An Optoelectronic MXene/HfO2 Memristor for In-Sensor Neuromorphic Visual Computing

Abstract

The increasing demands for artificial intelligence (AI) have exposed the limitations of conventional von Neumann architectures, motivating the development of brain-inspired neuromorphic computing systems. Here, a photonic synaptic device array based on an Ag/MXene/HfO2/FTO structure is demonstrated, exhibiting dual electrical and optical modulation as well as nonvolatile memory characteristics. The device intrinsically supports sensing, memory, and computation at the device level. At the computation and memory level, a Verilog A-based memristor model is developed to implement basic logical applications, including AND, OR, and 2-to-1 multiplexer operations. At the sensing level, the device emulates key synaptic plasticity behaviors under optical stimulation, including paired pulse facilitation (PPF), spike-number-dependent plasticity (SNDP), and spike-rate-dependent plasticity (SRDP). Leveraging the SRDP response of the memristor under optical pulse stimulation, a high-pass filtering behavior is demonstrated, which enables device-response-inspired image edge enhancement and preprocessing demonstration. Building upon this in-sensor preprocessing capability, a convolutional neural network (CNN) is employed for handwritten digit recognition, achieving an accuracy of 95.5%. These results demonstrate the potential of the Ag/MXene/HfO2/FTO memristive platform for multifunctional neuromorphic computing and vision applications.