Finding alloys with specific design properties is challenging due to the large number of possible compositions and the complex interactions between elements. This study introduces a multi-objective Bayesian optimization approach guiding molecular dynamics simulations for discovering high-performance refractory alloys with both targeted intrinsic static thermomechanical properties and also deformation mechanisms occurring during dynamic loading. The objective functions are aiming for excellent thermomechanical stability via a high bulk modulus, a low thermal expansion, a high heat capacity, and for a resilient deformation mechanism maximizing the retention of the BCC phase after shock loading. Contrasting two optimization procedures, we show that the Pareto-optimal solutions are confined to a small performance space when the property objectives display a cooperative relationship. Conversely, the Pareto front is much broader in the performance space when these properties have antagonistic relationships. Density functional theory simulations validate these findings and unveil underlying atomic-bond changes driving property improvements.
@article{osti_2426032,
author = {Startt, Jacob Kyle and McCarthy, Megan Jeanne and Wood, Mitchell Anthony and Donegan, Sean and Dingreville, Rémi Philippe Michel},
title = {Bayesian blacksmithing: discovering thermomechanical properties and deformation mechanisms in high-entropy refractory alloys},
annote = {Finding alloys with specific design properties is challenging due to the large number of possible compositions and the complex interactions between elements. This study introduces a multi-objective Bayesian optimization approach guiding molecular dynamics simulations for discovering high-performance refractory alloys with both targeted intrinsic static thermomechanical properties and also deformation mechanisms occurring during dynamic loading. The objective functions are aiming for excellent thermomechanical stability via a high bulk modulus, a low thermal expansion, a high heat capacity, and for a resilient deformation mechanism maximizing the retention of the BCC phase after shock loading. Contrasting two optimization procedures, we show that the Pareto-optimal solutions are confined to a small performance space when the property objectives display a cooperative relationship. Conversely, the Pareto front is much broader in the performance space when these properties have antagonistic relationships. Density functional theory simulations validate these findings and unveil underlying atomic-bond changes driving property improvements.},
doi = {10.1038/s41524-024-01353-z},
url = {https://www.osti.gov/biblio/2426032},
journal = {npj Computational Materials},
issn = {ISSN 2057-3960},
number = {1},
volume = {10},
place = {United States},
publisher = {Nature Publishing Group},
year = {2024},
month = {07}}
