In order to shift services from one wavelength to another without burdening a line card with thousands of semiconductor lasers, optical-transport OEMs needed a light source that could shift with ease from one frequency band to another. That need was met through the development of a key semiconductor-based component, the tunable laser.

This vision of "tunability" as the enabler of the fully connected optical mesh drove development programs at leading component manufacturers such as Agere Systems, Fujitsu Ltd. and Agilent Technologies, and also spurred venture financing for a host of startups, including Agility Communications Inc., Santur Corp., Bandwidth9 Inc. and Iolon Inc. In 2001 and 2002, some of the first true tunable lasers moved from prototypes to early production. Concurrent with the availability of components, however, came a downturn in network equipment demand, the likes of which the telecommunications industry had never seen. With large long-distance carriers declaring bankruptcy and small competitive carriers disappearing daily, the hopes of IP-over-fiber became a dream deferred. So does that make the tunable photonic component a solution to a problem that no longer exists? Some market analysts have even concluded that deteriorting market conditions means there is no immediate need for the tunable light source.

But this is not exactly the case, for these lasers can be valuably employed in applications that exist today, such as the replacement of large inventories of single-frequency spare lasers with generic spares in dense wave-division multiplexing (DWDM), equipment. And because a tunable laser embedded on a DWDM line card can allocate a service to a wavelength without requiring careful manual tuning of a service to a fixed frequency, a new customer for a service provider can be provided a service in a matter of minutes, usually without the carrier sending out a dedicated service truck — an application called remote provisioning. Also, every optical node in a carrier network can utilize generic spares when a laser goes out. If the laser was not tunable, the service provider or the OEM would have to provide field service personnel with scores of fixed lasers tuned to different frequencies.

But there's no doubt that the downturn hurts across the optical equipment industry, necessitating cutbacks in staff and marketing at virtually all tunable laser and filter companies. "There are customers today for certain applications," said Arlon Miller, vice president of marketing at Agility. "Nevertheless, caution means everything in this environment. We raised the cash to take us through mid-2005, but we have been very cautious on how to spend it."

The tunable laser is a derivative of the general-purpose semiconductor laser, based on a diode manufactured with III-V compounds such as gallium arsenide (GaAs) or indium phosphide (InP). In principle both the fixed-wavelength and the tunable semiconductor laser operate with an electronic current "pumping" a gain-medium material to create a gain region in which many electrons are in high-energy states. Tunable lasers also use many of the same basic process and architectural structures developed for fixed-frequency semiconductor lasers. The difference in the tunable laser is that an additional structure is utilized to adjust the frequency range of the light emitted. With the cavity in which coherent light resonates doubling as the microstructure where tuning takes place, various techniques of adjusting imprinted gratings within the cavity can change the frequency of emitted light. Alternatively, light can be adjusted externally to the laser source, using micromachined elements such as micromirrors or actuators. To understand how the tunable laser works, a better understanding of a fixed-frequency laser is necessary.

Waveguides and mirrors
By alternating refractive indices in layers above and below the active layer, in a so-called double-heterojunction architecture, the semiconductor laser can confine light in a stripe geometry that acts like a waveguide to send light along the chip in a single waveguide mode. The length of the cavity determines the wavelength of the semiconductor laser.

The addition of mirrors at both ends of the waveguide makes the semiconductor laser a resonator. In traditional semiconductor lasers, the mirrors are made from edges of the semiconductor wafer, and the resulting diode is known as an edge-emitting laser. The simplest form of resonant-cavity laser is called Fabry-Perot, referring to a device originally designed by Charles Fabry and Alfred Perot.

In both fixed and tunable worlds, a laser beam can be emitted from either the edge or surface of a semiconductor die. The vertical cavity surface-emitting laser, or VCSEL, has been favored in its fixed-frequency form in low-cost LAN applications, because arrays of lasers can be manufactured on a wafer more cheaply than edge-emitters. In these devices, the mirror reflectors are created directly in the layers of materials used in the wafers, structured to obtain an effect similar to a fiber grating. Bandwith9 Inc. has been in the forefront of applying VCSEL economies of scale to a tunable device (see "VCSELs can bring optics to the masses," page 151).

The simplest form of edge-emitting tunable laser was developed from a popular form of fixed-frequency laser called distributed feedback, or DFB. In this design, regularly spaced ridges are etched at the bottom of the semiconductor substrate, and this regular grating pattern operates like a more complex version of Fabry-Perot mirrors. The only way to tune a standard DFB laser is by varying the temperature, which reduces the range of tunability, and makes for slow tuning. However, because the thermally tuned DFB laser was available well before more complex types, it has been used in simpler DWDM provisioning architectures.

