Comparison of Erosion Behavior and Particle Contamination in Mass-Production CF4/O2 Plasma Chambers Using Y2O3 and YF3 Protective Coatings

Yttrium fluoride (YF3) and yttrium oxide (Y2O3) protective coatings prepared using an atmospheric plasma spraying technique were used to investigate the relationship between surface erosion behaviors and their nanoparticle generation under high-density plasma (1012–1013 cm−3) etching. As examined by transmission electron microscopy, the Y2O3 and YF3 coatings become oxyfluorinated after exposure to the plasma, wherein the yttrium oxyfluoride film formation was observed on the surface with a thickness of 5.2 and 6.8 nm, respectively. The difference in the oxyfluorination of Y2O3 and YF3 coatings could be attributed to Y–F and Y–O bonding energies. X-ray photoelectron spectroscopy analyses revealed that a strongly fluorinated bonding (Y–F bond) was obtained on the etched surface of the YF3 coating. Scanning electron microscopy and energy dispersive X-ray diffraction analysis revealed that the nanoparticles on the 12-inch wafer are composed of etchant gases and Y2O3. These results indicate that the YF3 coating is a more erosion-resistant material, resulting in fewer contamination particles compared with the Y2O3 coating.


Introduction
Silicon-based ceramics have been extensively used in semiconductor plasma processing equipment as plasma-facing materials, due to their hardness, high wear resistance, dielectric strength, high corrosion resistance, and chemical stability [1,2]. They are used mainly as a shield to protect the ceramic parts inside etchers or chemical vapor deposition reactor chambers from corrosion caused by fluorocarbon corrosive gases such as CF 4 , CHF 3 , C 4 F 6 , and C 2 F 6 [3][4][5]. These materials interact with plasma and are eroded, resulting in the production of contaminant particles on the wafer. As integrated circuits continue to scale down with wider use of high-density plasma for wafer processing, the particles generated in the plasma processing equipment cause serious problems, such as short current in integration circuit and lower production yield [6,7]. In order to solve this problem, yttrium oxide (Y 2 O 3 ) was adopted as plasma-facing inner wall materials in plasma processing equipment because their plasma erosion resistance values are much higher than those of conventional SiO 2 coatings [8][9][10]. Mass-production factories have found that the Y 2 O 3 inner walls have problems with significant erosion and particle generation [11]. Under fluorine-based plasma processing, a thin top carbonaceous polymer reaction layer has been identified depending on the etching conduction and the etched materials [12][13][14]. The polymer layer thickness is determined using polymer deposition and its removal rate, and consumption rate in substrate etching. This polymer layer etching rate is affected by the plasma incident ion kinetic energy. Some volatile etching products are produced during the etching process, such as carbon oxide, carbon oxyfluorides, and silicon fluorides [15]. Unlike silicon-based materials, the yttrium-based material etching mechanism is not fully understood. Yttrium fluoride (YF 3 ) coatings ave recently attracted substantial attention because of their high plasma erosion resistance, preventing the generation of fluoride particles on the chamber wall surface, reducing particulate contamination [16]. Thus, the YF 3 coating might be a new plasma-facing material that produces fewer contamination particles. The Y 2 O 3 and YF 3 coatings were deposited using atmospheric plasma spraying (APS). In this study, the mechanism of formation of yttrium oxyfluoride film and their particle trajectories in industrial plasma processing tools have been examined. Moreover, we compared the etching behavior of Y 2 O 3 and YF 3 coatings and their compounds under fluorocarbon plasma. The surface morphology, chemical reactions on the etched surface, microstructure, and particle contaminations of Y 2 O 3 and YF 3 coatings were investigated.

