Professor Aydogan Ozcan and his research team at UCLA have created a pyramid-structured diffractive optical network that scales its layers pyramidally to match the direction of image magnification or demagnification. Their study was reported in Light Science & Applications.
Image Credit: Bijie Bai, Xilin Yang, Tianyi Gan, Jingxi Li, Deniz Mengu, Mona Jarrahi & Aydogan Ozcan
Diffractive deep neural networks, or D2NNs, are optical systems that use deep learning to optimize successive transmissive layers to carry out computing tasks in an all-optical way.
The UCLA research team, under the guidance of Professor Aydogan Ozcan, has created a pyramid-structured diffractive optical network that scales its layers pyramidally to match the direction of image magnification or demagnification.
This concept achieves unidirectional imaging with fewer diffractive degrees of freedom by guaranteeing high-fidelity image generation in one direction while restricting it in the other. The researchers’ demonstration of the capacity to reach larger magnification factors by cascading additional P-D2NN modules further illustrated the technology's flexibility and scalability.
Terahertz (THz) illumination was used to experimentally evaluate the P-D2NN architecture. The diffractive layers created by 3D printing were examined under continuous THz illumination. The numerical calculations and the practical results, which included several magnification and demagnification designs, complemented each other.
The outputs in the forward direction properly mirrored the enlarged or demagnified input images, but the reverse way yielded low-intensity, non-informative findings, as expected for unidirectional imaging.
Applications and Future Prospects
The P-D2NN framework is a potential tool for various applications since it can reduce backward energy transfer while distributing the original signal into undetectable noise at the output. These include monitoring, privacy-protected optical communications, decoupling of transmitters and receivers in telecommunications, and optical isolation for photonic devices.
The system’s potential in various defense-related applications is highlighted by its polarization-insensitive functioning and capacity to project high-power structured beams onto target objects while shielding the source from counterattacks.
The other study authors are Bijie Bai, Xilin Yang, Tianyi Gan, Jingxi Li, Deniz Mengu, Mona Jarrahi, and Aydogan Ozcan, who are affiliated with the UCLA Electrical and Computer Engineering Department. Professor Ozcan also serves as an associate director of the California NanoSystems Institute (CNSI).
US Office of Naval Research (ONR) supported the study.
Journal Reference:
Bai, B., et. al. (2024) Pyramid diffractive optical networks for unidirectional image magnification and demagnification. Light Science & Applications. doi:10.1038/s41377-024-01543-w