A newer derivative of the standard DFB design, used by Agility Communications, is the distributed Bragg reflector (DBR), in which the grating is etched in a separate portion of the chip from the portion in which the electronic drive current "pumps" the laser. The reflection principle is similar to that used in a fiber Bragg grating, hence the name distributed Bragg reflector. In Agility's version of DBR, the grating alternates with blank areas, creating what is called a sampled-grating distributed Bragg reflector. This requires complex electronic control circuitry, but allows for wide tuning ranges, some 50 nanometers (nm) in Agility's first-generation device.

Agere has made more use of silicon optical bench integration technologies in its DBR designs. In work originally launched at Lucent Bell Labs and developed at Agere's former Breinigsville, Pa. facility, the company used an electro-absorption modulator integrated with a two-stage DBR laser and a semiconductor optical amp. The device uses a dual-waveguide structure in an InGaAsP layer to achieve 10-Gbit/second transmission over 82 km of fiber.

In some designs, such as those from Iolon and Intel Corp. (a design acquired from New Focus Communications), the edge-emitting chip is placed in a special external cavity that selects the wavelength. The reflective grating is in a different part of the hybrid device, and often is moveable via microelectromechanical systems-based (MEMS) technologies. Iolon uses deep reactive ion etching to fabricate an electrostatic comb-drive actuator to tune its laser chip. The actuator operates along a semicircular arc to actually change the waveguide channel along which light is emitted.

One advantage an external-cavity laser could realize is a better ability to tune to a variety of frequencies without experiencing drift, even in DWDM environments using tight channel spacing. The International Telecommunication Union has specified a standard 100-GHz channel spacing for DWDM, though a 25-GHz spacing has been on the drawing boards for at least two years. At last spring's Optical Fibers in Communications conference, Iolon claimed its new Apollo laser could support 25-GHz spacing, which would allow a single device to access 200 channels in the C band, and 200 channels in the L band.

While DFB, DBR, and external-cavity designs all are important in tunable applications, all three architectures were first used in fixed-frequency designs. In fact, Fujitsu Ltd. and Agere both entered the tunable market based on popular production-level DFB designs that were later modified for tunable applications.

The need for tunable devices arose in the mid-1990s, as DWDM equipment from vendors such as Ciena and Pirelli became important for increasing capacity of fibers without upgrading fibers to the next highest Sonet rate. When WDM equipment could only combine a handful of channels based on the color of the signal, multiplexing those wavelengths was not cost-effective. But when new equipment vendors could offer long-haul carriers a way to combine tens or even hundreds of channels in one transmission with extremely tight grid spacings, DWDM equipment sales increased exponentially.

However, the popularity of the equipment brought about a problem. Since DWDM equipment required scores of lasers, carriers needed plenty of spares on hand. And since many channels would operate on different frequencies, it would make sense to keep large inventories of a single basic semiconductor laser design, which could then be tuned to the application required. This concept gave rise to the first and simplest DFB-based tunable lasers. These lasers can be more cost-competitive as compared to fixed-frequency equivalents, but often are tunable over a range of only 5 nm or less.

The new generation of tunables is responding to more sophisticated applications, but not all of them involve active wavelength translation. Miller of Agility said that there is a second, simple but important DWDM application that takes carriers a step further than sparing. In order to relieve customers of the slow provisioning associated with services based on Sonet virtual circuits, carriers began offering "one-time provisioning" services, in which a tunable laser in DWDM equipment could be used to bring up a specific service that a customer would demand on a semipermanent basis. Indeed, Agility is depending to a certain extent on such simple provisioning applications to carry the company through the lean years.

Miller predicted that tunable-laser manufacturers also will find a near-term market in "remote provisioning," in which a carrier uses wavelength-selectable equipment to reprovision a service to a remote customer, either a new customer or one at a remote site who wishes to change a service. This kind of application can be justified easily, even by a carrier in financial straits, since it reduces operational costs. Miller said that successful use of remote provisioning would "replace many truck rolls with mouse rolls" — in other words, the truck of a service call is replaced with the click of a mouse from a remote site.

However, the target most companies hoped for when launching their companies was the all-optical mesh network, requiring wavelength translation in optical add-drop multiplexers and optical cross-connect switches. The original idea driving the shift from Sonet rings to meshes was to associate particular services with specific wavelengths. Then, if a video service was carried in a particular flow of IP packets, that service could be shifted from wavelength A to wavelength B if it made the most economic use of a particular optical channel. By contrast, Sonet rings use optical transport, but require services to be dedicated to particular channels, which often can make very inefficient use of bandwidth.

