{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,26]],"date-time":"2026-03-26T03:57:23Z","timestamp":1774497443453,"version":"3.50.1"},"reference-count":58,"publisher":"IOP Publishing","issue":"3","license":[{"start":{"date-parts":[[2025,7,14]],"date-time":"2025-07-14T00:00:00Z","timestamp":1752451200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2025,7,14]],"date-time":"2025-07-14T00:00:00Z","timestamp":1752451200000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/iopscience.iop.org\/info\/page\/text-and-data-mining"}],"funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"crossref","id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"crossref"}]},{"name":"State Key Laboratory of Physical Chemistry of Solid Surfaces"}],"content-domain":{"domain":["iopscience.iop.org"],"crossmark-restriction":false},"short-container-title":["Mach. Learn.: Sci. Technol."],"published-print":{"date-parts":[[2025,9,30]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    Active learning (AL) requires massive time for comprehensive sampling of complex potential energy surfaces to achieve desirable accuracy and stability of machine learning (ML) potentials. Here, we develop an active delta-learning (ADL) protocol for speeding up AL and building delta-learning models yielding stable simulations. ADL converges after a few iterations and needs tenfold fewer sampled points than without delta-learning while leading to models of similar accuracy, as we show on the test simulations of Diels\u2013Alder reactions. The test reactions include one small (ethene + 1,3-butadiene) and one relatively big (C\n                    <jats:sub>60<\/jats:sub>\n                    + 2,3-dimethyl-1,3-butadiene) system, treated with a target density functional theory level (U)B3LYP(-D4)\/6-31G* and a baseline semi-empirical quantum mechanical method, GFN2-xTB. The crucial advantage of the models built with the delta-learning protocol is their remarkable simulation stability: even models from the initial ADL iterations yield reasonable results. In contrast, the pure ML potentials built without delta-learning often lead to the collapse in simulations, i.e. to unphysical structures.\n                  <\/jats:p>","DOI":"10.1088\/2632-2153\/adeb46","type":"journal-article","created":{"date-parts":[[2025,7,2]],"date-time":"2025-07-02T19:01:36Z","timestamp":1751482896000},"page":"035004","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":4,"title":["Active delta-learning for fast construction of interatomic potentials and stable molecular dynamics simulations"],"prefix":"10.1088","volume":"6","author":[{"ORCID":"https:\/\/orcid.org\/0000-0001-6942-3499","authenticated-orcid":true,"given":"Yaohuang","family":"Huang","sequence":"first","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9188-5323","authenticated-orcid":true,"given":"Yi-Fan","family":"Hou","sequence":"additional","affiliation":[]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2975-9876","authenticated-orcid":true,"given":"Pavlo O","family":"Dral","sequence":"additional","affiliation":[]}],"member":"266","published-online":{"date-parts":[[2025,7,14]]},"reference":[{"key":"mlstadeb46bib1","doi-asserted-by":"publisher","first-page":"457","DOI":"10.1002\/andp.19273892002","article-title":"Zur Quantentheorie der Molekeln","volume":"389","author":"Born","year":"1927","journal-title":"Ann. Phys., NY"},{"key":"mlstadeb46bib2","doi-asserted-by":"publisher","first-page":"e1290","DOI":"10.1002\/wcms.1290","article-title":"Challenges in large scale quantum mechanical calculations","volume":"7","author":"Ratcliff","year":"2017","journal-title":"Wires. Comput. Mol. Sci."},{"key":"mlstadeb46bib3","doi-asserted-by":"publisher","first-page":"10142","DOI":"10.1021\/acs.chemrev.0c01111","article-title":"Machine learning force fields","volume":"121","author":"Unke","year":"2021","journal-title":"Chem. Rev."},{"key":"mlstadeb46bib4","doi-asserted-by":"publisher","first-page":"10218","DOI":"10.1021\/acs.chemrev.1c00033","article-title":"Machine learning for chemical reactions","volume":"121","author":"Meuwly","year":"2021","journal-title":"Chem. Rev."