Parallel exact diagonalization solver for quantum-electron models

TL;DR

This paper introduces a parallel eigenvalue solver for quantum-electron models using Arnoldi algorithm, enabling efficient large-scale computations and analysis of electronic properties in complex materials.

Contribution

It develops a novel parallel computation scheme with data transfer and storage optimizations for exact diagonalization in quantum-electron models.

Findings

Good performance and scalability demonstrated.

Applied to study Sr2RuO4's electronic properties.

Analyzed the impact of spin-flip terms.

Abstract

We present a parallel computation scheme based on the Arnoldi algorithm for exact diagonalization of quantum-electron models. It contains a selective data transferring method and distributed storage format for efficient computing of the matrix-vector product on distributed computing systems. The performed numerical experiments demonstrated good performance and scalability of our eigenvalue solver. The developed technique has been applied to investigate the electronic properties of Sr2RuO4 at experimental temperatures. The role of the spin-flip term in the electronic Hamiltonian was analyzed.