Efficiency Drive Delivers Greater Laser Reliability
According to Strategies Unlimited analyst Bob Steele, the market for diode lasers used in industrial applications dipped 13% in 2004, with rising unit shipments more than offset by declining prices. Despite this, the industrial market for diodes is widely regarded as a good long-term prospect, with German company Jenoptik Diode Lab now building a 500 m2 facility for 3 inch GaAs wafer production in Berlin. And the acquisition of diode-pumped-laser-system maker Lightwave Electronics by JDS Uniphase in March this year demonstrates the latter s commitment to becoming a bigger player on the industrial-laser scene.
Major industrial-laser OEMs such as Germany s Trumpf and Rofin-Sinar are increasingly looking to improve their products, which are used in applications as diverse as shaping medical devices and welding car-body parts. Predictably, these laser-system builders want more power from diodes and a reduction in the price per watt delivered. The cost issue is particularly sensitive, as it s estimated that the diodes in a laser system account for one-third of the overall system cost.
More watts per dollar
One way to improve the dollar-per-watt metric is to raise the efficiency of individual emitters. As well as increasing the overall power output of any laser system based on diode-pumped crystals or direct-diode emission, it also improves reliability.
In fact, this added benefit is arguably of greater importance than increased power output. This is because improvements in diode efficiency mean that less heat is produced per device. And excess heat is a killer for laser diodes. Toby Strite, the manager of high-power diode-laser marketing at JDS Uniphase, says that as a rule of thumb the lifetime of a diode is cut by two-thirds for every 20° temperature increase in the diode junction.
So while there has been an understandable focus on achievements made in pushing diode power conversion efficiency (PCE) to new highs in excess of 65% from many companies, especially those working under DARPA s super-high-efficiency diode sources (SHEDS) program, it perhaps makes more sense to look at the progress made in a different way: "We were wasting 50% [of the power input] in heat, and now we re only wasting 35%," is how Strite likes to think about it. "The relative [reduction] in waste-heat generation outstrips the PCE enhancement."
Under the SHEDS program, JDS Uniphase has taken two approaches to improve diode performance, both of which Strite says can be "productized". The first is in the management of the optical field, where the aim is to reduce the impact of dopant absorption. This can be done by changing the vertical structure of the diode - either depressing the mode downward into the wafer, or shifting it up toward the wafer surface - using optical engineering. In JDS Uniphase s developmental diodes, the center of the laser mode is also directed away from the most highly absorbing regions of the doped semiconductor.
Some of those refinements to diode design are now flowing back into JDS Uniphase s product pipeline, with prototyping currently in progress. Once commercialized, the devices will be manufactured at the firm s one remaining compound semiconductor wafer fab in San Jose, CA, before being shipped out along with telecom lasers for packaging in the company s China facility.
JDS Uniphase s "L-3" diode laser, which emits 5 W into 100 μm core, 0.22 numerical aperture fiber, has been shipping from China for a year. A 6.5 W version of the device is set to be released next month, with 8 and 10 W emitters to follow.
Made in China
A "made in China" stamp on JDS Uniphase s products presents an obvious advantage. "We ve actually commoditized the industrial [diode] laser market by virtue of that low-cost structure," said Strite. The diode-laser demand from telecom applications is sufficient to support the Chinese packaging base, and the industrial side of JDS Uniphase s business can exploit the same manufacturing infrastructure.
While JDS Uniphase is best known for its telecom products, Strite believes that the relative maturation of this industry means that it is now the industrial-laser business that offers more scope for growth. "Essentially, there s no longer a demand for a bigger, better 980 nm pump laser," he explained. "It s a little bit like Intel s problem, in that [once it had made] a 1 GHz chip, there wasn t really a need for a 2 GHz one, so they went into things like wireless connectivity instead."
According to Strite, the design of JDS Uniphase s single emitters provides the ideal combination of power coupled with reliability. Reliability requirements in the telecom market are demanding, and with innovations such as passivated mirrors, JDS Uniphase s devices come rated with a mean time to failure of anywhere between 100,000 and 2 million hours, depending on the application.
JDS Uniphase has its own fiber laser business and this type of laser design complements the company s high-power single emitters. "Imagine a 100 W fiber laser," said Strite. "This will require about 200 W of diode power. To do that with bars, you need four bars rated at 50 W each. If one of those bars fails you re a dead duck, so you ve got to add some redundancy to the system with extra bars."
