Probabilistic spin-torque switching of perpendicular magnetic tunnel junctions under short current pulses

True random number generators (TRNGs) are fundamental building blocks for various applications ranging from cryptography to stochastic sampling. Magnetic tunnel junctions (MTJs) utilizing low-barrier magnets (LBMs) are typically used to implement magnet-based TRNGs. LBMs have energy barriers of Δ ≤ 10 kT and therefore require low energy to operate. They, however, suffer from slow dynamics, extreme sensitivity to process variations, and are hard to build in practice. Recently, stochastic magnetic actuated random transducer (SMART) devices based on perpendicular MTJs (pMTJs) utilizing medium-barrier magnets (MBMs) have emerged as potential candidates to circumvent these drawbacks [1]. We present a systematic analysis of spin-torque-driven switching of the MBM-based pMTJs (Δ ~ 20 – 40 kT) under short current pulses by numerically solving their dynamics using a 1-D Fokker–Planck equation. We find that our device requires low switching energy (< 100 fJ) and operates at a fast switching speed (< 1 ns). We also analyze the impact of process variations, viz. temperature and material parameters, on the switching probability of our device and show its robustness to them. Our results agree well with the analytical equation, stochastic Landau–Lifshitz–Gilbert equation, and recent experiments and show a path toward creating a fast and energy-efficient TRNG hardware unit for solving optimization problems.

[1] L. Rehm et al., arXiv.2209.01480.