FigsharePub Date : 2021-10-04DOI: 10.1021/jacs.1c08794.s001
Tian Lan (186455), Qi An (331241)
{"title":"DiscoveringnCatalytic Reaction Networks Using DeepnReinforcement Learning from First-Principles","authors":"Tian Lan (186455), Qi An (331241)","doi":"10.1021/jacs.1c08794.s001","DOIUrl":"https://doi.org/10.1021/jacs.1c08794.s001","url":null,"abstract":"Determining the reaction pathways,nwhich is central to illustratingnthe working mechanisms of a catalyst, is severely hindered by thenhigh complexity of the reaction and the extreme scarcity of the data.nHere, we develop a novel artificial intelligence framework integratingndeep reinforcement learning (DRL) techniques with density functionalntheory simulations to automate the quantitative search and evaluationnon the complex catalytic reaction networks from zero knowledge. Ournframework quantitatively transforms the first-principles-derived freenenergy landscape of the chemical reactions to a DRL environment andnthe corresponding actions. By interacting with this dynamic environment,nour model evolves by itself from scratch to a complete reaction path.nWe demonstrate this framework using the Haber-Bosch process on thenmost active Fe(111) surface. The new path found by our framework hasna lower overall free energy barrier than the previous study basednon domain knowledge, demonstrating its outstanding capability in discoveringncomplicated reaction paths. Looking forward, we anticipate that thisnframework will open the door to exploring the fundamental reactionnmechanisms of many catalytic reactions.","PeriodicalId":522265,"journal":{"name":"Figshare","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2021-10-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147894349","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}