SchNet – A deep learning architecture for molecules and materials

2017/12/31 by K. T. Schütt, Kristof T. Schütt, Huziel E. Sauceda +8 · 2,235 citations
Biochemistry, Genetics and Molecular Biology · Computer Science · Materials Science · Physics and Astronomy · #Ab initio #Architecture #Artificial intelligence #Bioinformatics #Biology #Chemical space #Computational Drug Discovery Methods #Computer science #Deep learning #Machine Learning in Materials Science #Molecular dynamics #Physics #Protein Structure and Dynamics #Quantum #Quantum mechanics #Space (punctuation) #cond-mat.mtrl-sci #physics.chem-ph

paper · pdf · doi:10.1063/1.5019779

published in The Journal of Chemical Physics 148(24), 241722 (American Institute of Physics)

arxiv created 2018/03/22 · openalex publication_date 2018/03/29 · arxiv updated 2018/04/18 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05

Abstract

Deep learning has led to a paradigm shift in artificial intelligence, including web, text, and image search, speech recognition, as well as bioinformatics, with growing impact in chemical physics. Machine learning, in general, and deep learning, in particular, are ideally suitable for representing quantum-mechanical interactions, enabling us to model nonlinear potential-energy surfaces or enhancing the exploration of chemical compound space. Here we present the deep learning architecture SchNet that is specifically designed to model atomistic systems by making use of continuous-filter convolutional layers. We demonstrate the capabilities of SchNet by accurately predicting a range of properties across chemical space for molecules and materials, where our model learns chemically plausible embeddings of atom types across the periodic table. Finally, we employ SchNet to predict potential-energy surfaces and energy-conserving force fields for molecular dynamics simulations of small molecules and perform an exemplary study on the quantum-mechanical properties of C20-fullerene that would have been infeasible with regular ab initio molecular dynamics.