Materials and Methods
Commercially available YF 3 powders (25-50 µm, 99.99%, Shin-Etsu Chemical, Tokyo, Japan) and Y 2 O 3 powders (25-50 µm, 99.99%, Shin-Etsu Chemical, Tokyo, Japan) were used as the spraying materials. YF 3 and Y 2 O 3 coatings were prepared using APS with an F4-MB plasma gun (Sulzer Metco, Orelikon, Pfaeffikon, Switzerland). An alloy aluminum (A6061) substrate was used for the experiment. The specimen had a size of 400 mm 2 and a thickness of 20 mm. Before APS, the substrate was treated with sandblasting. The sandblasting material was SiO 2 . Acetone was used to clean the substrate. The stand-off distance was adjusted to 10 cm. The Ar and H 2 gas cylinders were opened when the air compressor was initiated. The Ar flow rate, H 2 flow rate, system voltage, gun movement rate, and feed rate were set to 45 L/min, 6 L/min, 50 V, 10 cm/s, and 15 g/min, respectively. The YF 3 and Y 2 O 3 spraying parameters are shown in Table 1. The erosion behaviors of both protective coatings were performed using an inductively coupled plasma (ICP) etcher (LAM 2300 Metal) under the routine plasma etching process; i.e., the same bias power and processing gases (CF 4 and O 2 ). High-density CF 4 /O 2 plasma with electron densities on the order of 10 12 to 10 13 cm −3 has been produced. Figure 1 shows a schematic diagram of the ICP etcher system employed in this study. The etch process details are shown in Table 2. Three-hundred millimeter blanket wafers with chemical vapor deposition titanium nitride layer/Si-substrate were prepared to evaluate the integrated circuit defective performance after dry etching process measured by Surfscan SP3 (Surfscan SP3, KLA-Tencor Corporation, Milpitas, CA, USA).
The surface morphology, microstructure, and elemental analysis of these coating samples were conducted using scanning electron microscopy (SEM, S-3000H, Hitachi, Tokyo, Japan) coupled with energy dispersive X-ray diffraction (EDX) and transmission electron microcopy (TEM, H-600, Hitachi, Tokyo, Japan). The composition of these samples were examined by X-ray photoelectron spectroscopy (XPS, PHI 5000 VersaProbe, ULVAC-PHI, Kanagawa, Japan) using a monochromatic Al Kα X-ray source at a passing energy of 20 eV with a spot size of 650 µm, then the sample surface was etched using focused argon ions sputtering to investigate the chemical compositional depth profile (ThermoScientific K-Alpha). A fitting software program (Thermo Fisher Scientific, Inc., Waltham, MA, USA) was used to deconvolute the photoelectron spectrum resulting from the core energy levels of Yttrium 3d states to estimate the contributions from bonding with fluorine elements. density CF4/O2 plasma with electron densities on the order of 10 12 to 10 13 cm −3 has been produced. Figure 1 shows a schematic diagram of the ICP etcher system employed in this study. The etch process details are shown in Table 2. Three-hundred millimeter blanket wafers with chemical vapor deposition titanium nitride layer/Si-substrate were prepared to evaluate the integrated circuit defective performance after dry etching process measured by Surfscan SP3 (Surfscan SP3, KLA-Tencor Corporation, Milpitas, CA, USA).    Figure 2 shows the surface and cross-sectional SEM images of Y 2 O 3 and YF 3 coatings under 15 kW plasma spraying powers. Figure 2a shows the Y 2 O 3 coating with poor surface roughness with laminar morphology. A horizontal crack and large cavities (within the size range of 5-10 µm) are also observed (Figure 2b). Dense YF 3 coating layers with less porosity are shown in Figure 2c,d).

