The Silicon-Sapphire Breakup

Anne L. Fischer, Senior Editor

Traditionally, LEDs are fabricated by growing heterostructures on either sapphire or silicon carbide substrates, but each of these has its own set of problems. Sapphire provides a poor lattice match and low thermal conductivity, silicon carbide is expensive, and both are difficult to dice. Naturally, this leads to the quest for a suitable replacement.

A group from Purdue University in West Lafayette, Ind., looked at silicon as a possible solution. It is fairly inexpensive, easily diced and thermally conductive. It also is easy to scale up. But the problems are a 20 percent mismatch with GaN and absorption of visible light. Also, when heated, it expands at a very different rate than GaN does.

Two 2-in. silicon wafers with zirconium nitride metal deposition reflect the ambient fluorescent lighting at Birck Nanotechnology Center at Purdue University. On the left is a zirconium nitride metal film showing a more silverlike reflectance than the stoichiometric zirconium nitride on the right. These wafers are ready for LED device growth.

As an alternative, the group looked at adding an intermediate layer of zirconium nitride. One problem, however, is that it is unstable when placed next to silicon. The researchers, led by Timothy D. Sands, the Basil S. Turner professor of materials engineering and electrical and computer engineering at Purdue, placed an insulating layer of aluminum nitride between the silicon substrate and the zirconium nitride. The layer was applied using reactive sputter deposition, a process in which a metal target is bombarded by argon ions, knocking the atoms from the target and transferring them to the substrate. Nitrogen gas in the vacuum chamber reacts with the sputtered aluminum and zirconium, resulting in the deposition of aluminum nitride and zirconium nitride onto the silicon surface. Then, using vapor phase epitaxy, gallium nitride was deposited at temperatures of about 1000 °C (1800 °F).

The investigators found that, as the metals were applied to the silicon, they arranged themselves into a crystalline structure – matching that of silicon. Two remaining challenges include figuring out how to reduce defects in the devices and preventing the gallium nitride layer from cracking as the silicon wafer cools down after manufacturing. Sands noted that they found the same spectrum of defects as when they used sapphire. What happens, he explained, is that when you cool down the wafer, the gallium nitride contracts faster than the silicon and can crack. But he sees these as minor issues compared with successfully determining the right metal to layer upon the silicon surface.

He believes that the use of silicon will allow the industry to mass-produce LEDs in a volume not possible with sapphire. And with thermal considerations always a challenge in traditional LED design, “extracting heat is critical for high brightness for any kind of lighting application,” he pointed out.

But cost also is important. Where today you can buy an LED spotlight for about $100, people just don’t want to spend that, according to Sands, “even though the life-cycle savings is there.” He noted that university-based researchers, together with industry, national laboratories and the US Department of Energy, are working on bringing the cost down to the $5 to $6 level. The cost of the substrate is critical to achieving this – and new designs, such as more colors coming from a single chip, will bring down the cost of white light.

Sands said he expects to see affordable LEDs on the market in 2015.