},{"key":"mlstadeb46bib5","doi-asserted-by":"publisher","first-page":"eadf0873","DOI":"10.1126\/sciadv.adf0873","article-title":"Accurate global machine learning force fields for molecules with hundreds of atoms","volume":"9","author":"Chmiela","year":"2023","journal-title":"Sci. Adv."},{"key":"mlstadeb46bib6","doi-asserted-by":"publisher","first-page":"139","DOI":"10.1016\/j.sbi.2019.12.016","article-title":"Machine learning approaches for analyzing and enhancing molecular dynamics simulations","volume":"61","author":"Wang","year":"2020","journal-title":"Curr. Opin. Struct. Biol."},{"key":"mlstadeb46bib7","doi-asserted-by":"publisher","first-page":"17","DOI":"10.1063\/1.4966192","article-title":"Perspective: machine learning potentials for atomistic simulations","volume":"145","author":"Behler","year":"2016","journal-title":"J. Chem. Phys."},{"key":"mlstadeb46bib8","doi-asserted-by":"publisher","first-page":"750","DOI":"10.1038\/s41563-020-0777-6","article-title":"Machine-learned potentials for next-generation matter simulations","volume":"20","author":"Friederich","year":"2021","journal-title":"Nat. Mater."},{"key":"mlstadeb46bib9","doi-asserted-by":"publisher","first-page":"10037","DOI":"10.1021\/acs.chemrev.0c00868","article-title":"Four generations of high-dimensional neural network potentials","volume":"121","author":"Behler","year":"2021","journal-title":"Chem. Rev."},{"key":"mlstadeb46bib10","doi-asserted-by":"publisher","first-page":"10187","DOI":"10.1021\/acs.chemrev.0c00665","article-title":"Neural network potential energy surfaces for small molecules and reactions","volume":"121","author":"Manzhos","year":"2021","journal-title":"Chem. Rev."},{"key":"mlstadeb46bib11","doi-asserted-by":"publisher","first-page":"9759","DOI":"10.1021\/acs.chemrev.1c00021","article-title":"Physics-inspired structural representations for molecules and materials","volume":"121","author":"Musil","year":"2021","journal-title":"Chem. Rev."},{"key":"mlstadeb46bib12","doi-asserted-by":"publisher","first-page":"14396","DOI":"10.1039\/D1SC03564A","article-title":"Choosing the right molecular machine learning potential","volume":"12","author":"Pinheiro Jr","year":"2021","journal-title":"Chem. Sci."},{"key":"mlstadeb46bib13","doi-asserted-by":"publisher","first-page":"315","DOI":"10.1613\/jair.1199","article-title":"Learning when training data are costly: the effect of class distribution on tree induction","volume":"19","author":"Weiss","year":"2003","journal-title":"J. Artif. Intell. Res."},{"key":"mlstadeb46bib14","doi-asserted-by":"publisher","first-page":"6589","DOI":"10.1021\/acs.jctc.3c00391","article-title":"Reducing the cost of neural network potential generation for reactive molecular systems","volume":"19","author":"Brezina","year":"2023","journal-title":"J. Chem. Theory. Comput."},{"key":"mlstadeb46bib15","first-page":"313","article-title":"Constructing machine learning potentials with active learning","author":"Shang","year":"2023"},{"key":"mlstadeb46bib16","doi-asserted-by":"publisher","first-page":"6338","DOI":"10.1021\/acs.chemmater.0c00768","article-title":"Quantum chemistry-informed active learning to accelerate the design and discovery of sustainable energy storage materials","volume":"32","author":"Doan","year":"2020","journal-title":"Chem. Mater."},{"key":"mlstadeb46bib17","doi-asserted-by":"publisher","first-page":"24","DOI":"10.1063\/1.5023802","article-title":"Less is more: sampling chemical space with active learning","volume":"148","author":"Smith","year":"2018","journal-title":"J. Chem. Phys."},{"key":"mlstadeb46bib18","doi-asserted-by":"publisher","DOI":"10.1016\/j.cpc.2020.107206","article-title":"DP-GEN: a concurrent learning platform for the generation of reliable deep learning based potential energy models","volume":"253","author":"Zhang","year":"2020","journal-title":"Comput. Phys. Commun."