Results and Discussion
Due to the small difference in thermal expansion between the YF 3 coating (28.5 × 10 −6 /K) and the Al substrate (23 × 10 −6 /K), no cracks were observed in any YF 3 coating samples. The erosion-resistance characteristics of Y 2 O 3 and YF 3 coatings were measured after exposure to the CF 4 /O 2 plasma, as discussed below. Figure 3a,b show the SEM image surface microstructure of the Y 2 O 3 and YF 3 coatings after etching in CF 4 /O 2 plasma for 60 min under a bias power of 500 W. There is a large difference in etched surface for both coating samples under the same etching condition. In Figure 3a, the surface of Y 2 O 3 coating is obviously cracked after the etching process. These Y 2 O 3 creaked pieces might be a particle contamination source during the wafer fabrication process. In Figure 3b, the YF 3 coating is revealed to have a relatively dense and smooth surface, indicating that the erosion of Y 2 O 3 coating was more severe than YF 3 coating after exposure to CF 4 /O 2 plasma. It is also consistent with our previous study regarding the film porosity [16], since the erosion resistance to CF 4 /O 2 plasma (chemical reaction) was enhanced by reducing film porosity [17]. From the SEM observation results, the YF 3 coating revealed that a clean and complete surface can be obtained by preventing the fluoride particles from attaching to the etching chamber sidewall after exposure to the fluorocarbon plasma during etching. Hence, the YF 3 coating is more favorable for application in plasma processing equipment. Similar results were consistent with the previously reported data by Kim et al. [18].   Figure 4 shows the compositional variation with the sputtering time from the Y2O3 and YF3 coating surfaces after exposure to the CF4/O2 plasma by using XPS. The Y2O3 and YF3 coating surfaces before exposure to the CF4/O2 plasma are shown in Figure 4a,b. The carbon content found on the surface decreased abruptly with the sputtering time, indicating that the carbon polymer layer is very thin (Figure 4c,d). This thin carbon polymer layer on the etched surface was previously reported in Si-based oxide materials etched under fluorocarbon plasma [19][20][21]. Moreover, the fluorine was verified on both Y2O3 and YF3 coatings and the higher fluorine content was detected on the YF3 surface (Figure 4c,d). It was found that the percentage of F atoms reached the maximum value of 35% and 48% on the Y2O3 and YF3 coatings after etching, causing a thicker fluorination layer to appear in the YF3 specimen. This is because YF3 is a fluorine-rich compound material; hence, the chemical composition of F atoms in the YF3 coating was unchanged accompanied with the depth from the surface. The stronger YF3 coating fluorination can also be confirmed by the XPS spectral analysis on etched sample surfaces.   Figure 4 shows the compositional variation with the sputtering time from the Y2O3 and YF3 coating surfaces after exposure to the CF4/O2 plasma by using XPS. The Y2O3 and YF3 coating surfaces before exposure to the CF4/O2 plasma are shown in Figure 4a,b. The carbon content found on the surface decreased abruptly with the sputtering time, indicating that the carbon polymer layer is very thin (Figure 4c,d). This thin carbon polymer layer on the etched surface was previously reported in Si-based oxide materials etched under fluorocarbon plasma [19][20][21]. Moreover, the fluorine was verified on both Y2O3 and YF3 coatings and the higher fluorine content was detected on the YF3 surface (Figure 4c,d). It was found that the percentage of F atoms reached the maximum value of 35% and 48% on the Y2O3 and YF3 coatings after etching, causing a thicker fluorination layer to appear in the YF3 specimen. This is because YF3 is a fluorine-rich compound material; hence, the chemical composition of F atoms in the YF3 coating was unchanged accompanied with the depth from the surface. The stronger YF3 coating fluorination can also be confirmed by the XPS spectral analysis on etched sample surfaces.   