3.1.

Reaction-Sintered Silicon Carbide Sample Processed by Plasma Chemical Vaporization Machining

The MRR of RS-SiC in PCVM was obtained by SWLI measurement, as shown in Fig. 2. The RF power, compositions and flow rates of the process gases, and machining time were 100 W, He ($750\mathrm{ml}/\min$), ${\mathrm{CF}}_4$ ($5\mathrm{ml}/\min$), ${\mathrm{O}}_2$ ($5\mathrm{ml}/\min$), and 90 s, respectively. It could be calculated that the peak-MRR was $0.533\mu \mathrm{m}/\min$, and the volume-MRR was $2.78\times {10}^{-3}{\mathrm{mm}}^3/\min$. Relative to the present methods for machining RS-SiC, such as the single-point diamond turning,¹⁸ electrolytic in-process dressing grinding,¹⁹ and magnetorheological finishing,²⁰ the MRR in PCVM of RS-SiC is considerable. What is more, PCVM is a noncontact figuring technique and the material removal is conducted by chemical reaction, so we suppose that there will be no subsurface damage or residual stresses on the processed sample, which are common in the processed RS-SiC sample obtained by the present mechanical material removal methods. Therefore, PCVM can be treated as an efficient rapid figuring method to process RS-SiC.

Fig. 2

Calculation of the MRR in PCVM of RS-SiC sample obtained by SWLI measurement: (a) typical shape of the removal spot and (b) cross-sectional line of the removal spot.

The comparisons of surface roughness of the RS-SiC sample before and after PCVM are shown in Fig. 3. It can be found that the initial surface, which was obtained by diamond lapping, had a fine surface roughness rms 0.910 nm. However, from Fig. 3(a), it can be found that there were scratches on the surface, because the hardness of the diamond is higher than that of the RS-SiC. Furthermore, there were small holes on the surface as shown in Fig. 3(a), because the hardnesses of SiC and Si grains in the RS-SiC substrate differ, which resulted in a different micro-MRR among the different grains. After PCVM, the RS-SiC surface was deteriorated, and the surface roughness root-mean-square (rms) and roughness average (Ra) were 382.116 and 242.427 nm, respectively. The major reason for this phenomenon might be that the chemical reactivities of SiC and Si grains differ. The possible chemical reactions that occurred in the PCVM of RS-SiC are as follows:

Fig. 3

Comparisons of surface roughness of the RS-SiC sample before and after PCVM: (a) the initial surface obtained by diamond lapping and (b) the processed surface obtained by PCVM.

The comparisons of the surface morphology of RS-SiC before and after PCVM at the same position were conducted by SEM, as shown in Fig. 4. From Fig. 4(a), it could be found that there were many small scratches and holes on the initial surface generated by diamond lapping, which was in accordance with the characters exhibited in Fig. 3(a). The surface of the processed RS-SiC sample in PCVM was bumpy, as shown in Fig. 4(b). Irregular holes were formed at the region of initial Si grains, and the areas of initial SiC grains had relatively smooth surfaces. It could be concluded that the micro-MRR of Si grain was larger than that of SiC grain in the RS-SiC substrate. This phenomenon coincided with the fact that the MRR in PCVM of single crystal Si wafer is larger than that of single crystal SiC wafer under the same experimental parameters.²¹

Fig. 4

Comparisons of the surface morphology of RS-SiC sample at the same position obtained by SEM: (a) before and (b) after PCVM.

The details of the processed RS-SiC sample in PCVM are shown in Fig. 5, which were obtained by SEM observation under high magnification. The outlines of the newly exposed SiC grains were very clear in Fig. 5(a), especially for some large grains. Meanwhile, there were obvious boundaries among the newly revealed Si grain and the SiC grains around it, as shown in Fig. 5(b), and the Si grain had the same characters as the amorphous silicon. What is more, sizes of the SiC grains were shrunken, as shown in Fig. 5(c), because the material removal in PCVM was conducted from the external to the internal on a SiC grain. Furthermore, the connects among different SiC grains were weakened, and there were some almost dissociative SiC grains, as shown in Figs. 5(c) and 5(d), because the chemical reaction rate in the boundary region was larger than that in the central area. Finally, these bumpy structures on the processed RS-SiC could decrease the surface hardness, and the contacting areas between the polishing particle and the abraded sample could be increased. All these features were propitious to improve the surface quality and the MRR in the further PAP process.

