Tunable Lasers

Despite the economic woes of the communications sector, some technologies can still stir great interest on Wall Street, as well as in the laboratory. Among the more exciting is the tunable laser. In recent years, several companies have entered this field, with some of them garnering impressive rounds of funding.

A couple of the more successful start-ups are Agility Communications Inc., Santa Barbara, Calif., which received a total of over US $168 million, and Bandwidth9, Fremont, Calif., which received some $110 million in funding over the last two years. Both have already announced products for the tunable laser market.

At the same time, some of the largest optical communications companies have begun investing heavily in the technology, both by funding in-house R&D and through acquisitions. In May 2000, for example, ADC, Eden Prairie, Minn., spent $872 million to buy Altitun AB, Kista, Sweden, one of the first companies to offer a tunable laser product for telecommunication applications. One month later, in June, Nortel Networks Corp., Brampton, Ont., Canada, acquired Coretek Inc., the Wilmington, Mass., company, for around $1.43 billion.

Why all the excitement? Replacing 80 or 160 units with one or a few laser types simplifies inventory management. Tunable lasers also make it easy to add or delete bandwidth by remote control--with no need to dispatch a service technician--thereby opening the door to a variety of new on-demand services.

Down with inventory

Today, single fiber-optic strands carry multiple wavelengths of infrared radiation across entire continents, with each wavelength channel carrying digital data at high bit-rates. Known as wavelength-division multiplexing (WDM), this process greatly expands the capacity of fiber-optic communications systems, making them one of the most important parts of the foundation on which the Internet rests.

Currently, WDM transponders, which include the laser, modulator, receiver, and associated electronics, incorporate fixed lasers operating in the near-infrared spectrum, at around 1550 nm. A 176-wavelength system uses one laser per wavelength, and must store 176 additional transponders as spares to deal with failures. These devices therefore account for a high percentage of total component costs in an optical network. Their prices vary, depending on their reach and data rate, but can go as high as $20 000 for a long-haul OC-192 (10-Gb/s) unit. In other words, the bill for maintaining and managing the inventory for WDM system spares can be steep, especially for high-speed systems with lots of channels.

Tunable lasers offer an alternative. A single tunable laser module can serve as a backup for multiple channels, so that fewer transponders need to be stocked as spares. The result: cost savings and simplification of the entire sparing process, including inventory management.

Prices for fixed and tunable lasers are not yet equivalent, however. Although some tunable types are priced like fixed-wavelength devices, they are tunable over only very narrow ranges, about 3-4 nm. Those that can be tuned across wide wavelength ranges remain at least two to three times as expensive as their fixed counterparts, because of the increased complexity of manufacturing them, the extra testing required, and the newness of the technology, which has yet to reach true volume demand.

As demand for tunables rises, their prices will come down. Ultimately, manufacturers claim, the price premium for a widely tunable laser will drop to about 15-20 percent above that of a fixed laser. With such a slight cost differential, it is easy to demonstrate the cost savings of using widely tunable lasers in place of fixed lasers in WDM networks.

The flexibility factor

While applications in inventory reduction will drive much of the initial demand for tunable lasers, the real revolution will come when they are applied to make optical networks more flexible.

Fiber-optic networks today are essentially fixed: the optical fibers are connected into pipes with huge capacity but little reconfigurability. It is well-nigh impossible to change how that capacity is deployed in real time. Part of the problem is the difficulty of choosing a wavelength for a channel: as traffic is routed through a network, certain wavelengths may be already in use across certain links. Tunable lasers will ease a switch to alternative channels without swapping hardware or reconfiguring network resources.

Tunable lasers can also provide flexibility at multiplexing locations, where wavelengths are added to and dropped from fibers, by letting carriers remotely reconfigure added channels as needed. Such lasers can help carriers more effectively manage wavelengths throughout a network, based on different customer requirements. The benefits gained are a far greater degree of flexibility in provisioning bandwidth and a reduction in the time it takes to actually deliver new services.

What's in a laser

The word laser is an acronym for light amplification by the stimulated emission of radiation, the process by which laser light is generated. From campfires to lasers, all light-emitting devices spit out photons when electrons drop from a higher to a lower energy state. Each photon emitted has an energy equal to the difference between the higher and lower energy levels.

