Spin Hall MTJ Devices for Advanced Neuromorphic Functions

Andrew W. Stephan; Steven J. Koester

doi:10.1109/TED.2019.2959732

Spin Hall MTJ Devices for Advanced Neuromorphic Functions

Andrew W. Stephan, Steven J. Koester

Electrical and Computer Engineering

Research output: Contribution to journal › Article › peer-review

6 Scopus citations

Abstract

We propose a new device based on standard spin Hall magnetic tunnel junction-based spintronic neurons, which allows them to compute multiple neural network functionalities simultaneously and in parallel, saving space and time. An approximation to the rectified linear unit transfer function and the local pooling function is computed simultaneously with the convolution operation itself. A proof-of-concept simulation thoroughly explores the behavior of the device, predicting that the operations can be performed with up to 99% precision at a cost of about 6 pJ per 3×3 pooling template. The simulations are remarkably robust to thermal noise, performing well even with very small magnetic layers.

Original language	English (US)
Article number	8957360
Pages (from-to)	487-492
Number of pages	6
Journal	IEEE Transactions on Electron Devices
Volume	67
Issue number	2
DOIs	https://doi.org/10.1109/TED.2019.2959732
State	Published - Feb 2020

Bibliographical note

Funding Information:
Manuscript received September 4, 2019; revised October 30, 2019; accepted December 10, 2019. Date of publication January 13, 2020; date of current version January 27, 2020. This work was supported by Seagate Technology PLC. The review of this article was arranged by Editor T.-H. Kim. (Corresponding authors: Andrew W. Stephan; Steven J. Koester.) The authors are with the Department of Electrical and Computer Engineering, University of Minnesota, Minneapolis, MN 55455 USA (e-mail: steph506@umn.edu; skoester@umn.edu).

Publisher Copyright:
© 1963-2012 IEEE.

Keywords

CMOS
magnetic tunnel junction (MTJ)
neuromorphic computing
spin Hall
spintronics

Access

10.1109/TED.2019.2959732

OpenUrl availability

Full text

Cite this

@article{d161848d2f2a4ddf86f354e60f0fe20f,

title = "Spin Hall MTJ Devices for Advanced Neuromorphic Functions",

abstract = "We propose a new device based on standard spin Hall magnetic tunnel junction-based spintronic neurons, which allows them to compute multiple neural network functionalities simultaneously and in parallel, saving space and time. An approximation to the rectified linear unit transfer function and the local pooling function is computed simultaneously with the convolution operation itself. A proof-of-concept simulation thoroughly explores the behavior of the device, predicting that the operations can be performed with up to 99% precision at a cost of about 6 pJ per 3×3 pooling template. The simulations are remarkably robust to thermal noise, performing well even with very small magnetic layers.",

keywords = "CMOS, magnetic tunnel junction (MTJ), neuromorphic computing, spin Hall, spintronics",

author = "Stephan, {Andrew W.} and Koester, {Steven J.}",

note = "Funding Information: Manuscript received September 4, 2019; revised October 30, 2019; accepted December 10, 2019. Date of publication January 13, 2020; date of current version January 27, 2020. This work was supported by Seagate Technology PLC. The review of this article was arranged by Editor T.-H. Kim. (Corresponding authors: Andrew W. Stephan; Steven J. Koester.) The authors are with the Department of Electrical and Computer Engineering, University of Minnesota, Minneapolis, MN 55455 USA (e-mail: steph506@umn.edu; skoester@umn.edu). Publisher Copyright: {\textcopyright} 1963-2012 IEEE.",

year = "2020",

month = feb,

doi = "10.1109/TED.2019.2959732",

language = "English (US)",

volume = "67",

pages = "487--492",

journal = "IEEE Transactions on Electron Devices",

issn = "0018-9383",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

number = "2",

}

TY - JOUR

T1 - Spin Hall MTJ Devices for Advanced Neuromorphic Functions

AU - Stephan, Andrew W.

AU - Koester, Steven J.

N1 - Funding Information: Manuscript received September 4, 2019; revised October 30, 2019; accepted December 10, 2019. Date of publication January 13, 2020; date of current version January 27, 2020. This work was supported by Seagate Technology PLC. The review of this article was arranged by Editor T.-H. Kim. (Corresponding authors: Andrew W. Stephan; Steven J. Koester.) The authors are with the Department of Electrical and Computer Engineering, University of Minnesota, Minneapolis, MN 55455 USA (e-mail: steph506@umn.edu; skoester@umn.edu). Publisher Copyright: © 1963-2012 IEEE.

PY - 2020/2

Y1 - 2020/2

N2 - We propose a new device based on standard spin Hall magnetic tunnel junction-based spintronic neurons, which allows them to compute multiple neural network functionalities simultaneously and in parallel, saving space and time. An approximation to the rectified linear unit transfer function and the local pooling function is computed simultaneously with the convolution operation itself. A proof-of-concept simulation thoroughly explores the behavior of the device, predicting that the operations can be performed with up to 99% precision at a cost of about 6 pJ per 3×3 pooling template. The simulations are remarkably robust to thermal noise, performing well even with very small magnetic layers.

AB - We propose a new device based on standard spin Hall magnetic tunnel junction-based spintronic neurons, which allows them to compute multiple neural network functionalities simultaneously and in parallel, saving space and time. An approximation to the rectified linear unit transfer function and the local pooling function is computed simultaneously with the convolution operation itself. A proof-of-concept simulation thoroughly explores the behavior of the device, predicting that the operations can be performed with up to 99% precision at a cost of about 6 pJ per 3×3 pooling template. The simulations are remarkably robust to thermal noise, performing well even with very small magnetic layers.

KW - CMOS

KW - magnetic tunnel junction (MTJ)

KW - neuromorphic computing

KW - spin Hall

KW - spintronics

UR - http://www.scopus.com/inward/record.url?scp=85078893055&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85078893055&partnerID=8YFLogxK

U2 - 10.1109/TED.2019.2959732

DO - 10.1109/TED.2019.2959732

M3 - Article

AN - SCOPUS:85078893055

SN - 0018-9383

VL - 67

SP - 487

EP - 492

JO - IEEE Transactions on Electron Devices

JF - IEEE Transactions on Electron Devices

IS - 2

M1 - 8957360

ER -

Spin Hall MTJ Devices for Advanced Neuromorphic Functions

Abstract

Bibliographical note

Keywords

Access

OpenUrl availability

Other files and links

Fingerprint

Cite this