Access Official Technical PDF

Share Summary:LinkedInEmail

InP-on-silicon tunable lasers offer a route to flexible optical transmitters by combining light generation in indium phosphide (InP) with passive waveguide functions on silicon. For engineers and technical decision-makers evaluating optical network architectures, the key design question is how to balance wavelength range, switching speed, and laser performance. Three demonstrated designs—a distributed Bragg reflector (DBR) laser, a distributed feedback (DFB) laser, and an arrayed waveguide grating (AWG)-based filtered-feedback laser—illustrate distinct ways to address that challenge.

The work is relevant to optical interconnects in and between data centers, where traffic patterns can make rapid wavelength changes useful. The reported devices are research-stage demonstrations, not a single design that already combines the widest tuning range with the fastest switching.

Why fast wavelength tuning matters

Optical packet switching and other flexible interconnect approaches can require transmitters that change wavelength quickly. Fixed wavelengths may not suit every switching or routing arrangement, while thermally tuned lasers can be too slow for applications requiring nanosecond-scale changes. Electronic tuning is attractive because it can be fast, but implementing it in silicon photonics has a materials challenge: the free-carrier plasma dispersion effect used to change a waveguide’s effective refractive index is weaker in silicon than in common InP-based compounds.

InP-on-silicon integration brings together InP-based active regions and silicon-on-insulator (SOI) waveguides. This supports laser designs that can connect with passive silicon photonic components. The three approaches below show how current injection and optical filtering can produce different forms of tunability.

InP-on-silicon tunable lasers: three design approaches

Three-section DBR: broad continuous tuning, with heating effects

The three-section DBR laser uses an InP/InGaAsP membrane bonded to an SOI waveguide circuit. Its sections perform different roles: the gain section supplies optical gain, while phase and Bragg sections provide tuning. Current injection into the tuning layer changes the waveguide mode’s effective index through both free-carrier effects and Joule heating. These effects shift the Bragg wavelength in opposite directions, so their relative influence matters to tuning behavior.

In the reported device, the threshold current was 55 mA, and maximum continuous-wave on-chip output power was 0.4 mW. Single-mode operation was measured at 1565 nm, with a side-mode suppression ratio (SMSR) of 32 dB. By adjusting the phase and Bragg currents together, the laser accessed wavelengths from 1565 to 1577 nm, a continuous range of 12 nm. SMSR stayed above 30 dB across that range, and output power varied within 2 dB. The observed wavelength shift was dominated by heating, rather than by the desired fast electronic effect. The work therefore identified changes to the epitaxial layer design as a potential way to strengthen electronic tuning.

DFB laser: electronic tuning across a narrower range

The second design is a tunable DFB laser that also uses a twin-guide membrane, but with a 200 nm tuning layer intended to increase optical confinement there. Preliminary measurements showed a 45 mA threshold current at room temperature and a maximum fiber-coupled output power of 0.1 mW for a device with a 600 µm grating.

Its tuning range was about 2.5 nm, with the measured wavelength shift indicating a net electronic tuning effect. That range was sufficient to address four channels on a dense wavelength-division multiplexing (DWDM) grid with 0.8 nm spacing. Compared with the DBR device, this result showed electronic tuning, but over a more limited wavelength span.

AWG filtered feedback: discrete, sub-nanosecond switching

The third architecture uses an AWG as a wavelength-selective filter outside the main Fabry-Pérot laser cavity. The AWG routes light into four branches, each containing an InP-on-silicon semiconductor optical amplifier (SOA) that acts as a gate. When a gate is biased at or above transparency, light can pass through it; when unbiased, it absorbs incident light. Reflectors at the branch ends return selected light to the laser, encouraging single-mode operation near the AWG channel with the greatest transmission.

This filtered-feedback arrangement provides discrete channel selection and does not use the more complex twin-guide tuning layer. The demonstrated laser produced four channels spaced 4 nm apart, each with an SMSR above 35 dB. At room temperature, its threshold current was 20 mA; maximum fiber-coupled output power was 0.1 mW, corresponding to 4 mW waveguide-coupled power. Continuous-wave operation was reported up to 65 °C.

For dynamic testing, the device was modulated with non-return-to-zero on-off keying (NRZ-OOK). Open eye diagrams were measured up to 12.5 Gbit/s across the four channels. The reported bit-error rate was below 3.8 × 10-3, a level stated to support error-free operation with 7% overhead hard-decision forward error correction. In a separate channel-switching test, alternating gate bias produced a rise time of 195 ps and a fall time of 190 ps—both under a nanosecond.

Engineering trade-offs and next steps

The results are not an apples-to-apples comparison: the devices differ in architecture, output-power measurement, tuning mechanism, and operating conditions. Still, they reveal useful design trade-offs. The DBR structure demonstrated the broadest continuous tuning range, but heating dominated its measured shift. The DFB structure showed an electronic shift across a narrower range. The AWG filtered-feedback laser demonstrated fast switching between discrete channels, rather than continuous tuning.

The next development goal described for these devices is to combine wide tuning range with fast switching and strong static and dynamic laser characteristics. For system designers, the practical takeaway is to match the laser architecture to the network’s requirements: continuous wavelength coverage and rapid channel selection are different capabilities, and no single result here demonstrates both at their maximum.

Conclusion

InP-on-silicon tunable lasers show how heterogeneous integration can support several approaches to wavelength agility in optical interconnects. The DBR, DFB, and AWG filtered-feedback designs each address a different part of the challenge, from broad continuous tuning to sub-nanosecond discrete switching. Further device optimization is needed to unite those strengths, but the demonstrations establish relevant design directions for future flexible optical networks.

Access Official Technical PDF

About The Author