Higher‐Order Temporal Dynamics in Complementary Charge Trap Memristor for High‐Dimensional Reservoir Computing

Abstract Reservoir computing (RC) is a brain‐inspired computational model that utilizes recurrent networks to transform temporal inputs into high‐dimensional representations with fewer trainable parameters. Hardware implementations require a reservoir capable of encoding complex temporal dynamics. While memristors with short‐term plasticity are widely investigated, most approaches rely on analog outputs, limiting encoding resolution. This work introduces higher‐order dynamics in a complementary charge‐trap memristor (CoCTM), based on a Ti/Nb ₂ O _5−x /Al ₂ O ₃ /Nb ₂ O _5−y /Pt structure. The device exhibits unique output current transients, characterized by an overshoot followed by a gradual relaxation, enabled by the interplay between the complementary charge‐trapping layers. Switching and current transient mechanisms are proposed and supported by electrical measurements and equivalent circuit simulations, revealing that the transient behavior results from the dual capacitive and electronic detrapping role of the Nb ₂ O _5−x/y layers. Both rising and decay time constants are tunable through programming conditions, producing complex short‐ and long‐term hybrid dynamics. These higher‐order temporal responses are leveraged to expand the reservoir state space and enhance the encoding capability for high‐dimensional, energy‐efficient reservoir computing applications.