Rings speed wavelength switching

Widely tunable lasers are key components in optical networks based on wavelength division multiplexing. They can be used to replace many fixed-wavelength lasers as back-up sources, so long as they can deliver a tuning range in excess of 40 nm and a wavelength switching speed of typically 1 s.

The best designs that meet these requirements use current injection into the semiconductor waveguide to tune the emission wavelength. Since a single-wavelength control region limits tuning to 15 nm, due to the maximum shift in refractive index that can be induced in this type of structure, many of today s commercial offerings feature filter-based structures.

Such devices, which can cover an entire optical communication band, employ a pair of filters with comb-like reflectivity spectra in differing free spectral ranges (FSRs). Several commercial lasers are based on this technology, including JDSU s sampled grating (SG) DBR laser, Syntune s Y-branched emitter, and a super-structure-grating distributed Bragg reflector (SSG-DBR) design manufactured by our subsidiary, NTT Electronics Corporation.

Although these lasers can serve today s optical communication networks, they are not necessarily suitable for next-generation infrastructure, which will be based on optical packet-switching systems. In these systems, wavelength switching will be performed through an arrayed waveguide grating (AWG) filter, with the tunable laser s wavelengths at the input port dictating the routing/switching of data to the desired output port. For such an approach, rapid switching and tuning of the lasing wavelength is demanded on a packet-by-packet basis.

Tunable lasers operating in this type of network also need to deliver a very stable emission wavelength – even very small frequency drifts can produce cross-talk into the neighboring output ports of the AWG filter. These drifts can occur over relatively long time-intervals and are often attributed to variations in injection current during tuning. These variations alter the device s temperature, leading to changes in refractive index and lasing wavelength.

We have recently invented and demonstrated a widely tunable laser that can fulfill all of these criteria (figure 1). This device consists of a gain section, a phase-control region and a double-ring-resonator-connected filter with two micro-ring resonators. The ring resonators are simple circular waveguide structures with several transmission peaks that can be tuned by varying the injection current. Each of these ring resonators, which are tens of microns in size, is coupled to a straight waveguide via a multi-mode interferometer coupler.

Our micro-ring-based filter is a small, simple structure that reduces overall device costs and increases production yield. In addition, it allows our lasers to be fabricated with the same set of processes that are used to make our other components for a wavelength-routing device. The two steps required to produce DBR-type lasers are also avoided – the regrowth process for burying the gratings and any form of high-resolution lithography.

Our ring resonator has superior spectral characteristics compared with SSG and SG lasers, and it features a narrow bandwidth filter with a Lorentzian-shaped response and an infinite number of resonant peaks with the same reflectivity (figure 2a). These characteristics are ideal for tunable lasers because they lead to a stable lasing mode and a high side-mode suppression ratio. In comparison, the grating-type filters that are employed in other tunable laser designs often have fewer than 10 peaks with similar reflectivity. Although that number can be raised by increasing the coupling coefficient of the grating, a penalty of higher bandwidth results, which reduces the laser s stability and side-mode suppression ratio (figure 2b).

We have carried out calculations that consider the effects of filter response on laser performance (see box "Designing the filters"). This has led to a design featuring two ring-resonators with FSR responses of 483 and 520 GHz. This corresponds to first and second ring radii of 13.6 and 11.9 µm, respectively.

These micro-rings form the key part of our monolithic 1.0 × 0.3 mm laser, which contains a high-reflection-coated back facet, an as-cleaved front facet and a stack-layer structure that enables fabrication with a single regrowth step. In this structure an active region – consisting of a multiple quantum well layer sandwiched between upper and lower separate confinement heterostructure layers – is grown on top of a 0.3 µm thick InGaAsP layer (λg=1.40 µm).

The gain and phase control sections of our micro-ring tunable laser are 400 and 200 µm long, respectively. They feature a ridge waveguide structure 1.6 µm wide and 2.0 µm high. The ring resonators contain a 1.2 µm wide, 4 µm high deep-ridge waveguide that is formed using chlorine-based reactive ion etching in an inductively coupled plasma.

This laser s emission wavelength can be switched within 3 ns and it delivers a tuning range in excess of 50 nm. The wavelength is selected by varying the second ring s injection current from 0 to 5.2 mA (figure 6). When no current is applied to either of the ring resonators and the injection current in the gain region is set to 94 mA, lasing occurs at 1533 nm.

Increasing the current in the second ring reduces the refractive index in its waveguide, leading to longer emission wavelengths. Increasing this current produces a range of lasing emission wavelengths with a fixed wavelength difference, and side-mode suppression ratios of more than 30 dB. The device s actual emission wavelength depends on the FSR of the first ring, because lasing occurs at the resonance wavelength of the two ring-resonators. The longest emission wavelength of 1572 nm occurred at 3.2 mA. At 3.7 mA the emission jumped to 1522 nm.

The very low currents required for wavelength switching improve our laser s wavelength stability. These variations are less than 5 GHz (figure 7).

These results show the importance of reducing injection current to improve the laser s speed and stability associated with switching emission wavelength. The results are encouraging, but our calculations show that better performance is possible if we fabricate a laser with an FSR of 200 GHz and a maximum refractive index change of 0.1%.