FET: Small: Collaborative Research: Integrated Spintronic Synapses and Neurons for Neuromorphic Computing Circuits - I(SNC)^2

FET：小型：协作研究：用于神经形态计算电路的集成自旋电子突触和神经元 - I(SNC)^2

• 批准号: 1910800
• 负责人: Joseph Friedman
• 金额: $ 19.11万
• 依托单位: University of Texas at Dallas
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-06-15 至 2023-05-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1910800&HistoricalAwards=false
• 关键词: FET; Small; Collaborative; Research; Integrated

There are many pressing problems today where data-intensive tasks are needed to be accomplished in real time. This can range from sequencing DNA, to self-driving cars recognizing a person walking by, to predicting the trajectory of a flying object. In these examples, traditional computing faces a performance wall where the computing time and energy is severely limited by memory access. If computers could be built closer to the way the brain computes, where memory and computation are densely connected together like the neurons (and synapses) of the brain, these tasks could be performed with a million times less energy. This requires doing research on designing and building artificial neurons and synapses, and research on connecting them together into neuromorphic circuits. Due to the many different kinds of problems this new type of computing will address, research in this area will have impact not only in the semiconductor industry, but also far-reaching impact in medicine, defense, and new technologies. This project will educate and train multiple Ph.D.-level and undergraduate students in this interdisciplinary field, with skills highly sought after in academia, national labs, and industry. It will also have significance for broadening participation of women and under-represented minorities in computing: the researchers seek to educate and train women and Hispanic students from their state of Texas.Nanodevices made from magnetic materials (such as iron) have many properties that make them uniquely suitable as artificial neurons and synapses to enable such computing. Nevertheless, a number of technical problems remain in using magnetic devices for neuromorphic computing, which this project aims to address: there has been little experimental study of circuits that combine spintronic neurons and synapses, the devices and circuits designed so far do not capture all the desired biological behaviors, and there have been no circuits designed that operate without external silicon-based devices. This interdisciplinary collaborative effort between experiment and circuit design will address these challenges by building and studying circuits using three-terminal magnetic tunnel junction devices. The research will result in design and fabrication of new types of these magnetic devices that more accurately represent the brain's functions, and in measurements of the magnetic devices' behavior in circuits. The project has the potential to establish magnetic materials as a platform for neuromorphic computing, similar to how silicon is the platform material for traditional computing.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

如今有许多紧迫的问题需要实时完成数据密集型任务。其范围包括从 DNA 测序到自动驾驶汽车识别路过的人，再到预测飞行物体的轨迹。在这些示例中，传统计算面临性能墙，其中计算时间和能量受到内存访问的严重限制。如果计算机的构建方式更接近大脑的计算方式，即内存和计算像大脑的神经元（和突触）一样紧密连接在一起，那么执行这些任务所需的能量可以减少一百万倍。这需要研究设计和构建人工神经元和突触，以及研究将它们连接在一起形成神经形态电路。由于这种新型计算将解决许多不同类型的问题，该领域的研究不仅会对半导体行业产生影响，而且还会对医学、国防和新技术产生深远的影响。该项目将在这一跨学科领域教育和培训多名博士级和本科生，他们的技能在学术界、国家实验室和工业界备受追捧。这对于扩大女性和代表性不足的少数族裔在计算机领域的参与也具有重要意义：研究人员寻求教育和培训来自德克萨斯州的女性和西班牙裔学生。由磁性材料（例如铁）制成的纳米设备具有许多特性，使得它们非常适合作为人工神经元和突触来实现此类计算。然而，使用磁性设备进行神经拟态计算仍然存在许多技术问题，该项目旨在解决这些问题：对结合自旋电子神经元和突触的电路的实验研究很少，迄今为止设计的设备和电路并没有捕获所有的信息。期望的生物行为，并且还没有设计出无需外部硅基设备即可运行的电路。实验和电路设计之间的跨学科合作将通过使用三端子磁性隧道结器件构建和研究电路来解决这些挑战。该研究将设计和制造新型磁性设备，更准确地代表大脑的功能，并测量磁性设备在电路中的行为。该项目有潜力将磁性材料建立为神经形态计算的平台，类似于硅作为传统计算的平台材料。该奖项反映了 NSF 的法定使命，并通过使用基金会的智力优势和更广泛的评估进行评估，认为值得支持。影响审查标准。

项目成果

Shape-Based Magnetic Domain Wall Drift for an Artificial Spintronic Leaky Integrate-and-Fire Neuron

人工自旋电子漏积分激发神经元的基于形状的磁畴壁漂移

• DOI: 10.1109/ted.2019.2938952
• 发表时间: 2019-05-14
• 期刊: IEEE Transactions on Electron Devices
• 影响因子: 3.1
• 作者: Wesley H. Brigner;J. Friedman;Naimul Hassan;L. Jiang;Xuan Hu;Diptish Saha;C. Bennett;M. Marinella;J. Incorvia;F. García
• 通讯作者: F. García

Lateral inhibition in magnetic domain wall racetrack arrays for neuromorphic computing

用于神经形态计算的磁畴壁跑道阵列的横向抑制

• DOI: 10.1117/12.2568870
• 发表时间: 2020-08
• 期刊: Spintronics XIII
• 作者: Cui, Can;Akinola, Otitoaleke G.;Hassan, Naimul;Bennett, Christopher H.;Marinella, Matthew J.;Friedman, Joseph S.;Incorvia, Jean Anne
• 通讯作者: Incorvia, Jean Anne

High-speed CMOS-free purely spintronic asynchronous recurrent neural network

高速无 CMOS 纯自旋电子异步循环神经网络

• DOI: 10.1063/5.0129006
• 发表时间: 2023-03
• 期刊: APL Machine Learning
• 作者: Mathews, Pranav O.;Duffee, Christian B.;Thayil, Abel;Stovall, Ty E.;Bennett, Christopher H.;Garcia;Marinella, Matthew J.;Incorvia, Jean Anne;Hassan, Naimul;Hu, Xuan;et al
• 通讯作者: et al

Process Variation Model and Analysis for Domain Wall-Magnetic Tunnel Junction Logic

畴壁磁隧道结逻辑的工艺变化模型与分析

• DOI: 10.1109/iscas45731.2020.9180675
• 发表时间: 2020-10-01
• 期刊: 2020 IEEE International Symposium on Circuits and Systems (ISCAS)
• 作者: Xuan Hu;Ale;er J. Edwards;er;T. Xiao;C. Bennett;J. Incorvia;M. Marinella;J. Friedman
• 通讯作者: J. Friedman

Shape‐Dependent Multi‐Weight Magnetic Artificial Synapses for Neuromorphic Computing

用于神经形态计算的形状依赖性多重量磁性人工突触

• DOI: 10.1002/aelm.202200563
• 发表时间: 2022-12-01
• 期刊: Advanced Electronic Materials
• 影响因子: 6.2
• 作者: Thomas Leonard;Samuel Liu;Mahshid Alamdar;Harrison Jin;Can Cui;Otitoaleke G. Akinola;Lin Xue;T. Xiao;J. Friedman;M. Marinella;C. Bennett;J. Incorvia
• 通讯作者: J. Incorvia