Brifing

Caltech's femtosecond light-steering chip enables ultrafast photonic computing

New meta-surface device redirects light in 74 quadrillionths of a second, breaking speed barriers

1Photonic computing's speed bottleneck

Current photonic technologies face fundamental speed limits due to electron relaxation times in conventional materials. Existing light-steering methods using liquid crystals or optical chips operate at nanosecond or picosecond scales, creating a bottleneck for ultrafast computing applications. The Caltech team's breakthrough addresses this by eliminating electronic state transitions entirely, achieving femtosecond-scale control through purely optical means.

2Meta-surface design breakthrough

The device uses amorphous silicon nano-pillars smaller than light wavelengths to trap and circulate photons. This meta-surface design extends light-matter interaction time, amplifying the normally weak optical Kerr effect. By precisely engineering pillar size and spacing, researchers achieved a 13-degree beam deflection - sufficient for practical photonic applications. The nanostructures enable control without electrical signals, removing traditional speed limitations.

374-femtosecond steering mechanism

The system uses a powerful 'pump' beam to temporarily alter a material's refractive index via the optical Kerr effect. A weaker 'probe' beam then passes through this modified region, changing direction according to the pump's pattern. Crucially, electrons aren't excited to higher states, avoiding relaxation delays. The current 74-femtosecond speed matches the pump pulse duration, suggesting potential for even faster operation with shorter pulses.

4Overcoming Kerr effect limitations

While the optical Kerr effect normally produces negligible refractive index changes, the meta-surface amplifies it sufficiently for practical beam steering. The nano-pillars' design causes light to circulate rather than pass straight through, increasing interaction strength. This engineered light trapping makes the effect strong enough to redirect probe beams without requiring electron state changes that would slow the process.

5Speed limited by lasers, not physics

The 74-femtosecond demonstration represents a system limitation from current laser technology, not fundamental material constraints. Researchers note the meta-material itself could potentially operate at even shorter timescales. This leaves room for improvement as laser pulse technologies advance, with future systems possibly reaching attosecond (quintillionth-second) domains for next-generation photonic applications.

6Potential photonic computing applications

The technology could enable ultrafast optical switches for photonic neural networks and quantum computing systems. Its speed surpasses conventional electronic components by orders of magnitude, making it suitable for processing high-bandwidth optical data streams. Potential implementations include time crystals and synthetic time-varying optical materials that require femtosecond-scale control precision.

7Research team and publication

The work was led by Claudio Hail during his postdoctoral research at Caltech's Atwater Lab, now at UC Berkeley. Collaborator Lior Michaeli has moved to Tel Aviv University. Their findings appear in Nature Nanotechnology, supported by multiple funding sources including the Air Force Office of Scientific Research and Swiss National Science Foundation. The Kavli Nanoscience Institute provided critical infrastructure.

8Future development directions

Next steps involve optimizing the meta-surface for different wavelength ranges and improving deflection angles for practical device integration. Researchers aim to develop arrays of these steering elements for complex optical routing applications. The technology's compatibility with existing semiconductor fabrication methods suggests potential for scalable production of ultrafast photonic chips in coming years.

Original article: Latest Science News -- ScienceDaily — Caltech's femtosecond light-steering chip enables ultrafast photonic computing