Energy Efficient Manipulation of Stochastic Switching Behavior in Dual-biased Magnetic Tunnel Junctions

Magnetic tunnel junctions (MTJs) have been identified as excellent candidates for generating stochastic signal, making them a promising solution for implementing stochastic computing (SC) due to their inherent randomness [1,2]. By leveraging the competition between the spin transfer torque and the external magnetic field (dual biasing method) [3], it becomes possible to adjust the average switching rate and average output level independently, thereby addressing the issue of device variation arising from the fabrication process. However, a significant increase in energy consumption is anticipated when attempting to simultaneously bias all the MTJs in the array network due to the limited current spin efficiency in each device. This counteracts the advantage of stochastic computing. In this report, the voltage-controlled exchange coupling (VCEC) switching mechanism [4] is utilized in place of the spin transfer torque as one of the two biases. It is observed that the tunability of stochastic switching is maintained, with the estimated energy consumption being around 2 orders of lower than the typical STT-MTJs in previous dual-biased MTJs study [3]. The results demonstrate the feasibility of VCEC in dual biasing method, taking a significant step toward the application of MTJs in stochastic computing.



[1] Yang Lv, et al, “A single magnetic-tunnel-junction stochastic computing unit”, IEDM, DOI: 10.1109/IEDM.2017.8268504 (2017)

[2] Brandon Zink, et al,. “Review of Magnetic Tunnel Junctions for Stochastic Computing.” IEEE Journal on Exploratory Solid-State Computational Devices and Circuits 8, no. 2 (December 2022): 173–84.

[3] Brandon Zink et al, “Influence of Intrinsic Thermal Stability on Switching Rate and Tunability of Dual-Biased Magnetic Tunnel Junctions for Probabilistic Bits.” IEEE Magnetics Letters 12 (2021): 1–5.

[4] D. Zhang, et al., Nano Lett. 22, 622–629 (2022), DOI: 10.1021/acs.nanolett.1c03395..