AlN Substrates: Bigger And Better
Will CrystAL-N's two inch aluminium nitride substrates trigger UV LED market growth, asks Compound Semiconductor.
Having just unveiled two inch AlN substrates, CrystAL-N is already looking to grow four inch material. Credit: CrystAL-N
With the ultraviolet LED market poised for massive growth, AlN substrate manufacturers are moving quickly to ensure LED makers choose these native, lattice-matched wafers over today's readily available, cheaper option, sapphire.
As established US-based heavyweights Crystal IS and Hexatech make great strides in growing substrates for in-house manufactured devices, Germany-based CrystAL-N has just unveiled two inch bulk AlN that promises to boost the performance of deep UV LEDs.
The move spells incredibly good news for device manufacturers. Until now, most of the substrates offered to the open market have been one inch diameter wafer, and a shift to two inch material can only help to cut the cost of making deep UV LEDs.
For the University of Erlangen-Nürnberg spin-off, the result is something of a breakthrough. The move from one inch to commercially acceptable two inch substrates has taken years of research, but chief executive, Paul Heimann, is confident his company's product is now ready for market.
“Our wafers are definitely competitive and right now there aren't too many AlN vendors worldwide," he says.
As Heimann points out, US-based AlN substrate incumbants, Crystal IS and Hexatech, also manufacture LEDs, with Crystal IS saving all of its home-grown crystals for its own devices.
And while Hexatech also sells AlN substrates, Heimann asserts these are smaller - some one to one and a half inches in diameter - and cost around $5000 per substrate. CrystAL-N's larger, two inch versions hit a similar price point, coming in at Euro 4000, so arguably offer better value for money.
Crucially, as CrystAL-N moves forward, other competitors have scaled back activities or are at an earlier stage of development. “I believe Nitride Crystals has stopped [crystal growth] because of funding issues and I haven't seen any material from a new start-up in Kansas called Nitride Solutions," says Heimann.
But its not just about tracking the competition, quality-wise CrystAL-N's substrates also appear to make the grade. Dislocation densities are coming in at less than 1X105cm-2, on a par with smaller AlN substrates, and importantly, the two inches wafers are UV transparent, helping to boost the light output through the substrate of the final LED. Hexatech currently offers one inch UV transparent substrates.
According to Heimann, many competitors grow crystals via physical vapour transport in tantalum carbide-based furnaces, which reduces transparency.
“Here the carbon content is so high that the material is no longer transparent at 300nm UV wavelengths," he says. “I know our competitors take the substrate, make the device on it and then etch away the substrate. This is a real pity as you struggle to produce such a perfect material and then you etch or grind it all away."
In contrast CrystAL-N grows its crystals in a tungsten furnace. A 3mm thick AlN-SiC layer is first grown on a low defect SiC seed wafer. The SiC seed is then cut away with the free-standing mixed crystal AlN-SiC layer then used as a seed for pure AlN growth. All in, growth takes up to ten days, and using the tungsten furnace circumvents the carbon contamination from tantalum carbide furnaces that plagues the competition.
“In our regime, crystal growth is aluminium polar, independent of the polarity of the seed," explains Heimann. “I see from [a competitor's] publications, crystal growth is nitrogen polar giving really nice looking crystals. Their specifications and XRD data look good and as a crystal grower, I like the shape of the crystals, but they are not UV transparent."
While Heimann will not be drawn on the performances of structures grown on his company's substrates, past tests on prototypes have shown “good" light output.
“I do know that competitor's devices have a similar light output to sapphire-based LEDs, but the lifetime is much better," he says. “I think it's down to basic physics. If you want nice devices with a high aluminium content, take a native seed wafer, do your epitaxy and you'll have less dislocations and defects, and a better output and lifetime. AlN will produce better results, but there is still work to do on the epitaxy."
However, CrystAL-N has now set its sights on four inch wafers. Comparing his native substrate's roadmap to the path followed by SiC - four inch diameters are commonplace and six inch substrates are emerging - Heimann is confident four inch bulk AlN is the future.
“It will take some time but technically speaking I see no limitations to getting to four inch diameters," he says. “Perhaps we will have to apply six inch SiC seeds, so we will need bigger furnaces. But now we've reached two inches, we won't stop."
But while CrystAL-N focuses on delivering bigger and better AlN substrates, will these efforts trigger wider market adoption of the more exotic, sapphire alternative? Pars Mukish, LED technology analyst at Yole Développement, echoes Heimann's words on epitaxy, stating issues here must be dealt with first.
“It's really difficult to make [deep] UV-C devices," he says. “The main epitaxy technology is HVPE, which is not as developed as the traditional MOCVD used for visible LEDs."
To increase HVPE experience and address epitaxy issues, both Hexatech and Crystal-IS have established vertically integrated operations, growing AlN substrates and using these to manufacture deep UV LEDs. CrystAL-N didn't comment on whether it would follow the same route, but industry sources suggest the company is forging partnerships with chip companies in this field.
In the meantime, the deep UV LED industry is blossoming, albeit slowly. Established electronics heavyweights including LG Innotek and Samsung are eyeing the market as new entrants join, which, could kick-start the market.
But as Mukish cautions: “If [these developments] trigger the market within the next two or so years, CrystAL-N can remain manufacturing only substrates. If not, they may well have to integrate vertically or close."