US researchers have unveiled an on-chip optical isolator that they say can be made using CMOS-compatible materials and processes. The team from Stanford University has filed a patent application for the device and believes that it could greatly reduce the cost of future optical chips (Nature Photonics doi:10.1038/nphoton.2008.273).
"There is intense interest in integrated photonic chips within the photonics community, however, integrated optical isolators have so far been a roadblock to their development," says Zongfu Yu, a researcher in Stanford's Department of Applied Physics group. "We have provided a promising solution to this issue by demonstrating that integrated isolators can be made without exotic or nonlinear materials."
Optical isolators allow light to travel in one direction only and play an essential role in fibre-optic communication networks. Optical isolators prevent interference between individual components in the networks, which would otherwise cause a complicated parasitic response. As researchers push to shrink the optical network onto a chip, a viable isolation mechanism that is suitable for on-chip integration is becoming increasingly important.
Until now, isolation mechanisms typically rely on magneto-optical materials or use external magnetic fields, while other schemes involving nonlinear materials cannot provide complete isolation. Crucially none of the conventional schemes are suitable for CMOS integration.
The Stanford team proposes a device based on inter-band photonic transitions in a silicon waveguide. Uniquely, the approach involves electro-optical effects, which means that it can be easily integrated into a CMOS process.
In the setup, a dynamic index modulation is applied a to a silicon waveguide. The modulation is designed so that it can only induce a photonic transition for forward propagating modes.
"As light passes through the waveguide in the forward direction it is converted into a different mode but in the backward direction, light passes through freely," explained Yu. "We then use a filter to absorb the converted mode and we achieve the desired result — light can only pass through the waveguide in one direction."
Yu admits that further performance optimization and experimental demonstrations are needed before commercialization. "Currently, our modulation technique in silicon provides a sizable operational bandwidth," he concludes. "However, to further improve the performance and reduce the footprint of the device, it is desirable to have stronger and faster index modulation."