How to validate machine-learned interatomic potentials.
TL;DR: In this paper , the authors review the basic principles behind ML potentials and their validation for atomic-scale material modeling and discuss the best practice in defining error metrics based on numerical performance, as well as physically guided validation.
read more
Abstract: Machine learning (ML) approaches enable large-scale atomistic simulations with near-quantum-mechanical accuracy. With the growing availability of these methods, there arises a need for careful validation, particularly for physically agnostic models-that is, for potentials that extract the nature of atomic interactions from reference data. Here, we review the basic principles behind ML potentials and their validation for atomic-scale material modeling. We discuss the best practice in defining error metrics based on numerical performance, as well as physically guided validation. We give specific recommendations that we hope will be useful for the wider community, including those researchers who intend to use ML potentials for materials "off the shelf."
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
Citations
How to train a neural network potential
Alea Miako Tokita,Jörg Behler +1 more
TL;DR: This Tutorial describes the necessary key steps for training reliable MLPs, from data generation via training to final validation, which is general and applicable to many types of MLPs.
Single-model uncertainty quantification in neural network potentials does not consistently outperform model ensembles
Aik Rui Tan,Shingo Urata,Samuel Goldman,Johannes C. B. Dietschreit,Rafael Gómez‐Bombarelli +4 more
TL;DR: Single-model uncertainty quantification in neural network potentials does not consistently outperform model ensembles.
Machine learned force-fields for an Ab-initio quality description of metal-organic frameworks
Sandro Wieser,Egbert Zojer +1 more
TL;DR: This work builds on existing types of machine-learned force fields belonging to the moment-tensor and kernel-based potential families to develop a recipe for their efficient parametrization that yields exceptionally accurate and computationally highly efficient force fields.
20
Stability and lifetime of diffusion-trapped oxygen in oxide-derived copper CO2 reduction electrocatalysts
Zan Lian,Federico Dattila,N. López +2 more
TL;DR: Oxide-derived Cu electrocatalysts exhibit pH-dependent oxygen concentration, with trapped oxygen stable under specific conditions, but surface oxygen atoms are unstable under common experimental conditions, influencing C-C coupling in CO2 reduction reactions.
18
Data Generation for Machine Learning Interatomic Potentials and Beyond
Maksim Kulichenko,Benjamin Nebgen,Nicholas Lubbers,Justin S. Smith,Kipton Barros,Alice Allen,Adela Habib,Emily Shinkle,Nikita Fedik,Ying Wai Li,Richard A. Messerly,Sergei Tretiak +11 more
TL;DR: This review explores the evolution of data-driven chemistry, focusing on machine learning-based interatomic potentials and the importance of high-quality training data for accurate modeling of chemical and structural properties at the atomic level.
17
References
Visualization and analysis of atomistic simulation data with OVITO–the Open Visualization Tool
TL;DR: The Open Visualization Tool (OVITO) as discussed by the authors is a 3D visualization software designed for post-processing atomistic data obtained from molecular dynamics or Monte Carlo simulations, which is written in object-oriented C++, controllable via Python scripts and easily extendable through a plug-in interface.
13.1K
Commentary: The Materials Project: A materials genome approach to accelerating materials innovation
Anubhav Jain,Shyue Ping Ong,Geoffroy Hautier,Wei-Wei Chen,William D. Richards,Stephen Dacek,Shreyas Cholia,Dan Gunter,David Skinner,Gerbrand Ceder,Kristin A. Persson +10 more
TL;DR: The Materials Project (www.materialsproject.org) is a core program of the Materials Genome Initiative that uses high-throughput computing to uncover the properties of all known inorganic materials as discussed by the authors.
First principles phonon calculations in materials science
Atsushi Togo,Isao Tanaka +1 more
TL;DR: In this paper, the authors demonstrate phonon properties with fundamental equations and show examples how the phonon calculations are applied in materials science, and demonstrate the importance of first principles phonon calculation in dynamical behaviors and thermal properties.
9.7K
Computer simulation of local order in condensed phases of silicon
TL;DR: A model potential-energy function comprising both two- and three-atom contributions is proposed to describe interactions in solid and liquid forms of Si, suggesting a temperature-independent inherent structure underlies the liquid phase, just as for ``simple'' liquids with only pair interactions.
5.3K
Generalized neural-network representation of high-dimensional potential-energy surfaces.
Jörg Behler,Michele Parrinello +1 more
TL;DR: A new kind of neural-network representation of DFT potential-energy surfaces is introduced, which provides the energy and forces as a function of all atomic positions in systems of arbitrary size and is several orders of magnitude faster than DFT.
4.3K