Monte Carlo Simulation of a Three-Terminal RRAM with Applications to Neuromorphic Computing

We developed a Monte Carlo simulator to compute the state-dependent I-V characteristics of three-terminal (3T) RRAM devices. State switching in these devices is modeled using a combination of vacancy migration and trap-assisted-tunneling mechanisms. We describe key elements of the simulator, compute hysteresis curves under typical voltage cycling conditions, and demonstrate agreement with experimental results. We then study the response of 2T- and 3T-RRAMs under pulsed operation and show that 3T-RRAM conductance values have both greater dynamic range than 2T-RRAMs and the potential to deliver superior inference accuracy in neuromorphic applications.