On operator estimates in homogenization of non-local operators of convolution type

TL;DR

This paper analyzes the homogenization of non-local convolution operators, showing their resolvent operators converge to a second-order elliptic differential operator as a small parameter tends to zero, with precise convergence rates.

Contribution

It establishes the operator-norm convergence of non-local operators to a local elliptic operator in homogenization, providing sharp estimates of the convergence rate.

Findings

Resolvent operators converge in norm to the elliptic operator's resolvent.

The limit operator is a second-order elliptic differential operator with constant coefficients.

Sharp estimates of the convergence rate are obtained.

Abstract

The paper studies a bounded symmetric operator ${\mathbf{A}}_\varepsilon$ in $L_2(\mathbf{R}^d)$ with $$ ({\mathbf{A}}_\varepsilon u) (x) = \varepsilon^{-d-2} \int_{\mathbf{R}^d} a((x-y)/\varepsilon) \mu(x/\varepsilon, y/\varepsilon) \left( u(x) - u(y) \right)\,dy; $$ here $\varepsilon$ is a small positive parameter. It is assumed that $a(x)$ is a non-negative $L_1(\mathbf{R}^d)$ function such that $a(-x)=a(x)$ and the moments $M_k =\int_{\mathbf{R}^d} |x|^k a(x)\,dx$, $k=1,2,3$, are finite. It is also assumed that $\mu(x,y)$ is $\mathbf{Z}^d$-periodic both in $x$ and $y$ function such that $\mu(x,y) = \mu(y,x)$ and $0< \mu_- \leq \mu(x,y) \leq \mu_+< \infty$. Our goal is to study the limit behaviour of the resolvent $({\mathbf{A}}_\varepsilon + I)^{-1}$, as $\varepsilon\to0$. We show that, as $\varepsilon \to 0$, the operator $({\mathbf{A}}_\varepsilon + I)^{-1}$ converges in the operator norm in $L_2(\mathbf{R}^d)$ to the resolvent $({\mathbf{A}}^0 + I)^{-1}$ of the effective operator ${\mathbf{A}}^0$ being a second order elliptic differential operator with constant coefficients of the form ${\mathbf{A}}^0= - \operatorname{div} g^0 \nabla$. We then obtain sharp in order estimates of the rate of convergence.