Physics-based modeling of volatile resistive switching memory (RRAM) for crosspoint selector and neuromorphic computing
2018 IEEE International Electron Devices Meeting (IEDM)
Volatile resistive switching memory (RRAM) is raising strong interest as potential selector device in crosspoint memory and short-term synapse in neuromorphic computing. To enable the design and simulation of memory and computing circuits with volatile RRAM, compact models are essential. To fill this gap, we present here a novel physics-based analytical model for volatile RRAM based on a detailed study of the switching process by molecular dynamics (MD) and finite-difference method (FDM). The
... alytical model captures all essential phenomena of volatile RRAM, e.g., threshold/holding voltages, on-off ratio, and size-dependent retention. The model is validated by extensive comparison with data from Ag/SiOx RRAM. To support the circuit-level capability of the model, we show simulations of crosspoint arrays and neuromorphic time-correlated learning. I. INTRODUCTION Resistive switching memory (RRAM) devices based on metallic filaments made of Cu and Ag generally offer high on-off ratio (> 10 7 ) [1, 2], steep switching slope (< 4 mV/dec), and high endurance . Although they were originally proposed as non-volatile switching memories , it was soon evidenced that they tend to display volatile switching due to the spontaneous turn-off within a short retention time from few µs to few ms . Thanks to its high nonlinearity (> 10 7 ), volatile RRAM has been proposed as select device in crosspoint arrays of memories or sensors  , and as artificial synapse to mimic long/short-term memory phenomena in the brain  . However, a comprehensive understanding of the volatile switching mechanism is still missing . Also, accurate compact models are needed to support all potential applications in memory and neuromorphic computing. In this work, we introduce a new physics-based model for volatile RRAM. First, we study the switching process in volatile RRAM by molecular dynamics (MD) and finite-difference method (FDM) simulations of the spontaneous filament rupture. From this study, we derive a universal equation for size-dependent retention time. By combining the new retention model with a previous RRAM model for voltage-controlled filament growth/dissolution , we develop the first physics-based analytical model for volatile RRAM. The model accurately captures all device characteristics, such as DC and AC switching curves and retention. By including the spread of the activation energy, stochastic variations are also described. We finally show simulations of potential applications such as crosspoint selectors and neuromorphic synapses.