Quantum conductance states in Ni/Mo/MoO3/Ni memristors: implications for multi-level neuromorphic computing

Abstract This study examined conductance quantization (CQ) in Ni/Mo/MoO 3 /Ni memristors using an electrothermal model in COMSOL Multiphysics. Oxygen ions moving in MoO 3 cause the formation and rupture of conductive filaments, making them very thin and leading to quantum effects on the device conductance. As a result, the memristor showed ten stable states near integer and half-integer multiples of the quantum conductance unit ( G 0 = 2 e 2 / h ) when a DC voltage was applied at room temperature. An analysis of the CQ states under different compliance currents ( I CC ) and voltage ramp rates ( V RR ) showed important behaviors: the number of CQ states increased with I CC until it leveled off, and a higher V RR resulted in a steady drop in the CQ states. These results provide important information for improving the memristor performance under different electrical conditions, aiding their use in neuromorphic computing systems. The CQ states in the Ni/Mo/MoO 3 /Ni memristors can potentially reduce the hardware requirements of neuromorphic systems by up to 62.5%. The simulation results also highlight the potential of using quantum conductance states for efficient neuromorphic computing, with the proposed memristor achieving approximately 97.56% accuracy in recognizing MNIST digits using a feedforward neural network. This study provides a promising direction for the development of advanced, multi-level neuromorphic systems.