A linear-scaling algorithm for rapid computation of inelastic transitions in the presence of multiple electron scattering

TL;DR

This paper presents a linear-scaling algorithm that significantly accelerates the computation of inelastic electron transitions in STEM-EELS simulations, enabling practical analysis of large, complex materials with high accuracy.

Contribution

The authors develop a novel linear-scaling algorithm for inelastic transition calculations in STEM-EELS, reducing computation time by an order of magnitude without sacrificing accuracy.

Findings

Achieved a speed-up from 80 days to 16 hours for a nanoparticle simulation.

Maintained accuracy comparable to traditional quadratic-scaling methods.

Demonstrated applicability to large-scale, heterogeneous systems in materials science.

Abstract

Strong multiple scattering of the probe in scanning transmission electron microscopy (STEM) means image simulations are usually required for quantitative interpretation and analysis of elemental maps produced by electron energy-loss spectroscopy (EELS). These simulations require a full quantum-mechanical treatment of multiple scattering of the electron beam, both before and after a core-level inelastic transition. Current algorithms scale quadratically and can take up to a week to calculate on desktop machines even for simple crystal unit cells and do not scale well to the nano-scale heterogeneous systems that are often of interest to materials science researchers. We introduce an algorithm with linear scaling that typically results in an order of magnitude reduction in compute time for these calculations without introducing additional error and discuss approximations that further improve computational scaling for larger scale objects with modest penalties in calculation error. We demonstrate these speed-ups by calculating the atomic resolution STEM-EELS map using the L-edge transition of Fe, for of a nanoparticle 80 \AA\ in diameter in 16 hours, a calculation that would have taken at least 80 days using a conventional multislice approach.