A novel approach to describe chemical environments in high-dimensional neural network potentials

A central concern of molecular dynamics simulations is the potential energy surfaces that govern atomic interactions. These hypersurfaces define the potential energy of the system and have generally been calculated using either predefined analytical formulas (classical) or quantum mechanical simulat...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:The Journal of chemical physics 2019-04, Vol.150 (15), p.154102-154102
Hauptverfasser: Kocer, Emir, Mason, Jeremy K., Erturk, Hakan
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 154102
container_issue 15
container_start_page 154102
container_title The Journal of chemical physics
container_volume 150
creator Kocer, Emir
Mason, Jeremy K.
Erturk, Hakan
description A central concern of molecular dynamics simulations is the potential energy surfaces that govern atomic interactions. These hypersurfaces define the potential energy of the system and have generally been calculated using either predefined analytical formulas (classical) or quantum mechanical simulations (ab initio). The former can accurately reproduce only a selection of material properties, whereas the latter is restricted to short simulation times and small systems. Machine learning potentials have recently emerged as a third approach to model atomic interactions, and are purported to offer the accuracy of ab initio simulations with the speed of classical potentials. However, the performance of machine learning potentials depends crucially on the description of a local atomic environment. A set of invariant, orthogonal, and differentiable descriptors for an atomic environment is proposed, implemented in a neural network potential for solid-state silicon, and tested in molecular dynamics simulations. Neural networks using the proposed descriptors are found to outperform ones using the Behler–Parinello and smooth overlap of atomic position descriptors in the literature.
doi_str_mv 10.1063/1.5086167
format Article
fullrecord <record><control><sourceid>proquest_scita</sourceid><recordid>TN_cdi_scitation_primary_10_1063_1_5086167</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2212721271</sourcerecordid><originalsourceid>FETCH-LOGICAL-c383t-96450111f2adad5c4454952086a0dd8ed8630b14dc506cd92ac02acbae9ba95f3</originalsourceid><addsrcrecordid>eNp90E1LxDAQBuAgiruuHvwDEvCiQtdJ2_TjuCx-wYIXPdc0SW3WtqlJuuK_N-uuCoIehiHwzJB5ETomMCWQRJdkSiFLSJLuoDGBLA_SJIddNAYISZAnkIzQgbVLACBpGO-jUUQAqB8do6cZ7vRKNpj1vdGM19hpLKTlRpUS81q2irMGy26ljO5a2TmLVYdr9VwHQvm3VbrzoJOD-WzuTZsX3GvnqWKNPUR7lW_yaNsn6PH66mF-Gyzub-7ms0XAoyxy_pcxBUJIFTLBBOVxTOOchv4sBkJkUmRJBCWJBaeQcJGHjIOvksm8ZDmtogk62-z1Z7wO0rqiVZbLpmGd1IMtwpCE6bqIp6e_6FIPxl-xVpCnNKZp5tX5RnGjrTWyKnqjWmbeCwLFOvaCFNvYvT3ZbhzKVopv-ZWzBxcbYLlyzPnM_t32J15p8wOLXlTRB10LmQ0</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2209754578</pqid></control><display><type>article</type><title>A novel approach to describe chemical environments in high-dimensional neural network potentials</title><source>AIP Journals Complete</source><source>Alma/SFX Local Collection</source><creator>Kocer, Emir ; Mason, Jeremy K. ; Erturk, Hakan</creator><creatorcontrib>Kocer, Emir ; Mason, Jeremy K. ; Erturk, Hakan</creatorcontrib><description>A central concern of molecular dynamics simulations is the potential energy surfaces that govern atomic interactions. These hypersurfaces define the potential energy of the system and have generally been calculated using either predefined analytical formulas (classical) or quantum mechanical simulations (ab initio). The former can accurately reproduce only a selection of material properties, whereas the latter is restricted to short simulation times and small systems. Machine learning potentials have recently emerged as a third approach to model atomic interactions, and are purported to offer the accuracy of ab initio simulations with the speed of classical potentials. However, the performance of machine learning potentials depends crucially on the description of a local atomic environment. A set of invariant, orthogonal, and differentiable descriptors for an atomic environment is proposed, implemented in a neural network potential for solid-state silicon, and tested in molecular dynamics simulations. Neural networks using the proposed descriptors are found to outperform ones using the Behler–Parinello and smooth overlap of atomic position descriptors in the literature.