Combining phonon accuracy with high transferability in Gaussian approximation potential models

Abstract

Machine learning driven interatomic potentials, including Gaussian approximation potential (GAP) models, are emerging tools for atomistic simulations. Here, we address the methodological question of how one can fit GAP models that accurately predict vibrational properties in specific regions of configuration space, whilst retaining flexibility and transferability to others. We use an adaptive regularization of the GAP fit that scales with the absolute force magnitude on any given atom, thereby exploring the Bayesian interpretation of GAP regularization as an "expected error", and its impact on the prediction of physical properties for a material of interest. The approach enables excellent predictions of phonon modes (to within 0.1-0.2 THz) for structurally diverse silicon allotropes, and it can be coupled with existing fitting databases for high transferability. These findings and workflows are expected to be useful for GAP-driven materials modeling more generally.

11 pages, 6 figures, 2 tables; submitted to The Journal of Chemical Physics

References in corpus (35)

• Gaussian Approximation Potentials: the accuracy of quantum mechanics, without the electrons
• On representing chemical environments
• Deep Potential Molecular Dynamics: a scalable model with the accuracy of quantum mechanics
• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• High-throughput electronic band structure calculations: challenges and tools
• Moment Tensor Potentials: a class of systematically improvable interatomic potentials
• Distribution of phonon lifetime in Brillouin zone
• Machine Learning of Accurate Energy-Conserving Molecular Force Fields
• A Spectral Analysis Method for Automated Generation of Quantum-Accurate Interatomic Potentials
• A Performance and Cost Assessment of Machine Learning Interatomic Potentials
• DScribe: Library of Descriptors for Machine Learning in Materials Science
• Machine-learning based interatomic potential for amorphous carbon
• Temperature dependent effective potential method for accurate free energy calculations of solids
• AiiDA: Automated Interactive Infrastructure and Database for Computational Science
• Active Learning of Uniformly Accurate Inter-atomic Potentials for Materials Simulation
• Machine learning a general purpose interatomic potential for silicon
• Discovery of low thermal conductivity compounds with first-principles anharmonic lattice dynamics calculations and Bayesian optimization
• Interatomic potentials for ionic systems with density functional accuracy based on charge densities obtained by a neural network
• High-Throughput Computational Screening of thermal conductivity, Debye temperature and Grüneisen parameter using a quasi-harmonic Debye Model
• Thermal physics of the lead chalcogenides PbS, PbSe, and PbTe from first principles
• De novo exploration and self-guided learning of potential-energy surfaces
• A Machine Learning Potential for Graphene
• Optimizing many-body atomic descriptors for enhanced computational performance of machine learning based interatomic potentials
• Complex low energy tetrahedral polymorphs of group IV elements from first-principles
• Feature Optimization for Atomistic Machine Learning Yields A Data-Driven Construction of the Periodic Table of the Elements
• Ab initio quality neural-network potential for sodium
• Low-energy tetrahedral polymorphs of carbon, silicon, and germanium
• Lattice thermal conductivities of two SiO${}_2$ polymorphs by first-principles calculation and phonon Boltzmann transport equation
• First-principles interatomic potentials for ten elemental metals via compressed sensing
• Low-density silicon allotropes for photovoltaic applications
• Machine-learning based interatomic potential for phonon transport in perfect crystalline Si and crystalline Si with vacancies
• Temperature Effect on Phonon Dispersion Stability of Zirconium by Machine Learning-driven Atomistic Simulations
• Quantifying Chemical Structure and Atomic Energies in Amorphous Silicon Networks
• Simulating lattice thermal conductivity in semiconducting materials using high-dimensional neural network potential
• Prediction of new sp3 silicon and germanium allotropes from the topology-based multiscale method

Cited by in corpus (11)

• Neuroevolution machine learning potentials: Combining high accuracy and low cost in atomistic simulations and application to heat transport
• How to validate machine-learned interatomic potentials
• BIGDML: Towards Exact Machine Learning Force Fields for Materials
• Accurate large-scale simulations of siliceous zeolites by neural network potentials
• Imperfections are not 0 K: free energy of point defects in crystals
• A reactive neural network framework for water-loaded acidic zeolites
• Indirect Learning of Interatomic Potentials for Accelerated Materials Simulations
• Machine learning based modeling of disordered elemental semiconductors: understanding the atomic structure of a-Si and a-C
• Broad-Range Directional Detection of Light Dark Matter in Cryogenic Ice
• Accurate Machine Learning Interatomic Potentials for Polyacene Molecular Crystals: Application to Single Molecule Host-Guest Systems
• Machine-Learned Interatomic Potential for Predictive Simulation of MoS2 Epitaxy