Come 2020, US analyst business, Lux Research, predicts sapphire-based LEDs will still dominate a multi-billion dollar LED lighting industry. Compound Semiconductor finds out why.
With LED lighting forecast to become a $80bn industry by the end of the decade, manufacturers of LEDs based on sapphire, silicon and silicon carbide substrates are jostling for market space. Today, the vast majority of devices are deposited on sapphire wafers, but could this change?
Both Plessey Semiconductor and Toshiba-Bridgelux recently unveiled GaN-on-silicon LEDs and have promised mass production before the end of this year while Cree is relentless in its delivery of GaN-on-SiC devices. But despite the progress, Lux Research analyst, Pallavi Madakasira, is certain GaN-on-sapphire LEDs will remain the industry’s leading lights.
In her recent report, “Dimming the Hype: GaN-on-silicon Fails to Outshine Sapphire by 2020”, she predicts GaN-on-sapphire will remain the entrenched incumbent come the end of the decade. Meanwhile the leading emerging technology, GaN-on-silicon, will snare only 10% market share while GaN-on-silicon carbide will grow to 18% of the market. This, she believes, is down to cost and performance.
“GaN-on-silicon developers tout very loudly that fully depreciated [CMOS] equipment can be used to make GaN-on-silicon [devices],” she says. “But can we really expect an organisation that has invested billions of dollars in its GaN-on-sapphire line for more than ten years to stop running that, and get into GaN-on-silicon [production].”
As she points out, industry players know how to make silicon substrates ‘with their eyes closed’ but depositing complex buffer layers onto silicon prior to GaN deposition, to overcome GaN and silicon lattice mismatches, adds time and cost to a manufacturing line.
“As much as these lines are full depreciated, they are still an additional investment for a company space- and time-wise,” she says.
“It’s common knowledge that almost all the big businesses, such as Philips and Osram, have in-house programmes looking at GaN-on-silicon… but it’s very unlikely you will see, say, Philips quit its GaN on sapphire programme and then move into GaN-on-silicon,” she adds.
Madakasira also highlights that the performance of GaN-on-silicon LEDs has yet to be favourably demonstrated compared to alternative technologies. Her figures put the luminance efficacy of GaN-on-SiC LEDs at 200 Lumens per Watt with GaN-on-sapphire devices coming in at between 150 to 180 Lumens per Watt.
Meanwhile, at 80 to 90 Lumens per Watts, Madakasira asserts GaN-on-silicon LEDs are ‘way behind’. “They have not yet achieved the luminous efficacy standards that GaN on sapphire devices had three years ago,” she asserts. “So there might be a substrate cost benefit, but manufacturers will still need to grow complex buffer solutions and, even then, not achieve a performance that matches that of sapphire LEDs.”
“And while early stage innovators [of GaN-on-silicon] LEDs argue they will be able to catch up, you have to remember they are catching up on a moving target. GaN-on-sapphire performances will keep improving and the costs are going to keep coming down,” she adds.
Indeed, six inch silicon substrates may only cost $25, but the two inch sapphire alternative is hardly expensive at between $8 to $10. And as Madakasira points out: “We now have so many Chinese players getting into the space and making LED dies. GaN-on-sapphire is something everyone is able to make.”
So as sapphire-based LED manufacturers settle in for a relatively secure future, where does this leave the manufacturers of GaN-on-SiC and, of course, GaN-on-silicon LEDs?
Cree, the dominant player in the GaN-on-SiC market is safe, reckons Madakasira.
“This technology is very unlikely to lose out to silicon. Cree has successfully made the transition from four inch to six inch diameter wafers and is continually improving performance and bringing costs down,” she says.
Meanwhile, the analyst believes its time GaN-on-silicon developers started forging strategic partnerships with industry heavyweights. Azzurro recently announced its intention to licence out its GaN-on-silicon strain engineering intellectual property to LED manufacturers while Bridgelux forged such a relationship with Toshiba some months ago now.
“Azzurro has made a smart move… and I think the way Bridgelux has formed its partnership is the way to go,” she says. “For start-ups in this space you need to have these critical, strategic partnerships or you really are going to go nowhere.”