When the process occurs naturally, it is called spontaneous emission. Photons that are spontaneously emitted can interact with high-energy electrons, triggering the electrons' transition to a lower-energy level, a process known as stimulated emission. A laser is created when a material with abnormally many high-energy electrons, called a gain medium, is placed between two reflective surfaces (typically a pair of mirrors), creating an optical cavity. Acting as a resonator for the photons that are emitted, the cavity reflects them back and forth within the gain medium, stimulating further laser emission and amplifying the optical signal.

The trick, of course, is to create the gain medium. Early lasers cast a ruby rod in that role. Consisting of crystalline aluminum oxide doped with about 0.05 percent chromium, ruby can absorb electromagnetic energy and retain it for about 4 ms, remaining in a semi-stable, excited (high-energy) state. The energy for those lasers came from optical "pumps"--powerful flash tubes wrapped around the ruby rods [see drawing, page 1].

More modern semiconductor lasers are usually pumped electrically, by an applied bias voltage. The material and pumping method are different, but the result is the same: the establishment of a gain region with what is known as a population inversion, abnormally many electrons in high-energy states.

Most lasers in today's optical communications networks are semiconductors--laser diodes. Their gain material is usually a compound of elements from columns III and V of the periodic table, like indium phosphide or gallium arsenide. Typically they operate in the 1310-nm or 1550-nm region of the spectrum. Their wavelength is determined both by the properties of the III-V gain medium and by the physical structure of the laser cavity surrounding the gain medium.

To form cavities in semiconductor lasers, a variety of techniques are used. One involves cleaving and polishing the edges of the semiconductor material itself to form mirrors (called facets). Another employs mirrors or diffraction gratings external to the semiconductor. (A diffraction grating is a collection of reflecting elements separated from one another by a distance comparable to the wavelength of the laser. In essence, the grating acts like a wavelength-selective mirror, reflecting only a specific wavelength in a desired direction.)

A laser's wavelength is determined by its optical cavity, or resonator. Like an organ pipe, it resonates at a wavelength determined by two parameters: its length--the distance between the mirrors--and the speed of light within the gain medium that fills the cavity. Accordingly, the wavelength of a semiconductor laser can be varied either by mechanically adjusting the cavity length or by changing the refractive index of the gain medium. The second approach is most easily done by changing the temperature of the medium or injecting current into it.

Recently a wide range of tunable lasers have emerged in the 1550-nm region of the infrared for use in WDM optical communication systems. There are four types: distributed feedback, distributed Bragg reflector, external cavity diode, and vertical-cavity surface-emitting lasers.

All but the last are based on edge-emitting devices, which emit light at the substrate edges rather than at the surface of the laser diode chip. Vertical-cavity structures do the opposite.

The distributed feedback laser

Among the most common diode lasers used in telecommunications today are distributed feedback (DFB) lasers. They are unique in that they incorporate a diffraction grating directly into the laser chip itself, usually along the length of the active layer (the gain medium) [see drawing, right]. As used in DFB lasers, the grating reflects a single wavelength back into the cavity, forcing a single resonant mode within the laser, and producing a stable, very narrow-bandwidth output.

DFB lasers are tuned by controlling the temperature of the laser diode cavity. Because a large temperature difference is required to tune across only a few nanometers, the tuning range of a single DFB laser cavity is limited to a small range of wavelengths, typically under 5 nm. DFB lasers with wide tuning ranges therefore incorporate multiple laser cavities.

One laser producer, Fujitsu Ltd., Tokyo, has developed a four-channel tunable DFB laser, which has been deployed in operational networks. More recently, the company announced a 22-channel device. The four-channel device has one cavity, which changes of temperature can tune to four standard communications wavelengths spaced 0.8 nm (100 GHz) apart.

The 22-channel device is fabricated using multiple cavities. Each of the cavities is connected by a waveguide to a semiconductor optical amplifier, which boosts the laser output power to 10-20 mW. Control electronics are used to select and operate one of the multiple cavities and to adjust the temperature of the cavity, depending on the desired wavelength. The electronics used for stability control and wavelength selection are integrated on a single silicon substrate with the laser diode chip itself.

