Abstract:Machine learning potentials (MLPs) offer the potential to accurately model the energy and free energy landscapes of molecules with the precision of quantum mechanics and an efficiency similar to classical simulations. This research focuses on using equivariant graph neural networks MLPs due to their proven effectiveness in modeling equilibrium molecular trajectories. A key issue addressed is the capability of MLPs to accurately predict free energies and transition states by considering both the energy and the diversity of molecular configurations. We examined how the distribution of collective variables (CVs) in the training data affects MLP accuracy in determining the free energy surface (FES) of systems, using Metadynamics simulations for butane and alanine dipeptide (ADP). The study involved training forty-three MLPs, half based on classical molecular dynamics data and the rest on ab initio computed energies. The MLPs were trained using different distributions that aim to replicate hypothetical scenarios of sampled CVs obtained if the underlying FES of the system was unknown. Findings for butane revealed that training data coverage of key FES regions ensures model accuracy regardless of CV distribution. However, missing significant FES regions led to correct potential energy predictions but failed free energy reconstruction. For ADP, models trained on classical dynamics data were notably less accurate, while ab initio-based MLPs predicted potential energy well but faltered on free energy predictions. These results emphasize the challenge of assembling an all-encompassing training set for accurate FES prediction and highlight the importance of understanding the FES in preparing training data. The study points out the limitations of MLPs in free energy calculations, stressing the need for comprehensive data that encompasses the system's full FES for effective model training.

Machine Learning Potentials with the Iterative Boltzmann Inversion: Training to Experiment

Machine learning potentials with Iterative Boltzmann Inversion: training to experiment

Learning together: Towards foundation models for machine learning interatomic potentials with meta-learning

Lifelong Machine Learning Potentials

Learning Together: Towards foundational models for machine learning interatomic potentials with meta-learning

Tutorial: How to Train a Neural Network Potential

Learning Interatomic Potentials at Multiple Scales

Considerations in the use of ML interaction potentials for free energy calculations

Simple and efficient algorithms for training machine learning potentials to force data

Training Quantum Boltzmann Machines with the $β$-Variational Quantum Eigensolver

The Potential of Neural Network Potentials

Training Multilayer Perceptrons by Sampling with Quantum Annealers

Synthetic pre-training for neural-network interatomic potentials

Inverting the Kohn-Sham equations with physics-informed machine learning

Generative and discriminative training of Boltzmann machine through Quantum annealing

Learning QM/MM Potential using Equivariant Multiscale Model

Active learning of Boltzmann samplers and potential energies with quantum mechanical accuracy

A dual-cutoff machine-learned potential for condensed organic systems obtained via uncertainty-guided active learning

Transferability of datasets between Machine-Learning Interaction Potentials

The Energy-saving Anaerobic Baffled Reactor-Membrane Bioreactor ( EABR-MBR ) System for High-rise Building Wastewater Recycling

Stable and Accurate Atomistic Simulations of Flexible Molecules using Conformationally Generalisable Machine Learned Potentials