Machine Learning and First-Principles Discovery of Ternary Superhard Materials
Lists
Only you see this
Abstract
Superhard compounds defined as those with a Vickers hardness H V > 40 GPa have a wide range of industrial and extreme-environment applications. However, unlike other mechanical properties such as bulk and shear moduli, the hardness of a material is less well-documented, which poses a challenge for data-driven machine learning approaches. Here, we discuss using machine learning and first-principles simulations to discover new ternary superhard compounds. We first introduce various existing superhard systems, and then discuss evolutionary crystal structure searches and machine learning algorithms. Finally, we review using the above computational tools to study new superhard B-C-N systems, where machine learning is performed by correlating the target hardness with the more widely available bulk and shear moduli. We also discuss the possibility of using an iterative process of machine learning and evolutionary search to create training data for superhard B-N-O systems. The applications of machine learning and first-principles calculations to discover superhard carbon allotropes and model load-dependent hardness measurements are also briefly reviewed.
BibTeX
@inproceedings{chen2022machine,
author = {Wei-Chih Chen and Da Yan and Cheng-Chien Chen},
title = {{Machine Learning and First-Principles Discovery of Ternary Superhard Materials}},
booktitle = {ACS},
year = {2022},
doi = {10.1021/bk-2022-1416.ch009},
}Loading related content...