Chemical–Mechanical Polishing of 4H-SiC Using Multi-Catalyst Synergistic Activation of Potassium Peroxymonosulfate

Abstract

This study optimized the proportions of synergistic catalysts to efficiently activate potassium peroxymonosulfate (Oxone), generate more reactive oxygen species, and accelerate the chemical oxidation of silicon carbide (4H-SiC) wafers during chemical–mechanical polishing (CMP) for an improved material removal rate (MRR) and surface quality. The Oxone was activated using ultraviolet (UV) catalysis with a photocatalyst (TiO₂) and transition metal (Fe₃O₄) to enhance the oxidation capacity of the polishing slurry through the production of strong oxidizing sulfate radicals (${\mathrm{S}\mathrm{O}}_4^{·-}$). First, the effects of the TiO₂, Fe₃O₄, and Oxone concentrations on the MRR were studied by conducting multiple single-factor experiments. Next, 4H-SiC wafers were polished using different catalyst combinations to verify the synergistic activation of Oxone by multiple catalysts. Finally, the roughnesses, physical features, and elemental compositions of the wafer surfaces were observed before and after polishing. The results showed that CMP with a TiO₂ concentration of 0.15 wt%, Fe₃O₄ concentration of 0.75 wt%, and Oxone concentration of 48 mM decreased the wafer surface roughness from Sa 134 to 8.251 nm and achieved a maximum MRR of 2360 nm/h, which is significantly higher than that associated with traditional CMP methods. The surface of a 4H-SiC wafer polished using CMP with the optimal catalytic system was extremely smooth with no scratches and exhibited many oxides that reduced its hardness. In summary, the proposed UV-TiO₂-Fe₃O₄-Oxone composite catalytic system for 4H-SiC CMP exhibited significant synergistic enhancements and demonstrated excellent surface quality, indicating considerable potential for the polishing of hard materials.

1. Introduction

Single-crystal silicon carbide (SiC) is widely utilized in various fields, such as aerospace, industrial power sources, and new-energy vehicles, because of its exceptional characteristics, which include a wide bandgap, high critical breakdown strength, high carrier saturation mobility, superior thermal conductivity, and excellent high-temperature and radiation resistance. However, stringent control of SiC substrate surface quality is required as it directly affects device performance. Despite the numerous advantages of SiC, its high hardness, brittleness, and chemical inertness pose significant challenges to achieving efficient and high-quality surface polishing through traditional chemical or mechanical processing methods. In addition, the Si-face of SiC is more suitable for subsequent epitaxial film growth compared to the C-face and it poses greater challenges for processing removal [1]. Therefore, investigating an efficient and low-damage polishing technique for the Si-face SiC wafer assumes paramount significance.

Chemical–mechanical polishing (CMP) is a pivotal technology for achieving wafer planarization. However, it encounters numerous challenges when applied to SiC workpieces, including the introduction of surface and subsurface damage, difficulties in post-polishing cleaning, and a low material removal rate (MRR). Researchers have proposed and explored various auxiliary CMP techniques, such as photocatalysis, electrochemical polishing, and plasma-assisted polishing (PAP), to address these issues [1]. According to Deng et al. [2], electrochemically assisted CMP under anodic oxidation significantly reduced the surface hardness of 4H-SiC; however, this approach required a high doping concentration, limiting its application to different SiC materials. Furthermore, Yamamura et al. [3] noted that although PAP enhances polishing efficiency, the practical application of this technique is hindered by the complexity of its operational process.

Researchers have attempted to reduce the complexity and cost of CMP by employing hydroxyl radical-based advanced oxidation processes to enhance the SiC substrate oxidation rate. Although the employed Fenton reaction has demonstrated potential for use in CMP, its further application is limited by the instability of the hydrogen peroxide (H₂O₂) required by these processes [4,5,6]. In contrast, sulfate radical-based advanced oxidation processes (SR-AOPs) can effectively degrade organic pollutants and produce ${\mathrm{S}\mathrm{O}}_4^{·-}$, which possesses a longer half-life and a broader pH range of applicability than ∙OH, thereby offering greater potential in the CMP of SiC [7]. Indeed, Wang et al. [8] investigated the CMP of SiC substrates based on the SR-AOP principle and discovered that ${\mathrm{S}\mathrm{O}}_4^{·-}$ significantly enhanced the MRR.

