Source-linked AI summary
Data-driven learning of total and local energies in elemental boron
Volker L. Deringer, Chris J. Pickard, Gábor Csányi
TL;DR
Boron lacks a reliable interatomic potential spanning multiple phases, while DFT-based structure searching is computationally expensive. The paper iteratively combines GAP fitting with GAP-driven random structure searching and single-point DFT data to construct such a potential, which describes multiple allotropes and provides local-energy insight into β-boron.
Problem
Boron lacks an interatomic potential that reliably describes the potential-energy surface across multiple phases, limiting efficient structure searching.
Method
The authors iteratively alternate single-point DFT energy and force calculations, GAP fitting, and GAP-driven random structure searching to explore and fit boron's potential-energy surface.
Results
The resulting potential describes the energetics of multiple boron polymorphs and captures qualitative and quantitative trends for unseen β-boron occupation models.
Takeaways & Limitations
Local-energy predictions identify unfavorable fully occupied B13 sites and show how vacancies alter neighboring and central-atom energetics in β-boron.
Abstract
from arXiv · showhide
The allotropes of boron continue to challenge structural elucidation and solid-state theory. Here we use machine learning combined with random structure searching (RSS) algorithms to systematically construct an interatomic potential for boron. Starting from ensembles of randomized atomic configurations, we use alternating single-point quantum-mechanical energy and force computations, Gaussian approximation potential (GAP) fitting, and GAP-driven RSS to iteratively generate a representation of the element's potential-energy surface. Beyond the total energies of the very different boron allotropes, our model readily provides atom-resolved, local energies and thus deepened insight into the frustrated $β$-rhombohedral boron structure. Our results open the door for the efficient and automated generation of GAPs and other machine-learning-based interatomic potentials, and suggest their usefulness as a tool for materials discovery.