The future's bright, the future's blue

Gallium nitride is a new semiconductor that promises to outshine silicon, eclipse gallium arsenide and revolutionise our lives by reducing demand for electricity. John Emsley reports.

There are hundreds of millions of traffic lights around the world, all endlessly and steadily consuming electricity. If all of them used light emitting diodes (LEDs), instead of standard bulbs, the planet would save massive amounts of energy. Until now, what has stopped this from happening has been the lack of a cheap LED that emits green, short-wavelength light.

Enter the new kid in town - gallium nitride: a semiconducting material with revolutionary physical and electronic properties. Its use in traffic signals is already saving money in China, Singapore and California and sales of gallium-nitride devices presently total more than $1 billion. Gallium nitride (GaN) is making short-wavelength LEDs commonplace and could mean that, in future, electronic devices become robust enough to be used under very harsh conditions.

Tim Jones at Imperial College London, specialises in semiconductor materials that have applications in optoelectronics. What does he think is so special about GaN? ’Silicon and gallium arsenide semiconductors are still very important,’ he explains, ’but they have their limitations. GaN is 10 times more powerful and can withstand a much higher power density.The UK semiconductor industry is active in four areas: basic research; chemical precursors; semiconductor manufacturing equipment; and the packaging of LED arrays from components. However, we are not yet manufacturing GaN devices themselves. We clearly lag behind our overseas competitors and we must invest if we want to strengthen our current research base and innovate new technologies for exploitation.’

Potential applications
As a material GaN is exceptionally robust, so it is perhaps unsurprising that it has attracted the attention of the defence industry, for whom expense is of secondary importance. Weapons manufacturers use GaN in phased-array radar systems, which have a longer range than conventional radar and do not reveal their location to the enemy. Additionally, radar systems with GaN components are unaffected by bursts of nuclear radiation.

The robust nature of GaN also makes it an attractive material for electronic sensors, both on the battle field and in Civvy Street. It is a very heat-tolerant material and retains its semiconductor properties at temperatures up to 1000 ?C. Silicon-based semiconductors fail under these conditions. GaN sensors could monitor, for example, the efficiency in a car engine or the performance of a chemical plant. GaN is also able to withstand more than seven times the electric field of other semiconductors. If voltage and current are considered together, then GaN can deliver 50 times the power performance of gallium arsenide (GaAs).

Let there be light!
Lasers based on GaN emit in the blue-violet region of the spectrum (pure GaN emits in the UV). This has caught the attention of companies such as Sony, Pioneer and Thompson: the ’Blu-ray’ consortium of nine electronics manufacturers is developing a new generation of DVD players, which will use blue laser light to read disks, rather than red light which is used at the moment. Because the wavelength of blue light is shorter than that of red, the information-storing ’pits’ on a DVD’s surface can be made much smaller, meaning more of them can be crammed onto each disc. The end result of this miniaturisation is that nearly 30 Gb of information could be stored on a single-layer disc, as opposed to the 4.7 Gb limit of current-generation DVDs.

Incorporating aluminium nitride into GaN means that the frequency of emitted electromagnetic radiation includes microwaves, which is exciting the mobile-phone industry. As mobile phones move away from voice-based communication and mobile video conferencing becomes more popular, communication devices will demand greater signal purity from their solid-state power amplifiers; this is what GaN could deliver. Furthermore, the large breakdown voltages inherent in GaN make such technology well-suited to high-power wireless networking applications. By using GaN, mobile-phone base stations could be 10 times further apart than they are at present.

Tom Foxton of Nottingham University knows a lot about the UK-based research into GaN. He is head of the UK Nitride Consortium, whose aim is to bring the nitride material community closer together and promote discussion and communication. There are currently six groups producing GaN in the UK. These are based at the universities of Bath, Cambridge, Liverpool, Nottingham, Sheffield and Strathclyde. Sheffield is also home to the National Central Facility for III/Vs - III/V referring to semiconductor combinations of elements in groups 13 and 15 of the periodic table.