Citations

• Quantum-chemical insights from deep tensor neural networks

Cited by

• Operators in quantum machine learning: Response properties in chemical space
• Predicting molecular dipole moments by combining atomic partial charges and atomic dipoles
• Coarse Graining Molecular Dynamics with Graph Neural Networks
• TorchANI 2.0: An Extensible, High-Performance Library for the Design, Training, and Use of NN-IPs
• Multi-body effects in a coarse-grained protein force field
• Machine learning for nonadiabatic molecular dynamics: best practices and recent progress
• Analyzing Atomic Interactions in Molecules as Learned by Neural Networks
• Alchemical Free-Energy Calculations at Quantum-Chemical Precision
• A Transdisciplinary Review of Deep Learning Research and Its Relevance for Water Resources Scientists
• Machine Learning for Molecular Simulation
• Learning on compressed molecular representations
• Machine Learning of Reactive Potentials
• Neural Network Potentials: A Concise Overview of Methods
• OrbitAll: A Unified Quantum Mechanical Representation Deep Learning Framework for All Molecular Systems
• GEqTrain: A Configuration-Driven Framework for Retargeting Equivariant Graph Neural Networks Across 3D Scientific Tasks
• Point Group Equivariant Graph Neural Networks for Materials
• ChemFusion: A Multimodal Cross-Attention Network for Reaction Yield Prediction
• Machine learning in NMR spectroscopy
• El Agente: An autonomous agent for quantum chemistry
• Extending the RANGE of Graph Neural Networks: Relaying Attention Nodes for Global Encoding
• Transformative Applications of Machine Learning for Chemical Reactions
• Generalizing Convolutional Neural Networks for Equivariance to Lie Groups on Arbitrary Continuous Data
• Energy-Guided Flow Matching Enables Few-Step Conformer Generation and Ground-State Identification
• Investigating 3D Atomic Environments for Enhanced QSAR
• Crash testing machine learning force fields for molecules, materials, and interfaces: model analysis in the TEA Challenge 2023
• Navigating protein landscapes with a machine-learned transferable coarse-grained model
• Neural Network Potential with Multi-Resolution Approach Enables Accurate Prediction of Reaction Free Energies in Solution
• Dual-Level Atomic and Coordination Geometry Learning for Crystal Property Prediction Using Graph Neural Networks
• mCGCNN: A Dual-Stream Crystal Graph Convolutional Neural Network for the Efficient Prediction of Magnetic Properties of Crystalline Materials
• Crystal Diffusion Variational Autoencoder for Periodic Material Generation
• PhononBench-MP40: a spectrum-resolved benchmark dataset for phonon stability
• qsGW quasiparticle and GW-BSE excitation energies of 133,885 molecules
• Infusing Experimental Reality into Complex Many-Body Hamiltonians: The Observable-Constrained Variational Framework (OCVF)
• Mapping Still Matters: Coarse-Graining with Machine Learning Potentials
• Equivariant Diffusion for Crystal Structure Prediction
• Knowing when to trust machine-learned interatomic potentials
• Predicting Kovats Retention Indices Using Graph Neural Networks
• QuantumCanvas: A Multimodal Benchmark for Visual Learning of Atomic Interactions
• Solar-GECO: Perovskite Solar Cell Property Prediction with Geometric-Aware Co-Attention
• Rotation Invariant Graph Neural Networks using Spin Convolutions
• Machine learning and the physical sciences
• Using pretrained graph neural networks with token mixers as geometric featurizers for conformational dynamics
• Incorporating electronic information into Machine Learning potential energy surfaces via approaching the ground-state electronic energy as a function of atom-based electronic populations
• Chemistry-Enhanced Diffusion-Based Framework for Small-to-Large Molecular Conformation Generation
• Toward bilipshiz geometric models
• Benchmarking GNNs for OOD Materials Property Prediction with Uncertainty Quantification
• Compact Artificial Neural Network Models for Predicting Protein Residue -- RNA Base Binding
• Quantum-centric machine learning for molecular dynamics
• Magnitude-Modulated Equivariant Adapter for Parameter-Efficient Fine-Tuning of Equivariant Graph Neural Networks
• Connecting the concepts of quantum state tomography and molecular representations for machine learning
• Stability and Dynamics of Sn-based Halide Perovskites: Insights from MACE-MP-0 and Molecular Dynamics Simulations
• Atomistic Simulations of Oxide-Water Interfaces using Machine Learning Potentials
• S pai NN: equivariant message passing for excited-state nonadiabatic molecular dynamics
• Uncertainty Quantification for In Silico Chemistry
• Ab initio machine learning in chemical compound space
• Learning to Make Chemical Predictions: the Interplay of Feature Representation, Data, and Machine Learning Algorithms
• Deep Learning in Protein Structural Modeling and Design
• MLatom Software Ecosystem for Surface Hopping Dynamics in Python with Quantum Mechanical and Machine Learning Methods
• Global geometry of chemical graph neural network representations in terms of chemical moieties
• A machine-learning framework for accelerating spin-lattice relaxation simulations
