Boundary estimates in elliptic homogenization (Q524235)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Boundary estimates in elliptic homogenization
scientific article

    Statements

    Boundary estimates in elliptic homogenization (English)
    0 references
    0 references
    0 references
    2 May 2017
    0 references
    The author proves sharp boundary estimates which are uniform down to the microscopic scale for the solution of systems of linear elasticity with rapidly oscillating periodic coefficients and with Dirichlet or Neumann boundary conditions, without assuming smoothness hypotheses on the coefficients. He considers the operator \[ \mathcal{L}_{\varepsilon }=-\mathrm{div} (A(x/\varepsilon )\nabla ), \] where \(A\) is a second-order tensor which satisfies the classical uniform continuity, coercivity and symmetry properties occurring in linear elasticity. Moreover \(A(y)\) is supposed to be 1-periodic. The first main result of the paper proves that if \(\Omega \) is a bounded Lipschitz domain in \(\mathbb{R}^{d}\) and if \(u_{\varepsilon }\in H^{1}(\Omega ;\mathbb{R}^{d})\) is the weak solution to the Dirichlet problem \(\mathcal{L}_{\varepsilon }u_{\varepsilon }=F\) in \(\Omega \), \(u_{\varepsilon }=f\) on \(\partial \Omega \), with \(F\in L^{p}(\Omega ;\mathbb{R}^{d})\) with \( p<2d/(d+1)\) and \(f\in H^{1}(\partial \Omega ;\mathbb{R}^{d})\), then for every \(\varepsilon <r<\operatorname{diam}(\Omega )\) the estimate \[ \left( \frac{1}{r} \int_{\Omega _{r}}\left| \nabla u_{\varepsilon }\right| ^{2}dx\right) ^{1/2}\leq C(\left\| F\right\| _{L^{p}(\Omega )}+\left\| f\right\| _{H^{1}(\partial \Omega )}) \] holds true for some positive constant \(C\), where \(\Omega _{r}=\{x\in \Omega :d(x,\partial \Omega )<r\}\).\ A similar estimate also occurs in the case of a Neumann boundary condition \(\frac{\partial u_{\varepsilon }}{\partial \nu }=g\) on \(\partial \Omega \), with \(g\in L^{2}(\partial \Omega ;\mathbb{R}^{d})\) and assuming that \(u_{\varepsilon }\perp R\), where \(R\) is the space of rigid displacements on \(\mathbb{R}^{d}\). The author first recalls the results already available and connected to the problems under consideration. The main difference with such previous results is that the coefficients \(A\) of the elasticity operator are here only supposed to be measurable. For the proof of its two main results, the author essentially uses convergence rates in \(H^{1}\) between the solutions of the original elasticity problem and of its homogenized version. He indeed proves an estimate \[ \left \| u_{\varepsilon }-u_{0}-\varepsilon \chi _{j}^{\beta }(x/\varepsilon )K_{\varepsilon }^{2}(\frac{\partial u_{0}^{\beta }}{\partial x_{j}}\eta _{\varepsilon })\right\| _{H_{0}^{1}(\Omega )}\leq C\varepsilon ^{1/2}\{\left\| f\right\| _{H^{1}(\partial \Omega )}+\left\| F\right\| _{L^{p}(\Omega )}\}, \] where \((\chi _{j}^{\beta }(y))\) is the matrix of correctors in \(H_{loc}^{1}( \mathbb{R}^{d};\mathbb{R}^{d})\), \(K_{\varepsilon }\) is the convolution product between \(f\) and a smooth and nonnegative function \(\varphi _{\varepsilon }\in C_{0}^{\infty }(B(0,1/4))\) with integral equal to 1, and \( \eta _{\varepsilon }\in C_{0}^{\infty }(\Omega )\) is a smooth truncation function with support in \(\{x\in \Omega :d(x,\partial \Omega )\geq 3\varepsilon \}\) and equal to 1 in \(\{x\in \Omega :d(x,\partial \Omega )\geq 4\varepsilon \}\). Here \(F\in L^{p}(\Omega ;\mathbb{R}^{d})\) with \(p=2d/(d+1)\) and \(f\in H^{1}(\partial \Omega ;\mathbb{R}^{d})\). This estimate is valid for both the Dirichlet and the Neumann problems, with a restriction in the latter case associated to the rigid displacements. The proofs of the main theorems are obtained as straightforward consequences of \(L^{2}\) estimates concerning the convolution operator \(K_{\varepsilon }\) and its gradient. From these estimates, the author derives \(L^{q}\) error estimates with \( q=2d/(d-1)\) between the solutions of the original elasticity problem and of the associated homogenized problem. In the last parts of the paper, the author presents quite similar estimates in the case of smooth domains and/or assuming \(A\in VMO(\mathbb{R}^{d})\).
    0 references
    0 references
    0 references
    0 references
    0 references
    sharp boundary estimates
    0 references
    Rellich estimates
    0 references
    Lipschitz estimates
    0 references
    convergence rates
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references