Other companies that have developed tunable DFB lasers include Nortel Networks and Agere Systems, Allentown, Pa., formerly the Microelectronics Division of Lucent Technologies.

A variation of the DFB laser is the distributed Bragg reflector (DBR) laser. It operates in a similar manner except that the grating, instead of being etched into the gain medium, is positioned outside the active region of the cavity. Lasing occurs between two grating mirrors or between a grating mirror and a cleaved facet of the semiconductor.

Tunable DBR lasers are made up of a gain section, a mirror (grating) section, and a phase section, the last of which creates an adjustable phase shift between the gain material and the reflector. Tuning is accomplished by injecting current into the phase and mirror sections, which changes the carrier density in those sections, thereby changing their refractive index.

The tuning range in a standard DBR laser seldom exceeds about 10 nm. But wider tuning ranges can be achieved using a specialized grating, called a sampled grating, which incorporates periodically spaced blank areas. A tunable sampled-grating DBR (SG-DBR), for instance, uses two such gratings with different blank area spacing. During tuning, the gratings are adjusted so that the resonant wavelengths of each grating are matched. The difference in blank spacings of each grating means that only a single wavelength can be tuned at any one time.

Since tuning with this sampled-grating technique is not continuous, the circuitry for controlling the multiple sections is far more complex than for a standard DFB laser. Also, the output power is typically less than 10 mW. On the plus side is the SG-DBR laser's wide tuning range. Agility Communications has announced a 4-mW SG-DBR laser capable of tuning from 1525 to 1565 nm--enough to span 50 channels at the standard channel spacing of 0.8 nm.

Another variation of the DBR laser is a grating-assisted codirectional coupler with rear sampled reflector. Patented by Altitun, ADC's Swedish acquisition, the structure uses a three-section DBR tunable across 40 channels, from 1529 to 1561 nm. Several other companies also at work on different variations of the tunable DBR laser include Agere Systems, Alcatel, JDS Uniphase, Marconi, and NTT/NEL.

The external cavity diode laser

The third edge-emitting tunable version is the external cavity diode laser (ECDL). It uses a conventional laser chip and one or two mirrors, external to the chip, to reflect light back into the laser cavity. To tune the laser output, a wavelength-selective component, such as a grating or prism, is adjusted in a way that produces the desired wavelength.

This type of tuning involves physically moving the wavelength-selective element. One ECDL implementation, for example, is the Littman-Metcalf external cavity laser, which uses a diffraction grating and a movable reflector [see illustration, right]. ECDLs can achieve wide tuning ranges (greater than 40 nm), although the tuning speed is fairly low--it can take tens of milliseconds to change wavelengths. External cavity lasers are widely used in optical test and measurement equipment.

ECDLs are attractive for some applications because they are capable of very high output powers and extremely narrow spectral widths over a broad range of wavelengths. Whether they will prove cost-effective in telecommunications applications remains to be seen. Still, last year New Focus Inc., in San Jose, Calif., introduced an external cavity diode laser for such applications. The fairly high-power (20-mW) device can tune across 40 nm (50 channels). It includes a wavelength locker, power control, and control electronics. Iolon, also in San Jose, and Blue Sky Research Inc., in Milpitas, Calif., are also developing ECDLs.

The vertical-cavity diode laser

The alternative to edge-emitting lasers is the vertical-cavity surface-emitting laser (VCSEL). Rather than incorporating the resonator mirrors at the edges of the device, the mirrors in a VCSEL are located on the top and bottom of the semiconductor material. This setup causes the light to resonate vertically in the laser chip, so that laser light is emitted through the top of the device, rather than through the side. As a result, VCSELs emit much more nearly circular beams than edge-emitting lasers do. What's more, the beams do not diverge as rapidly. These benefits enable VCSELs to be coupled to optical fibers more easily and efficiently.

Since fabricating VCSELs requires only a single process growth phase, manufacturing them is much simpler than producing edge emitters. VCSEL manufacturers can also exploit wafer-stage testing, thus eliminating defective devices early in the manufacturing process, saving time, and improving overall component manufacturing yields. (Edge-emitting lasers cannot be tested until the wafer is separated into individual dice because only then do the light-emitting edges become accessible.) Because of these features, VCSEL chips can be produced far less expensively than edge-emitting lasers.

