Including physics-informed atomization constraints in neural networks for reactive chemistry

Shuhao Zhang,Michael Chigaev,Olexandr Isayev,Richard Messerly,Nicholas Lubbers

DOI: https://doi.org/10.26434/chemrxiv-2024-lsp3k

2024-12-10

Abstract:Machine learning interatomic potentials (MLIPs) have emerged as powerful tools for investigating atomistic systems with high accuracy and relatively low computational cost. However, a common challenge with current MLIPs is their limited ability to accurately predict the relative energies of systems containing isolated or nearly isolated atoms, which often appear in various chemical processes. To address this limitation, we present a technique for modifying existing neural network architectures in a minimal way which accounts for the energies of isolated atoms and the atomization energy (AE) of a system. Using this technique, we build a model architecture we call HIP-NN-AE, an enhanced version of the Hierarchically Interacting Particle Neural Network (HIP-NN). Our results demonstrate that the HIP-NN-AE model significantly outperforms the previous HIP-NN model in multiple scenarios involving isolated atoms, such as complete atomization and bond dissociation processes. We also show that this innovation does not compromise the model performance on other tasks, including barrier height and conformational energy variation. The HIP-NN-AE model thus offers a robust solution to the challenges posed by isolated atoms in energy prediction tasks.

Chemistry

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that when using machine - learning interatomic potentials (MLIPs) for chemical reaction research, the existing models have limited ability in predicting the relative energies of systems containing isolated or nearly isolated atoms. Specifically, the paper points out that although MLIPs perform excellently in studying atomic systems with high precision and relatively low computational cost, they have significant flaws when dealing with the energy prediction of isolated atoms or nearly isolated atoms. This is a common problem in various chemical processes, because these processes often involve the states of isolated atoms or nearly isolated atoms. To solve this problem, the author proposes a technique. By minimally modifying the existing neural network architecture, considerations for the isolated - atom energy and the system atomization energy (AE) are introduced. This method constructs a new model architecture called HIP - NN - AE, which is an enhanced version of the Hierarchically Interacting Particle Neural Network (HIP - NN). The experimental results show that the HIP - NN - AE model significantly outperforms the previous HIP - NN model in multiple scenarios involving isolated atoms, such as complete atomization and bond dissociation processes. Moreover, this innovation does not affect the performance of the model on other tasks, including the prediction of potential barrier heights and conformational energy changes. Therefore, the HIP - NN - AE model provides a robust solution to the challenges of isolated atoms in energy prediction tasks.

Including physics-informed atomization constraints in neural networks for reactive chemistry

Machine Learning Interatomic Potentials and Long-Range Physics

AIMNet2: A Neural Network Potential to Meet your Neutral, Charged, Organic, and Elemental-Organic Needs

Lightweight and effective tensor sensitivity for atomistic neural networks

Exploring the frontiers of condensed-phase chemistry with a general reactive machine learning potential

Ab initio Accuracy Neural Network Potential for Drug-like Molecules

A Physics-Aware Neural Network for Protein-Ligand Interactions with Quantum Chemical Accuracy

Ab Initio Molecular Dynamics Simulations of Atmospheric Molecular Clusters Boosted by Neural Networks

Benchmarking of machine learning interatomic potentials for reactive hydrogen dynamics at metal surfaces

Accurate modeling of the potential energy surface of atmospheric molecular clusters boosted by neural networks

Extending the atomic decomposition and many-body representation, a chemistry-motivated monomer-centered approach for machine learning potentials

Neural Network Potential with Multi-Resolution Approach Enables Accurate Prediction of Reaction Free Energies in Solution

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

Machine learning interatomic potentials for reactive hydrogen dynamics at metal surfaces based on iterative refinement of reaction probabilities

Analyzing Atomic Interactions in Molecules as Learned by Neural Networks

A physics-aware neural network for protein–ligand interactions with quantum chemical accuracy

Modeling of Metal Nanoparticles: Development of Neural-Network Interatomic Potential Inspired by Features of the Modified Embedded-Atom Method

Enhancing Molecular Energy Predictions with Physically Constrained Modifications to the Neural Network Potential

Constructing accurate and efficient general-purpose atomistic machine learning model with transferable accuracy for quantum chemistry

Incorporating Neural Networks into the AMOEBA Polarizable Force Field

Emergence of Accurate Atomic Energies from Machine Learned Noble Gas Potentials