SANTA CLARA, CA — The first all-silicon Raman laser on a chip has been demonstrated by scientists at Intel, bringing closer to reality the eventual development of an all-optical computer that uses photons instead of electrons to process information.
Although an all-silicon Raman laser was demonstrated last year by researchers at the University of California at Los Angeles (see Laser Focus World, Dec. 2004, p. 9), it required an external fiber ring cavity to produce lasing. The Intel laser is entirely integrated in a silicon chip, with the laser cavity consisting of a waveguide fabricated on the surface of an undoped silicon-on-insulator (SOI) substrate.
According to Kevin Teixeira, Intel R&D media relations, the Silicon Photonics group within Intel has a four- to five-year history of advancing the practicality of silicon photonics in commercial applications.
“The promise of the Intel Silicon Photonics group is the development of silicon-based devices using standard CMOS-like fabrication techniques that reduce costs to the point where better, faster information processing by light in silicon can be commercially realized,” Teixeira said.
Using standard projection photolithography and plasma reactive-ion etching fabrication techniques, the SOI waveguide (with a width, rib height, and etch depth of 1.5, 1.55, and 0.7 mm, respectively) was formed into an s-shaped curve with a bend radius of 400 mm and a total length of 4.8 cm. The laser optical cavity was prepared by coating one of the waveguide facets with a multi-layer coating designed to be broadband with a high reflectivity (~90%) for the pump wavelength of 1.536 mm and the Raman/Stokes wavelength of 1.67 mm, while the other facet was left uncoated for coupling-in the pump beam. Although the laser beam is emitted from both facets of the silicon chip, the laser output was measured from the uncoated facet.
To reduce the effect of two-photon free- carrier absorption in silicon that both attenuates the pump beam and increases the optical loss for the Raman signal, the researchers designed a p-i-n diode structure along the rib waveguide. When a reverse bias is applied to this region, the electron-hole pairs that are generated can be swept out of the silicon waveguide by the applied electric field, thus reducing the effective carrier lifetime and reducing the optical loss of the signal. For the Intel silicon laser, lasing occurred with a pump-power threshold of 0.4 mW, at a wavelength of 1669.5 nm, before lasing was halted by the two-photon effect. Optimization of the waveguide cavity and further reductions in the carrier lifetime could lead to continuous-wave (CW) Raman lasing in silicon.
According to news reports, the laser was fabricated in an existing fab using standard high-volume manufacturing processes. While the Intel device currently achieves pulsed lasing, Teixeira is optimistic that the Silicon Photonics group will achieve CW output, “and that’s going to be the big news,” he said. Intel researchers are now working to solve the two-photon absorption problem, which is critical to achieving CW output. No doubt the fervor for development of an all-optical computer will be renewed based on this Raman laser breakthrough.
Complete details of Intel’s silicon-laser research are to be published in Nature magazine later this year.
-Gail Overton