Atomic-Device Hybrid Modeling of Relaxation Effect in Analog RRAM for Neuromorphic Computing

Conductance relaxation effect in analog RRAM devices poses a great challenge in building neuromorphic system with high computing accuracy. Due to the broad spatial-temporal scale and the stochastic nature of the related physical processes, it is very difficult to clarify the mechanism of the relaxation effect. Here, we develop a new atomic-device hybrid modeling technique to investigate the physical mechanism of resistive switching and relaxation effect in amorphous-HfO
_x
based analog RRAM. The technique utilizes a synergistic multi-scale simulation framework consisting of molecular dynamics (MD), density functional theory (DFT), kinetic Monte Carlo (KMC), and finite difference method (FDM). With the proposed modeling technique, the key microscopic events that are responsible for the relaxation effect are identified. This work provides valuable design guidelines for improving the reliability of analog RRAM.