Plasma Science and Technology, Vol.15, No.10, Oct. 2013 Effect of Low-Frequency Power on Etching Characteristics of 6H-SiC in C 4 F 8 /Ar Dual-Frequency Capacitively Coupled Plasma XU Yijun ( ) 1,2, WU Xuemei ( ) 1,2,3, YE Chao ( ) 1,2 1 Department of Physics, Soochow University, Suzhou 215006, China 2 The Key Laboratory of Thin Films of Jiangsu, Soochow University, Suzhou 215006, China 3 State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China Abstract Dry etching of 6H silicon carbide (6H-SiC) wafers in a C 4F 8/Ar dual-frequency capacitively coupled plasma (DF-CCP) was investigated. Atomic force microscopy (AFM) and X-ray photoelectron spectroscopy (XPS) were used to measure the SiC surface structure and compositions, respectively. Optical emission spectroscopy (OES) was used to measure the relative concentration of F radicals in the plasma. It was found that the roughness of the etched SiC surface and the etching rate are directly related to the power of low-frequency (LF) source. At lower LF power, a smaller surface roughness and a lower etching rate are obtained due to weak bombardment of low energy ions on the SiC wafers. At higher LF power the etching rate can be efficiently increased, but the surface roughness increases too. Compared with other plasma dry etching methods, the DF-CCP can effectively inhibit C xf y films deposition, and reduce surface residues. Keywords: SiC, plasma etching, dual-frequency capacitively coupled plasma, X-ray photoelectron spectroscopy, optical emission spectroscopy PACS: 52.80.Pi, 52.70.Ds, 52.70.Kz DOI: 10.1088/1009-0630/15/10/19 1 Introduction Because of its unique properties, such as large band gap, large breakdown field, excellent thermal conductivity, great hardness, and resistance to chemical attack and radiation, silicon carbide (SiC) is considered to be an excellent candidate for high-temperature, high-power and high-frequency electronic devices [1], micro-sensor and micro-actuator applications in microelectromechanical systems (MEMS) [2], and applications in harsh environments [3]. The application of SiC to nano-devices requires precise etching of SiC and little surface residue after etching [4 6]. Unfortunately, the large Si-C bond energy makes SiC etching very difficult, therefore the powerful plasma-based dry etching becomes the main practical way to etch SiC in the fabrication of MEMS [7]. However, the surface residues resulting from dry etching can significantly affect the properties of materials, subsequent processing and device performance [8,9]. Therefore, it is very important to develop a new method to etch SiC with high etching rate, small surface roughness and little surface residue [7]. Dual-frequency capacitively coupled plasma (DF- CCP) is an important plasma etching tool which can be used for manufacturing ultra-large scale integrated (ULSI) circuits and other nano-materials. To achieve separate control of plasma density and energy of ions onto the wafer, two power sources with large discrepancy in frequency are usually used in the dual-frequency CCP reactor. The high-frequency (HF) source from 13.56 MHz to 500 MHz is used to control the plasma density, while the low-frequency (LF) source from 800 khz to 2 MHz is used to adjust the energy of ions onto the wafer [10 14]. In our previous works on etching of SiCOH low-k films in CHF 3 DF-CCP, we have found that increasing the low-frequency power can effectively suppress the surface C x F y residues [15,16]. However, the etching of SiC by DF-CCP is seldom reported, and the effect of low-frequency power on SiC etching is still unclear. Therefore, this work investigated the effect of power of low-frequency source on etching behavior of SiC in 60 MHz/2 MHz C 4 F 8 