Among the available oxidizing agents, potassium peroxymonosulfate (Oxone) is a stable white crystalline solid that is nontoxic, inexpensive, and easily soluble in water [7,9]. So et al. [10] conducted experiments on the degradation of naphthalene catalyzed by transition metals and ultraviolet (UV) light-assisted activation of Oxone and reported that the Fe²⁺ ions from the transition metal catalyzed the activation of Oxone to generate ${\mathrm{S}\mathrm{O}}_4^{·-}$; the degradation efficiency peaked when the molar ratio of Fe²⁺ to Oxone was 1:2. Lu et al. determined that a dual-catalytic system comprising photocatalysts and transition metals under UV light significantly increased the concentration of active radicals in the polishing slurry to 6 times that of a single-catalytic system and 1.38 times that of a catalytic system using only UV light with a photocatalyst. Furthermore, Rastogi et al. [11] discovered that the efficiency of Oxone activation by Fe²⁺ gradually decreased as the pH increased. Finally, Chen et al. [12] reported that the auxiliary effect of UV light inhibited the concentration of Fe³⁺ in the polishing slurry, thereby promoting the efficiency of Oxone catalysis by Fe²⁺ to increase the concentration of active radicals in the solution. This is critical because the Fe³⁺ generated by the Fenton reaction spontaneously hydrolyzes in an aqueous solution, forming monohydroxy complexes such as Fe(OH)²⁺ that directly generate ferrous ions and ∙OH through photosensitization as follows.

$${\mathrm{Fe}}^{3+}+{\mathrm{OH}}^{-}+\mathrm{h}\mathrm{\nu }\to {\mathrm{Fe}}^{2+}+•\mathrm{OH}$$

(1)

$$\mathrm{Fe}{(\mathrm{O}\mathrm{H})}^{2+}+\mathrm{h}\mathrm{\nu }\to {\mathrm{Fe}}^{2+}+•\mathrm{OH}$$

(2)

This study accordingly applied the SR-AOP principle to propose a green, efficient polishing slurry oxidation system employing Oxone, which is more stable than hydrogen peroxide, activated by UV irradiation of a TiO₂ photocatalyst with a Fe₃O₄ transition metal for the CMP of SiC wafers. The primary objectives of this study were to use this approach to improve the MRR of SiC wafer CMP and enhance the resulting surface quality. First, multiple single-factor experiments were conducted to investigate the effects of the TiO₂, Fe₃O₄, and Oxone concentrations on the MRR. Next, the feasibility of Oxone activation through the synergistic action of multiple catalysts was verified by comparing the effects of different catalyst combinations. Finally, detailed analyses of the surface quality and chemical composition of each SiC wafer before and after polishing were performed using X-ray photoelectron spectroscopy (XPS), scanning electron microscopy (SEM), and white light interferometry (WLI). The results not only indicate that the proposed multiple-catalyst oxidation method can enhance the processing efficiency of SiC, but also offer new insights and solutions for the surface treatment of similar difficult-to-process materials.

2. Experimental Approach

2.1. Activation of Oxone and Removal Mechanism of 4H-SiC

The Oxone activation process under synergistic UV-TiO₂-Fe₃O₄ catalysis is illustrated in Figure 1. The transition-metal-activated Oxone reaction process is shown in Figure 1a. This reaction is primarily catalyzed through an electron transfer mechanism between the peroxymonosulfate anion (${\mathrm{H}\mathrm{S}\mathrm{O}}_5^{-}$) provided by Oxone and Fe²⁺ ions, resulting in the generation of highly oxidizing ${\mathrm{S}\mathrm{O}}_4^{·-}$ as follows [13]:

$${\mathrm{Fe}}^{2+}+{\mathrm{HSO}}_5^{-}\to {\mathrm{Fe}}^{3+}+{\mathrm{SO}}_4^{·-}+{\mathrm{OH}}^{-}({\mathrm{k}}_1=3.0\times {10}^4{\mathrm{M}}^{-1}{\mathrm{s}}^{-1})$$

(3)

In Equation (4), Fe³⁺ ions are generated by the decomposition of Fe₃O₄ particles and undergo a secondary reaction with ${\mathrm{H}\mathrm{S}\mathrm{O}}_5^{-}$ to produce Peroxysulfate (${\mathrm{S}\mathrm{O}}_5^{·-}$) as follows [14]:

$${\mathrm{Fe}}^{3+}+{\mathrm{HSO}}_5^{-}\to {\mathrm{SO}}_5^{·-}+{\mathrm{Fe}}^{2+}+{\mathrm{H}}^{+}$$

(4)

In Equation (5), this ${\mathrm{S}\mathrm{O}}_5^{·-}$ can react with itself to generate ${\mathrm{S}\mathrm{O}}_4^{·-}$ as follows:

$${\mathrm{SO}}_5^{-}+{\mathrm{SO}}_5^{-}\to 2{\mathrm{SO}}_4^{·-}+{\mathrm{O}}_2$$

(5)

However, the rate of the reaction in Equations (4) and (5) is significantly lower than that in Equation (3). Therefore, the concentration of Fe²⁺ ions in the polishing slurry gradually decreases as the reaction progresses, thereby limiting the efficiency of free radical generation [15,16].

