GPUMD: A package for constructing accurate machine-learned potentials and performing highly efficient atomistic simulations

Abstract

We present our latest advancements of machine-learned potentials (MLPs) based on the neuroevolution potential (NEP) framework introduced in [Fan et al., Phys. Rev. B 104, 104309 (2021)] and their implementation in the open-source package GPUMD. We increase the accuracy of NEP models both by improving the radial functions in the atomic-environment descriptor using a linear combination of Chebyshev basis functions and by extending the angular descriptor with some four-body and five-body contributions as in the atomic cluster expansion approach. We also detail our efficient implementation of the NEP approach in graphics processing units as well as our workflow for the construction of NEP models, and we demonstrate their application in large-scale atomistic simulations. By comparing to state-of-the-art MLPs, we show that the NEP approach not only achieves above-average accuracy but also is far more computationally efficient. These results demonstrate that the GPUMD package is a promising tool for solving challenging problems requiring highly accurate, large-scale atomistic simulations. To enable the construction of MLPs using a minimal training set, we propose an active-learning scheme based on the latent space of a pre-trained NEP model. Finally, we introduce three separate Python packages, GPYUMD, CALORINE, and PYNEP, which enable the integration of GPUMD into Python workflows.

29 pages, 15 figures, code and data available

References in corpus (23)

• Canonical sampling through velocity-rescaling
• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• A Spectral Analysis Method for Automated Generation of Quantum-Accurate Interatomic Potentials
• Pressure control using stochastic cell rescaling
• Machine learning for molecular simulation
• Machine-learning based interatomic potential for amorphous carbon
• Accurate sampling using Langevin dynamics
• Efficient and Accurate Machine-Learning Interpolation of Atomic Energies in Compositions with Many Species
• Physically Motivated Recursively Embedded Atom Neural Networks: Incorporating Local Completeness and Nonlocality
• Accelerated molecular dynamics force evaluation on graphics processing units for thermal conductivity calculations
• PiNN: A Python Library for Building Atomic Neural Networks of Molecules and Materials
• Recursive evaluation and iterative contraction of $N$-body equivariant features
• Kapitza thermal resistance across individual grain boundaries in graphene
• Reduced Thermal Conductivity of Supported and Encased Monolayer and Bilayer MoS${}_2$
• Bimodal grain-size scaling of thermal transport in polycrystalline graphene from large-scale molecular dynamics simulations
• Thermal conductivity reduction in carbon nanotube by fullerene encapsulation: A molecular dynamics study
• Pushing the limit of molecular dynamics with ab initio accuracy to 100 million atoms with machine learning
• Mode Localization and Suppressed Heat Transport in Amorphous Alloys
• Heat transport across graphene/hexagonal-BN tilted grain boundaries from phase-field crystal model and molecular dynamics simulations
• A minimal Tersoff potential for diamond silicon with improved descriptions of elastic and phonon transport properties
• Interpretation of apparent thermal conductivity in finite systems from equilibrium molecular dynamics simulations
• Exactly equivalent thermal conductivity in finite systems from equilibrium and nonequilibrium molecular dynamics simulations
• REANN: A PyTorch-based End-to-End Multi-functional Deep Neural Network Package for Molecular, Reactive and Periodic Systems

Cited by in corpus (59)

