Snakes detect prey in darkness using sensory organs that sense infrared heat. Thermal imaging cameras contain sensors serving the same function, and a special material capable of enhancing these sensors has now been developed using artificial intelligence technology. This breakthrough enables the production of high-performance thermal cameras and nighttime pedestrian detection systems for vehicles.
A research team led by Professors Son Chang-hee and Park Hyung-ryul from the Department of Physics at Ulsan National Institute of Science and Technology (UNIST) announced on the 25th that they have developed a multilayer thin-film material for microbolometer sensors with performance more than 20 times superior to commercial materials using AI technology.
A microbolometer is a sensor that detects infrared heat emitted by objects and converts it into electrical signals processable by electronic devices. Since the sensor operates on the principle that electrical resistance changes when special materials inside absorb infrared heat, creating high-performance sensors requires materials sensitive enough to show resistance changes even with minor temperature variations.
The material developed by the research team is based on vanadium dioxide, known for its high sensitivity. It consists of four stacked thin-film layers of vanadium dioxide doped with tungsten. Each layer is designed with different tungsten content and thickness, reducing the abrupt signal changes and hysteresis phenomena that have been persistent problems with vanadium dioxide.
Sensors have higher signal reliability when electrical resistance changes linearly with temperature changes, similar to a first-degree function. However, pure vanadium dioxide exhibits abrupt resistance changes in certain ranges and shows hysteresis, where electrical resistance values differ when temperature rises versus when it falls back down. This caused measurement values to vary even at the same temperature, reducing sensor signal reliability.
The theoretical number of thickness combinations for these thin-film layers exceeds 1.3 million. The research team used an AI technique called genetic algorithm to find the optimal thickness combination. This method mimics the "natural selection" principle of biological evolution, repeatedly selecting only high-performing combinations from randomly generated thickness variations, recombining and modifying them to arrive at the optimal configuration.
Experimental results showed the material achieved a temperature coefficient of resistance (TCR) of 7.3% in the room temperature range (20-45°C), more than three times higher than existing commercial materials. The beta (β) index improved 23.6-fold. The beta index is a comprehensive performance metric that evaluates actual sensor performance by considering not only sensitivity but also signal accuracy and reliability.
Additionally, the material can be directly deposited onto existing semiconductor circuits (CMOS) through a low-temperature process at 300°C, significantly enhancing commercialization potential. Microbolometers must be deposited on semiconductor circuits that read resistance change signals. Previous vanadium dioxide-based technologies required high-temperature processes above 500°C, risking damage to already completed circuits from excessive heat.
This research was led by researcher Choi Jin-hyun and Dr. Lee Hyung-taek from UNIST's Department of Physics as co-first authors. The research team explained, "This is meaningful research in that we dramatically shortened what would have taken an estimated 750 years if humans had to manually test each material combination one by one, using artificial intelligence technology, and directly synthesized the designed materials in thin-film form."
Professor Son Chang-hee stated, "This will be widely applied in fields requiring high-performance thermal detection technology, such as nighttime obstacle detection for autonomous vehicles, night and long-range surveillance using drones, and large-scale virus infection monitoring that detects body temperature changes in large numbers of people."
The research was published on January 28 in the prestigious journal "Advanced Science." The study was supported by the Ministry of Science and ICT's National Research Foundation of Korea (NRF) through the Nano Materials Technology Development Program, Basic Research Laboratory Program, and Mid-career Researcher Program, as well as the Institute of Information & Communications Technology Planning & Evaluation (IITP) through the Information Technology Research Center (ITRC) support program.