Caltech Brings 20-Fold Lower-Loss Visible Photonics to Silicon Chips
Updated
Updated · ScienceDaily · Aug 17
Caltech Brings 20-Fold Lower-Loss Visible Photonics to Silicon Chips
3 articles · Updated · ScienceDaily · Aug 17
Summary
Caltech researchers built germano-silicate waveguides on standard 8- and 12-inch silicon wafers, moving light with losses at visible wavelengths that approach optical-fiber performance.
At visible wavelengths, the platform beats silicon nitride's record by a factor of 20 because the glass can be furnace-reflowed to atomically smooth surfaces that sharply cut scattering loss.
Near-infrared performance already matches some of the best silicon-nitride devices, while lasers made on the new platform show more than a 100-fold improvement in coherence over earlier designs.
The low-loss paths are patterned as compact spirals and ring resonators, letting light circulate over effective distances of meters or kilometers on chips only 2 centimeters across.
That combination could enable chip-scale atomic clocks, gyroscopes, quantum and ion-trap systems, and lower-energy optical links for AI data centers.
If this near-perfect glass chip makes visible-light photonics twenty times better, what hidden commercial applications will it suddenly make possible?
How does a simple heating trick heal flawed silicon chips to unlock the ultrafast optical networks needed for tomorrow's quantum computers?
Could wrapping glass into microscopic spirals finally solve the massive energy bottlenecks threatening the future of AI data centers?
From Fiber to Chip: Caltech’s Ultralow-Loss Germano-Silicate Photonic Platform and the Future of Optical Computing
Overview
In February 2026, Caltech researchers achieved a major breakthrough by printing ultralow-loss optical circuits made from germano-silicate glass onto standard silicon wafers. By shaping the glass into spiral waveguides and using a special thermal process to smooth their surfaces, they created chips that guide light with almost no energy loss. This innovation allows lasers on these chips to maintain coherence over 100 times longer than before and enables seamless light transfer between optical fibers and semiconductor lasers. The technology promises to lower energy use in AI data centers and supports compact, high-precision devices for quantum computing and sensing, all while being compatible with existing chip manufacturing lines.