• Knowledge Distillation of a Protein Language Model Yields a Foundational Implicit Solvent Model
• Molecular Simulations with a Pretrained Neural Network and Universal Pairwise Force Fields
• Four Generations of High-Dimensional Neural Network Potentials
• Data-driven complete basis set limit estimates from a minimal auxiliary basis
• QSAR without borders
• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems
• GeoMol: Torsional Geometric Generation of Molecular 3D Conformer Ensembles
• Gaussian Process Regression for Materials and Molecules
• Machine learning interatomic potentials in biomolecular modeling: principles, architectures, and applications
• Learning 3D Anisotropic Noise Distributions Improves Molecular Force Field Modeling
• Structure-Aware Fusion with Progressive Injection for Multimodal Molecular Representation Learning
• Accelerated Simulations of Molecular Systems through Learning of their Effective Dynamics
• Ab-Initio Potential Energy Surfaces by Pairing GNNs with Neural Wave Functions
• Committors without Descriptors
• Transfer learning graph representations of molecules for pKa, 13 C-NMR, and solubility
• Deep Learning-Based Extraction of Promising Material Groups and Common Features from High-Dimensional Data: A Case of Optical Spectra of Inorganic Crystals
• ProtoMol: Enhancing Molecular Property Prediction via Prototype-Guided Multimodal Learning
• 3D-GSRD: 3D Molecular Graph Auto-Encoder with Selective Re-mask Decoding
• GO-Diff: Data-free and amortized global structure optimization
• Universally Invariant Learning in Equivariant GNNs
• Ab-Initio Solution of the Many-Electron Schrödinger Equation with Deep Neural Networks
• Reasoning-Enhanced Large Language Models for Molecular Property Prediction
• Incorporating Neural Networks into the AMOEBA Polarizable Force Field
• Reinforcement Learning for Traversing Chemical Structure Space: Optimizing Transition States and Minimum Energy Paths of Molecules
• Analysis of Atomistic Representations Using Weighted Skip-Connections
• Accelerated prediction of dielectric functions in solar cell materials with graph neural networks
• Deep potential molecular dynamics simulations of ion-enhanced etching of silicon by atomic chlorine
• When Does Global Attention Help? A Unified Empirical Study on Atomistic Graph Learning
• Fundamental Limits of Crystalline Equivariant Graph Neural Networks: A Circuit Complexity Perspective
• Enhancing NMR Shielding Predictions of Atoms-in-Molecules Machine Learning Models with Neighborhood-Informed Representations
• 3DMolNet: A Generative Network for Molecular Structures
• Flow-Matching Based Refiner for Molecular Conformer Generation
• Rotational equivariant graph neural networks via local eigenbasis transformations
• On the design space between molecular mechanics and machine learning force fields
• Structure and transport properties of LiTFSI-based deep eutectic electrolytes from machine-learned interatomic potential simulations
• Shoot from the HIP: Hessian Interatomic Potentials without derivatives
• Scalable Machine Learning Model for Energy Decomposition Analysis in Aqueous Systems
• Efficient Grand Canonical Global Optimization with On-the-fly-trained Machine-learning Interatomic Potentials
• Advancing Universal Deep Learning for Electronic-Structure Hamiltonian Prediction of Materials
• Automated Discovery of Algorithms for Molecular Electronic Structure Calculations Using Physics-Informed Program Synthesis
• Incorporating Scientific Knowledge into Neural Network Density Functionals
• Deep learning in-depth analysis of crystal graph convolutional neural networks: A new era in materials discovery and its applications
• Beyond Numerical Hessians: Higher-Order Derivatives for Machine Learning Interatomic Potentials via Automatic Differentiation
• Geometric deep learning for molecular property predictions with chemical accuracy across chemical space
• Deep potential molecular dynamics simulations of low-temperature plasma-surface interactions
• Machine learning coarse-grained potentials of protein thermodynamics
• A foundation model for atomistic materials chemistry
• Machine-Learning Potentials for Efficient Simulations of Anisotropic Colloids
• How are Scientific Concepts Birthed? Typing Rules of Concept Formation in Theoretical Physics Reasoning
• Generative Quasi-Continuum Modeling of Confined Fluids at the Nanoscale
• Topotein: Topological Deep Learning for Protein Representation Learning
• Global properties of the energy landscape: a testing and training arena for machine learned potentials
• Machine learning force fields and coarse-grained variables in molecular dynamics: application to materials and biological systems
• MACE4IRmol: An uncertainty-aware foundation model for molecular infrared spectroscopy
• LFaB: Low fidelity as Bias for Active Learning in the chemical configuration space
• Symmetry-adapted generation of 3d point sets for the targeted discovery of molecules
• Autonomous thermodynamically informed database generation for machine-learned interatomic potentials and application to magnesium
• Topological Feature Compression for Molecular Graph Neural Networks
• Benchmarking Pretrained Molecular Embedding Models For Molecular Representation Learning