The photocatalytic activation of Oxone is illustrated in Figure 1b. Upon exposure to UV light, electrons in the valence band of the nano-TiO₂ particles are excited and transferred to the conduction band, resulting in the formation of free electrons (e⁻) and generation of positive holes (h+) in the former [6,16], which is expressed as follows:

$$\mathrm{S}\mathrm{e}\mathrm{m}\mathrm{i}\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{t}\mathrm{o}\mathrm{r}\mathrm{+}\mathrm{h}\mathrm{v}\to {\mathrm{e}}_{\mathrm{C}\mathrm{B}}^{-}+{\mathrm{h}}_{\mathrm{V}\mathrm{B}}^{+}$$

(6)

These free electrons participate in the decomposition of Oxone, generating ${\mathrm{S}\mathrm{O}}_4^{·-}$ or ∙OH as follows [17,18]:

$$\mathrm{H}\mathrm{S}{\mathrm{O}}_{\mathrm{5}}^{-}+{\mathrm{e}}_{\mathrm{C}\mathrm{B}}^{-}\to \mathrm{S}{\mathrm{O}}_{\mathrm{4}}^{·-}+\mathrm{O}{\mathrm{H}}^{-}\mathrm{o}\mathrm{r}\mathrm{S}{\mathrm{O}}_{\mathrm{4}}^{2-}+•\mathrm{OH}$$

(7)

The generated holes (h⁺) react with Oxone to produce ${\mathrm{S}\mathrm{O}}_5^{·-}$ as follows [18,19]:

$$\mathrm{H}\mathrm{S}{\mathrm{O}}_5^{-}+{\mathrm{h}}_{\mathrm{V}\mathrm{B}}^{+}\to \mathrm{S}{\mathrm{O}}_5^{·-}+{\mathrm{H}}^{+}$$

(8)

and the ${\mathrm{S}\mathrm{O}}_5^{·-}$ reacts with itself to form ${\mathrm{S}\mathrm{O}}_4^{·-}$ as per Equation (5) [20].

The material removal model for the CMP of SiC is illustrated in Figure 2. Under the Fe₃O₄-Oxone and UV-TiO₂-Oxone catalytic systems, activated Oxone generates highly oxidizing ${\mathrm{S}\mathrm{O}}_4^{·-}$ and ·OH radicals that attach to the TiO₂ and Fe₃O₄ particles. Under the squeezing and rotating actions of the polishing pad and head, these particles uniformly adhere to the surface of the SiC wafer where they undergo oxidative reactions that transform the hard and difficult-to-process SiC surface material into a softer and more easily removable SiO₂ oxide layer. The abrasive in the polishing slurry subsequently removes this soft layer through mechanical action. The alternating effects of mechanical removal and chemical reactions enable the precise and efficient polishing of SiC.

2.2. Preparation of Polishing Slurry

The 4H-SiC specimens used in these experiments were procured from Hangzhou Qianjing Semiconductor Co., Ltd. (Hangzhou, China) and their properties are listed in

Table 1. Parameters of 4H-SiC specimens.

Parameter	Value
Diameter (mm)	50 ± 2
Thickness (um)	350 ± 25
Surface roughness (nm)	110.844
Range of resistivity (Ω·cm)	0.015–0.028
Doping element	Nitrogen

. The diamond abrasive particles were purchased from Dongguan Lizhi Abrasive Technology Co., Ltd. (Dongguan City, China) and had an average particle size of 100 nm. The P25 TiO₂ used as the photocatalyst was obtained from Shanghai Macklin Biochemical Technology Co., Ltd. (Shanghai, China) and had an average particle size of 20 nm. The Fe₃O₄ used as the transition metal catalyst was also acquired from Shanghai Macklin Biochemical Technology Co., Ltd. (Shanghai, China) and had a particle size of 100 nm. The Oxone used as the oxidant was purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). The sodium hexametaphosphate ((NaPO₃)₆) used as the dispersant was obtained from Shanghai Titan Technology Co., Ltd. (Shanghai, China). The KOH used as an acid–base regulator was purchased from Guoyao Group Chemical Reagent Co., Ltd. (Shanghai, China). Finally, laboratory-produced deionized water was employed as the base for the polishing slurry.

The basic polishing slurry consisted of 1 wt% diamond abrasive particles and 0.06 wt% (NaPO₃)₆ with the remainder comprising deionized water. First, these components were mixed and subjected to ultrasonic treatment for 5 min to ensure uniform dispersion of the abrasive particles. Next, the TiO₂, Fe₃O₄, and Oxone were weighed according to the required concentrations and added to the slurry. Finally, a 10% mass fraction potassium hydroxide (KOH) solution was gradually added while continuously monitoring the pH of the slurry until it reached a value of approximately 3.

