Geometric regularity estimates for quasi-linear elliptic models in non-divergence form with strong absorption

by, João Vitor da Silva · 2025

Read Geometric regularity estimates for quasi-linear elliptic models in non-divergence form with strong absorption by by, João Vitor da Silva (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-2503-21899.txt.

Introduction

Claudemir Alcantara, João Vitor da Silva, and Ginaldo Sá study how smooth the solutions of a class of hard equations must be. The equations are quasi-linear and elliptic, a family used to model diffusion-like processes. Their version can behave badly where the slope of the solution vanishes, and it includes a strong absorption term that eats the solution away.

Because of that term, regions can appear where the solution is exactly zero, like a plateau, with an unknown border. The authors prove improved smoothness estimates right along that border. They pin down the best possible smoothness rate in terms of the equation's settings. They also show the solution grows away from the border at a controlled rate, and they prove a Liouville theorem restricting whole-space solutions. Such results matter because the border is a free boundary, an edge analysts must control in many models.

Catalog record

CREATORby, João Vitor da Silva
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE90 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-21899

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Geometric regularity estimates for quasi-linear elliptic models in non-divergence form with strong absorption by, João Vitor da Silva Abstract In this manuscript, we investigate geometric regularity estimates for problems governed by quasi-linear elliptic models in non-divergence form, which may exhibit either degenerate or singular behavior when the gradient vanishes, under strong absorption conditions of the form: | ∇ u ​ ( x ) | γ ​ Δ p N ​ u ​ ( x ) = f ⁡ ( x , u ) in B 1 , |\nabla u(x)|^{\gamma}\Delta_{p}^{\mathrm{N}}u(x)=f(x,u)\quad\text{in}\quad B_{1}, where γ > − 1 \gamma>-1 , p ∈ ( 1 , ∞ ) p\in(1,\infty) , and the mapping u ↦ f ⁡ ( x , u ) ≲ 𝔞 ⁡ ( x ) ​ u + m u\mapsto f(x,u)\lesssim\mathfrak{a}(x)u_{+}^{m} (with m ∈ [ 0 , γ + 1 ) m\in[0,\gamma+1) ) does not decay sufficiently fast at the origin. This condition allows for the emergence of plateau regions, i.e., a priori unknown subsets where the non-negative solution vanishes identically. We establish improved geometric C loc κ \mathrm{C}^{\kappa}_{\text{loc}} regularity along the set ℱ 0 = ∂ { u > 0 } ∩ B 1 \mathscr{F}_{0}=\partial\{u>0\}\cap B_{1} (the free boundary of the model) for a sharp value of κ ≫ 1 \kappa\gg 1 , which is explicitly determined in terms of the structural parameters. Additionally, we derive non-degeneracy results and other measure-theoretic properties. Furthermore, we prove a sharp Liouville theorem for entire solutions exhibiting controlled growth at infinity. Keywords : Geometric regularity estimates, quasi-linear elliptic models, non-divergence form operators. AMS Subject Classification: Primary 35B65, 35J60, 35J70, 35J75; Secondary 35D40 Pontifícia Universidade Católica do Rio de Janeiro - PUC-Rio - Departamento de Matemática. Rio de Janeiro - RJ, Brazil E-mail address: [email protected] Universidade Estadual de Campinas - UNICAMP. Departamento de Matemática. Campinas - SP, Brazil. E-mail address: [email protected] 1 Introduction In this article, we investigate sharp regularity estimates for diffusion problems governed by quasi-linear operators in non-divergence form (with singular or degenerate characteristics) for which a Minimum Principle does not apply. Specifically, we focus on prototypical models arising in Tug-of-War