876.txt.
Introduction
Giacomo Colombo proves smoothness results for the Stefan problem, a classic model of melting and freezing. The paper appeared in two thousand twenty-five. The model tracks a moving boundary between phases, such as ice and water. That free boundary can have singular points where it kinks. Colombo asks how bad those points can be over time. The setting here is four space dimensions plus time.
He shows the boundary is a smooth three-dimensional surface at almost every time. The singular set splits into two parts with distinct shapes. One part lies on a single smooth surface whose bad times form a tiny set. The other part is thin in a parabolic counting sense. The proof extends known results to this higher dimension. It sharpens the picture of phase change geometry.
Catalog record
| CREATOR | Giacomo Colombo |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 77 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2504-11876
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Generic regularity in time for solutions of the Stefan problem in 4+1 dimensions Giacomo Colombo Abstract . We show that the free boundary of a solution of the Stefan problem in ℝ 4 + 1 \mathbb{R}^{4+1} is a 3 3 -dimensional manifold of class C ∞ C^{\infty} in \R 4 \R^{4} for almost every time. This is achieved by showing that for all dimensions n n the singular set Σ ⊂ \R n + 1 \Sigma\subset\R^{n+1} can be decomposed in two parts Σ = Σ ∞ ∪ Σ ∗ \Sigma=\Sigma^{\infty}\cup\Sigma^{*} , where Σ ∞ \Sigma^{\infty} is covered by one ( n − 1 ) (n-1) -dimensional manifold of class C ∞ C^{\infty} in \R n + 1 \R^{n+1} and its projection onto the time axis has Hausdorff dimension 0, while Σ ∗ \Sigma^{*} is parabolically countably ( n − 2 ) (n-2) -rectifiable. 1. Introduction The Stefan problem [ Ste90 ] describes the evolution of a block of ice melting in water. More precisely, the temperature \theta\geq 0 satisfies \partial_{t}\theta=\Delta\theta in \{\theta>0\} (the region occupied by water) with the additional boundary condition \theta_{t}=|\nabla_{x}\theta|^{2} on \partial\{\theta>0\} (the interface ice/water). Assuming non-zero speed of the free boundary \partial\{\theta>0\} , the Stefan problem is locally equivalent through the change of variables u(x,t)=\int_{0}^{t}\theta(x,s)ds (see [ Duv73 ] ) to the parabolic obstacle problem (1) \begin{cases}(\Delta-\partial_{t})u=\chi_{\{u>0\}}&\text{in }\Omega\times[0,T],\\ u\geq 0,\partial_{t}u\geq 0&\text{in }\Omega\times[0,T],\\ \partial_{t}u>0&\text{in }\{u>0\},\end{cases}\quad\Omega\times[0,T]\subset\R^{n}\times\R. The regularity of the free boundary for ( 1 ) was developed in the groundbreaking [ Caf77 ] . The main result shows that the free boundary is smooth outside a closed set \Sigma of singular points, which present a “cusp-like” behaviour, namely such that the contact set \{u=0\} has zero density. The center of further studies has been the singular part of the free boundary. In particular, there has been interest on the study of the size of the singular set and its structure [ LM15 , Bla06 , FROS24 ] . If we denote by \Sigma_{t} the singular set at time t , variants of the techniques used in the elliptic counterpart show that \Sigma_{t} is contained in an (n-1) -dimensional C^{1} manifold. Moreover, the whole singular set \Sigma is contained in an (n-1) -dimensional manifold which is C^{1} in space and C^{1/2} in time [ Bla06 , LM15 ] . This results are optimal in space, in the sense that for a single time t the free boundary can be (n-1) -dimensional. However, it is still open the question of what is the size in time of the singular set, i.e. for how many times the set \Sigma_{t} is not empty. In the seminal paper [ FROS24 ] the authors show the sharp bound on the parabolic dimension 1 1 1 The parabolic Hausdorff dimension of a set is the Hausdorff dimension in by the parabolic distance \delta((x_{1},t_{1}),(x_{2},t_{2}))=\sqrt{|x_{1}-x_{2}|^{2}+|t_{1}-t_{2}|} . of the singular set \dim_{\pb}(\Sigma)\leq n-1 . Moreover, if n\leq 3 , the authors show that for almost every time t the singular set \Sigma_{t} is empty 2 2 2 More precisely, if u solves ( 1 ) in n+1 dimensions and \cS denotes the set of times such that \Sigma_{t} is not empty, then \dim_{\cH}(\cS)=0 when n=2 and \dim_{\cH}(\cS)\leq 1/2 when n=3 . . Our main result is the following. {teo} Let \Omega\subset\R^{4} , let u\in L^{\infty}(\Omega\times(0,T)) be