Towards Linearly Scaling and Chemically Accurate Global Machine Learning Force Fields for Large Molecules

Abstract

Machine learning force fields (MLFFs) are gradually evolving towards enabling molecular dynamics simulations of molecules and materials with ab initio accuracy but at a small fraction of the computational cost. However, several challenges remain to be addressed to enable predictive MLFF simulations of realistic molecules, including: (1) developing efficient descriptors for non-local interatomic interactions, which are essential to capture long-range molecular fluctuations, and (2) reducing the dimensionality of the descriptors in kernel methods (or a number of parameters in neural networks) to enhance the applicability and interpretability of MLFFs. Here we propose an automatized approach to substantially reduce the number of interatomic descriptor features while preserving the accuracy and increasing the efficiency of MLFFs. To simultaneously address the two stated challenges, we illustrate our approach on the example of the global GDML MLFF; however, our methodology can be equally applied to other models. We found that non-local features (atoms separated by as far as 15$\mathring{A}$ in studied systems) are crucial to retain the overall accuracy of the MLFF for peptides, DNA base pairs, fatty acids, and supramolecular complexes. Interestingly, the number of required non-local features in the reduced descriptors becomes comparable to the number of local interatomic features (those below 5$\mathring{A}$). These results pave the way to constructing global molecular MLFFs whose cost increases linearly, instead of quadratically, with system size.

References in corpus (19)

• E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials
• Machine Learning Force Fields
• Big Data of Materials Science - Critical Role of the Descriptor
• Combining Machine Learning and Computational Chemistry for Predictive Insights Into Chemical Systems
• A Fourth-Generation High-Dimensional Neural Network Potential with Accurate Electrostatics Including Non-local Charge Transfer
• Physics-inspired structural representations for molecules and materials
• Machine learning for electronically excited states of molecules
• SpookyNet: Learning Force Fields with Electronic Degrees of Freedom and Nonlocal Effects
• Resolving transition metal chemical space: feature selection for machine learning and structure-property relationships
• A deep potential model with long-range electrostatic interactions
• Raman Spectrum and Polarizability of Liquid Water from Deep Neural Networks
• Self-Consistent Determination of Long-Range Electrostatics in Neural Network Potentials
• Recursive evaluation and iterative contraction of $N$-body equivariant features
• BIGDML: Towards Exact Machine Learning Force Fields for Materials
• Challenges for Machine Learning Force Fields in Reproducing Potential Energy Surfaces of Flexible Molecules
• Efficient implementation of atom-density representations
• Towards Linearly Scaling and Chemically Accurate Global Machine Learning Force Fields for Large Molecules
• A Hessian-Based Assessment of Atomic Forces for Training Machine Learning Interatomic Potentials
• Learning intermolecular forces at liquid-vapor interfaces

Cited by in corpus (6)

• Evaluation of the MACE Force Field Architecture: from Medicinal Chemistry to Materials Science
• Accurate machine learning force fields via experimental and simulation data fusion
• Towards Linearly Scaling and Chemically Accurate Global Machine Learning Force Fields for Large Molecules
• Adaptive energy reference for machine-learning models of the electronic density of states
• Interpolating many-body wave functions for accelerated molecular dynamics on the near-exact electronic surface
• Photothermal Expansion of Nanostructures in Photo-induced Force Microscopy