Diffraction Casting: A New Hope for Optical Computing

A novel optical computing technique, diffraction casting, promises to address the limitations of existing electronic systems and unlock potential for faster, more efficient computing.

In the ever-evolving world of computing, increasing performance demands from applications like artificial intelligence are driving the search for more efficient processing methods. Traditional electronic computers are reaching their physical limits, generating excessive heat and consuming vast amounts of power. Optical computing, which utilises light instead of electricity, is one potential solution. Despite its potential, the technology has faced significant hurdles preventing it from mainstream adoption. However, a new design architecture, diffraction casting, could offer a way forward.

At present, all modern computers rely on electronic technology. This has inherent drawbacks: higher performance comes at the cost of increased heat generation, and fabrication methods are approaching their theoretical limits. To overcome these challenges, researchers have been investigating alternatives, with optical computing standing out as one promising avenue.

Optical computing takes advantage of light waves, which can travel at high speeds and interact in complex ways without producing heat. This could lead to computers that are both fast and power-efficient. Theoretically, multiple wavelengths of light could be used simultaneously, allowing for massively parallel computation. Yet, despite these benefits, optical computing has struggled to make the leap from concept to reality.

“In the 1980s, researchers in Japan explored an optical computing method called shadow casting, which could perform simple logical operations,” explains Associate Professor Ryoichi Horisaki from the University of Tokyo’s Information Photonics Lab. “However, their implementation relied on bulky geometric optical forms, akin to the vacuum tubes used in early digital computers. While they worked in principle, they lacked flexibility and integration, preventing practical use.”

Horisaki and his team have developed a new technique, diffraction casting, which builds on the earlier shadow casting method but with significant improvements. “Shadow casting was based on the interaction of light rays with geometric forms,” says Horisaki. “Diffraction casting, on the other hand, leverages the properties of the light wave itself. This results in more spatially efficient and functionally flexible optical elements, which are crucial for creating a universal computer.”

The team has already demonstrated promising results with simulations, using small 16-by-16 pixel images to perform logic operations. These images, smaller than icons on a smartphone screen, acted as input data for the system. Their approach is to develop an all-optical system, where only the final output is converted into an electronic signal. Prior to this stage, all processing steps remain optical. This could significantly improve efficiency and speed in tasks such as image processing, with potential applications in machine learning.

“Think of it like layers in an image-editing application such as Adobe Photoshop,” says lead author Ryosuke Mashiko. “We have an input layer — a source image — and can overlay other layers that manipulate the data in various ways. The final layer represents the processed output. In our case, light passes through these layers, casting an image onto a sensor, which is then converted into digital data.”

The diffraction casting technique is still in its early stages, but the researchers believe it could eventually become a key component in future computing systems. “This is just one building block,” Mashiko notes. “Like graphical processing units (GPUs), which handle specific tasks such as graphics and machine learning, diffraction casting could play a specialised role within a larger system. While the technology won’t be commercially available for another decade, there’s potential for its use in areas like quantum computing.”

Although much work remains to be done before diffraction casting can be realised in hardware, the technique could one day help overcome the limitations of electronic systems and pave the way for more powerful, energy-efficient computing solutions.

You can read the team’s paper “Diffraction casting” in Advanced Photonics.