10 Mar 2008
Intel's optical engineers have notched up another first: a cascaded Raman laser built in silicon that they believe could be used in gas sensing applications.
Researchers at Intel have demonstrated for the first time a cascaded Raman laser in silicon that produces laser light at 1850 nm. "Our laser is efficient, works at room temperature and could take silicon photonics technology to lots of everyday devices," said Mario Paniccia, director of Intel's silicon photonics lab.
Cascaded lasing is a unique attribute of Raman lasers, which use an optical pump to produce coherent light emission in wavelength regions that are hard to reach with conventional lasers. Another big advantage is that Raman lasers can be made from materials – such as silicon –that do not have the right energy band structure to produce laser light by stimulated emission.
"Silicon is particularly suitable as a Raman laser material in the near and mid-infrared regions because of its high Raman gain and optical transparency in these regions," explained Haisheng Rong, who leads Intel's research into Raman-based lasers and amplifiers.
The first continuous wave silicon Raman laser, which operated at a wavelength of 1686 nm, was demonstrated by Intel back in February 2005. Since then, however, the group has made significant improvements to the laser's design and performance.
"Instead of applying optical coatings on the chip facets to form a laser cavity, this design uses a mirror-less ring cavity which is monolithically integrated on a chip and fabricated entirely in a CMOS fab," said Rong. "We have achieved a 10x reduction in lasing threshold to 20 mW and a 5x increase in output power to more than 50 mW. Another important characteristic of the silicon Raman laser is its extraordinary spectral purity."
Cascaded lasing enables longer wavelengths
It's these improvements to the basic laser design that has enabled the group to achieve cascaded Raman lasing. Such cascaded designs exploit the fact that the output from a Raman laser is always at a longer wavelength than the pump. By feeding this first-order laser output back into the system, a second-order beam is produced at an even longer wavelength.
In Intel's prototype design, a continuous-wave pump laser at 1550 nm is directed through a polarization controller and then coupled to the ring-shaped waveguide cavity. Photons circulating in the ring acquire sufficient gain to undergo lasing, and emerge as a coherent beam at the output terminal.
The first laser beam to emerge from the cavity has a wavelength of 1686 nm, and feeding it back into the system then produces laser output at 1848 nm. "Through cascaded Raman lasing in silicon, one can convert pump wavelengths in the near-infrared region to wavelengths in the mid-infrared region," commented Rong.
The Intel team has also shown how its prototype device could be used in gas sensing, one of the most promising applications for a small, cheap and efficient mid-infrared laser. In the experiments, the laser beam was passed through a sample cell that contains either water vapour or methane, with the laser output set to match the wavelength at which these two gases absorb most light.
"For methane we used the first-order output at 1687 nm, and for water vapour the second-order output at 1847 nm," said Rong. "By comparing the measured absorption spectra with those in a database that contains the fingerprints of all known gas molecules, we can identify the gas molecule, its concentration, and other characteristics."
According to Rong, the measured absorption profiles agree very well with calculations based on the HITRAN (High-resolution Transmission Molecular Absorption) database. What's more, the high spectral purity of the laser ensure that the spectra are well resolved, with the width of the absorption lines only limited by the thermal motion of the molecules.
Impressive though these results are, extending the design to operate at still longer wavelengths will be needed to further improve the efficiency. "Current Raman lasers are inefficient because of two-photon absorption," commented Bahram Jalali, a silicon photonics expert at the University of California, Los Angeles. "As you get closer to mid-IR, as Intel has done with their impressive on-chip cavity design, two-photon absorption is reduced. If the team manages to extend the wavelength beyond 2.3 µm, then the two-photon problem disappears and the laser will become very efficient."
The Intel engineers believe that this could be achieved with higher-ordered cascaded lasing. "We were able to get second-order lasing in this prototype, and it's only a matter of time before higher order lasing can be achieved as well," said Paniccia.
However, higher-order lasing presents some serious engineering challenges. As resonance effects lie at the core of the laser mechanism, it is crucial to design and fabricate the laser cavity so that it has the right resonance conditions at the lasing wavelengths – and increasing the wavelength requires high-precision design and fabrication. Unforeseen material properties might also hinder development in the longer wavelength regime.