3.1 Morphology of H-etched SiC substrates

Our 6H-SiC(0001) substrates are prepared by H-etching, as illustrated in Figure 1. Figures 1(a) and (b) review results for singular substrates Reference Ramachandran, Brady, Smith, Feenstra and Greve[8]: The morphology of as-received substrates [Figure 1(a)] displays many polishing scratches. Following etching, full-unit-cell height (15 Å) steps, arising from an unintentional miscut of the substrate, are seen on the surface [Figure 1(b)]. On a larger scale the surface is found to break into low-angle facets, with neighboring facets having a different <1 

 0 0> step direction and the angle of each facet being less than the overall miscut of the wafer.

Figure 1. AFM images of 6H-SiC(0001) surfaces: (a) on-axis, as-received; (b) on-axis, H-etched; (c) off-axis, 3.5° towards [1 [1 

 0 0], H-etched; (d) off-axis, 3.5° towards [1 1 

 0], H-etched. Gray scale ranges are (a) 10 nm, (b) 0.7 nm, (c) 7 nm, and (d) 1 nm. Note the varying lateral scales.

Results for a H-etched substrate miscut by 3.5° towards [1 

 0 0] 0 0] are shown in Figure 1(c). In this case the surface is seen to consist of a series of ≈ 300 nm wide strips, with normal vector oriented along the miscut direction. These strips are separated by step bunches, typically 3 nm high. In contrast, results for a H-etched substrates miscut by 3.5° towards [1 1  0] are shown in Figure 1(d). We now observe half-unit-cell height (7.5 Å) steps separated by ≈ 12 nm wide terraces. The steps, on average, have normal vector oriented in the miscut direction. However, it is clearly seen in Figure 1(d) that on a small scale the steps have a different orientation, i.e. the step edges have a zig-zag or chevron morphology. The precise orientation of this small-scale structure of the step edges is difficult to discern from Figure 1(d), but we believe the normal vectors to be [0 1  0] and [1 0  0] (±30° from [1 1  0]) since (i) this is a low energy step orientation for SiC steps Reference Ramachandran, Brady, Smith, Feenstra and Greve[8], and (ii) the same step orientation is seen on GaN films grown on these substrates as shown below.

3.2 GaN films on H-etched SiC

Before discussing results for GaN films grown on vicinal substrates, we first review results for growth on singular material as shown in Figure 2 Reference Lee, Ramachandran, Sagar, Feenstra, Greve, Sarney, Salamanca-Riba, Look, Bai, Choyke and Devaty[7]. For samples grown with Ga/N ratio only slightly greater than unity, the surface is covered with pits [Figure 2(a)]. The pits are formed with facetted sidewalls and flat-topped (0001) oriented ridges, as illustrated by the linecut in Figure 2(a). From AFM linescans the sidewalls are found to have an angle of 30° ± 5° relative to the (0001) surface. From facet-derived features in RHEED patterns this angle is found more precisely to be 33° ± 2°. The facets thus correspond to {10

3} planes, which have an angle of 32.0° relative to the (0001) surface. This result is similar to that of Heying et al. who report {10onebar3} or {104} facets Reference Heying, Averbeck, Chen, Haus, Riechert and Speck[12]. An analysis of facet angles in the AFM image (of the type done e.g. in Ref. Reference Lutz, Feenstra, Mooney, Tersoff and Chu[13]) reveals that all six equivalent {103} facets are present in approximately equal amounts on the surface. As the Ga/N ratio is increased to about 1.1 fewer pits appear on the surface, and those which are present tend to merge into trenches which separate plateaus of atomically flat morphology [Figure 2(b)]. For Ga/N ratios near 1.3, isolated pits with typical separation of about 1 μm are seen in the morphology [Figure 2(c)]. TEM studies reveal that these pits are associated with dislocations intersecting the surface Reference Lee, Ramachandran, Sagar, Feenstra, Greve, Sarney, Salamanca-Riba, Look, Bai, Choyke and Devaty[7] Reference Lee, Sagar, Feenstra, Inoki, Kuan, Sarney and Salamanca-Riba[14]; some type of preferential film decomposition or reduced growth rate apparently occurs at these points. Finally, for growth under very Ga-rich conditions, Ga/N ratio above about 1.5, the morphology is flat with only monatomic steps revealed in AFM [Figure 2(d)]. On a large scale, the GaN film morphology follows that of the SiC, with low-angle facets on the surface Reference Lee, Sagar, Feenstra, Sarney, Salamanca-Riba and Hsu[10]. Gallium droplets are observed on these surfaces grown with high Ga/N ratios, in agreement with the observations of Heying et al. Reference Heying, Averbeck, Chen, Haus, Riechert and Speck[12].

Figure 2. AFM images of GaN(0001) films grown on H-etched 6H-SiC, with Ga/N flux ratios of (a) 1.05, (b) 1.1, (c) 1.3, and (c) 1.6. The gray scale ranges are 210 nm, 86 nm, 5 nm, and 5 nm for (a)-(d) respectively. The film pictured in (a) was grown at 800°C and the other films were grown at 750°C. In (a), a line cut taken between the arrows is shown in the inset.

