Forming high-quality GaN via growth through holes and coalescence, using sputtering to form high-quality nitride films on metal foil and increasing the versatility of this class of material with the addition of scandium all featured at the recent IWN meeting. Stephan Knoll from the University of Cambridge reports.

Japan has now played hosted to three of the seven biannual International Workshops on Nitride Semiconductors. This first ever meeting in this series was held in Nagoya in 2000, six year’s later Kyoto was chosen and this year delegates headed to Japan’s most northerly island, Hokkaido. Although they were hit by chilly autumn winds on arrival, they couldn’t fail to notice the heartfelt welcome they received at Sapporo.

The weeklong workshop that they attended kicked off on 14 October, slightly later than is traditional for this conference, so that they had the chance to admire and enjoy the colourful autumn foliage. Pushing out the meeting meant that the academic year had already started in many universities, but that did not quash attendance: The conference attracted over 900 delegates. A wide range of excellent posters and presentations were delivered at the meeting, and following on from pervious IWN workshops, talks were divided into four topics: In this case growth, optoelectronic properties and characterisation, optical devices and electronic devices.

This year’s meeting showcased the large volume of work, both from industry and academia, that is focusing on the development of bulk nitrides. The benefits of having readily available bulk substrates are already widely appreciated, with the absence of lattice mismatch freeing device makers and developers from problems associated with high dislocation densities and low radiative recombination efficiencies. However, these advantages come with a penalty – high substrate prices, which stem from high fabrication costs. 

The good news is that prices could tumble, thanks to significant leaps in progress made by many groups using various approaches. It will be interesting to see which one of these methods establishes itself as the best – this is sure to be one of the hot topics in future conferences. One strong contender to driving down costs is the sodium flux method. A team at Osaka University, Japan, are pioneers of this approach, and at the IWN meeting Yusuke Mori described how he and his co-workers had learnt to yield low-dislocation-density, high-quality GaN from a poor-quality seed crystal.

This team has recently employed two technologies that have led to the fabrication of centimetre-sized bulk GaN single crystals with dislocation densities below 103. The first is described as necking, and involves forcing a GaN crystal to grow through a small hole in a sapphire plate placed over the GaN seed – dislocations propagating through this aperture bend and terminate at sample edges. The other is to coalescence multiple GaN crystals without generating dislocations at their boundaries.

Improvements in AlN also featured at the most recent IWN meeting, with wide bandgap materials developer HexaTech unveiling its low-defect-density, free-standing AlN. This company that is based in Morrisville, NC, produces these pieces of AlN via an iterative Boule expansion process, using AlN substrates as seed crystals. The process is scalable, with crystal quality maintained over many generations of boules (average defect densities is less than 103 cm-2).



Figure 1  As-grown AlN boule, grown by physical vapour transport. Epi-ready AlN wafers are fabricated directly from these boules. 

One of the areas of improvement for these crystals is their level of transparency, which is hampered by sub-band transitions that are attributed to carbon point defects. However, this does not prevent Hexatech’s AlN from providing a platform for impressive UV LED performance. Device structures produce an internal quantum efficiency of 70 percent, which is claimed to be the highest ever figure for this class of LED.

Arguably the most famous LED developer of devices built on native substrates is Soraa, a start-up founded by Shuji Nakamura, Steven DenBaars and James Speck from the University of California, Santa Barbara (UCSB). The firm’s chief technology officer, Mike Krames, claims that their GaN-on-GaN approach has two big advantages over traditional LEDs: Dislocation density in the epilayers is three orders of magnitude less than that for GaN grown on sapphire; and droop, the decline in device efficiency as current is cranked up, diminishes by 15-25 percent. In February 2012, this west-coast outfit launched it MR16s flagship LED, which is based on this homoepitaxial technology and delivers 2400 cd at 12 Watts. This is claimed to produce a similar lumen output to a 50W halogen lamp.

To spur widespread deployment of LED lighting, chip costs must fall. One way to do this is to reduce the growth temperature, possibly by switching the deposition technology from MOCVD to pulsed sputtering deposition (PSD), a technique that enhances surface migration. Hiroshi Fujioka from the University of Tokyo outlined this technology, explaining that it has produced device-quality III-nitrides at room temperature. According to him, PSD can be scaled up relatively easily, it allows fabrication on large area substrates, and thanks to the low growth temperatures associated with this technique, it is ideal for forming high indium-content devices, such as long wavelength LEDs and solar cells. The team have used PSD to form full-colour RGB LEDs on a metal foil and mica sheets.

GaN on anything?

Growth of GaN on substrates such as these also reduces LED chip costs, and for that reason it is currently attracting significant attention. Another low-cost platform is glass, which is widely used in the thin-film PV industry and has, amongst its attributes, large size and transparency. The main challenges associated with the growth of single-crystalline GaN films on glass are the low melting point of the substrate and the lack of a well-defined global epitaxial relationship between it and the epitaxial layer.

