Different faces fit for GaN researchers

The locations for the last two nitride conferences couldn t have been more different. A year ago delegates found themselves in the heart of the world s gambling capital, Las Vegas. But this year it was the turn of quiet, picturesque Montreux, which sits in the Swiss Alps on the eastern shore of Lake Geneva.

Any attendees with an interest in mountaineering will have enjoyed looking up at the snow-capped peaks as they walked along the shoreline. And it s possible that they would have asked themselves: what is the easiest route to the top? OK, that face is smooth, but it s covered in snow and prone to avalanche, so the more rugged alternative might actually be the better bet.

The question of the most promising face is not just an issue for those heading off in search of the summit. It s also a central question for researchers developing GaN light emitters, who may have the choice of growing on either non-polar, semi-polar or more conventional polar substrates (figure 1).

Rohm of Japan and the University of California, Santa Barbara (UCSB), are currently leading the development of light-emitting structures on the semi- and non-polar faces, and delegates at the International Workshop on Nitride Semiconductors were given an overview of the latter s efforts by Steven DenBaars. He began by reminding the audience that the benefits of moving to these planes include a substantial reduction in the internal electric fields that pull apart the electrons and holes in conventional nitride devices. This produces a blue-shift in emission that increases with current density. With semi-polar and non-polar structures, wavelength shifts with current are far smaller, and the performance can be improved through the introduction of far thicker quantum wells. "But LEDs still suffer from droop," warned DenBaars, who didn t elaborate on the potential causes of the decline in external quantum efficiency (EQE).

UCSB researchers have been developing non-polar devices for several years, but early progress was held back by r-plane sapphire substrates that led to a very high density of stacking faults in the nitride epilayers. A switch to Mitsubishi Chemical s GaN substrates overcame that hurdle and opened the door to a range of high-performance devices.

DenBaars team has made good progress with its violet lasers over the last two years. The latest designs avoid the need for the problematic AlGaN cladding layer, which causes cracking through tensile strain, a higher voltage operation and reduced yield. Mirror quality has also improved thanks to a switch to focused ion beam etching, which has helped to cut the pulsed threshold current density from 4.6 kA/cm2 to 2.2 kA/cm2.

Rohm and UCSB want to increase the operating wavelength of their emitters, and DenBaars explained that his group was having a lot of success on the semi-polar plane. This provided a better basis than its non-polar cousin for growing the highly indium-rich layers needed to push emission out to the green and yellow, but didn t suffer from the large blue-shift that plagues conventional polar c-plane devices.

In a separate talk, Anurag Tyagi from UCSB detailed recent efforts relating to green-laser development on the (1122) plane. His device is similar to a c-plane laser, but incorporates a thick p-type AlGaN layer rather than a GaN/AlGaN p-type superlattice. Pumping this laser structure with a titanium-sapphire laser revealed a stimulated emission peak at 514 nm.

Even-longer wavelengths have been reached with LEDs on (1122) substrates, according to another UCSB speaker, Hitoshi Sato. He told delegates about a device with a 3.5 nm thick InGaN quantum well that produced 563 nm emission. Under pulsed operation with a 10% duty cycle, output powers and EQEs at 20 and 200 mA were 5.9 mW and 13.4%, and 29.2 mW and 6.4%, respectively.

GaN boule growth
Although these results are very encouraging, the tiny pieces of GaN used as a growth platform are clearly unsuitable for high-volume manufacturing. However, Mitsubishi Chemical s Toshinari Fujimori said that 18 mm × 18 mm pieces should be available in the near future. And more importantly, the company expects to develop 2 inch material by 2010.

Hitting that goal would be a tremendous achievement, but production costs would still prevent non-polar substrates from being used for LED manufacturing. However, Steve Hersee from the University of New Mexico is pioneering a novel method for non-polar LED fabrication on polar substrates.

Hersee s trick is a scalable, uniform process that promises to build LED arrays based on nanowires. He s developed an interferometric process to create the template for these nanocolumns, which can be 0.5–20 µm long and 90 nm to 0.7 µm in diameter.

The big attraction with these nanowires is the absence of any defects, despite growth on typical GaN-on-sapphire surfaces, which have defect densities of 109–1010 cm–2. Some future goals include the production of highly n-doped columns that are coated with an additional active region and p-type layers to make an LED array.

