Homoepitaxial growth and investigation of stacking faults of 4H-SiC C-face epitaxial layers with a 1° off-angle

First, we studied the uniformity of the thickness and carrier concentration, and the surface morphology of 4-in. C-face epitaxial layers with a 1° off-angle to confirm that these fundamental properties are acceptable for device applications. We used 137 points and 56 points on the entire wafer to estimate the uniformity of the thickness and carrier concentration, respectively. There was little change in these properties when we varied such growth parameters as the in-situ H₂ etching depth and C/Si ratio. The typical uniformities of the thickness and carrier concentration were 1–2% (σ/mean) and 10–15% (σ/mean), respectively. The root mean square value of the surface roughness was less than 0.2 nm and there was no step bunching. It has been reported that there is no step bunching on C-face epitaxial layers because the surface energy is lower than that of a Si-face.^2,8) Therefore, C-face epitaxial layers have smooth surfaces even when the off-angle is 1°, although with Si-face epitaxial growth step bunching is easily generated when the off-angle is decreased.^15–17) These results suggest that all the epitaxial layers grown in this experiment had acceptable uniformities and surface morphologies for device applications.

Moreover, we confirmed that the background carrier concentration was reduced by varying the C/Si ratio. When we increased the C/Si ratio from 0.8 to 2.0, the background carrier concentration decreased from 2.1 × 10¹⁵ to 5.2 × 10¹⁴ cm⁻³ presumably due to the site competition effect.⁶⁾ We consider that this background carrier concentration generated at a C/Si ratio of 2.0 is acceptable for 3.3 kV MOSFETs.

We identified three types of SFs by performing PL measurements. Figure 1 shows the PL spectra of the three types of SFs, and their PL images are inset. The luminescence at about 500–600 nm in the PL spectra is emitted from the substrates due to impurities in the substrates. The PL images suggest that the three types of SFs have triangular shapes. The PL spectra reveal luminescence at 389 nm, which corresponds to the 4H-SiC bandgap. This luminescence at 389 nm is emitted from the substrates and 4H-SiC parts of the epitaxial layers around these SFs. In addition, Figs. 1(a)–1(c) show the luminescence at 539, 457, and 483 nm, respectively. These luminescences indicate that the three types of SFs are (a) 3C inclusions, (b) 8H-SFs and (c) $(3,5)$-SFs.^3,18) Thus, we found that these three types of SFs were generated in the C-face epitaxial layers with a 1° off-angle. Other types of SFs were rarely generated and their densities were much lower than those of 3C inclusions, 8H-SFs and $(3,5)$-SFs. In the following sections, we describe how we observed 3C inclusions, 8H-SFs and $(3,5)$-SFs, and investigated the dependence of the growth conditions on their densities to discuss the important factors related to decreasing these defects.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We observed a 3C inclusion using a SEM to look for the presence of some kind of trigger at its origination point. Figures 2(a)–2(c) show a surface SEM image of the 3C inclusion, a magnified bird's eye view of its origination point, and a cross-sectional bird's eye view of the origination point, respectively. Figure 2(a) shows that part of the 3C inclusion is brighter than that of the area without SFs presumably due to the difference in the carrier concentration.¹⁹⁾ It is clear that there is a particle at the origination point of the 3C inclusion as shown in Fig. 2(b). Figure 2(c) indicates that the 3C inclusion is generated from the particle. The brightness of the particle is the same as that of the 3C inclusion part, which emits the 3C-SiC luminescence shown in Fig. 1(a). Moreover, other reports show that the cause of 3C inclusions is 3C-SiC particles.^3,20) Therefore, the particle must be a 3C-SiC polytype.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We estimated the points at which the 3C inclusions were generated by multiplying tangent 1° by the length along the $[11\bar{2}0]$ direction obtained from the PL images. The generation points were defied as points located less or more than 1 µm from the interface between the substrates and epitaxial layers. The 3C inclusions were considered to be generated near the interface between the substrates and epitaxial layers when the generation points were located less than 1 µm from the interface, and to be generated during epitaxial growth when the generation points were located more than 1 µm from the interface. Figure 3 shows a schematic drawing of this method for estimating the generation point. As a result, both types of 3C-inclusions were observed, which suggests that the 3C inclusions were caused by 3C-SiC particles which were present on the substrates prior to epitaxial growth, or which had fallen on the substrates during epitaxial growth from the inside walls of the chemical vapor deposition reactor.

Zoom In Zoom Out Reset image size

We investigated the relationship between the in-situ H₂ etching depth and 3C-inclusion density, because we assumed that the particles which were present on the substrates before epitaxial growth were removed by in-situ H₂ etching. Figure 4 shows the 3C-inclusion density for the in-situ H₂ etching depth. The red and blue bars indicate the density of the 3C inclusions generated near the interface between the substrates and epitaxial layers and during epitaxial growth, respectively. The C/Si ratio was 2.0. The 3C-inclusion density is approximately 0.5 cm⁻² when the in-situ H₂ etching depth exceeds 0.4 µm, although the density is 1.6 cm⁻² when the depth is 0.1 µm. An in-situ H₂ etching depth of over 0.4 µm is needed to remove the 3C-SiC particles which were present on the substrates before epitaxial growth.

