The thermodynamic calculation confirmed that Si, Si2C, Si2N and SiC2 are the main Si-containing species in the gas phase, and the importance of Si2C, SiC and SiC2 increased dramatically (one order of magnitude for Si2C, 2 orders of magnitude for SiC, and 2–3 orders of magnitude for SiC2) with temperature. Figure 1 displays the predicted change of mole ratio of Al (in AlN) to Si (in Si, Si2C, Si2N, SiC2, SiC, and SiN) in the gas phase with temperature from 1800 °C to 2000 °C. It is evident that the mole ratio of Al:Si increases with temperature. After reaching a maximum (at ~1910–1920 °C), it decreases gradually. The mole ratio of Al to Si in the gas phase was different from the original composition ratio of Al:Si in the source material; the gas phase contained a higher Al:Si ratio under most conditions.
Figure 2 shows optical micrographs of an AlN-SiC alloy crystal grown on 8° off-axis 6H-SiC substrate at 1860 °C for 24 h. The resultant AlN-SiC alloy crystal was colorless and transparent. The film was cracked, probably due to the large stress introduced during the cooling process from sublimation temperature to room temperature, which was caused by the thermal expansion mismatch between the SiC substrate and the AlN-SiC alloys. EPMA measurements confirmed that all four constituents, Si, Al, C, and N, were present in every region investigated, without any indication of phase separation of SiC and AlN or any segregation of individual elements.
Figure 3 displays a SEM image of an AlN-SiC alloy crystal grown on 8° off-axis 6H-SiC substrate at 1860 °C for 20 h. The AlN-SiC alloy was initially deposited in an island growth mode which was evident from the hexagonal pyramids with flat tops. The hexagonal grains then merged together to form a continuous layer.
AlN-SiC alloy crystals grown on off-axis SiC substrates were basically similar in morphology under optical microscope (as in Figure 2), independent of the different polytypes (6H or 4H) and miscut angles (8° or 3.68 ° ). They only differed slightly in the width of terraces and heights of steps. The growth developed layer by layer, with individual hexagonal grains around the sample edges. The steps originally present at the miscut SiC substrates provided the initial nucleation sites, leading to a three-dimensional nucleation of islands, followed by a two-dimensional layer-by-layer growth due to the island coalescence. In contrast, AlN-SiC alloy crystals formed on on-axis 6H-SiC substrate were rough with individual hexagonal islands all over the surface (demonstrated in Figure 4). Therefore, growth on on-axis SiC substrate was not further investigated.
The lattice constants for one AlN-SiC alloy were accurately measured by XRD. The alloy crystal grown on 8° off-axis 6H-SiC substrate at 1985 °C for 24 h (with the original mole composition Al:Si as 3:1) was ground to powder so as to obtain the complete XRD spectra. Figure 5 shows the XRD spectra of the ternary AlN-SiC alloy crystal and a binary AlN crystal. The diffraction peak of the AlN-SiC alloy crystal was shifted toward AlN, since AlN was the main component. The XRD patterns confirm that as expected, the SiC from the substrate was the 6H polytype. The lattice constants a and c measured for 6H-SiC were 3.083(9) Å, and 15.04(2) Å, which were somewhat altered from the values of a and c for a ground 6H-SiC substrate without AlN-SiC alloy layer, measured on the same instrument as 3.0818(7) Å and 15.111(4) Å. The AlN-SiC alloy had the wurtzite crystal structure. The lattice constants a and c calculated for the grown AlN-SiC alloy crystal were 3.098(7) Å and 4.996(9) Å, respectively. The appearance of SiO2 peaks was probably artifacts introduced by the grinding process.
