Computation-Less, Bias-Free, and Ultra-Broadband Photodetector for On-Chip Edge Extraction

Edges represent the critical structural features in visual information, and their extraction is essential for image processing. However, mainstream edge extraction technologies either rely on numerical computation implemented in software or hardware, or require auxiliary delicate optical components, thereby facing challenges in energy consumption, structural complexity, or operational spectral range. Here, a physical-mechanism-empowered, edge-sensitive photodetector is proposed to achieve computation-less on-chip edge extraction, featuring bias-free operation and an ultra-broadband photoresponse. The device exhibits a position-dependent bi-directional photoresponse driven by the diffusion and asymmetrical transport of photocarriers, thereby realizing an intrinsic spatial differentiation. Consequently, a single device can directly perceive both edge intensity and orientation at the pixel level. Experimental results also demonstrate that the device can operate without any external power supply and exhibits an ultra-broad operating spectral range from the near-infrared to the long-wave infrared at room temperature, with commendable sensitivity to blackbody radiation. The results highlight the feasibility of exploiting intrinsic physical mechanisms for edge extraction, offering an insightful pathway toward the development of compact, low-power, and functional optoelectronic devices.