Improving Reliability of Machine Learned Interatomic Potentials With Physics-Informed Pretraining
College of Computing, Georgia Institute of Technology(佐治亚理工学院计算学院);School of Computational Science and Engineering, Georgia Institute of Technology(佐治亚理工学院计算科学与工程学院)
Machine learned interatomic potentials (MLIPs) have emerged as powerful tools for molecular dynamics (MD) simulations with their competitive accuracy and computational efficiency. However, MLIPs are often observed to exhibit un-physical behavior when encountering configurations which deviate significantly from their training data distribution, leading to simulation instabilities and unreliable dynamics, thus limiting the reliability of MLIPs for materials simulations. We present a physics-informed pretraining strategy that leverages simple physical potentials which can improve the robustness and stability of graph-based MLIPs for MD simulations. We demonstrate this approach by deploying a pretraining-finetuning pipeline where MLIPs are initially pretrained on data labelled with embedded atom model potentials and subsequently finetuned on the quantum mechanical ground truth data. By evaluating across three diverse material systems (phosphorus, silica, and a subset of Materials Project) and three representative MLIP architectures (CGCNN, M3GNet, and TorchMD-NET), we find that this physics-informed pretraining consistently improves both prediction accuracy as well as stability in MD compared to the baselines.
Machine-learning octet $AB$-type binary compounds across chemical space with domain knowledge of the interatomic bond
Ruhr-Universität Bochum(波鸿鲁尔大学);Federal Institute for Materials Research and Testing(联邦材料研究与测试研究所);Friedrich-Schiller-Universität Jena(耶拿弗里德里希·席勒大学);Karlsruhe Institute of Technology(卡尔斯鲁厄理工学院)
Global Universal Scaling and Ultra-Small Parameterization in Machine Learning Interatomic Potentials with Super-Linearity
Southern University of Science and Technology(南方科技大学);Shanghai University(上海大学);Materials Genome Institute(材料基因组研究院);Institute of Innovative Materials(创新材料研究院)
Chemo-mechanical coupling stabilizes mixed $\mathrm{Ag}_{x}\mathrm{Cu}_{1-x}\mathrm{GaSe}_{2}$ solar-cell absorbers: Insights from Monte-Carlo simulations assisted by ab initio informed machine-learning potentials
Institute of Materials Science, Technische Universitat Darmstadt(达姆施塔特工业大学材料科学研究所)
DPA4: Pushing the Accuracy-Cost Frontier of Interatomic Potentials with EMFA SO(2) Convolution
Peking University(北京大学);AI for Science Institute(人工智能科学研究院);Tsinghua University(清华大学);HEDPS, CAPT, College of Engineering, Peking University(北京大学工学院理论力学与计算力学系、高超声速空气动力学国家重点实验室);DP Technology(深势科技);Academy for Advanced Interdisciplinary Studies, Peking University(北京大学前沿交叉学科研究院);National Key Laboratory of Computational Physics, Institute of Applied Physics and Computational Mathematics(中国科学院应用物理与计算数学研究所)
FIRE-GNN: Force-informed, Relaxed Equivariance Graph Neural Network for Rapid and Accurate Prediction of Surface Properties
Northeastern University(东北大学);Department of Mathematics(数学系);Khoury College of Computer Sciences(计算机科学学院);Department of Mechanical and Industrial Engineering(机械与工业工程系)
CHGNet: Pretrained universal neural network potential for charge-informed atomistic modeling
University of California, Berkeley(加州大学伯克利分校);Lawrence Berkeley National Laboratory(劳伦斯伯克利国家实验室);University of Cambridge(剑桥大学);University of Minnesota(明尼苏达大学)
Physics-Informed Tree Search for High-Dimensional Computational Design
Department of Mechanical and Industrial Engineering, University of Illinois, Chicago(机械与工业工程系,伊利诺伊大学芝加哥分校);Center for Nanoscale Materials, Argonne National Laboratory(纳米材料中心,阿贡国家实验室);Mathematics and Computer Science Division, Argonne National Laboratory(数学与计算机科学部,阿贡国家实验室)
Boston College(波士顿学院);NIST Center for Neutron Research, National Institute of Standards and Technology(美国国家标准与技术研究院中子研究中心);Carnegie Mellon University(卡内基梅隆大学);University of California, Berkeley(加州大学伯克利分校);University of Maryland, College Park(马里兰大学帕克分校)
PSI Center for Scientific Computing, Theory and Data(瑞士保罗谢勒研究所科学计算、理论与数据中心);National Centre for Computational Design and Discovery of Novel Materials (MARVEL)(国家新材料计算设计与发现中心(MARVEL));Universität Bremen(不来梅大学);École Polytechnique Fédérale de Lausanne(洛桑联邦理工学院);University of Cambridge(剑桥大学)
Fine-tuning MLIP foundation models: strategies for accuracy and transferability
University of Cambridge(剑桥大学);Science and Technology Facilities Council–Daresbury Laboratory(科学和技术设施委员会-达尔斯伯里实验室);U.S. Naval Research Laboratory(美国海军研究实验室)
Breaking Bottlenecks in Solid Electrolyte Discovery with Large Artificial Intelligence Models
Tohoku University(东北大学);National University of Singapore(新加坡国立大学);University at Buffalo(布法罗纽约州立大学);London School of Economics(伦敦政治经济学院);Karlsruhe Institute of Technology(卡尔斯鲁厄理工学院);University of Minho(米尼奥大学);UPV/EHU(巴斯克大学);Suzhou MatSource Technology Co., Ltd.(苏州迈特源科技有限公司);Gusu Laboratory of Materials(姑苏材料实验室);Indian Institute of Technology Hyderabad(印度海得拉巴理工学院);Yonsei University(延世大学)
Exploring Charge Density Waves in two-dimensional NbSe2 with Machine Learning
John A. Paulson School of Engineering and Applied Sciences, Harvard University(哈佛大学约翰·A·保尔森工程与应用科学学院);Imperial College London(伦敦帝国理工学院);Aalto University(阿尔托大学);Department of Physics, Harvard University(哈佛大学物理系);University of Oxford(牛津大学)
Fast and Accurate Prediction of Lattice Thermal Conductivity via Machine Learning Surrogates
University of Notre Dame(圣母大学);The University of Tokyo(东京大学);Nanyang Technological University(南洋理工大学);Berkeley Education Alliance for Research in Singapore(新加坡伯克利教育研究联盟);Research Organization of Information and Systems(信息系统研究机构);Nara Institute of Science and Technology(奈良先端科学技术大学院大学);RIKEN Center for Advanced Intelligence Project(理化学研究所先进智能研究中心)
Data-Driven Thermal and Mechanical Modeling of Defective Covalent Organic Frameworks
Institute for Materials Science and Max Bergmann Center of Biomaterials, TUD Dresden University of Technology(德累斯顿工业大学材料科学与马克斯·伯格曼生物材料中心);Leibniz Universität Hannover(汉诺威莱布尼茨大学);Cluster of Excellence PhoenixD, Leibniz Universität Hannover(汉诺威莱布尼茨大学凤凰D卓越集群);Dresden Center for Computational Materials Science (DCMS), TUD Dresden University of Technology(德累斯顿计算材料科学中心(DCMS),德累斯顿工业大学);Cluster of Excellence CARE, TU Dresden and RWTH Aachen(Aachen工业大学与德累斯顿工业大学CARE卓越集群);Cluster of Excellence CeTI, TU Dresden(德累斯顿工业大学CeTI卓越集群)