2.3. Polishing Experiments

The polishing experiments utilized a Shenyang Kejing UNIPOL-1200S (Shenyang kejing automation equipment Co., Ltd, Shenyang, China) automatic grinding and polishing machine to perform CMP of different 4H-SiC wafer specimens. A schematic of the experimental setup is shown in Figure 3. First, each wafer was affixed to the center of the upper polishing disk using paraffin wax, and then the upper polishing disk was adjusted to align the wafer with the edge of the lower polishing disk. A 175 W mercury lamp irradiated UV light with a wavelength of 365 nm onto the polishing pad and slurry outflow from approximately 3 cm above the polishing pad. The UV light source was preheated for 3 min before each experiment to ensure a stable wavelength. All experiments maintained the process parameters listed in

Table 2. Experimental conditions.

Parameter	Value
Polishing pad	Polyurethane
Abrasive	100 nm diamond
Rotational speed of the polishing head (rpm)	30
Rotational speed of the polishing pad (rpm)	60
Polishing slurry flow rate (mL/min)	80
Load (kg)	12
pH	3

. The feasibility of multi-catalyst synergistic Oxone activation was verified by altering only the catalyst concentrations in the experiments without changing the concentrations of the other components in the polishing slurry. Thus, three sets of single-factor experiments were established to investigate the impacts of the TiO₂, Fe₃O₄, and Oxone concentrations on the MRR using the parameters listed in

Table 3. Parameters of single-factor experiments.

Experiment Group	TiO2 Concentration (wt%)	Fe3O4 Concentration (wt%)	Oxone Concentration (mM/L)
1	0.05, 0.1, 0.15, 0.2	0.5	32
2	0.1	0.25, 0.5, 0.75, 1	32
3	0.1	0.5	16, 32, 48, 64

. Each experiment applied a short-duration CMP treatment of 30 min to ensure the efficiency of the polishing process and quality of the resulting wafer surface. The polishing pad surface was conditioned for 10 min prior to each experiment using a conditioning ring. After polishing, the SiC wafer was removed by from the polishing disc by heating and subsequently subjected to repeated ultrasonic cleaning in anhydrous ethanol, and then dried using high-purity N2 before observation.

A precision balance (MSA225S-CE; 0.01 g accuracy) was employed to measure the change in workpiece mass three times before and after polishing, and the average value was applied to calculate the MRR as follows:

$$MRR={\displaystyle \frac{m_1-m_2}{\rho \times S\times T}}$$

(9)

where m₁ (g) denotes the initial mass of the 4H-SiC wafer prior to polishing; m₂ (g) denotes the mass of the 4H-SiC wafer after polishing; ρ (3.2 g/cm³) denotes the density of 4H-SiC; S (cm²) denotes for the surface area of the wafer; and T (h) denotes the polishing duration.

Finally, the chemical compositions of the wafer surfaces were analyzed before and after polishing using XPS (Shimadzu-KRA, TOS, Kyoto, Japan), and the surface morphologies of the wafers were observed before and after polishing using SEM (SU8010, HITACHI, Tokyo, Japan) and WLI (Super View W1, Chotest, Shenzhen, China).

3. Results and Discussion

3.1. Effect of TiO₂ Concentration on 4H-SiC MRR

The ability of the TiO₂ photocatalyst to activate Oxone has attracted considerable attention. The impact of TiO₂ concentration on the SiC MRR provided by the UV-activated catalytic system was investigated in this study using different concentrations TiO₂ with constant concentrations of Oxone and Fe₃O₄ as listed in Table 3. The results are illustrated in Figure 4, in which the SiC removal rate exhibits a parabolic trend that reached a maximum MRR of 1942 nm/h at a TiO₂ concentration of 0.15 wt% before decreasing sharply to 1603 nm/h at a TiO₂ concentration of 0.2 wt%. This phenomenon is consistent with similar patterns observed in applications such as UV- TiO₂ photocatalytic organic degradation and wastewater treatment [16,21].

These results indicate that an increase in TiO₂ dosage initially increased, and then decreased the photocatalytic oxidation efficiency. Thus, the SiC MRR was relatively low when the catalytic system had an insufficient concentration of TiO₂, as this prevented the provision of adequate free radicals, such as holes (h⁺) and electrons (e⁻) (Equation (6)) [22], to generate oxides (Equation (7)) for SiC oxidation. Once the TiO₂ dosage exceeded a certain threshold, the SiC removal rate likely decreased owing to the increased turbidity of the solution induced by excess TiO₂, which reduced the light transmittance of the solution, affecting the photoelectron transfer efficiency and ultimately weakening the photocatalytic oxidation capability of the entire solution.

3.2. Effect of Fe₃O₄ Concentration on 4H-SiC MRR

The Fe₃O₄ functions as a transition metal catalyst activator in the proposed Oxone system. The impact of the transition metal-generated Fe²⁺ concentration on the SiC MRR provided by the UV-activated catalytic system was evaluated using different concentrations of Fe₃O₄ with constant concentrations of Oxone and TiO₂ as shown in Table 3. The results are illustrated in Figure 5, which shows a peak SiC MRR of 1858 nm/h at an Fe₃O₄ concentration of 0.5 wt% that declined to 1620 nm/h and 1530 nm/h at Fe₃O₄ concentrations of 0.75 wt% and 1 wt%, respectively.

