{"title":"用于预测固态特性的图神经网络力场的通用性","authors":"Shaswat Mohanty, Yifan Wang, Wei Cai","doi":"arxiv-2409.09931","DOIUrl":null,"url":null,"abstract":"Machine-learned force fields (MLFFs) promise to offer a computationally\nefficient alternative to ab initio simulations for complex molecular systems.\nHowever, ensuring their generalizability beyond training data is crucial for\ntheir wide application in studying solid materials. This work investigates the\nability of a graph neural network (GNN)-based MLFF, trained on Lennard-Jones\nArgon, to describe solid-state phenomena not explicitly included during\ntraining. We assess the MLFF's performance in predicting phonon density of\nstates (PDOS) for a perfect face-centered cubic (FCC) crystal structure at both\nzero and finite temperatures. Additionally, we evaluate vacancy migration rates\nand energy barriers in an imperfect crystal using direct molecular dynamics\n(MD) simulations and the string method. Notably, vacancy configurations were\nabsent from the training data. Our results demonstrate the MLFF's capability to\ncapture essential solid-state properties with good agreement to reference data,\neven for unseen configurations. We further discuss data engineering strategies\nto enhance the generalizability of MLFFs. The proposed set of benchmark tests\nand workflow for evaluating MLFF performance in describing perfect and\nimperfect crystals pave the way for reliable application of MLFFs in studying\ncomplex solid-state materials.","PeriodicalId":501162,"journal":{"name":"arXiv - MATH - Numerical Analysis","volume":"21 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Generalizability of Graph Neural Network Force Fields for Predicting Solid-State Properties\",\"authors\":\"Shaswat Mohanty, Yifan Wang, Wei Cai\",\"doi\":\"arxiv-2409.09931\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Machine-learned force fields (MLFFs) promise to offer a computationally\\nefficient alternative to ab initio simulations for complex molecular systems.\\nHowever, ensuring their generalizability beyond training data is crucial for\\ntheir wide application in studying solid materials. This work investigates the\\nability of a graph neural network (GNN)-based MLFF, trained on Lennard-Jones\\nArgon, to describe solid-state phenomena not explicitly included during\\ntraining. We assess the MLFF's performance in predicting phonon density of\\nstates (PDOS) for a perfect face-centered cubic (FCC) crystal structure at both\\nzero and finite temperatures. Additionally, we evaluate vacancy migration rates\\nand energy barriers in an imperfect crystal using direct molecular dynamics\\n(MD) simulations and the string method. Notably, vacancy configurations were\\nabsent from the training data. Our results demonstrate the MLFF's capability to\\ncapture essential solid-state properties with good agreement to reference data,\\neven for unseen configurations. We further discuss data engineering strategies\\nto enhance the generalizability of MLFFs. The proposed set of benchmark tests\\nand workflow for evaluating MLFF performance in describing perfect and\\nimperfect crystals pave the way for reliable application of MLFFs in studying\\ncomplex solid-state materials.\",\"PeriodicalId\":501162,\"journal\":{\"name\":\"arXiv - MATH - Numerical Analysis\",\"volume\":\"21 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-16\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - MATH - Numerical Analysis\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.09931\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - MATH - Numerical Analysis","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.09931","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Generalizability of Graph Neural Network Force Fields for Predicting Solid-State Properties
Machine-learned force fields (MLFFs) promise to offer a computationally
efficient alternative to ab initio simulations for complex molecular systems.
However, ensuring their generalizability beyond training data is crucial for
their wide application in studying solid materials. This work investigates the
ability of a graph neural network (GNN)-based MLFF, trained on Lennard-Jones
Argon, to describe solid-state phenomena not explicitly included during
training. We assess the MLFF's performance in predicting phonon density of
states (PDOS) for a perfect face-centered cubic (FCC) crystal structure at both
zero and finite temperatures. Additionally, we evaluate vacancy migration rates
and energy barriers in an imperfect crystal using direct molecular dynamics
(MD) simulations and the string method. Notably, vacancy configurations were
absent from the training data. Our results demonstrate the MLFF's capability to
capture essential solid-state properties with good agreement to reference data,
even for unseen configurations. We further discuss data engineering strategies
to enhance the generalizability of MLFFs. The proposed set of benchmark tests
and workflow for evaluating MLFF performance in describing perfect and
imperfect crystals pave the way for reliable application of MLFFs in studying
complex solid-state materials.