41.4 Plasma-Assisted Polishing

Concept of Plasma-Assisted Polishing

In the PCVM process, a damage-free surface can be obtained because of its chemical removal mechanism (Mori et al. 2000b). However, the ability of PCVM to achieve atomic-level flatness is lower than that of conventional polishing because surface atoms are removed by isotropic etching under the atmospheric-pressure plasma condition. Yamamura et al. have recently proposed a novel dry polishing technique named plasma-assisted polishing (PAP) that combines the atmosphericpressure plasma process with fixed abrasive polishing (Yamamura et al. 2011). In this technique, the irradiation of reactive plasma modifies the surface of a hard material to form a soft layer, and the subsequent polishing using a soft abrasive preferentially removes the soft layer. Atmospheric-pressure water vapor plasma is used to soften the 4H-SiC (0001) surface by oxidation in this study. The atmospheric-pressure plasma is generated by applying an RF (f =13.56 MHz) electric power, and helium-based water vapor (ca. 2 %) with a flow rate of 1.5 L/min is supplied as a process gas. Water vapor is introduced into the process gas by bubbling helium through ultrapure water (UPW), and its concentration is measured using a dew point meter (DPM). The copper electrode used for generating the plasma is covered with a closed-end quartz glass tube to prevent arc discharge through the generation of dielectric barrier discharge. In the mechanical removal part, which is separate from the electrode, a polishing experiment using a polishing film (φ8 mm) placed at the tip of a spindle was conducted. A specimen was placed on a rotary table, and surface modification by plasma irradiation and dry polishing using a polishing film were sequentially conducted. The coaxial electrode and polishing film were located at the same distance from the center of the rotary table. Thus, the polished field was not the whole surface of the specimen but a ring with a width of about 8 mm. We used a commercially available 2-in. n-type 4H-SiC (0001) 0.29o off wafer with a specific resistance of 0.110 Ω cm as the specimen, and the polishing film contained CeO2 abrasive particles with an average grit size of 0.5 μm that were fixed on a polyethylene terephthalate (PET) film using a polyester polyurethane adhesive. In contrast to CMP, PAP is a dry process without the use of slurry and/or chemicals. Therefore, this technique is promising as an environmentally benign finishing process for difficult-to-machine materials.

Finishing of 4H-SiC (0001)

Nanoindentation measurements were conducted to evaluate the hardness of the surface modified by the irradiation of water vapor plasma (Lucca et al. 2010). Figure 12a shows load–displacement curves of the 4H-SiC surface measured by the nanoindentation method. A Berkovich-type indenter made of diamond was used, and the number of measurement points and the maximum load for each sample were 5 and 0.5 mN, respectively. The hardnesses of the surface shown in Fig. 12b were calculated by the Oliver–Pharr method (Oliver and Pharr 1992). From this figure, it was found that the irradiation of water vapor plasma reduces the average hardness of 4H-SiC from 37.4 to 4.5 GPa, indicating that the irradiation of water vapor plasma is effective for forming a soft layer on the hard SiC surface.

Fig. 12 Nanoindentation measurement results of 4H-SiC (0001) surface. (a) Load-displacement curves of the as-received and water vapor plasma irradiated surfaces. (b) Hardnesses calculated by using the Oliver-Pharr method

Fig. 13 XPS spectra of the processed 4H-SiC surfaces, which correspond to (a) Si2p, (b) Si2p (in detail), (c) C1s. (i) as-received, (ii) after irradiated water vapor plasma (iii) after PAP using CeO2 abrasive, (iv) after SPM cleaning followed by HF dipping

Figure 13a–c show X-ray photoelectron spectra (XPS) of the 4H-SiC surface processed by PAP corresponding to Si2p, Si2p (in detail), and C1s, respectively. In each figure, (i), (ii), (iii), and (iv) denote the as-received surface, the surface after water vapor plasma irradiation without polishing, the surface after PAP, and the surface cleaned with a mixture of H2SO4 and H2O2 (SPM cleaning) after PAP, respectively. The composition of the SPM solution was H2SO4 (97 wt%):H2O2 (30 wt%) = 4:1, and the immersion time in the solution was 10 min. The SPM cleaning for the removal of organic and metal contaminants was followed by dipping in 25 wt% HF for 5 min to remove the oxide film. A peak corresponding to the Si-O bond (102.9 eV) can be observed in Fig. 13a (ii), b (ii). The peaks observed at 101.5 and 283.9 eV were identified as corresponding to the interface oxide Si4C4xO2 (Hornetz et al. 1994). These results indicate that the irradiation of water vapor plasma oxidized the surface of SiC, and it is considered that the oxidation species in this reaction system are hydroxyl radicals because strong optical emission from excited hydroxyl radicals was observed in the optical emission spectroscopy measurement of the plasma (Yamamura et al. 2011). Therefore, it is assumed that the formation of the oxide layer led to the decrease in hardness observed in the nanoindentation measurements. After PAP, the peak intensity of the Si-O bond decreased as shown in Fig. 13b (iii), and a peak from C–H/C–C bonds on the surface (285.0 eV) was observed, as shown in Fig. 13c (iii). It is considered that the detection of the C–H/C–C peak is due to the contamination of the adhesive component of the polishing film or ambient. This contamination can be easily removed by SPM cleaning as shown in Fig. 13c (iv). Figure 13b (iv) shows that the residual oxide was completely removed by HF dipping after SPM cleaning.