These results indicate that excessively low or high concentrations of Fe₃O₄ decreased the reaction rate. The MRR was likely smaller when the Fe₃O₄ concentration was 0.25 wt% because of an insufficient content of Fe²⁺ in the polishing slurry, which resulted in insufficient activation of Oxone, thereby leading to a lower concentration of ${\mathrm{S}\mathrm{O}}_4^{·-}$. Conversely, the MRR likely decreased significantly when the Fe₃O₄ concentration exceeded 0.5% because of the scavenging effect as excess Fe²⁺ reacted with strongly oxidizing ${\mathrm{S}\mathrm{O}}_4^{·-}$ to form other non-oxides as follows [1]:

$${\mathrm{Fe}}^{2+}+{\mathrm{SO}}_4^{·-}\to {\mathrm{Fe}}^{3+}+{\mathrm{SO}}_4^{2-}(k_3=3.0\times {10}^8{\mathrm{M}}^{-1}{\mathrm{s}}^{-1})$$

(10)

A comparison of the reaction rate constant k₃ between ${\mathrm{S}\mathrm{O}}_4^{·-}$ and Fe²⁺ in Equation (10) with the reaction rate constant k₁ between Fe²⁺ and activated Oxone to generate ${\mathrm{S}\mathrm{O}}_4^{·-}$ in Equation (3) [2] reveals that the quenching of ${\mathrm{S}\mathrm{O}}_4^{·-}$ by Fe²⁺ is a major side reaction contributing to the observed significant decrease in the SiC MRR with increasing Fe₃O₄.

These results indicate that when maintaining a constant ratio between Fe²⁺ and Oxone, the concentration of Fe²⁺ required to activate Oxone should be minimized because excess Fe²⁺ significantly inhibits the SiC removal efficiency. Furthermore, experimental observations suggest that when the concentration of Fe²⁺ released from the transition metal catalyst approaches that of Oxone, the reaction rate of the catalytic system reaches its optimum value, resulting in the fastest SiC MRR. This finding is consistent with other research results on the activation of Oxone by transition metals, further validating the underlying theory [10,14,23,24,25,26].

3.3. Effect of Oxone Concentration on 4H-SiC MRR

The impact of Oxone concentration on the SiC MRR provided by the UV-activated catalytic system was investigated using different Oxone concentrations with constant concentrations of Fe₃O₄ and TiO₂ as shown in Table 3. The results are illustrated in Figure 6, which shows an MRR of only 1494 nm/h at an Oxone concentration of 16 mM. This was a result of the limited number of Oxone molecules available to be effectively activated, which led to a significant deficiency in the concentration of free radicals with strong oxidizing ability in the polishing slurry, as well as the side reactions between the Fe₃O₄ transition metal oxide and Oxone, which further reduced the effective concentration of active free radicals, collectively resulting in a low MRR. The MRR increased significantly to 2070 nm/h at an Oxone concentration of 48 mM, indicating that the promotion of the oxidation reaction and increase in the number of active free radicals accelerated the MRR. However, the SiC MRR decreased slightly at an Oxone concentration of 64 mM because excessive Oxone can also consume ${\mathrm{S}\mathrm{O}}_4^{·-}$ as follows [27]. The first item is shown below.

$${\mathrm{HSO}}_5^{-}+{\mathrm{SO}}_4^{·-}\to {\mathrm{SO}}_4^{2-}+{\mathrm{SO}}_5^{·-}+{\mathrm{H}}^{+}$$

(11)

Therefore, an appropriate Oxone concentration is required.

The significant enhancement in MRR observed when the concentration of Oxone increased from 32 mM to 48 mM can be attributed to several factors. During CMP, the heat generated by the UV lamp and polishing promotes the thermal decomposition of excess ${\mathrm{H}\mathrm{S}\mathrm{O}}_5^{-}$ in the polishing slurry as follows [28,29]:

$$\mathrm{H}\mathrm{S}{\mathrm{O}}_5^{-}+\mathrm{h}\mathrm{e}\mathrm{a}\mathrm{t}\to \mathrm{S}{\mathrm{O}}_4^{·-}+•\mathrm{O}\mathrm{H}$$

(12)

This phenomenon not only consumes the excess ${\mathrm{H}\mathrm{S}\mathrm{O}}_5^{-}$, thereby maintaining the concentration balance between the catalyst and Oxone in the polishing slurry, but also generates new active radicals, thereby enhancing its oxidizing ability. The concentration of Oxone in the polishing slurry approaches the optimal catalyst concentration at this point, thereby improving the efficiency of the catalytic reaction.

