Comprehensive predictive modeling of resistive switching devices using a bias-dependent window function approach

Development of accurate models for resistive switching devices (memristors) is a research topic of utmost interest. Behavioral models usually employ window functions (WFs) to capture the dependency of the resistance switching-rate on the bias conditions. Several WFs have been published so far, all of them being functions of just the state variable(s), ignoring the effect of the applied signal magnitude in dynamic behavior. In this context, we describe in an extended manner a generalized concept of bias-dependent WFs, designed to enhance behavioral models in capturing rich dynamic time-response of memristors. We present a specific WF formulation and evaluate its effect on the performance of threshold-type models of voltage-controlled bipolar memristor, in simulations with LTSPICE. The obtained results not only reflect the accumulated effect of the applied signal and the proper saturation of the device at voltage-dependent levels, but are also quantitatively in line with experimental data taken from commercial self-directed channel (SDC) memristors of Knowm Inc.