The fractional \(p\)-Laplacian emerging from homogenization of the random conductance model with degenerate ergodic weights and unbounded-range jumps (Q2278126)

The authors describe the asymptotic behavior when \(\varepsilon \rightarrow 0\) of the energy functional \(\mathcal{E}_{p,s,\varepsilon }\) defined on \(\mathcal{H}_{\varepsilon }=\{u:\mathbb{Z}_{\varepsilon }^{d}=\varepsilon\mathbb{Z}^{d}\rightarrow \mathbb{R}\}\) as \[\mathcal{E}_{p,s,\varepsilon}(u)=\varepsilon ^{2d}\sum_{(x,y)\in \mathbb{Z}_{\varepsilon }^{2d}}c_{\frac{x}{\varepsilon },\frac{y}{\varepsilon }}\frac{V(u(x)-u(y))}{\left\vert x-y\right\vert ^{d+ps}}+\varepsilon ^{d}\sum_{x\in \mathbb{Z}_{\varepsilon}^{d}}G(u(x))-\varepsilon ^{d}\sum_{x\in \mathbb{Z}_{\varepsilon}^{d}}u(x)f_{\varepsilon }(x).\] Here the random field \(c\) is stationary ergodic in \(\mathbb{Z}^{d}\times \mathbb{Z}^{d}\) with expectation \(\mathbb{E}(c)<\infty \) and is such that for any \(s\in (0,1)\) and \(p>1\) there exist \(\mathfrak{q}\in (\frac{d}{ps},+\infty )\) and \(r\in (1,p)\) such that \(\mathbb{E}(c^{-\mathfrak{q}})<\infty \) and \(\mathfrak{q}\geq \frac{r}{p-r}>\frac{d}{ps}\), the function \(V:\mathbb{R}^{m}\rightarrow \mathbb{R}\) is continuous, and there exist \(\alpha ,\beta ,c>0\) and \(p\in \lbrack 1,\infty ]\) such that \(\alpha \left\vert \xi \right\vert ^{p}\leq V(\xi )\leq c+\beta \left\vert\xi \right\vert ^{p}\) and \(\xi \rightarrow \left\vert \xi \right\vert^{-p}V(\xi )\) is continuous in 0, the function \(G:\mathbb{R}^{m}\rightarrow\mathbb{R}\) is non-negative and continuous and satisfies \(G(\xi )\leq \alpha \left\vert \xi \right\vert ^{k}\), for \(\alpha >0\), \(k < p_{\mathfrak{q}}^{\star }=\frac{dp\mathfrak{q}}{2d+d\mathfrak{q}-sp\mathfrak{q}}\), and \(f_{\varepsilon }\in \mathcal{H}_{\varepsilon }\) is such that \(\mathcal{R}_{\varepsilon }^{\ast }f_{\varepsilon }\) weakly converges to \(f\) in \(L^{r^{\star }}(Q)\). Here the operator \(\mathcal{R}_{\varepsilon }^{\ast }\) is defined through \(\mathcal{R}_{\varepsilon }^{\ast }u(x)=u(x_{i})\) if \(x_{i}\in \mathbb{Z}^{d}\) and \(x\in x_{i}+[-\frac{\varepsilon }{2},\frac{\varepsilon }{2})^{d}\). This allows to compare discrete and continuous functions. The main result proves that the sequence \(\mathcal{E}_{p,s,\varepsilon }\) restricted to \(\mathcal{H}_{\varepsilon }(Q)\), where \(Q\) is any bounded domain in \(\mathbb{R}^{d}\), Mosco-converges almost surely to the functional \(\mathcal{E}_{p,s}\) defined as \[\mathcal{E}_{p,s}(u) = \mathbb{E}(c)\int\int_{Q\times Q}\frac{V(u(x)-u(y))}{\left\vert x-y\right\vert^{d+ps}}+\int_{Q}G(u(x))-\int_{Q}u(x)f(x).\] This functional may be associated to an extension of the fractional \(p\)-Laplace equation. Assuming further conditions on the function \(G\), the authors also prove that the sequence \(\mathcal{E}_{p,s,\varepsilon }\) restricted to \(\mathcal{H}_{\varepsilon }\) Mosco-converges almost surely to \(\mathcal{E}_{p,s}\). For the proof, the authors first gather properties of discrete and continuous Sobolev-Slobodeckij spaces. They also define the notion of uniform extension domain and they prove a discrete Sobolev inequality on \(\mathbb{Z}_{\varepsilon }^{d}\) and embedding theorems. Key tools of the proof of the homogenization results are an ergodic theorem proved by \textit{D.J. Daley} and \textit{D. Vere-Jones} in [An introduction to the theory of point processes. New York: Springer-Verlag (1988; Zbl 0657.60069)] and an extended ergodic theorem proved by \textit{F. Flegel} et al. [Ann. Inst. Henri Poincaré, Probab. Stat. 55, No. 3, 1226--1257 (2019; Zbl 1442.60071)]. The authors prove a \(\lim \inf \) property of the first term of the functional \(\mathcal{E}_{p,s,\varepsilon }\) for every sequence \(u_{\varepsilon }:\mathbb{Z}_{\varepsilon }^{d}\rightarrow \mathbb{R}^{m}\) of functions having support in \(Q_{\varepsilon }\) such that \(\mathcal{R}_{\varepsilon }^{\ast }u_{\varepsilon }\rightarrow u\) pointwise a.e. and \(\sup_{\varepsilon }\left\Vert u_{\varepsilon }\right\Vert _{\infty }<\infty \) and they analyze the behavior of the other terms of \(\mathcal{E}_{p,s,\varepsilon }\). From these homogenization results, the authors prove the convergence of solutions of associated minimization and eigenvalue problems.