BIGDML: Towards Exact Machine Learning Force Fields for Materials

Abstract

Machine-learning force fields (MLFF) should be accurate, computationally and data efficient, and applicable to molecules, materials, and interfaces thereof. Currently, MLFFs often introduce tradeoffs that restrict their practical applicability to small subsets of chemical space or require exhaustive datasets for training. Here, we introduce the Bravais-Inspired Gradient-Domain Machine Learning (BIGDML) approach and demonstrate its ability to construct reliable force fields using a training set with just 10-200 geometries for materials including pristine and defect-containing 2D and 3D semiconductors and metals, as well as chemisorbed and physisorbed atomic and molecular adsorbates on surfaces. The BIGDML model employs the full relevant symmetry group for a given material, does not assume artificial atom types or localization of atomic interactions and exhibits high data efficiency and state-of-the-art energy accuracies (errors substantially below 1 meV per atom) for an extended set of materials. Extensive path-integral molecular dynamics carried out with BIGDML models demonstrate the counterintuitive localization of benzene--graphene dynamics induced by nuclear quantum effects and allow to rationalize the Arrhenius behavior of hydrogen diffusion coefficient in a Pd crystal for a wide range of temperatures.

arXiv comment15 pages, 8 figures, development of methodology and applications

References in corpus (46)

• Quantum ESPRESSO: a modular and open-source software project for quantum simulations of materials
• Advanced capabilities for materials modelling with Quantum ESPRESSO
• Gaussian Approximation Potentials: the accuracy of quantum mechanics, without the electrons
• On representing chemical environments
• Fast and Accurate Modeling of Molecular Atomization Energies with Machine Learning
• SchNet - a deep learning architecture for molecules and materials
• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• Machine Learning Force Fields
• Machine Learning of Accurate Energy-Conserving Molecular Force Fields
• The interlayer cohesive energy of graphite from thermal desorption of polyaromatic hydrocarbons
• PhysNet: A Neural Network for Predicting Energies, Forces, Dipole Moments and Partial Charges
• Negative Thermal Expansion Coefficient of Graphene Measured by Raman Spectroscopy
• Machine learning for molecular simulation
• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems
• Machine Learning Unifies the Modelling of Materials and Molecules
• Towards Exact Molecular Dynamics Simulations with Machine-Learned Force Fields
• Long-range correlation energy calculated from coupled atomic response functions
• Machine Learning of Molecular Electronic Properties in Chemical Compound Space
• A Fourth-Generation High-Dimensional Neural Network Potential with Accurate Electrostatics Including Non-local Charge Transfer
• Physics-inspired structural representations for molecules and materials
• Nuclear quantum effects enter the mainstream
• How to represent crystal structures for machine learning: towards fast prediction of electronic properties
• i-PI 2.0: A Universal Force Engine for Advanced Molecular Simulations
• Sieving hydrogen isotopes through two dimensional crystals
• SpookyNet: Learning Force Fields with Electronic Degrees of Freedom and Nonlocal Effects
• Alchemical and structural distribution based representation for improved QML
• Physisorption of Nucleobases on Graphene
• Interatomic potentials for ionic systems with density functional accuracy based on charge densities obtained by a neural network
• An Accurate and Transferable Machine Learning Potential for Carbon
• sGDML: Constructing Accurate and Data Efficient Molecular Force Fields Using Machine Learning
• A Machine Learning Potential for Graphene
• Density-functional model for van der Waals interactions: Unifying many-body atomic approaches with nonlocal functionals
• Atom-Density Representations for Machine Learning
• Evidence for supercritical behavior of high-pressure liquid hydrogen
• Constructing first-principles phase diagrams of amorphous LixSi using machine-learning-assisted sampling with an evolutionary algorithm
• Coupled cluster theory in materials science
• Molecular Force Fields with Gradient-Domain Machine Learning: Construction and Application to Dynamics of Small Molecules with Coupled Cluster Forces
• Fast and accurate quantum Monte Carlo for molecular crystals
• Gaussian approximation potentials for body-centered-cubic transition metals
• Autonomous robotic nanofabrication with reinforcement learning
• Inverse Temperature Dependence of Nuclear Quantum Effects in DNA Base Pairs
• Stability of Complex Biomolecular Structures: Vander Waals, Hydrogen Bond Cooperativity, and Nuclear Quantum Effects
• Equation of state of fluid methane from first principles with machine learning potentials
• Dynamical Strengthening of Covalent and Non-Covalent Molecular Interactions by Nuclear Quantum Effects at Finite Temperature
• Combining phonon accuracy with high transferability in Gaussian approximation potential models
• Charge transfer energies of benzene physisorbed on a graphene sheet from constrained density functional theory

Cited by in corpus (9)

• Universal Machine Learning for the Response of Atomistic Systems to External Fields
• The Evolution of Machine Learning Potentials for Molecules, Reactions and Materials
• Toward Accurate Interpretable Predictions of Materials Properties within Transformer Language Models
• Towards Linearly Scaling and Chemically Accurate Global Machine Learning Force Fields for Large Molecules
• Decomposing Chemical Space: Applications to the Machine Learning of Atomic Energies
• Efficient moment tensor machine-learning interatomic potential for accurate description of defects in Ni-Al Alloys
• A Study on Quantum Graph Neural Networks Applied to Molecular Physics
• Reconstructing Kernel-based Machine Learning Force Fields with Super-linear Convergence
• Enhancing ab initio diffusion calculations in materials through Gaussian process regression