image: Figure 1. Architecture design and operational mechanism. a. （i）Schematic illustration showing the snake vision systems during daytime and night. (ii) Snakes focus on their target objects through the eyes during the daytime. (iii) Snakes generate a “thermal image” of prey through the pit organ for detecting infrared radiation. (iv) Schematic of the pit organ structure. b. Schematic illustration of the artificial vision perception through integrating the CMOS silicon sensors with infrared-to-visible upconverters enabling infrared visualization. c. Picture of an 8-inch silicon wafer, a zoomed view of the silicon sensor, and the optical micrograph image of the pixel region. d. (i) Schematic of the architecture and (ii) cross-sectional TEM image of CMOS sensors-integrated upconverters. e. Energy band structure diagram of upconverters without and with infrared illumination. f. Response and emission spectra of upconverters.
Credit: Xin Tang et al.
In the natural world, snakes possess a unique “superpower”. Using specialized “pit organs”, they can detect infrared radiation, generating thermal images of prey even in total darkness. This biological wisdom has long inspired scientists seeking to expand human vision beyond the visible spectrum.
In modern society, the demand for “seeing the invisible”—whether for night vision, autonomous driving through fog, or industrial defect detection—is growing rapidly. However, existing high-performance infrared imaging systems face significant hurdles. They often rely on expensive materials (like InGaAs) or bulky cryogenic cooling systems to suppress noise, making them difficult to miniaturize or produce at a low cost. However, the ubiquitous and affordable silicon-based CMOS cameras found in smartphones are blind to infrared light.
In a new paper published in Light: Science & Applications, a team of scientists led by Professor Xin Tang and Dr. Ge Mu from the Beijing Institute of Technology has developed a snake-inspired artificial vision system that bridges this gap. They innovatively integrated an infrared-to-visible upconverter directly onto a standard silicon CMOS sensor, enabling 4K ultra-high-resolution imaging of short-wave infrared (SWIR) and mid-wave infrared (MWIR) at room temperature.
“To endow artificial systems with snake-like thermal perception, we had to overcome the noise issues inherent in room-temperature infrared detection,” the researchers explained. They designed a unique “barrier heterojunction” architecture using mercury telluride (HgTe) colloidal quantum dots. Unlike traditional structures, this design utilizes zinc oxide (ZnO) and polymer layers to build an efficient barrier. This effectively blocks the “dark current” (noise) caused by heat while allowing signal carriers to pass through, ensuring high sensitivity without the need for cooling.
Furthermore, to convert the captured infrared signals into visible light that the CMOS sensor can read, the team introduced a “co-hosted” light-emitting unit. This optimized structure balances the transport of electrons and holes, significantly boosting the conversion efficiency. The system allows weak infrared signals to be converted into bright, visible images.
The result is a game-changer for imaging technology. By integrating this upconverter with a standard 8-inch CMOS wafer, the team achieved a resolution of 3840 × 2160 pixels (4K) with a pixel pitch of just 1.55 microns. The system can see through silicon wafers and generate thermal images of heat sources, capabilities previously reserved for expensive, specialized equipment.
“This technology opens a low-cost, high-performance path for infrared imaging,” the scientists forecast. “By extending the detection range from visible light out to 4.5 microns—widening the spectrum by 14 times—we enable all-weather vision that can penetrate smoke and fog.”. This breakthrough holds immense potential for intelligent manufacturing, autonomous vehicles, and medical diagnostics, bringing "super-vision" from the animal kingdom into our daily lives.
Article Title
Infrared visualized snakes-inspired artificial vision systems with CMOS sensors-integrated upconverters