Paralleled and multiplexed all-optical logic operation

In recent years, the rapid growth of artificial intelligence and the Internet of Things (IoT) has driven a surge in computational demand. Traditional electronic computing, however, is approaching its performance limits as predicted by Moore’s law, which forecasts the doubling of transistors on a microchip approximately every two years. This has led to a search for new computing paradigms that can meet the increasing demands for speed, scale, and energy efficiency. One promising avenue is optical computing, which leverages the unique properties of light to perform computations.

Researchers at the Information Photonics Lab at the University of Tokyo have made significant strides in this field with the development of a novel optical computation architecture called “diffraction casting.” As reported in Advanced Photonics, this method builds on the concept of spatial parallelism of light, a principle first explored in the 1980s with a technique known as “shadow casting.” While shadow casting provided valuable insights, it was limited by its reliance on geometrical optics, which restricted its flexibility and integration capabilities.

The new diffraction casting technique overcomes these limitations by using wave optics. Layers of diffractive optical elements are trained to exploit the spatial parallelism and wave properties of light, such as diffraction and interference. This allows for scalable and parallel logic operations with high flexibility and integration capability. The operations can be altered simply by changing the illumination patterns, eliminating the need for encoding and decoding of inputs and outputs.

Numerical demonstrations of diffraction casting have shown impressive results, achieving all sixteen logic operations on two arbitrary 256-bit parallel binary inputs without error, and at the speed of light. This architecture offers significant advantages in scalability and integration, making it a promising candidate for next-generation computing systems. Its flexible and reconfigurable nature also opens a wide range of applications, from image processing to optical computing accelerators.

Results of the numerical experiment of diffraction casting. (a) Example of a 256-bit parallel binary input pair and its corresponding (b) 16 logic operation outputs. (c) Computational errors associated with varying the number of diffractive optical elements (DOEs) when parameterizing the number of computation bits and operation types. This analysis indicates the computational performance, scalability, and integration capability of diffraction casting, which drastically reduced the number of DOEs leading to high practicality. Credit: Ryosuke Mashiko et al., doi DOI 10.1117/1.AP.6.5.056005

This research, conducted as part of the Grant-in-Aid for Transformative Research Areas project led by Professor Tetsuya Kawanishi at Waseda University, highlights the potential of optical computing using spatial parallelism as a building block for future computing systems. It also lays the groundwork for a new information processing framework that integrates imaging, sensing, and computing, potentially expanding into various fields.

For details, see the original Gold Open Access article by Ryosuke Mashiko, Makoto Naruse, and Ryoichi Horisaki, “Diffraction casting,” Adv. Photon. 6(5), 056005 (2024), doi 10.1117/1.AP.6.5.056005