Vertical conduction strategy cranks up UV LED output power

Mercury lamps are the dominant source in numerous applications requiring UV light of less than 340 nm. They are used to purify air, water and food, cure polymers, and provide the emission that pumps the blue-white phosphors of fluorescent lamps.

But success has not resulted from a product that totally fulfills the customer s wish list. In fact, this lamp is simply the only option available today and users are forced to put up with its bulky size, high operating voltage, and its use of mercury. This toxic element raises environmental concerns, particularly for the food and medical industries.

What s needed is a solid-state replacement – a high-efficiency LED operating within the UVB and UVC parts of the solar spectrum (see figure 1 for details of the spectral ranges). This would provide an ideal replacement for all of the applications outlined, thanks to its compactness, lack of mercury, low driving voltage and straightforward integration with silicon control electronics. Pulsed-mode operation could also be used to drive the LEDs, leading to significantly enhanced performance in many optoelectronic systems, such as those used for biochemical detection and identification.

Developing a solution
Efforts have been under-way to develop such a source since the beginning of this decade. In the US, research has been driven by the Defense Advanced Research Projects Agency, which guided a program entitled Semiconductor UV Optical Sources (SUVOS). Our team at the University of South Carolina was involved in this four-year, $50 million project. At about the same time Japan started to develop UVB and UVC LEDs through research at various institutions, including RIKEN, the Tokyo Institute of Technology, Meijo University and the University of Tokushima.

This work initially employed sapphire substrates and AlGaN or AlInGaN multiple-quantum-well-based p-n junctions (figure 2a). Sapphire is available in large sizes and is completely transparent in the deep UV, which enables the use of a flip-chip packaging geometry that features light extraction through the substrate.

However, this combination of materials has its drawbacks. It is difficult to dope AlGaN, and this nitride has a large lattice mismatch with sapphire, particularly at the high aluminum mole fractions of 30–70% needed to produce deep-UV emission.

Between 2000 and 2004 our group developed three innovative approaches to overcome these problems. The first of these was the introduction of AlN buffer layers on sapphire, which were grown by using a pulsed MOCVD approach. This growth process produced films with excellent crystal quality and a very smooth surface.

Our pulsed MOCVD process was also used to fabricate AlN/AlGaN superlattices that reduce epilayer strain and enable the growth of crack-free thick layers of n-type AlGaN. These layers improve device conductivity and reduce current crowding. On top of this we introduced a heterojunction comprising p-GaN and p-AlGaN, which increased p-type conductivity and hole injection.

These improvements led to the fabrication of 280–340 nm devices producing continuous-wave (CW) output power in excess of 1 mW at 20 mA. These LEDs feature flip-chip device packaging that improves thermal management, something that was previously limited by sapphire s poor thermal conductivity (figure 2b).

These first-generation deep-UV LEDs suffered from small output powers, low efficiencies and relatively short lifetimes. Although the flip-chip geometry improves thermal handling, CW output powers saturate at relatively low drive currents due to device heating. Efficiencies are also poor – typically just 1–2% – due to the high dislocation densities in the AlGaN-on-sapphire films. These weaknesses combine to cause premature device degradation – CW lifetimes are only about 200–400 hours at 20 mA. Yet, despite these major shortcomings, these devices are being sold for air, water and food purification, deployment in miniaturized biomedical instruments, polymer curing and deep-UV spectroscopy. Sensor Electronic Technology (SET) – a spin-off of our group that manufactures these devices – is selling them, alongside its strategic partner Seoul Opto-Devices Company of Korea, which also undertakes processing and packaging.

Today we are aiming to address all of these weaknesses, along with other researchers in this field. Considerable effort has been directed at processes for making templates with thick layers of GaN or AlGaN on sapphire, which can form a better platform for UV LEDs. Our research has involved the development of two growth methods: pulsed air-bridge assisted lateral epitaxy (PLOG) and metal-organic hydride vapor phase epitaxy (MOHVPE), which can produce low-defect AlN buffer layers on sapphire that are at least 15 µm thick.

PLOG is a technique that is similar to epitaxial layer overgrowth, but involves pulsing the delivery of one reactant. This variation, which is compatible with both MOCVD and HVPE growth, prevents the growth of adducts, a major cause of defects in high-aluminum-content nitride epilayers.

The MOHVPE technique can bring together the benefits of MOCVD and HVPE in a single chamber. MOCVD can be used where slow, controlled growth is required, such as the deposition of the buffer, and HVPE can form thick layers in a reasonable time.

Combining PLOG and MOHVPE has led to the production of deep-UV LEDs on AlN-on-sapphire templates that deliver a significantly superior performance to our first-generation devices. Peak power output is increased as higher pump currents can be tolerated, thanks to the better thermal conductivity of AlN epilayers. LED lifetimes also receive a boost, due to reductions in defect densities and improvements in thermal handling capabilities. And 280 nm emitters hardly deteriorated during a 2000 hour lifetime test (figure 3).

Vertical conduction
Recently we have made further improvements to our 280 nm LEDs by introducing a vertical conduction geometry that is similar to that employed by several makers of visible-range LEDs. Again, this approach begins with the growth of the epilayer structure on an AlN-on-sapphire template, but in this case the top surface is bonded to a p-contact connected to a heat sink. The sapphire is then removed by laser lift-off (figures 4a and 4b).

The vertical conduction design is ideal for producing large-area LEDs, including deep-UV lamps. High output powers are possible because this type of device is capable of handling much higher drive currents. In our case, we have built devices that can operate at 1 A. We are currently packaging and testing emitters that will have similar efficiencies to our lateral current LEDs, but total output powers that are 1–2 orders of magnitude higher, thanks to the higher drive currents.

Our improvements in UV performance, which have been driven by optimization of the design, will be implemented in commercial products. Nitek Inc, a spin-off of our research group that is developing deep-UV technologies, is planning to market second-generation deep-UV LEDs made with the PLOG and MOHVPE processes. Patents have been filed for these and other Nitek proprietary technologies. (SET, in comparison, employs a migration-enhanced MOCVD approach). These devices should deliver lifetimes that are well in excess of 5000 hours, thanks to the incorporation of new device geometries that will improve thermal management.

Progress to date has raised the performance of deep-UV LEDs to a level comparable with that of visible nitride LEDs of the mid-1990s. In that community, large-area visible LED lamps were one of the next steps and UV-LED researchers seem to be following a similar path. Indeed, we are not the only research group to develop vertically conducting deep-UV LED lamps. There is a massive market pull for commercialization of such devices, which would offer a welcome alternative to the mercury-based lamps that are employed today.

Further reading
A Khan et al. 2005 Jpn. J. Appl. Phys. 44 7191.
M Iwaya et al. 2003 Jpn. J. Appl. Phys. 42 400.
H Hirayama et al. 2002 Appl. Phys. Lett. 80 1589.
A Khan et al. 2008 Nature Photonics 2 77.
K Balakrishnan et al. 2007 Jpn. J. Appl. Phys. 46 L307.
K Kawasaki et al. 2006 Appl. Phys. Lett. 89 261114.