High-power semiconductor lasers were introduced more than 15 years ago following the invention of quantum-well (QW) layering. QW layering is a process of creating highly efficient active regions for turning electrical signals into photonic signals at room temperature. Applying the QW concept to the semiconductor laser technology of the day yielded high-power output, but the devices were edge-emitting lasers that produced highly multimode, highly divergent beams from the edge of semiconductor chips. Since that time, companies developing edge-emitting lasers have created designs that emit sufficient power and reasonable beam quality for use in transmission and amplification systems in fiber-optic telecommunications.
Ten years ago, researchers developed the electrically pumped, vertical-cavity surface-emitting laser (VCSEL). These devices offer a rounder, less-divergent beam and narrower linewidth compared to edge emitters. A number of technical issues kept VCSELs from emitting more than a few milliwatts of power in a fundamental spatial mode, howeverfor example carrier crowding, in which the injected carriers would tend to crowd in a circular ring as the device diameter exceeded 10 µm. As a result, in optical networking, VCSEL-based transmitters were generally relegated to low-power datacom applications, lacking sufficient power to serve as pump lasers, for example.
Telecommunication systems today are designed around the more practical edge-emitting diode lasers, despite their obvious drawbacks. The divergent, elliptical output beam produced by edge emitters requires special handling not only for the beam and lenses but also for the external fiber Bragg gratings and modulators. A surface-emitting approach with sufficiently high beam quality could reduce the number of subcomponents required to produce a practical laser module. The size of systems could shrink along with the number of subcomponents and the overall costs.
One way to circumvent some of the power limitations of VCSELs is to use an extended cavity design that allows extraction of high power in a fundamental spatial mode from devices with diameters of hundreds of microns. We use an external cavity in our extended cavity surface-emitting laser design. Such a design also removes heat from the gain region efficiently, because the gain region is adjacent to the heat sink. This approach yields a highly manufacturable laser with high-power, narrow-linewidth output, and good beam quality with a near-round profile. These devices are well suited to many laser-based products including fiber-optic amplifiers and transmission systems, as well as displays, specialty lighting, medicine, and sensors.
Figure 1. The large-diameter Novalux extended-cavity surface-emitting laser (NECSEL) contains three mirrors. Two distributed Bragg reflectors (DBRs) sandwich the active region, and a third curved mirror defines the external cavity.extended-cavity design
The extended cavity controls both the optical mode and the wavelength of the devices (see figure 1). The laser emits more than 1 W of output power from a multimode device, more than 500 mW in single-mode operation, and several watts under pulsed operation at 980 nm. By incorporating an extended cavity, the devices combine the attributes of solid-state conventional lasers and semiconductor lasers.
The design incorporates a three-mirror coupled-cavity configuration. Epitaxial distributed Bragg reflector (DBR) layers form two of the mirrors, while the third is a curved external mirror. The gain region and the Bragg mirrors are grown via metal-organic chemical vapor deposition (MOCVD) processing on gallium-arsenide (GaAs) wafers with a number of gallium-indium-arsenide (GaInAs) QWs serving as the active gain medium.
To drive the devices, we inject electrical current through a circular aperture in the p-doped region, defined by a silicon-nitride layer, while the n-contact has an anti-reflection-coated circular aperture to allow the beam to exit to the external mirror. The epitaxial n-mirror helps to reduce the loss from absorption in the substrate and also controls the output wavelength by locking the output wavelength to the Fabry-Perot frequency produced by the two internal Bragg mirrors.
These external-cavity VCSELs can be driven to very high current levels without degradation, with powers limited only by junction temperature. Because of the large aperture, the optical intensity is far below the catastrophic limit (tens of megawatts per square centimeter for edge emitters), and the output beam is near circular for all power levels. For example, peak output power levels for a 300-µm-diameter device could exceed several tens of kilowatts in a short pulse of several nanoseconds. The beam quality, as measured by M2, is roughly 1.2 (a perfectly Gaussian TEM00 beam, by definition, has an M2 of 1). Packaged external-cavity VCSELs can achieve a spectral linewidth of less than 0.01 nm and wavelength stability to within ± 0.1 nm over the case operating temperature range. manufacturability of the design
Figure 2. Map of spectral absorption wavelengths across a wafer shows an absorption dip indicating the Fabry-Perot wavelength at which the laser will oscillate.