• Efficient Prediction of SO(3)-Equivariant Hamiltonian Matrices via SO(2) Local Frames
• Deriving effective electrode-ion interactions from free-energy profiles at electrochemical interfaces
• Study of Deep Generative Models for Inorganic Chemical Compositions
• Atomistic Generative Diffusion for Materials Modeling
• Classical and reactive molecular dynamics: Principles and applications in combustion and energy systems
• Predicting Material Properties Using a 3D Graph Neural Network with Invariant Local Descriptors
• On-the-Fly Fine-Tuning of Foundational Neural Network Potentials: A Bayesian Neural Network Approach
• Data-Driven Approach to Encoding and Decoding 3-D Crystal Structures
• Generating valid Euclidean distance matrices
• MP-ALOE: An r2SCAN dataset for universal machine learning interatomic potentials
• Implicit Machine Learning Force Fields Accelerate Molecular Dynamics Simulations
• Zero-Shot Cyclic Peptide Design via Composable Geometric Constraints
• Flexible dual-branched message passing neural network for quantum mechanical property prediction with molecular conformation
• Tensor Decomposition Networks for Fast Machine Learning Interatomic Potential Computations
• Augmenting Molecular Graphs with Geometries via Machine Learning Interatomic Potentials
• Rotational Sampling: A Plug-and-Play Encoder for Rotation-Invariant 3D Molecular GNNs
• A Benchmark for Quantum Chemistry Relaxations via Machine Learning Interatomic Potentials
• Fine-Tuning Universal Machine-Learned Interatomic Potentials: A Tutorial on Methods and Applications
• A Survey of AI for Materials Science: Foundation Models, LLM Agents, Datasets, and Tools
• Machine Learning Accelerates Raman Computations from Molecular Dynamics for Materials Science
• Fast and Distributed Equivariant Graph Neural Networks by Virtual Node Learning
• Operator Forces For Coarse-Grained Molecular Dynamics
• Leveraging Influence Functions for Resampling Data in Physics-Informed Neural Networks
• CLOUD: A Scalable and Physics-Informed Foundation Model for Crystal Representation Learning
• Atomistic Graph Neural Networks for metals: Application to bcc iron
• Understanding multi-fidelity training of machine-learned force-fields
• GeoRecon: Graph-Level Representation Learning for 3D Molecules via Reconstruction-Based Pretraining
• Graph Neural Networks in Modern AI-aided Drug Discovery
• Leveraging active learning-enhanced machine-learned interatomic potential for efficient infrared spectra prediction
• Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges
• Efficient dataset generation for machine learning perovskite alloys
• A neural network-based four-body potential energy surface for parahydrogen
• chemtrain-deploy: A parallel and scalable framework for machine learning potentials in million-atom MD simulations
• Molecular modeling with machine-learned universal potential functions
• Neural Message Passing with Edge Updates for Predicting Properties of Molecules and Materials
• GraphITE: Estimating Individual Effects of Graph-structured Treatments
• On Universality Classes of Equivariant Networks
• Self-Refining Training for Amortized Density Functional Theory
• HMAE: Self-Supervised Few-Shot Learning for Quantum Spin Systems
• From Invariant Representations to Invariant Data: Provable Robustness to Spurious Correlations via Noisy Counterfactual Matching
• Representing local protein environments with machine learning force fields
• Equivariant Spherical Transformer for Efficient Molecular Modeling
• Machine-Learned Potentials for Solvation Modeling
• Apax: A Flexible and Performant Framework For The Development of Machine-Learned Interatomic Potentials
• DistMLIP: A Distributed Inference Platform for Machine Learning Interatomic Potentials
• Machine Learning Interatomic Potentials: library for efficient training, model development and simulation of molecular systems
• Learning simple heuristic rules for classifying materials based on chemical composition
• Scalable Generation of Spatial Transcriptomics from Histology Images via Whole-Slide Flow Matching
• High-order Equivariant Flow Matching for Density Functional Theory Hamiltonian Prediction
• Deep Neural Network Computes Electron Densities and Energies of a Large Set of Organic Molecules Faster than Density Functional Theory (DFT)
• Local-Global Associative Frames for Symmetry-Preserving Crystal Structure Modeling
• Loss-Guided Auxiliary Agents for Overcoming Mode Collapse in GFlowNets
• Properties of α-Brass Nanoparticles II: Structure and Composition
• EDBench: Large-Scale Electron Density Data for Molecular Modeling
• Stable crystal structure prediction using machine learning-based formation energy and empirical potential function
• Efficient Parallelization of Message Passing Neural Network Potentials for Large-scale Molecular Dynamics
• BOOM: Benchmarking Out-Of-distribution Molecular Property Predictions of Machine Learning Models
• Artificial Intelligence based Autonomous Molecular Design for Medical Therapeutic: A Perspective
• The evolution of machine learning potentials for molecules, reactions and materials
• ChargeFlow: Flow-Matching Refinement of Charge-Conditioned Electron Densities
