The Localization Dichotomy for gapped periodic quantum systems

TL;DR

This paper establishes a fundamental dichotomy in the localization properties of gapped periodic quantum systems in dimensions up to three, linking topological triviality to the finiteness of Wannier functions' localization.

Contribution

It proves a general localization dichotomy for such systems, showing that topologically non-trivial phases necessarily have delocalized Wannier functions, applicable across various models.

Findings

Topologically trivial systems have localized Wannier functions.

Topologically non-trivial systems have delocalized Wannier functions.

The localization functional diverges in non-trivial phases.

Abstract

We investigate the localization properties of gapped periodic quantum systems, modeled by a periodic or covariant family of projectors, as e.g. the orthogonal projectors on the occupied orbitals at fixed crystal momentum for a gas of non-interacting electrons. We prove a general localization dichotomy for dimension $d\leq 3$: either the system is topologically trivial i.e. all the Chern numbers vanish, or any arbitrary choice of composite Wannier functions yields an infinite expectation value for the squared position operator. Equivalently, in the topologically non-trivial phase, the localization functional introduced by Marzari and Vanderbilt diverges, as already noticed in the case of the Haldane model. Our result is formulated by using only the relevant symmetries of the system, and it is thus largely model-independent. Possible applications include both tight-binding and continuous models of crystalline solids, cold gases in optical lattices as well as flat band superconductivity.