PhysNet: A Neural Network for Predicting Energies, Forces, Dipole Moments and Partial Charges

Abstract

In recent years, machine learning (ML) methods have become increasingly popular in computational chemistry. After being trained on appropriate ab initio reference data, these methods allow to accurately predict the properties of chemical systems, circumventing the need for explicitly solving the electronic Schrödinger equation. Because of their computational efficiency and scalability to large datasets, deep neural networks (DNNs) are a particularly promising ML algorithm for chemical applications. This work introduces PhysNet, a DNN architecture designed for predicting energies, forces and dipole moments of chemical systems. PhysNet achieves state-of-the-art performance on the QM9, MD17 and ISO17 benchmarks. Further, two new datasets are generated in order to probe the performance of ML models for describing chemical reactions, long-range interactions, and condensed phase systems. It is shown that explicitly including electrostatics in energy predictions is crucial for a qualitatively correct description of the asymptotic regions of a potential energy surface (PES). PhysNet models trained on a systematically constructed set of small peptide fragments (at most eight heavy atoms) are able to generalize to considerably larger proteins like deca-alanine (Ala${}_{10}$): The optimized geometry of helical Ala${}_{10}$ predicted by PhysNet is virtually identical to ab initio results (RMSD = 0.21 Å). By running unbiased molecular dynamics (MD) simulations of Ala${}_{10}$ on the PhysNet-PES in gas phase, it is found that instead of a helical structure, Ala${}_{10}$ folds into a wreath-shaped configuration, which is more stable than the helical form by 0.46 kcal mol${}^{-1}$ according to the reference ab initio calculations.

23 pages, 9 figures, 7 tables

References in corpus (6)

• ANI-1: An extensible neural network potential with DFT accuracy at force field computational cost
• On the Convergence of Adam and Beyond
• Recurrent Models of Visual Attention
• A Model of Inductive Bias Learning
• Efficient and Accurate Machine-Learning Interpolation of Atomic Energies in Compositions with Many Species
• Influence of the Leaving Group on the Dynamics of a Gas Phase SN2 Reaction

Cited by in corpus (229)

• E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials
• Machine Learning Force Fields
• Machine learning for molecular simulation
• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems
• Atomistic Line Graph Neural Network for Improved Materials Property Predictions
• Big-Data Science in Porous Materials: Materials Genomics and Machine Learning
• DeePMD-kit v2: A software package for Deep Potential models
• A Fourth-Generation High-Dimensional Neural Network Potential with Accurate Electrostatics Including Non-local Charge Transfer
• Deep neural network solution of the electronic Schrödinger equation
• GPUMD: A package for constructing accurate machine-learned potentials and performing highly efficient atomistic simulations
• Towards Universal Neural Network Potential for Material Discovery Applicable to Arbitrary Combination of 45 Elements
• Machine learning for electronically excited states of molecules
• Machine-learning interatomic potentials for materials science
• SpookyNet: Learning Force Fields with Electronic Degrees of Freedom and Nonlocal Effects
• FCHL revisited: faster and more accurate quantum machine learning
• Neural Network Based in Silico Simulation of Combustion Reactions
• OrbNet: Deep Learning for Quantum Chemistry Using Symmetry-Adapted Atomic-Orbital Features
• A deep potential model with long-range electrostatic interactions
• Scalable Parallel Algorithm for Graph Neural Network Interatomic Potentials in Molecular Dynamics Simulations
• Combining SchNet and SHARC: The SchNarc machine learning approach for excited-state dynamics
• Deep Potentials for Materials Science
• Inverse design of 3d molecular structures with conditional generative neural networks
• Evaluation of the MACE Force Field Architecture: from Medicinal Chemistry to Materials Science
• Committee neural network potentials control generalization errors and enable active learning
• Quantum Machine Learning for Chemistry and Physics
• Representations of molecules and materials for interpolation of quantum-mechanical simulations via machine learning
• Automated Fitting of Neural Network Potentials at Coupled Cluster Accuracy: Protonated Water Clusters as Testing Ground
• Neural Network Potentials for Chemistry: Concepts, Applications and Prospects
• General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian
• Machine Learning Coarse-Grained Potentials of Protein Thermodynamics
• Predicting molecular dipole moments by combining atomic partial charges and atomic dipoles
• Toward Explainable AI for Regression Models
• TeaNet: universal neural network interatomic potential inspired by iterative electronic relaxations
• Gaussian Moments as Physically Inspired Molecular Descriptors for Accurate and Scalable Machine Learning Potentials
• Advances of Machine Learning in Materials Science: Ideas and Techniques
• OrbNet Denali: A machine learning potential for biological and organic chemistry with semi-empirical cost and DFT accuracy
• End-to-End Differentiable Molecular Mechanics Force Field Construction
• Informing Geometric Deep Learning with Electronic Interactions to Accelerate Quantum Chemistry
• DeePMD-kit v3: A Multiple-Backend Framework for Machine Learning Potentials
• Automated discovery of a robust interatomic potential for aluminum
• Physically Motivated Recursively Embedded Atom Neural Networks: Incorporating Local Completeness and Nonlocality
• Strategies for the Construction of Machine-Learning Potentials for Accurate and Efficient Atomic-Scale Simulations
• Single-model uncertainty quantification in neural network potentials does not consistently outperform model ensembles
• High-Dimensional Neural Network Potentials for Magnetic Systems Using Spin-Dependent Atom-Centered Symmetry Functions
• SchNetPack 2.0: A neural network toolbox for atomistic machine learning
• Efficient and Accurate Simulations of Vibrational and Electronic Spectra with Symmetry-Preserving Neural Network Models for Tensorial Properties
• 3D Infomax improves GNNs for Molecular Property Prediction
• Transferable E(3) equivariant parameterization for Hamiltonian of molecules and solids
• Graph Neural Networks for Temperature-Dependent Activity Coefficient Prediction of Solutes in Ionic Liquids
• Graph Neural Networks for Molecules
• Tutorial: How to Train a Neural Network Potential
• PiNN: A Python Library for Building Atomic Neural Networks of Molecules and Materials
• Recursive evaluation and iterative contraction of $N$-body equivariant features
• BIGDML: Towards Exact Machine Learning Force Fields for Materials
• Fast Uncertainty Estimates in Deep Learning Interatomic Potentials
• AI-driven inverse design of materials: Past, present and future
• Universal Machine Learning for the Response of Atomistic Systems to External Fields
• MLatom 3: Platform for machine learning-enhanced computational chemistry simulations and workflows
• Flow-matching -- efficient coarse-graining of molecular dynamics without forces
• Reactive Dynamics and Spectroscopy of Hydrogen Transfer from Neural Network-Based Reactive Potential Energy Surfaces
• The Evolution of Machine Learning Potentials for Molecules, Reactions and Materials
• Machine learning and excited-state molecular dynamics
• Graph Neural Networks Accelerated Molecular Dynamics
• Equivariant representations for molecular Hamiltonians and N-center atomic-scale properties
• PANNA: Properties from Artificial Neural Network Architectures
• Physical Pooling Functions in Graph Neural Networks for Molecular Property Prediction
• TorchMD-Net 2.0: Fast Neural Network Potentials for Molecular Simulations
• A Perspective on Deep Learning for Molecular Modeling and Simulations
• High-Dimensional Potential Energy Surfaces for Molecular Simulations
• Permutationally Invariant, Reproducing Kernel-Based Potential Energy Surfaces for Polyatomic Molecules: From Formaldehyde to Acetone
• Insights into lithium manganese oxide-water interfaces using machine learning potentials
• Molecular Force Fields with Gradient-Domain Machine Learning (GDML): Comparison and Synergies with Classical Force Fields
• Improved accuracy and transferability of molecular-orbital-based machine learning: Organics, transition-metal complexes, non-covalent interactions, and transition states
• Atomic Permutationally Invariant Polynomials for Fitting Molecular Force Fields
• Permutationally invariant polynomial regression for energies and gradients, using reverse differentiation, achieves orders of magnitude speed-up with high precision compared to other machine learning methods
• Molecular Mechanics-Driven Graph Neural Network with Multiplex Graph for Molecular Structures
• Rotation Invariant Graph Neural Networks using Spin Convolutions
• Accurate machine learning force fields via experimental and simulation data fusion
• Geometric Deep Learning for Molecular Crystal Structure Prediction
• Force-Field-Enhanced Neural Network Interactions: from Local Equivariant Embedding to Atom-in-Molecule properties and long-range effects
• Lifelong Machine Learning Potentials