DF-CCP. 2 Experimental setup The n-type 6H-SiC wafers with the size of 5 mm 5 mm (from Tankblue Co. Ltd) were used as wafers. All the wafers were ultrasonically cleaned in HCl and NH 3 before being introduced into the vacuum chamber. After cleaning, wafers were immediately sent into the vacuum chamber. The etching of 6H-SiC wafers was carried out in a DF-CCP system. The plasma was produced between two symmetrical supported by National Natural Science Foundation of China (Nos. 10975105, 11275136, 10975106, 11175126, 11204266 and 11075114), the National Magnetic Confinement Fusion Science Program of China (Nos. 2010GB106000, 2010GB106009), the Open Project of State Key Laboratory of Functional Materials for Information and Qing Lan Project, a Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions and the Program for graduates Research & Innovation in University of Jiangsu Province, China (No. CX10B-031Z)
XU Yijun et al.: Effect of Low-Frequency Power on Etching Characteristics of 6H-SiC in C 4 F 8 /Ar DF-CCP parallel-plate stainless steel electrodes. The Comdel CV500 RF generator (60 MHz, 300 W) was used as the HF source and applied to the top electrode. The HF power was obtained by subtracting the reflected power from the incident power. The 6H-SiC wafers were placed at the bottom electrode which was driven by Comdel CX600 generator (2 MHz) in the power range of 150 W to 300 W. The diameter of the two electrodes was 200 mm and the gap between them was 50 mm. The base pressure of the chamber was approximately 10 3 Pa. It is well known that F atom can react with Si atom to produce a volatile species [17], therefore, C 4 F 8 /Ar 2 mixed gas was chosen as the etching gas. The flow rates of C 4 F 8 and Ar 2 were chosen as 10 sccm and 0.5 sccm, respectively. The 0.5 sccm Ar 2 was used as the actinometry for the measurement of relative concentration of F radicals. The working pressure was kept at 40 Pa, and the etching time was 5 min. The surface microstructure of etched 6H-SiC wafers was analyzed by atomic force microscopy (AFM) (Solver Pro SPM) with semi-contact operating mode. All the measurements were performed on an area of 500 500 nm 2. Chemical states of the pristine and etched 6H-SiC wafers were identified by a Karotos XSAN-800 X-ray photoelectron spectroscopy (XPS) with an Al Kα irradiation (1486.6 ev) powered at 168 W with a resolution of 0.9 ev. The pressure in the ion-pumped analysis chamber was 6 10 7 Pa during data acquisition. The instrument was calibrated by Au 4f 7/2 binding energy at 84.0 ev. Surface cleaning was executed on the etched 6H-SiC in the XPS chamber with 3 kev Ar + ions (10 µa/cm 2, 5 min). In order to analyze the effect of plasma on etching behavior, relative concentration of F radicals was measured using optical emission spectroscopy (OES), which was widely used in low pressure plasmas [18 21]. In our experiment the optical emission was measured in the wavelength range of 200 800 nm with a resolution of 0.05 nm to 0.13 nm using AvaSpec-2048 fiber optical spectrometer equipped with a CCD detector. The optical emission signal was collected via a fiber located at approximately 2 mm above the surface of 6H-SiC wafer. Fig. 1 shows the optical emission spectrum in the wavelength range of 680 710 nm. A very sharp and no Doppler broadening fluorine (F) line (703.8 nm) can be observed. Because the Doppler broadening originates from dissociative excitation, the result means that the F line originates from electron impact excitation rather than dissociative excitation [22]. Let n F be the concentration of the free radical F and let I F be the optical emission intensity, the excitation emission from the ground state F can be written as [23] where α F = k D (λ) 0 I F = α F n F, (1) 4πν 2 dνq F σ F (ν)νf e (ν). (2) Fig.1 Optical emission spectrum of C 4F 8 mixed with 5% Ar 2 in the wavelength range of 680 710 nm Here f e is the electron distribution function, σ F is the cross section for emission of a photon of wavelength λ due to the electron impact excitation of F, Q F is the quantum yield for photon emission from the excited state (0 Q F 1), and k D is the detector response constant. Due to a small range of overlap of f e with σ, the cross sections with values near the threshold can be replaced by σ F C F (ν ν thr ), where the C F is proportionality constant. Normalizing the F line (703.8 nm) intensity by the Ar line (750.4 nm) intensity, the concentration n F of the free radical F is where n F = C F Ar n Ar I F I Ar, (3) C F Ar = k D(λ F )Q F C F k D (λ Ar )Q Ar C Ar. (4) For the F (703.8 nm) and Ar (750.4 nm) emissions, there are λ F λ Ar, k D (λ F ) k D (λ Ar ) and Q F Q Ar. Thus, Eq. (4) can be written as C F Ar C F /C Ar, which is related to the threshold behavior of two cross sections. Because the excitation threshold energy for F and Ar is nearly the same, 14.74 ev and 13.47 ev, respectively, a good estimation of relative concentration of F radicals n F can be directly obtained by emissions intensity ratio of I F /I Ar. 3 Results and discussions The variation of etching rate with the increase of LF power is shown in Fig. 2. It can be seen that the etching rate increases with the increase of LF power, and reaches the maximum 152 nm/min at the LF power of 300 W. Fig. 3 shows the 3D AFM images of SiC wafers before and after etching. The pristine SiC wafer has a very smooth surface with the root mean square (RMS) roughness of 0.33 nm, as shown in Fig. 3(a). After etching at the LF power of 150 W, the surface of the etched SiC wafer becomes slightly rougher and the RMS roughness increases to 0.4 nm. This result indicates that the plasma etching at the LF power of 150 W causes little roughness to the wafer. If the LF power is further increased, the surfaces of etched SiC wafers become rougher and rougher. Fig. 4 shows the change of surface 1067
Plasma Science and Technology, Vol.15, No.10, Oct. 2013 RMS roughness of etched SiC wafers with the increase of LF power. It can be seen that the RMS roughness increases with the increase of LF power. When the LF power increases to 300 W, the surface RMS roughness of the etched SiC wafer reaches 0.67 nm, which is twice that of the pristine wafer. Therefore, plasma etching at higher LF power can cause serious roughness to SiC wafers. Fig.4 Dependence of RMS roughness on low-frequency power Fig.2 Dependence of etching rate on low-frequency power The dominant mechanisms related to etching behavior of 6H-SiC include physical and chemical mechanisms [17]. The physical mechanism is energetic ion bombardment. In the DF-CCP, the increase of LF power can lead to the increase of ion energy. The ion bombardment with high energy can cause more damage to SiC and result in a rougher surface. The chemical mechanism is the reaction between fluorine and silicon atoms to form volatile species [17]. In order to analyze chemical compositions of etched SiC, XPS measurements were carried out. Fig. 5 shows the C 1s corelevel spectra and their Gaussian fitting peaks of (a) the pristine SiC wafer, and SiC wafers etched at the LF power of (b) 150 W to (e) 300 W, respectively. For the pristine SiC wafer, the C 1s core-level spectrum can be deconvoluted into two components. The peaks located at the binding energies of 282.6 ev and 284.4 ev are due to Si-C and C-C bonds, respectively [24]. For the SiC wafers etched at the LF power of 150 W to 300 W, C 1s core-level spectra can be fitted into three components. The peaks located at