Stochastic p-Bits Based on Spin-Orbit Torque Magnetic Tunnel Junctions

TL;DR

This paper compares different spin-orbit torque magnetic tunnel junctions for stochastic p-Bits, highlighting the superior tunability and robustness of Y-type SOT-MTJs, which produce high-quality random numbers suitable for computational applications.

Contribution

It provides a comparative analysis of SOT-MTJ switching schemes, demonstrating the advantages of Y-type SOT-MTJs for stochastic p-Bit implementation through experimental and theoretical methods.

Findings

Y-type SOT-MTJs show the gentlest dependence of switching probability on voltage.

Y-type SOT-MTJs exhibit superior tunability and robustness.

Generated random numbers pass NIST SP800-22 tests.

Abstract

Stochastic p-Bit devices play a pivotal role in solving NP-hard problems, neural network computing, and hardware accelerators for algorithms such as the simulated annealing. In this work, we focus on Stochastic p-Bits based on high-barrier magnetic tunnel junctions (HB-MTJs) with identical stack structure and cell geometry, but employing different spin-orbit torque (SOT) switching schemes. We conducted a comparative study of their switching probability as a function of pulse amplitude and width of the applied voltage. Through experimental and theoretical investigations, we have observed that the Y-type SOT-MTJs exhibit the gentlest dependence of the switching probability on the external voltage. This characteristic indicates superior tunability in randomness and enhanced robustness against external disturbances when Y-type SOT-MTJs are employed as stochastic p-Bits. Furthermore, the random numbers generated by these Y-type SOT-MTJs, following XOR pretreatment, have successfully passed the National Institute of Standards and Technology (NIST) SP800-22 test. This comprehensive study demonstrates the high performance and immense potential of Y-type SOT-MTJs for the implementation of stochastic p-Bits.