Active Δ-learning with universal potentials for global structure optimization

Abstract

Universal machine learning interatomic potentials (uMLIPs) have recently been formulated and shown to generalize well. When applied out-of-sample, further data collection for improvement of the uMLIPs may, however, be required. In this work we demonstrate that, whenever the envisaged use of the MLIPs is global optimization, the data acquisition can follow an active learning scheme in which a gradually updated uMLIP directs the finding of new structures, which are subsequently evaluated at the density functional theory (DFT) level. In the scheme, we augment foundation models using a Δ-model based on this new data using local SOAP-descriptors, Gaussian kernels, and a sparse Gaussian Process Regression model. We compare the efficacy of the approach with different global optimization algorithms, Random Structure Search, Basin Hopping, a Bayesian approach with competitive candidates (GOFEE), and a replica exchange formulation (REX). We further compare several foundation models, CHGNet, MACE-MP0, and MACE-MPA. The test systems are silver-sulfur clusters and sulfur-induced surface reconstructions on Ag(111) and Ag(100). Judged by the fidelity of identifying global minima, active learning with GPR-based Δ-models appears to be a robust approach. Judged by the total CPU time spent, the REX approach stands out as being the most efficient.

References in corpus (35)

• Gaussian Approximation Potentials: the accuracy of quantum mechanics, without the electrons
• On representing chemical environments
• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials
• Parallel Tempering Algorithm for Conformational Studies of Biological Molecules
• A Universal Graph Deep Learning Interatomic Potential for the Periodic Table
• Big Data meets Quantum Chemistry Approximations: The $\Delta$-Machine Learning Approach
• DScribe: Library of Descriptors for Machine Learning in Materials Science
• On-the-fly machine learning force field generation: Application to melting points
• Active Learning of Uniformly Accurate Inter-atomic Potentials for Materials Simulation
• Machine-learning interatomic potentials for materials science
• GPAW: An open Python package for electronic-structure calculations
• The first-principles phase diagram of monolayer nanoconfined water
• De novo exploration and self-guided learning of potential-energy surfaces
• Data-driven learning of total and local energies in elemental boron
• Efficient global structure optimization with a machine learned surrogate model
• $\Delta$-Machine Learning for Potential Energy Surfaces: A PIP approach to bring a DFT-based PES to CCSD(T) Level of Theory
• Performance Assessment of Universal Machine Learning Interatomic Potentials: Challenges and Directions for Materials' Surfaces
• Updates to the DScribe Library: New Descriptors and Derivatives
• Replica-exchange molecular dynamics simulation for supercooled liquids
• Global optimization of atomistic structure enhanced by machine learning
• IrO2 Surface Complexions Identified Through Machine Learning and Surface Investigations
• The Evolution of Machine Learning Potentials for Molecules, Reactions and Materials
• Learning Smooth and Expressive Interatomic Potentials for Physical Property Prediction
• Machine Learning and Data-Driven Methods in Computational Surface and Interface Science
• Machine-learning interatomic potentials from a users perspective: A comparison of accuracy, speed and data efficiency
• Atomistic structure search using local surrogate mode
• A machine learning potential for simulating infrared spectra of nanosilicate clusters
• Thermal Conductivity Predictions with Foundation Atomistic Models
• Augmentation of Universal Potentials for Broad Applications
• Foundation for the ΔSCF Approach in Density Functional Theory
• Cascading symmetry constraint during machine learning-enabled structural search for sulfur induced Cu(111)-$(\sqrt{43}\times \sqrt{43})$ surface reconstruction
• $\Delta$-model correction of Foundation Model based on the models own understanding
• Replica exchange nested sampling
• RAFFLE: Active learning accelerated interface structure prediction