Graph neural networks (GNN) have been shown to provide substantial performance improvements for atomistic material representation and modeling compared with descriptor-based machine learning models. While most existing GNN models for atomistic predictions are based on atomic distance information, they do not explicitly incorporate bond angles, which are critical for distinguishing many atomic structures. Furthermore, many material properties are known to be sensitive to slight changes in bond angles. We present an Atomistic Line Graph Neural Network (ALIGNN), a GNN architecture that performs message passing on both the interatomic bond graph and its line graph corresponding to bond angles. We demonstrate that angle information can be explicitly and efficiently included, leading to improved performance on multiple atomistic prediction tasks. We ALIGNN models for predicting 52 solid-state and molecular properties available in the JARVIS-DFT, Materials project, and QM9 databases. ALIGNN can outperform some previously reported GNN models on atomistic prediction tasks by up to 85% in accuracy with better or comparable model training speed.
Taming Multi-Domain, -Fidelity Data: Towards Foundation Models for Atomistic Scale Simulations
Graduate School of Engineering Science, The University of Osaka(大阪大学工学研究科);Center for Quantum Information and Quantum Biology, The University of Osaka(大阪大学量子信息与量子生物学中心)
Atomistic Machine Learning with Irreducible Cartesian Natural Tensors
Institute of Fundamental and Frontier Sciences, University of Electronic Science and Technology of China(电子科技大学基础与前沿研究院);Department of Mechanical Engineering, The University of Hong Kong(香港大学机械工程系)
Inverse design of glass structure with deep graph neural networks
直接操纵原子结构以实现特定性质是材料领域的长期追求。然而,受到无序的、非原型的玻璃结构以及结构与性能之间复杂相互作用的阻碍,这种逆向设计对于玻璃来说是令人生畏的困难。在这里,结合两个前沿技术,图神经网络和交换蒙特卡罗,我们开发了一个数据驱动的,面向属性的逆向设计路线,设法以可控的方式提高铜锆金属玻璃的塑性阻力。Swap Monte Carlo作为“采样器”有效地探索了玻璃景观,而图形神经网络在预测塑性阻力方面具有很高的回归精度,作为“决策器”来指导配置空间中的搜索。通过一种非常规的强化机制,一种几何上超稳定但能量上亚稳定的状态被解开,与通常认为的能量越高,塑性阻力越低相反。这表明了一个巨大的配置空间,可以很容易地被传统的原子模拟忽略。数据驱动技术、结构搜索方法和优化算法整合形成一个工具箱,为玻璃材料的设计铺平了新的道路。
DScribe: Library of Descriptors for Machine Learning in Materials Science
Aalto University(阿尔托大学);Nanolayers Research Computing Ltd.(纳米层研究计算有限公司);Norwegian University of Science and Technology(挪威科技大学);Technische Universität München(慕尼黑工业大学);Kanazawa University(金泽大学)
Mixture of Experts Framework in Machine Learning Interatomic Potentials for Atomistic Simulations
Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, MA, USA(材料科学与工程系,麻省理工学院,剑桥,马萨诸塞州,美国);John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, USA(约翰·A·波尔森工程与应用科学学院,哈佛大学,剑桥,马萨诸塞州,美国);Robert Bosch LLC Research and Technology Center, Watertown, MA, USA(罗伯特·博世 LLC 研究与技术中心,沃特敦,马萨诸塞州,美国)
Fast and Accurate Equivariant Foundation Models for Atomistic Simulation
Yusuf Hamied Department of Chemistry, University of Cambridge(剑桥大学);John A. Paulson School of Engineering and Applied Sciences, Harvard University(哈佛大学);Department of Materials Science and Engineering, Massachusetts Institute of Technology(麻省理工学院);Department of Molecular Science and Software Engineering, University of California, Berkeley(加州大学伯克利分校);Sandia National Laboratories(桑迪亚国家实验室)
optimade-maker: Automated generation of interoperable materials APIs from static datasets
PSI Center for Scientific Computing, Theory and Data(瑞士保罗谢尔研究所科学计算、理论与数据中心);National Centre for Computational Design and Discovery of Novel Materials (MARVEL)(新材料计算设计与发现国家中心(MARVEL));UCLouvain, Institute of Condensed Matter and Nanosciences(乌克斯洛万大学凝聚态与纳米科学研究所);Matgenix SRL(Matgenix有限责任公司);datalab industries ltd.(datalab工业有限公司);Theory and Simulation of Materials (THEOS), École Polytechnique Fédérale de Lausanne(洛桑联邦理工学院材料理论与模拟(THEOS))