Wide Bandgap Materials Draw Interest

Ruth DeJule, Associate Editor -- Semiconductor International, 2/1/1999

The unique properties of wide bandgap semiconductor materials such as silicon carbide (SiC) and gallium nitride (GaN) are drawing greater interest as an alternative for enhancing device performance. The high breakdown voltage and saturated electron drift velocity make them attractive for a wide variety of high-frequency and switch-mode power devices. Silicon carbide's high thermal conductivity further adds to its appeal.

Both SiC and GaN are epitaxially grown. GaN is typically grown on SiC or sapphire (Al₂O₃) substrates, SiC on SiC substrates. Since SiC has a homoepitaxial layer, vertical as well as horizontal devices can be built. Therefore, Schottky and MOSFET vertical switch-mode power devices can be fabricated with SiC. GaN, having a heteroepitaxial process, is cu rrently limited to horizontal devices, only RF power devices. However, because a quantum well forms at the AlGaN/GaN heterojunction, very high sheet carrier densities can result, yielding higher current densities than in SiC.

The numbers have been impressive: maximum frequencies of oscillation (f_max) exceeding 100 GHz and power densities of 5.3 W/mm at 10 GHz with AlGaN HFETs, power densities of 3.3 W/mm at 850 MHz (CW) and 10 GHz (pulsed) with 4H-SiC MESFETs and output power of 450 W (pulsed) at 600 MHz and 38 W (pulsed) at 3 GHz with 4H-SiC SITs. The RF power densities of SiC MESFETs and AlGaN HFETs are superior to those of Si MOSFETs and GaAs MESFETs (see Figure).

Furthermore, the high breakdown voltage capability of SiC has been demonstrated in a number of switch-mode power devices such as Schottky diodes, PIN diodes, MOSFETs and GTOs. For 4H-SiC PIN diodes, breakdown voltages as high as 5.5 KV have specific on-resistances two orders of magnitude lower than those of comparable Si devices.

Despite the promising performance of these devices and technologies, several major challenges must be overcome before commercialization is possible, said Charles E. Weitzel, member of technical staff at Motorola (Tempe, Ariz.). He targeted three areas of concern. Clearly, the availability of material takes precedence. Several viable material suppliers must be able to provide high quality, low defect density, large area, standardized starting material at a reasonable cost.

Fig. 1. Comparison of RF power densities indicates better performance of GaN- and SiC-based devices.

SiC epitaxial layers on SiC substrates are typically fraught with micropipes that appear as holes through the substrate. Since the SiC epitaxial layer does not fill the holes, device defects result, impacting yield. Currently, the best micropipe densities are <1/cm². Similarly, GaN epitaxial layers can have dislocation densities in excess of 1 x 10⁹/cm². Surprisingly, they still make lasers. Since GaN is heteroepitaxial, good crystallographic matching to the substrate is essential. One method for reducing dislocations is lateral epitaxial overgrowth. Using a SiC substrate with stripes of SiO₂ or Si₃N₄, the GaN is grown from the substrate but not on the dielectric. However, lateral growth occurs over the dielectric and notably contains fewer dislocations. This process is repeated over the region grown from the substrate. The resulting film has a significantly lower dislocation density.

Once the material issues are resolved, fabrication alternatives must be developed. For example, SiC requires very high temperature processes such as implant anneals at 1500°C. Finally, new packaging technology will have to be developed. In these as well as all semiconductor devices, heat accelerates device failures. While SiC has high thermal conductivity, operating temperatures for high-power applications will quickly offset this advantage. To realize the full potential of wide bandgap devices, packaging to accommodate higher power densities and temperatures will be necessary.

The greatest impact of wide bandgap materials, however, may be in photonics, in particular, blue GaN LEDs and lasers. With a wavelength more than a factor of two smaller than GaAs, data storage densities can be quadrupled. Clearly, whether in the realm of electronic or optical devices, wide bandgap materials show distinct advantages. Could manufacturable solutions for these unique materials be far behind?