Machine Learning Force Fields (2020-10-14T00:00:00.000000Z)

TL;DR

An overview of applications of ML-FFs and the chemical insights that can be obtained from them is given, and a step-by-step guide for constructing and testing them from scratch is given.

Abstract

In recent years, the use of machine learning (ML) in computational chemistry has enabled numerous advances previously out of reach due to the computational complexity of traditional electronic-structure methods. One of the most promising applications is the construction of ML-based force fields (FFs), with the aim to narrow the gap between the accuracy of ab initio methods and the efficiency of classical FFs. The key idea is to learn the statistical relation between chemical structure and potential energy without relying on a preconceived notion of fixed chemical bonds or knowledge about the relevant interactions. Such universal ML approximations are in principle only limited by the quality and quantity of the reference data used to train them. This review gives an overview of applications of ML-FFs and the chemical insights that can be obtained from them. The core concepts underlying ML-FFs are described in detail, and a step-by-step guide for constructing and testing them from scratch is given. The text concludes with a discussion of the challenges that remain to be overcome by the next generation of ML-FFs.

Authors

A. Tkatchenko

11 papers

Kristof T. Schütt

6 papers

M. Gastegger

4 papers

TL;DR

Abstract

Authors

References329 items

Machine learning of solvent effects on molecular spectra and reactions

Machine Learning in QM/MM Molecular Dynamics Simulations of Condensed-Phase Systems.

Quantum chemical accuracy from density functional approximations via machine learning

Neural Network Potential Energy Surfaces for Small Molecules and Reactions.

A fourth-generation high-dimensional neural network potential with accurate electrostatics including non-local charge transfer

Quantum machine learning using atom-in-molecule-based fragments selected on the fly

Machine Learning Approaches toward Orbital-free Density Functional Theory: Simultaneous Training on the Kinetic Energy Density Functional and Its Functional Derivative

Molecular force fields with gradient-domain machine learning (GDML): Comparison and synergies with classical force fields.

On the role of gradients for machine learning of molecular energies and forces

Machine learning the ropes: principles, applications and directions in synthetic chemistry.

Machine Learning for Electronically Excited States of Molecules

Dynamical strengthening of covalent and non-covalent molecular interactions by nuclear quantum effects at finite temperature

Comprehensive Investigations of the Cl + CH 3 OH → HCl + CH 3 O/CH 2 OH Reaction: Validation of Experiment and Dynamic Insights

Accurate Many-Body Repulsive Potentials for Density-Functional Tight Binding from Deep Tensor Neural Networks.

Equivariant Flows: exact likelihood generative learning for symmetric densities

A deep neural network for molecular wave functions in quasi-atomic minimal basis representation.

Ensemble learning of coarse-grained molecular dynamics force fields with a kernel approach.

Pushing the Limit of Molecular Dynamics with Ab Initio Accuracy to 100 Million Atoms with Machine Learning

Raman spectrum and polarizability of liquid water from deep neural networks.

Isomerization and decomposition reactions of acetaldehyde relevant to atmospheric processes from dynamics simulations on neural network-based potential energy surfaces.

Directional Message Passing for Molecular Graphs

Reinforcement Learning for Molecular Design Guided by Quantum Mechanics

Combining SchNet and SHARC: The SchNarc Machine Learning Approach for Excited-State Dynamics

Thermal activation of methane by MgO+: temperature dependent kinetics, reactive molecular dynamics simulations and statistical modeling.

Free energy of proton transfer at the water–TiO2 interface from ab initio deep potential molecular dynamics†

Improved protein structure prediction using potentials from deep learning

PyTorch: An Imperative Style, High-Performance Deep Learning Library

Simulating protein–ligand binding with neural network potentials

Exploring chemical compound space with quantum-based machine learning

Machine learning for protein folding and dynamics.

Reactive dynamics and spectroscopy of hydrogen transfer from neural network-based reactive potential energy surfaces

Machine learning for molecular simulation

High-dimensional potential energy surfaces for molecular simulations: from empiricism to machine learning

Generating valid Euclidean distance matrices

Deep-neural-network solution of the electronic Schrödinger equation

Boltzmann generators: Sampling equilibrium states of many-body systems with deep learning

Machine Learning Interatomic Potentials as Emerging Tools for Materials Science

FCHL revisited: Faster and more accurate quantum machine learning.

Embedded Atom Neural Network Potentials: Efficient and Accurate Machine Learning with a Physically Inspired Representation.

Unifying machine learning and quantum chemistry with a deep neural network for molecular wavefunctions

Reactive atomistic simulations of Diels-Alder reactions: The importance of molecular rotations.

High-Resolution Markov State Models for the Dynamics of Trp-Cage Miniprotein Constructed over Slow Folding Modes Identified by State-Free Reversible VAMPnets.

Cormorant: Covariant Molecular Neural Networks

Exhaustive state-to-state cross sections for reactive molecular collisions from importance sampling simulation and a neural network representation.

Symmetry-adapted generation of 3d point sets for the targeted discovery of molecules

Reactive molecular dynamics for the [Cl–CH3–Br]− reaction in the gas phase and in solution: a comparative study using empirical and neural network force fields

DScribe: Library of Descriptors for Machine Learning in Materials Science

Kernel methods for detecting coherent structures in dynamical data.

Coarse-graining molecular systems by spectral matching.

Efficient multi-objective molecular optimization in a continuous latent space

Targeted Adversarial Learning Optimized Sampling.

Machine learning meets quantum physics

Unmasking Clever Hans predictors and assessing what machines really learn

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

Sampling reactive regions in phase space by following the minimum dynamic path.

Nonlinear Discovery of Slow Molecular Modes using Hierarchical Dynamics Encoders

Molecular force fields with gradient-domain machine learning: Construction and application to dynamics of small molecules with coupled cluster forces.

Solvation Free Energy Calculations with Quantum Mechanics/Molecular Mechanics and Machine Learning Models.

sGDML: Constructing accurate and data efficient molecular force fields using machine learning

Machine Learning of Coarse-Grained Molecular Dynamics Force Fields

Machine learning enables long time scale molecular photodynamics simulations

Deep Learning for Nonadiabatic Excited-State Dynamics.

Generating equilibrium molecules with deep neural networks

Nuclear Quantum Effects in Sodium Hydroxide Solutions from Neural Network Molecular Dynamics Simulations.

Elucidating the Nuclear Quantum Dynamics of Intramolecular Double Hydrogen Transfer in Porphycene

Accurate molecular polarizabilities with coupled cluster theory and machine learning

Converged Colored Noise Path Integral Molecular Dynamics Study of the Zundel Cation Down to Ultralow Temperatures at Coupled Cluster Accuracy.

SchNetPack: A Deep Learning Toolbox For Atomistic Systems.

A Density Functional Tight Binding Layer for Deep Learning of Chemical Hamiltonians.

i-PI 2.0: A universal force engine for advanced molecular simulations

Operators in quantum machine learning: Response properties in chemical space.

3D Steerable CNNs: Learning Rotationally Equivariant Features in Volumetric Data

Quantum-chemical insights from interpretable atomistic neural networks

Predicting molecular properties with covariant compositional networks.

Learning Effective Molecular Models from Experimental Observables.

End-to-end Symmetry Preserving Inter-atomic Potential Energy Model for Finite and Extended Systems