Hi, Steve. The wafer news really is "big progress."
A conducting SiC substrate enables a vertical electrode geometry, but is not active as far as light emission is concerned. Lattice mismatch with GaN must be taken up by transition layers to achieve low defect concentration in the epitaxially grown active layers (as described in one of Cree's patents, US 5393993, one such structure consists of two discrete layers of AlN+GaN in different proportions). These layers must be conducting as well for vertical current flow. Buffer layers are also required with sapphire substrates. With homoepitaxy on a GaN substrate, the need for buffer layers between the substrate and the (AlIn)GaN device structure is eliminated.
The thermal conductivity of GaN is about the same as Si, four times higher than sapphire, and less than half that of SiC. With respect to dissipation of heat generated within GaN-based diodes, the thermal conductivity of GaN must be somewhat of a limiting property regardless of the substrate material.
It will probably take a while for Sumitomo and others to bring down the cost of GaN substrates. Initially they will market them for blue laser applications and even that won't be until next year:
Sumitomo Electric plans to confirm the practicality of single-crystal for epitaxial growth for lasers; to establish a mass production system which will include process technology; and to begin sales by 2001.
However, the Japanese have targeted GaN substrates for development, and now that they know it can be done, I'm sure they'll redouble their efforts.
I would expect those using techniques such as Nakamura's ELOG overgrowth technique to switch over, since that is just a complicated way of making a GaN substrate.
WT |