But simply adding one extra bar will not be sufficient, explained Strite: "A typical bar is rated at 20,000 hours mean time to failure. If you have five in your system, you haven t guaranteed that the system will run for five years continuously. So you need six or seven bars, and that s a huge incremental expense."
The advantage with ultrareliable single emitters, says Strite, is that you d deploy perhaps 50 5 W L-3s to power a 100 W fiber laser, and that 20% added system-redundancy cost is affordable. "That s why we re really making some hay in the low-to-mid-power solid-state-laser market," said Strite. "Some day a fiber laser producing 4 kW will be out there and then you will be able to enjoy bar economies of scale."
Clearly, concentrating on the fiber-laser market plays to JDS Uniphase s strengths and experience in the fiber-optic communications sector, where it is the leading manufacturer of 980 nm pump lasers. However, Strite says that the other markets also present the firm with an opportunity. "We could bring to market a bar that s four times as powerful [as a competitor s product] and are currently in discussion with customers willing to pay only twice as much for the technology."
Alfalight readies new processes
Like JDS Uniphase, Alfalight is now looking at how to implement advances in diode structure in its commercial products. Through incremental improvements, Alfalight s best research effort has produced a 976 nm device with 73% PCE. According to the company s vice-president of research and development Manoj Kanskar, the SHEDS-inspired reductions in diode turn-on voltage, free-carrier absorption and Joule-effect heating will be implemented in the production environment first, since they are not expected to have any negative effect on device reliability.
Secondary improvements will follow once sufficient lifetime-testing data have been collected. So far, Alfalight says it has demonstrated 797, 808 and 915 nm laser bars with a peak PCE of 66% by introducing the initial changes to its production process.
However, the SHEDS program calls for the demonstration of diodes with 80% PCE. JDS Uniphase and Alfalight are taking different approaches to meeting the target. Strite believes that JDS Uniphase can get close to the 80% figure with practical, commercializable technology. Meanwhile, Kanskar is of the opinion that while Alfalight could push a little higher than its current best of 73% with similar efforts, some revolutionary technology will be needed to hit 80%. Critically, Kanskar says, reducing the laser threshold current will require a radical approach.
That revolutionary technology could come in the form of quantum-dot structures or by growing quantum wells in a different crystal orientation, both of which Alfalight is working on. "Quantum dots is a high-risk, high-pay-off method," admitted Kanskar. "But by growing the quantum wells in the 110 direction, where the electric field is in line with the dipole moment, we expect a two-fold enhancement in gain, and a reduced [lasing] threshold."
The problem with current quantum-dot gain media, says Kanskar, is that these structures require 7-10 layers of nanostructures to produce sufficient gain, and the overall diode efficiency suffers from a high lasing threshold. "Our goal is to make a single-layer [quantum-dot] gain medium with a much higher density and uniformity, and that s what we re working on to reduce the threshold current."
So far, Kanskar and colleagues have made quantum-dot lasers in aluminum-free high-bandgap material, without incorporating the wetting layer that most developers have used.
In the laboratory, Kanskar and colleagues have also grown quantum wells in the 110 orientation, and he says that the photoluminescence of these devices is much higher than that of conventional diodes. "The challenge is that the growth temperature required is significantly different," he said.
A separate development that Alfalight has been working on under a US Air Force Research Laboratory project could lead to precision pumping applications of diode lasers, and also have benefits in terms of wafer yields. The goal is to wavelength-stabilize a multimode diode laser, and the company has begun implementing a new process to do this. By using holographically defined gratings and wavelength-dependent feedback, the wavelength drift is effectively limited to less than 0.065 nm/°C instead of the usual 0.32 nm/°C. This means that lasing can be tuned to within ± 0.2 nm, and so a larger proportion of the die on the wafer can be used.
Currently, a wavelength tolerance of around ± 3 nm is required to get a high device yield, but with customers increasingly demanding tighter tolerances, that yield can take a big hit. "I think we re going to see some huge benefits," said Kanskar, who adds that because the holography step is a wafer-level process, the per-device added cost of incorporating the extra manufacturing step is tiny.