Proceedings of MATSUS Spring 2025 Conference (MATSUSSpring25)
Publication date: 16th December 2024
Memristors have emerged as one of the most promising components in the fields of memory and computation. These memristors exhibit unique characteristics by getting both the memory and computing in the same device, storing the information as a modulated conductance. Hence, the benefits of this configuration position them as highly efficient artificially intelligent hardware that can mimic the functions of the human brain. These characteristics vary from nonvolatile binary switching for digital in-memory computing and spiking neural networks to volatile analog switching for brain-inspired computing and artificial neural networks.
In these devices, a state variable is typically defined, which imparts the memory capability of the device. Changes in this state variable, driven by variations in the applied excitation voltage, enable the system to transition between different conductance states, thereby generating memory states. While this behavior is well understood, the vast diversity of memristor types and their distinct electrical responses make it challenging to develop a unified model that can describe all these responses simply by adjusting modeling parameters.
Here, we describe the SET and RESET processes using a Conductance-Activated Quasi-Linear Memristor (CALM) model, along with its dynamic behavior driven by a single voltage-dependent relaxation time of the memory variable. By tuning the parameters of this unified model, we can represent a wide variety of memristor responses. Additionally, by modifying the relaxation time of the memory variable, we can simulate both volatile and non-volatile behaviors.
Our study focuses on how this model can characterize the responses of various halide perovskite memristors and their intersections with neuro-inspired computing, because, in recent years, halide perovskite memristors have become some of the most promising candidates for such applications due to their extensive range of properties and tunability.
This model is a potent tool for understanding and probing the dynamical response of memristors by indicating the relaxation properties that control observable responses and will serve as a crucial tool for investigating regular memristors and as starting point for more complicated memristive behaviors.
This work was funded by the European Research Council (ERC) via Horizon Europe Advanced Grant, grant agreement nº 101097688 (“PeroSpiker”)