Probabilistic computing with nanoscale voltage-controlled magnetic tunnel junctions

Probabilistic (p-) computing is a physics-based approach to addressing hard computational problems which cannot be efficiently solved using deterministic algorithms running on von Neumann computers. A key requirement for p-computing is the realization of fast, compact, and energy-efficient probabilistic bits. Stochastic magnetic tunnel junctions (S-MTJs), which are designed to have a small energy barrier between their two states [1], have been used widely to implement p-bits. This approach presents scaling challenges due to the need for precise control of a small energy barrier across large numbers of S-MTJs. Here we demonstrate an alternative p-bit design based on perpendicular voltage-controlled magnetic tunnel junctions (V-MTJs) that use optimized material stacks with a large voltage-controlled magnetic anisotropy (VCMA) effect [2] to create the random state of a p-bit on demand [3]. The V-MTJs are stable (i.e., have large energy barriers) in the absence of voltage, and VCMA-induced dynamics are used to generate random numbers in less than 10 ns/bit [4]. We then show a compact method of implementing p-bits by using V-MTJs without a bias current. As a demonstration of the functionality of the proposed p-bits in solving a representative hard optimization problem, we solve up to 40-bit integer factorization problems using experimental bit-streams generated by nanoscale V-MTJs [4]. We also discuss perspectives for integration of CMOS and V-MTJs for probabilistic computing. Finally, we discuss perspectives on the use of emerging all-antiferromagnetic tunnel junctions [5] in probabilistic computing applications.



References:

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

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

[3] Y. Shao et al., Advanced Electronic Materials 9, 2300195 (2023)

[4] Y. Shao et al., Nanotechnology 34, 495203 (2023)

[5] J. Shi et al., Advanced Materials 36, 2312008 (2024)