},{"key":"mlstadeb46bib19","doi-asserted-by":"publisher","first-page":"7744","DOI":"10.1021\/acs.jctc.4c00821","article-title":"Physics-informed active learning for accelerating quantum chemical simulations","volume":"20","author":"Hou","year":"2024","journal-title":"J. Chem. Theory. Comput."},{"key":"mlstadeb46bib20","doi-asserted-by":"publisher","first-page":"738","DOI":"10.1016\/j.chempr.2020.12.009","article-title":"Active learning accelerates ab initio molecular dynamics on reactive energy surfaces","volume":"7","author":"Ang","year":"2021","journal-title":"Chemistry"},{"key":"mlstadeb46bib21","doi-asserted-by":"publisher","first-page":"10944","DOI":"10.1039\/D1SC01825F","article-title":"A transferable active-learning strategy for reactive molecular force fields","volume":"12","author":"Young","year":"2021","journal-title":"Chem. Sci."},{"key":"mlstadeb46bib22","doi-asserted-by":"publisher","first-page":"23467","DOI":"10.1039\/d3cp03515h","article-title":"Energy-conserving molecular dynamics is not energy conserving","volume":"25","author":"Zhang","year":"2023","journal-title":"Phys. Chem. Chem. Phys."},{"key":"mlstadeb46bib23","doi-asserted-by":"publisher","first-page":"3720","DOI":"10.1039\/d4sc06529h","article-title":"Crash testing machine learning force fields for molecules, materials, and interfaces: model analysis in the TEA challenge 2023","volume":"16","author":"Poltavsky","year":"2025","journal-title":"Chem. Sci."},{"key":"mlstadeb46bib24","doi-asserted-by":"publisher","first-page":"225","DOI":"10.1038\/s41524-023-01180-8","article-title":"Single-model uncertainty quantification in neural network potentials does not consistently outperform model ensembles","volume":"9","author":"Tan","year":"2023","journal-title":"npj Comput. Mater."},{"key":"mlstadeb46bib25","doi-asserted-by":"publisher","first-page":"4451","DOI":"10.1021\/acs.jpclett.4c00746","article-title":"Tell machine learning potentials what they are needed for: simulation-oriented training exemplified for glycine","volume":"15","author":"Ge","year":"2024","journal-title":"J. Phys. Chem. Lett."},{"key":"mlstadeb46bib26","doi-asserted-by":"publisher","first-page":"2087","DOI":"10.1021\/acs.jctc.5b00099","article-title":"Big data meets quantum chemistry approximations: the \u0394-machine learning approach","volume":"11","author":"Ramakrishnan","year":"2015","journal-title":"J. Chem. Theory. Comput."},{"key":"mlstadeb46bib27","doi-asserted-by":"publisher","DOI":"10.1088\/2632-2153\/abcc44","article-title":"Enabling robust offline active learning for machine learning potentials using simple physics-based priors","volume":"2","author":"Shuaibi","year":"2020","journal-title":"Mach. Learn.: Sci. Technol."},{"key":"mlstadeb46bib28","author":"McQuarrie","year":"1973"},{"key":"mlstadeb46bib29","doi-asserted-by":"publisher","first-page":"15041","DOI":"10.1021\/acs.joc.4c01763","article-title":"Surprising dynamics phenomena in the Diels-Alder reaction of C60 uncovered with AI","volume":"89","author":"Hou","year":"2024","journal-title":"J. Org. Chem."},{"key":"mlstadeb46bib30","doi-asserted-by":"publisher","first-page":"764","DOI":"10.1016\/j.jesp.2013.03.013","article-title":"Detecting outliers: do not use standard deviation around the mean, use absolute deviation around the median","volume":"49","author":"Leys","year":"2013","journal-title":"J. Exp. Soc. Psychol."},{"key":"mlstadeb46bib31","doi-asserted-by":"publisher","first-page":"3192","DOI":"10.1039\/C6SC05720A","article-title":"ANI-1: an extensible neural network potential with DFT accuracy at force field computational cost","volume":"8","author":"Smith","year":"2017","journal-title":"Chem. Sci."},{"key":"mlstadeb46bib32","doi-asserted-by":"publisher","DOI":"10.1103\/PhysRevLett.98.146401","article-title":"Generalized neural-network representation of high-dimensional potential-energy surfaces","volume":"98","author":"Behler","year":"2007","journal-title":"Phys. Rev. Lett."},{"key":"mlstadeb46bib33","doi-asserted-by":"publisher","first-page":"7607","DOI":"10.1038\/s41467-024-51886-6","article-title":"E(n)-equivariant cartesian tensor message passing interatomic potential","volume":"15","author":"Wang","year":"2024","journal-title":"Nat. Commun."