The speed of SSL adoption, however, will depend on how quickly the cost of ownership can be brought down. Today, typical LED replacement bulbs provide many years of energy savings over their lifetime, but not enough to reclaim their original cost. Sandia National Labs estimates that today’s LED lighting has twice the lifetime cost of ownership of incandescents and 10 times that of fluorescents. But in the decade starting in 2020, the lab predicts, SSL will reach lifetime costs that are one-tenth that of incandescents and one-half that of fluorescents.
The primary engines for cost savings in SSL are the same ones driving cost reduction in any semiconductor technology: higher yields, fewer materials and larger wafers. The near-term throttle holding prices high, according to Veeco, is process nonuniformities that affect yields. In fact, there is so much nonuniformity in LEDs today that they have to be tested and “binned” for color, brightness, forward voltage and other variations that accumulate during the 10-hour, thousand-degree vacuum deposition of a hundred semiconductor layers. Veeco’s MOCVD tool was recently upgraded with a “uniform flow flange” that the company claims keeps tighter controls on temperature, flows and materials composition, giving LED makers up to 90 percent yields for a 5-nanometer bin.
Veeco and other MOCVD makers, such as Aixtron SE (Herzogenrath, Germany) and Taiyo Nippon Sanso Corp. (Tokyo), will continue to increase yields incrementally; those advances will combine with steady improvements in assembly, packaging and the driver chips for LEDs. Meantime, SSL pioneers are inventing multi-LED architectures that they say sidestep today’s yield and binning problems.
Cree, the current SSL leader, had a leg up on crafting such architectural features, thanks to the white LEDs it makes as backlights for mobile phones. Indeed, Cree holds intellectual property that prevents other LED makers from using its secret sauce: silicon carbide (SiC) wafers whose lattice mismatch with gallium nitride gives Cree a claimed competitive advantage over competitors using sapphire wafers.
The architectural features Cree uses to sidestep the yield and binning problems let the company prequalify LEDs internally before combining them into multidie LED components. Cree uses two architectures, TrueWhite and EasyWhite, that together “allow us to make use of almost every LED we make,” said Scheidt. “We can use LEDs that would ordinarily be too far off [spec], but which can be made usable by mixing them with other LEDs in same package.”
TrueWhite is a hybrid approach using a phosphor-corrected LED (pc-LED), tuned to provide a greenish light, matched to a separate, red LED. Together they provide warm white light that Cree claims is even more appealing than the light from some expensive halogen bulbs, achieving a color rendering accuracy of 90 on the CRI scale of 0 to 100. Cree proved its TrueWhite architecture in its downlights, which have outsold all other solid-state lights combined.
For less expensive luminaires, Cree’s EasyWhite architecture matches LEDs binned as overly greenish with others binned as overly reddish, then combines them into the same package with a dozen or more complementary dice to create a consistently pleasing hue of white light at a relatively low cost. Color rendering accuracy comes in at about 80 CRI.
|While the conventional light bulb is an electrical device, Cree's TrueWhite A19 prototype is chock full of electronics.
|Cree's LED prototype swap-in for the A19 incandescent bulb won't burn your fingers during replacements.
The final hurdle to cost reduction in LEDs—manufacturing on larger wafers—will be cleared only as SSL volumes ramp to a level that justifies the transition. Most of the LEDs sold in today’s $10 billion market are manufactured on 2- to 4-inch wafers (about 10,000 LEDs fit on a 4-inch wafer). A series of milestones is planned for transitioning to 6-inch, 8-inch and larger wafers as volumes ramp.
“Today the primary engineering problems with LEDs involve reducing cost, since brightness is already approaching the 150-lumen/watt level, which makes them the brightest available light source,” said Alexei Erchak, chief technology officer at Luminus Devices. “Unfortunately, LEDs are still way more expensive than conventional lighting technologies. But the gap is decreasing.”
Luminus Devices has already migrated its LEDs to 4-inch sapphire wafers and will move to 6-inch wafers over the next few years. Rather than use multiple LED dice in the same package, the company prefers to cut costs by using extra-large dice so that fewer LEDs are needed inside each bulb.
Luminus began its life as a projector company and thus has amassed expertise in making the extra-large LEDs required to power projectors. Borrowing that technology for SSL, it aims to use as few dice as possible—preferably just one—per incandescent or halogen replacement bulb.
The endgame industrywide is to cut LEDs’ cost by as five times by 2020, according to the U.S. Department of Energy’s ”Solid-State Lighting Research and Development: Manufacturing Roadmap.”