The landscape law for tight binding Hamiltonians (Q2099806)

In the remarkable paper [``Universal mechanism for Anderson and weak localization'', Proc. Natl. Acad. Sci. USA 109, No. 37, 14761--14766 (2012; \url{doi:10.1073/pnas.112043210})], \textit{M. Filoche} and \textit{S. Mayboroda} have given an effective theory to predict the localization of eigenfunctions of the operator \(L=-\operatorname{div}A\nabla+V\). According to this theory, the precise spatial location of quantum states in a potential can be predicted using the solution of the associated Dirichlet problem \(Lu = 1\), called the \textit{localization landscape}. In the subsequent paper by \textit{G. David} et al. [Adv. Math. 390, Article ID 107946, 34 p. (2021; Zbl 1479.35267)] the authors proposed to use \(1/u\) to estimate the integrated density of states for the continuous model with operator \(-\Delta+V\). The paper under review is devoted to the \textit{discrete tight binding Hamiltonian} \(H\). This is the linear operator on \(\ell^2({\Lambda})\) defined by \[ (H \varphi)_n=-\sum_{|m-n|_1=1}\left(\varphi_m-\varphi_n\right)\,+ v_n \varphi_n,\ \ n\in {\Lambda}, \] where \(\Lambda= (\mathbb{Z}/K\mathbb{Z})^d\) is an integer torus; \(3 \leq K\in\mathbb{N}\); \(|n|_1:=\sum_{i=1}^d|n_i|\). It is assumed that \(\{v_n\}_{n\in {\Lambda}}\in \ell^\infty(\Lambda)\) be a real-valued, non-constant, non-negative potential. The authors investigate the normalized integrated density of states (IDOS) of \(H\), i.e., the eigenvalue counting function per unit volume: \[ N(\mu):=K^{-d} \times \left\{\text{the number of eigenvalues } \lambda \text{ of }H\text{ such that }\lambda\le \mu \right\} . \] The following upper and lower bounds for IDOS in terms of the landscape box counting function \(N_u\) are proved \[ c_3N_u(c_2\mu)\leq N(\mu) \leq N_u(C_1\mu),\ \forall \mu > 0, \] where \(C_1,c_2,c_3\) are positive dimensional constants. The lower bound is proved under the additional assumption that \(u\) satisfies a doubling type condition (this condition holds, for example, whenever \(V\) is periodic). The function \(N_u\) is defined as follows. For any positive integer \(s\) let \(\mathcal{P}(s)\) be a partition of the set \(\{1,\dots,K\}^d\) into subsets which are mostly boxes of side length \(s\). For the operator \(H\) the discrete localization landscape function is the unique solution \(u=\{u_n\}_{n\in \Lambda}\in \ell^2({\Lambda})\) to the equation \((H u)_n=1\). For given \(\mu>0\) set \(s(\mu)={\mu^{-1/2}}\) and define \(N_u(\mu)\) as the number of boxes on which the minimum of \(1/u_n\) does not exceed \(\mu\), normalized by the size of the set \(\Lambda\): \[ N_u(\mu)=K^{-d}\times \left\{ \text{the number of } Q\in \mathcal{P}\big(s(\mu)\big) \text{ such that }\min_{n\in Q}\frac{1}{u_n}\le \mu \right\}. \] The upper and lower bounds are also proved for disordered potentials, i.e., the values \(\{v_n\}_{n\in\Lambda}\) are given by independent, identically distributed random variables, with common probability measure \(P_0\) on \(\mathbb R\). In this case, bounds are proved for the expectations of IDOS with respect to the product measure on \(\mathbb{R}^{|\Lambda|}\) generated by \(P_0\).