Reverse Engineering in Memristors for Neuromorphic Applications

In this study, we employ ML-driven numerical modeling to elucidate intrinsic behavior of switching mechanisms of candidate layered in memristors. Being natively ill-posed mathematical problem, we resolve the inherent internal structures of the memristors by employing simulated pulses through a variety of layer designs of the memristors containing various types of oxides with a range of defect concentrations. By utilizing a relatively simple but discretized circuit design to represent these layers, we were able to reverse engineer the detailed structures of these circuits through overall pulse signals. This approach can be potentially device-agnostic and modular enough to be used in many types of memristor layered structures.