a solution of the Stefan problem ( 1 ), and let \cS\coloneqq\{t\in(0,T)\,:\,\exists x\in\Omega\text{ s.t. }(x,t)\in\Sigma\}. Then \cH^{1}(\cS)=0. In particular, for almost every time t\in(0,T) , the free boundary is a 3-dimensional manifold of class C^{\infty} in \R^{4} . Theorem 1 is based on the following result, valid in all dimensions. {teo} Let \Omega\subset\R^{n} and let u be a bounded solution of the Stefan problem ( 1 ) in \Omega\times[0,T] . Then there exists \Sigma^{\infty}\subset\Sigma that can be covered by one (n-1) -dimensional manifold of class C^{\infty} in \R^{n+1} and such that \dim_{\cH}(\{t\in(0,T)\,:\,\exists(x,t)\in\Sigma^{\infty}\})=0, while \Sigma\setminus\Sigma^{\infty} is parabolically countably (n-2) -rectifiable 3 3 3 A set is parabolically countably m -rectifiable if it can be covered by countably many sets of the form f(E) , where E\subset\R^{m} and f\colon E\to\R^{n+1} is Lipschitz with respect to the parabolic distance. By [ Mat22 , Theorem 6.1] , this is stronger than rectifiability with respect to the euclidean structure. . {oss} The approach is solid and we expect it to apply also to less regular right hand sides. More precisely, assume that u\geq 0 satisfying \partial_{t}u\geq 0 and \partial_{t}u>0 in \{u>0\} solves (\Delta-\partial_{t})u=f\chi_{\{u>0\}} with f>0 and f\in C^{h,\beta} for some h\geq 3 and \beta\in(0,1) . Then the first part of Theorem 1 should hold with an (n-1) -dimensional manifold of class C^{h+1,\alpha} for all \alpha<\beta . Similarly, Theorem 1 is expected to hold assuming 0<f\in C^{\infty} . 1.1. Ideas of the proof The proof of Theorems 1 and 1 are based on a refinement of the results and the techniques used in [ FROS24 ] . In particular, once Theorem 1 is proven, to show Theorem 1 it is only necessary to control the size of the projection onto the time axis of \Sigma\setminus\Sigma^{\infty} . Since for solutions in \R^{4+1} this set is 2 -dimensional, it is sufficient to show a sharp “quadratic cleaning” of the free boundary, namely that at all singular points (x_{0},t_{0}) (2) \partial\{u(\cdot,t_{0}+Cr^{2})>0\}\cap B_{r}(x_{0})=\emptyset\quad\forall r>0, and combine this with a covering argument. We now illustrate the strategy adopted to prove Theorem 1 . Let (x_{0},t_{0}) be a singular point, then (3) u(x_{0}+x,t_{0}+t)=p_{2,x_{0},t_{0}}+o(|x|^{2}+|t|), where p_{2,x_{0},t_{0}} is a quadratic polynomial of the form \frac{1}{2}x^{t}\cdot Ax , with A\geq 0 and \tr A=1 . In particular, the singular set can be stratified as \Sigma=\cup_{k=0}^{n-1}\Sigma_{k} , where \Sigma_{k}\coloneqq\{\dim(\{p_{2,x_{0},t_{0}}=0\})=k\}. One of the consequences of ( 3 ) is that each stratum \Sigma_{k} can be covered by a k -dimensional manifold of class C^{1} in space and C^{1/2} in time. Hence the main challenge to show Theorem 1 is to prove that the stratum \Sigma_{n-1} can be covered by a smooth (i.e. of class C^{\infty} in space and time, with respect to the euclidean structure) (n-1) -dimensional manifold. In order to increase the regularity of the covering manifolds, one needs to improve the expansion ( 3 ) to higher order, thus Theorem 1 is based on a pointwise smooth expansion at all points in the maximal stratum \Sigma_{n-1} . Note that a solution of the Stefan problem is not smooth (as the optimal regularity for a solution of ( 1 ) is C^{1,1} in space and C^{1} in time), so one needs to introduce suitable functions (namely two-sided polynomial Ansätze, see Definition 15 ) to approximate the solution at all orders. {teo} For all \rho,M,c>0 , \alpha\in(0,1) and k\geq 3 there exist \bar{r},\beta>0 such that for all u solving the Stefan problem ( 1 ) in B_{1}\times[-1,1] and in the class \cS(M,c,\rho) (see Definition 3 ) and for all (x_{0},t_{0})\in\Sigma_{n-1}(u)\cap B_{1-\rho}\times[-1+\rho^{2},1] there is a two-sided polynomial Ansatz \mathscr{P}_{k} (see Definition 15 ) such that \|u(x_{0}+\cdot,t_{0}+\cdot)-\mathscr{P}_{k}\|_{L^{2}(B_{r}\times[-r^{2},-r^{2+\beta}])}\leq r^{k+\alpha}\quad\forall r<\bar{r}. One of the many contributions of [ FROS24 ] consists in noting that to prove Theorem 1.1 it is actually sufficient to show a C^{3+\beta} expansion, namely to prove that for all (x_{0},t_{0})\in\Sigma_{n-1} there is a cubic two-sided polynomial ansatz \mathscr{P}_{3} such that (4) u(x_{0}+x,t_{0}+t)=\mathscr{P}_{3}(x,t)+O((|x|+\sqrt{|t|})^{3+\beta}). As explained in [ FROS24 ]