A parallel algorithm for Hamiltonian matrix construction in electron-molecule collision calculations: MPI-SCATCI

TL;DR

This paper presents a parallel algorithm for constructing Hamiltonian matrices in electron-molecule collision calculations, significantly speeding up computations and enabling larger, more complex simulations using modern multi-CPU architectures.

Contribution

The authors re-engineer a previous Hamiltonian construction algorithm to leverage parallel computing, improving performance in electron-molecule collision calculations.

Findings

Significant speed-ups with multiple CPUs.

Enables higher collision energies and larger molecules.

Applicable to various molecular physics studies.

Abstract

Construction and diagonalization of the Hamiltonian matrix is the rate-limiting step in most low-energy electron -- molecule collision calculations. Tennyson (J Phys B, 29 (1996) 1817) implemented a novel algorithm for Hamiltonian construction which took advantage of the structure of the wavefunction in such calculations. This algorithm is re-engineered to make use of modern computer architectures and the use of appropriate diagonalizers is considered. Test calculations demonstrate that significant speed-ups can be gained using multiple CPUs. This opens the way to calculations which consider higher collision energies, larger molecules and / or more target states. The methodology, which is implemented as part of the UK molecular R-matrix codes (UKRMol and UKRMol+) can also be used for studies of bound molecular Rydberg states, photoionisation and positron-molecule collisions.