</description><identifier>ISSN: 0021-9606</identifier><identifier>EISSN: 1089-7690</identifier><identifier>DOI: 10.1063/1.5086167</identifier><identifier>PMID: 31005106</identifier><identifier>CODEN: JCPSA6</identifier><language>eng</language><publisher>United States: American Institute of Physics</publisher><subject>Artificial intelligence ; Atomic interactions ; Computer simulation ; Hyperspaces ; Machine learning ; Material properties ; Molecular dynamics ; Neural networks ; Organic chemistry ; Potential energy ; Quantum mechanics ; Simulation</subject><ispartof>The Journal of chemical physics, 2019-04, Vol.150 (15), p.154102-154102</ispartof><rights>Author(s)</rights><rights>2019 Author(s). Published under license by AIP Publishing.</rights><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c383t-96450111f2adad5c4454952086a0dd8ed8630b14dc506cd92ac02acbae9ba95f3</citedby><cites>FETCH-LOGICAL-c383t-96450111f2adad5c4454952086a0dd8ed8630b14dc506cd92ac02acbae9ba95f3</cites><orcidid>0000-0003-4207-8220 ; 0000-0001-5564-3845 ; 0000-0002-0425-9816 ; 0000000204259816 ; 0000000155643845 ; 0000000342078220</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://pubs.aip.org/jcp/article-lookup/doi/10.1063/1.5086167$$EHTML$$P50$$Gscitation$$H</linktohtml><link.rule.ids>314,780,784,794,4512,27924,27925,76384</link.rule.ids><backlink>$$Uhttps://www.ncbi.nlm.nih.gov/pubmed/31005106$$D View this record in MEDLINE/PubMed$$Hfree_for_read</backlink></links><search><creatorcontrib>Kocer, Emir</creatorcontrib><creatorcontrib>Mason, Jeremy K.</creatorcontrib><creatorcontrib>Erturk, Hakan</creatorcontrib><title>A novel approach to describe chemical environments in high-dimensional neural network potentials</title><title>The Journal of chemical physics</title><addtitle>J Chem Phys</addtitle><description>A central concern of molecular dynamics simulations is the potential energy surfaces that govern atomic interactions. These hypersurfaces define the potential energy of the system and have generally been calculated using either predefined analytical formulas (classical) or quantum mechanical simulations (ab initio). The former can accurately reproduce only a selection of material properties, whereas the latter is restricted to short simulation times and small systems. Machine learning potentials have recently emerged as a third approach to model atomic interactions, and are purported to offer the accuracy of ab initio simulations with the speed of classical potentials. However, the performance of machine learning potentials depends crucially on the description of a local atomic environment. A set of invariant, orthogonal, and differentiable descriptors for an atomic environment is proposed, implemented in a neural network potential for solid-state silicon, and tested in molecular dynamics simulations. Neural networks using the proposed descriptors are found to outperform ones using the Behler–Parinello and smooth overlap of atomic position descriptors in the literature.</description><subject>Artificial intelligence</subject><subject>Atomic interactions</subject><subject>Computer simulation</subject><subject>Hyperspaces</subject><subject>Machine learning</subject><subject>Material properties</subject><subject>Molecular dynamics</subject><subject>Neural networks</subject><subject>Organic chemistry</subject><subject>Potential energy</subject><subject>Quantum mechanics</subject><subject>Simulation</subject><issn>0021-9606</issn><issn>1089-7690</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2019</creationdate><recordtype>article</recordtype><recordid>eNp90E1LxDAQBuAgiruuHvwDEvCiQtdJ2_TjuCx-wYIXPdc0SW3WtqlJuuK_N-uuCoIehiHwzJB5ETomMCWQRJdkSiFLSJLuoDGBLA_SJIddNAYISZAnkIzQgbVLACBpGO-jUUQAqB8do6cZ7vRKNpj1vdGM19hpLKTlRpUS81q2irMGy26ljO5a2TmLVYdr9VwHQvm3VbrzoJOD-WzuTZsX3GvnqWKNPUR7lW_yaNsn6PH66mF-Gyzub-7ms0XAoyxy_pcxBUJIFTLBBOVxTOOchv4sBkJkUmRJBCWJBaeQcJGHjIOvksm8ZDmtogk62-z1Z7wO0rqiVZbLpmGd1IMtwpCE6bqIp6e_6FIPxl-xVpCnNKZp5tX5RnGjrTWyKnqjWmbeCwLFOvaCFNvYvT3ZbhzKVopv-ZWzBxcbYLlyzPnM_t32J15p8wOLXlTRB10LmQ0</recordid><startdate>20190421</startdate><enddate>20190421</enddate><creator>Kocer, Emir</creator><creator>Mason, Jeremy K.</creator><creator>Erturk, Hakan</creator><general>American Institute of Physics</general><scope>NPM</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>8FD</scope><scope>H8D</scope><scope>L7M</scope><scope>7X8</scope><orcidid>https://orcid.org/0000-0003-4207-8220</orcidid><orcidid>https://orcid.org/0000-0001-5564-3845</orcidid><orcidid>https://orcid.org/0000-0002-0425-9816</orcidid><orcidid>https://orcid.org/0000000204259816</orcidid><orcidid>https://orcid.org/0000000155643845</orcidid><orcidid>https://orcid.org/0000000342078220</orcidid></search><sort><creationdate>20190421</creationdate><title>A novel approach to describe chemical environments in high-dimensional neural network potentials</title><author>Kocer, Emir ; Mason, Jeremy K. ; Erturk, Hakan</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c383t-96450111f2adad5c4454952086a0dd8ed8630b14dc506cd92ac02acbae9ba95f3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2019</creationdate><topic>Artificial intelligence</topic><topic>Atomic interactions</topic><topic>Computer simulation</topic><topic>Hyperspaces</topic><topic>Machine learning</topic><topic>Material properties</topic><topic>Molecular dynamics</topic><topic>Neural networks</topic><topic>Organic chemistry</topic><topic>Potential energy</topic><topic>Quantum mechanics</topic><topic>Simulation</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Kocer, Emir</creatorcontrib><creatorcontrib>Mason, Jeremy K.</creatorcontrib><creatorcontrib>Erturk, Hakan</creatorcontrib><collection>PubMed</collection><collection>CrossRef</collection><collection>Technology Research Database</collection><collection>Aerospace Database</collection><collection>Advanced Technologies Database with Aerospace</collection><collection>MEDLINE - Academic</collection><jtitle>The Journal of chemical physics</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Kocer, Emir</au><au>Mason, Jeremy K.</au><au>Erturk, Hakan</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>A novel approach to describe chemical environments in high-dimensional neural network potentials</atitle><jtitle>The Journal of chemical physics</jtitle><addtitle>J Chem Phys</addtitle><date>2019-04-21</date><risdate>2019</risdate><volume>150</volume><issue>15</issue><spage>154102</spage><epage>154102</epage><pages>154102-154102</pages><issn>0021-9606</issn><eissn>1089-7690</eissn><coden>JCPSA6</coden><abstract>A central concern of molecular dynamics simulations is the potential energy surfaces that govern atomic interactions. These hypersurfaces define the potential energy of the system and have generally been calculated using either predefined analytical formulas (classical) or quantum mechanical simulations (ab initio). The former can accurately reproduce only a selection of material properties, whereas the latter is restricted to short simulation times and small systems. Machine learning potentials have recently emerged as a third approach to model atomic interactions, and are purported to offer the accuracy of ab initio simulations with the speed of classical potentials. However, the performance of machine learning potentials depends crucially on the description of a local atomic environment. A set of invariant, orthogonal, and differentiable descriptors for an atomic environment is proposed, implemented in a neural network potential for solid-state silicon, and tested in molecular dynamics simulations. Neural networks using the proposed descriptors are found to outperform ones using the Behler–Parinello and smooth overlap of atomic position descriptors in the literature.</abstract><cop>United States</cop><pub>American Institute of Physics</pub><pmid>31005106</pmid><doi>10.1063/1.5086167</doi><tpages>9</tpages><orcidid>https://orcid.org/0000-0003-4207-8220</orcidid><orcidid>https://orcid.org/0000-0001-5564-3845</orcidid><orcidid>https://orcid.org/0000-0002-0425-9816</orcidid><orcidid>https://orcid.org/0000000204259816</orcidid><orcidid>https://orcid.org/0000000155643845</orcidid><orcidid>https://orcid.org/0000000342078220</orcidid></addata></record>
fulltext fulltext
identifier ISSN: 0021-9606
ispartof The Journal of chemical physics, 2019-04, Vol.150 (15), p.154102-154102
issn 0021-9606
1089-7690
language eng
recordid cdi_scitation_primary_10_1063_1_5086167
source AIP Journals Complete; Alma/SFX Local Collection
subjects Artificial intelligence
Atomic interactions
Computer simulation
Hyperspaces
Machine learning
Material properties
Molecular dynamics
Neural networks
Organic chemistry
Potential energy
Quantum mechanics
Simulation
title A novel approach to describe chemical environments in high-dimensional neural network potentials
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-22T20%3A25%3A46IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_scita&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=A%20novel%20approach%20to%20describe%20chemical%20environments%20in%20high-dimensional%20neural%20network%20potentials&rft.jtitle=The%20Journal%20of%20chemical%20physics&rft.au=Kocer,%20Emir&rft.date=2019-04-21&rft.volume=150&rft.issue=15&rft.spage=154102&rft.epage=154102&rft.pages=154102-154102&rft.issn=0021-9606&rft.eissn=1089-7690&rft.coden=JCPSA6&rft_id=info:doi/10.1063/1.5086167&rft_dat=%3Cproquest_scita%3E2212721271%3C/proquest_scita%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2209754578&rft_id=info:pmid/31005106&rfr_iscdi=true