There’s no question that bringing down the cost of HB-LEDs would accelerate market penetration. The US Department of Energy (DOE) in fact is tentatively targeting a 20x cost reduction over the next five to six years to achieve its energy savings goals for the sector. And key to reaching that ambitious goal is things like better chip yields and more efficient automated production.
The same manufacturing efficiencies that reduced semiconductor cost over the past three decades will enable high-brightness LEDs to penetrate the general purpose lighting market, argues SEMI president Stan Myers, and industry collaboration—facilitated by appropriate standards—is a big part of that.
HB-LED manufacturing now, with its highly proprietary processes running on 3-inch wafers at throughputs of ~50 wafers per hour, with limited yields and limited equipment uptime, looks very much like the semiconductor industry in 1975, when it too was running proprietary processes on 2-3-inch wafers at ~50 wafers per hour with limited yields and limited uptimes.
Since then the semiconductor industry has achieved a 4,000,000x cost reduction, 500x of that from manufacturing efficiencies, thanks to10%-30% improvements in productivity and 10%-15% improvements in yield every year, and a 30%-50% increase in wafer size every 10-15 years. Manufacturing cost of flat panel displays has followed a similar curve, for similar reasons.
The LED sector is making fast progress on improving LED efficiency and drive current requirements, but judging from the recent progress of the technology, practical people looking at the trends can see a 1.5-2x improvement in efficiency, and about a 4x-6x improvement in getting more light out with higher drive currents, notes Ultratech CTO Andy Hawryluk. “The rest is going to have to come from improvement in manufacturing productivity and utilization,” he says.
Now that a real market is developing, establishing some industry standards would enable equipment makers to devote the resources to developing more efficient tools tailored specifically to the needs for the sector. Currently, manufacturers use either re-purposed semiconductor equipment or specially designed custom tools. Though MOCVD (metalorganic vapor phase deposition) may have the most headroom for improvement, there’s also potential for improving the other process steps.
In lithography, for example, HB-LED designs are getting to feature sizes of 5 micron and below and to volume production requirements that drove the MEMS and thin-film-head industries to move from contact aligners to projection lithography, and could likely get the same gains in throughput, critical dimension control and overlay that these other sectors experienced with the next generation equipment.
But semiconductor industry reduction steppers are costly overkill, and not well suited to LED issues. Instead each company wants its own customized tool. “Up to now there was no motivation to do this design for one or two customers,” says Hawryluk. “We’re hoping for some industry standards, so instead of just selling one, we can sell multiple tools with a similar configuration.”
For example, Ultratech is demonstrating an HB-LED projection stepper with a new loading system for small substrates and an image recognition alignment system to enable automation. More unique is software that pre-maps the surface and compensates for warp by tilting the stage to keep the wafer focus, since the thicker sapphire wafers can’t be pulled flat by vacuum.
Epitaxial growth issues
The LED industry is very different than CMOS, as Rainer Beccard, Aixtron VP of marketing, reminds the silicon folks. “It’s a very complex process, with thick structures consisting of large numbers of different layers, and a lot of manual handling,” he says. “But it is not finally optimized yet, and we will see the same learning curve as silicon as it matures.” Beccard notes the young LED industry, though, is currently somewhat ahead of its expected cost reduction curve.
A key milestone on the way towards better performance at lower cost is the implementation of large wafer size processes, argues Beccard. Current MOCVD tools already can be quickly changed over to run larger substrates, such as multi 6-inch instead of multi 2-inch wafers, and actually get better results with more efficient use of the epi reactor and fewer edge-related defects with the larger wafers. However, strain from depositing compound semiconductor layers on sapphire warps the wafers if they’re too thin, so 6-inch wafers typically are about 1mm thick. That makes the 6-inch wafers cost significantly more per square centimeter than 4-inch wafers, so it’s not yet economic to make the switch. Equipment designed especially for these LED issues will eventually help, such as systems for optical monitoring to better control the process to reduce strain, and adding compensating layers, now in development.
But another issue may be the conservatism of the sector. “It’s only the more established makers that are doing 4-inch, and only few of them are thinking about 6-inch,” says Beccard. “It’s technically possible to get higher throughput and lower cost, if people would only dare to go for more disruptive technology,” he argues. “All those silicon people getting interested and about to get in could stimulate the industry. And maybe we can learn from them and not have to make all the same mistakes silicon made.”
HB-LEDs at SEMICON West
The DOE will also report on the progress of its planned initiative to encourage industry efforts towards efficiencies, followed by an HB-LED Meetup. See the Extreme Electronics program at SEMICON West website for complete details.
LEDs Magazine readers can register online for SEMICON West and gain FREE access to the exhibits and Extreme Electronics HB-LED session, by using the following promotional code: W9EXX