Domestic researchers have developed the world's first "chiral magnetic quantum dot" that simultaneously possesses optical chirality and magnetic properties. This innovation is condensed into a single device capable of processing, judging, storing, and resetting information like the human brain, presenting a new paradigm for high-performance artificial intelligence (AI) hardware.
Professor Yeom Ji-hyun of the Korea Advanced Institute of Science and Technology (KAIST) noted that her research team developed a special nanoparticle known as a quantum dot (CFQD) that exhibits both "chirality" and magnetism in response to light asymmetrically. They also succeeded in implementing an AI neuromorphic device (ChiropS) that mimics human brain structure and functions with low power usage.
The newly developed chiral magnetic quantum dot is synthesized by introducing chiral organic materials into inorganic nanoparticles based on silver sulfide, reacting differently according to the polarization direction of light. In particular, it shows varying responses across the entire visible spectrum at wavelengths of 405, 488, and 532 nm, which can be utilized as a platform for neural synapse devices capable of multi-channel recognition. Additionally, it stands out commercially for being environmentally friendly and highly stable, synthesized with a water-based process.
The research team fabricated a synapse transistor structure by stacking a silver sulfide layer utilizing chiral magnetic quantum dots on silicon with an organic semiconductor known as pentacene. This device demonstrates long-term potentiation (LTP) characteristics when illuminated with light, and it also incorporates an electrical erasure function, successfully mimicking learning and adaptation, akin to the human brain, through the use of light.
Moreover, when a laser light pulse is repeatedly flashed for a very short duration, the current gradually accumulates, forming a multi-level state that increases incrementally. This implies the ability to adjust synaptic weights, allowing the AI to learn like a brain and enabling multi-learning.
The researchers confirmed that when they created a 2×3 device array and illuminated it with light of different polarizations and wavelengths, the response currents of each device were distinctly differentiated. They revealed that the array could sense and process a total of nine pieces of information in parallel through six channels, allowing for at least nine times the information processing compared to existing technologies.
Furthermore, the device reacted like a smart sensor that performs complex judgments even when exposed to constant light. For instance, it has an auto-filtering role, which incorporates a function to filter out noise and amplify signals within the device itself. In practice, when noise is added to handwritten (MNIST) data and passed through the device, it was confirmed that high-frequency noise decreased, leaving only the essential information. It was noted that this device could operate with up to 30% less power compared to existing computing technologies.
This research merges optical chirality and magnetic spin characteristics into a single nanomaterial, achieving polarization differentiation and long-term memory capabilities that have not been realized before. The integration of detection (seeing), processing (judging), memory (storing), and initialization (erasing) functions into a single device is expected to significantly enhance the potential for creating high-performance artificial intelligence hardware that is smaller and more efficient.
In particular, the photonic synapse transistor developed by the research team, utilizing chiral quantum dots, is a key technology that consolidates various functions such as polarization differentiation, multi-wavelength recognition, and electrical erasure into a single device. This technology could also be applied in optical encryption, secure communications, and quantum information processing in the future.
Professor Yeom Ji-hyun emphasized, "To overcome the limitations of existing quantum dots, we designed a novel concept of quantum dots that fuse optical chirality and magnetic spin characteristics. Given that a single device can handle multiple polarizations and wavelengths, and incorporates the ability to be initialized with electrical signals, it can serve as an innovative platform for implementing low-power, high-precision AI systems."
This paper was published online on the 7th in the international journal "Advanced Materials."
References
Advanced Materials (2025), DOI: https://doi.org/10.1002/adma.202415366