Long-wavelength VCSELs are slow to 'materialize'
Several companies have put their 1310-nm VCSEL development on the back burner or completely abandoned it—not surprising given today's market climate in which many companies' resources are drying up or are being devoted to the technologies generating revenue. Gore, an early proponent of long-wavelength VCSELs, sold its 850-nm VCSEL assets to Optical Communications Products (OCP) in January. OCP had previously purchased Cielo Communications' 1310-nm VCSEL assets last October. Other companies are rumored to have cut back on similar efforts.
Whether the market sensation that 850-nm VCSELs created as a low-cost alternative for data communications links will be repeated when the long-wavelength technology finally reaches fruition remains uncertain. Unlike edge-emitting semiconductor lasers, VCSELs emit photons from the surface of the wafer, creating a circular beam that couples to the fiber more efficiently. With thousands grown on a single wafer, VCSELs promise to lower testing, manufacturing, and packaging costs relative to edge-emitting devices. But 1310-nm edge-emitting devices have a significant headstart in the market.
"I think the commercialization of 1310-nm VCSELs is slower than people expected, mainly due to a materials problem," says Yong-Hang Zhang, chief executive of Lytek (Phoenix), a startup working on the technology. "This year, we are seeing a lot of progress from Lytek as well as from Agilent and Honeywell, but it is still at the R&D level. To make these devices really, really manufacturable, and also for higher-end applications, it may take a bit longer."
The semiconductor materials commonly used in shorter-wavelength 850-nm VCSELs, aluminum gallium arsenide and gallium arsenide, are harder to work with at the longer wavelengths. Many companies pursuing long-wavelength VCSELs are adding nitride to the materials' structure.
Lytek is taking a different approach by working with antimonide-based material. "We believe that the antimonide has better manufacturing, better uniformity, and we expect a better lifetime," says Zhang, who is also a professor at Arizona State University. Founded in May 2000, Lytek obtained a license to commercialize technology developed by Zhang and others at Arizona State. Intel Capital was one of the early investors in the company. Lytek expects to deliver sample devices to select customers this quarter.
The materials structure is not the only challenge that developers face. "A lot of the problems have to do with the two different approaches right now, the laser pump and the electrical pump," explains Bob Pierce, vice president of marketing at E2O Communications (Calabasas, CA). "Inherently with the laser pump, you can get the power out, but you have a more sophisticated assembly, and then mirror technology and materials that you have to overcome. With the electrical pump, you can get the frequency responses, but you can't get the necessary power out of these devices."
Founded in 1998, E2O Communications has been developing laser-pump technology for several years. The company's long-wavelength VCSEL technology will serve as the base for a 10-Gbit/sec transceiver and transponder product family for 10-Gigabit Ethernet and OC-192 applications. These products will be offered in 300-pin transponder, XENPAK, and XGP form factors. Work remains on the long-wavelength VCSEL technology, however.
"In essence, we have been able to get the power out and some of the alignment and some of the mirror technology straightened out," says Pierce. "And we have lasing operation in the entire industrial temperature range right now, which is another problem faced by developers. We are reasonably ready to show some technology demonstrators this year."
Although progress is being made, as evidenced by several post deadline papers presented at Photonics West in January, 1310-nm VCSEL devices may not reach the market in 2003, and when the technology is finally ready for commercialization, the market itself may present barriers to acceptance of the technology. "Our take on it is that they are pursuing a moving target at 1310 nm," says Tom Hausken, director of the Optical Communications Components Practice at market researcher Strategies Unlimited (Mountain View, CA). "They have technological barriers to overcome, but by the time they do, there is still the marketing hurdle. They have to provide a cost advantage, and right now the advantage that they could provide is so small that the [edge-emitter] competitors will underprice them and keep them out."
For this reason, Hausken suggests that many companies are hedging their bets. "Agilent uses 850-nm VCSELs, and OCP gets Fabry-Perots from a third party," he says. "For these little startups, that's not the case, but for some of these other companies, they are realizing now that they better have a backup."
If the suppliers of long-wavelength VCSELs can solve the technical issues and are able to seize substantial market share from competing edge-emitters, as much as $150 million worth of transceivers could be sold in 2007 with long-wavelength VCSELs inside, according to Hausken. This number is more than all the shipments of high-speed 1310-nm transceivers in 2002, but it would still leave a substantial market for longer-reach lasers.
Meantime, Photodigm (Richardson, TX) is developing 1310-nm grating surface-emitting (GSE) laser technology based on aluminum indium gallium arsenide/indium phosphide (AlInGaAs/InP), a proven edge-emitting material structure. Distributed Bragg reflector gratings are on each end of the device. In the center is an active ridge guide where the light is generated in a horizontal plane. Between the ridge guide and one of the gratings is a second-order grating where a portion of the light is emitted vertically on each pass of the horizontal beam.
"In effect, it acts like a VCSEL," says Jack Mattis, vice president of business development, "with the performance of a DFB [distributed-feedback] laser." The GSE laser offers singlemode output up to 6 mW, according to Mattis. The technology is in the alpha sampling stage.