期刊
MATERIALS TODAY
卷 51, 期 -, 页码 126-135出版社
ELSEVIER SCI LTD
DOI: 10.1016/j.mattod.2021.08.012
关键词
Materials discovery; Bayesian optimization; Graph neural network; Deep learning
资金
- National Science Foundation Materials Research Science and Engineering Center program through the UC Irvine Center for Complex and Active Materials [DMR-2011967]
- U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, Materials Sciences and Engineering Division [DE-AC02-05-CH11231]
- Extreme Science and Engineering Discovery Environment (XSEDE) [ACI-1548562]
The utilization of Bayesian optimization with symmetry constraints and a graph deep learning energy model allows for DFT-free relaxations of crystal structures, significantly improving the accuracy of ML-predicted properties and leading to the discovery and synthesis of novel ultra-incompressible hard crystals with exceptional properties.
Machine learning (ML) models utilizing structure-based features provide an efficient means for accurate property predictions across diverse chemical spaces. However, obtaining equilibrium crystal structures typically requires expensive density functional theory (DFT) calculations, which limits ML-based exploration to either known crystals or a small number of hypothetical crystals. Here, we demonstrate that the application of Bayesian optimization with symmetry constraints using a graph deep learning energy model can be used to perform DFT-free relaxations of crystal structures. Using this approach to significantly improve the accuracy of ML-predicted formation energies and elastic moduli of hypothetical crystals, two novel ultra-incompressible hard MoWC2 (P6(3)=mmc) and ReWB (Pca2(1)) were identified and successfully synthesized via in situ reactive spark plasma sintering from screening 399,960 transition metal borides and carbides. This work addresses a critical bottleneck to accurate property predictions for hypothetical materials, paving the way to ML-accelerated discovery of new materials with exceptional properties.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据