• Adapting Evidential Neural Networks to Test-Time Neighbor Fusion Improves Molecular Property Prediction
• Learning Hamiltonian Flow Maps: Mean Flow Consistency for Large-Timestep Molecular Dynamics
• Heterogeneous Molecular Graph Neural Networks for Predicting Molecule Properties
• Spatial statistics for screening molecular structures
• MLIPilot: LLM-Driven Auto-Research for Machine-Learned Interatomic Potentials
• MatMMFuse: Multi-Modal Fusion model for Material Property Prediction
• Geometric learning of the conformational dynamics of molecules using dynamic graph neural networks
• Utilising Graph Machine Learning within Drug Discovery and Development
• AI-driven discovery of new materials: breaking data bottlenecks and transforming research paradigms
• PCEvo: Path-Consistent Molecular Representation via Virtual Evolutionary
• Molecular force fields with gradient-domain machine learning (GDML): Comparison and synergies with classical force fields
• Self-Supervised Graph Neural Networks for Accurate Prediction of Néel Temperature
• Efficient Global Structure Optimization with a Machine-Learned Surrogate Model
• Finding symmetry breaking order parameters with Euclidean neural networks
• An accurate and transferable machine learning potential for carbon
• Accurate and generalizable protein-ligand binding affinity prediction with geometric deep learning
• Construction of Machine Learned Force Fields with Quantum Chemical Accuracy: Applications and Chemical Insights
• Predicting Methane Adsorption in Metal-Substituted MOFs: A Comparative Study between Density Functional Theory and Machine Learning
• Inverse Design of a Graphene-Based Quantum Transducer via Neuroevolution
• Symmetrical Graph Neural Network for Quantum Chemistry with Dual Real and Momenta Space
• Bayesian force fields from active learning for simulation of inter-dimensional transformation of stanene
• Learning properties of ordered and disordered materials from multi-fidelity data
• Complex reaction processes in combustion unraveled by neural network-based molecular dynamics simulation
• Deep learning for UV absorption spectra with SchNarc: First steps toward transferability in chemical compound space
• Supervised learning through physical changes in a mechanical system
• Machine learning on neutron and x-ray scattering and spectroscopies
• Machine learning enabled discovery of application dependent design principles for two-dimensional materials
• Physically informed artificial neural networks for atomistic modeling of materials
• Structural Phase Transitions in SrTiO3 from Deep Potential Molecular Dynamics
• ElemNet: Deep Learning the Chemistry of Materials From Only Elemental Composition. [europepmc]
• Machine Learning of Coarse-Grained Molecular Dynamics Force Fields. [europepmc]
• Accurate and transferable multitask prediction of chemical properties with an atoms-in-molecules neural network. [europepmc]
• Data-Driven Materials Science: Status, Challenges, and Perspectives. [europepmc]
• Unifying machine learning and quantum chemistry with a deep neural network for molecular wavefunctions. [europepmc]
• Combining SchNet and SHARC: The SchNarc Machine Learning Approach for Excited-State Dynamics. [europepmc]
• QSAR without borders. [europepmc]
• The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for molecules. [europepmc]
• Big-Data Science in Porous Materials: Materials Genomics and Machine Learning. [europepmc]
• Machine Learning Force Fields and Coarse-Grained Variables in Molecular Dynamics: Application to Materials and Biological Systems. [europepmc]
• Quantum chemical accuracy from density functional approximations via machine learning. [europepmc]
• Machine Learning for Electronically Excited States of Molecules. [europepmc]
• Coarse graining molecular dynamics with graph neural networks. [europepmc]
• Deep Learning in Protein Structural Modeling and Design. [europepmc]
• A fourth-generation high-dimensional neural network potential with accurate electrostatics including non-local charge transfer. [europepmc]
• Machine Learning Force Fields. [europepmc]
• TorchMD: A Deep Learning Framework for Molecular Simulations. [europepmc]
• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems. [europepmc]
• Gaussian Process Regression for Materials and Molecules. [europepmc]
• DNA-Encoded Chemical Libraries: A Comprehensive Review with Succesful Stories and Future Challenges. [europepmc]
• GEOM, energy-annotated molecular conformations for property prediction and molecular generation. [europepmc]
• Bottom-up Coarse-Graining: Principles and Perspectives. [europepmc]
• Graph neural networks for materials science and chemistry. [europepmc]
• DeePMD-kit v2: A software package for deep potential models. [europepmc]
• Chemprop: A Machine Learning Package for Chemical Property Prediction. [europepmc]

Related

• SchNet: A continuous-filter convolutional neural network for modeling quantum interactions
• Learning representations of molecules and materials with atomistic\n neural networks
• Quantum-chemical insights from interpretable atomistic neural networks
• Analyzing Atomic Interactions in Molecules as Learned by Neural Networks
• Analyzing Atomic Interactions in Molecules as Learned by Neural Networks