DFB laser array

A DFB laser array is a row of DFB lasers made side by side on one InP chip, each with its own grating so that each emits at a different wavelength. Arrays are used in two ways. In a wavelength-selectable array only one laser runs at a time, and temperature tuning of that laser covers the gap to its neighbors, so the chip as a whole works as a widely tunable laser. In a multi-wavelength array all the lasers run at once, each carrying its own channel, and an on-chip multiplexer combines them into one output for wavelength-division multiplexing.

Setting the wavelengths

Each laser emits close to the Bragg wavelength of its grating,

$${\lambda }_B=2n_{\text{eff}}\mathrm{\Lambda },$$

where $\mathrm{\Lambda }$ is the grating period and $n_{\text{eff}}$ the effective index of the laser waveguide, so the wavelength can be stepped from laser to laser by changing either. At 1550 nm with $n_{\text{eff}}\approx 3.2$ the period is about 242 nm, and moving the wavelength by one 100 GHz DWDM channel, 0.8 nm, takes a period change of only 0.125 nm; gratings with steps this fine are written by electron-beam lithography. The other approach keeps a single grating period and steps the width of each laser's ridge, which changes $n_{\text{eff}}$.

No laser lands exactly on its design wavelength, because $n_{\text{eff}}$ depends on layer thicknesses and ridge widths that vary across a wafer. Each laser is brought onto its channel by temperature, which moves a DFB by about 0.09 nm/K. A thermoelectric cooler under the chip moves all the lasers together; correcting them one by one takes a heater beside each laser, or a change in its drive current, which also heats it.

Wavelength-selectable arrays

A temperature range of 35 K tunes a DFB over about 3 nm, so about a dozen lasers spaced 3 nm apart cover the 35 nm of the C-band: the choice of laser selects a 3 nm range, and the temperature selects the wavelength within it. Within each range the tuning is continuous and the laser keeps the single-mode spectrum of a DFB; moving between ranges means switching to another laser.

The light from whichever laser is on has to reach one fiber. An N-to-1 MMI coupler joins the waveguides from all the lasers but passes only $1\mathrm{/}N$ of the power of any one of them, a loss of $10{\log }_{10}N$ dB, or 10.8 dB for 12 lasers. The coupler is commonly followed by an integrated semiconductor optical amplifier that makes up the loss, as in the chip of Kudo et al., which also integrated an electro-absorption modulator. The combining loss can be avoided altogether by selecting the laser optically: in the module of Pezeshki et al., a MEMS mirror steers the beam of the chosen laser into the fiber.

Selectable arrays are one of the established ways of making a widely tunable telecom laser. The others tune a single cavity, by the Vernier effect between two sampled-grating mirrors in a DBR laser or with a tunable filter in an external cavity.

Multi-wavelength arrays

With every laser running, an MMI combiner would cost $10{\log }_{10}N$ dB on every channel, 9.0 dB for eight. A wavelength-selective combiner avoids that loss: an arrayed waveguide grating routes each input wavelength to the common output, so each channel sees only the AWG's insertion loss, which does not grow with $N$ the way the MMI's loss does. The lasers must then sit within the AWG's passbands, and temperature control holds the two in step. Nagarajan et al. described a transmitter chip of this kind with ten DFB lasers, each followed by its own electro-absorption modulator, and an AWG multiplexer, sending 10 × 10 Gb/s into one fiber.

Channel spacings follow the WDM grids of the CWDM and DWDM standards: 100 GHz (0.8 nm) or 50 GHz near 1550 nm for DWDM, and 20 nm for CWDM. A continuous-wave laser for each wavelength, discrete or in an array, is also one of the candidate light sources for co-packaged optics, alongside comb lasers; Co-packaged optics light sources compares them.

Yield

Every laser in an array has to work, and has to land within tuning reach of its channel, for the chip to be usable. If each laser is good with probability $p$, an array of $N$ lasers is good with probability $p^N$; at $p=0.95$ that is 81% for 4 lasers, 66% for 8 and 44% for 16. One remedy is redundancy, with more lasers than channels and the working ones chosen at test. The alternative to an array is a set of separate lasers combined by an external multiplexer, which lets each laser be tested and replaced on its own at the cost of more parts, more alignment and a larger package.

Common questions

How is a DFB laser array tuned?

In two steps: selecting a laser picks a range of a few nanometers, and the temperature of the chip, or of that laser, sets the wavelength within the range. Within a range the tuning is continuous.

How does a DFB array differ from a sampled-grating DBR laser?

A DFB array holds many separate lasers, each covering a small range, with one running at a time. A sampled-grating DBR laser is a single cavity whose two mirrors reflect at combs of wavelengths with slightly different spacings; tuning the currents in the mirrors picks the wavelength at which the two combs coincide, and a phase section fine-tunes the cavity mode. Across the full range the output moves in steps between cavity modes, which is called quasi-continuous tuning.

References: K. Kudo et al., "1.55-µm wavelength-selectable microarray DFB-LD's with monolithically integrated MMI combiner, SOA, and EA-modulator," IEEE Photonics Technology Letters (2000). B. Pezeshki et al., "20-mW widely tunable laser module using DFB array and MEMS selection," IEEE Photonics Technology Letters (2002). R. Nagarajan et al., "Large-scale photonic integrated circuits," IEEE Journal of Selected Topics in Quantum Electronics (2005). L. A. Coldren et al., "Tunable semiconductor lasers: a tutorial," Journal of Lightwave Technology 22, 193 (2004). J. Buus, M.-C. Amann and D. J. Blumenthal, Tunable Laser Diodes and Related Optical Sources, 2nd ed. (Wiley-IEEE Press, 2005). L. A. Coldren, S. W. Corzine and M. L. Mašanović, Diode Lasers and Photonic Integrated Circuits, 2nd ed. (Wiley, 2012).