Sampling and equidistribution theorems for elliptic second order operators, lifting of eigenvalues, and applications (Q2306686)

The authors consider the uniformly elliptic operator \(\mathcal{H}:C_0^\infty({\mathbb R}^d) \to L^2({\mathbb R}^d)\) \[ \mathcal{H}u=-\operatorname{div}(Au\nabla u)+b^T\nabla u + cu \] with Lipschitz continuous leading order coefficients and bounded coefficients \(b,c\). Let \(H\) be the Friedrichs extension of \(\mathcal{H}\). Another operator considered in this article is the corresponding operator \(H_L\) defined on open centered cube \((-L/2,L/2)^d\) of side length \(L\). The authors investigate properties of the functions \(\psi\) satisfying \(|H\psi|\leq |V\psi|+|\zeta|\), \(V:{\mathbb R}^d\to {\mathbb R}\), \(V\in L^\infty({\mathbb R}^d)\), \(\zeta\in L^2({\mathbb R}^d)\) (and the corresponding functions \(\psi_L\) for the operator \(H_L\)). The main results of the paper are a quantitative sampling theorem (for \(\psi\)) and an equidistribution theorem (for \(\psi_L\)). They compare the \(L^2\)-norm of a solution on the full domain to the \(L^2\)-norm on balls distributed evenly throughout the domain. The quantitative dependence of the constants in these theorems of the coefficients of the operators is obtained. From authors' abstract: ``The estimates are scale-free, in the sense that for a sequence of growing cubes we obtain uniform estimates. These results are applied to prove lifting of eigenvalues as well as the infimum of the essential spectrum, and an uncertainty relation (aka spectral inequality) for short energy interval spectral projectors. Several applications including random operators are discussed. In the proof we have to overcome several challenges posed by the variable coefficients of the leading term.''