Advanced physical modeling of SiO<inf>x</inf> resistive random access memories

Abstract

Abstract-We apply a three-dimensional (3D) physical simulator, coupling self-consistently stochastic kinetic Monte Carlo descriptions of ion and electron transport, to investigate switching in silicon-rich silica (SiO x ) redox-based resistive random-access memory (RRAM) devices. We explain the intrinsic nature of resistance switching of the SiO x layer, and demonstrate the impact of self-heating effects and the initial vacancy distributions on switching. We also highlight the necessity of using 3D physical modelling to predict correctly the switching behavior. The simulation framework is useful for exploring the little-known physics of SiO x RRAMs and RRAM devices in general. This proves useful in achieving efficient device and circuit designs, in terms of performance, variability and reliability.

Physical Insights into the Transport Properties of RRAMs Based on Transition Metal Oxides

“…We employ an in-house 3D device simulator, which has been previously used for gaining insight into the operation of SiO x structures [3], [8], to study HfO x -based RRAMs, a widely used transition metal oxide (TMO) in memristor technology. Hafnia is highly suitable for high-density CMOS integration due to their high dielectric constants.…”

“…The dynamic nature of the vacancy formation and annihilation, and electron trapping is considered accurately, as discussed rigorously in Refs. [3], [8], [12]. The ion and vacancy time-dependent dynamics (drift, diffusion, generation and recombination) are also modeled carefully, as discussed in Refs.…”

“…The ion and vacancy time-dependent dynamics (drift, diffusion, generation and recombination) are also modeled carefully, as discussed in Refs. [3], [8]. The effect of all the dominant electron transport mechanisms are carefully considered, including trap-assisted tunneling, trap-to-trap tunneling, Fowler-Nordheim tunneling, Poole-Frenkel emission, and direct tunneling mechanisms [3].…”

“…1(b). As discussed extensively in literature, the memristive behaviour of oxide-based RRAMs is a direct consequence of the forming and destruction of conductive filaments, which are formed by direct electrical conductive paths between the cathode and the anode [3], [8], [13]. These filaments are created by the generation of oxygen ion-vacancy pairs, whose rates and transport are in general governed by the electric field (potential) and temperature distribution within the oxide.…”

“…In general, the elevated local temperatures, resulting mainly from Joule heating, affect significantly the device behavior, as they can boost the probability of vacancy generation and ion hopping, but also electron transport via trap-assisted tunneling and other relevant mechanisms [3], [8], [12], [15].…”