Zoom In Zoom Out Reset image size

In addition, we investigated the relationship between the C/Si ratio and the 3C-inclusion density; however, the 3C-inclusion density was not changed by varying the C/Si ratio from 0.8 to 2.0. As reported for Si-face epitaxial growth with a low off-angle of less than 1°, the 3C inclusion density increases as the C/Si ratio increases because 3C-SiC nucleation occurs due to the short diffusion length of adatoms and the wide terrace width caused by step bunching generation.^11,21,22) Compared with the reported Si-face epitaxial growth, the 3C-inclusion density on the C-face did not increase as the C/Si ratio increased. With C-face epitaxial growth, we consider that 3C inclusions, which were attributed to 3C-SiC nucleation occurred by the wide terrace width, were not generated since the terrace width was not increased by the suppression of step bunching because of its lower surface energy compared with a Si-face.⁸⁾

The 3C-inclusion density was not affected by the growth parameters but was affected by the chamber conditions. The 3C-inclusion density decreased when in-situ H₂ etching depth exceeded 0.4 µm because it removed the 3C-SiC particles that were present on the substrates before epitaxial growth.

We observed an 8H-SF and a $(3,5)$-SF using SEM, but did not observe any particles at their origination points in contrast to the 3C inclusions. Using the method described in Sect. 3.3, we found that almost all these SFs were generated near the interface between the substrate and epitaxial layer. In addition, the densities of both the 8H-SF and $(3,5)$-SF did not depend on the growth parameters of either the in-situ H₂ etching depth or the C/Si ratio.

It is possible that defects in substrates cause SFs in epitaxial layers. To investigate the dependence of the crystal quality of the substrates on these SF densities, we grew two epitaxial layers under the same conditions (in-situ H₂ etching depth of 0.4 µm and C/Si ratio of 2.0) on substrates cut from different boules. Figures 5(a) and 5(b) show microscopic images of the Si-face substrates cut from two different boules after KOH etching. Dislocations on the Si-face SiC materials reveal etch pits after KOH etching.²³⁾ The etch pit density obtained from Figs. 5(a) and 5(b) is 4.1 × 10³ and 8.3 × 10³ cm⁻². It means that the dislocation density of the substrate shown in Fig. 5(b) is slightly higher than that of the substrate shown in Fig. 5(a). Moreover, the substrate shown in Fig. 5(a) has many high dislocation density areas unlike the substrate shown in Fig. 5(b), which has evenly-distributed dislocations. We grew epitaxial layers on these C-face substrates. The epitaxial layer grown on the substrate shown in Fig. 5(a) had an 8H-SF density of 0.38 cm⁻² and a $(3,5)$-SF density of 0.43 cm⁻². The epitaxial layer grown on the substrate shown in Fig. 5(b) had an 8H-SF density of 0.05 cm⁻² and a $(3,5)$-SF density of 0.01 cm⁻². The 8H-SF and $(3,5)$-SF densities changed depending on the boules used as substrates even under the same growth conditions. We believe that these SF densities were affected by the dislocation distribution of the substrates.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We investigated the dislocations of a 9 µm epitaxial layer grown on the substrate shown in Fig. 5(a) around these SFs by using PL images detected using a long-pass (>750 nm) filter. Figure 6 shows the PL images of three positions of the epitaxial layer where 8H-SFs and $(3,5)$-SFs are generated at relatively high densities. 8H-SFs and $(3,5)$-SFs are outlined with red and orange dashed lines, respectively. The triangular defects with no marks are 3C inclusions. Both the epitaxial layer and the substrate were excited because we employed 295–370 nm Hg-lamp excitation through a bandpass filter and the penetration depth was approximately 100 µm at 370 nm.²⁴⁾ The dislocations in the substrates and epitaxial layers have dark and bright spots, respectively, in PL images detected using this long-pass filter.^25–28) However, the bright spots are not visible because of the low magnification measurement and weak luminescence of the dislocations in this study. In these figures, many dark spots are observed that correspond to dislocations in substrates. Areas with lower and higher dislocation densities are separated by a white dashed line and the areas inside the white dashed lines are the high dislocation density areas. The 8H-SFs and $(3,5)$-SFs are generated in or near the high dislocation density areas on the substrate. It is possible that the high dislocation density areas of the substrates rather than the dislocations themselves cause the 8H-SFs and $(3,5)$-SFs. It has been reported that nucleation of multiple 2D islands on the terraces may lead to the formation of faulted Si-C bilayers and in-grown SFs, and the nucleation is not connected to any other defect, inclusion, particulate, or any type of heterogeneity in the epitaxial layer or the substrate.²⁹⁾ In addition, it has been reported that a cluster of threading dislocations exist near the starting point of the 8H-SFs and such irregularity can increase the chance of a polytype error during epitaxial growth.³⁰⁾ Therefore, we think that the nucleation of multiple two-dimensional islands occurred in or near the high dislocation density areas because the diffusion length of adatoms became short as a result of the increase in dangling bonds. Further investigation of the 8H-SF and $(3,5)$-SF generation mechanism is needed but it is important to suppress generation of the high dislocation density areas on substrates if we are to reduce 8H-SF and $(3,5)$-SF densities.

Zoom In Zoom Out Reset image size