Fig. 5

Details of the processed RS-SiC sample in PCVM obtained by SEM observation under high magnification: (a) the newly exposed SiC grains, (b) the newly revealed Si grains, (c) the shrunken SiC grains, and (d) the weakened connects among different SiC grains.

Quantitative analysis of surface morphology of the processed RS-SiC sample in PCVM was obtained by AFM, as shown in Fig. 6. Surfaces of the processed SiC grains in RS-SiC substrate almost remained in a plane, which could be judged from the profile of the cross-sectional line shown in Fig. 6(b). What is more, there were many holes formed at the initial Si regions, and depths of the holes varied from 20 to 60 nm, which further validated the characteristics in PCVM of RS-SiC sample.

Fig. 6

Quantitative analysis of surface morphology of the processed RS-SiC sample in PCVM by AFM: (a) surface morphology and (b) profile of the cross-sectional line.

From research on the MRR, surface quality, and surface morphology of the processed RS-SiC sample in PCVM, it could be concluded that PCVM could be treated as an efficient method for the rapid figuring of RS-SiC, and the surface quality should be further improved.

3.2.

Reaction-Sintered Silicon Carbide Sample Processed by Plasma-Assisted Polishing

The PCVM processed RS-SiC sample was further processed by PAP to improve the surface quality. The investigated RS-SiC sample was shown in Fig. 7. The process parameters in PCVM were summarized in Table 1. The size of the RS-SiC sample was $30\mathrm{mm}\times 30\mathrm{mm}$. The size of the processed area in PCVM was $20\mathrm{mm}\times 20\mathrm{mm}$, which was realized by controlling the apparatus in Fig. 1(a) in the raster scanning mode with the scan speed of $200\mathrm{mm}/\min$ and feed pitch of 1 mm.

Fig. 7

Illustrations of the initial surface, processed area in PCVM, and processed region in PAP.

Table 1

Experimental parameters in processing RS-SiC by PCVM.

The parameters in processing RS-SiC by the PAP were summarized in Table 2. The schematic diagram of the PAP process was shown in Fig. 8. The autorotation speed ${\omega }_1$ and revolution speed ${\omega }_2$ were adjusted to compare the machining results when other parameters were confirmed.

Table 2

The summarized experimental parameters in processing RS-SiC by PAP.

Fig. 8

Schematic diagram of the PAP process with the parameters corresponding to Table 2.

The generation rate of the oxide layer was established when the RF power, process gases, and their flow rates were confirmed. The MRR of the oxide layer could be controlled by adjusting the autorotation speed and revolution speed when the type of grinding stone (polishing particle) and its size, polishing head and its size, and eccentricity were assigned. In the experimental apparatus of PAP as shown in Fig. 1(b), there were five levels for both the autorotation motor and the revolution motor. The selected experiments were organized in 11 groups corresponding to the different collocations of autorotation speed and revolution speed, as shown in Table 3, which aimed to seek the optimal processing parameters.

Table 3

The summarized collocations of autorotation speed and revolution speed in PAP.

