Chicago Section

Zoom link: https://argonne.zoomgov.com/j/1616096375?pwd=QllQWStYbWk1dGZUWmdTNzhzY3NxUT09

Speaker: Professor Pedram Khalili

Department of Electrical and Computer Engineering
Northwestern University, Evanston, IL, USA 

Date: Friday, October 20, 2023
Time: 11:00 AM CDT

Abstract

The emergence of embedded magnetic random-access memory (MRAM) provides an unprecedented opportunity to develop unconventional computing architectures, which go far beyond using MRAM as a mere replacement for existing memory solutions (e.g., embedded Flash or SRAM).

This talk will consist of two parts: First, we review the current state of development of ferromagnet-based MRAM, which uses current-induced spin-transfer torque (STT) to switch the magnetic state. We then discuss how emerging device concepts based on new physics and new materials may enable significant advances beyond today’s STT-MRAM: (i) As an example of new materials, we examine memory devices based on antiferromagnetic (AFM) materials, which may offer advantages such as picosecond switching, improved scalability, and immunity to external magnetic fields. We review recent progress in manipulating the Néel vector of such materials by current-induced spin-orbit torque (SOT) [1-3] and discuss perspectives for their further development. (ii) As an example of new physics, we discuss electric-field-controlled MRAM devices that utilize the voltage-controlled magnetic anisotropy (VCMA) effect for switching, and present recent results on developing the first VCMA-MRAM devices with sub-1V write voltage [4, 5].

Second, we will discuss how appropriately designed stochastic MRAM cells with low retention time can be used to fulfill unconventional roles within a computing system [5, 6], notably as electrically controlled stochastic bitstream (SBS) generators. We then discuss the application of such MRAM-based SBS generators to true random number generation and stochastic computing (SC), and present our recent results on the implementation of an SC-based artificial neural networks [6] and physically unclonable functions [5] using a series of stochastic MRAM cells. Finally, we then show examples of how a network of stochastic MRAM bits with appropriately designed control/readout circuitry – referred to as probabilistic (p-) bits – can be used to solve difficult optimization problems.

[1] J. Shi et al., Nature Electronics 3, 92 (2020)

[2] S. Arpaci et al., Nature Communications 12, 3828 (2021)

[3] Z. Zheng et al., Nature Communications 12, 4555 (2021)

[4] Y. Shao et al., Communications Materials 3, 87 (2022)

[5] Y. Shao et al., Advanced Electronic Materials, 2300195 (2023)

[6] Y. Shao et al., IEEE Magnetics Letters 12, 4501005 (2021)

Biography:

Pedram Khalili is Associate Professor of Electrical and Computer Engineering at Northwestern University, where he is also affiliated with the Applied Physics program. Prior to joining Northwestern, he was an adjunct assistant professor at the University of California, Los Angeles. He leads research programs on voltage-controlled MRAM, antiferromagnetic spintronics, magnonics, and spintronics-based computing. He received the B.Sc. degree from Sharif University of Technology in 2004, and the Ph.D. degree (cum laude) from Delft University of Technology (TU Delft), The Netherlands, in 2008, both in electrical engineering. Pedram received the Northwestern University ECE department’s Best Teacher Award in 2020. He serves on the Editorial Board of Journal of Physics: Photonics. He has served on the technical program committees and organizing committees of several conferences, including the Joint MMM/Intermag Conference and the SPIE Spintronics Conference, and is a member of the Flash Memory Summit conference advisory board. He has served as Chair of the Chicago Chapter of the IEEE Magnetics Society (2020-2023), and represents the IEEE Magnetics Society on the IEEE Task Force for Rebooting Computing (TFRC) Executive Committee. He is a Senior Member of the IEEE.

9700 S Cass Ave

LEMONT, IL 60439