The leading LED manufacturers continue to advance the performance of polar devices. Osram Opto Semiconductors, Cree and Philips Lumileds all took the opportunity to unveil their personal bests.

Nathan Gardner from Lumileds unveiled the Californian chip maker s hero results in a talk that focused on the origins of droop – the reduction in EQE at higher drive currents. Interestingly, Gardner did not reveal any detail about the active region, even though his presentation detailed the benefits of a thicker well that combats droop through a reduction in the carrier density.

Lumileds best results in the lab for a 1 mm × 1 mm white-emitting chip with a color temperature of 4600 K are 140 lm/W at 350 mA, 112 lm/W at 1 A and 100 lm/W at 1.5 A. These are all driven with 10 µs pulses and a duty factor of 1%. Cranking the current up to 2 A increased output to 562 lm, which Gardner claimed to be comparable to the output of a 40 W incandescent bulb.

Cree s James Ibbetson decided to focus his talk on the performance of the company s best LED products. Many of the company s 1 mm × 1 mm chips are housed in the 7 × 9 mm XR-E package and put into a performance bin matched to the lumen output at 350 mA. The top bin contains 107–114 lm LEDs that are based on 450–460 nm blue-emitting chips delivering 500–540 mW. This equates to an EQE of 54%, which falls by 14% as the current is ramped to 1 A.

Ibbetson also mentioned the company s current lab record – a 141 lm white LED, which featured a 620 mW blue chip. This LED suffered from less droop than the company s commercial LEDs, but Ibbetson admitted that this partly resulted from the use of a larger chip with undisclosed dimensions, which cuts the current density.

Johannes Baur revealed that Osram s best lab result at 350 mA is slightly better than Cree s – the German chip maker can hit 643 mW. Baur also detailed the results of another 1 mm × 1 mm developmental chip with an output of 601 mW at this drive current, which can form the backbone of a 155 lm output white LED. This blue emitter delivers 1.97 W at 1.4 A and 3.2 W at 3 A.

Non-polar nitrides on sapphire are being used as the basis for the latest ultraviolet (UV) LEDs that are targeting polymer curing, and air, water and food purification. Vinod Adivarahan from Nitek, a spin-off of Asif Khan s group at the University of South Carolina, revealed a new contact geometry that can boost device output. He explained that the effects of current crowding can be minimized by switching from a traditional single contact to an array of independent pixels.

A UV LED with this design that features an array of 10 × 10 contacts, each 20–30 µm in diameter, operates at 5.2–5.5 V at 20 mA. Light output power saturates at a higher value than standard LEDs and lifetime "looks good", according to Adivarahan.

The company has also built a 3 mm × 3 mm chip that features several 10 × 10 arrays of contacts and delivers 35–37 mW at 800 mA. "Lifetime tests on this LED are underway," said Adivarahan.

Sensor Electronic Technology, an earlier offshoot of the University of South Carolina, also detailed its latest advances in chip design. Yuri Bilenko said that the firm has been developing LED lamps that can incorporate 90 individual chips. These require massive heat sinks, but can deliver a continuous output of more than 80 mW at 275 nm and a drive current of 1.4 A. "265–285 nm lamps will be delivered to customers in their hundreds," claimed Bilenko.

The week-long conference also included several reports of developments of nitride materials and devices on the cheapest of all platforms – silicon. Sudhiranjan Tripathy from Singapore s Institute of Materials Research and Engineering is pioneering a novel twist on this theme, by switching the growth platform to a silicon-on-insulator (SOI) wafer. This has one major advantage over silicon for blue and green light emitters – any light entering the substrate is reflected rather than absorbed.

Tripathy s team has produced crack-free epiwafers on 6 inch SOI substrates and he sees no reason why this process can t be scaled up to 8 and even 12 inch material. His team has produced 0.3 mm × 0.3 mm LEDs with a 5-period InGaN/GaN multi-quantum well that emits at 525–530 nm, but delivers only 1.2 mW at 20 mA. Tripathy admits that his devices are not all that bright: "We need more research and development efforts to improve their internal and external quantum efficiencies." But there s no doubt that this platform, and the more exotic surfaces of GaN, are already offering a host of opportunities to extend the performance of nitride emitters.   

View pdf of article