Silicon Raman Laser Cascades Toward Mid-IR Spectral Region

Breck Hitz

The mid-IR spectral region, often defined as wavelengths between 2 and 5 μm, is important to spectroscopy because virtually all molecules have unique rotational-vibrational “fingerprints” in that region. However, building compact and efficient laser sources in the region is challenging. Today, most mid-IR laser sources are based on bulky and expensive solid-state lasers, whose nominal 1-μm outputs are converted via nonlinear optics, or on lead-salt diode lasers that typically operate at cryogenic temperatures.

Figure 1. Nine ring resonators, such as the one diagrammed in Figure 2, are fabricated on the upper half of this chip. Each one has a 3-cm circumference. On the lower quarter of the chip are six 1.5-cm rings. The waveguide structures in the middle are for device characterization.

But recently, Haisheng Rong and his colleagues at Intel in both Santa Clara, Calif., and in Jerusalem demonstrated what they believe is the first cascaded Raman laser in silicon, and they showed that it is likely to be a viable laser source in the mid-IR spectral region.

Figure 2. When the 1550-nm pump light was tuned into resonance with the ring cavity, gain and lasing occurred at both the first and second Stokes lines. In principle, the process can be extended to the third Stokes line and beyond, producing outputs all the way up to 3 μm. The dotted lines in the lower diagram indicate how wavelength-tuning the pump laser results in a wavelength shift of the Stokes lines. Reprinted from Nature Photonics with permission of the researchers.

To date, their Raman laser has not generated output in the mid-IR, but their experiments clearly show the way to get there. They constructed a silicon-on-insulator rib waveguide in the shape of a ring resonator and coupled light into it from an adjacent bus waveguide (Figure 2). Starting with a pump wavelength of 1550 nm, they obtained output at the first and second Stokes wavelengths, 1686 and 1848 nm, respectively. Theoretical considerations indicate that the process eventually could be iterated all the way to the sixth Stokes wavelength at ∼3 μm.

Figure 3. The first Stokes achieved threshold and increased rapidly until the second Stokes reached threshold and saturated it. Reprinted from Nature Photonics with permission of the researchers.

As the scientists increased the pump power, the first Stokes reached its threshold at about 80 mW of launched pump power and grew quickly with increasing pump until the second Stokes reached threshold at about 120 mW (Figure 3). At that point, the first Stokes saturated and remained relatively independent of increasing pump, as the extra pump power was converted to the second Stokes.

Figure 4. An optical spectrum analyzer showed the pump wavelength and the two Stokes wavelengths, but its resolution was too low to display the true width of the Stokes line. (The small spike next to the pump is amplified spontaneous emission.) Reprinted from Nature Photonics with permission of the researchers.

The scientists observed the Raman laser’s output using an optical spectrum analyzer, which displayed the pump wavelength and the two Stokes lines (Figure 4). The Stokes lines were narrower than the analyzer’s resolution, so the scientists built a fiber ring interferometer whose free spectral range was 15 MHz, and they coupled the second Stokes line into it. From the resulting data, they concluded that the bandwidth of the Stokes line was less than 2.5 MHz.

Figure 5. The methane spectrum obtained with the output of the Raman silicon laser showed excellent agreement with theory. (The Fabry-Perot interference fringes at the bottom served as a wavelength marker for calibration purposes.) Reprinted from Nature Photonics with permission of the researchers.

They not only wanted to demonstrate cascaded Raman lasing in silicon but also to show that the technique produced a laser whose power, stability and wavelength tunability were suitable for spectroscopic applications. To do so, they illuminated a methane cell with the first Stokes output and recorded the spectrum. They tuned the laser’s wavelength smoothly — that is, without mode hopping — by tuning the pump laser wavelength while thermally keeping the same lasing mode in resonance with the ring Raman resonator. The resulting spectroscopic data were in excellent agreement with theory (Figure 5). A similar experiment, using the second Stokes line as well as a water vapor cell, showed similar agreement with theory.

Nature Photonics, March 2008, pp.170-174.