Physics-based RRAM compact model for multilevel programming across multiple timescales

Abstract Resistive random access memories (RRAMs) are emerging as a key enabling technology for cost-effective, energy-efficient and secure chips, especially in the framework of edge computing. In particular, their electrically programmable resistance has been widely exploited in several in-memory computing and neuromorphic architectures. By adjusting the applied voltages and compliance currents ( I C ), RRAM devices can be programmed to multiple resistance states during set and reset procedures, enabling multilevel functionality. While the multilevel behavior of the reset phase is generally well captured by existing compact models, only a few account for the multilevel characteristics of the set operations. Moreover, such models are rarely validated against comprehensive experimental datasets capturing device dynamics across multiple timescales. In this work, we present a physics-based compact model that enhances the UniMORE RRAM framework by incorporating the dynamic lateral evolution of the conductive filament (CF), thereby enabling accurate simulation of set operations at varying I C values. The model is calibrated to experimental data from IHP 130 nm 1T1R RRAM technology and reproduces device behavior across several operating conditions using a single set of parameters. The results highlight the potential of the proposed compact model in design optimization workflows of RRAM-based circuits.