Professor Hyong-Ryeol Park and his research team at UNIST have developed technology to amplify terahertz (THz) electromagnetic waves more than 30,000 times. Combined with artificial intelligence algorithms based on physical modeling, this amplification of terahertz waves could enable large-scale commercial utilization of 6G frequencies.
Professor Park's team, in collaboration with Professor Joon Sue Lee from the University of Tennessee and Professor Mina Yoon from Oak Ridge National Laboratory, successfully developed a terahertz nanoresonator optimized for 6G transmission. Thanks to the integration of artificial intelligence models based on physical theoretical foundations, the researchers succeeded in efficiently designing terahertz nanoresonators on a personal computer. This process, which used to be time-consuming even with supercomputers, has now been greatly sped up.
AI is more than 30,000 times more efficient than older terahertz nanoresonators
Through terahertz electromagnetic wave transmission experiments, the team evaluated the performance of the nanoresonator it developed. The results were astonishing: the resulting electric field was more than 30,000 times greater than that of conventional waves. Compared with previous terahertz nanoresonators, this new artificial intelligence technology improves performance by more than 300% for 6G.
"While traditional AI reverse design techniques focus on optical devices, their application at 6G frequencies (0.075 to 0.3 THz) poses significant challenges. At these scales, which are a million times smaller than the wavelength, the approach needs to be thorough Reform," Professor Parker explained.
Artificial intelligence significantly reduces nanoresonator optimization time
To address these challenges, researchers led by Professor Hyong-Ryeol Park developed a method to combine novel terahertz nanoresonators for 6G with artificial intelligence inverse design methods. This new method is actually based on theoretical physical modeling.
This process allowed them to optimize the device on a PC within 40 hours. This is much lower than a single simulation that would require tens of hours of optimization, or even hundreds of years for complete optimization.
Dr. Young-Taek Lee, first author of the study, also highlighted the versatility of the obtained nanoresonators. Its properties open up major perspectives in different fields. For example, the design of miniature molecular sensors, ultra-precision detectors, and even the study of bolometers. "The developed method can be applied to a variety of research objects. However, it involves theoretical physical modeling of various wavelengths or structures," he explains.
Professor Park emphasized the complementarity between artificial intelligence and detailed understanding of phenomena. "While artificial intelligence may appear to be a catch-all solution, mastering basic physical principles is still critical," he added.
