Different forms of linear transformations, such as the Fourier transform, are widely employed in processing of information in various appli­cations. These transformations are generally imple­mented in the digital domain using elec­tronic processors, and their compu­tation speed is limited with the capacity of the electronic chip being used, which sets a bottle­neck as the data and image size get large. A remedy of this problem might be to replace digital processors with optical counter­parts and use light to process information.

Now, a team of optical engineers, led by Aydogan Ozcan from the Cali­fornia Nano­Systems Institute (CNSI) and the Electrical and Computer Engi­neering Department at the University of Cali­fornia, Los Angeles (UCLA), and co-workers have developed a deep learning-based design method for all-optical compu­tation of an arbitrary linear transform. This all-optical processor uses spatially-engi­neered diffractive surfaces in manipulating optical waves and computes any desired linear transform as the light passes through a series of diffrac­tive surfaces. This way, the computation of the desired linear transform is completed at the speed of light propa­gation, with the trans­mission of the input light through these diffractive surfaces. In addition to its computational speed, these all-optical processors also do not consume any power to compute, except for the illu­mination light, making it a passive and high-throughput computing system.

The analyses indicate that deep learning-based design of these all-optical diffrac­tive processors can accu­rately synthesize any arbitrary linear trans­formation between an input and output plane, and the accuracy as well as the diffrac­tion effi­ciency of the resulting optical transforms signi­ficantly improve as the number of diffrac­tive surfaces increases, revealing that deeper diffrac­tive processors are more powerful in their computing capa­bilities.

The success of this method has been demons­trated by performing a wide range of linear trans­formations including for example randomly generated phase and amplitude trans­formations, the Fourier transform, image permu­tation and filtering operations. This computing framework can be broadly applied to any part of the electro­magnetic spectrum to design all-optical processors using spatially-engi­neered diffractive surfaces to univer­sally perform an arbitrary complex-valued linear transform. It can also be used to form all-optical infor­mation processing networks to execute a desired computa­tional task between an input and output plane, providing a passive, power-free alter­native to digital processors. (Source: ULCA)

Reference: O. Kulce et al.: All-optical synthesis of an arbitrary linear transformation using diffractive surfaces, Light Sci. Appl. 10, 196 (2021); DOI: 10.1038/s41377-021-00623-5

Link: Electrical and Computer Engineering Dept., University of California, Los Angeles, USA