Deep Potential: A General Representation of a Many-Body Potential Energy Surface

DOI10.4208/CICP.OA-2017-0213MaRDI QIDQ5159341zbMATH OpenOpenAlexWikidataFDO

Authors Jiequn Han, Linfeng Zhang, Roberto Car, Weinan E

Publication date 27 October 2021

Published in Communications in Computational Physics (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1707.01478

zbMATH Keywords

deep learningmolecular simulationpotential energy surface

Mathematics Subject Classification ID

Artificial intelligence (68T99) Molecular physics (81V55) Molecular structure (graph-theoretic methods, methods of differential topology, etc.) (92E10)

Abstract: We present a simple, yet general, end-to-end deep neural network representation of the potential energy surface for atomic and molecular systems. This methodology, which we call Deep Potential, is "first-principle" based, in the sense that no ad hoc approximations or empirical fitting functions are required. The neural network structure naturally respects the underlying symmetries of the systems. When tested on a wide variety of examples, Deep Potential is able to reproduce the original model, whether empirical or quantum mechanics based, within chemical accuracy. The computational cost of this new model is not substantially larger than that of empirical force fields. In addition, the method has promising scalability properties. This brings us one step closer to being able to carry out molecular simulations with accuracy comparable to that of quantum mechanics models and computational cost comparable to that of empirical potentials.

Recommendations

• The role of depth and flatness of a potential energy surface in chemical reaction dynamics
• Potential energy surface exploration with equilibrial paths. I: Theory
• Potential energy surface exploration with equilibrial paths. II: Application
• Exploring the potential energy landscape over a large parameter-space
• The many-body problem with an energy-dependent confining potential
• scientific article; zbMATH DE number 2010155
• Born-Oppenheimer potential energy surfaces for Kohn-Sham models in the local density approximation
• scientific article; zbMATH DE number 17115

Cites work

• A Liouville-operator derived measure-preserving integrator for molecular dynamics simulations in the isothermal–isobaric ensemble
• Deep learning
• Fast parallel algorithms for short-range molecular dynamics

Cited in

(27)

• SwitchNet: a neural Network Model for forward and inverse scattering problems
• Machine learning materials physics: integrable deep neural networks enable scale bridging by learning free energy functions
• Proper orthogonal descriptors for multi-element chemical systems
• Machine learning and computational mathematics
• Deep Learning Unravels a Dynamic Hierarchy While Empowering Molecular Dynamics Simulations
• FLAME: a library of atomistic modeling environments
• An equivariant neural operator for developing nonlocal tensorial constitutive models
• Subspace decomposition based DNN algorithm for elliptic type multi-scale PDEs
• A general code for fitting global potential energy surfaces via CHIPR method: direct-fit diatomic and tetratomic molecules
• Deep Density: circumventing the Kohn-Sham equations via symmetry preserving neural networks
• Molecular modeling by machine learning
• A Framework for a Generalization Analysis of Machine-Learned Interatomic Potentials
• Solving inverse problems with deep learning
• Better approximations of high dimensional smooth functions by deep neural networks with rectified power units
• Frequency Principle: Fourier Analysis Sheds Light on Deep Neural Networks
• Frame-independent vector-cloud neural network for nonlocal constitutive modeling on arbitrary grids
• DP-GEN: a concurrent learning platform for the generation of reliable deep learning based potential energy models
• Solving traveltime tomography with deep learning
• Proper orthogonal descriptors for efficient and accurate interatomic potentials
• Physics-constrained, data-driven discovery of coarse-grained dynamics
• Data-driven, physics-based feature extraction from fluid flow fields using convolutional neural networks
• Neural networks-based variationally enhanced sampling
• Solving many-electron Schrödinger equation using deep neural networks
• Multi-scale deep neural network (MscaleDNN) for solving Poisson-Boltzmann equation in complex domains
• The deep Ritz method: a deep learning-based numerical algorithm for solving variational problems
• Using neural networks to accelerate the solution of the Boltzmann equation
• A mathematical perspective of machine learning

Describes a project that uses

Uses Software

• ReaxFF
• i-pi
• Adam
• ImageNet
• Quantum Espresso

This page was built for publication: Deep Potential: A General Representation of a Many-Body Potential Energy Surface

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5159341)