Originally, asynchronous transfer mode switching was supposed to provide the simplest way to carry differentiated services in the electrical domain, which then could be segmented out to different optical tranport networks. But when ATM failed to catch on as a unified network from user to long-distance carrier, developers looked for ways to shrink the protocol stack. In "reduced protocol-stack" equipment, intervening provisioning layers such as ATM or Sonet, were removed, allowing services to be provisioned directly and immediately from the IP routing layer, translated to wavelength switching.

As an example, a popular form of defining IP flows in a pure electronic routing environment debuted in the mid-1990s, called multi-protocol label switching, or MPLS. A group of researchers at WorldCom/UUnet working under Daniel Awduche developed an extension of MPLS for the optical domain, which was later standardized as generalized-MPLS, or G-MPLS.

This protocol would allow a system with tunable lasers to instantly associate a particular video conference, or a particular voice-over IP (VOPI) call, with a specific wavelength in an optically routed network. Many such advanced concepts in dynamic wavelength provisioning were pioneered between 1999 and 2000 by companies that either no longer exist, or have moved into deep hibernation.

"There's no denying that a lot more work was being done in dynamic provisioning a couple years ago," said Charles Duvall, applications consultant with Bandwidth9. "If those scenarios had come to pass, tunable lasers could have been used at every port."

But executives at Agility, Bandwidth9 and Iolon agree that the industry had gotten ahead of itself, in terms of prototyping equipment that carriers might not feel comfortable using, for applications that end customers might not know they needed. While the success of such equipment could have led to a huge boom (and possible shortages and allocation) in tunable lasers, many in the industry are somewhat grateful for the breather, in that dynamic provisioning may now get a chance to have bugs shaken out in early academic and industrial prototypes, before being taken directly to market.

Even if the loftier ideas of all-optical provisioning and long-haul mesh networks are postponed, incumbent carriers can make use of dynamic provisioning as a means of cobbling together super-regional and even national networks that never make use of an interexchange carrier.

Some architectures being touted by network equipment providers would allow wavelength translation over large supermetro fiber rings (Sonet or otherwise) that could let incumbent local-exchange carrier coalitions build nationwide networks that bypass the IXC. In the aftermath of the WorldCom scandal, this could prove appealing to ILECs in the short term.

"True dynamic provisioning could come through the ILECs in a shorter time span than anticipated," Miller of Agility said. "With consolidation, many of the ILECs are getting big enough that they don't want to send their traffic to someone like MCI. You could see tunability move into some of these interconnection points."

As dynamic-provisioning applications become a reality, integration of multiple devices on a common substrate could determine the future of tunable components. Laser manufacturers tend to see III-V materials, particularly GaAs or InP, as the best substrates for integration. Developers working with silicon-based passive devices, such as NanoOpto Corp., argue that it is often more cost-effective to integrate III-V laser chips on a silicon substrate for lower manufacturing and packaging costs.

Duvall of Bandwidth9 said that the use of a VCSEL architecture for a tunable laser will allow integration of other devices alongside VCSEL arrays, with a modulator being particularly desirable for the future. Miller of Agility pointed out that his company has worked with UC Santa Barbara on development of advanced modulators, and that Agility considers "photonic integration to be the most important thing we could be doing, in both the short and the long term."

From whence support?
The issue then becomes how to support substrate integration at a time of industry contraction. Nortel Networks, for example, paid $1.4 billion for tunable-laser specialist Coretek Inc. in 2000, but laid off the bulk of the Coretek staff in early 2002 as part of across-the-board reductions in optoelectronic component markets. Agere remains a leader in integration at its New Jersey facilities, though the company closed down its separate Breinigsville, Pa., operation.

Among the startups, both the tunable-laser specialists, as well as substrate-integration companies like Bookham Technologies and NanoOpto, recognize the value of joint efforts. Yet those efforts may have to come about through ad hoc alliances, as few companies have the available capital to consider the type of mergers and acquisitions that were common only two to three years ago.

Duvall said that most laser specialists recognize that their primary realm of expertise is in growing the lasers on III-V substrates, and in coming up with innovative laser designs. Many other integration and packaging functions can be outsourced. Making new strides on the integration front will be possible, he said, but only if a clear OEM customer base is identified for longer-term integrated applications.

If there's a great concept out there for something that may not have a true application for five years, many companies will be too cautious."