Researchers worldwide are pursuing several approaches to building tunable lasers that will enable future telecom systems to be quickly reconfigured for maximum transmission efficiency. Some approaches involve changing the feedback frequency of the laser's resonator, and others combine multiple lasers to span the frequencies of the transmission system. The wavelength accuracy of the multiple laser designs is often insufficient, and additional elements are required to eliminate inaccuracies.A new multilaser concept promises much greater wavelength accuracy in a compact design, with all elements integrated onto a single chip. But the concept behind this approach is subtle.Figure 1. Research groups are investigating new methods to achieve tunable lasers for telecommunications applications. One approach exploits the frequency comb produced by an arrayed waveguide grating. The transmission through an arrayed waveguide grating at a fixed angle to its exit aperture, like that of any interferometer, is a "comb" of frequencies whose "teeth" are separated by the device's free spectral range. At a different fixed angle, the transmission is another comb with the same free spectral range, but the teeth are slightly offset from those of the first (Figure 1).If two arrayed waveguide gratings with slightly different free spectral ranges are arranged back-to-back, the angle-specific transmission through each of them will be the combs shown in Figure 2, but the only frequencies that can pass through both gratings are those that simultaneously match one of the teeth of both combs. In the figure, only light whose frequency matches the far-left tooth of both combs is transmitted through both devices. In effect, the pair of arrayed waveguide gratings is a very narrow bandpass filter.Figure 2. When two arrayed waveguide gratings with slightly different free spectral ranges are connected, only a frequency that matches one of the teeth on both of the combs is transmitted. Suppose the observation angles for the two gratings are changed. The situation essentially is the same, but the combs have shifted slightly, and so a different frequency will pass through the two gratings.That's the principle behind the efforts at several laboratories to develop a digitally tunable semiconductor laser. A group at Bell Labs in Holmdel, N.J., has integrated two arrayed waveguide gratings and 16 semiconductor optical amplifiers on a single chip (Figure 3). Each pair of semiconductor optical amplifiers -- one from the upper array and one from the lower -- oscillates only at the frequency that can pass through the gratings.Figure 3. A laser produced at Bell Labs promises 56 channels with a spacing of 100 GHz. When semiconductor optical amplifier No. 1 of the top array and amplifier No. 1 of the bottom array are turned on, one wavelength is emitted from the laser. Similarly, when semiconductor optical amplifier No. 1 of the top array and amplifier No. 2 of the bottom array are turned on, a different wavelength is emitted, and so on. An advantage of this is that the number of channels available from the laser scales as the square of the number of semiconductor optical amplifiers, but the device size scales linearly.The Bell Labs researchers designed the laser to operate at 56 channels with a channel spacing of 100 GHz, although they have to date reported experimental results for only eight channels. Another group, from Eindhoven University of Technology and JDS Uniphase, both in Eindhoven, the Netherlands, has integrated eight semiconductor optical amplifiers and a pair of arrayed waveguide gratings on a chip in a design that differs somewhat from the Bell Labs design. This team obtained oscillation at 16 different wavelengths, but the device's long-term stability was questionable.