However, it is possible to address these issues by first depositing a titanium pre-orienting layer. Researchers at the Samsung Advanced Institute of Technology, Korea simply evaporate titanium onto glass, and use this layer to provide an intermediary for epitaxial growth of GaN. This determines a common c-direction for the epitaxial layer, but in-plane rotation between different GaN islands occurs. Up until now, it has proved impossible to get them to coalesce, so deposition creates an array of randomly rotated GaN islands.

Growth of this GaN film is a two-step process that begins with the growth of a low-temperature GaN nucleation layer on the titanium pre-orienting film. A SiO2 hole mask is added, before high-temperature GaN growth selectively allows certain grains to grow through the mask and forms GaN pyramids aligned with the c-axis. Each pyramid remain un-coalesced and functions as an individual micro-LED, delivering emission characterised by excellent uniformity.

Auger falls out of favour

In the on-going debate surrounding the physical origin of efficiency droop, it seems that Auger recombination is no longer the most popular explanation for its cause – many at IWN 2012 now seem to side with the view that carrier leakage out of the active region is the primary cause of this mysterious malady. One of the first team’s to argue that this is the culprit is Fred Schubert’s team from Rensselear Polytechnic Institute, NY, and at the meeting they detailed their an analytical model for efficiency droop in LEDs. According to them, this model provides excellent agreement with experimental values.

Support for carrier leakage is coming from, amongst others, a partnership between the National Taiwan University and UCSB. Yuh-Renn Wu from NTU described efforts to understand the effects of nanoscale composition fluctuations in InGaN quantum wells, as found by atom probe measurements. Wu claims that although the indium fluctuations boost radiative recombination via enhanced local carrier confinement, they also lead to shallower junctions within the quantum well that result in greater electron overflow. He and his co-workers argue that it is possible to account for droop entirely by just considering charge leakage.

Different planes

Another highlight of the workshop was the talk given UCSB’s James Speck, who detailed his team’s progress in non-polar and semipolar GaN materials and devices. He began by reminding the audience of the biggest problem inherently associated with fabricating devices on c-plane GaN: Due to the polar nature of GaN, devices suffer from large internal fields that drive electrons and holes apart due to the quantum confined Stark effect, and ultimately limiting radiative recombination efficiencies. To yield devices with respectable efficiencies, wells must not be too thick, which adversely affects the LED’s droop behaviour. 

Speck explained that to overcome the limitations imposed by the QCSE, researchers have turned to GaN growth on nonpolar and semi-polar planes. Devices grown on such planes benefit from either reduced or no electronic polarisation, but they produce broad luminescence linewidths and low efficiencies at higher wavelengths – these are areas for improvement. Speck spoke of the benefits of selecting optimal semi-polar planes, such as (2011) and (1122), which show a high tendency for indium uptake. The downside of growth on semi-polar planes is that it often results in dislocation slip on the inclined basal plane, as well as the generation of misfit dislocation at heterostructure interfaces. However, despite the challenges, Speck and his co-workers have fabricated blue laser diodes on (1122) semi-polar planes on relaxed buffer layers, which effectively suppress the effect of misfit dislocations near the active region.

 

Figure 2 Schematic representation for the growth of a GaN pyramid array on a glass substrate. 

A unique feature of the non-polar m-plane is that it intrinsically emits highly polarised light, which could find utility in a wide range of applications, such as efficient colour displays and quantum encryption. Until now, techniques for increasing LED light extraction, such as surface roughening or pattering, have paid the penalty of halting polarised light emission. But it is possible to combine high light extraction with polarized emission by building photonic crystal nonpolar LEDs. Speck’s team has just made the first device of this kind.

Laser and LED performance continues to improve on every plane of GaN. Nichia used conventional GaN substrates to launch its first green laser diodes in August 2010, and it has subsequently increased their output power and wavelength. The first lasers emitted 510 nm at 50 mW, while at the IWN meeting they showcased 80 mW, 515 nm variants.

Novel nitrides

One way to improve the versatility of the nitrides is to increase the number of alloys that are available. A new entrant to the nitride portfolio is scandium nitride, which has been pioneered by Michelle Moram’s team from Imperial College, UK. She explained that one of the strengths of scandium, compared with other transition metals, is that it is soluble in the nitrides across the entire composition range. What’s more, it retains both a direct band gap and the wurtzite structure at low scandium fractions.

The London-based team have investigated the growth and properties of ScGaN and ScAlN alloys grown heteroepitaxially by MBE. At low scandium contents, a hexagonal structure is maintained and growth occurs mostly in a three-dimensional mode, with coalescence beginning at film thicknesses of 300 nm or more. According to Moram, although growth challenges remain, ScN is on track as a potential new member of the existing nitride family, thanks to its wide range of accessible lattice parameters and bandgaps.

These efforts, plus those of others in the nitride community, are helping to enhance the performance of GaN-based devices, whether they are grown on native material or low cost substrates, such as glass. More progress will be unveiled at next international nitride conference – the tenthInternational Conference on Nitride Semiconductors, which will be held on the outskirts of Washington DC in late August.