Figure 4a,b. The carbon content found on the surface decreased abruptly with the sputtering time, indicating that the carbon polymer layer is very thin (Figure 4c,d). This thin carbon polymer layer on the etched surface was previously reported in Si-based oxide materials etched under fluorocarbon plasma [19][20][21]. Moreover, the fluorine was verified on both Y 2 O 3 and YF 3 coatings and the higher fluorine content was detected on the YF 3 surface (Figure 4c,d). It was found that the percentage of F atoms reached the maximum value of 35% and 48% on the Y 2 O 3 and YF 3 coatings after etching, causing a thicker fluorination layer to appear in the YF 3 specimen. This is because YF 3 is a fluorine-rich compound material; hence, the chemical  Figure 5 represents the XPS spectra for the yttrium atoms from the (a) Y2O3 and (b) YF3 coating samples after CF4/O2 plasma etching. In the curve-fitted XPS spectra of the Y2O3 coating, two peaks mean two sources of bonding for cations from Y3d split into two Y3d5/2 states (high peaks) and Y3d3/2 (low peaks) [22]. In Figure 5a, the etched surface of the Y2O3 coating consisted of Y3d5/2 and Y3d3/2, with an intensity ratio of 3:2 and peak shift difference in the binding energy of 2 eV, according to the XPS standard [23,24]. This peak shifting to higher binding energy (located at 160 and 162 eV) could be attributed to Y-F bonding in the Y2O3 sample. The higher electronegativity of fluorine (4.0) compared to that of oxygen (3.5) causes higher electron binding energy from the cation [25]. This high fluorine concertation demonstrated that the Y2O transformed into a YOxFy (x + y = 1.5) and/or YFx (x < 3) surface by the exposed fluorine-based plasma [26]. Meanwhile, the two binding energy peaks located at 158.5 and 160.5 eV correspond to a Y-O bond, resulting in a low binding energy. Figure 5b shows the YF3 coating XPS spectra on the etched surface deconvoluted into four peaks, corresponding to Y-F bonding (located at 159.5 and 161.5 eV) and Y-O bonding (located at 157 and 159 eV) [27,28]. Furthermore, the intensity ratio of Y-F to Y-O peaks on the YF3 coating reached 2.9, which was much higher than that of the Y-F to Y-O peak on the Y2O3 coating (0.73), indicating stronger fluorination on the etched YF3 coating surface. These results indicate that the YF3 coating exhibited superior inherent chemical stability after fluorocarbon plasma treatment [29]. The reactions for CF4/O2 plasma chemical etching of Y2O3 and YF3 coatings can be expressed as follows: Y2O3 + 3CF2* → 2YF3 + 3CO; YF3 + 2CF2* → YF + 2CF3.
when the CF4/O2 plasma chemical reaction dominates the etching process, and the chemical stability YF3 might be effective in the suppression of particle generation during the etching process, as mentioned in the SEM results.  In the curve-fitted XPS spectra of the Y 2 O 3 coating, two peaks mean two sources of bonding for cations from Y3d split into two Y3d 5/2 states (high peaks) and Y3d 3/2 (low peaks) [22]. In Figure 5a, the etched surface of the Y 2 O 3 coating consisted of Y3d 5/2 and Y3d 3/2 , with an intensity ratio of 3:2 and peak shift difference in the binding energy of 2 eV, according to the XPS standard [23,24]. This peak shifting to higher binding energy (located at 160 and 162 eV) could be attributed to Y-F bonding in the Y 2 O 3 sample. The higher electronegativity of fluorine (4.0) compared to that of oxygen (3.5) causes higher electron binding energy from the cation [25]. This high fluorine concertation demonstrated that the Y 2 O transformed into a YO x F y (x + y = 1.5) and/or YF x (x < 3) surface by the exposed fluorine-based plasma [26]. Meanwhile, the two binding energy peaks located at 158.5 and 160.5 eV correspond to a Y-O bond, resulting in a low binding energy. Figure 5b shows the YF 3 coating XPS spectra on the etched surface deconvoluted into four peaks, corresponding to Y-F bonding (located at 159.5 and 161.5 eV) and Y-O bonding (located at 157 and 159 eV) [27,28]. Furthermore, the intensity ratio of Y-F to Y-O peaks on the YF 3 coating reached 2.9, which was much higher than that of the Y-F to Y-O peak on the Y 2 O 3 coating (0.73), indicating stronger fluorination on the etched YF 3 coating surface. These results indicate that the YF 3 coating exhibited superior inherent chemical stability after fluorocarbon plasma treatment [29]. The reactions for CF 4 /O 2 plasma chemical etching of Y 