The growth of data-intensive computing tasks requires processing units with higher performance and energy efficiency, but these requirements are increasingly difficult to achieve with conventional semiconductor technology. One potential solution is to combine developments in devices with innovations in system architecture. Here we report a tensor processing unit (TPU) that is based on 3,000 carbon nanotube field-effect transistors and can perform energy-efficient convolution operations and matrix multiplication. The TPU is constructed with a systolic array architecture that allows parallel 2 bit integer multiply–accumulate operations. A five-layer convolutional neural network based on the TPU can perform MNIST image recognition with an accuracy of up to 88% for a power consumption of 295 µW. We use an optimized nanotube fabrication process that offers a semiconductor purity of 99.9999% and ultraclean surfaces, leading to transistors with high on-current densities and uniformity. Using system-level simulations, we estimate that an 8 bit TPU made with nanotube transistors at a 180 nm technology node could reach a main frequency of 850 MHz and an energy efficiency of 1 tera-operations per second per watt. Carbon nanotube networks made with high purity and ultraclean interfaces can be used to make a tensor processing unit that contains 3,000 transistors in a systolic array architecture to improve energy efficiency in accelerating neural network tasks.
{"title":"A carbon-nanotube-based tensor processing unit","authors":"Jia Si, Panpan Zhang, Chenyi Zhao, Dongyi Lin, Lin Xu, Haitao Xu, Lijun Liu, Jianhua Jiang, Lian-Mao Peng, Zhiyong Zhang","doi":"10.1038/s41928-024-01211-2","DOIUrl":"10.1038/s41928-024-01211-2","url":null,"abstract":"The growth of data-intensive computing tasks requires processing units with higher performance and energy efficiency, but these requirements are increasingly difficult to achieve with conventional semiconductor technology. One potential solution is to combine developments in devices with innovations in system architecture. Here we report a tensor processing unit (TPU) that is based on 3,000 carbon nanotube field-effect transistors and can perform energy-efficient convolution operations and matrix multiplication. The TPU is constructed with a systolic array architecture that allows parallel 2 bit integer multiply–accumulate operations. A five-layer convolutional neural network based on the TPU can perform MNIST image recognition with an accuracy of up to 88% for a power consumption of 295 µW. We use an optimized nanotube fabrication process that offers a semiconductor purity of 99.9999% and ultraclean surfaces, leading to transistors with high on-current densities and uniformity. Using system-level simulations, we estimate that an 8 bit TPU made with nanotube transistors at a 180 nm technology node could reach a main frequency of 850 MHz and an energy efficiency of 1 tera-operations per second per watt. Carbon nanotube networks made with high purity and ultraclean interfaces can be used to make a tensor processing unit that contains 3,000 transistors in a systolic array architecture to improve energy efficiency in accelerating neural network tasks.","PeriodicalId":19064,"journal":{"name":"Nature Electronics","volume":"7 8","pages":"684-693"},"PeriodicalIF":33.7,"publicationDate":"2024-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141737000","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-17DOI: 10.1038/s41928-024-01198-w
Vladimir Bruevich, Vitaly Podzorov
Hall effect measurements are important in determining the electronic properties of emerging semiconductor materials, but care must be taken in their use and analysis.
霍尔效应测量对于确定新兴半导体材料的电子特性非常重要，但在使用和分析时必须小心谨慎。
{"title":"Reporting Hall effect measurements of charge carrier mobility in emerging materials","authors":"Vladimir Bruevich, Vitaly Podzorov","doi":"10.1038/s41928-024-01198-w","DOIUrl":"10.1038/s41928-024-01198-w","url":null,"abstract":"Hall effect measurements are important in determining the electronic properties of emerging semiconductor materials, but care must be taken in their use and analysis.","PeriodicalId":19064,"journal":{"name":"Nature Electronics","volume":"7 7","pages":"510-512"},"PeriodicalIF":33.7,"publicationDate":"2024-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141725845","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-07-17DOI: 10.1038/s41928-024-01201-4
Graphene plasmon polaritons are expected to enable rapid data transfer and processing; however, these plasmons are difficult to access. Terahertz electronics now facilitate the efficient generation, manipulation and on-chip detection of wave packets lasting as little as 1.2 ps. This advance could lead to the development of nanoscale terahertz circuits.
{"title":"Terahertz electronics generate and detect graphene plasmon polaritons","authors":"","doi":"10.1038/s41928-024-01201-4","DOIUrl":"10.1038/s41928-024-01201-4","url":null,"abstract":"Graphene plasmon polaritons are expected to enable rapid data transfer and processing; however, these plasmons are difficult to access. Terahertz electronics now facilitate the efficient generation, manipulation and on-chip detection of wave packets lasting as little as 1.2 ps. This advance could lead to the development of nanoscale terahertz circuits.","PeriodicalId":19064,"journal":{"name":"Nature Electronics","volume":"7 7","pages":"523-524"},"PeriodicalIF":33.7,"publicationDate":"2024-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141725843","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Ultrashort polariton wave packets, such as terahertz graphene plasmon polaritons, could be used for fast information processing in integrated circuits. However, conventional optical techniques have struggled to integrate the components for controlling polariton signals and have a low conversion efficiency. Here, we show that graphene plasmon wave packets can be generated, manipulated and read out on-chip using terahertz electronics. Electrical pulses injected into a graphene microribbon through an ohmic contact can be efficiently converted into a plasmon wave packet with a pulse duration as short as 1.2 ps and a three-dimensional spatial confinement of 2.1 × 10−18 m3. The conversion efficiency between the electrical pulses and plasmon wave packets can also reach 35% due to the absence of a momentum mismatch. The transport properties of graphene plasmons are studied by changing the dielectric environments, which provides a basis for designing graphene plasmonic circuits. Terahertz electronics that can create and control ultrashort graphene plasmon wave packets with durations as short as 1.2 ps can offer on-chip handling of plasmonic signals.
{"title":"On-chip transfer of ultrashort graphene plasmon wave packets using terahertz electronics","authors":"Katsumasa Yoshioka, Guillaume Bernard, Taro Wakamura, Masayuki Hashisaka, Ken-ichi Sasaki, Satoshi Sasaki, Kenji Watanabe, Takashi Taniguchi, Norio Kumada","doi":"10.1038/s41928-024-01197-x","DOIUrl":"10.1038/s41928-024-01197-x","url":null,"abstract":"Ultrashort polariton wave packets, such as terahertz graphene plasmon polaritons, could be used for fast information processing in integrated circuits. However, conventional optical techniques have struggled to integrate the components for controlling polariton signals and have a low conversion efficiency. Here, we show that graphene plasmon wave packets can be generated, manipulated and read out on-chip using terahertz electronics. Electrical pulses injected into a graphene microribbon through an ohmic contact can be efficiently converted into a plasmon wave packet with a pulse duration as short as 1.2 ps and a three-dimensional spatial confinement of 2.1 × 10−18 m3. The conversion efficiency between the electrical pulses and plasmon wave packets can also reach 35% due to the absence of a momentum mismatch. The transport properties of graphene plasmons are studied by changing the dielectric environments, which provides a basis for designing graphene plasmonic circuits. Terahertz electronics that can create and control ultrashort graphene plasmon wave packets with durations as short as 1.2 ps can offer on-chip handling of plasmonic signals.","PeriodicalId":19064,"journal":{"name":"Nature Electronics","volume":"7 7","pages":"537-544"},"PeriodicalIF":33.7,"publicationDate":"2024-07-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"141726018","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：