The results of the CMP experiments indicate that excessively low or high concentrations of Oxone can affect the SiC MRR. However, the adverse effects of an excessively high Oxone concentration on the oxidation system performance were relatively minor compared with those of an excessively low Oxone concentration. Although excess Oxone consumes ${\mathrm{S}\mathrm{O}}_4^{·-}$, this does not affect the activation reaction between Fe₃O₄-Oxone and UV-TiO₂-Oxone. Additionally, Equation (11) indicates that ${\mathrm{S}\mathrm{O}}_5^{·-}$, which has a weaker oxidizing power than ${\mathrm{S}\mathrm{O}}_4^{·-}$, is generated. According to Equation (5), these weakly oxidizing groups can transform into each other and regenerate ${\mathrm{S}\mathrm{O}}_4^{·-}$. In contrast, insufficient ${\mathrm{H}\mathrm{S}\mathrm{O}}_5^{-}$ is available when the Oxone concentration is too low, which prevents the generation of sufficient ${\mathrm{S}\mathrm{O}}_4^{·-}$ and other oxidizing groups in the polishing slurry. Furthermore, excess transition metal Fe²⁺ scavenges free radicals, further limiting the resulting SiC MRR.

3.4. Comparison of Different Catalytic Methods

The synergistic enhancement of the oxidizing properties of a polishing slurry catalyzed by UV light, TiO₂, and Fe₃O₄ in conjunction with Oxone was investigated using different combinations of 0.75 wt% Fe₃O₄, 0.15 wt% TiO₂, and 48 mM Oxone with a pH value of 3 to conduct CMP experiments for 30 min and comparing the resulting MRR values. Four combinations were considered: Specimen 1 was polished using only EV light, 48 mM Oxone, and 1 wt% diamond at a pH of 3; Specimen 2 was polished using UV light, 48 mM Oxone, 0.15 wt% TiO₂, and 1 wt% diamond at a pH of 3; Specimen 3 was polished using UV light, 48 mM Oxone, 0.5 wt% Fe₃O₄, and 1 wt% diamond at a pH of 3; and Specimen 4 was polished using UV light, 48 mM Oxone, 0.15 wt% TiO₂, 0.75 wt% Fe₃O₄, and 1 wt% diamond at a pH of 3. The results presented in Figure 7 indicate that the complete UV-TiO₂-Fe₃O₄-Oxone combination exhibited the best experimental effect.

The use of the UV–Oxone catalytic system on Specimen 1 exhibited the lowest SiC MRR. This may be attributed to the poor absorption of UV light energy by Oxone, which led to a slow activation rate and low concentration of generated free radicals, thereby reducing the reaction efficiency. The addition of TiO₂ to the UV–Oxone catalytic system used on Specimen 2 significantly enhanced the SiC MRR because the TiO₂ underwent photocatalytic reactions when exposed to the 365 nm UV light. These reactions excited its valence band electrons to the conduction band, forming electrons and holes that promoted the decomposition of ${\mathrm{H}\mathrm{S}\mathrm{O}}_5^{-}$ to generate highly reactive ${\mathrm{S}\mathrm{O}}_4^{·-}$, thereby improving the oxidation efficiency of SiC and the corresponding MRR.

The SiC MRR achieved using the UV-Fe₃O₄-Oxone catalytic system on Specimen 3 was higher than that achieved on Specimens 1 or 2. This can be attributed to the continuous release of Fe²⁺ by the Fe₃O₄, which ensured the stable progression of the reaction given by Equation (3). Additionally, the auxiliary effect of UV light enabled the generation of ${\mathrm{S}\mathrm{O}}_4^{·-}$ from a small quantity of Oxone in the polishing slurry under the influence of temperature and wavelength energy, thereby enhancing the polishing efficiency and quality of the SiC surface. The UV-TiO₂-Fe₃O₄-Oxone catalytic system used on Specimen 4 achieved the highest SiC MRR, significantly surpassing that achieved using any other system. This indicates that the UV-TiO₂-Fe₃O₄ system exhibited a pronounced synergistic enhancement effect when catalyzing Oxone. Indeed, this significant increase in MRR is predictable as Oxone can generate free radicals through multiple pathways as follows [9,16,26,29]:

$$\mathrm{H}\mathrm{S}{\mathrm{O}}_5^{-}\stackrel{\mathrm{h}\mathrm{v}/\mathrm{h}\mathrm{e}\mathrm{a}\mathrm{t}/\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{n}\mathrm{s}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\cdot \mathrm{m}\mathrm{e}\mathrm{a}\mathrm{t}\mathrm{a}\mathrm{l}/\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{d}\mathrm{u}\mathrm{c}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\cdot \mathrm{e}\mathrm{l}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{n}}{\to }\mathrm{S}{\mathrm{O}}_4^{·-}+•\mathrm{O}\mathrm{H}$$

(13)