},{"key":"mlstadeb46bib34","doi-asserted-by":"publisher","first-page":"7022","DOI":"10.1038\/s41467-021-27340-2","article-title":"Artificial intelligence-enhanced quantum chemical method with broad applicability","volume":"12","author":"Zheng","year":"2021","journal-title":"Nat. Commun."},{"key":"mlstadeb46bib35","doi-asserted-by":"publisher","DOI":"10.26434\/chemrxiv-2024-j8pxp","article-title":"AIQM2: better reaction simulations with the 2nd generation of general-purpose AI-enhanced quantum mechanical method","author":"Chen","year":"2024"},{"key":"mlstadeb46bib36","doi-asserted-by":"publisher","first-page":"554","DOI":"10.1002\/cber.19290620318","article-title":"The Diels-Alder reaction","volume":"62","author":"Diels","year":"1929","journal-title":"Ber. Dtsch. Chem. Ges."},{"key":"mlstadeb46bib37","doi-asserted-by":"publisher","first-page":"405","DOI":"10.1021\/cr60219a002","article-title":"Diels-Alder syntheses with heteroatomic compounds","volume":"62","author":"Needleman","year":"1962","journal-title":"Chem. Rev."},{"key":"mlstadeb46bib38","doi-asserted-by":"publisher","DOI":"10.1063\/1.4993215","article-title":"Extension of the D3 dispersion coefficient model","volume":"147","author":"Caldeweyher","year":"2017","journal-title":"J. Chem. Phys."},{"key":"mlstadeb46bib39","doi-asserted-by":"publisher","DOI":"10.1063\/5.0137101","article-title":"Benchmark of general-purpose machine learning-based quantum mechanical method AIQM1 on reaction barrier heights","volume":"158","author":"Chen","year":"2023","journal-title":"J. Chem. Phys."},{"key":"mlstadeb46bib40","doi-asserted-by":"publisher","first-page":"4365","DOI":"10.1021\/acs.jctc.5c00347","article-title":"ANI-1xBB: an ANI-based reactive potential for small organic molecules","volume":"21","author":"Zhang","year":"2025","journal-title":"J. Chem. Theory. Comput."},{"key":"mlstadeb46bib41","doi-asserted-by":"publisher","first-page":"12860","DOI":"10.1073\/pnas.1209316109","article-title":"Dynamics, transition states, and timing of bond formation in Diels-Alder reactions","volume":"109","author":"Black","year":"2012","journal-title":"Proc. Natl Acad. Sci."},{"key":"mlstadeb46bib42","doi-asserted-by":"publisher","first-page":"1652","DOI":"10.1021\/acs.jctc.8b01176","article-title":"GFN2-xTB\u2014an accurate and broadly parametrized self-consistent tight-binding quantum chemical method with multipole electrostatics and density-dependent dispersion contributions","volume":"15","author":"Bannwarth","year":"2019","journal-title":"J. Chem. Theory. Comput."},{"key":"mlstadeb46bib43","doi-asserted-by":"publisher","first-page":"2903","DOI":"10.1038\/s41467-019-10827-4","article-title":"Approaching coupled cluster accuracy with a general-purpose neural network potential through transfer learning","volume":"10","author":"Smith","year":"2019","journal-title":"Nat. Commun."},{"key":"mlstadeb46bib44","doi-asserted-by":"publisher","first-page":"3479","DOI":"10.1021\/acs.jpclett.2c00734","article-title":"Toward chemical accuracy in predicting enthalpies of formation with general-purpose data-driven methods","volume":"13","author":"Zheng","year":"2022","journal-title":"J. Phys. Chem. Lett."},{"key":"mlstadeb46bib45","doi-asserted-by":"publisher","first-page":"727","DOI":"10.1038\/s41557-023-01427-3","article-title":"Exploring the frontiers of condensed-phase chemistry with a general reactive machine learning potential","volume":"16","author":"Zhang","year":"2024","journal-title":"Nat. Chem."},{"key":"mlstadeb46bib46","doi-asserted-by":"publisher","first-page":"1546","DOI":"10.1021\/acs.jctc.8b00832","article-title":"Boosting quantum machine learning models with a multilevel combination technique: Pople diagrams revisited","volume":"15","author":"Zaspel","year":"2019","journal-title":"J. Chem. Theory. Comput."},{"key":"mlstadeb46bib47","doi-asserted-by":"publisher","DOI":"10.1088\/2632-2153\/ad4ae5","article-title":"Reducing training data needs with minimal multilevel machine learning (M3L)","volume":"5","author":"Heinen","year":"2024","journal-title":"Mach. Learn.: Sci. Technol."