A Divide-and-Conquer Algorithm for Disordered and Interacting Few-Particle Systems in One Dimension

TL;DR

This paper introduces a scalable divide-and-conquer algorithm for analyzing large one-dimensional disordered and interacting few-particle systems, enabling detailed eigenfunction studies up to billion-site scales.

Contribution

The authors develop a linear-scaling algorithm exploiting eigenfunction localization, allowing analysis of extremely large disordered systems and extending to interacting particles.

Findings

Able to compute all eigenfunctions for systems with up to one billion sites

Revealed detailed eigenfunction property histograms showing rich structures

Extended the method to interacting two-particle systems in disorder environments

Abstract

We present an algorithm to solve very large one-dimensional disordered and interacting few-particle systems. Our approach exploits the localized nature of the eigenfunctions in real space to achieve a linear scaling with the total system size $L$. This allows us to solve for all eigenfunctions of single-particle systems with different types of disorder up to one billion sites. Based on this technology we collect very detailed histograms of properties of eigenfunctions, such as the localization length or the participation ratio as a function of their energy. These histograms reveal surprisingly rich fine structures, whose origins we discuss. We also apply the algorithm to single particle problems where not all eigenfunctions are localized and show how this is diagnosed. Finally we extend the algorithm to interacting two-particle problems in the presence of disorder and demonstrate that our algorithm is well suited to analyze the effect of interactions on wavefunctions.