Chip-level voltage controlled magnetic p-bits array with variation compensation for stochastic neural network computation

Chip-level voltage controlled magnetic p-bits array with variation compensation for stochastic neural network computation | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Chip-level voltage controlled magnetic p-bits array with variation compensation for stochastic neural network computation Yi Cao, Jiachen Bao, Songsong Li, Ruizhi Ren, Zheng Zhu, Di Wu, and 10 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8549564/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The increasing computational and energy-efficiency demands of artificial intelligence are driving the development of in-memory computing based on emerging memory devices. Probabilistic bits (p-bits) based on magnetic tunnel junction (MTJ), benefiting from non-volatility, CMOS compatibility, and fast sampling, show broad potential in stochastic neural networks. However, unavoidable device-to-device variations in MTJ switching probabilities hiders the realization of scalable p-bit chips. Here, we integrate a high-density array of voltage-controlled magnetic anisotropy MTJs (VCMA-MTJs) and employ a tail truncation strategy to compensate for their variations in probabilistic switching ranges, enabling chip-level p-bit functionality. Our VCMA-MTJ chip (VM-chip) successfully implements restricted Boltzmann machines (RBM) and deep belief network (DBN) for image recognition, anomaly detection, image completion, and image generation, demonstrating the practical utility of this chip in stochastic neural networks. Furthermore, we propose a dual-control strategy that tunes both the amplitude and duration of the voltage pulse, significantly enhancing network performance and surpassing the results obtained using conventional software-generated pseudo-random number. Our work enhances the system tolerance to variations in MTJ switching probabilities, thereby reducing manufacturing costs and facilitating large-scale production, ultimately accelerating the chip-level hardware deployment of p-bits in artificial intelligence applications. Physical sciences/Materials science/Materials for devices/Electronic devices Physical sciences/Physics/Electronics, photonics and device physics/Electronic and spintronic devices Physical sciences/Nanoscience and technology/Nanoscale devices/Magnetic devices Full Text Additional Declarations There is NO Competing Interest. Supplementary Files NEsupplementaryinformation.docx Chip-level voltage controlled magnetic p-bits array with variation compensation for stochastic neural network computation Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8549564","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":580628729,"identity":"34b0d470-292e-4728-82a8-ce7514e13781","order_by":0,"name":"Yi Cao","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxElEQVRIiWNgGAWjYLCChAogIQHEPMRrOQNUTZoWxjZStPC3Hz4m8XDeYXt76QbGB2/bGOTNCWmROJOWJpG47XBij8wBZsO5bQyGOxsIaDFgyDEDaUngkUhgk+ZtY0gwOEBIC/8boJY5h+2BWth/E6dFAmRLw2HGHqAtzERpkbjxLNki4Vh6Ys+dg82Sc85JGG4gpIW/P/ngzR811vbss5sPfnhTZiNP0BYgYJGA0IwNDJA0QBgwfyBK2SgYBaNgFIxcAAAyPjuXfDF6QQAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0001-7718-0394","institution":"University of Science and Technology Beijing","correspondingAuthor":true,"prefix":"","firstName":"Yi","middleName":"","lastName":"Cao","suffix":""},{"id":580628730,"identity":"3d3bbbab-68b7-4be2-aec2-dfd5ae032494","order_by":1,"name":"Jiachen Bao","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Jiachen","middleName":"","lastName":"Bao","suffix":""},{"id":580628731,"identity":"9f822a38-de7c-403b-81d0-74247c913222","order_by":2,"name":"Songsong Li","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Songsong","middleName":"","lastName":"Li","suffix":""},{"id":580628732,"identity":"e698863d-f74f-426e-b06e-3308fe21a458","order_by":3,"name":"Ruizhi Ren","email":"","orcid":"","institution":"College of Sciences, Northeastern University","correspondingAuthor":false,"prefix":"","firstName":"Ruizhi","middleName":"","lastName":"Ren","suffix":""},{"id":580628733,"identity":"54a6f45e-9266-44a8-a664-e8b673ac57ca","order_by":4,"name":"Zheng Zhu","email":"","orcid":"","institution":"Institute of Integrated Circuits，Shanghai University","correspondingAuthor":false,"prefix":"","firstName":"Zheng","middleName":"","lastName":"Zhu","suffix":""},{"id":580628734,"identity":"d4d61b45-7f9f-40a7-981b-c9edbc76fb7b","order_by":5,"name":"Di Wu","email":"","orcid":"","institution":"Institute of Integrated Circuits，Shanghai University","correspondingAuthor":false,"prefix":"","firstName":"Di","middleName":"","lastName":"Wu","suffix":""},{"id":580628735,"identity":"fc4041a1-7048-4fa0-b360-3fb6ad279536","order_by":6,"name":"Jiayong Wang","email":"","orcid":"","institution":"Institute of Integrated Circuits，Shanghai University","correspondingAuthor":false,"prefix":"","firstName":"Jiayong","middleName":"","lastName":"Wang","suffix":""},{"id":580628736,"identity":"bdd034ed-817f-43f3-abc1-45acc899c809","order_by":7,"name":"Minggao Zuo","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Minggao","middleName":"","lastName":"Zuo","suffix":""},{"id":580628737,"identity":"0309c219-e6ec-4f55-bce6-3e5f35bb17a4","order_by":8,"name":"Mingke Li","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Mingke","middleName":"","lastName":"Li","suffix":""},{"id":580628738,"identity":"e1d1ca85-a977-412b-9d07-75314386275f","order_by":9,"name":"Duanzihua