Unfortunately for VCSEL manufacturers, the dominant cost of a telecommunications laser is not the chip but the package that houses it. According to Tim Day, chief technology officer at New Focus, laser chips themselves account for no more than 30 percent of the cost. Most of the rest goes for the precision-machined hermetic package in which the chips are mounted.

Another plus is that VCSELs need less power and can be directly modulated at relatively high speeds--up to 10 Gb/s. With no need for an external modulator, direct modulation leads to simpler drive circuitry and lower-cost transmitter modules.

While VCSELs outdo the edge-emitters in many respects, they do have a weak spot: their inability to generate a lot of optical power. Because the beam in a VCSEL traverses the thin dimension of the wafer--typically less than 500 µm--it gets to interact with only a thin layer of gain medium, and therefore can build up only a little power. Edge emitters, in contrast, are limited by wafer diameter, usually more than 100 mm across. Thus, today VCSELs are used mostly in enterprise data communications applications that run at 850 nm. Optical output power for 1550-nm tunable VCSELs is just a fraction of a milliwatt, whereas many of the standard 1550-nm edge-emitting lasers now used in telecommunications deliver 10-20 mW.

Still, VCSELs operating at 850 nm have proven ideal in short-reach applications--in buildings, say--where low output power is not an issue. Several companies are currently working on commercializing VCSELs at 1310 and 1550 nm for telecommunications networks.

Tuning with micromechanical elements

In tuning VCSELs, the technique used is based on mechanical modification of the laser cavity using microelectromechanical systems (MEMS) technology. With MEMS, a movable mirror can be fabricated at one end of the laser cavity. This approach enables VCSEL/MEMS devices to achieve a relatively wide tuning range--preliminary specifications from manufacturers quote tuning ranges of 28-32 nm, enough to cover 35-40 channels at the standard 0.8-nm channel spacing.

One concern with using MEMS is that their long-term mechanical reliability has yet to be proved. So before these devices are incorporated into actual telecommunications systems, they must pass stringent reliability testing by Telcordia Technologies Inc., Morristown, N.J. Many tunable laser manufacturers are now involved in these reliability tests, both at their own labs and in trials at networking systems manufacturers.

To boost a VCSEL's optical output power, some manufacturers are including an optical pump source (typically a laser diode at a slightly lower wavelength). Nortel Networks and Princeton Optronics Inc., Princeton, N.J., use this approach to produce devices capable of 10-20 mW of output power in the 1550-nm range. Using pump lasers, though, makes the laser module more complex, increases power requirements, and raises costs.

The California start-up, Bandwidth9, is currently developing a tunable VCSEL laser using MEMS as the tuning element. But without an optical pump, the laser is capable of producing only 100-200 µW of output power.

Higher output powers are possible. Bandwidth9 claims to have exceeded 1 mW in the laboratory with a device fabricated with an integrated MEMS-based cantilever arm [see illustration, top of this page]. The arm is used to adjust the length of the laser cavity and thus tune the output wavelength. The company told IEEE Spectrum that it will focus on metro networking applications, where distances are rarely more than 200 km and the power requirements are much lower than those for long-haul applications.

Will tunable lasers revolutionize optical networks? With many of the technologies just now becoming commercially available, it is still too early to say. What is evident is that tunable lasers can dramatically improve network efficiency and will play an important role in enabling future dynamically reconfigurable optical networks, along with optical switches [see "In Search of Transparent Networks," Spectrum, October 2001, pp. 47-51] and semiconductor optical amplifiers. One recent advance especially worth keeping an eye on is the work done by some manufacturers in integrating laser diodes with other functional elements, such as the wavelength locker, modulators, and optical amplifiers--all on a single chip.

—Michael J. Riezenman, Editor

Acknowledgments

The author thanks the following people for their contributions to this article: Ray Moyer, formerly of Fujitsu Network Communications; Arlon Martin, Agility Communications; Tim Day, New Focus; and Charles Duvall, Bandwidth9.