AlScN-based ferroelectric memristor for electrical synapse emulation and light-stimulated reservoir computing

Abstract

In this study, we present a multifunctional indium tin oxide (ITO)/aluminum scandium nitride (AlScN)/n+ Si ferroelectric memristor for integrated electrical–optical neuromorphic computing. The device, fabricated using radio frequency sputtering, exhibits robust ferroelectricity with an average remanent polarization of 48.46 μC/cm2 and stable endurance over 105 cycles. Electrical measurements confirm core synaptic behaviors, including potentiation and depression, with improved linearity and recognition accuracy using incremental pulse schemes. Spike-dependent plasticity modulated by pulse number, amplitude, and width is also demonstrated. In addition, the device exhibits a volatile photoresponse under 405 nm illumination conditions, enabling optically induced potentiation and depression depending on light intensity, mimicking short-term synaptic plasticity. Leveraging this dual electrical–optical modulation, we implemented a physical reservoir computing system using optically stimulated devices to process 4-bit encoded Modified National Institute of Standards and Technology inputs, achieving a classification accuracy of 96.35%. These results highlight the potential of the ITO/AlScN/n+ Si memristor as a compact, energy-efficient platform for next-generation optoelectronic neuromorphic systems.