Vertical Cavity Lasers Push into the Future
Traditional semiconductor lasers are reliable and small, but VCSELs offer advantages in optical coupling, electrical power usage, and flexible array formatting that keep interest in the field high.
Semiconductor lasers are reliable, small, and easy to operate. The advantages over larger scale lasers are clear, including the advantage in cost, as many manufacturing steps take place on the wafer scale. But there are packaging difficulties associated with traditional laser diodes, which are “edge-emitters,” with the light exiting each laser perpendicular to the direction of the layer structure. The laser cavity is formed by cleaving the end faces of the diodes, and coupling the light out is not necessarily straightforward. Vertical cavity surface emitting lasers (VCSELs) avoid many of the issues associated with edge-emitters.
VCSELs are designed with mirrors at the top and bottom, surrounding a thin active region. Because the active layer is thin, the optical cross-section is small, and the mirrors must be highly reflective—above 90% as opposed to the 30% of the cleaved faces in edge emitters. Early designs included metallic mirrors, but to reduce threshold current, the top and bottom mirrors are now typically composed of alternating layers of semiconductor materials of different indices of refraction, creating a distributed Bragg reflector (DBR) with reflectivity typically around 99%.
A mounted VCSEL chip is ready to be coupled to a fiber, which can directly interface with the top surface of the VCSEL. |
An eye diagram illustrates the cleanliness of an optical communications signal. This open area within the orange lines illustrates a high-quality signal being transmitted through 320 m of fiber with the JDSU Mesa VCSEL operating at 95°C. Images: JDSU |
|
Another step typically taken to reduce threshold current is creating a buried oxide structure in a layer of aluminum gallium arsenide, AlGaAs, which restricts the lateral area through which current will flow, channeling the current through a small active region. These design improvements, along with many others, have enabled VCSEL technology to demonstrate several key characteristics for a variety of applications.
For example, because the light is coupled vertically out of the semiconductor surface, a fiber normal to the surface will couple light out. In addition, because the output apertures are larger than for edge emitters, the divergence angle is smaller. And perhaps one of the most important advantages of the VCSEL approach is the ability to perform wafer-scale testing of the devices. To test an edge emitter, the devices must be diced out of the wafer and some effort must be directed to getting the light out. If there is a problem with the material, process, or design of the device, there is no way to tell without a significant expenditure of time. With wafer scale testing, VCSELs can be addressed prior to investing additional time into preparing each device.
The emission characteristics of VCSELs are uniquely suited to many telecommunications applications, which is one reason why there has been, and continues to be, significant progress towards shifting the emission wavelength into regions around 1300 and 1550 nm. There is a unique challenge in the VCSEL design, because the active region gain window must match the DBR reflectance window, and the materials must also match their lattice constants. Even with all the challenges facing VCSEL development, there continue to be innovative designs.
On the mesa
A “typical” 850-nm VCSEL will have a bottom DBR constructed from a few tens of layers of AlGaAs, then a few GaAs quantum wells, a layer of AlGaAs with 97-98% Al, and a final couple dozen layers of AlGaAs for the top Bragg reflector. The middle AlGaAs layer is modified by a wet oxidation step. Ramana Murty, a development engineer with JDSU, Milpitas, Calif., describes the complications due to wet oxidation: “the oxidation rate of AlxGa1-xAs (with x around 0.98) is extremely sensitive to the Al composition and requires tight control over the epitaxial and process parameters.” Still, it is a required step, because the current must be limited to the optically useful area of the active region to maintain performance with low electrical power consumption. It also introduces non-standard wet chemistry into a standard semiconductor production process. Essentially, this necessary step introduces additional cost and production delays, and potentially decreases the yield as well. But is this step necessary?
At this year’s Optical Fiber Communication Conference (OFC) in Anaheim, Calif., JDSU described their mesa VCSEL technology, which eliminates the lateral oxidation step, making it simpler and more robust for manufacturing. Rather than chemically modifying a region of each VCSEL, the top DBR is physically etched away. The remaining etched mesa provides current confinement, resulting in a device with a low threshold current. The mesa structure also provides photon confinement. The mesa diameter in the VCSEL is 10µm. The typical threshold current and slope efficiency at room temperature are under 1 mA and 0.3 W/A, respectively. All lateral dimensions in the device are defined lithographically with standard integrated circuit fabrication processes, and real-time laser monitoring is used to control the mesa etch depth. Precise treatment of all surfaces and interfaces leads to a long device lifetime that exceeds published data for oxide VCSELs.
Reducing cost and improving reliability are always good actions, but the timing is especially good right now. VCSELs at 850 nm have dominated datacom links over short distances, says Murty. “The combination of on-wafer testing, low-power consumption, and the ability to fabricate 1- and 2-D arrays permits the creation of low-cost, high-bandwidth links critical for the short reach optical interconnects envisioned in the emerging IEEE 100 GbE standard.”
An arrayed solution
The ability to easily create arrays of transmitters is another advantage for VCSEL technology. Fabricating VCSEL transmitters, photodiode receivers, and their associated send and receive electronics opens up the possibilities for unprecedented integration of telecommunications capabilities. In another OFC presentation, IBM, Armonk, N.Y., presented results from a monolithically integrated 16-channel optical transceiver. The chip contained a 4 x 4 array of 985-nm VCSELs. The VCSELs were built on a 250 µm x 350 µm pitch, with a 62.5 µm offset between rows. The arrangement is designed to accommodate optical coupling to a set of arrayed waveguides.
Each of the channels, integrated with transceiver electronics, is capable of modulation at speeds of better than 10 Gb/sec. Together, the integrated package demonstrated transmission of optical data at rates greater than 160 Gb/sec. Not only is the single-channel data rate unprecedented, but the key figures of merit for power dissipation (15.6 mW/Gb/s) and density (9.4 Gb/mm2) are also unprecedented. IBM is developing this transceiver as part of a Defense Advanced Research Projects Agency (DARPA)-sponsored chip-to-chip program designed to speed up communications between supercomputers.
Stretching it out
VCSEL technology is most advanced at 850 nm. That's the wavelength achieved with GaAs quantum wells confined with AlGaAs DBR layers. However, there is great interest in extending to other wavelengths of interest for optical telecommunications, specifically 1300 and 1550 nm. For these wavelengths, other materials must be selected, with special care given to addressing the lattice mismatch issues.
Although VCSELs have their own set of unique technological challenges, the advantages of low power requirements, straightforward optical coupling, flexible array architecture, and on-chip testing all are keeping the field energetic.
—Richard Gaughan
Gaughan is founder and Chief Engineer at www.mountainoptical.com
|
