Global gradient estimates for the borderline case of double phase problems with BMO coefficients in nonsmooth domains (Q526028): Difference between revisions

The authors consider general stationary equations written in divergence form as \(\mathrm{div\,}A(x,Du)=\mathrm{div\,}G(x,F)\) in \(\Omega \) with homogeneous Dirichlet boundary conditions \(u=0\) on \(\partial \Omega \), where \(A:\mathbb{R }^{n}\times \mathbb{R}^{n}\rightarrow \mathbb{R}^{n}\) is measurable in \(x\), differentiable with respect to \(\xi \neq 0\) and satisfies \[ \left| A(x,\xi )\right| +\left| \xi \right| \left| D_{\xi }A(x,\xi )\right| \leq L(\left| \xi \right| ^{p-1}+a(x)\left| \xi \right| ^{p-1}\log (e+\left| \xi \right| )), \] \[ \left\langle D_{\xi }A(x,\xi )\eta ,\eta \right\rangle \geq \nu (\left| \xi \right| ^{p-2}+a(x)\left| \xi \right| ^{p-2}\log (e+\left| \xi \right| ))\left| \eta \right| ^{2}, \] for a.e. \(x\in \mathbb{R }\), \(\xi ,\eta \in \mathbb{R}^{n}\), where \(0<\nu \leq L<+\infty \) and \(a\) is a nonnegative and bounded function. The purpose of the paper is to prove bounds on the gradient of a weak solution to the above problem under appropriate hypotheses on the data. \(A\) is indeed supposed to be \((\delta ,R_{0})\)-vanishing with \(0<\delta <1/8\) satisfying \[ \begin{aligned}\sup_{0<r\leq R_{0}}\sup_{y\in \mathbb{R}^{n}}\frac{1}{\left| B_{r}(y)\right| } \int_{B_{r}(y)}\sup_{\xi \in \mathbb{R}^{n}\setminus \{0\}}\bigg| \frac{ A(x,\xi )}{\left| \xi \right| ^{p-1}+a(x)\left| \xi \right| ^{p-1}\log (e+\left| \xi \right| )}\,\,-\\ \left( \frac{A(\cdot ,\xi )}{ \left| \xi \right| ^{p-1}+a(\cdot )\left| \xi \right| ^{p-1}\log (e+\left| \xi \right| )}\right) _{B_{r}(y)}\bigg| dx\leq \delta .\end{aligned} \] \(\Omega \) is supposed to be \((\delta ,R_{0})\)-Reifenberg flat, that is, there exists \(x\in \partial \Omega \) such that, for each \(r\in (0,R_{0}]\), there exists a coordinate system \(\{y^{1},\ldots ,y^{n}\}\) such that \(x=0\) in this coordinate system and \[ B_{r}(0)\cap \{y^{n}>\delta r\}\subset B_{r}(0)\cap \Omega \subset B_{r}(0)\cap \{y^{n}>-\delta r\}. \] The authors define the notion of a distributional solution \(u\in W_{0}^{1,1}(\Omega )\) which satisfies \[ \int_{\Omega }\left\langle A(x,Du),D\varphi \right\rangle dx=\int_{\Omega }\left\langle G(x,F),D\varphi \right\rangle dx \] for every \(\varphi \in C_{0}^{\infty }(\Omega )\). The main result of the paper is that if \(H(x,Du),H(x,F)\in L^{1}(\Omega )\) with \(H(x,\xi )=\left| \xi \right| ^{p}+a(x)\left| \xi \right| ^{p}\log (e+\left| \xi \right| )\) and \(H(x,F)\in L^{\gamma }(\Omega )\) for some \(\gamma \in (1,\infty )\), then there exists \(\delta >0\) which only depends on the data of the problem such that \(H(x,Du)\in L^{\gamma }(\Omega )\) with a bound of its norm in \(L^{\gamma }(\Omega )\) in terms of that of \(H(x,F)\). For the proof, the author requires properties of Orlicz and Musielak-Orlicz spaces, among which are Caccioppoli-type estimates and comparison estimates.