Cheng","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Duanzihua","middleName":"","lastName":"Cheng","suffix":""},{"id":580628739,"identity":"f6d57030-b728-4356-830c-06a1f5b83945","order_by":10,"name":"Jianyu Lv","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Jianyu","middleName":"","lastName":"Lv","suffix":""},{"id":580628740,"identity":"209d8e1d-adc4-4748-9fb9-3fcd4e95337d","order_by":11,"name":"Lintong Zou Zou","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Lintong","middleName":"Zou","lastName":"Zou","suffix":""},{"id":580628741,"identity":"ff22e56f-7462-4c05-bd2d-f2c581a60b1f","order_by":12,"name":"Chuyue Liu","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Chuyue","middleName":"","lastName":"Liu","suffix":""},{"id":580628742,"identity":"52801810-c119-42e5-8a2c-a624695623f8","order_by":13,"name":"Wangyang Hu","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Wangyang","middleName":"","lastName":"Hu","suffix":""},{"id":580628743,"identity":"b8ee0a97-3602-48e0-b14a-0bc13fbe8b12","order_by":14,"name":"Yuesheng Li Li","email":"","orcid":"","institution":"Institute of Integrated Circuits，Shanghai University","correspondingAuthor":false,"prefix":"","firstName":"Yuesheng","middleName":"Li","lastName":"Li","suffix":""},{"id":580628744,"identity":"46d6e886-72b3-42e9-8198-deed0b84ddee","order_by":15,"name":"Guanghua Yu","email":"","orcid":"","institution":"University of Science and Technology Beijing","correspondingAuthor":false,"prefix":"","firstName":"Guanghua","middleName":"","lastName":"Yu","suffix":""}],"badges":[],"createdAt":"2026-01-08 09:25:56","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8549564/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8549564/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108823251,"identity":"a462cc9e-f37b-48ed-85e3-c5c78cc32a3d","added_by":"auto","created_at":"2026-05-08 16:52:58","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1210341,"visible":true,"origin":"","legend":"","description":"","filename":"NEmanuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8549564/v1_covered_ea1641f6-edfb-48d2-ae56-9ccf7124881c.pdf"},{"id":101311566,"identity":"71d9d193-1003-4d98-99ac-84e137cdf897","added_by":"auto","created_at":"2026-01-28 10:59:28","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":4653582,"visible":true,"origin":"","legend":"Chip-level voltage controlled magnetic p-bits array with variation compensation for stochastic neural network computation","description":"","filename":"NEsupplementaryinformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-8549564/v1/ba4aec7ea7da7c7984311e35.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Chip-level voltage controlled magnetic p-bits array with variation compensation for stochastic neural network computation","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8549564/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8549564/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The increasing computational and energy-efficiency demands of artificial intelligence are driving the development of in-memory computing based on emerging memory devices. Probabilistic bits (p-bits) based on magnetic tunnel junction (MTJ), benefiting from non-volatility, CMOS compatibility, and fast sampling, show broad potential in stochastic neural networks. However, unavoidable device-to-device variations in MTJ switching probabilities hiders the realization of scalable p-bit chips. Here, we integrate a high-density array of voltage-controlled magnetic anisotropy MTJs (VCMA-MTJs) and employ a tail truncation strategy to compensate for their variations in probabilistic switching ranges, enabling chip-level p-bit functionality. Our VCMA-MTJ chip (VM-chip) successfully implements restricted Boltzmann machines (RBM) and deep belief network (DBN) for image recognition, anomaly detection, image completion, and image generation, demonstrating the practical utility of this chip in stochastic neural networks. Furthermore, we propose a dual-control strategy that tunes both the amplitude and duration of the voltage pulse, significantly enhancing network performance and surpassing the results obtained using conventional software-generated pseudo-random number. Our work enhances the system tolerance to variations in MTJ switching probabilities, thereby reducing manufacturing costs and facilitating large-scale production, ultimately accelerating the chip-level hardware deployment of p-bits in artificial intelligence applications.","manuscriptTitle":"Chip-level voltage controlled magnetic p-bits array with variation compensation for stochastic neural network computation","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-28 10:59:17","doi":"10.21203/rs.3.rs-8549564/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"6ca95116-d6fc-43e9-b1de-b78280c57282","owner":[],"postedDate":"January 28th, 2026","published":true,"recentEditorialEvents":[{"type":"decision","content":"Reject after peer review","date":"2026-05-08T16:24:24+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":61767720,"name":"Physical sciences/Materials science/Materials for devices/Electronic devices"},{"id":61767721,"name":"Physical sciences/Physics/Electronics, photonics and device physics/Electronic and spintronic devices"},{"id":61767722,"name":"Physical sciences/Nanoscience and technology/Nanoscale devices/Magnetic devices"}],"tags":[],"updatedAt":"2026-05-08T16:50:37+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-28 10:59:17","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8549564","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8549564","identity":"rs-8549564","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}