Foxton worked for Mullards (now part of Philips) from 1969 to 1991, starting the first European research into molecular beam epitaxy (MBE). In 1991 he moved to the department of physics at Nottingham University and started work on nitrides in conjunction with the department of engineering.

Mind the gap
To be a useful semiconductor, the energy needed to promote an electron from a material’s valence band into its conduction band must be of the right magnitude. Materials researchers call this energy difference the ’band gap’. Diamond has a band gap of around 6 eV, making it suitable for use as an insulator. However the band gap values of gallium arsenide (GaAs) and silicon are 1.4 and 1.1 eV, respectively, which accounts for these materials’ extensive use in semiconductor devices (the solid state laser in a CD player is made from GaAs, which has red emission). With GaN, the band gap widens to around 3 eV, meaning the material can emit light at the short-wavelength end of the spectrum. It is this property in particular that has opened up new applications, including being able to generate white light from LEDs.

Some modification of GaN is needed before it can act as an LED or laser. Positive ’holes’ are created by adding electron-deficient atoms (eg magnesium) and electrons are introduced by adding species such as silicon. Where positive, hole-containing GaN (pGaN) meets negative, electron-rich GaN (nGaN) a so-called ’p-n junction’ is formed. A layer of indium gallium nitride (InGaN) is sandwiched in this p-n junction to act as a quantum well, into which holes and electrons migrate and combine to form ’excitons’, which release a quantum of light.

Excitons need to be localised in quantum wells of about 2 nm in size. In Si and GaAs the location of these wells is easily disrupted by crystal defects, which can render a light-emitting device completely useless. The puzzling thing about GaN is that it is riddled with crystal dislocations and shouldn’t work as a microelectronic material at all. If Si or GaAs have more than 1000 dislocations per cm2 they are unable to function as LEDs, yet GaN can have 109 dislocations per cm2 and still perform very well. How does GaN accomplish this mission impossible?

Colin Humphreys, Goldsmiths Professor of Materials Science at Cambridge University, UK recently challenged the conventional explanation of why GaN works. Orthodoxy has it that the InGaN quantum well has clusters of indium atoms, which are observable using an electron microscope, and it is these that provide the nanoscale regions for excitons and explain why GaN is less sensitive to crystal dislocations.

Yet when Humphreys’ PhD student, Tim Smeeton, examined InGaN with an electron microscope operating on its weakest possible beam setting, the regions of indium could not be seen. It was only as the intensity of the beam is raised that they appeared, suggesting that the clusters are simply experimental artefacts, formed by electron beam damage to the InGaN material. Furthermore, not only did Smeeton not find indium clusters in the material made at Cambridge, he could not find them in samples of commercially available GaN either.

So how does GaN function as a semiconductor? According to Humphreys, the answer lies in the material’s piezoelectric properties. Post-doctorial researcher Jonathan Barnard has measured the piezoelectric effect across quantum wells of known thickness using a technique called electron holography. He has shown that while the interface between the n-GaN and the InGaN is perfectly flat, the surface of the interface with the p-GaN is not and displays monolayer thickness variations of the quantum well. These fluctuations localise the excitons and the emissions, explaining why GaN is relatively insensitive to crystal dislocations. For other III/V devices the piezoelectric effect is either non-existent or very weak; they must be free from crystal imperfections in order to work correctly.

Humphreys began studying GaN in 1995, originally relying on material supplied by colleagues. But his group now makes its own GaN-based materials using metal organic chemical vapour deposition (MOCVD). This is done on a ?500 000 instrument donated by Thomas Swan Scientific Equipment (TSSE), Cambridge, in 1999. TSSE is a major player in the manufacture of MOCVD reactors for nitride synthesis. They export equipment for GaN preparation with a large proportion going to the Far East where the main manufacturing centres are located. Within the last two years TSSE has introduced its first production scale system, aimed at growing material LEDs. This has enabled them to secure about 30 per cent of this very buoyant market; the current market for GaN LEDs is $2 billion with a compound annual growth rate of 56 per cent.