Fig. 14 AFM images of 4H-SiC (0001) surface. As-received surface: (a1) 13.40 nm p-v 1.31 nm rms (a2) 3.39 nmp-v 0.53 nm rms, After PAP with residual scratches: (b1) 2.26 nmp-v (b2) 0.20 nm rms, After PAP without scratch: (c1) 1.48 nm p-v 0.25 nm rms (c2) 1.74 nm p-v 0.26 nm rms

Figure 14 shows atomic force microscopy (AFM) images of 4H-SiC (0001) surfaces. In these figures, (a1) and (a2) show the as-received surface, and (b1), (b2) and (c1), (c2) show the surfaces processed by PAP using CeO2 abrasive, respectively. The as-received surface has many scratches and is highly undulating. In the initial stage of PAP smoothing, deep scratches remain as shown in Fig. 14b1, b2. However, the rest of the surface is very smooth. Finally, as shown in Fig. 14c1, c2, the whole surface has an atomically flat step and terrace structure, which corresponds to the inclination of the crystal plane (0.29), without any newly formed scratches.

Subsurface damage, such as microcracks and lattice strain, remaining on the SiC substrate causes the degradation of quality of the SiO2/SiC interface and epitaxial layer, which are formed to fabricate electronic devices. Therefore, an atomically smooth damage-free SiC substrate is essential for fabricating high-performance SiC power devices. The residual lattice strains of the surface observed before and after PAP were evaluated by reflection high-energy electron diffraction (RHEED) measurement. We evaluated two commercially available 4H-SiC wafers supplied by two different companies, which were sliced on-axis (0.29 tilt: sample 1) and off-axis (8 tilt: sample 2) from the (0001) basal plane. The acceleration voltage and beam current in the RHEED measurements were 15 kV and 20 μA, respectively. The inset of Fig. 15 shows the RHEED pattern of sample 1 after PAP. Kikuchi lines were clearly observed and the background intensity was lower than that of the as-received surface. These results indicate that both surface roughness and crystallinity are improved by PAP. Figure 15 shows the lattice constants of the two samples calculated from their diffraction spot patterns. The lattice constants of the surfaces processed by PAP approached the ideal a-axis value of 0.307 nm, which is indicated by a dashed line in Fig. 15, and positive strains (0.7–2.9 %) were observed in both as-received wafers.

Fig. 15 Lattice constants of as-received and PAP processed 4H-SiC (0001) surfaces obtained by RHEED measurements. Inset is a RHEED pattern of the PAP processed surface

Fig. 16 XTEM images of the surface processed by PAP. (a) Low magnification image. (b) Closeup image and schematic of 4H-SiC structure

Figure 16a, b shows the XTEM images of the surface processed by PAP using ceria abrasive for 1 h. Oxide layer having amorphous structure is not observed on the PAP processed surface. Since the hardness of ceria is almost the same with that of SiO2 (Shorey et al. 2000), the oxide layer formed by irradiation of water vapor plasma is preferentially removed by soft ceria abrasive as shown in Fig. 16a. It seems that moderate removal of the oxidized surface atom is very effective to obtain an atomically smooth surface. On the other hand, the deep contrast layer remains on the surface, and this layer is considered as an interface silicon oxycarbide layer (Si4C4-xO2) SiO2 (Hornetz et al. 1994). As a product of insufficient oxidation, it is very reasonable to assume that the hardness of Si4C4-xO2 layer is between that of SiC which is the base material and SiO2 which is the sufficient oxidation product. As the hardness of ceria is almost the same with that of SiO2, only the top amorphous SiO2 layer can be removed, and the silicon oxycarbide layer will remain on the surface. Figure 16b shows the close-up image of the PAP processed surface. A periodical well-ordered structure, which corresponds to the structure of 4H-SiC, is continuously observed from the bulk region to the top surface. These observation results lead to the conclusion that PAP technique enables us to obtain an atomically smooth surface of single-crystal SiC substrate without introducing crystallographical defect in the subsurface region.
These results lead to the conclusion that PAP finishing makes it possible to obtain an atomically smooth surface on a single-crystal SiC substrate without introducing crystallographical subsurface damage.