[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-128100-en":3,"doc-seo-128100-105":30,"detail-sidebar-cat-0-en-105":96},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},128100,687207022233,"Riley","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","Crash testing machine learning force fields for molecules, materials, and interfaces - Molecular dynamics in the TEA challenge 2023 - rigorous evaluation of modern MLFFs","Second-part rigorous evaluation of modern machine learning force fields (MLFFs) within the TEA Challenge 2023. The work analyzes performance of MACE, sGDML, SO3krates, SOAP/GAP, and FCHL19 for modeling molecules, molecule–surface interfaces, and periodic materials. Identical-condition molecular-dynamics simulations produce comparable observables across MLFFs, benchmarked against density-functional theory or experiment when available. Without DFT, results are compared across MLFF architectures. The study finds practitioners’ choice of ML model matters only when the problem is outside the model’s scope; otherwise, results show weak architecture dependence, while long-range noncovalent interactions remain challenging.","Open Access Art 2025. Down on 3/31/2025icle . Published on 03 February loaded 3:36:24 PM .  \nhicle is licensed under a Creative C mmons A 3 0 U d[ttr .](ttr .)ibution npor e nce.  \nChemical Science  \n| EDGE ARTICLE | View Article Online\u003Cbr>View Journal | View Issue |\n| --- | --- |\n| \u003Cbr>Cite this: Chem. Sci., 2025, 16, 3738\u003Cbr>\u003Cbr> All publication charges for this article | Crash testing machine learning force fields for molecules, materials, and interfaces: molecular dynamics in the TEA challenge 2023† |\n| have been paid for by the Royal Society of Chemistry | Igor Poltavsky,  *a Mirela Puleva,‡ab Anton Charkin-Gorbulin,‡ac Grgory Fonseca,‡a Ilyes Batatia,d Nicholas J. Browning,e Stefan Chmiela,fg Mengnan Cui,h\u003Cbr>J. Thorben Frank, fg Stefan Heinen,i Bing Huang,j Silvan Kser,  k Adil Kabylda,  a Danish Khan,il Carolin Müller,  m Alastair J. A. Price,no Kai Riedmiller,  q Kai Töpfer,  k Tsz Wai Ko,p Markus Meuwly,  k Matthias Rupp,  r Gbor Csnyi, d\u003Cbr>O. Anatole von Lilienfeld, fginost Johannes T. Margraf,u Klaus-Robert Müller fguvwx and Alexandre Tkatchenko  *ab |\n| Received 26th September 2024 Accepted 25th December 2024\u003Cbr>DOI: 10.1039/d4sc06530a | We present the second part of the rigorous evaluation of modern machine learning force fields (MLFFs) within the TEA Challenge 2023 . This study provides an in-depth analysis of the performance of MACE, SO3krates, sGDML, SOAP/GAP, and FCHL19 * in modeling molecules, molecule-surface interfaces, and periodic materials. We compare observables obtained from molecular dynamics (MD) simulations using different MLFFs under identical conditions. Where applicable, density-functional theory (DFT) or experiment serves as a reference to reliably assess the performance of the ML models. In the absence of DFT benchmarks, we conduct a comparative analysis based on results from various MLFF architectures. Our findings indicate that, at the current stage of MLFF development, the choice of ML model is in the hands of the practitioner. When a problem falls within the scope of a given MLFF architecture, the resulting simulations exhibit weak dependency on the specific architecture used. Instead, emphasis should be placed on developing complete, reliable, and representative training datasets. Nonetheless, long-range noncovalent interactions remain challenging for all MLFF models, necessitating special caution in simulations of physical systems where such interactions are prominent, such as molecule- |\n| [rsc.li/chemical-science](rsc.li/chemical-science) | surface interfaces. The findings presented here reflect the state of MLFF models as of October 2023 . |\n\naDepartment of Physics and Materials Science, University of Luxembourg, L-1511 Luxembourg, Luxembourg. E-mail: [alexandre.tkatchenko@uni.lu](alexandre.tkatchenko@uni.lu); igor.poltavskyi@ [uni.lu](uni.lu)  \nbInstitute for Advanced Studies, University of Luxembourg, Campus Belval, L-4365 Esch-sur-Alzette, Luxembourg  \ncLaboratory for Chemistry of Novel Materials, University of Mons, B-7000 Mons, Belgium  \ndDepartment of Engineering, University of Cambridge, Trumpington Street, Cambridge, CB2 1PZ, UK  \neSwiss National Supercomputing Centre (CSCS), 6900 Lugano, Switzerland fMachine Learning Group, Technical University Berlin, Berlin, Germany gBIFOLD, Berlin Institute for the Foundations of Learning and Data, Berlin, Germany hFritz-Haber-Institut der Max-Planck-Gesellscha􀀁, Berlin, Germany  \niVector Institute for Arti􀀁cial Intelligence, Toronto, ON, M5S 1M1, Canada jWuhan University, Department of Chemistry and Molecular Sciences, 430072 Wuhan, China  \nkDepartment of Chemistry, University of Basel, Klingelbergstrasse 80, CH-4056 Basel, Switzerland  \nlChemical Physics Theory Group, Department of Chemistry, University of Toronto, St. George Campus, Toronto, ON, Canada  \nmFriedrich-Alexander-Universitt Erlangen-Nürnberg, Computer-Chemistry-Center, Ngelsbachstraße 25, 91052 Erlangen, Germany  \nnDepartment of Chemistry, University of Toronto, St. George campus,","cbCaiqh6WyGP4khJ","https://ap.wps.com/l/cbCaiqh6WyGP4khJ","pdf",4492913,1,17,"English","en",105,"# Introduction\n## Rigorous evaluation framework in TEA Challenge 2023\n## MLFF architectures compared\n## Benchmarking strategy (DFT/experiment or cross-MLFF comparisons)\n## Practical implications and limitations","[{\"question\":\"Which MLFF models are evaluated in the TEA Challenge 2023 study?\",\"answer\":\"The study evaluates MACE, SO3krates, sGDML, SOAP/GAP, and FCHL19 for multiple target systems.\"},{\"question\":\"How are MLFFs compared in this work?\",\"answer\":\"MLFFs are assessed via molecular-dynamics simulations under identical conditions, using consistent observables for comparison.\"},{\"question\":\"What does the study say about model architecture choice?\",\"answer\":\"When a problem fits the scope of a given MLFF architecture, simulations show weak dependence on the specific architecture, shifting emphasis toward complete and representative training datasets.\"},{\"question\":\"What remains difficult across MLFF models?\",\"answer\":\"Long-range noncovalent interactions are reported as challenging for all MLFF models, requiring extra caution for systems where these interactions are prominent, such as molecule–surface interfaces.\"}]","Crash testing machine learning force fields for molecules, materials, and interfaces - Molecular dynamics in the TEA challenge 2023 - rigorous evaluation of modern MLFFs | PDF",1785944808,43,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":91,"head_meta":93,"extra_data":95,"updated_unix":28},"crash-testing-machine-learning-force-fields-for-molecules-materials-and-interfaces-molecular-dynamics-in-the-tea-challenge-2023-rigorous-evaluation-of-modern-mlffs","",{"@graph":36,"@context":90},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/crash-testing-machine-learning-force-fields-for-molecules-materials-and-interfaces-molecular-dynamics-in-the-tea-challenge-2023-rigorous-evaluation-of-modern-mlffs/128100/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-25","2026-08-05",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82,86],{"name":73,"@type":74,"acceptedAnswer":75},"Which MLFF models are evaluated in the TEA Challenge 2023 study?","Question",{"text":76,"@type":77},"The study evaluates MACE, SO3krates, sGDML, SOAP/GAP, and FCHL19 for multiple target systems.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"How are MLFFs compared in this work?",{"text":81,"@type":77},"MLFFs are assessed via molecular-dynamics simulations under identical conditions, using consistent observables for comparison.",{"name":83,"@type":74,"acceptedAnswer":84},"What does the study say about model architecture choice?",{"text":85,"@type":77},"When a problem fits the scope of a given MLFF architecture, simulations show weak dependence on the specific architecture, shifting emphasis toward complete and representative training datasets.",{"name":87,"@type":74,"acceptedAnswer":88},"What remains difficult across MLFF models?",{"text":89,"@type":77},"Long-range noncovalent interactions are reported as challenging for all MLFF models, requiring extra caution for systems where these interactions are prominent, such as molecule–surface interfaces.","https://schema.org",{"og:url":52,"og:type":92,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":94,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":97},[98,102,106,110,115,120,125,128,133,136,140],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},"Exam",70,"exam",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},5,"Comic",60,"comic",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},6,"Technology",50,"technology",{"id":121,"doc_module":4,"doc_module_name":46,"category_name":122,"show_sort_weight":123,"slug":124},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":126,"slug":127},30,"research-report",{"id":129,"doc_module":4,"doc_module_name":46,"category_name":130,"show_sort_weight":131,"slug":132},9,"Religion & Spirituality",20,"religion-spirituality",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":134,"show_sort_weight":131,"slug":135},"World Cup","world-cup",{"id":137,"doc_module":4,"doc_module_name":46,"category_name":138,"show_sort_weight":137,"slug":139},10,"Lifestyle","lifestyle",{"id":141,"doc_module":4,"doc_module_name":46,"category_name":142,"show_sort_weight":111,"slug":143},19,"General","general"]