An official website of the United States government
Abstract Machine learning interatomic potential (MLIP) has been widely adopted for atomistic simulations. While errors and discrepancies for MLIPs have been reported, a comprehensive examination of the MLIPs’ performance over a broad spectrum of material properties has been lacking. This study introduces an analysis process comprising model sampling, benchmarking, error evaluations, and multi-dimensional statistical analyses on an ensemble of MLIPs for prediction errors over a diverse range of properties. By carrying out this analysis on 2300 MLIP models based on six different MLIP types, several properties that pose challenges for the MLIPs to achieve small errors are identified. The Pareto front analyses on two or more properties reveal the trade-offs in different properties of MLIPs, underscoring the difficulties of achieving low errors for a large number of properties simultaneously. Furthermore, we propose correlation graph analyses to characterize the error performances of MLIPs and to select the representative properties for predicting other property errors. This analysis process on a large dataset of MLIP models sheds light on the underlying complexities of MLIP performance, offering crucial guidance for the future development of MLIPs with improved predictive accuracy across an array of material properties.
more »
« less
Discrepancies and error evaluation metrics for machine learning interatomic potentials
Abstract Machine learning interatomic potentials (MLIPs) are a promising technique for atomic modeling. While small errors are widely reported for MLIPs, an open concern is whether MLIPs can accurately reproduce atomistic dynamics and related physical properties in molecular dynamics (MD) simulations. In this study, we examine the state-of-the-art MLIPs and uncover several discrepancies related to atom dynamics, defects, and rare events (REs), compared to ab initio methods. We find that low averaged errors by current MLIP testing are insufficient, and develop quantitative metrics that better indicate the accurate prediction of atomic dynamics by MLIPs. The MLIPs optimized by the RE-based evaluation metrics are demonstrated to have improved prediction in multiple properties. The identified errors, the evaluation metrics, and the proposed process of developing such metrics are general to MLIPs, thus providing valuable guidance for future testing and improvements of accurate and reliable MLIPs for atomistic modeling.
more »
« less
- Award ID(s):
- 2004837
- PAR ID:
- 10465282
- Publisher / Repository:
- Nature Publishing Group
- Date Published:
- Journal Name:
- npj Computational Materials
- Volume:
- 9
- Issue:
- 1
- ISSN:
- 2057-3960
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
High-entropy alloys (HEAs) have attracted considerable attention due to their exceptional properties and outstanding performance across various applications. However, the vast compositional space and complex high-dimensional atomic interactions pose significant challenges in uncovering fundamental physical principles and effectively guiding alloy design. Traditional experimental approaches, often reliant on trial-and-error methods, are time-consuming, cost-prohibitive, and inefficient. To accelerate progress in this field, advanced simulation techniques and data-driven methodologies, particularly machine learning (ML) with a particular interest in nanoscale phenomena, have emerged as transformative tools for composition design, property prediction, and performance optimization. By leveraging extensive materials databases and sophisticated learning algorithms, ML facilitates the discovery of intricate patterns that conventional methods may overlook, and enables the design of HEAs with targeted properties. This review paper provides a comprehensive overview of recent advancements in ML applications for HEAs. It begins with a brief introduction of the fundamental principles of HEAs and ML methodologies, including key algorithms, databases, and evaluation metrics. The critical role of materials representation and feature engineering in ML-driven HEA design is thoroughly discussed. Furthermore, state-of-the-art developments in the integration of ML with HEA research, particularly in composition optimization, property prediction, and phase identification, are systematically reviewed. Special emphasis is placed on cutting-edge deep learning techniques, such as generative models and computer vision, which are revolutionizing the f ield. This study explores the application of machine learning (ML) in developing highly accurate ML interatomic potentials (MLIPs) for molecular dynamics (MD) simulations. These MLIPs have the potential to enhance the accuracy and efficiency of simulations, enabling a more precise representation of the fundamental physics governing high-entropy alloys (HEAs) at the atomic level. A critical discussion is provided, addressing both the potential advantages and the inherent limitations of this approach. This review aims to provide insights into the future directions of ML-driven HEA research, offering a roadmap for advancing material design through data-driven innovation.more » « less
-
Abstract We present machine‐learning interatomic potentials (MLIPs) for simulations of Si–C–O–H compounds. The MLIPs are constructed from moment tensor potentials (MTPs) and were trained to a library of configurations that included polysiloxane structures, hypothetical crystalline and amorphous SiCOH structures, and trajectories of Si–C–O–H systems obtained via ab initio molecular dynamic (aiMD) simulations at elevated temperatures. Passive, active, and hybrid learning strategies were implemented to develop the MLIPs. The MLIPs reproduce vibrational properties of polymers and SiCOH structures obtained from aiMD simulations, thus providing a tool to identify chemical units and distinct structural characteristics through their vibrational properties. Simulations of the polymer‐to‐ceramic transformation show the development of mixed tetrahedra in SiCO ceramics and align with experimental observations. Million‐atom simulations for several nanoseconds highlight the precipitation of graphitic nanosheets from a carbon‐rich SiCO precursor. Atomistic simulations with the MLIPs deliver details of chemical reaction mechanisms during the pyrolysis of polysiloxanes, including methane abstraction and Kumada‐like rearrangements that transform the siloxane backbone. While the MLIPs still leave room for systematic improvement, they deliver simulations with “density functional theory (DFT)‐like” quality at low and high temperatures.more » « less
-
Calculations with high accuracy for atomic and inter-atomic properties, such as nuclear magnetic resonance (NMR) spectroscopy and bond dissociation energies (BDEs) are valuable for pharmaceutical molecule structural analysis, drug exploration, and screening. It is important that these calculations should include relativistic effects, which are computationally expensive to treat. Non-relativistic calculations are less expensive but their results are less accurate. In this study, we present a computational framework for predicting atomic and inter-atomic properties by using machine-learning in a non-relativistic but accurate and computationally inexpensive framework. The accurate atomic and inter-atomic properties are obtained with a low dimensional deep neural network (DNN) embedded in a fragment-based graph convolutional neural network (F-GCN). The F-GCN acts as an atomic fingerprint generator that converts the atomistic local environments into data for the DNN, which improves the learning ability, resulting in accurate results as compared to experiments. Using this framework, the 13C/1H NMR chemical shifts of Nevirapine and phenol O–H BDEs are predicted to be in good agreement with experimental measurement.more » « less
-
null (Ed.)The application of machine learning models and algorithms towards describing atomic interactions has been a major area of interest in materials simulations in recent years, as machine learning interatomic potentials (MLIPs) are seen as being more flexible and accurate than their classical potential counterparts. This increase in accuracy of MLIPs over classical potentials has come at the cost of significantly increased complexity, leading to higher computational costs and lower physical interpretability and spurring research into improving the speeds and interpretability of MLIPs. As an alternative, in this work we leverage “machine learning” fitting databases and advanced optimization algorithms to fit a class of spline-based classical potentials, showing that they can be systematically improved in order to achieve accuracies comparable to those of low-complexity MLIPs. These results demonstrate that high model complexities may not be strictly necessary in order to achieve near-DFT accuracy in interatomic potentials and suggest an alternative route towards sampling the high accuracy, low complexity region of model space by starting with forms that promote simpler and more interpretable inter- atomic potentials.more » « less