the binding energies of 282.6 ev, 284.4 ev and 288.0 ev correspond to Si-C, C-C and C-F bonds, respectively. However, compared with the C-C peak, the Si-C and C-F peaks are very small. For wafers (b) to (e), ratios of C-C to Si-C peak area are 7.2, 4.2, 5.6 and 6.6, respectively, while the ratios of C-C to C-F peak area for wafers (b) to (e) are 3.3, 2, 2.5 and 2.4, respectively. The small Si-C peak is due to the removal of Si atoms, while the weak C-F peak indicates small Cx Fy residues on the surface of etched SiC wafers. In the C 1s core-level spectra, no C-F2 or C-F3 related bonds, which are located respectively at the binding energies of 289 ev [25] and 293 ev [26], are found. These results indicate that only very small Cx Fy residues were left at the SiC surface after etching. The etching characteristics of SiC wafers are mainly controlled by the fluorocarbon plasma chemistry. For the C4 F8 gas, the dissociations by electron-neutral collisions are shown as follows [21,27]. C4 F8 + e F + C4 F7, H = 3.00 ev, Fig.3 AFM micrographs of SiC wafers before and after etching. (a) Before etching, (b) to (e) Etching at lowfrequency power of 150 W, 200 W, 250 W and 300 W, respectively (color online) 1068 C4 F7 + e C2 F4 + C2 F3 + e, H = 2.42 ev, (5) (6) C4 F8 + e 2C2 F4 + e, H = 2.42 ev, (7) C2 F4 + e 2CF2 + e, H = 3.06 ev, (8)
XU Yijun et al.: Effect of Low-Frequency Power on Etching Characteristics of 6H-SiC in C4 F8 /Ar DF-CCP Cx Fy films deposition on the etched materials can usually be observed in the fluorocarbon plasma etching, which can influence the etching characteristics [16]. But for SiC wafers etched by the DF-CCP, the XPS results show that very small Cx Fy films deposit on the surface of SiC wafers. Therefore, the SiC etching with small Cx Fy residues can be obtained by C4 F8 DF-CCP. Fig.6 Dependence of relative concentration of F radicals on LF power 4 Fig.5 C 1s XPS spectra and their Gaussian fitting peaks of (a) original SiC, and the SiC etched at LF power of (b) 150 W, (c) 200 W, (d) 250 W and (e) 300 W, respectively (color online) The weak Cx Fy residues on the surface of the SiC wafer may depend on the Cx Fy radials in the plasma and the energy of ions onto the SiC wafer. By OES measurement, the relative concentration of F radicals is found to increase with the increase of LF power, as shown in Fig. 6. Because the F radicals are produced mainly by the dissociations of C4 F8 gas, the increase of relative concentration of F radicals means the increase of C4 F8 dissociation. As a result, more Cx Fy radials can be produced in the plasma and deposit on the SiC wafers. However, the ion energy also depends on the LF power. LEE and HUANG s results show that the increase of LF power can lead to the increase of ions energy and the broadening of ion energy distribution function (IEDF) to the high energy region [28,29]. Therefore, the ions bombardment onto the SiC surface can increase with the increase of LF power. As a result, the more the Cx Fy radicals sputtered, the less the Cx Fy films left on the etched surface. Therefore, the SiC etching involves the Cx Fy films deposition, the ions energy and the relative concentration of F radicals. With the increase of LF power, both the ions energy and relative concentration of F radicals increase. The increase of ions energy can lead to more sputtering of Cx Fy films, and the increase of F radicals leads to more removal of Si atoms in SiC by reacting with F radicals to form volatile products. Thus the etching rate increases with the increase of LF power. Conclusions In this work, the etching behavior of SiC wafers in the C4 F8 /Ar 60 MHz/2 MHz dual-frequency CCP was investigated. It is found that the etching rate and the surface