Cree R&D achieves 1,000 lm from single-chip LED (update)

Sept. 7, 2007
A single-die, cool-white LED driven at 4 A has produced 1050 lumens in Cree's R&D lab.
In its R&D lab, LED manufacturer Cree has demonstrated a single-chip white LED with an output of more than 1,000 lumens, equivalent to the output level of a standard household light bulb.

Cree says that historically its R&D demonstrations generally have been commercialized within 12 to 24 months.

A single-die LED, driven at four amps, produced 1,050 lumens in cool white, with an efficacy of 72 lumens per watt. The input power was 14.6 W and the forward voltage was 3.65 V.

A warm-white version produced 760 lumens with efficacy of 52 lumens per watt.

Both LED versions operated at substantially higher efficacy levels than those of today's conventional light bulbs.

"Cree's XLamp LEDs are the best-performing commercially available LEDs, but we won't be satisfied until light bulbs are obsolete," comments John Edmond, Cree co-founder and director of advanced optoelectronics.

"We've worked 20 years to achieve lighting-class LED performance, and we still have plenty of ways to advance the technology further."

More details

LEDs Magazine spoke with John Edmond, who confirmed that the results are based on a next generation chip in which almost all aspects if technology – from epitaxial layer design to optical extraction to phosphor conversion – are not currently in production. A mix of evolutionary improvements and brand new, innovative approaches have been used. Individually, each of these improved technologies will begin to appear in production devices in the next 1-2 years.

Edmond would not comment on the shape of the chip or confirm its size, which he said was between 1 and 2 mm per edge. Cree's EZBright chip (currently in production) measures 0.98 x 0.98 mm and is 100 micron high. The R&D device has a similar height, with the epitaxial layers transferred to a thin silicon substrate.

Measurements were made instantaneously, meaning the device would not experience a significant rise in junction temperature. Such a rise would decrease the measured performance values. A standard XR-E package (used for production devices) was used. Edmond said that a package with better thermal performance, for example metal-core based rather than ceramic, would be required for this chip. Thermal issues continue to be an area requiring constant improvement, particularly in devices with high current density.

Another area is "droop", or the fall-off of efficacy at higher drive currents. Edmond said the efficacy of this LED would be significantly higher, in fact almost double, at the standard 350 mA operating current. Possible solutions to "droop" might include adjusting the barrier layers in quantum well structures or the use of non-polar substrates for growth.

In future, will it be necessary to build significantly bigger chips? "It could be done, although there will be limitations relating to defect density and current saturation," says Edmond. "However, we will get so much light from a single chip [with edge size between 1 and 2 mm] that there probably will be no need."

After the latest series of R&D improvements have been moved into production, there is still plenty of room for further enhancement, says Edmond. "It should be possible to reach 150 lm/W at this current level [around 4 A] in 5-10 years," he says. "This will require a huge amount of work, but I think it can be done."

Due to phosphor technology, warm white is always likely to lag behind, but should reach 100 lm/W at least. At roughly 15 W, such an LED would produce 1500 lumens. In terms of both light output and efficacy, this will overshadow almost all other lighting technologies for most applications.