摘要
The superparamagnetic tunnel junction (SP-MTJ) represents an attractive solution for stochastic computing applications, providing substantial advantages in terms of area efficiency over traditional complementary metal oxide semiconductor (CMOS) technologies. This study proposes an effective technique to enable SP-MTJ functionality by incorporating a molybdenum (Mo) insertion layer above the free layer using the standard spin-transfer-torque magnetoresistive random-access memory (STT-MRAM) structure. Through precise modulation of the Mo layer thickness, the perpendicular magnetic anisotropy (PMA) can be systematically controlled, resulting in intrinsic stochastic switching behavior. Consequently, the engineered SP-MTJ functions as an efficient, compact probabilistic bits (p-bits), demonstrating intrinsic random oscillatory behavior for implementing stochastic discrete Hopfield neural network (DHNN). Simulation results demonstrate the potential and effectiveness in addressing challenging combinatorial optimization problems (COPs).