2 O 3 and YF 3 coatings can be expressed as follows: when the CF 4 /O 2 plasma chemical reaction dominates the etching process, and the chemical stability YF 3 might be effective in the suppression of particle generation during the etching process, as mentioned in the SEM results. The microstructures of both coated samples irradiated by high-density CF4/O2 plasma were revealed by TEM. Figure 6a,b shows cross-sectional TEM micrographs of the plasma-etched Y2O3 and YF3 coatings, respectively. A yttrium oxyfluoride (YOxFy) film 5.2 nm in thickness was observed on the Y2O3 surface, while the YF3 sample showed a thicker YOxFy of 6.8 nm. The slightly lesser fluorination of Y2O3 than YF3 is explained by comparing the bonding energies of Y-O (685 KJ/mol) and Y-F (605 KJ/mol). Because the bonding energy of Y-O is higher than Y-F, it results in less-efficient reactions between the Y-O bonding and the fluorocarbon film. The formation of an oxyfluoride layer on the surface of Y2O3 and YF3 coatings might act as a protecting layer to prevent the coating's surface from further erosion by CF4/O2 plasma.  Figure 7 illustrates the schematic formation mechanism of YOxFy on Y2O3 and YF3 surfaces. The procedure of YOxFy formation on the Y2O3 and YF3 surfaces is as follows: the first step is deposition of the fluorocarbon film by the adsorption of CFx radicals on these two coatings' sample surfaces. The second step is carbon reactions with oxygen (Y-O) and fluorine (Y-F) to form volatile CO and CFx, resulting in the decomposition of the Y-O and Y-F bondings. Subsequently, the YOxFy film is formed in the Y2O3 and YF3 coatings surface (third step), whereas a part of YFx may desorb from the coating surface. Similar reactions and formation of YOxFy were also investigated previously [30,31]. It is believed that the YOxFy layer is effective in reducing practical generation and thus contamination of the integrated circuit [32]. The microstructures of both coated samples irradiated by high-density CF4/O2 plasma were revealed by TEM. Figure 6a,b shows cross-sectional TEM micrographs of the plasma-etched Y2O3 and YF3 coatings, respectively. A yttrium oxyfluoride (YOxFy) film 5.2 nm in thickness was observed on the Y2O3 surface, while the YF3 sample showed a thicker YOxFy of 6.8 nm. The slightly lesser fluorination of Y2O3 than YF3 is explained by comparing the bonding energies of Y-O (685 KJ/mol) and Y-F (605 KJ/mol). Because the bonding energy of Y-O is higher than Y-F, it results in less-efficient reactions between the Y-O bonding and the fluorocarbon film. The formation of an oxyfluoride layer on the surface of Y2O3 and YF3 coatings might act as a protecting layer to prevent the coating's surface from further erosion by CF4/O2 plasma.  Figure 7 illustrates the schematic formation mechanism of YOxFy on Y2O3 and YF3 surfaces. The procedure of YOxFy formation on the Y2O3 and YF3 surfaces is as follows: the first step is deposition of the fluorocarbon film by the adsorption of CFx radicals on these two coatings' sample surfaces. The second step is carbon reactions with oxygen (Y-O) and fluorine (Y-F) to form volatile CO and CFx, resulting in the decomposition of the Y-O and Y-F bondings. Subsequently, the YOxFy film is formed in the Y2O3 and YF3 coatings surface (third step), whereas a part of YFx may desorb from the coating surface. Similar reactions and formation of YOxFy were also investigated previously [30,31]. It is believed that the YOxFy layer is effective in reducing practical generation and thus contamination of the integrated circuit [32].  Figure 7 illustrates the schematic formation mechanism of YO x F y on Y 2 O 3 and YF 3 surfaces. The procedure of YO x F y formation on the Y 2 O 3 and YF 3 surfaces is as follows: the first step is deposition of the fluorocarbon film by the adsorption of CF x radicals on these two coatings' sample surfaces. The second step is carbon reactions with oxygen (Y-O) and fluorine (Y-F) to form volatile CO and CF x , resulting in the decomposition of the Y-O and Y-F bondings. Subsequently, the YO x F y film is formed in the Y 2 O 3 and YF 3 coatings surface (third step), whereas a part of YF x may desorb from the coating surface. Similar reactions and