%0 Journal Article %1 Gubaev2023 %A Gubaev, Konstantin %A Zaverkin, Viktor %A Srinivasan, Prashanth %A Duff, Andrew Ian %A Kästner, Johannes %A Grabowski, Blazej %D 2023 %J npj Computational Materials %K EXC2075 PN3 PN3A-4 selected %N 1 %P 129 %R 10.1038/s41524-023-01073-w %T Performance of two complementary machine-learned potentials in modelling chemically complex systems %U https://doi.org/10.1038/s41524-023-01073-w %V 9 %X Chemically complex multicomponent alloys possess exceptional properties derived from an inexhaustible compositional space. The complexity however makes interatomic potential development challenging. We explore two complementary machine-learned potentials---the moment tensor potential (MTP) and the Gaussian moment neural network (GM-NN)---in simultaneously describing configurational and vibrational degrees of freedom in the Ta-V-Cr-W alloy family. Both models are equally accurate with excellent performance evaluated against density-functional-theory. They achieve root-mean-square-errors (RMSEs) in energies of less than a few meV/atom across 0 K ordered and high-temperature disordered configurations included in the training. Even for compositions not in training, relative energy RMSEs at high temperatures are within a few meV/atom. High-temperature molecular dynamics forces have similarly small RMSEs of about 0.15 eV/\AA for the disordered quaternary included in, and ternaries not part of training. MTPs achieve faster convergence with training size; GM-NNs are faster in execution. Active learning is partially beneficial and should be complemented with conventional human-based training set generation.

@article{Gubaev2023, abstract = {Chemically complex multicomponent alloys possess exceptional properties derived from an inexhaustible compositional space. The complexity however makes interatomic potential development challenging. We explore two complementary machine-learned potentials---the moment tensor potential (MTP) and the Gaussian moment neural network (GM-NN)---in simultaneously describing configurational and vibrational degrees of freedom in the Ta-V-Cr-W alloy family. Both models are equally accurate with excellent performance evaluated against density-functional-theory. They achieve root-mean-square-errors (RMSEs) in energies of less than a few meV/atom across 0 K ordered and high-temperature disordered configurations included in the training. Even for compositions not in training, relative energy RMSEs at high temperatures are within a few meV/atom. High-temperature molecular dynamics forces have similarly small RMSEs of about 0.15 eV/{\AA} for the disordered quaternary included in, and ternaries not part of training. MTPs achieve faster convergence with training size; GM-NNs are faster in execution. Active learning is partially beneficial and should be complemented with conventional human-based training set generation.}, added-at = {2025-02-14T11:17:09.000+0100}, author = {Gubaev, Konstantin and Zaverkin, Viktor and Srinivasan, Prashanth and Duff, Andrew Ian and K{\"a}stner, Johannes and Grabowski, Blazej}, biburl = {https://puma.ub.uni-stuttgart.de/bibtex/2012abfee1a3b3d63508bb45b51639816/simtechpuma}, day = 25, doi = {10.1038/s41524-023-01073-w}, interhash = {77490c81feaefdacd3f9a22e021dcbd6}, intrahash = {012abfee1a3b3d63508bb45b51639816}, issn = {2057-3960}, journal = {npj Computational Materials}, keywords = {EXC2075 PN3 PN3A-4 selected}, month = {07}, number = 1, pages = 129, timestamp = {2025-02-14T11:17:09.000+0100}, title = {Performance of two complementary machine-learned potentials in modelling chemically complex systems}, url = {https://doi.org/10.1038/s41524-023-01073-w}, volume = 9, year = 2023 }