Thermally Tunable Photonic Synaptic Transistor with Bioinspired Temporal Dynamics for Task-Adaptive Neuromorphic Vision

Inspired by temperature-associated temporal redistribution in biological systems, we demonstrate a thermally tunable photonic synaptic transistor (TT-PST) based on a magnesium-doped zinc oxide (MZO)/indium-gallium-zinc oxide (IGZO) heterostructure. By using temperature as a simple external control parameter, we enable continuous and reversible reshaping of synaptic temporal behavior without structural reconfiguration. As the temperature increases from 20 to 80 °C, the maximum photoresponsivity (Max. PR) rises from 0.43 to 2.25 A/W, while the temporal persistence decreases sharply, resulting in a transition from integration-dominant to response-dominant operation. Leveraging this capability, temperature-dependent temporal encoding is evaluated in two representative visual tasks using experimentally extracted device dynamics in a convolutional neural network (CNN) framework. Weak-signal recognition under low-light conditions achieves higher accuracy at 20 °C (82.86%) compared to 80 °C (69.23%), whereas dynamic motion analysis shows superior performance at 80 °C (88.03%) versus 20 °C (74.37%). These findings establish temperature as a bioinspired, physically accessible knob for on-demand temporal dynamics engineering in a single hardware platform, providing an efficient strategy for task-adaptive neuromorphic vision systems.