Unlike edge emitters, surface-emitting devices can be tested at a wafer level prior to die separation, mounting, and placement in the module package. Wafer-level laser testing allows us to identify good dies, which are then picked and passed along to assembly (see figure 2).
The fiber-to-laser mounting of surface emitters is generally easier than that for edge-emitting diode lasers due to the circular TEM00 beams. This automated and monitored process allows us to greatly reduce cycle times: laser testing occurs only a few days after the GaAs substrates enter the processing line. This rapid cycle time, in turn, allows us to optimize the device manufacturing parameters rapidly.
Figure 3. Temperature modeling of a vertical-cavity laser chip allows developers to optimize designs.
Modeling and simulation are also an integral part of the manufacturing process (see figure 3). It is critical to optimize all designs with a device modeling cycle that gives information on optical, thermal, mechanical, and electrical properties.
The most obvious application for the high-power external-cavity VCSEL design is as a pump device for erbium-doped fiber amplifiers (EDFA). The lasers are most competitive used at powers of more than 500 mW. They have a niche market for use at low power in metropolitan-area-network amplifiers. The devices are also compatible with small, low-power amplifiers designed for single-channel/band systems, which can boost the signal in degraded legacy networks, in optical cross-connects, and loss compensators for 40 Gb/s systems.
With the advent of dense wavelength division multiplexing, the demand for amplification systems has skyrocketed. External-cavity VCSELs offer a method for lowering the size, number of components, and cost of amplifier modules. Next-generation products, which will incorporate fewer optical-electrical-optical conversions, will need signal regeneration and boosting all along the network. Scalable and flexible component platforms, such as the surface emitter described above, are key to lowering the cost of fiber networks and thus expanding their reach within high-volume, cost-sensitive metropolitan networks. oe
oxide-confined VCSELs for 10 gigabit ethernet
Vertical-cavity surface-emitting lasers (VCSELs) have traditionally been fabricated using proton implantation. The use of selective oxidation to produce oxide-confined devices can significantly improve performance.
Because high-energy implantation induces vacancies and areas for non-radiative recombination at the perimeter of the active region, proton implantation results in a planar device structure. The thermal conductivity of the implanted material is reduced by a factor of two compared to non-irradiated material. The device features thermally-induced index guiding that is caused by a gain superposition; this effect varies under different driving conditions.
Selective oxidation, on the other hand, can yield small, single mode devices with diameters below 5 µm and threshold currents under 100 µA. In addition, the technique results in an oxide with a lowered refractive index, which means that the current aperture also induces lateral optical index guiding. The tradeoff for oxide-confined devices is that the contract area and shading of the outcoupling facet have to be accepted.
The maximum output power of the implanted VCSEL with a differential quantum efficiency ηd=30% is limited by thermal rollover. The output power of the oxidized VCSEL (ηd=71%) is three times higher due to higher optical gain, reduced dissipated power, and improved thermal conductivity in the device. The differential resistance of the oxidized device is a factor of nearly two, which is smaller than that of the implanted VCSEL. Maximum wallplug efficiency is ηd=33% in oxidized types and only ηimpl=10% in implanted VCSELs. The reliability of both types of VCSELs are comparable. While the capable speed (modulation bandwidth) of implanted devices is mostly limited to 1.25 Gb/s, oxide confined devices run up to 10 Gb/s per channel, mainly because of higher efficiency.
The low power, circular and low-divergence beam profile, high modulation bandwidth, and scalability of VCSELs make them good sources for data communications. The challenge of increasing the aggregate bandwidth of tomorrow's optical transceivers can be met by increasing the speed of the single channel and/or increasing the numbers of channels with high density. One and two-dimensional VCSEL arrays appear to be key components in reaching the highest aggregate bandwidths of tomorrow's parallel optical transceivers.
While oxide-confined VCSELs offer a good solution for 10Gb/s applications, questions remain regarding high frequency packaging for single channel and arrays. Flip chip technology with bottom emitting devices could help alleviate the high frequency demands. In flip-chip bonding, the VCSEL chip is equipped with solder bumps on top that connect to submounts (or even direct to the driver circuitry) and emit in the opposite direction. Flip chip technology not only facilitates high frequency operation, but also cuts down on assembly time while increasing yield and optimizing heat flow.
Burghard Schneider, CEO, ULM Photonics
Aram Mooradian is chief technical officer of Novalux, Sunnyvale, CA.