Figure 6 shows the XPS spectra for the AlN-SiC alloy crystal grown on 8° off-axis 6H-SiC substrate at 1865 °C for 24 h. The measurement indicated the presence of Al, N, Si, C and O elements on the surface. Al 2p, N 1s peaks of the alloy crystal were compared with those of a pure AlN substrate, and Si 2p, C 1s peaks were compared with those of pure SiC substrate. Both Al 2p and N 1s peaks shifted to a slightly higher binding energy in the AlN-SiC alloy crystal in comparison to those in pure AlN substrate. However, C 1s peak shifted to a lower binding energy in the alloy crystal, whereas Si 2p peak shifted to a much higher binding energy than that in pure SiC substrate. A higher binding energy was generally associated with a higher positive oxidation state. The shift of Si 2p, C 1s, Al 2p, and N 1s peaks as compared with bulk SiC and AlN confirmed no phase segregation of AlN and SiC, and the formation of AlN-SiC alloy crystals.
Figure 7 shows the Si 2p XPS spectra for the AlN-SiC alloy crystal grown on 8° off-axis 6H-SiC substrate at 1850 °C for 24 h. The Si 2p spectrum was decomposed into four components at 99.07, 100.48, 101.25, and 102.13 eV. The employed decomposition parameters were as follows: the full width at half maximum (FWHM) was 1.34 eV, and the broadening function, 70% Gaussian. The 100.48 eV peak can be ascribed to Si-C, the 101.25 eV peak to Si-N, and the 102.13 eV to Si-O . The peak fitting results are listed in Table 1. In the AlN-SiC alloy crystal, Si was not only bonded to C, but also to N to a large extent. This provides further confirmation that AlN-SiC crystal was actually an alloy without phase separation of SiC and AlN. However, it is difficult to apply the same procedure to the Al 2p, C 1s and N 1s spectra to obtain quantitative results, because of the large number of overlapping peaks.
The lattice image of AlN-SiC alloy crystal (which was grown on 8° off-axis 6H-SiC substrate at 1985 °C for 24 h) at <\( < 11\bar 20 > \)> projection and its corresponding diffraction pattern shown in Figure 8 confirms that it was a wurtzite structured crystal. EDX measurements indicate that the crystals were really alloys of AlN and SiC, with a rough estimate of the SiC composition of several tens of percentage.
The origin and paths of defects were examined by TEM. Figure 9 demonstrates the dislocation density in the initial growth was on the order of 1010 cm−2. The threading dislocations bent over as the alloy layer became thicker, and propagated horizontally (as shown by Figure 10). The defect density was reduced significantly at the surface region to less than 108 cm−2. In a second sample repeated at the same growth condition, the dislocations propagated in a zigzag manner from the beginning. The dislocation density at the alloy surface was less than 106 cm−2, which exceeded the examination limit of cross-section sampling. It was 3 orders of magnitude lower than the dislocation density at the early stage of growth of 109 cm−2.
The composition of the alloy crystal should depend on the vapor phase composition, and the diffusion rate of each vapor species. Diffusion rates are generally inversely proportional to the molecular weight of species involved. Therefore, Al and Si should diffuse at rates with the same order of magnitude, with the diffusion rates of Si2C, Si2N and SiC2 much lower. Consequently, the overall diffusion rate of Al-containing species (only Al) should be higher than that of Si-containing species (Si, Si2C, Si2N, and SiC2). An approximately uniform composition was achieved from the interface to the crystal surface, as confirmed by both XPS and EPMA. When the source composition AlN:SiC was 3:1 (growth was repeated for three times under the same growth conditions of 1850 °C and 24 h), XPS measurements demonstrated non-stoichiometric ratios of Al:N and Si:C. The N concentration was always slightly higher than the Al concentration (5–15%). The ratio of Al:Si was higher than 3 (3.6–4.4), comparable to 4.0 in the gas phase as predicted by JANAF calculation in Figure 1. In comparison, the N concentration measured by EPMA was also higher than Al (10–14%), however, the ratio of Al:Si was lower than 3 (1.8–2.4), independent of the wide range of growth temperature from 1860 °C to 1990 °C. The Si concentration by EPMA differed significantly from that by XPS, with EPMA giving a much higher Si content. When the mole ratio of Al:Si in the source material was 5:1 and 2:1, the ratio of Al:Si in the alloy crystals changed with the growth temperature, and it was different from the original composition, which agreed well with the results calculated by JANAF data.