An Hrem Study of the Microstructure of Al Contact on GaN/AlN/SiC Thin Films
- PDF / 2,716,134 Bytes
- 7 Pages / 414.72 x 648 pts Page_size
- 26 Downloads / 208 Views
Mat. Res. Soc. Symp. Proc. Vol. 355 01995 Materials Research Society
simulation is used. In both structures there is an atom plane at the half-period (0.259nm for GaN) position of the (0001) plane group. This half-period atom plane is the same as the integral plane but shifted slightly. It results half-period fringes in the HREM images. The intensity of the half period fringes is equal to that of the integral fringes in the [ 1010] view. In the [ 1120] view, when the sample is not very thin and has a small amount of tilting toward the [0001] direction the half-period fringes show different contrast from that of integral fringes. This phenomenon can be seen in fig. 1 and 2 clearly and was confirmed by the simulation. Therefore, the contrast changes in the half-period (0001) fringes are solely a electron optical phenomenon and do not imply any local chemical or structural change. The single crystals in the AIN and GaN layers are not free of defects; they have a columnar structure which can be seen clearly in the figures. The columns are 10 to 40 nm in width and are perpendicular to the film plane. Some of the column boundaries in AIN layer start from the steps on SiC substrate surface. Many of them extend from the AIN layer into the GaN layer. Actually, the columns can be considered as grains slightly different in lattice orientation. Microdiffraction of the columns indicates that the orientation difference between adjacent columns may be as large as 20. The comparison of the images from the AD sample and the RTA sample indicates that the columnar morphology is not affected by the annealing. The annealing does have some effect on the AIN buffer layer. Moir6 fringes can be seen clearly in the AIN layer of the RTA sample (fig3a), but not in the AD sample (fig.3b). The direction and the spacing of the fringes in the RTA sample are not totally random. The inset of fig.3a schematically portraits these fringes. The appearance of the Moir6 fringes indicates that the annealing introduces some local structural change in the AIN layer. The microstructure of the GaN layer is more complicated than that of AIN layer. First, there is a quite high density of planar defects perpendicular to the [0001] direction. These caused [0001] streaking in the (1120) diffraction pattern and were directly observed in the HREM images as stacking faults. It is interesting that in the region next to the GaN/AIN interface there is a layer 35 nm thick with very few planar defects. Beyond this low-defect zone lots of stacking faults appear. (fig.3) Second, in the high-defect region, areas with lattice fringes different from the wurtzite fringes are seen in the HREM image. Fig.4 gives a closer view of such an image, in which the areas labeled by "Z' have different fringes from the areas labeled "W". The "Z" areas occupy about 3-4% of the total area. A reasonable explanation for the different lattice fringes is that there are two different structures of GaN, zincblende and wurtzite. Multislice simulation of the images of these two structures confirmed
Data Loading...
