Etched facets set to cut the cost of diode arrays

The Los Angeles-based firm, which was founded in November 2000, has demonstrated prototype semiconductor chips that emit up to 100 W in continuous-wave mode from monolithic arrays containing 66 surface-emitting elements. However, the firm believes that the technology is scalable and could ultimately produce chips containing hundreds of elements that emit a total power of 1kW or more.

QPC hopes to have something on the market soon, according to its chief executive officer Jeffrey Ungar. "We plan to have products at a prototype level by the end of the year, with commercial products following soon after," said Ungar. "Initial products will be replacements for diode stacks - say a few hundred watts for pumping thin-disc lasers or conventional YAG lasers, for example. However, prices will be somewhat less than what people are paying right now for stacks."

The development could be an important step forward in the creation of cost-effective and practical kilowatt solid-state lasers, which are sought by the military and commercial sectors. The US military, in particular, is keen to develop solid-state lasers in the 100 kW to megawatt range and with today s diode technology that is simply not feasible due to their high cost and poor reliability.

"The economics of today s diode pumps in the rack and stack configurations are such that they make it totally impractical to have a field weapon at these sorts of power levels," explained Ungar.

"If you want a 100 kW solid-state laser then you need about 250-300 kW of diode power to pump it. With today s style of diode arrays you re looking at several millions of dollars for just one of these weapons."

Currently, high-power diode arrays are usually made from a stack of edge-emitting discrete bars - a strip of semiconductor wafer containing several individual laser diodes - which each emit a few tens of watts.

Although stacks can achieve total output powers ranging from a few hundred watts to a few kilowatts by combining the output from many bars, they are tricky to package and cool. Each bar has to be cleaved, individually cooled and electrically connected to its neighbors. As a result, the stacks tend to be expensive to scale to higher powers.

High-power stacks are also vulnerable to reliability problems, especially a cooling failure. Each bar, or layer of the stack, has to be connected to a microchannel cooler that pumps water through a series of miniature channels, each just tens of microns wide. The problem is that these channels can clog up or corrode.

"If you imagine a stack with 50-100 W per layer then you re looking at having hundreds of layers [bars] to create a multikilowatt pump and then the likelihood of a failure is very high," commented Ungar. "To a large extent the reliability issues are not specifically semiconductor issues, but are in many cases due to the cooling and mounting technologies."

QPC seems to have come up with an ingenious solution to the problem. When making its wafer of edge-emitting laser diodes, it etches a series of 45° mirrors next to the diodes to reflect their light out of the surface of the wafer. As a result, 2D surface-emitting arrays of laser diodes can be made without the need for cleaving the wafer into a series of discrete bars.

The beauty of QPC s monolithic design is that devices are potentially much easier to package and cool. This should in turn mean that the technology is very cost competitive when it hits the market. Instead of having to mount a cooler to each cleaved diode bar, with QPC s design a single more powerful cooler can be attached to a substrate of the entire chip. This cooler removes heat from all of the emitters simultaneously.

"The cost benefits come through the simplicity of the cooling and the packaging. Instead of, say, a 20-layer [bar] stack with 20 microchannel coolers, you have a single semiconductor chip with a single cooler," explained Ungar. "We re striving to make large arrays where the price target for a packaged array is $1-2 per watt."

The firm s latest test results were unveiled at the recent meeting on diode lasers in Albuquerque, New Mexico, US (2004 solid-state and diode-laser technology review).

To date, QPC s demonstration chips are based on several rows of 22 surface-emitting elements and emit light at around 980 nm. At the event, QPC announced a 1 x 22 array that produced 50 W at a drive current of about 110 A and a 3 x 22 array that generated 100 W at a drive current of about 220 A. Typically each emitting element is square, 100 µm long and emits between 1 and 2 W, although higher powers may be possible

"We have also made a 10-row device [10 x 22] containing 220 emitters, but that has not yet operated in lasing mode," said Ungar. "At the moment it s mounted on a piece of copper and we re waiting for a custom-made water cooler to support the high power densities that it will generate."

However, with appropriate cooling it is likely that the device could generate 300-400 W of optical power, a figure comparable to many stacks.

In all of QPC s devices, light from edge-emitting laser elements is flipped by 90° so that it comes out from the top of the wafer. Light leaves the AlGaAs active region parallel to the wafer surface and is then bent upwards by 90° using an etched 45° mirror that acts as a total internal reflector (see figure).

The laser s cavity mirrors are formed by two distributed Bragg reflectors (DBRs) made from thin layers of varying refractive index. One of these is placed after the 45° mirror and also serves as an output coupler letting light out of the wafer.

The use of DBR mirrors also helps to stabilize the emission spectrum from the laser, and experiments indicate that the beam from a 980 nm array can have a full-width half maximum (FWHM) value of just 0.6 nm.

As with conventional edge-emitting lasers, the beam leaving each element of the array is highly non-symmetric with a typical divergence of 35° (FWHM) in one dimension and 5° (FWHM) in the other. This causes the beams from neighboring emitters along a row to overlap and gives the emission an appearance of stripes.

"The critical component in all of this is being able to make a perfect 45° mirror," explained Ungar. "The idea of using such a reflector has occurred to others but was never a practical proposition. The mirror surface must be absolutely perfect, if it s rough or out by a degree then you re sunk."

Although QPC s etching process is proprietary and Ungar was reluctant to go into detail, the firm appears to have refined a wet-etch process so that it can reproducibly fabricate a 45° total internal reflector that has a beautifully smooth optical surface. It claims that the orientation of its etched reflector is accurate to within 0.5°.

In effect, the new type of laser combines the power of edge-emitting laser technology with the easy processing of vertical-cavity surface-emitting lasers (VCSELs).

According to Ungar, the advantage of the new design is that the output power of each surface emitter can be scaled to several watts per emitter instead of just a few milliwatts, which is the typical limit for a VCSEL.

"The VCSEL design is completely inappropriate for high powers because its geometry has high thermal and electrical impedance," said Ungar. "It s a fine structure for producing a milliwatt or so at 800 nm, but it s a difficult proposition to get higher powers."

At the same time, QPC s design maintains the cost-effective on-wafer testing that is often touted as a benefit of surface emitters such as VCSELs. In effect, the emitting elements can be tested on the wafer with no need for any cleaving of individual devices.

In fact it was searching for an alternative to long-wavelength VCSELs that first inspired QPC s research in the area.

"Our initial thought was that this would be a great way to make 1300-1550 nm vertical emitters where one could get much higher powers than with a VCSEL," explained Ungar. "However, we soon realized that this design was probably even more appropriate for high-power pumping."

Before founding QPC, Ungar and several members of his technical team were working at Ortel, a firm specializing in lasers for telecom applications. Ortel was purchased by Lucent for $3 billion at the height of the telecom boom in 2000 and Ungar left to form QPC in November of the same year.

However, another start-up in the US is now pursuing QPC s initial ideas for a long-wavelength VCSEL alternative. Binoptics of Ithaca, US, has developed an InP 1310 nm laser based on a ridge waveguide with an etched 45° mirror at one end. The surface-emitting device produces 20 mW of optical power and is aimed at telecom applications.