The synergistic reactions that may occur during this process are hypothesized to include the combined catalytic effects of the photocatalyst and transition metal, which facilitate the decomposition of Oxone and generate more ${\mathrm{S}\mathrm{O}}_4^{·-}$ and ·OH. According to Equation (3), the Fe³⁺ generated by the reaction of Fe²⁺ with Oxone can act as an electron acceptor to capture electrons, thereby reducing the recombination of electron/hole pairs (e⁻/h⁺) in the TiO₂ photocatalytic reaction [30], enhancing the ability of the TiO₂ to generate free radicals and augmenting the catalytic efficiency between the transition metal and Oxone. Furthermore, under UV light irradiation, the Fe³⁺ in the Fe₃O₄-Oxone catalytic reaction system is reduced to Fe²⁺, generating a small quantity of ·OH (as shown in Equations (1) and (2)), which not only accelerates the regeneration of Fe²⁺ and promotes catalytic efficiency between the transition metal and Oxone, but also increases the concentration of highly oxidizing free radicals in the polishing slurry.

As a result, the UV-TiO₂-Fe₃O₄-Oxone catalytic system exhibited a higher reaction rate and more diverse pathway for the generation of active free radicals compared to those of any single-catalyst system. This resulted in a superior SiC MRR and surface quality, further substantiating the synergistic enhancement effect of the UV-TiO₂-Fe₃O₄-Oxone catalytic system on SiC removal.

3.5. Surface Analyses

The mechanism by which UV, Fe₃O₄, and TiO₂ synergistically catalyze the activation of Oxone to enhance the MRR was investigated by conducting an XPS analysis on polished SiC Specimens 1, 2, and 3. The O 1s, C 1s, and Si 2p peaks in the full spectrum shown in Figure 8 appeared at 532, 285, and 101 eV, respectively [31]. The chemical states and relative concentrations of Si, O, and C elements were analyzed by XPS through the excitation of their core-level electrons (Si 2p, O 1s, and C 1s orbitals) using a monochromatic Al Kα X-ray source, whereby the kinetic energies of emitted photoelectrons were measured to calculate binding energies through the Einstein relation, with subsequent chemical state identification and quantitative analysis performed via reference to standard spectral databases following rigorous binding energy calibration. After CMP using the UV-TiO₂-Fe₃O₄-Oxone oxidation system (Specimen 3), the O 1s peak intensity showed a significant increase, while the C 1s peak intensity exhibited a slight decrease, compared to those obtained from the UV-TiO₂-Oxone and UV-Fe₃O₄-Oxone systems (Specimens 1 and 2). This suggests that a more significant oxidation reaction occurred on the surface of Specimen 3, further demonstrating the synergistic effects of UV light, photocatalysts, and transition metal when catalyzing Oxone.

Narrow-scan analyses of the C and Si elements in the specimens were conducted to obtain the C 1s and Si 2p spectra, which were subsequently subjected to data fitting with the results presented in Figure 9 and Figure 10, respectively. Following CMP, the C 1s spectra of the 4H-SiC surfaces exhibited six peaks, namely C-C/C-H, C-Si, C-O, C=O, Si₄C₄O₄, and Si₄C₄-xO₂ corresponding to binding energies of 284.7, 282.3, 286, 288.3, 285.3, and 282.8 eV, respectively [32,33,34]. The Si₄C₄O₄ and Si₄C₄-xO₂ peaks represent the oxides generated by the reaction between the Oxone and catalysts in the polishing slurry. The Si 2p spectra exhibited three peaks: Si-C, Si-C-O, and Si-O₂ corresponding to binding energies of 100.2, 100.8, and 103 eV, respectively [31,35,36]. The Si-C-O and Si-O₂ peaks represent oxides formed after CMP following the activation of Oxone by the catalysts. Comparing Figure 9a,b, the intensity of the C-Si peak is clearly lower in the latter, whereas those of the Si₄C₄O₄ and C-O peaks are higher, indicating that the oxidation capability of the UV-Fe₃O₄-Oxone system (Specimen 2) was stronger than that of the UV-TiO₂-Oxone system (Specimen 1). Furthermore, comparing Figure 9a,c, the intensities of the C-C and C-Si peaks are significantly lower in the latter (Specimen 3), whereas the intensities of the oxide peaks of Si and C are significantly higher, confirming that the oxidation capability of the UV-TiO₂-Fe₃O₄-Oxone system was higher than that of any single-catalyst system.

Figure 10 shows that the Si-C peak slightly decreased, whereas the Si-O₂ peak slightly increased with the introduction of the TiO₂ and Fe₃O₄ catalysts. This indicates that the addition of these catalysts during CMP promoted the decomposition of Oxone, releasing more active free radicals and thereby enhancing the oxidative performance of the polishing slurry, weakening the Si-C bond, and promoting the formation of oxides on the SiC substrate. This new oxide layer was continuously cycled through mechanical removal and chemical reformation, facilitating the removal of damaged layers on the SiC wafer surface. Therefore, the magnitude of oxidation intensity directly affected the MRR realized by CMP. Finally, the XPS analysis results indicate that the oxidation intensities for the different catalytic systems decreased from UV-TiO₂-Fe₃O₄-Oxone (Specimen 3) to UV-Fe₃O₄-Oxone (Specimen 2) to UV-TiO₂-Oxone (Specimen 1), which is consistent with the CMP experiment results.