• Molecular dynamics simulation with finite electric fields using Perturbed Neural Network Potentials
• Predicting Properties of Periodic Systems from Cluster Data: A Case Study of Liquid Water
• Infrared spectra at coupled cluster accuracy from neural network representations
• Fast and Sample-Efficient Interatomic Neural Network Potentials for Molecules and Materials Based on Gaussian Moments
• Machine Learned Hückel Theory: Interfacing Physics and Deep Neural Networks
• ML Models of Vibrating H${}_2$CO: Comparing Reproducing Kernels, FCHL and PhysNet
• Reactive Atomistic Simulations of Diels-Alder Reactions: the Importance of Molecular Rotations
• DenseGNN: universal and scalable deeper graph neural networks for high-performance property prediction in crystals and molecules
• Graph Machine Learning for Design of High-Octane Fuels
• The MD17 Datasets from the Perspective of Datasets for Gas-Phase "Small" Molecule Potentials
• E(n)-Equivariant Cartesian Tensor Passing Potential
• ArcaNN: automated enhanced sampling generation of training sets for chemically reactive machine learning interatomic potentials
• On the design space between molecular mechanics and machine learning force fields
• Perspective: Non-conventional Force Fields for Applications in Spectroscopy and Chemical Reaction Dynamics
• Spherical Message Passing for 3D Graph Networks
• MLatom software ecosystem for surface hopping dynamics in Python with quantum mechanical and machine learning methods
• Transferability of machine learning potentials: Protonated water neural network potential applied to the protonated water hexamer
• PhysNet Meets CHARMM: A Framework for Routine Machine Learning / Molecular Mechanics Simulations
• Thermally Averaged Magnetic Anisotropy Tensors via Machine Learning Based on Gaussian Moments
• Isomerization and Decomposition Reactions of Acetaldehyde Relevant to Atmospheric Processes from Dynamics Simulations on Neural Network-Based Potential Energy Surfaces
• Routine Molecular Dynamics Simulations Including Nuclear Quantum Effects: from Force Fields to Machine Learning Potentials
• Neural network with optimal neuron activation functions based on additive Gaussian process regression
• Towards fully ab initio simulation of atmospheric aerosol nucleation
• Fewest-Switches Surface Hopping with Long Short-Term Memory Networks
• Atoms in molecules from alchemical perturbation density functional theory
• Machine Learning and Data-Driven Methods in Computational Surface and Interface Science
• Modeling electronic response properties with an explicit-electron machine learning potential
• Lightweight and Effective Tensor Sensitivity for Atomistic Neural Networks
• Heat flux for semi-local machine-learning potentials
• Denoise Pretraining on Nonequilibrium Molecules for Accurate and Transferable Neural Potentials
• Double Proton Transfer in Hydrated Formic Acid Dimer: Interplay of Spatial Symmetry and Solvent-Generated Force on Reactivity
• SE(3)-equivariant prediction of molecular wavefunctions and electronic densities
• A Universal Framework for Accurate and Efficient Geometric Deep Learning of Molecular Systems
• Impact of the characteristics of quantum chemical databases on machine learning predictions of tautomerization energies
• High-dimensional neural network potentials for accurate vibrational frequencies: The formic acid dimer benchmark
• $\Delta$-Machine Learning to Elevate DFT-based Potentials and a Force Field to the CCSD(T) Level Illustrated for Ethanol
• Innate Dynamics and Identity Crisis of a Metal Surface Unveiled by Machine Learning of Atomic Environments
• Transfer-Learned Potential Energy Surfaces: Towards Microsecond-Scale Molecular Dynamics Simulations in the Gas Phase at CCSD(T) Quality
• Reinforcement learning for traversing chemical structure space: Optimizing transition states and minimum energy paths of molecules
• Fast and flexible long-range models for atomistic machine learning
• Neural Network Potential with Multi-Resolution Approach Enables Accurate Prediction of Reaction Free Energies in Solution
• 3DReact: Geometric deep learning for chemical reactions
• Accelerating fourth-generation machine learning potentials by quasi-linear scaling particle mesh charge equilibration
• Machine-Learning Interatomic Potentials for Long-Range Systems
• Dynamics on Multiple Potential Energy Surfaces: Quantitative Studies of Elementary Processes Relevant to Hypersonics
• chemtrain: Learning Deep Potential Models via Automatic Differentiation and Statistical Physics
• Ab Initio Generalized Langevin Equation
• Degeneration of kernel regression with Matern kernels into low-order polynomial regression in high dimension
