Disorder, Path Integrals and Localization

TL;DR

This paper derives Anderson localization directly from the path integral formulation of quantum mechanics with a Gaussian random potential, providing explicit formulas for localization length and connecting path integrals to Green's functions.

Contribution

It introduces a novel derivation of Anderson localization from path integrals and explicitly relates localization length to potential correlation properties.

Findings

Localization length formula in 1D derived from path integrals

Explicit demonstration of exponential decay via closed loops in paths

Connection established between path integral approach and Green's function methods

Abstract

Anderson localization is derived directly from the path integral representation of quantum mechanics in the presence of a random potential energy function. The probability distribution of the potential energy is taken to be a Gaussian in function space with a given autocorrelation function. Averaging the path integral itself we find that the localization length, in one-dimension, is given by (E_{{\xi}}/{\sigma})(KE_{cl}/{\sigma}){\xi} where E_{{\xi}} is the "correlation energy", KE_{cl} the average classical kinetic energy, {\sigma} the root-mean-square variation of the potential energy and {\xi} the autocorrelation length. Averaging the square of the path integral shows explicitly that closed loops in the path when traversed forward and backward in time lead to exponential decay, and hence localization. We also show how, using Schwinger proper time, the path integral result can be directly related to the Greens function commonly used to study localization.