formation of YO x F y were also investigated previously [30,31]. It is believed that the YO x F y layer is effective in reducing practical generation and thus contamination of the integrated circuit [32]. One of the prime concerns of production yields is the particles generated from the plasma processing equipment, resulting in open-or short-circuits. We have investigated the possible particles generated during the etching process using an in-situ particle monitoring system which can detect the particle trajectories. Figure 8a shows an SEM image of a typical particle observed on the wafer surface. The circular particle is composed of etchant gases and a Y2O3-coated chamber wall after exposed to CF4/O2 plasma-it can be called a partial etch defect. An EDX analysis was carried out to clarify the particle compositions. As shown in Figure 8b, it was found that the flaking particles were composed mainly of yttrium, oxide, and fluoride elements, indicating the particle source from the Y2O3 coating. Therefore, the YF3 coating can behave as a new plasma-facing material that produces fewer contamination particles.

Conclusions
During the plasma etching process, particles generated from the Y2O3 and YF3 protective coatings of the ICP chamber wall were investigated in this study. The particle generation mechanism could be due to the fact that the bonding strength of Y-O was weaker than that of Y-F when the chamber-wall surface suffered irradiation from high-density CF4/O2 plasmas. From the SEM examination results, YF3 was also confirmed to be more robust than Y2O3 against CF4/O2 plasma irradiation. The YF3 coatings for the ICP etching chamber components and materials can play an One of the prime concerns of production yields is the particles generated from the plasma processing equipment, resulting in open-or short-circuits. We have investigated the possible particles generated during the etching process using an in-situ particle monitoring system which can detect the particle trajectories. Figure 8a shows an SEM image of a typical particle observed on the wafer surface. The circular particle is composed of etchant gases and a Y 2 O 3 -coated chamber wall after exposed to CF 4 /O 2 plasma-it can be called a partial etch defect. An EDX analysis was carried out to clarify the particle compositions. As shown in Figure 8b, it was found that the flaking particles were composed mainly of yttrium, oxide, and fluoride elements, indicating the particle source from the Y 2 O 3 coating. Therefore, the YF 3 coating can behave as a new plasma-facing material that produces fewer contamination particles. One of the prime concerns of production yields is the particles generated from the plasma processing equipment, resulting in open-or short-circuits. We have investigated the possible particles generated during the etching process using an in-situ particle monitoring system which can detect the particle trajectories. Figure 8a shows an SEM image of a typical particle observed on the wafer surface. The circular particle is composed of etchant gases and a Y2O3-coated chamber wall after exposed to CF4/O2 plasma-it can be called a partial etch defect. An EDX analysis was carried out to clarify the particle compositions. As shown in Figure 8b, it was found that the flaking particles were composed mainly of yttrium, oxide, and fluoride elements, indicating the particle source from the Y2O3 coating. Therefore, the YF3 coating can behave as a new plasma-facing material that produces fewer contamination particles.

Conclusions
During the plasma etching process, particles generated from the Y 2 O 3 and YF 3 protective coatings of the ICP chamber wall were investigated in this study. The particle generation mechanism could be due to the fact that the bonding strength of Y-O was weaker than that of Y-F when the chamber-wall surface suffered irradiation from high-density CF 4 /O 2 plasmas. From the SEM examination results, YF 3 was also confirmed to be more robust than Y 2 O 3 against CF 4 /O 2 plasma irradiation. The YF 3 coatings for the ICP etching chamber components and materials can play an important role in decreasing the extreme number of particles in the fluorine-based plasma environment.