Multiscale Modeling of Defect Phenomena in Platinum Using Machine Learning of Force Fields
- PDF / 1,989,345 Bytes
- 13 Pages / 593.972 x 792 pts Page_size
- 46 Downloads / 204 Views
https://doi.org/10.1007/s11837-020-04385-0 Ó 2020 The Minerals, Metals & Materials Society
AUGMENTING PHYSICS-BASED MODELS IN ICME WITH MACHINE LEARNING AND UNCERTAINTY QUANTIFICATION
Multiscale Modeling of Defect Phenomena in Platinum Using Machine Learning of Force Fields JAMES CHAPMAN1 and RAMPI RAMPRASAD1,2 1.—School of Materials Science and Engineering, Georgia Institute of Technology, North Ave NW, Atlanta, GA 30332, USA. 2.—e-mail: [email protected]
Computational methodologies have been critical to our understanding of defects at nanometer scales. These methodologies have been dominated by two classes: quantum mechanics (QM)-based methods and semiempirical/classical methods. The former, while accurate and versatile, are time consuming, while the latter are efficient but limited in versatility and transferability. Recently, machine learning (ML) methods have shown initial promise in bridging these two limitations due to their accuracy and flexibility. In this work, the true capability of ML methods is explored by simulating defects in platinum over several length/time scales. We compare our results with density functional theory (DFT) for atomic-level defect behavior and with experiments for nanolevel behavior. We also compare our predictions with several classical potentials. This work aims to showcase the length/time scales attainable using ML, as well as the complexity they are capable of capturing, demonstrating that these methodologies may be effectively used, in the future, to bridge experiments and QM methods.
INTRODUCTION Atomistic computational techniques have been instrumental in the exploration of material defects at both atomic and nanometer length scales.1–7 These methods have historically fallen into two broad categories: QM-based methods, e.g., density functional theory (DFT),8,9 and semiempirical methods, e.g., the embedded atom method.10–16 While both groups have been used to study a multitude of materials phenomena,17–19 they both suffer from serious shortcomings. QM methods, while able to capture properties at a high level of fidelity, are computationally expensive, which severely restricts both the time and length scales that can be reliably accessed through such simulations. Semiempirical/classical methods significantly reduce this cost burden and allow for the exploration of length and time scales that are not attainable with QM. However, as such methods are parameterized to tackle specific problems, they do not attain the same
(Received May 20, 2020; accepted September 14, 2020)
level of versatility as QM methods, showing a steep decline in accuracy away from their respective reference data. To this end, data-driven machine learning (ML) methods have shown promise as a reliable alternative, bridging the gap in cost, accuracy, and transferability.20–28 Unlike their counterparts, ML methods rely on functional forms, and parameterizations of these functional forms, that are derived from statistics, rather than physics. The accuracy of these models will still decline duri
Data Loading...
