SiC Supports Blue LEDs and More

Ruth DeJule, Associate Editor -- Semiconductor International, 7/1/1999

The greatest impact of wide bandgap materials may be in photonics, according to Charles Weitzel, of Motorola's technical staff. While several technologies support high power chips -- Si bipolar, Si MOSFETs as well as wide bandgap devices -- blue light emitting diodes (LED) and lasers currently rely on gallium nitride (GaN). Since the status of the once-promising zinc selenide is uncertain, GaN appears the only option; it's shorter wavelength (410 nm), provides four times the data storage capacity of GaAs devices. Though the decision to use a GaN-based active region has been made, the choice of a su bstrate material is still being debated: sapphire or silicon carbide (SiC).

Sapphire has been used successfully as a substrate by notables such as Nichia Chemical and Hewlett Packard. Blue lasers from Nichia have demonstrated 10,000 hr lifetimes at 20°C, dropping to 1000 hrs at 50°C. Though not a show-stopper, sapphire substrates require two top surface contacts, which increases device complexity and somewhat offsets the cost savings of this economical material.

Rivaling the more established sapphire is SiC. Interest in SiC stems from its good thermal conductivity (better than copper) and homoepitaxial properties. According to Lynn Tellefsen, marketing director at MicroPatent (East Haven, Conn.), there has been nearly a 25% increase in worldwide patents covering SiC over the past two years, with 60% coming from Japan. Among these are inventions from Cree Research and Japan's Sharp Kabushiki Kaisha demonstrating the fabrication of blue LEDs using GaN-based structures, grown on SiC substrates (see Figure).

Fig 1 An LED structure grown on a SiC substrate allows for top and bottom ohmic contacts. (Source: MicroPatent)

Sharp began by investigating a GaN LED structure grown on a sapphire C-surface substrate. However, lattice mismatch of ~16% caused misalignment at the interface and subsequent device degradation at high temperatures, despite the addition of buffer layers. The close lattice matching between GaN and SiC made SiC substrates a better choice. The LED was grown on a 6H-SiC substrate and had an InGaN active region.

In terms of lattice constant and thermal expansion coefficient, AlN and AlGaN are almost perfectly matched to GaN, making them ideal buffers. But when an AlN layer was deposited on the SiC substrate, defect levels at the interface remained at ~10⁸ to 10¹⁰ cm^-2, only a slight improvement over sapphire substrates. Alternatively, a SiC epitaxial layer, 10 to 40X the thickness of the damaged region, proved better at absorbing grating defects and damage. A blue LED, emitting at 432 nm with threshold currents of 40 mA, was achieved.

Other substrate applications may be on the horizon for SiC. Power densities on SiC can be as much as 200X that of silicon, according to Tom Sullivan, president at Sullivan & Co. The resulting higher threshold to avalanche breakdown makes SiC an attractive alternative to silicon substrates for high power applications. Growing SiC expitaxial layers on a SiC substrate instead of Si eliminates a 20% difference in lattice constant and an 8% difference in thermal expansion coefficient. Because of the mismatch, SiC layers on Si substrates can warp. According to Sullivan, SiC substrates open the door to high power 15 kV devices for electric motors not currently addressed by solid state devices.

What remains to be sorted out is the selection of the most effective SiC material phase, alpha or beta. Alpha material consists of 170 polyforms separated by thermal dynamic differences, and many are available in single crystal form. Beta is the only cubic form of SiC, but is only available as a polycrystal. Micropipes that penetrate the surface can be as large as 5 µm; they are present in alpha wafers and absent in beta. Which will endure, only time will tell.