Title : Augmented quantum dot-enhanced IGZO-Te heterogeneous photodiode enabling synaptic in-sensor image processing
Abstract:
This study presents a quantum-dot–enhanced heterogeneous photodiode based on an IGZO-Telluirum (Te) junction, engineered to function as an optoelectronic synapse for neuromorphic vision applications. By incorporating CdSe-ZnS core-shell quantum dots (QDs) onto the IGZO-Te interface, the device simultaneously performs light sensing and synaptic learning, enabling in-sensor computing without the need for external data transfer between sensing and processing units.
Optical and structural analyses confirm the multilayer configuration of the device. Approximately 45 nm IGZO, 11 nm Te, and a 16 nm thick QDs layer. The optical bandgaps of IGZO, Te, and QDs were measured to be 3.95 eV, 0.65 eV, and 1.92 eV, respectively. The large bandgap of IGZO ensures low leakage current and high stability, whereas the narrow bandgap of Te extends detection into the near-infrared (NIR) region. QDs play a dual role by enhancing light absorption and acting as charge trapping sites that modulate the carrier transport within the heterojunction. As a result, the IGZO-Te-QD device exhibits a remarkable 9.78 times increase in photocurrent compared to IGZO-Te without QDs.
The device demonstrates robust synaptic performance under optical pulse stimuli (405 nm, 80 μWcm?²). Excitatory postsynaptic current (EPSC), paired-pulse facilitation (PPF), and pulse number and duration dependent plasticity are clearly observed. The photocurrent retention exceeds 300 seconds, confirming long-term memory functionality analogous to biological synapses. Furthermore, a 6×6 photodiode array was fabricated to evaluate pattern recognition capability. The array successfully distinguished an illuminated letter ‘O’, demonstrating both short-term and long-term synaptic weight modulation with uniform pixel response and negligible crosstalk.
Overall, this IGZO-Te-QDs optoelectronic synaptic photodiode integrates sensing, memory, and learning within a single device. The demonstrated broadband optical response, long-term retention, and scalable array characteristics highlight its strong potential for future energy-efficient neuromorphic vision hardware.
