Global gradient estimates for elliptic equations of \(p(x)\)-Laplacian type with BMO nonlinearity (Q294266)

The authors consider the elliptic boundary value problem \[ -\operatorname{div} a(Du,x)=\operatorname{div}(| F(x)|^{p(x)-2F(x)})\text{ in }\Omega,\quad u=0\text{ on }\partial \Omega,\tag{P} \] where \(\Omega\) is a bounded open domain in \(\mathbb R^n (n\geq 2)\), \(p:\mathbb R^n\rightarrow (1,\infty)\) is a log-Hölder continuous bounded function, with \(\inf_{\mathbb R^n}p>1\), \(a:\mathbb R^n\times \mathbb R^n\rightarrow \mathbb R^n\) is a function differentiable in the first variable, measurable in the second one, and satisfying an ellipticity condition for the variable exponent \(p(x)\), and \(F\in L^{p(x)}(\Omega,\mathbb R^n)\) is a given vector field. The authors prove the following global gradient estimate for the weak solution \(u\in W_0^{1,p(x)}(\Omega)\) of problem (P): Assume that \(| F|^{p(x)}\in L^q(\Omega)\), for some \(q>1\). Then, there exists a constant \(\tilde{\sigma}>0\) with the following property: for every \(\sigma \in (0,\tilde{\sigma})\) there exists \(\delta>0\) such that, if \(a\) is \((\delta,R_0)\)-vanishing and \(\Omega\) is \((\delta,R_0)\)-Reifenberg flat for some \(R_0>0\), then \(| Du|^{p(x)}\in L^q(\Omega)\) and there exists a constant \(c>0\), independent of \(u\), such that \[ \left(\int_\Omega | Du|^{p(x)q}dx\right)^{\frac{1}{q}}\leq c\left(\int_\Omega (| F|^{p(x)}+1)dx+1\right)^\sigma\left(\int_\Omega (| F|^{p(x)q}+1)dx\right)^{\frac{1}{q}}.\tag{1} \] The authors emphasize that the \((\delta,R_0)\)-vanishing condition required on \(a\) is weaker than the continuous condition and that the \((\delta,R_0)\)-Reifenberg flatness condition on \(\Omega\) is weaker than the Lipschitz regularity condition usually assumed in the literature. The main result of this paper extends the results in [\textit{E. Acerbi} and \textit{G. Mingione}, ibid. 584, 117--148 (2005; Zbl 1093.76003)], where local gradient estimates for the solution of problem (P) were established for equations with continuous coefficients. The proof is based on a perturbation argument: the authors consider a limiting problem where the coefficients are averaged with respect the variables and establish gradient comparison estimates between the solution to the original problem and the solution to the limiting problem. These comparison estimates and some properties of the level sets of the gradient of the solution to the limiting problem allow to obtain, using a Vitali-type covering lemma, an estimate for the upperlevel sets for the gradient of the solution to the original problem. From this estimate, after integration, one obtains inequality (1).