After PAP, the processed RS-SiC samples were observed by SWLI to evaluate the surface roughness, and the results were shown in Fig. 9. For the purpose of conveniently comparing the surface quality, the obtained surface roughness rms in Fig. 9 was summarized in Table 4. It could be found that the optimal collocation was group 2 (corresponding to ${\omega }_1=100$ and ${\omega }_2=75$), in which the obtained surface roughness rms was 2.186 nm, as shown in Fig. 9(b). Meanwhile, the worst result appeared in the group 11 (corresponding to ${\omega }_1=160$ and ${\omega }_2=715$), in which the achieved surface roughness rms was 13.080 nm, as shown in Fig. 9(k). The experimental results indicated that surface quality of the processed RS-SiC in PAP was obviously better than that of the surface obtained by PCVM [Fig. 3(b)] and slightly worse than that of the surface obtained by diamond lapping [Fig. 3(a)]. Meanwhile, there were no visible subsurface damages or scratches on the processed sample in PAP, because the hardness of the ceria abrasive grain was smaller than that of the RS-SiC substrate and almost equal to that of the oxide layer. Relative to the obtained ultrasmooth surface (roughness rms 0.629 nm) in the divided PAP (D-PAP) process,⁵ the achieved surface quality in the simultaneous PAP (S-PAP) could be improved by optimizing the process parameters. Meanwhile, the rate in plasma oxidation of RS-SiC was low and the oxide layer prevented the former oxidation of subsurface layer. What is more, in D-PAP, the oxidation and the polishing processes were divided, which indicated that the MRR in D-PAP was obviously smaller than that in S-PAP. Therefore, S-PAP was more suitable for the fine finishing of the RS-SiC sample.

Fig. 9

Evaluations of surface qualities of the processed RS-SiC in PAP corresponding to the different groups: (a) ${\omega }_1=70$ and ${\omega }_2=75$, (b) ${\omega }_1=100$ and ${\omega }_2=75$, (c) ${\omega }_1=200$ and ${\omega }_2=75$, (d) ${\omega }_1=100$ and ${\omega }_2=234$, (e) ${\omega }_1=130$ and ${\omega }_2=234$, (f) ${\omega }_1=160$ and ${\omega }_2=234$, (g) ${\omega }_1=70$ and ${\omega }_2=386$, (h) ${\omega }_1=200$ and ${\omega }_2=386$, (i) ${\omega }_1=130$ and ${\omega }_2=544$, (j) ${\omega }_1=200$ and ${\omega }_2=544$, (k) ${\omega }_1=160$ and ${\omega }_2=715$, and (l) the summarized data.

Table 4

The obtained surface roughness rms of the polished RS-SiC sample in each group.

The surface quality of the processed RS-SiC sample in PAP was determined by matching the MRR of oxide layer $v_1$ with its generation rate $v_2$. In this study, the plasma oxidation rate $v_2$ of the RS-SiC sample was confirmed by the experimental parameters listed in Table 2, and the MRR of the oxide layer $v_1$ could be controlled by adjusting the autorotation speed ${\omega }_1$ and the revolution speed ${\omega }_2$. Although the actual values of $v_1$ and $v_2$ are difficult to obtain, they could be confirmed under certain experimental conditions. If $v_1=v_2$, the generation of oxide layer entirely matched its removal, which could obtain a relatively smooth surface roughness, as shown in Fig. 9(b), and the PAP of RS-SiC could simultaneously obtain a high MRR and smooth surface in this condition. When $v_1<v_2$, the oxide layer could not be removed completely and the polishing process was equal to the polished silica (${\mathrm{SiO}}_2$) component by ceria abrasion, which was difficult to obtain an ultrasmooth surface, as shown in Fig. 9(a). Otherwise if $v_1>v_2$, after removal of the oxide layer, the ceria abrasive continued to polish the RS-SiC substrate. The ceria abrasive was seriously abraded, because the hardness of the ceria abrasive grain was distinctly smaller than that of the RS-SiC. The wear degree of the ceria abrasive grain was aggravated along with the increasing of ${\omega }_1$ and ${\omega }_2$, which resulted in a worse surface quality of the processed RS-SiC sample in PAP, as shown in Figs. 9(c)–9(k). The experimental results indicated that parameters in PAP of RS-SiC should be strictly selected, and the optimal parameters simultaneously could obtain high MRR and smooth surface.