Engineering stochastic magnetic tunnel junction for probabilistic computing
ORAL · Invited
Abstract
Probabilistic computing has attracted significant attention due to its ability to solve nondeterministic polynomial time (NP) challenges effectively. A recent development involves the utilization of stochastic magnetic tunnel junctions (s-MTJs) characterized by a low barrier energy [1,2]: by applying s-MTJs for spintronics-based probabilistic computing, integer factorization [2], machine learning [3], and quantum simulation [4] have been demonstrated with lower energy consumption than the probabilistic computing fully composed of CMOS-based hardware.
Here, I will talk about our recent works on the physics of the s-MTJs engineered for stochastic computing: first, I will discuss the fundamental s-MTJ characteristics defined by time-domain response, which determines the calculation speed [5,6,7], and time-averaged response, which determines the input-output response and precision [8,9]. Next, I experimentally demonstrate external-magnetic-field-robust s-MTJs [10] and input-voltage-robust s-MTJs [11], which are device designs essential for withstanding input disturbances and/or deliberate attacks.
[1] K. Y. Camsari et al., Phys. Rev. X 7, 031014 (2017).
[2] W. A. Borders et al., Nature 573, 390-393 (2019).
[3] J. Kaiser et al., Phys. Rev. Appl. 17, 014016 (2022).
[4] A. Grimaldi et al., 68th Annual IEEE International Electron Devices Meeting (2022).
[5] S. Kanai et al., Phys. Rev. B 103, 094423 (2021).
[6] K. Hayakawa et al., Phys. Rev. Lett. 126, 117202 (2021).
[7] S. Kanai et al., 68th Annual Conference on Magnetism and Magnetic Materials (2023).
[8] K. Kobayashi et al., Appl. Phys. Lett. 119, 132406 (2021).
[9] T. Funatsu et al., Nat. Commun. 13, 4079 (2022).
[10] K. Kobayashi et al., Phys. Rev. Appl. 18, 054085 (2022).
[11] R. Ota et al., Appl. Phys. Lett. 125, 022406 (2024).
Here, I will talk about our recent works on the physics of the s-MTJs engineered for stochastic computing: first, I will discuss the fundamental s-MTJ characteristics defined by time-domain response, which determines the calculation speed [5,6,7], and time-averaged response, which determines the input-output response and precision [8,9]. Next, I experimentally demonstrate external-magnetic-field-robust s-MTJs [10] and input-voltage-robust s-MTJs [11], which are device designs essential for withstanding input disturbances and/or deliberate attacks.
[1] K. Y. Camsari et al., Phys. Rev. X 7, 031014 (2017).
[2] W. A. Borders et al., Nature 573, 390-393 (2019).
[3] J. Kaiser et al., Phys. Rev. Appl. 17, 014016 (2022).
[4] A. Grimaldi et al., 68th Annual IEEE International Electron Devices Meeting (2022).
[5] S. Kanai et al., Phys. Rev. B 103, 094423 (2021).
[6] K. Hayakawa et al., Phys. Rev. Lett. 126, 117202 (2021).
[7] S. Kanai et al., 68th Annual Conference on Magnetism and Magnetic Materials (2023).
[8] K. Kobayashi et al., Appl. Phys. Lett. 119, 132406 (2021).
[9] T. Funatsu et al., Nat. Commun. 13, 4079 (2022).
[10] K. Kobayashi et al., Phys. Rev. Appl. 18, 054085 (2022).
[11] R. Ota et al., Appl. Phys. Lett. 125, 022406 (2024).
–
Presenters
-
Shun Kanai
Tohoku University, Japan
Authors
-
Shun Kanai
Tohoku University, Japan