Iterative training set refinement enables reactive molecular dynamics via machine learned forces

Atomistic Simulation of HF Etching Process of Amorphous Si3N4 Using Machine Learning Potential

An atomistic understanding of dry-etching processes with reactive molecules is crucial for achieving geometric integrity in highly scaled semiconductor devices. Molecular dynamics (MD) simulations are instrumental, but the lack of reliable force fields hinders the widespread use of MD in etching simulations. In this work, we develop an accurate neural network potential (NNP) for simulating the etching process of amorphous Si3N4 with HF molecules. The surface reactions in diverse local environments are considered by incorporating several types of training sets: baseline structures, reaction-specific data, and general-purpose training sets. Furthermore, the NNP is refined through iterative comparisons with the density functional theory results. Using the trained NNP, we carry out etching simulations, which allow for detailed observation and analysis of key processes such as preferential sputtering, surface modification, etching yield, threshold energy, and the distribution of etching products. Additionally, we develop a simple continuum model, built from the MD simulation results, which effectively reproduces the surface composition obtained with MD simulations. By establishing a computational framework for atomistic etching simulation and scale bridging, this work will pave the way for more accurate and efficient design of etching processes in the semiconductor industry, enhancing device performance and manufacturing precision.

Reducing the Cost of Neural Network Potential Generation for Reactive Molecular Systems

Although machine learning potentials have recently had a substantial impact on molecular simulations, the construction of a robust training set can still become a limiting factor, especially due to the requirement of a reference ab initio simulation that covers all the relevant geometries of the system. Recognizing that this can be prohibitive for certain systems, we develop the method of transition tube sampling that mitigates the computational cost of training set and model generation. In this approach, we generate classical or quantum thermal geometries around a transition path describing a conformational change or a chemical reaction using only a sparse set of local normal mode expansions along this path and select from these geometries by an active learning protocol. This yields a training set with geometries that characterize the whole transition without the need for a costly reference trajectory. The performance of the method is evaluated on different molecular systems with the complexity of the potential energy landscape increasing from a single minimum to a double proton-transfer reaction with high barriers. Our results show that the method leads to training sets that give rise to models applicable in classical and path integral simulations alike that are on par with those based directly on ab initio calculations while providing the computational speedup we have come to expect from machine learning potentials.

Quality of uncertainty estimates from neural network potential ensembles

Neural network potentials (NNPs) combine the computational efficiency of classical interatomic potentials with the high accuracy and flexibility of the ab initio methods used to create the training set, but can also result in unphysical predictions when employed outside their training set distribution. Estimating the epistemic uncertainty of a NNP is required in active learning or on-the-fly generation of potentials. Inspired from their use in other machine-learning applications, NNP ensembles have been used for uncertainty prediction in several studies, with the caveat that ensembles do not provide a rigorous Bayesian estimate of the uncertainty. To test whether NNP ensembles provide accurate uncertainty estimates, we train such ensembles in four different case studies and compare the predicted uncertainty with the errors on out-of-distribution validation sets. Our results indicate that NNP ensembles are often overconfident, underestimating the uncertainty of the model, and require to be calibrated for each system and architecture. We also provide evidence that Bayesian NNPs, obtained by sampling the posterior distribution of the model parameters using Monte Carlo techniques, can provide better uncertainty estimates.

Molecular Dynamics and Machine Learning in Catalysts

Given the importance of catalysts in the chemical industry, they have been extensively investigated by experimental and numerical methods. With the development of computational algorithms and computer hardware, large-scale simulations have enabled influential studies with more atomic details reflecting microscopic mechanisms. This review provides a comprehensive summary of recent developments in molecular dynamics, including ab initio molecular dynamics and reaction force-field molecular dynamics. Recent research on both approaches to catalyst calculations is reviewed, including growth, dehydrogenation, hydrogenation, oxidation reactions, bias, and recombination of carbon materials that can guide catalyst calculations. Machine learning has attracted increasing interest in recent years, and its combination with the field of catalysts has inspired promising development approaches. Its applications in machine learning potential, catalyst design, performance prediction, structure optimization, and classification have been summarized in detail. This review hopes to shed light and perspective on ML approaches in catalysts.

Differentiable sampling of molecular geometries with uncertainty-based adversarial attacks

Neural network (NN) interatomic potentials provide fast prediction of potential energy surfaces, closely matching the accuracy of the electronic structure methods used to produce the training data. However, NN predictions are only reliable within well-learned training domains, and show volatile behavior when extrapolating. Uncertainty quantification methods can flag atomic configurations for which prediction confidence is low, but arriving at such uncertain regions requires expensive sampling of the NN phase space, often using atomistic simulations. Here, we exploit automatic differentiation to drive atomistic systems towards high-likelihood, high-uncertainty configurations without the need for molecular dynamics simulations. By performing adversarial attacks on an uncertainty metric, informative geometries that expand the training domain of NNs are sampled. When combined with an active learning loop, this approach bootstraps and improves NN potentials while decreasing the number of calls to the ground truth method. This efficiency is demonstrated on sampling of kinetic barriers, collective variables in molecules, and supramolecular chemistry in zeolite-molecule interactions, and can be extended to any NN potential architecture and materials system.