• MGNN: Moment Graph Neural Network for Universal Molecular Potentials
• Accurate quantum Monte Carlo forces for machine-learned force fields: Ethanol as a benchmark
• Quantum chemical roots of machine-learning molecular similarity descriptors
• Stress and heat flux via automatic differentiation
• Deep Learning in Protein Structural Modeling and Design
• Optimized Multifidelity Machine Learning for Quantum Chemistry
• A machine learning potential for simulating infrared spectra of nanosilicate clusters
• Formation and Stabilization of Ground and Excited State Singlet O${}_2$ upon Recombination of ${}^3$P Oxygen on Amorphous Solid Water
• Graph Nets for Partial Charge Prediction
• FeNNol: an Efficient and Flexible Library for Building Force-field-enhanced Neural Network Potentials
• Artificial Intelligence for Science in Quantum, Atomistic, and Continuum Systems
• Predicting the Electronic Density Response of Condensed-Phase Systems to Electric Field Perturbations
• Algorithmic Differentiation for Automated Modeling of Machine Learned Force Fields
• Decomposing Chemical Space: Applications to the Machine Learning of Atomic Energies
• Analytical Gradients for Molecular-Orbital-Based Machine Learning
• Thermal Activation of Methane by MgO${}^{+}$: Temperature Dependent Kinetics, Reactive Molecular Dynamics Simulations and Statistical Modeling
• Heterogeneous Molecular Graph Neural Networks for Predicting Molecule Properties
• Augmentation of Universal Potentials for Broad Applications
• Mean-Field Density Matrix Decompositions
• No Headache for PIPs: A PIP Potential for Aspirin Outperforms Other Machine-Learned Potentials
• Similarity Equivariant Graph Neural Networks for Homogenization of Metamaterials
• A chemical bond-based machine learning model for dipole moment: Application to dielectric properties of liquid methanol and ethanol
• Quantum Gaussian process model of potential energy surface for a polyatomic molecule
• Molecular-orbital-based Machine Learning for Open-shell and Multi-reference Systems with Kernel Addition Gaussian Process Regression
• Predicting Rigid Body Dynamics using Dual Quaternion Recurrent Neural Networks with Quaternion Attention
• Conformational and state-specific effects in reactions of 2,3-dibromobutadiene with Coulomb-crystallized calcium ions
• Low-data deep quantum chemical learning for accurate MP2 and coupled-cluster correlations
• A Hessian-Based Assessment of Atomic Forces for Training Machine Learning Interatomic Potentials
• Prediction rigidities for data-driven chemistry
• Machine learning potentials for complex aqueous systems made simple
• Towards Symbolic XAI -- Explanation Through Human Understandable Logical Relationships Between Features
• Efficient force field and energy emulation through partition of permutationally equivalent atoms
• Metatensor and metatomic: foundational libraries for interoperable atomistic machine learning
• Gradient domain machine learning with composite kernels: improving the accuracy of PES and force fields for large molecules
• End-to-end AI framework for interpretable prediction of molecular and crystal properties
• Discovery of structure-property relations for molecules via hypothesis-driven active learning over the chemical space
• The Wetting of H${}_2$O by CO${}_2$
• Extending the atomic decomposition and many-body representation, a chemistry-motivated monomer-centered approach for machine learning potentials
• Iterative charge equilibration for fourth-generation high-dimensional neural network potentials
• Machine learning at the atomic-scale
• $\Delta$-model correction of Foundation Model based on the models own understanding
• SAIBench: Benchmarking AI for Science
• Reliable computational prediction of supramolecular ordering of complex molecules under electrochemical conditions
• Graph-neural-network predictions of solid-state NMR parameters from spherical tensor decomposition
• Neural Network Approach for Predicting Infrared Spectra from 3D Molecular Structure
• Molecular Dipole Moment Learning via Rotationally Equivariant Gaussian Process Regression with Derivatives in Molecular-orbital-based Machine Learning
• Decomposed Mean-Field Simulations of Local Properties in Condensed Phases
• Efficient Generation of Stable Linear Machine-Learning Force Fields with Uncertainty-Aware Active Learning
• Equivariant Machine Learning Interatomic Potentials with Global Charge Redistribution
• Efficient ensemble uncertainty estimation in Gaussian Processes Regression
• Molecular relaxation by reverse diffusion with time step prediction
• EOSnet: Embedded Overlap Structures for Graph Neural Networks in Predicting Material Properties
• Scaling Graph Neural Networks to Large Proteins
