Torque Optimization for Voltage-Controlled Magnetic Tunnel Junctions as Memory and Stochastic Signal Generators

We propose an optimization principle for the operation of voltage-controlled magnetic tunnel junctions (VC-MTJs): optimization via magnetic torque, in contrast to the conventional optimization via magnetization. For memory applications, we minimize the magnetic torque at the end of the write pulse, whereas for stochastic applications, we attempt to find a balance between damping and precessional torque at the beginning of the pulse. We simulate the method using a physics-based VC-MTJ model and show that, for memory applications, optimization for torque reduces ringing by 43% and improves pulse shape tolerance by a factor of 2.1, whereas for stochastic applications, it improves convergence time by 61% and energy by 37%.