June 21, 2013
Will a new buffer material for gallium nitride tempt LED manufacturers away from sapphire and onto silicon? Compound Semiconductor investigates.
More than a decade ago, researchers from Australia-based Queensland Micro and Nanotechnology Facility (QMF) of Griffith University were depositing silicon carbide on silicon as part of a venture capital funded program to develop novel non-volatile memory cells. Come the global financial crisis, funding waned and the project was shelved. The process, however, was not.
In 2011, the QMF researchers joined forces with micro-device equipment manufacturer, SPTS Technologies, to develop a production reactor targeted at producing SiC coated Si wafers for GaN for LEDs.
As Alan Iacopi, QMF director of operations, explains: “Silicon carbide makes a great buffer layer for GaN, but we knew that if the industry was going to take us seriously, we needed a production vehicle to commercialize the technology.”
And, today, this is what they have. Sited at QMF, the team's vertical reactor - dubbed EpiFlx - is designed for high temperature vacuum processing of large batches of wafers sized from 50mm-300mm.
Right now, the team is finalising the baseline SiC-on-Si growth processes on the reactor to produce a cost-effective buffer material. Will this at last enable LED manufacturers to switch from sapphire to silicon substrates? Iacopi and the rest of the team thinks so.
“Since March of this year we've been transferring our process from the original research reactor to the EpiFlx and the initial uniformities are very encouraging,” says Iacopi. “We've already achieved SiC film thickness uniformities of around 1% on 300mm wafers from the EpiFlx and are now characterising the processes on smaller wafers.”
At the same time, SPTS and QMF have been working with industry partners to validate the technology within the LED and Power GaN on Si markets. Structures fabricated on silicon are already showing promising results and as SPTS chief executive, Bill Johnson, says: “These manufacturers feel that a SiC buffer layer offers a real value proposition and like the ability to have a batch tool capable of supporting multiple GaN MOCVD reactors.”
Both Iacopi and Johnson agree, there is no demand for 300mm wafers right now, but as Johnson says: “We wanted to say, ‘look, here's a 300mm wafer with industry-leading uniformity’. It clearly shows the industry that a very uniform buffer layer can be epitaxially grown on smaller wafer sizes.”
Today, LED manufacturers are migrating from 100mm to 150mm sapphire substrates to boost chip yields per wafer, but the partners are looking to enter the market with 200mm wafers.
“The leading LED manufacturers are now interested in 150mm substrates and we believe that by the time we are ready with our technology, 150mm sapphire substrates will have a usable cost structure. So 200mm would be a logical insertion point,” says Johnson.
Initial estimates suggest the team's SiC on silicon coating process, in volume production, will add no more than $35 to the cost of a silicon wafer.
This, without a doubt, would draw LED manufacturers away from large sapphire substrates, costing hundreds of dollars, but what about competition from, say, Azzurro, Translucent and Kyma? The team reckons its technology offers greater appeal than commercially-available templates.
Johnson asserts EpiFlx produced wafers would be much cheaper than templates with exotic layers produced via more expensive deposition processes while Iacopi believes his team offers a more flexible proposition.
As he highlights, many makers of buffer materials claim templates are good for LED manufacturers to 'play with', but is this the case?
“LED manufacturers can't 'tune' these and will not be in control of their manufacturing process; they also can't control how much they pay for those wafers,” he says. “With an EpiFlx system, manufacturers will get a baseline process and you can imagine them tuning this for their own applications to develop a competitive advantage.”
So as SPTS-QMF team continues to fine-tune its process, where next for the partners? “By early next year customers will be engaged in beta tests with us,” says Johnson. “This is not years away, it's quarters away. And from a performance and cost of ownership point of view, EpiFlx will be far and away the clear winner.”