roughness of etched SiC wafers are directly related with the low-frequency power. At lower LF power, the etched SiC surface is almost as smooth as the original one, and the etching rate is low. With the increase of LF power, the etching rate increases but the etched surface becomes rougher. It is due to the increase of energy of ions onto the etched SiC wafers. XPS and OES measurement results prove that the SiC etching depends on the Cx Fy films deposition, the ions energy and the relative concentration of F radicals. With the increase of LF power, both ions energy and relative concentration of F radicals increase. The increase of ions energy can lead to more sputtering of Cx Fy films, and the increase of F radicals results in more removal of Si atoms in SiC by reacting with F radicals to form volatile products, thus the etching rate increases with the increase of LF power. Compared with other plasma dry etching methods, DF-CCP can effectively inhibit Cx Fy films deposition. Therefore, the DF-CCP technique can be adopted as a useful tool for SiC etching with small surface residues. References 1 2 3 4 5 6 Choyke W J, Matsunami H, Pensl G. 1997, In Silicon Carbide, A Review of Fundamental Questions and Applications to Current Device Technology. Academic, New York Nabki F, Dusatko T A, Vengallatore S, et al. 2011, Journal of Microelectromechanical Systems, 20: 720 Hitoshi H, Yusuke K, Yutaka M, et al. 2009, Mater. Sci. Forum., 603: 655 Kim D W, Lee H Y, Kyoung S J, et al. 2004, IEEE Trans. Plas. Sci., 32: 1362 Khan F A, Foor B, Zhou L. 2001, J. Elctron Mater., 30: 212 Kim B, Lee BT. 2004, J. Electronic Mat., 33: 1308 1069
Plasma Science and Technology, Vol.15, No.10, Oct. 2013 7 So M S, Lim S G, Jackson T N. 1999, J. Vac. Sci. Technol. B, 17: 2055 8 Pang S W, Shul R J, Pearton S J. 2000, Chapter 8, Surface Damage Induced by Dry Etching. Handbook of Advanced Plasma Processing Technologies, Springer- Verlag, Berlin 9 Van der Drift E, Cheung R, Zijlstra T. 1996, Microelectron. Eng., 32: 241 10 Miyauchi M, Miyoshi Y. 2007, Solid-State Electronics, 51: 1418 11 Ishihara K, Shimada T, Yagisawa T. 2006, Plasma Phys. Control. Fusion B, 48: 99 12 Denda T, Miyoshi Y, Komukai Y, et al. 2004, J. Appl. Phys., 95: 870 13 Uchida S, Takashima S, Hori M, et al. 2008, J. Appl. Phys., 103: 073303 14 Kinoshita K, Noda S, Morishita S, et al. 1999, J. Vac. Sci. Technol. A, 17: 1520 15 Ye C, Xu Y J, Huang X J, et al. 2009, Microelectron. Eng., 86: 421 16 Shi G F, Ye C, Xu Y J, et al. 2010, Plasma Sci. Technol., 12: 437 17 Yih P H, Saxena V, Steckl A J. 1997, Phys. Status. Solidi. B, 202: 605 18 Coburn J W, Chen M. 1980, J. Appl. Phys., 51: 3134 19 Cruden B A, Gleason K K, Sawin H H. 2002, J. Appl. Phys., 91: 9547 20 Gaboriau F, Cartry G, Peignon M C, et al. 2006, J. Phys. D: Appl. Phys., 39: 1830 21 Kokkoris G, Goodyear A, Cooke M, et al. 2008, J. Phys. D: Appl. Phys., 41: 195211 22 Chen F F, Chang J P. 2002, Lecture notes on principles of plasma processing Plenum/Kluwer Publishers 23 Lieberman M A, Lichtenberg A J. 2005, Principles of Plasma Discharges and Materials Processing. John Wiley & Sons, Inc., Hoboken, New Jersey 24 Socha R P, Vaeyrynen J. 2003, Applied Surface Science, 212: 636 25 Yun S M, Chang H Y, Kang M S, et al. 1999, Thin Solid Films, 341: 109 26 Durrant S F, Rangel E C, Cruz N C, et al. 1996, Surf. Coat. Technol., 86: 443 27 Font G I, Morgan W L, Mennenga G. 2002, J. Appl. Phys., 91: 3530 28 Lee J K, Manuilenko O V, Babaeva N Y, et al. 2005, Plasma Sources Sci. Technol., 14: 89 29 Huang X J, Xin Y, Yuan Q H, et al. 2008, Phys. Plasmas, 15: 073501 (Manuscript received 30 November 2011) (Manuscript accepted 29 March 2013) E-mail address of corresponding author WU Xuemei: xmwu@suda.edu.cn 1070