A know-how that surpasses the constraints of present sensors, which did not distinguish between water and asphalt on darkish roads, has emerged to reinforce the accuracy of autonomous driving and medical diagnostics. Our college’s analysis workforce has developed a next-generation polarization sensor that may learn the “direction” of sunshine and alter its personal response. KAIST introduced on May twelfth {that a} analysis workforce led by Professor Joonki Suh from the Department of Chemical and Biomolecular Engineering has developed a “self-reconfigurable” polarization sensor array know-how that regulates its operation by discovering the optimum state utilizing “polarization” info—the property of sunshine vibrating in a particular path. With the latest explosive improve in knowledge and the fast improvement of synthetic intelligence know-how, the necessity for next-generation imaginative and prescient programs that may effectively course of huge quantities of knowledge with low power is rising. However, present picture sensors solely detect the depth (brightness) of sunshine, limiting their means to exactly grasp the orientation or floor construction of objects. To overcome these limitations, the analysis workforce developed a polarization-based sensor know-how able to recognizing the vibration path of sunshine. In specific, by using a “heterostructure” that mixes two completely different supplies—tellurium (Te) and rhenium disulfide (ReS₂)—they successfully carried out traits the place the response to gentle varies relying on the crystal orientation.
.png)
To exactly stack the 2 supplies so that they cross one another, the analysis workforce utilized “Epitaxial Atomic Layer Deposition,” a course of that controls crystal constructions by stacking supplies exactly on the atomic layer degree. By guaranteeing the crystal constructions of the 2 supplies interlock precisely, they secured increased reproducibility and steady efficiency in comparison with earlier strategies. In this construction, when gentle is irradiated, interfacial service switch and trapping (a phenomenon the place electrons transfer or keep at particular areas) happen on the materials boundary. As a consequence, a “bipolar photoresponse”—a light-induced response the place the present path flips relying on circumstances equivalent to gentle depth, wavelength, and path—seems. A key function is that the sensor’s working state may be freely adjusted utilizing solely gentle, with out exterior electrical alerts. Furthermore, this know-how may be utilized to “in-sensor computing” constructions the place the sensor itself processes knowledge, permitting for the environment friendly processing of multi-dimensional optical info that modifications over time with out advanced calculation processes. In precise experiments, it recorded a excessive accuracy of over 95% in recognizing transferring objects, proving its potential for functions in varied fields equivalent to autonomous driving and medical prognosis.
Professor Joonki Suh said, “This research presents a new foundation for AI vision technology that can secure richer visual information by utilizing polarization information. It is expected to play an important role in implementing low-power, high-efficiency AI systems in the future.” Wenxuan Zhu (Postdoctoral Researcher) and Changhwan Kim (Ph.D. pupil) participated as first authors on this research, with Professor Joonki Suh collaborating because the corresponding writer. The analysis outcomes have been printed on April 14 within the worldwide educational journal Nature Sensors.
- Paper Title: Self-reconfigurable polarization notion in dual-anisotropy heterostructures for high-dimensional in-sensor computing
- Authors: Wenxuan Zhu, Changhwan Kim, Ruofan Zhang, Mingchun Lu, Namwook Hur, Hanbin Cho, Jihyun Kim, Jiacheng Sun, Joohoon Kang, Junchi Yan, Yuan Cheng & Joonki Suh
- DOI: https://doi.org/10.1038/s44460-026-00057-9
Meanwhile, this analysis was carried out with the help of the PIM AI Semiconductor Core Technology Development (Device) Project and the Individual Basic Research Project of the National Research Foundation of Korea, funded by the Ministry of Science and ICT, and the Industrial Innovation Talent Growth Support Project of the Korea Institute for Advancement of Technology (KIAT).