[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-117578-en":3,"doc-seo-117578-105":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},117578,1374391974468,"Eden","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",8,"Research & Report","Introduction to Machine Learning Potentials for Atomistic Simulations","Machine learning potentials provide an efficient, accurate surrogate for the potential energy surface in molecular simulations, enabling interpolation at costs comparable to force-field methods while retaining ab initio accuracy. The work delivers a structured overview covering practical development of machine learning potentials: chemical descriptors, regression model families, strategies for generating and validating training data, and limitations such as locality and descriptor completeness. It contrasts early approaches like HD-NNPs and GAP with modern advances including graph neural networks and general-purpose foundational models, concluding with showcase applications in atomistic simulations.","arXiv :2410 .00626v1 [physics .chem-ph] 1 Oct 2024  \nIntroduction to machine learning potentials for atomistic simulations  \nFabian L. Thiemann, 1, 2 Niamh O’Neill,3, 2, 4 Venkat Kapil,3, 4, 5, 6 Angelos Michaelides,3, 4 and Christoph Schran2, 4, a)  \n1) IBM Research Europe, Daresbury, Warrington, WA4 4AD, UK  \n2) Cavendish Laboratory, Department of Physics, University of Cambridge, Cambridge, CB3 0HE, UK  \n3) Yusuf Hamied Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge, CB2 1EW, UK  \n4) Lennard-Jones Centre, University of Cambridge, Trinity Ln, Cambridge, CB2 1TN, UK  \n5) Department of Physics and Astronomy, University College London, London, UK  \n6) Thomas Young Centre and London Centre for Nanotechnology, London, UK, London, UK  \n(Dated: 2 October 2024)  \nMachine learning potentials have revolutionised the field of atomistic simulations in recent years and are becoming a mainstay in the toolbox of computational scientists. This paper aims to provide an overview and introduction into machine learning potentials and their practical application to scientific problems. We provide a systematic guide for developing machine learning potentials, reviewing chemical descriptors, regression models, data generation and validation approaches. We begin with an emphasis on the earlier generation of models, such as high-dimensional neural network potentials (HD-NNPs) and Gaussian approximation potential (GAP), to provide historical perspective and guide the reader towards the understanding of recent developments, which are discussed in detail thereafter. Furthermore, we refer to relevant expert reviews, open-source software, and practical examples – further lowering the barrier to exploring these methods. The paper ends with selected showcase examples, highlighting the capabilities of machine learning potentials and how they can be applied to push the boundaries in atomistic simulations.  \nI. Introduction 2  \nII. Chemical Descriptors 3  \nA. Atom-Centred Symmetry Functions 3  \nB. Smooth Overlap of Atomic Positions 5  \nC. Discussion and Outlook on Chemical Descriptors 6  \nIII. Regression Models 8  \nA. Artificial neural networks 8  \n1. High–dimensional Neural Network Potentials 9  \n2. Obtaining Analytical Derivatives 10  \n3. Training 10  \nB. Gaussian Process and Kernel Ridge Regression 11  \n1. Weight-Space View 11  \n2. Function-Space View 13  \n3. Gaussian Approximation Potentials 14  \nC. Discussion and Outlook of Regression Models 16  \nIV. Current Developments 17  \nA. Completeness of Descriptors 17  \nB. Learnable Descriptors: Graph Neural Networks 18  \na) Electronic mail: [cs2121@cam.ac.uk](cs2121@cam.ac.uk)  \nC. Beyond Locality 18  \nD. General Purpose and Foundational  \nModels 19  \nV. Data Set Generation 19  \nA. Structural Selection Techniques 20  \nB. Active Learning 20  \nC. Reinforcement Workflows 21  \nVI. Validation 21  \nA. Primary Properties and Numerical Errors 22  \nB. Validation of Secondary Properties 23  \nVII. Showcase Examples 24  \nVIII. Summary and Outlook 26  \nAcknowledgments 28  \nCompeting interests 28  \nSupporting Information 28  \nOverview of Machine Learning  \nConcepts S1  \nOverview of Open-Source Code S2  \n2  \nI. INTRODUCTION  \nMost of the chemistry and physics of molecular systems and materials is governed by the potential energy surface (PES) . Within the Born-Oppenheimer approximation, the properties of a system of interest can thus be obtained from its thermally weighted population on the ground state PES, as sampled either by molecular dynamics or Monte Carlo techniques. Having access to an accurate but efficient representation of the system’s PES is therefore of paramount importance for the computational study of material properties, reactions, and molecular processes. While ab initio techniques such as density functional theory (DFT) can provide the required accuracy for a large variety of complex systems, they are usually relatively expensive as the electronic structure of each sampled config","cbCaich8vSMG8MU5","https://ap.wps.com/l/cbCaich8vSMG8MU5","pdf",6047343,1,35,"English","en",105,"# Introduction\n# Chemical Descriptors\n## Atom-Centred Symmetry Functions\n## Smooth Overlap of Atomic Positions\n# Regression Models\n## Artificial neural networks\n## Gaussian Process and Kernel Ridge Regression\n# Current Developments\n# Data Set Generation\n## Structural Selection Techniques\n## Active Learning\n# Validation\n## Primary Properties and Numerical Errors\n# Showcase Examples\n# Summary and Outlook","[{\"question\":\"Why are accurate potential energy surface representations crucial for atomistic simulations?\",\"answer\":\"Most physical chemistry and materials behavior depends on the potential energy surface (PES). Accurate PES representations allow reliable thermally weighted sampling using molecular dynamics or Monte Carlo, supporting studies of properties, reactions, and molecular processes.\"},{\"question\":\"What are the two core components of machine learning potentials (MLPs)?\",\"answer\":\"MLPs typically consist of an encoding strategy that transforms molecular structure into descriptors, and a regression method mapping atomic configurations to the PES.\"},{\"question\":\"How does the paper guide readers from earlier models to current developments?\",\"answer\":\"It begins with earlier model generations such as high-dimensional neural network potentials (HD-NNPs) and Gaussian approximation potential (GAP) for historical context, then details later developments including learnable descriptors (graph neural networks) and advances beyond locality and toward general-purpose foundational models.\"}]","Introduction to Machine Learning Potentials for Atomistic Simulations | PDF",1785677080,88,{"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},"introduction-to-machine-learning-potentials-for-atomistic-simulations","",{"@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/introduction-to-machine-learning-potentials-for-atomistic-simulations/117578/",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","2026-08-02",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},"Why are accurate potential energy surface representations crucial for atomistic simulations?","Question",{"text":75,"@type":76},"Most physical chemistry and materials behavior depends on the potential energy surface (PES). Accurate PES representations allow reliable thermally weighted sampling using molecular dynamics or Monte Carlo, supporting studies of properties, reactions, and molecular processes.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What are the two core components of machine learning potentials (MLPs)?",{"text":80,"@type":76},"MLPs typically consist of an encoding strategy that transforms molecular structure into descriptors, and a regression method mapping atomic configurations to the PES.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the paper guide readers from earlier models to current developments?",{"text":84,"@type":76},"It begins with earlier model generations such as high-dimensional neural network potentials (HD-NNPs) and Gaussian approximation potential (GAP) for historical context, then details later developments including learnable descriptors (graph neural networks) and advances beyond locality and toward general-purpose foundational models.","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},[93,97,101,105,110,115,120,123,128,131,135],{"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":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]