[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-126655-en":3,"doc-seo-126655-105":30,"detail-sidebar-cat-0-en-105":91},{"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":4,"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},126655,549768064778,"Finn","https://ap-avatar.wpscdn.com/davatar_6f874abed73319feea01a86fa6f0fab8",8,"Research & Report","Accurate Energy Barriers for Catalytic Reaction Pathways - An Automatic Training Protocol for Machine Learning Force Fields - Supplementary Information","Supplementary material details an automatic machine-learning force-field training protocol for catalytic reaction pathways, with validation against a referenced DFT reaction path. Adsorption-energy errors are tracked across active-learning and optimization iterations, including geometry optimization and nudged elastic band steps, and are assessed using convergence and termination criteria. The document provides adsorption-energy convergence statistics, minimum-characterization via single-point DFT forces, and several visualizations of reference and newly found minimum-energy paths, alongside dataset overview and methodology settings comparisons.","Supplementary Information  \nAccurate Energy Barriers for Catalytic Reaction Pathways: An Automatic Training Protocol for Machine Learning Force Fields  \nLars L Schaaf1 , Edvin Fako2 , Sandip De2 , Ansgar Schäfer2 , and Gábor Csányi 1  \n1 Engineering Laboratory, University of Cambridge, Cambridge, CB2 1PZ UK  \n2 BASF SE, Carl-Bosch-Straße 38, 67056 Ludwigshafen, Germany  \nSupplementary Figures  \nAdsorption energies errors E!\"\\# -E $%& averaged across intermediaries [eV]  \n: geometry optimization iterations  \n: nudged elastic bands iterations  \nSupplementary Figure 1: Adsorption energy error across iterations. For validating our method against the Ref [1] path, we show how the GAP adsorption energies converge to the true DFT ones . Aobve we see the adsorption energy error across geometry optimization and nudged elastic band iterations. The tables show the mean, standard deviation, minimum, and maximum energy errors across all 12 intermediates for each iteration. The graphs give more detailed insight into each individual intermediate.  \nmax D FT force [eV/A]  \n1.6  \n1.4  \n1.2  \n1.0  \n0.8  \n0.6  \n0.4  \n0.2  \n0.0  \nintermediate  O vac.  CO2  \n CO2 + H  HCOOmono  \n HCOObi  \n HCOO + H  \n H2COO  \n H2CO  \n H2CO + HO + H  \n H3CO + H  \n H3CO + H(2)  CH3OH  \nSupplementary Figure 2: GAP minima characterisation. At each iteration the minima found by GAP are evaluated using single-point DFT. The DFT force quantifies how close the GAP minima are to true DFT minima. After the 11th iteration, all minima are below the 0.2 eV/A cutoff, and hence the termination criteria for the geometry optimization active-learning block is reached.  \nSupplementary Figure 3: Ref [1] path for HCOO + H  H2COO Note that this is simply a trajectory, that follows a similar path to Ref [1] . It is not the exact images of the Ref [1] NEB, which include four images. Here we show all 10 images of the NEBs found during this investigation. Note on all visualisations: the surface atoms are scaled down to 70% of their default size and connected by bonds. The molecular atoms are kept at 100% size. The difference in the oxygen coordination of the two indium atoms of this oxygen vacancy can be easily distinguished in this view. Note that in the paper, surface atoms are colored in grey for clarity. In the main text, Indium and Oxygen atoms can be distinguished by their size.  \nSupplementary Figure 4: Optimised path for HCOO + H  H2COO This is the newly found minimum energy path for the reaction step that was thought to be rate limiting.  \nSupplementary Figure 5: Ref [1] path for H2 COO  H2 CO + O Note that this is simply a trajectory, that follows a similar path to Ref [1] It is not the exact images of the Ref [1] NEB, which included four images. Here we show all 10 images of the NEBs found during this investigation.  \nSupplementary Figure 6: Thermodynamically relevant minimum energy path for H2 COO  H2 CO + O.  \nThis is the newly found minimum energy path for the third reaction step.  \nSupplementary Figure 7: Visualisation of the top-layer reduced reaction pathway  \nSupplementary Notes  \nSupplementary Note 1: Dataset overview  \nThe dataset accumulated throughout active learning can be obtained from an online repository as detailed in the data availability section. The table below outlines the number of configurations in the training set throughout the training protocol for the training dataset curated for validation against Dang et al [1] . Each configuration in the online repository is labeled with details on which active learning (AL) block and which iteration the configurations originate from.  \n\n| AL block | block name | iterations | counts |\n| --- | --- | --- | --- |\n| 0 | initial dataset | 0: isolated | 4 |\n| 1 | bulk | 0: material project | 1 |\n|  |  | 0: rattled + deformed | 10 |\n|  |  | 1 | 7 |\n| 2 | surfaces | 0\u003Cbr>1 | 2\u003Cbr>20 |\n| 3 | molecules-md | 0: approximate intermediates | 11 |\n|  |  | 0: catkit | 36 |\n|  |  | 0: dimers | 24 |\n| 4 | molecules-geom-opt | 1-6\u003Cbr>1","cbCaidQsbBNG0aWP","https://ap.wps.com/l/cbCaidQsbBNG0aWP","pdf",12607972,1,12,"English","en",105,"# Supplementary Figures\n## Adsorption energy error across iterations\n## GAP minima characterisation\n## Reference and newly optimised reaction paths\n# Supplementary Notes\n## Dataset overview\n## Obtaining approximate intermediates\n# Supplementary Methods\n## Vasp vs QE","[{\"question\":\"How is adsorption energy error evaluated across iterations in the protocol?\",\"answer\":\"Adsorption energy errors are averaged across intermediates for each iteration, then summarized with mean, standard deviation, minimum, and maximum values. Convergence is inspected across both geometry-optimization and nudged elastic band iterations.\"},{\"question\":\"What criterion determines when the geometry optimization active-learning block terminates?\",\"answer\":\"After the 11th iteration, all minima are reported to fall below a 0.2 eV/Å cutoff using DFT forces, indicating the termination criteria for the active-learning geometry optimization block are met.\"},{\"question\":\"How are approximate intermediates obtained for benchmarking against the referenced DFT study?\",\"answer\":\"Minima from the referenced path are geometry optimized using less accurate, lower-cost DFT settings. The reduced settings aim to emulate early-path-search results while keeping the starting configurations sufficiently away from true minima so adsorption-energy convergence remains meaningful.\"}]","Accurate Energy Barriers for Catalytic Reaction Pathways - An Automatic Training Protocol for Machine Learning Force Fields - Supplementary Information | PDF",1785934061,30,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":28},"accurate-energy-barriers-for-catalytic-reaction-pathways-an-automatic-training-protocol-for-machine-learning-force-fields-supplementary-information","",{"@graph":36,"@context":85},[37,54,68],{"@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/accurate-energy-barriers-for-catalytic-reaction-pathways-an-automatic-training-protocol-for-machine-learning-force-fields-supplementary-information/126655/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-05",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"How is adsorption energy error evaluated across iterations in the protocol?","Question",{"text":75,"@type":76},"Adsorption energy errors are averaged across intermediates for each iteration, then summarized with mean, standard deviation, minimum, and maximum values. Convergence is inspected across both geometry-optimization and nudged elastic band iterations.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What criterion determines when the geometry optimization active-learning block terminates?",{"text":80,"@type":76},"After the 11th iteration, all minima are reported to fall below a 0.2 eV/Å cutoff using DFT forces, indicating the termination criteria for the active-learning geometry optimization block are met.",{"name":82,"@type":73,"acceptedAnswer":83},"How are approximate intermediates obtained for benchmarking against the referenced DFT study?",{"text":84,"@type":76},"Minima from the referenced path are geometry optimized using less accurate, lower-cost DFT settings. The reduced settings aim to emulate early-path-search results while keeping the starting configurations sufficiently away from true minima so adsorption-energy convergence remains meaningful.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":29,"slug":121},"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]