},{"key":"mlstadeb46bib48","doi-asserted-by":"publisher","DOI":"10.1063\/5.0006498","article-title":"Hierarchical machine learning of potential energy surfaces","volume":"152","author":"Dral","year":"2020","journal-title":"J. Chem. Phys."},{"key":"mlstadeb46bib49","doi-asserted-by":"publisher","DOI":"10.26434\/chemrxiv-2024-ng3ws","article-title":"All-in-one foundational models learning across quantum chemical levels","author":"Chen","year":"2025"},{"key":"mlstadeb46bib50","doi-asserted-by":"publisher","DOI":"10.26434\/chemrxiv-2025-j207x","article-title":"OMNI-P2x: a universal neural network potential for excited-state simulations","author":"Martyka","year":"2025"},{"key":"mlstadeb46bib51","first-page":"491","article-title":"Learning from multiple quantum chemical methods: \u0394-learning, transfer learning, co-kriging, and beyond","author":"Dral","year":"2023"},{"key":"mlstadeb46bib52","doi-asserted-by":"publisher","first-page":"1193","DOI":"10.1021\/acs.jctc.3c01203","article-title":"MLatom 3: a platform for machine learning-enhanced computational chemistry simulations and workflows","volume":"20","author":"Dral","year":"2024","journal-title":"J. Chem. Theory. Comput."},{"key":"mlstadeb46bib53","author":"Dral","year":"2024"},{"key":"mlstadeb46bib54","article-title":"Gaussian 16 Rev. B.01","author":"Frisch","year":"2016"},{"key":"mlstadeb46bib55","article-title":"Semiempirical extended tight-binding program package xtb","author":"Atkinson","year":"2024"},{"key":"mlstadeb46bib56","doi-asserted-by":"publisher","first-page":"8499","DOI":"10.1039\/D0CP00502A","article-title":"Extension and evaluation of the D4 London-dispersion model for periodic systems","volume":"22","author":"Caldeweyher","year":"2020","journal-title":"Phys. Chem. Chem. Phys."},{"key":"mlstadeb46bib57","author":"Caldeweyher","year":"2024"},{"key":"mlstadeb46bib58","doi-asserted-by":"publisher","first-page":"3408","DOI":"10.1021\/acs.jcim.0c00451","article-title":"TorchANI: a free and open source pytorch-based deep learning implementation of the ANI neural network potentials","volume":"60","author":"Gao","year":"2020","journal-title":"J. Chem. Inf. Model."}],"container-title":["Machine Learning: Science and Technology"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adeb46","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adeb46\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adeb46","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adeb46\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adeb46\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adeb46\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adeb46\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adeb46\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,7,14]],"date-time":"2025-07-14T09:46:19Z","timestamp":1752486379000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adeb46"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,7,14]]},"references-count":58,"journal-issue":{"issue":"3","published-online":{"date-parts":[[2025,7,14]]},"published-print":{"date-parts":[[2025,9,30]]}},"URL":"https:\/\/doi.org\/10.1088\/2632-2153\/adeb46","relation":{"has-preprint":[{"id-type":"doi","id":"10.26434\/chemrxiv-2024-fb02r","asserted-by":"object"}]},"ISSN":["2632-2153"],"issn-type":[{"value":"2632-2153","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,7,14]]},"assertion":[{"value":"Active delta-learning for fast construction of interatomic potentials and stable molecular dynamics simulations","name":"article_title","label":"Article Title"},{"value":"Machine Learning: Science and Technology","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2025 The Author(s). Published by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2025-04-02","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2025-07-01","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2025-07-14","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}