"Our indium phosphide laser can be used by itself for simple communications tasks, or can serve as the on-chip optical pump for our all-silicon Raman laser," said UCSB professor John Bowers. "Raman lasing is very tunable over a broad range, but our problem in the past was all the cost and fiber associated with an off-chip pump."
The new hybrid InP/silicon laser has a fixed wavelength, making it more scalable than the previous Raman laser, which had a variable wavelength. The InP laser's wavelength is fine-tuned using the already highly refined etching and lithography tools employed for the rest of the CMOS chip. Intel plans to build arrays of such lasers and individually tune each one for a different communications wavelength to realize photonic multiplexers for future optical interconnects.
To combine indium phosphide and silicon, Intel first etched the waveguide on a silicon wafer, then separately fabricated the InP laser on another wafer. The indium phosphide wafer was then flipped and attached to the silicon wafer with a glass-glue plasma process--a technique pioneered to bond gallium arsenide lasers to gallium nitride wafers.
Next, Intel plans to improve the performance of its indium phosphide laser by shrinking the device to lower its threshold from the 65 milliamps required today to initiate lasing. The team also wants to move to an integrated waveguide with less diffraction to achieve higher efficiency.
"When we start shrinking the device, that will lower its threshold and improve its other performance characteristics. Once we have the threshold down and an integrated cavity, then we will move to putting the laser alongside modulators and other optical components on a photonic chip," said Paniccia.
Bowers and Paniccia will present their photonics strategy for combining indium phosphide lasers with CMOS chip processing at the Frontiers in Optics conference, slated for Oct. 8-12 in Rochester, N.Y.