Critical Exponent of the Anderson Transition using Massively Parallel Supercomputing

TL;DR

This paper introduces a parallelizable method for estimating the critical exponent of the Anderson transition, enabling larger system simulations and more precise results using supercomputing resources.

Contribution

The authors develop a scalable simulation approach using random orthogonal vectors, improving the estimation of the critical exponent in three-dimensional Anderson transition studies.

Findings

Extended the largest simulated system size to L=64

Achieved more precise critical exponent estimation

Method is adaptable to correlated disorder systems

Abstract

To date the most precise estimations of the critical exponent for the Anderson transition have been made using the transfer matrix method. This method involves the simulation of extremely long quasi one-dimensional systems. The method is inherently serial and is not well suited to modern massively parallel supercomputers. The obvious alternative is to simulate a large ensemble of hypercubic systems and average. While this permits taking full advantage of both OpenMP and MPI on massively parallel supercomputers, a straight forward implementation results in data that does not scale. We show that this problem can be avoided by generating random sets of orthogonal starting vectors with an appropriate stationary probability distribution. We have applied this method to the Anderson transition in the three-dimensional orthogonal universality class and been able to increase the largest $L\times L$ cross section simulated from $L=24$ (New J. Physics, 16, 015012 (2014)) to $L=64$ here. This permits an estimation of the critical exponent with improved precision and without the necessity of introducing an irrelevant scaling variable. In addition, this approach is better suited to simulations with correlated random potentials such as is needed in quantum Hall or cold atom systems.