Growth of Silicon-Doped and High Quality, Highly Resistive GaN for FET Applications

  • PDF / 294,199 Bytes
  • 5 Pages / 414.72 x 648 pts Page_size
  • 95 Downloads / 138 Views

DOWNLOAD

REPORT


897 Mat. Res. Soc. Symp. Proc. Vol. 395 0 1996 Materials Research Society

RESULTS AND DISCUSSION Growth of highly resistive GaN The growth of thin, high mobility GaN deposited directly on sapphire is not possible because of the large lattice mismatch between GaN and sapphire. Therefore, the first growth obstacle to be overcome before GaN FET structures can be grown is the ability to grow highly resistive GaN, which is also of high quality. Often times when the growth conditions are not optimized, highly resistive GaN can be grown. However, if this material is highly resistive because it is heavily compensated, then upon doping these films with Si, high mobilities are not obtained. We have been successful in obtaining high quality, highly resistive GaN by continuing to optimize parameters that we have previously shown to be critical for the growth of high quality GaN, e.g. nucleation layer temperature[6j, nucleation layer thickness, NH 3/TMG ratio, and GaN growth temperature[3]. For the MESFET structures described in this paper, a-plane sapphire (1120) was used as the substrate because we have previously shown that for our growth conditions, we can more easily obtain high mobility GaN films on a-plane sapphire over a wider range of conditions than the corresponding growth on c-plane sapphire[3]. Irrespective of whether c-plane or aplane sapphire is used as a substrate, the resulting GaN films are oriented with the c-plane parallel to the substrate. Characterization of highly resistive GaN Since it is difficult to make contacts to GaN when the carrier concentration is below low 10 6cmn-, Hall measurements cannot be used to characterize these highly resistive films. X-ray rocking curves are also not very useful since the FWHM of the (0002) peak does not show a direct correlation with the electrical properties. For GaN films grown that are typically 3pm thick, the FWHM (typically 250-350 arcsec) does not track changes in mobilities, and cannot distinguish the difference between low mobility compensated films and higher quality films. In addition, the FWHM of the (0002) peak cannot distinguish the difference between films with a high background carrier concentration (mid-1016cm 3 or higher) or highly resistive films. The mobility of the doped channel layers grown on the highly resistive GaN thick buffer layer is a powerful characterization tool and gives some indication of the quality of the underlying highly resistive GaN layer (high quality highly resistive GaN exhibits higher mobilities when doped than highly compensated lower quality GaN). In the present study, the highly resistive GaN that was grown can support a two-point probe voltage of >1000V without exhibiting breakdown. This does confirm that the GaN films are highly resistive (estimated to be approximately 10 °(i-cm from I-V measurements on samples with large area In contacts), but it cannot be used, however, to determine the difference between low quality heavily compensated GaN and high quality highly resistive GaN. Therefore, we have used low temperature