Modica type gradient estimates for an inhomogeneous variant of the normalized p-Laplacian evolution

From MaRDI portal
Publication:2349503



Abstract: In this paper, we study an inhomogeneous variant of the normalized p-Laplacian evolution which has been recently treated in cite{BG1}, cite{Do}, cite{MPR} and cite{Ju}. We show that if the initial datum satisfies the pointwise gradient estimate eqref{e:main1} a.e., then the unique solution to the Cauchy problem eqref{main5} satisfies the same gradient estimate a.e. for all later times, see eqref{e:main} below. A general pointwise gradient bound for the entire bounded solutions of the elliptic counterpart of equation eqref{main5} was first obtained in cite{CGS}. Such estimate generalizes one obtained by L. Modica for the Laplacian, and it has connections to a famous conjecture of De Giorgi.


In this paper, the authors consider the following Cauchy problem in \(\mathbb{R}^n\times [0,T]\): \[ u_t=|Du|^{2-p}\{\mathrm{div}(|Du|^{p-2}Du)-F'(u) \}\quad u(\cdot,0)=g, \] where \(p\in (1,2]\), \(F\in C^{2,\beta}_{\mathrm{loc}}(\mathbb{R})\) for some \(\beta > 0\) and \(F\geq 0\). Firstly, they show that the problem is well posed in the space \(H^T=\{u\in C(\mathbb{R}^n \times [0,T]): x\mapsto u(x,t)\in C^{0,1}(\mathbb{R}^n),\;\|u \|_{L^\infty(\mathbb{R}^n \times [0,T])},\;\|Du \|_{L^\infty(\mathbb{R}^n \times [0,T])}<\infty\} \) for \(g\in C^{0,1}(\mathbb{R}^n)\), \(\|g \|_{L^\infty(\mathbb{R}^n)}\), \(\|Dg \|_{L^\infty(\mathbb{R}^n)}<\infty\) and any \(T>0\) (if \(p=2\), they also assume a further assumption on \(F\)). Their main result, a parabolic analog of Modica's gradient estimate, says that if such a \(g\) satisfies \(|Dg(x)|^p\leq \frac{p}{p-1}F(g(x))\) a.e., then, for any \(t>0\), it holds \(|Du(x,t)|^p\leq \frac{p}{p-1}F(u(x,t ))\) a.e. Their approach is based on a regularization of the problem and on the use of barriers. As they remark, if \(p=2\), the solution is smooth and the above inequality holds for any \(x\in \mathbb{R}^n\). Moreover, under a further minor hypothesis on \(g\), they obtain better regularity in two dimensions. Lastly, they prove a Liouville type result: if \(F(u(x_0,t_0) )=0\) at some point \((x_0,t_0 )\), then \(u\) is a constant.



Cites work









This page was built for publication: Modica type gradient estimates for an inhomogeneous variant of the normalized \(p\)-Laplacian evolution

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2349503)