“In conventional hexagonal gallium nitride you end up with electric fields on the c-plane across the crystal,” said Dave Wallis, funded programme manager at Plessey. “This is good for making transistors because you get a two-dimensional gas without the need for dopant, but in LEDs it separates electrons and holes which would have combined to make photons.”
What goes on is called the ‘quantum confinement Stark effect’ (QCSE), said Wallis.
QCSE pushes electrons and holes apart so they fail to combine in the LEDs’ GaN-InGaN quantum wells where photons are made. The effect increases as the amount of indium increases.
As increasing the amount of indium is the mechanism used to lengthen the wavelength of GaN LEDs – from violet, through various blues to green, QCSE is most serious in green GaN LEDs making them the least effective at emitting photons.
However, if you shift to cubic lattice GaN “the symmetry changes and the electric fields are completely removed”, said Wallis – the QCSE no longer holds effects photon creation.
Is QCSE the entire reason for the green gap? No one is sure, said Wallis, but it is part of the cause, and both internal and external quantum efficiency will be better in cubic GaN green LEDs – they will make more photons per unit of electricity.
Shifting to cubic GaN has a second advantage when it comes to green LEDs: its bandgap is 200mV lower than hexagonal GaN. THis means “if you want to go for green, you don’t have to use as much indium, and there are a lot of problems getting enough indium into quantum wells”, said Wallis.
So why isn’t everyone using cubic GaN for green?
In GaN, the 3C crystal lattice is thermodynamically unstable so, at the temperatures required for epitaxial growth, only hexagonal crystals will form – unless the energy balance is artificially tilted. And Anvil Semiconductor of Coventry has found a way to do just that.
It has invented a way to grow cubic silicon carbide, and the lattice constant (atom spacing) in 3C SiC is close enough to that of 3C GaN that, given a cubic SiC substrate, the hexagonal form of GaN will be discouraged and cubic crystals will grow.
The University of Cambridge, at its Cambridge Centre for GaN, has done just that and grown GaN with >99% cubic structure, said Wallis. Not only that, but it has grown quantum wells in the material, and these have been stimulated into blue and green photo-luminescence.
So far, the structures need photos to make photons but, according to Wallis, Cambridge will be growing n and p-type layers around the wells to make diodes that can be biased to convert electrons to photons – LEDs.
Another advantage of the Anvil process is that it grows cubic SiC on cheap silicon wafers rather than expensive SiC wafers – the stress that would build-up from the large lattice mis-match is dissipated by dividing the surface into squares by trenching.
The Cambridge Centre for GaN knows all about growing hexagonal GaN on silicon wafers – a process which it sold to Plessey which, in turn, now uses it to make blue and white LEDs at its Devon fab. The Centre’s cubic GaN is grown on cubic SiC-on-Si wafers, which are very close to the wafers Plessey is used to handling.
Once the n and p-layers have been incorporated in Cambridge, the wafers will be shipped to Plessey to have electrodes deposited to form working green LEDs – at least, that is the plan.
So far, the troika is three months into its programme, which will end in September when green cubic GaN LEDs are expected.
How efficient will they be?
“I wouldn’t like to put a number on it. The hope is that green leds will be approximately as efficient as blue leds,” said Wallis pointing out that all three organisations can work with 150mm (6inch) wafers, opening a route green LED production on 150mm.
With white LEDs already made from blue die and amber (or amber/red) phosphor, is there a market for green LEDs, however efficient?
Yes, said Wallis, in lighting that combines red, green and blue LEDs to make colour-tuneable light.
While better greens will only improve the efficiency of one of the three sources in an RGB lamp, there will be a valuable reduction in heat generation, he said, as the other two sources are already quite efficient – and heat removal is the bugbear of LED lighting.