Homogenisation for the Stokes equations in randomly perforated domains under almost minimal assumptions on the size of the holes (Q2334991): Difference between revisions

The authors prove a homogenization result for the Stokes problem \[-\Delta u_{\varepsilon}+\nabla p_{\varepsilon}=f,\;\;\; \nabla \cdot u_{\varepsilon}=0,\] posed in \(D^{\varepsilon}=D\setminus H^{\varepsilon}\), where \(D\subset \mathbb{R}^{d}\), \(d>2\), is an open and bounded domain, which is star-shaped with respect to the origin, and \(H^{\varepsilon}\) is the union of a random number of small balls \(B_{\varepsilon ^{\frac{d}{d-2}}\rho _{i}}(\varepsilon z_{i})\) with \(z_{i}\in \Phi +\frac{1}{\varepsilon}D\), \(\Phi \) being a Poisson point process on \(\mathbb{R}^{d}\) of homogeneous intensity rate \(\lambda >0\), and \(\{\rho _{i}\}\) are identically and independently distributed unbounded random variables. The source term \(f\) of the Stokes equation is supposed to belong to \(H^{-1}(D;\mathbb{R}^{d})\) and the fluid is supposed to perfectly adhere to the balls and to the boundary of \(D\). The probability space associated to the joint process \((\Phi ,\mathcal{R})\) of the centres and radii is denoted as \((\Omega ,\mathcal{F},\mathbb{P})\). The main result of the paper proves that, for \(\mathbb{P}\)-almost every \(\omega \in \Omega \), \(u_{\varepsilon}(\omega ,\cdot )\) weakly converges in \(H_{0}^{1}(D;\mathbb{R}^{d})\) to the solution \(u_{h}\) of Brinkman's equation \[ -\Delta u_{h}+\nabla p_{h}+C_{d}\lambda \left\langle \rho^{d-2}\right\rangle u_{h}=f, \] in \(D\) with the divergence property \(\nabla\cdot u_{h}=0\) and the boundary condition \(u_{h}=0\) on \(\partial D\). In the extra term of this homogenized problem, \(\left\langle \cdot \right\rangle\) denotes the expectation under the probability measure on the radii \(\rho_{i}\) and the positive constant \(C_{d}\) only depends on the dimension \(d\) and is equal to \(6\pi \) in the case \(d=3\). The authors prove a similar result for the stationary Navier-Stokes equation. Concerning the pressure, the authors prove the existence of a set \(E^{\varepsilon}\subseteq \mathbb{R}^{d}\) satisfying \(E^{\varepsilon}\supset H^{\varepsilon}\) and \(Cap(E^{\varepsilon}\setminus H^{\varepsilon})\rightarrow 0\) when \(\varepsilon \) goes to 0 such that for every compact set \(K\Subset D\), the modification of the pressure \(\widetilde{p}_{\varepsilon}=p_{\varepsilon}-\frac{1}{\left\vert K\setminus E^{\varepsilon}\right\vert}\int_{K\setminus E^{\varepsilon}}p_{\varepsilon}\) in \(K\setminus E^{\varepsilon}\) and 0 outside weakly converges in \(L_{0}^{q}(K;\mathbb{R})\) to \(p_{h}\) for every \(q<\frac{d}{d-1}\). The main tool of the proof of the first main result is the construction of a linear operator \(R_{\varepsilon}:\{v\in C_{0}^{\infty}(D;\mathbb{R}^{d}):\nabla \cdot v=0\}\rightarrow H^{1}(D;\mathbb{R}^{d})\) such that \(\int \nabla R_{\varepsilon}v:\nabla u_{\varepsilon}dx\rightarrow\int \nabla v:\nabla udx+C_{d}\lambda \left\langle \rho ^{d-2}\right\rangle\int v\cdot udx\), for all \(u_{\varepsilon}\in H^{1}(D^{\varepsilon}; \mathbb{R}^{d})\) satisfying \(\nabla \cdot u_{\varepsilon}=0\) in \(D\) and \(u_{\varepsilon}\rightharpoonup u\) weakly in \(H_{0}^{1}(D;\mathbb{R}^{d})\), among other properties. This construction is obtained splitting the set \(H^{\varepsilon}\) into a ``good'' set \(H_{g}^{\varepsilon}\) which contains holes which are small and well-separated and a ``bad'' set \(H_{b}^{\varepsilon}\) which contains big and overlapping holes. The authors build \(R_{\varepsilon}v\) such that it vanishes on \(H_{g}^{\varepsilon}\) following ideas by \textit{G. Allaire} in [Arch. Ration. Mech. Anal. 113, No. 3, 209--259 (1991; Zbl 0724.76020)] and by \textit{L. Desvillettes} et al. in [J. Stat. Phys. 131, No. 5, 941--967 (2008; Zbl 1154.76018)]. On \(H_{b}^{\varepsilon}\), the authors introduce a suitable covering \(\overline{H}_{b}^{\varepsilon}\) of the set \(H_{b}^{\varepsilon}\) selecting some of the balls in \(H_{b}^{\varepsilon}\), dilating them by a uniformly bounded factor \(\lambda _{\varepsilon}\leq\Lambda \), solving different Stokes problems in disjoint annuli and iterating this procedure a finite number of steps. This requires the derivation of geometric properties of the holes and improves a procedure previously introduced by the same authors and \textit{J. J. L. Velazquez} in [Commun. Partial Differ. Equations 43, No. 9, 1377--1412 (2018; Zbl 1448.60136)]. The last part of the paper gives some probabilistic results on the random set \(H^{\varepsilon}\), in terms of the size of the clusters generated by overlapping balls of comparable size.