A white LED breakthrough at the University of Cambridge could lead to mass production in the UK.
"Its it is a way of making GaN LED die that is a factor of 10 cheaper: growing them on 150mm silicon wafers rather than 50mm sapphire," Professor Colin Humphries told Electronics Weekly.
"The cost of processing a 150mm wafer and a 50mm wafer is the same, or even cheaper in 150mm because a lot of 150mm fabs exist," and raw 150mm silicon wafers are much cheaper to produce than 50mm sapphire wafers.
Everyone would be doing this if it was easy. The critical step is reliably growing GaN/InGaN structures on silicon where the mismatch in lattice constant means wafers bend - preventing lithography - or even break.
When researchers have made flat GaN-on-Si wafers, the high number of dislocations caused in the GaN makes it useless for electronics.
"LEDs are much more sensitive to high dislocation densities that electronics," said Humphries.
To get a flat wafer, the University grows inter-layers to add compressive stress - a standard technique, but the devil is in the detail.
"We have now done that so we have a very flat wafer with 1011/cm2 defect density," said Humphries. "Then we have a way to knock that down a hundred fold, to below 109."
It is this last step that makes LEDs viable.
"We introduce a thin SiN layer before making the InGaN quantum wells," said Humphries. "We believe this layer has pin holes and the SiN blocks dislocations coming up."
So good GaN crystals grow out from the pores, only developing flaws when they hit the crystals growing from surrounding pores.
According to Humphries, the resultant defect density is probably not as low as existing power LED makers achieve on their sapphire and SiC wafers.
"Our devices have an internal quantum efficiency of 40%. We think Cree's is about 70% grown on SiC, and growth on sapphire is also more efficient," he said. "But we have only been working for a year or so and we are still on a steeply rising curve. Our way is so much cheaper, I think it is probably commercially viable even now."
With die one tenth of the price, Humphries estimates packaged devices will be a quarter of the cost of existing LEDs.
Cambridge is operating within a consortium part-funded by the Government's Technology Strategy Board which also includes GaAs chip-maker RF Micro Devices (RFMD) - which claims to process more GaAs wafers than any other company in the world. GaAs and GaN production is very similar.
RFMD runs the GaAs fab in County Durham, currently Europe's largest, that was once a Fujitsu 150mm silicon fab - and was well known as Filtronic's GaAs fab.
"We are investing in the consortium and if it is successful we would manufacture," said RFMD's Steve Clements, who is overall project leader.
He cautions that manufacture is not a foregone conclusion. "I don't want to pour cold water on it. We would like to be mass-producers and are used to large commercial consumer applications, but to make something a commercial success requires more than some technical results."
Other organisations involved are: QinetiQ, which aims to turn Cambridge's science into a commercial process: and German growth equipment maker Aixtron.
Cumbria-based LED product maker Forge Europa is looking at marketing and applications.
Professor Phil Dawson at the University of Manchester provides independent measurement of quantum efficiency.
Fundamental science in the programme is funded by the Engineering and Physical Sciences Research Council (EPSRC).