Anomalously Localized States in the Anderson Model

TL;DR

This paper investigates anomalously localized states (ALS) in the Anderson lattice model, revealing lattice-specific ALS near the band center and clarifying discrepancies with continuous models through analytical and numerical analysis.

Contribution

It introduces a new type of lattice-specific ALS near the band center and provides an analytical calculation of their likelihood, explaining previous numerical inconsistencies.

Findings

Lattice-specific ALS exist near the band center, E=0.

Analytical likelihood of these ALS is derived.

Discrepancies with continuous models are explained by lattice effects.

Abstract

In a diffusive conductor the eigenstates are spread over the entire sample. However, with certain probability, an anomalously localized state (ALS) can occur, i.e. the wave function assumes anomalously large values in some region of space. Existing analytical theories of ALS are based on models described by a continuous (Gaussian) random potential. In the present paper we study ALS in a lattice (Anderson) model. We demonstrate that close to the center of the band, E=0, a new type of ALS exist and calculate analytically their likelihood. These ALS are lattice-specific and have no analog in the continuum. Our findings are relevant to numerical simulations, which are necessarily performed on a lattice. We demonstrate that inconsistencies with "continuous" results reported in the previous numerical work on ALS can be explained within our analytical theory. Finally, we point out that, in order to compare the numerics with the "continuous" ALS theories, simulations must be carried out not too far from the band edges, within the band, where the continuous description applies. Simulations performed for $E$ close to the band center reveal lattice-specific ALS that do not exist in continuous models.