HfAlO_x-based optical ferroelectric memristor with transparent electrode for RGB color image classification via physical reservoir

The swift advancement of artificial intelligence is driving the increasing complexity of neuromorphic computing, presenting new challenges for conventional hardware. Significant progress has been achieved in advancing neuromorphic hardware through various memory devices. This study presents the development and characterization of an optical ferroelectric memristor (OFM) device for reservoir computing (RC) for more efficient data processing. We explore the electrical properties of OFM device using indium tin oxide (ITO) as the transparent top electrode and HfAlO_x (HAO) as the ferroelectric layer. The maximum remnant polarization (2 P_r) and tunneling electroresistance (TER) are achieved by the positive-up-negative-down (PUND) methods for synaptic memory operation. The synaptic and spike characteristics of the device was conducted by examining paired pulse facilitation (PPF) and its recognition capabilities using reservoir computing technology making it a promising candidate for artificial neural network applications. The device's optical response, influenced by light-induced oxygen vacancy ionization, enabled short-term plasticity and synaptic weight modulation under light stimulation. Simulations of optical reservoir computing (ORC) using the Fruits-360 dataset highlight its capability to efficiently process both RGB and grayscale inputs. The classification accuracy for RGB inputs outperform grayscale inputs by approximately 10 % for datasets with distinct color characteristics, underscoring the advantage of color information in complicated neuromorphic tasks. These findings demonstrate the potential of the ITO/HAO/n⁺ Si device for energy efficient and flexible neuromorphic platform.