Figshare最新文献

筛选

英文中文

EOSnet: EmbeddednOverlap Structures for Graph NeuralnNetworks in Predicting Material Properties 图神经网络在预测材料性能中的嵌入重叠结构

Figshare Pub Date : 2025-01-11 DOI: 10.1021/acs.jpclett.4c03179.s001

Shuo Tao (14625633), Li Zhu (67404)

{"title":"EOSnet: EmbeddednOverlap Structures for Graph NeuralnNetworks in Predicting Material Properties","authors":"Shuo Tao (14625633), Li Zhu (67404)","doi":"10.1021/acs.jpclett.4c03179.s001","DOIUrl":"https://doi.org/10.1021/acs.jpclett.4c03179.s001","url":null,"abstract":"Graph Neural Networks (GNNs) have emerged as powerfulntools fornpredicting material properties, yet they often struggle to capturenmany-body interactions and require extensive manual feature engineering.nHere, we present EOSnet (Embedded Overlap Structures for Graph NeuralnNetworks), a novel approach that addresses these limitations by incorporatingnGaussian Overlap Matrix (GOM) fingerprints as node features withinnthe GNN architecture. Unlike models that rely on explicit angularnterms or human-engineered features, EOSnet efficiently encodes many-bodyninteractions through orbital overlap matrices, providing a rotationallyninvariant and transferable representation of atomic environments.nThe model demonstrates superior performance across various predictionntasks of materials’ properties, achieving particularly notablenresults in properties sensitive to many-body interactions. For bandngap prediction, EOSnet achieves a mean absolute error of 0.163 eV,nsurpassing previous state-of-the-art models. The model also excelsnin predicting mechanical properties and classifying materials, withn97.7% accuracy in metal/nonmetal classification. These results demonstratenthat embedding GOM fingerprints into node features enhances the abilitynof GNNs to capture complex atomic interactions, making EOSnet a powerfulntool for materials’ discovery and property prediction.","PeriodicalId":522232,"journal":{"name":"Figshare","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2025-01-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147893796","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0