As shown in Figure 2, the roughness of films grown with Ga/N flux ratios near unity is significantly greater than for those with higher Ga/N ratios. Rms roughness values are 35, 20, 4, and 3 nm for the images shown in Figs. 2(a)-(d) respectively. However, for these singular substrates, we find that even though the morphology at low Ga/N flux ratio is worse (i.e. rougher) than that at high Ga/N flux ratio, the structural properties of the low-flux-ratio films is much improved. In particular, the FWHM of asymmetric (11

0) x-ray rocking curves are about 2x less for films with flux ratio near unity compared to that of films with flux ratio of 1.3 - 1.5 Reference Lee, Sagar, Feenstra, Inoki, Kuan, Sarney and Salamanca-Riba[14]. TEM reveals an order-of-magnitude reduction in dislocation density for the low-flux-ratio films. The mechanism for this improvement in film quality is found to be the tendency of dislocations to cluster near the topographic minima, where their probability of annihilation is increased due to their closer proximity Reference Lee, Sagar, Feenstra, Inoki, Kuan, Sarney and Salamanca-Riba[14].

The surface morphology of our GaN films grown on vicinal SiC is pictured in Figure 3. These results can be seen to reflect the morphology of the underlying SiC substrates. The films grown on a substrate miscut towards [1 

 0 0] display similar size terraces (with [1  0 0] normal vectors) as the substrate, although these GaN strips are not completely continuous along their length [Figure 3(a)]. The films grown on the substrate miscut towards [1 1  0] display steps with average orientation having [1 1  0] normal, but with [0 1  0] and [1 0  0] oriented step edges seen on a smaller scale [Figure 3(b)]. Note however that these features are much larger than the corresponding ones seen in Figure 1(d), indicating significant step bunching and terrace growth during the GaN film growth. Rms roughness is 3 nm for both Figure 3 (a) and (b), similar to the values found for Figure 2 (c) and (d). In all cases those films are probably smooth enough so as to not to produce any deleterious effects in device processing or operation (e.g. room temperature mobility values in field-effect transistors would generally not be affected by this level of surface or interface roughness, although low temperature mobility values may be impacted Reference Monroe, Xie, Fitzgerald, Silverman and Watson[15]).

Figure 3. AFM images of GaN(0001) films grown on H-etched miscut SiC, with Ga/N ratio of 1.5 for both films. Substrates used are: (a) off-axis, 3.5° towards [1 

 0 0], and (b) off-axis, 3.5° towards [1 1 

 0]. Gray scale ranges are (a) 15 nm and (b) 12 nm. Linescan profiles taken across the middle of each image are shown below the images (a linear background subtraction has been applied to all images and linescans).

X-ray results for our films are listed in Table 1. We report FWHM of rocking curves acquired in both symmetric (0002) and asymmetric (10

2) configurations. The former reflects the screw dislocation density in the material and the latter reflects a combination of screw and edge dislocation densities. Actually, since in our material the edge dislocation density is 1−2 orders of magnitude larger than the screw dislocation density (and there are very few dislocations of mixed character), the (102) width reflects primarily the edge dislocation density. It is this (102) width which is the important parameter in Table 1 for distinguishing the various films.

The first two entries in Table 1 list films grown on on-axis SiC with Ga/N ratio near unity. These films have (10

2) widths of 400−500 arcsec. The surface morphology for those films is somewhat rough, as described above. The next three entries in Table 1 list films grown with higher flux ratios. The (102) widths are correspondingly larger, for the reason discussed in the previous section. The final four entries in Table 1 give results for growth on the vicinal substrates, for Ga/N ratios of 1.5. We find that the (102) widths for these films are significantly less than those for the films grown on singular substrates using the same flux ratios. In fact, the results for the films grown on vicinal substrates are comparable to those obtained for our best (i.e. low-flux-ratio) growth on the singular substrates, whereas the morphology of the growth on the vicinal substrates as seen in Figure 3 is improved compared to that on those low-flux-ratio films on the singular substrates.

For several of the films grown on miscut substrates listed in Table 1 we measured the x-ray rocking curves for two inequivalent orientations of the vicinal substrates relative to the incident beam. For [1 1

0] or [0 1 0] miscut directions the first value listed in Table 1 corresponds to a rocking curve scan with in-plane component perpendicular or parallel, respectively, to the miscut direction. The second value listed in Table 1 then corresponds to a scan direction with in-plane component rotated by 60° relative to the first. Not surprisingly some variation in the width is seen, implying that the mechanism for the dislocation reduction in these films does indeed involve some specific aspect of their vicinality. TEM measurements have not been performed to date, so no further information in available concerning the mechanism for dislocation reduction in these films. Recently Xie et al. Reference Xie, Zheng, Cheung, Ng, Wu, Tong and Ohtani[6] have reported results for dislocation reduction in GaN films grown on vicinal SiC. In their case the initial SiC substrates were not H-etched, so that their screw dislocation density for growth on singular substrates was very high. This value was greatly reduced for the growth on vicinal substrates, which they interpret as being due to nucleation, growth and coalescence of GaN islands at step edges rather than on terraces Reference Xie, Zheng, Cheung, Ng, Wu, Tong and Ohtani[6]. They also observed a reduction in edge dislocation density, similar to that reported here, although no specific mechanism was suggested for that reduction. Their results are thus consistent with those reported here, although their model for the reduction in screw dislocation density is not relevant for our situation of H-etched substrates in which our screw dislocation density is low for both singular and vicinal substrates.