A team of scientists has developed a new way to design photonic microchips that could make future computers, artificial intelligence systems, and communication networks much faster and more compact.
Instead of carefully designing each chip component by hand, the researchers used a powerful computer algorithm to create tiny structures that are much smaller than traditional designs while still working extremely well.
The research was carried out by scientists from the Max Planck Institute for the Science of Light in Germany and Harvard University in the United States.
Their findings were published in the journal Nature Communications.
Photonic microchips use light instead of electricity to move and process information.
Because light travels much faster than electrical signals and generates less heat, these chips are becoming increasingly important in modern technology.
They are already used in high-speed internet systems, AI data centers, advanced sensors, and emerging quantum technologies.
Inside these chips, light travels through extremely small pathways called waveguides. These pathways are only a few micrometers wide, yet they can carry huge amounts of information. The chips also contain tiny components that direct, split, reflect, or temporarily store light so it can perform different tasks.
Traditionally, engineers have designed these components by starting with familiar shapes and gradually adjusting them until they achieve the desired result. While this method works well, it can take a long time and limits how small the components can become.
In the new study, the researchers used a different method called inverse design. Instead of drawing the component first, they told a computer exactly what they wanted the light to do. The algorithm then searched through countless possible shapes until it found one that could perform the job.
Many of the final designs look unusual, with irregular patterns of tiny holes and ridges that no human engineer would likely have imagined.
Despite their strange appearance, these structures guide light very accurately while fitting into an incredibly small space. Some are built in an area up to 1,000 times smaller than the width of a human hair.
The researchers focused on a material called thick silicon nitride, which is widely used in high-performance photonic chips because it allows light to travel with very little loss. It can also help generate stable, laser-like light in many different colors. Until now, most components made from this material were still designed by hand.
The team successfully created and tested three important types of photonic devices. The first, called wavelength splitters, separate different colors of light. The smallest versions occupy 50 to 300 times less space than traditional designs. They also developed mode sorters, which separate light into different spatial channels while using about 500 times less space than conventional devices.
The third device was a tiny mirror that reflects up to 98.5% of incoming light while blocking unwanted light patterns. When two of these mirrors are placed together, they create a miniature optical cavity where light can bounce back and forth more than 100 times before leaving. These cavities are important for improving the performance of many photonic systems.
A major advantage of the new approach is that the algorithm also considers how the devices will actually be manufactured. The software includes limits on feature size and accounts for small production variations, making the finished designs compatible with commercial chip manufacturing processes.
The researchers now plan to combine these compact components with advanced optical circuits that can generate optical frequency combs. These special light sources produce many evenly spaced colors of light and are used in precision measurements, telecommunications, and quantum technologies.
The study shows how computer algorithms can help create photonic chips that are much smaller, easier to manufacture, and capable of supporting the next generation of high-speed technologies. As demand continues to grow for faster computing and communication, this new design approach could play an important role in shaping future photonic devices.