games (stochastic models with noise), where the presence of dead core regions plays a fundamental role in understanding the nature of such models (cf. [ 9 ] , [ 16 ] , [ 32 ] , and [ 36 ] ). The model under consideration is given by (1.1) |\nabla u(x)|^{\gamma}\Delta_{p}^{\mathrm{N}}u(x)=f(x,u)\lesssim\mathfrak{a}(x)u_{+}^{m}(x)\quad\text{in}\quad\Omega, where ( A0 ) p\in(1,\infty) is a fixed parameter (related to the elliptic nature of the operator); ( A1 ) \gamma\in(-1,\infty) governs the singularity/degeneracy of the model; ( A2 ) m\in[0,\gamma+1) represents the strong absorption exponent of the model. Additionally, we impose the boundary condition (1.2) u(x)=g(x)\quad\text{on }\partial\Omega, where the boundary datum is assumed to be continuous and non-negative, i.e., 0\leq g\in\mathrm{C}^{0}(\partial\Omega) . Throughout this manuscript, for a fixed p\in(1,\infty) , the operator \Delta^{\mathrm{N}}_{p}u=|\nabla u|^{2-p}\operatorname{div}(|\nabla u|^{p-2}\nabla u)=\Delta u+(p-2)\left\langle D^{2}u\frac{\nabla u}{|\nabla u|},\frac{\nabla u}{|\nabla u|}\right\rangle denotes the Normalized p -Laplacian operator or Game-theoretic p -Laplace operator , which arises in certain stochastic models (see [ 16 ] , [ 32 ] , and [ 36 ] for comprehensive surveys on Tug-of-War games and their associated PDEs). Furthermore, the fully nonlinear operator \Delta_{\infty}^{\mathrm{N}}u(x)\coloneqq\left\langle D^{2}u\frac{\nabla u}{|\nabla u|},\frac{\nabla u}{|\nabla u|}\right\rangle is referred to as the Normalized infinity-Laplacian , which also plays a significant role in the interplay between nonlinear PDEs and Tug-of-War games (cf. [ 16 ] and [ 36 ] ). We assert that the Normalized p -Laplacian operator is “uniformly elliptic” in the sense that \mathcal{M}^{-}_{\lambda,\Lambda}(D^{2}u)\leq\Delta_{p}^{\mathrm{N}}u\leq\mathcal{M}^{+}_{\lambda,\Lambda}(D^{2}u), where (Pucci) \mathcal{M}^{-}_{\lambda,\Lambda}(D^{2}u)\coloneqq\inf_{\mathfrak{A}\in\mathcal{A}_{\lambda,\Lambda}}\text{Tr}(\mathfrak{A}D^{2}u)\quad\text{and}\quad\mathcal{M}^{+}_{\lambda,\Lambda}(D^{2}u)\coloneqq\sup_{\mathfrak{A}\in\mathcal{A}_{\lambda,\Lambda}}\text{Tr}(\mathfrak{A}D^{2}u) are the Pucci extremal operators , and \mathcal{A}_{\lambda,\Lambda} denotes the set of symmetric n\times n matrices whose eigenvalues lie within the interval [\lambda^{\mathrm{N}}_{p},\Lambda^{\mathrm{N}}_{p}] . Indeed, the Normalized p -Laplacian operator can be expressed in the form \Delta_{p}^{\mathrm{N}}u=\mathrm{Tr}\left[\left(\mathrm{Id}_{n}+(p-2)\frac{\nabla u}{|\nabla u|}\otimes\frac{\nabla u}{|\nabla u|}\right)D^{2}u\right], from which it is straightforward to verify that \lambda^{\mathrm{N}}_{p}=\min\{1,p-1\}\quad\text{and}\quad\Lambda^{\mathrm{N}}_{p}=\max\{1,p-1\}. Next, we detail our assumption, which pertains to the bounded Thiele modulus or modulating function \mathfrak{a} . \mathrm{A} 1 ( Thiele modulus ) . We assume that \mathfrak{a}:\Omega\to\mathbb{R}_{+} (referred to as the Thiele modulus or modulating function) is a continuous and bounded function (strictly positive and finite). Specifically, there exist constants \Lambda_{0}\geq\lambda_{0}>0 such that \Lambda_{0}\geq\sup_{\Omega}\mathfrak{a}(x)\geq\mathfrak{a}(x)\geq\inf_{\Omega}\mathfrak{a}\geq\lambda_{0}. In this framework, equation ( 1.1 ) is classified as an equation with a strong absorption condition, and the set \partial\{u>0\}\cap\Omega constitutes the (physical) free boundary of the problem. The mathematical analysis of equation ( 1.1 ) is significant not only due to its applications but also because of its connection to various mathematical free boundary problems extensively studied in the literature (cf. Alt–Phillips [ 1 ] , Díaz [ 25 ] , Friedman–Phillips [ 26 ] , and Phillips [ 38 ] , as well as Shahgholian et al. [ 28 ] and [ 31 ] for related problems with variational structure; additionally, for the non-variational counterpart of these models, see Da Silva et al. [ 20 ] and Teixeira [ 45 ] ). A fundamental aspect of free boundary problems like ( 1.1 ) is determining the optimal regularity of viscosity solutions. For instance, if we fix 0<m<\gamma+1 and \mathrm{R}>r>0 , the radial profile \omega:B_{\mathrm{R}}(x_{0})\to\mathbb{R}^{+} given by (1.3) \omega(x)=\omega_{\gamma}^{\mathrm{R}}(x)=\mathrm{c}_{n,\gamma,m,p}\left(|x-x_{0}|-r\right)^{\beta}_{+},\quad\text{for}\quad\beta=\frac{\gamma+2}{\gamma+1-m}, where \mathrm{c}_{n,\gamma,m,p}=\mathrm{c}(n,\gamma,m,p)=\frac{1}{\beta^{\frac{\gamma+1}{\gamma+1-m}}}\frac{1}{\left[n-1+(p-1)(\beta-1)\right]^{\frac{1}{\gamma+1-m}}}>0 is a suitable constant, is a viscosity solution to the equation -|\nabla\omega(x)|^{\gamma}\Delta_{p}^{\mathrm{N}}\omega(x)+\omega^{q}(x)=0\quad\text{in}\quad B_{\mathrm{R}}(x_{0}). It is well-known that, in general, solutions to ( 1.1 ) belong to C^{1,\kappa}_{\text{loc}} for some \kappa\in(0,1) (cf. Attouchi–Ruosteenoja [ 9 ] for further details). However, for this particular example, with \gamma>-1 fixed, one observes that \omega\in C^{\beta}_{\text{loc}} at free boundary points, where (1.4) \beta=\beta(\gamma,m):=\frac{\gamma+2}{\gamma+1-m}. Notably, when m>0 , we obtain \beta(\gamma,m)=\frac{\gamma+2}{\gamma+1-m}>1+\frac{1}{\gamma+1}\quad\Rightarrow\quad\text{Gain of smoothness along the free boundary.} In other words, within a small radius, viscosity solutions satisfy the sharp growth estimates (1.5) \varrho^{\beta}\lesssim\sup_{B_{\varrho}(x_{0})}\omega(x)\lesssim\varrho^{\beta}. 1.1 Main results and some consequences Inspired by the preceding motivation, we will demonstrate that bounded viscosity solutions of ( 1.1 ) satisfy a growth estimate of the form ( 1.5 ) near free boundary points (see Theorems 1.1 and 1.4 for further details). Next, we establish an improved regularity estimate (near free boundary points) for solutions of ( 1.1 ), a sort of \beta- power-like decay (for \beta>1 as in ( 1.4 )). Theorem 1.1 ( Improved Regularity). Suppose that (\mathrm{A}0)-(\mathrm{A}2) , and \mathrm{A}{\ref{Assumption_lambda}} (upper bound) hold true, and let u\in\mathrm{C}^{0}(B_{1}) be a non-negative bounded viscosity solution to |\nabla u(x)|^{\gamma}\Delta_{p}^{\mathrm{N}}u(x)=\mathfrak{a}(x)u_{+}^{m}(x)\quad\text{in}\quad B_{1} and let x_{0}\in\partial\{u>0\}\cap B_{1/2} . Then, for any point z\in\{u>0\}\cap B_{1/2} , there exists a universal constant 1 1 1 Throughout this work, we will refer to universal constants when they depend only on the dimension and structural parameters of the problem, i.e., n,\,p,\,\gamma,\,m , and the bounds of \mathfrak{a} . \mathrm{C}_{\mathcal{G}}>0 such that u(z)\leq\mathrm{C}_{\mathcal{G}}\|u\|_{L^{\infty}(B_{1})}|z-x_{0}|^{\beta}. As a consequence of our findings, we establish a more precise control for viscosity solutions of ( 1.1 ) near their free boundaries. This information is crucial for various quantitative aspects of numerous free boundary problems (see, e.g., [ 3 ] and [ 46 ] ). More precisely, we prove that, close to their free boundaries, solutions decay at a rate proportional to an appropriate power of \mathrm{dist}(\cdot,\partial\{u>0\}) . Corollary 1.2 . Let u be a nonnegative, bounded viscosity solution to ( 1.1 ) in \Omega . Given x_{0}\in\{u>0\}\cap\Omega^{\prime} with \Omega^{\prime}\Subset\Omega , then u(x_{0})\leq\mathrm{C}_{\sharp}\,\mathrm{dist}\left(x_{0},\partial\{u>0\}\right)^{\beta}, where \mathrm{C}_{\sharp}>0 is a universal constant. Remark 1.3 .

Read the complete text