• DeePMD-kit v2: A software package for Deep Potential models
• General-purpose machine-learned potential for 16 elemental metals and their alloys
• Advances of Machine Learning in Materials Science: Ideas and Techniques
• Molecular dynamics simulations of heat transport using machine-learned potentials: A mini review and tutorial on GPUMD with neuroevolution potentials
• DeePMD-kit v3: A Multiple-Backend Framework for Machine Learning Potentials
• Transferable E(3) equivariant parameterization for Hamiltonian of molecules and solids
• Quantum-corrected thickness-dependent thermal conductivity in amorphous silicon predicted by machine-learning molecular dynamics simulations
• Anisotropic and high thermal conductivity in monolayer quasi-hexagonal fullerene: A comparative study against bulk phase fullerene
• Phase transitions in inorganic halide perovskites from machine learning potentials
• Sub-micrometer phonon mean free paths in metal-organic frameworks revealed by machine-learning molecular dynamics simulations
• The Evolution of Machine Learning Potentials for Molecules, Reactions and Materials
• Accurate prediction of heat conductivity of water by a neuroevolution potential
• Mechanisms of temperature-dependent thermal transport in amorphous silica from machine-learning molecular dynamics
• Limits of the phonon quasi-particle picture at the cubic-to-tetragonal phase transition in halide perovskites
• Tuning the lattice thermal conductivity in van-der-Waals structures through rotational (dis)ordering
• Quantifying Dynamic Tilting in Halide Perovskites: Chemical Trends and Local Correlations
• Correcting force error-induced underestimation of lattice thermal conductivity in machine learning molecular dynamics
• E(n)-Equivariant Cartesian Tensor Passing Potential
• Advances in modeling complex materials: The rise of neuroevolution potentials
• Revealing the free energy landscape of halide perovskites: Metastability and transition characters in CsPbBr${}_3$ and MAPbI${}_3$
• Moiré-Driven Interfacial Thermal Transport in Twisted Transition Metal Dichalcogenides
• Machine Learning for Polaritonic Chemistry: Accessing chemical kinetics
• Highly efficient path-integral molecular dynamics simulations with GPUMD using neuroevolution potentials: Case studies on thermal properties of materials
• Tensorial properties via the neuroevolution potential framework: Fast simulation of infrared and Raman spectra
• PYSED: A tool for extracting kinetic-energy-weighted phonon dispersion and lifetime from molecular dynamics simulations
• Combining linear-scaling quantum transport and machine-learning molecular dynamics to study thermal and electronic transports in complex materials
• Accurate estimation of interfacial thermal conductance between silicon and diamond enabled by a machine learning interatomic potential
• Variable thermal transport in black, blue, and violet phosphorene from extensive atomistic simulations with a neuroevolution potential
• Combining the D3 dispersion correction with the neuroevolution machine-learned potential
• chemtrain: Learning Deep Potential Models via Automatic Differentiation and Statistical Physics
• Dissimilar thermal transport properties in $\kappa$-Ga${}_2$O${}_3$ and $\beta$-Ga${}_2$O${}_3$ revealed by machine-learning homogeneous nonequilibrium molecular dynamics simulations
• Probing the ideal limit of interfacial thermal conductance in two-dimensional van der Waals heterostructures
• Phonon coherence and minimum thermal conductivity in disordered superlattice
• Impact of Surface Adsorbates and Dimensionality on Templating of Halide Perovskites
• Revealing the Low Temperature Phase of FAPbI${}_3$ using A Machine-Learned Potential
• Copper delocalization leads to ultralow thermal conductivity in chalcohalide CuBiSeCl2
• Dynasor 2: From Simulation to Experiment Through Correlation Functions
• Efficient moment tensor machine-learning interatomic potential for accurate description of defects in Ni-Al Alloys
• Exploring the energy landscape of aluminas through machine learning interatomic potential
• Machine learning based modeling of disordered elemental semiconductors: understanding the atomic structure of a-Si and a-C
• Optical line shapes of color centers in solids from classical autocorrelation functions
• Unique Hierarchical Rotational Dynamics Induces Ultralow Lattice Thermal Conductivity in Cyanide-bridged Framework Materials
• Consistency between the Green-Kubo formula and Lorentz model for predicting the infrared dielectric function of polar materials
• Predicting neutron experiments from first principles: A workflow powered by machine learning
• Anisotropic and isotropic elasticity and thermal transport in monolayer C${}_{24}$ networks from machine-learning molecular dynamics
• Guest Editorial: Special Topic on Software for Atomistic Machine Learning
• Efficient GPU-Accelerated Training of a Neuroevolution Potential with Analytical Gradients
• Modular hybrid machine learning and physics-based potentials for scalable modeling of van der Waals heterostructures
• Role of octahedral tilting induced acoustic softening on limiting thermal transport in SrSnO3
• Accelerating the Discovery of Materials with Expected Thermal Conductivity via a Synergistic Strategy of DFT and Interpretable Deep Learning
• Decoding lithium's subtle phase stability with a machine learning force field
• Uniaxial stress tuning of interfacial thermal conductance in cubic BAs/4H-SiC heterostructures
• Thermal conductivities of monolayer graphene oxide from machine learning molecular dynamics simulations
• LLM-driven discovery for carbon allotropes with bond-network entropy
• Suitability of available interatomic potentials for Sn to model its 2D allotropes
• Interface phonon modes governing the ideal limit of thermal transport across diamond/cubic boron nitride interfaces
• Sliding and superlubric moiré twisting ferroelectric transition in HfO2
• A review of simulation, measurement techniques, and development in chip thermal design
• Predicting co-segregation in alloys with solute-solute interactions