Although Humphreys applied to the EPSRC for a grant to make optical materials and devices using the MOCVD instrument, he was initially turned down. But, in 2000, along with Philip Dawson of the University of Manchester Institute of Science and Technology, and Jeff Harris of University College London, he obtained ?800 000 to cover installation and running costs. This allowed three crystal-growing runs to be carried out each week, the limit being the high cost of the ultra-pure chemicals such as trimethyl gallium, trimethyl indium, and dicyclopentadienyl magnesium.

Not that the instrument was under-used: TSSE was more than willing to take up the slack. Indeed, part of the deal was that TSSE’s crystal growth expert Ted Thrush would spend 25 per cent of his time in Humphreys’ lab. Thrush is very supportive of what they have achieved, admitting that the collaboration has been productive for both parties: ’The combination of excellent results, and the ability to bring potential customers to the university to see our equipment in action has had a major impact on the performance of the company’.

Humphreys is equally appreciative of the support of TSSE: ’Although we were novices in the GaN field, we soon caught up with the rest of the world, producing LEDs that could emit bright green light, which we produced by combining GaN with indium nitride. We have produced some extremely energy-efficient devices, with an internal quantum efficiencies up to 46 per cent. Humphreys is currently developing a wide range of GaN-based materials and devices, including GaN quantum dots for use in lasers and quantum computers.

In 2003, EPSRC granted Humphreys a further ?1.2 million for a three-year programme synthesising GaN materials and studying their optical properties. The aim is to make blue and green multiquantum-well LEDs with an efficiency of 50 per cent and UV emitters with an efficiency of 15 per cent. So when will Humphreys’ group start developing semiconductor devices based on their research?

’Our proposal submitted to the EPSRC was to work on both materials and devices,’ he explains ’but the EPSRC’s visiting panel asked us to remove the device work and concentrate on the materials side, and that is only what we have been funded to do. In my opinion their decision didn’t make sense.’ Nor does it, but Humphreys and his colleagues have secured alternative funding for researching GaN devices via a Link Displays Initiative grant, also involving TSSE and a company in Cumbria called Forge-Europa. He has also established strong links with engineers Ian White and Richard Penty at the University of Cambridge.

Will they one day find a way to mass-produce GaN as cheaply as silicon or GaAs? Quite possibly, but it won’t be easy. However, the economic rewards of such an achievement would be truly great.

Epitaxy
Of all the III/V semiconductors produced by crystal growth, GaN is the most challenging because there is no perfect substrate on which to grow it. Silicon was researched as a substrate in the 1990s, but this approach was abandoned because it was realised that the lattice mismatch of around 17 per cent, and thermal expansion mismatch of 50 per cent, could not be overcome. Sapphire was considered a better alternative, having lattice and thermal mismatches of 13 and 33 per cent, respectively. As a benefit it is relatively cheap and has the added bonus of being transparent, although a low thermal conductivity does present some problems.

It is silicon carbide, though, which is the susbrate of choice when it comes to growing GaN. Silicon carbide has lattice and thermal mismatches of 4 and 25 per cent respectively, and good thermal conductivity. It is sometimes used with a barrier layer of GaAs, ZnO or SiO₂  applied to its surface.

Of course the ideal material upon which to grow GaN would be GaN itself. Sumitomo Electric Industries in Japan has looked into this approach and developed single-crystal GaN substrates, with a low dislocation density, especially for this purpose. The company is currently producing quantities of material suitable for GaN growth at a rate of 500 units per week, each of which can be used to make 10,000 blue laser diodes.

Progress in this area may have seemed slow until now but the pace of development is accelerating. Indeed, the first blue LEDs were only made in 1995 and now millions are sold every year.

The future’s bright, the future’s blue.

Acknowledgements

John Emsley is a science writer and author of several popular chemistry books, including his most recent Vanity, Vitality, and Virility, published by Oxford University Press.