The surfaces of the 4H-SiC specimens were examined before and after polishing using WLI; typical results are provided in Figure 11(a₁) and (a₂), respectively. As shown in Figure 11(a₂), the unpolished surface exhibited numerous sharp peaks and valleys, a rough texture with surface roughnesses R_a and S_a of 110 and 135 nm, respectively, and a highest point of 1315.2 nm; the surface height range after polishing was significantly reduced with the sharp peaks and valleys smoothed, surface roughnesses R_a and S_a of 0.3 and 8.3 nm, respectively, and a highest point of only 92.7 nm. This result confirms that the CMP process effectively eliminated microscopic irregularities on the wafer surface. The appearance of low-lying areas on the specimen surface after polishing was primarily due to the adsorption of strong oxidizing radicals on the surfaces of the Fe₃O₄ and TiO₂ catalysts. Under the rotational compression of the polishing head and pad, the catalysts carrying strong oxidizing radicals came into contact with the Si surface, triggering oxidation reactions at the contact points and forming a softer oxidized layer that was mechanically removed by abrasive particles in the polishing slurry. Critically, because the hardness of the catalysts in the polishing slurry was lower than that of the Si surface, the oxidation process occurred through point contact and no scratches were generated on the wafer surface.

Further examination of the Si specimen surfaces before and after polishing was conducted using SEM; typical results are provided in Figure 11(b₁) and (b₂), respectively. As shown in Figure 11(b₁), the unpolished wafer surface exhibited distinct depressions and irregular surface features. After 1 h of CMP using the UV-TiO₂-Fe₃O₄-Oxone oxidation system, the surface became quite uniform with almost no visible irregularities or granular structures, indicating that the polishing process effectively eliminated microscopic irregularities on the surface. Indeed, not only did the surface become flat and smooth, but no new microscopic defects were observed, as shown in Figure 11(b₂). The comparison of the physical images before and after polishing is shown in Figure 12.

XPS analysis confirmed the formation of oxide species on the wafer surface through characteristic binding energy shifts in O 1s spectra, while SEM characterization revealed a smooth and featureless post-polished surface morphology with sub-nanometer roughness, indicating effective removal of oxide layers without residual accumulation. This apparent contradiction between oxide generation and surface cleanliness demonstrates the synergistic mechanism of chemical oxidation and mechanical abrasion in the CMP process, where transiently formed oxides are continuously eliminated by abrasive action, thereby achieving atomic-level planarization.

4. Conclusions

This study evaluated a novel oxidative CMP process to enhance the MRR and surface quality of 4H-SiC wafers. This process utilizes the synergistic effects of UV light, a photocatalyst (TiO₂), and a transition metal (Fe₃O₄) to activate an Oxone oxidant and thereby generate strongly oxidizing ${\mathrm{S}\mathrm{O}}_4^{·-}$ that significantly improves the performance of the polishing slurry. The following primary conclusions were drawn from the experiments and observations in this study:

(1)

The principle underlying the multi-catalyst activation of Oxone was briefly discussed relative to its application as a novel approach for polishing 4H-SiC. Multiple sets of single-factor experiments were conducted to investigate the impact of the Oxone, TiO₂, and Fe₃O₄ concentrations on the SiC MRR accordingly. The results indicated that the highest MRR of 2360 nm/h was achieved when using 0.75 wt% Fe₃O₄, 0.15 wt% TiO₂, and 48 mM Oxone.

(2)

Comparative experiments and XPS analyses were undertaken to confirm that the synergistic effect of UV irradiation, TiO₂, and Fe₃O₄ significantly enhanced the oxidizing properties of the polishing slurry, thereby improving the MRR and surface quality of the polished 4H-SiC. Furthermore, observations conducted using WLI and SEM indicated a supersmooth surface with a surface roughness R_a of 0.304 nm.

Although this study explored MRR trends through a series of experiments, the optimal ratio between the catalyst and oxidant concentrations has not yet been fully determined. Furthermore, the experiment results suggested that, in addition to the quantities of catalyst and oxidant used, factors such as temperature, UV light wavelength, and pH also influence the activation of Oxone. Therefore, future studies should prioritize synergistic optimization of these critical parameters to advance both efficiency and surface integrity in 4H-SiC wafer CMP processes, while leveraging the fundamental similarities in mechanical properties and material removal mechanisms between SiC and other hard brittle materials to extend these optimized strategies to their precision manufacturing scenarios.