Higher-order localization landscape theory of Anderson localization

Abstract

For a Hamiltonian ${\hat H}$ containing a position-dependent (disordered) potential, we introduce a sequence of landscape functions $u_n(\vec{r})$ obeying ${\hat H} u_n(\vec{r}) = u_{n-1}(\vec{r})$ with $u_0(\vec{r}) = 1$. For $n \to \infty$, $1/v_n(\vec{r}) = u_{n-1}(\vec{r})/u_{n}(\vec{r})$ converges to the lowest eigenenergy $E_1$ of ${\hat H}$ whereas $u_{\infty}(\vec{r})$ yields the corresponding wave function $\psi_1(\vec{r})$. For large but finite $n$, $v_n(\vec{r})$ can be approximated by a piecewise constant function $v_n(\vec{r}) \simeq v_n^{(m)}$ for $\vec{r} \in \Omega_m$ and yields progressively improving estimations of eigenenergies $E_m = 1/v_n^{(m)}$ of locally fundamental eigenstates $\psi_m(\vec{r}) \propto u_{n}(\vec{r})$ in spatial domains $\Omega_m$. These general results are illustrated by a number of examples in one dimension: box potential, sequence of randomly placed infinite potential barriers, smooth and spatially uncorrelated random potentials, quasiperiodic potential, as well as for the uncorrelated random potential in two dimensions.