High Performance Switchable Dual-Mode Photoresponse Enabled by Defect States in ReS2/PdSe2 Heterojunctions

Two-dimensional (2D) semiconductors exhibit significant potential for next-generation optoelectronic and neuromorphic applications. In particular, intrinsic defect states in 2D materials not only modulate the doping level but also play key roles in the photoresponse time of optoelectronic devices. Depending on the role of the defect states, either high-speed photodetection or low-speed defect-mediated synaptic functions can be realized within individual device. However, the integration of these two functions into a single device remains a challenge. Here, we report a dual-mode optoelectronic device based on an n-ReS2/p-PdSe2 heterojunction, enabling bias-switchable operation between self-driven photodetection and synaptic functions. At zero bias, the strong built-in electric field enables self-driven broadband detection with fast response times of 22.3 μs/22.8 μs. Under an applied bias, intrinsic defects and interfacial traps at the 2D material/substrate interface act as charge-trapping centers to emulate synaptic behaviors. Furthermore, the device emulates the "perception-memory-processing" capabilities of an artificial visual system and exhibits polarization-sensitive neuromorphic responses for spatial-domain image processing. Combined with artificial neural networks, it achieves efficient image denoising and contrast enhancement, reducing training epochs for high-precision handwritten digit recognition by 70%. This work presents a universal strategy for integrating high-speed detection and synaptic functions in multifunctional optoelectronics, showing promise for next-generation artificial vision and neuromorphic computing.