• Molecule3D: A Benchmark for Predicting 3D Geometries from Molecular Graphs
• Learning intermolecular forces at liquid-vapor interfaces
• Reconstructions and Dynamics of $\beta$-Lithium Thiophosphate Surfaces
• Apax: A Flexible and Performant Framework For The Development of Machine-Learned Interatomic Potentials
• Atomistic Simulations of Oxide-Water Interfaces using Machine Learning Potentials
• A simple approach to rotationally invariant machine learning of avector quantity
• Accelerated lignocellulosic molecule adsorption structure determination
• Machine learning-guided construction of an analytic kinetic energy functional for orbital free density functional theory
• Transformative Applications of Machine Learning for Chemical Reactions
• Machine learning interatomic potential for the low-modulus Ti-Nb-Zr alloys in the vicinity of dynamical instability
• Accurate neural-network-based fitting of full-dimensional two-body potential energy surfaces
• Free energy profiles for chemical reactions in solution from high-dimensional neural network potentials: The case of the Strecker synthesis
• Ab initio machine learning in chemical compound space
• Molecular Dynamics and Machine Learning Unlock Possibilities in Beauty Design -- A Perspective
• Symmetry- and Gradient-enhanced Gaussian Process Regression for the Active Learning of Potential Energy Surfaces in Porous Materials
• Atomistic Graph Neural Networks for metals: Application to bcc iron
• Hybrid localized graph kernel for machine learning energy-related properties of molecules and solids
• Transfer Learned Potential Energy Surfaces: Accurate Anharmonic Vibrational Dynamics and Dissociation Energies for the Formic Acid Monomer and Dimer
• Reactive atomistic simulations of Diels-Alder-type reactions: Conformational and dynamic effects in the polar cycloaddition of 2,3-dibromobutadiene radical ions with maleic anhydride
• Transferability of the chemical bond-based machine learning model for dipole moment: the GHz to THz dielectric properties of liquid propylene glycol and polypropylene glycol
• Machine Learning Potential Energy Surfaces
• Flexible dual-branched message passing neural network for quantum mechanical property prediction with molecular conformation
• FastCHGNet: Training one Universal Interatomic Potential to 1.5 Hours with 32 GPUs
• GeoT: A Geometry-aware Transformer for Reliable Molecular Property Prediction and Chemically Interpretable Representation Learning
• Reconstructing Kernel-based Machine Learning Force Fields with Super-linear Convergence
• Orbital Mixer: Using Atomic Orbital Features for Basis Dependent Prediction of Molecular Wavefunctions
• Machine-Learned Potentials for Solvation Modeling
• Unified theory of atom-centered representations and message-passing machine-learning schemes
• General Learning of the Electric Response of Inorganic Materials
• Managing Expectations and Imbalanced Training Data in Reactive Force Field Development: an Application to Water Adsorption on Alumina
• Incorporating electronic information into Machine Learning potential energy surfaces via approaching the ground-state electronic energy as a function of atom-based electronic populations
• Representing spherical tensors with scalar-based machine-learning models
• Interpretation of Crystal Energy Landscapes with Kolmogorov-Arnold Networks
• Molecular electrostatic potentials from machine learning models for dipole and quadrupole predictions
• Gradient-based grand canonical optimization enabled by graph neural networks with fractional atomic existence
• Aitomia: Your Intelligent Assistant for AI-Driven Atomistic and Quantum Chemical Simulations
• Quantitative Molecular Simulations
• REANN: A PyTorch-based End-to-End Multi-functional Deep Neural Network Package for Molecular, Reactive and Periodic Systems
• Detect the Interactions that Matter in Matter: Geometric Attention for Many-Body Systems
• Learning Atomic Multipoles: Prediction of the Electrostatic Potential with Equivariant Graph Neural Networks
• Thermal and Vibrationally Activated Decomposition of the syn-CH${}_3$CHOO Criegee Intermediate
• Active learning of potential-energy surfaces of weakly-bound complexes with regression-tree ensembles
• Active Learning of a Neural Network Potential for Large-Scale Atomistic Simulations of Polymer Electrolyte Membranes
• Graph Convolutional Neural Networks for (QM)ML/MM Molecular Dynamics Simulations
• Geometric Algebra Attention Networks for Small Point Clouds
• Macro-Dipole-Constrainted Learning of Atomic Charges for Accurate Electrostatic Potentials at Electrochemical Interfaces
• Knowledge Distillation of a Protein Language Model Yields a Foundational Implicit Solvent Model