Forbes (6-14-99) The laser blue(s)
forbes.com
...Silicon carbide's natural cleave points are ideal mirrors.
Sapphire, the material Nichia and most other laser developers use, is cheaper and easier to come by but more difficult to process correctly to provide the necessary mirrors.
These paragraphs are about lasers, for which the mirror is more critical, but any advantage of SiC over sapphire would also apply to LEDs. The article is intended for a business audience and only touches on the technical issues. After looking at Cree's structure, I have my doubts about this argument; it is not quite so simple, since they have additional layers in between.
Cree's SiC approach clearly has other advantages, such as high thermal conductivity and a vertical structure (à la HBT transistors) which is possible with a conducting substrate. The latter results in a smaller footprint. However, that may not be very important for an application like automotive lighting. They achieve a low defect concentration by means of intermediate buffer layers, which take up lattice mismatch, but this adds some complexity and cost back in.
The old proprietary buffer layer trick... <g>
A buffer layer described in one of their patents, US 5393993, is a discrete step transition comprised of two alloy layers with different mole percentages of two components, gallium nitride and aluminum nitride, then a pure AlN layer adjacent to the SiC.
Think of this as layers X | (100-a%)X+(a%)Y | (100-b%)X+(b%)Y | Y | SiC
a,b are 30,90 and X,Y are GaN,AlN in dependent claims covering a preferred structure.
The portion of the buffer layer adjacent to the lower GaN layer can be doped to have the same conductivity type as that GaN layer, so that the electrical path of the lower contact is mostly through this layer instead of through the substrate. Doesn't this suggest that while they must use SiC to grow high-quality GaN, it is not essential that it be a conductor?
WT |
