[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-127282-en":3,"doc-seo-127282-105":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},127282,2336475104736,"Quinn","https://ap-avatar.wpscdn.com/avatar/22000c4c5e0e5b17e70?x-image-process=image/resize,m_fixed,w_180,h_180&k=1786591360781797222",8,"Research & Report","From Electrons to Phase Diagrams with Machine Learning Potentials - Using Pyiron-Based Automated Workflows - Supplementary Information","Supplementary information details domain-driven training dataset generation and active learning for an ACE machine-learning interatomic potential using pyiron-based automated workflows. It describes unary and binary crystal structures, random deformation protocols, elastic and vibrational calculations, phonon computation, and vacancy supercells. Active learning employs MD in the NPT ensemble across 1–1500 K, selection via extrapolation grades, DFT retraining, and two active-learning rounds. It also outlines random-structure filtering, dataset statistics, and functional definitions underpinning the EAM-style potential used.","Supplementary Information for ‘From electrons to phase diagrams with machine learning potentials using pyiron based automated workflows’  \nSarath Menon , 1, ∗ Yury Lysogorskiy ,2 Alexander L. M. Knoll ,3, 4 Niklas Leimeroth  \n,5 Marvin Poul , 1 Minaam Qamar ,2 Jan Janssen , 1 Matous Mrovec ,2 Jochen Rohrer ,5 Karsten Albe ,5 J¨org Behler ,3, 4 Ralf Drautz ,2 and J¨org Neugebauer 1,†  \n1 Max-Planck-Institut f¨ur Eisenforschung GmbH, 40237 D¨usseldorf, Germany  \n2 ICAMS, Ruhr-Universit¨at Bochum, 44801 Bochum, Germany  \n3 Lehrstuhl f¨ur Theoretische Chemie II, Ruhr-Universit¨at Bochum, 44780 Bochum, Germany  \n4 Research Center Chemical Sciences and Sustainability,  \nResearch Alliance Ruhr, 44780 Bochum, Germany  \n5 Technische Universit¨at Darmstadt, Fachbereich Material und Geowissenschaften,  \nFachgebiet Materialmodellierung, 64287 Darmstadt, Germany  \n(Dated: July 17, 2024)  \nSupplementary Note 1 . DOMAIN-DRIVEN DATA  \nSET GENERATION  \nFirst generation of the training dataset contains following structures:  \n1. Perfect unary crystals (fcc/bcc/hcp) of Al and Li. For each structure, the following properties were computed: an energy-nearest neighbor distance curve (from 2 to 6 .5 ˚A with a 0 .5 ˚A step), full structural relaxation, an energy-volume curve around equilibrium volume (±10% with a 2% step), elastic matrix calculations (with 5 points along each deformation mode in a ±0.5% strain range), phonons (as determined by Phonopy [1, 2]), and a supercell with a single vacancy.  \n2. Binary prototypes from the Materials Project, that contains Al and Li: Li2Al mp-1210753, LiAl mp- 1067, LiAl3 mp-10890, Li9Al4 mp-568404, LiAlmp-1079240, LiAl mp-1191737, Li3Al2 mp-16506) . For each structure, the same steps as in p.1 were performed.  \n3. Randomly deformed supercells. For each of the optimized structures from p. 1 and 2, a new supercell was constructed in such a way that its minimal length of the cell vector was more than 7 ˚A . For each supercell, five random deformations were generated. Each deformation consists of random normal atom displacements with σ = 0 .05 ˚A anda random normal cell deformation with σ = 0 .05. For each of the five random deformations, 11 uniformly isotropic deformations from-10% to +10% with a 2% step were generated.  \nThis dataset was utilized to train the zeroth generation of the ACE potential. An active learning procedure  \n∗ [s.menon@mpie.de](s.menon@mpie.de)[ ](s.menon@mpie.de)† [neugebauer@mpie.de](neugebauer@mpie.de)  \nwas employed with this potential to sample more configurations. New configurations were generated through MD simulations in the NPT ensemble with zero pressure and increasing temperatures from 1 to 1500K over 15,000 steps of supercells from p.3 . Only structures with a maximum per-atom extrapolation grade exceeding 5 were selected. If the extrapolation grade exceeded 20, simulations were halted. Extrapolation grades were computed every 5th MD step. The number of captured configurations for different crystal structure types ranged from 6 to 261 . In total, 491 structures were collected during the first round of active learning. These structures were computed with DFT, added to the training set, and the ACE potential was retrained. In the second round of active learning, the same procedure as before was applied, but MD ran for 50,000 steps with a steady temperature increase from 1 to 1500K, followed by an additional 50,000 steps at T=1500K. A total of 225 configurations were collected, ranging from 4 to 146 configurations per crystal structure.  \nSupplementary Note 2 . RANDOM CRYSTAL  \nSTRUCTURES  \nSupplementary Table 1 show the parameters used to generate the training data. During all steps of this procedure we remove structures that have atomic distances below 1 .9 ˚A to avoid overlapping PAW spheres, which would negatively impact the quality of the training data.  \nSupplementary Figure 1 shows the distribution of Li concentration in the full training set, Supplementary Figure 2 the distr","cbCaihmDJ7nozVzA","https://ap.wps.com/l/cbCaihmDJ7nozVzA","pdf",1990745,1,7,"English","en",105,"# Supplementary Note 1: Domain-Driven Data Set Generation\n## Training structures and computed properties\n## Active learning workflow and configuration selection\n# Supplementary Note 2: Random Crystal Structures\n## Dataset generation parameter constraints\n## Training set distributions and convex hull\n# Supplementary Note 3: Functions in EAM Potential\n## Pair function V and cutoff\n## Electron density ρ and embedding term\n# Supplementary Table 1: Hyperparameters","[{\"question\":\"What structures are included in the first training dataset generation?\",\"answer\":\"The first generation includes perfect unary fcc/bcc/hcp crystals of Al and Li, and binary prototypes from the Materials Project containing Al and Li, plus randomly deformed supercells derived from the optimized structures.\"},{\"question\":\"How does the active learning procedure generate new configurations?\",\"answer\":\"The ACE potential is used to run MD simulations in the NPT ensemble with zero pressure while temperatures increase from 1 to 1500 K. Configurations are selected based on a per-atom extrapolation grade, and simulations are stopped when the grade exceeds 20.\"},{\"question\":\"What key functions define the EAM-style potential described in the supplementary note?\",\"answer\":\"The note defines a pair function V based on Morse-function components with short-range repulsive terms and a cutoff function, an electron density ρ constructed from Gaussian-like terms, and an embedding term F(ρ) defined separately for ρ \\u003c 1 and ρ \\u003e 1.\"}]","From Electrons to Phase Diagrams with Machine Learning Potentials - Using Pyiron-Based Automated Workflows - Supplementary Information | PDF",1785938088,18,{"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},"from-electrons-to-phase-diagrams-with-machine-learning-potentials-using-pyiron-based-automated-workflows-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/from-electrons-to-phase-diagrams-with-machine-learning-potentials-using-pyiron-based-automated-workflows-supplementary-information/127282/",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},"What structures are included in the first training dataset generation?","Question",{"text":75,"@type":76},"The first generation includes perfect unary fcc/bcc/hcp crystals of Al and Li, and binary prototypes from the Materials Project containing Al and Li, plus randomly deformed supercells derived from the optimized structures.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the active learning procedure generate new configurations?",{"text":80,"@type":76},"The ACE potential is used to run MD simulations in the NPT ensemble with zero pressure while temperatures increase from 1 to 1500 K. Configurations are selected based on a per-atom extrapolation grade, and simulations are stopped when the grade exceeds 20.",{"name":82,"@type":73,"acceptedAnswer":83},"What key functions define the EAM-style potential described in the supplementary note?",{"text":84,"@type":76},"The note defines a pair function V based on Morse-function components with short-range repulsive terms and a cutoff function, an electron density ρ constructed from Gaussian-like terms, and an embedding term F(ρ) defined separately for ρ \u003C 1 and ρ > 1.","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},[93,97,101,105,110,115,119,122,127,130,134],{"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":21,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"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"]