File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Article: Thousands of conductance levels in memristors integrated on CMOS

TitleThousands of conductance levels in memristors integrated on CMOS
Authors
Issue Date2023
Citation
Nature, 2023, v. 615, n. 7954, p. 823-829 How to Cite?
AbstractNeural networks based on memristive devices1–3 have the ability to improve throughput and energy efficiency for machine learning4,5 and artificial intelligence6, especially in edge applications7–21. Because training a neural network model from scratch is costly in terms of hardware resources, time and energy, it is impractical to do it individually on billions of memristive neural networks distributed at the edge. A practical approach would be to download the synaptic weights obtained from the cloud training and program them directly into memristors for the commercialization of edge applications. Some post-tuning in memristor conductance could be done afterwards or during applications to adapt to specific situations. Therefore, in neural network applications, memristors require high-precision programmability to guarantee uniform and accurate performance across a large number of memristive networks22–28. This requires many distinguishable conductance levels on each memristive device, not only laboratory-made devices but also devices fabricated in factories. Analog memristors with many conductance states also benefit other applications, such as neural network training, scientific computing and even ‘mortal computing’25,29,30. Here we report 2,048 conductance levels achieved with memristors in fully integrated chips with 256 × 256 memristor arrays monolithically integrated on complementary metal–oxide–semiconductor (CMOS) circuits in a commercial foundry. We have identified the underlying physics that previously limited the number of conductance levels that could be achieved in memristors and developed electrical operation protocols to avoid such limitations. These results provide insights into the fundamental understanding of the microscopic picture of memristive switching as well as approaches to enable high-precision memristors for various applications.[Figure not available: see fulltext.].
Persistent Identifierhttp://hdl.handle.net/10722/335448
ISSN
2023 Impact Factor: 50.5
2023 SCImago Journal Rankings: 18.509
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorRao, Mingyi-
dc.contributor.authorTang, Hao-
dc.contributor.authorWu, Jiangbin-
dc.contributor.authorSong, Wenhao-
dc.contributor.authorZhang, Max-
dc.contributor.authorYin, Wenbo-
dc.contributor.authorZhuo, Ye-
dc.contributor.authorKiani, Fatemeh-
dc.contributor.authorChen, Benjamin-
dc.contributor.authorJiang, Xiangqi-
dc.contributor.authorLiu, Hefei-
dc.contributor.authorChen, Hung Yu-
dc.contributor.authorMidya, Rivu-
dc.contributor.authorYe, Fan-
dc.contributor.authorJiang, Hao-
dc.contributor.authorWang, Zhongrui-
dc.contributor.authorWu, Mingche-
dc.contributor.authorHu, Miao-
dc.contributor.authorWang, Han-
dc.contributor.authorXia, Qiangfei-
dc.contributor.authorGe, Ning-
dc.contributor.authorLi, Ju-
dc.contributor.authorYang, J. Joshua-
dc.date.accessioned2023-11-17T08:25:58Z-
dc.date.available2023-11-17T08:25:58Z-
dc.date.issued2023-
dc.identifier.citationNature, 2023, v. 615, n. 7954, p. 823-829-
dc.identifier.issn0028-0836-
dc.identifier.urihttp://hdl.handle.net/10722/335448-
dc.description.abstractNeural networks based on memristive devices1–3 have the ability to improve throughput and energy efficiency for machine learning4,5 and artificial intelligence6, especially in edge applications7–21. Because training a neural network model from scratch is costly in terms of hardware resources, time and energy, it is impractical to do it individually on billions of memristive neural networks distributed at the edge. A practical approach would be to download the synaptic weights obtained from the cloud training and program them directly into memristors for the commercialization of edge applications. Some post-tuning in memristor conductance could be done afterwards or during applications to adapt to specific situations. Therefore, in neural network applications, memristors require high-precision programmability to guarantee uniform and accurate performance across a large number of memristive networks22–28. This requires many distinguishable conductance levels on each memristive device, not only laboratory-made devices but also devices fabricated in factories. Analog memristors with many conductance states also benefit other applications, such as neural network training, scientific computing and even ‘mortal computing’25,29,30. Here we report 2,048 conductance levels achieved with memristors in fully integrated chips with 256 × 256 memristor arrays monolithically integrated on complementary metal–oxide–semiconductor (CMOS) circuits in a commercial foundry. We have identified the underlying physics that previously limited the number of conductance levels that could be achieved in memristors and developed electrical operation protocols to avoid such limitations. These results provide insights into the fundamental understanding of the microscopic picture of memristive switching as well as approaches to enable high-precision memristors for various applications.[Figure not available: see fulltext.].-
dc.languageeng-
dc.relation.ispartofNature-
dc.titleThousands of conductance levels in memristors integrated on CMOS-
dc.typeArticle-
dc.description.naturelink_to_subscribed_fulltext-
dc.identifier.doi10.1038/s41586-023-05759-5-
dc.identifier.pmid36991190-
dc.identifier.scopuseid_2-s2.0-85151197734-
dc.identifier.volume615-
dc.identifier.issue7954-
dc.identifier.spage823-
dc.identifier.epage829-
dc.identifier.eissn1476-4687-
dc.identifier.isiWOS:000961772700001-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats