Journal of the Korean Physical Society, Vol. 51, No. 3, September 2007, pp. 978 983 Plasma Kinetic Study of Silicon-Dioxide Removal with Fluorocompounds in a Plasma-Enhanced Chemical Vapor Deposition Chamber Heeyeop Chae Department of Chemical Engineering, Sungkyunkwan University, Suwon 440-746 Hebert H. Sawin Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, U.S.A. (Received 13 November 2006) Plasma kinetics of silicon dioxide plasma cleaning is investigated with C 2F 6 and CF 3CF 2OCF=CF 2 (perfluoroethyl vinyl ether, PEVE) in this work. Perfluorocompounds (PFCs) are widely used in the semiconductor industry, and they are known to have long atmospheric lifetimes and high global warming potentials (GWP) due to their chemical stability and large cross sections for infrared radiation absorption. The silicon dioxide (SiO 2) cleaning process in a plasmaenhanced chemical vapor deposition (PECVD) chamber is known to be the largest PFC emission source for the semiconductor industry. Silicon-dioxide removal rates by surface reactions in plasmas were measured in the temperature range from room temperature to 400 C by using laser interferometry, and the fluorine density was determined by using an optical emission spectroscopy (OES) actinometry technique. The ion density and the ion energy were determined by using an impedance analysis and an equivalent circuit model. The activation energy was determined from Arrehenius plot to be 0.163 ev and 0.169 ev for C 2F 6 and PEVE plasmas, respectively, in the PECVD chamber cleaning condition. It is shown here that silicon-dioxide removal rate is linearly proportional to the fluorine concentration for the PECVD chamber cleaning condition of 300 400 C and that the removal rate depends on the ion flux and the ion energy below 100 C. The combined etch rate model is suggested to explain this temperature dependence in this work. From this plasma kinetic study, we can conclude that the atomic fluorine concentration is the dominant factor in silicon-dioxide removal for the PECVD chamber cleaning condition. PACS numbers: 52.77.Bn, 81.65.Cf, 82.33.Xj Keywords: Silicon-dioxide etching, Silicon-dioxide cleaning, Actinometry, Impedance analysis I. INTRODUCTION Perfluorocompounds (PFCs), such as CF 4, C 2 F 6, NF 3, and SF 6 are identified as global warming gases, along with carbon dioxide (CO 2 ), methane (CH 4 ), and nitrous oxide (N 2 O) [1]. These PFCs are widely used in various processing of semiconductor and MEMS devices [2 4]. Although the amount of PFC emission is small when it is compared with CO 2, the global warming potential (GWP) is very high and the lifetimes of PFCs are long, as shown in Figure 1. Figure 1 shows global warming potentials of various PFCs calculated on a 100-year integrated time horizon (ITH) relative to CO 2 [5]. These gases have long lifetimes of 200 to 50,000 years. The plasma cleaning of silicon dioxide in a plasmaenhanced chemical vapor deposition (PECVD) chamber is known to be the largest single PFC emission source in the semiconductor industry [1]. In the typical PECVD E-mail: hchae@skku.edu; Fax: +82-31-290-7272 -978- SiO 2 deposition processes, silicon dioxide is deposited not only on the wafer but also on the surface of chamber wall. The accumulated SiO 2 on the wall can generate particles that can cause critical defects in microelectronic devices. To remove SiO 2 in the chamber, one typically uses PFC-based plasmas. Atomic F and various ions Fig. 1. Global warming potentials and lifetimes of various perfluorocompoudns (PFCs) [5].
Plasma Kinetic Study of Silicon-Dioxide Removal with Heeyeop Chae and Hebert H. Sawin -979- Fig. 2. Experimental Setup. are considered to be the primary reactants that remove SiO 2. Typically oxygen is often introduced to increase the atomic fluorine concentration by reacting with CF x (x = 1 to 3) species. It is known that about 30 % of C 2 F 6 is converted into non-pfc in the process and about 70 % is emitted to atmosphere. In this work, the plasma cleaning kinetics of silicon dioxide in PECVD processes is studied with C 2 F 6 and CF 3 CF 2 OCF=CF 2 (perfluoroethyl vinyl ether, PEVE) by measuring the cleaning rate as a function of temperature. In this work the term removal is used to mean both etching and cleaning. This kinetic study of the SiO 2 cleaning process can contribute to process optimization for PFC reduction by providing quantitative models. II. EXPERIMENTS For the plasma cleaning study, a capacitively-coupled plasma (CCP) system was built to simulate the industrial application, as shown in Figure 2. Plasmas were generated for both SiO 2 deposition and removal between two 3-inch-diameter aluminum plates separated by a 0.5- inch gap. Both thermally grown oxide and TEOS (Si(OC 2 H 5 ) 4, Tetraethylorthosilicate) oxide deposited by using PECVD process were used. In PECVD process, SiO 2 was deposited by Alfa 99.9999 % (metal basis) TEOS/O 2 plasma. TEOS was delivered through a bubbler with He as carrier gas. About 7000 Å thick SiO 2 was deposited with deposition rate of about 2000 Å/min with the following process conditions of 5 sccm of TEOS/He mixture, 25 sccm of O 2, 2.0 Torr of pressure and 385 C of temperature. TEOS oxide is used to find the optimum cleaning condition in the wide range of process parameters of pressure, oxygen composition, and total flowrate. For the activation energy study, a 2-cm 2 -area 5000 Å-thick thermal oxide was used. For SiO 2 cleaning, C 2 F 6 /O 2 and PEVE/O 2 plasmas were generated between two aluminum electrodes. The surface temperature was controlled using resistive heaters located under the bottom electrode. A designed set of experiments was carried out by varying the oxygen composition (10 90 % for C 2 F 6 /O 2 and 50 90 % for PEVE/O 2 ), the pressure (1.2 4.0 Torr), the power (40 80 watt) and the total flow rate (20 40 sccm) to determine representative process conditions. For the activation energy determination, the following optimal process conditions were used: i) C 2 F 6 /O 2 plasma with O 2 = 50 %, 1.5 Torr, 60-watt RF power, and a total flow rate of 30 sccm, or ii) PEVE/O 2 plasma with O 2 = 70 %, 2.5 Torr, 60 watt and a total flowrate of 30 sccm. For the in situ oxide cleaning rate measurement, a He- Ne laser(uniphaser 1508) was used as the light source and a 0.27 m monochromator (Jarrell-AshR) was used as the signal detector. Data acquisition was carried out by a personal computer with a National InstrumentR Lab-PC + board. The laser signal was generated by using a He-Ne laser and was delivered to the oxide surface through fiber optics (NewportR, F-MMC). The intensity of reflected signal changed due to interference as the film thickness was varied. With a known light incident angle (θ = 90 ), wavelength (λ = 632.8 nm), and reflective index of oxide (n = 1.5), the removal rate, R, was determined with the following equation: R = λ 2n(sin θ) t. (1) where t is the etching or the deposition processing time. For the kinetic study, optical emission spectra (OES) were taken using another monochromator (ISAR HR640) from the side of the reactor. The relative concentration of atomic F was determined by using actinometry technique that uses inert Ar gas as a reference. The assumption of actinometry is that the concentration of excited species is proportional to that of ground state [7]. If the excitation cross-sections are similar between the reference gas and a radical and if the operating pressure is low enough, the relative concentration of a radical can be determined by using the peak intensity ratio. Actinometry has been developed for F, O, CO, CO 2, CF and CF 2 in various plasmas [7 9]. In this work, the relative concentration of atomic F (7037 Å) was determined using the Ar (7504 Å) peak as a reference signal: [F] = I F I Ar [Ar] (2) The ion energy and flux were estimated by using an impedance analysis technique [10]. The impedance analysis provides information on the electron density and the
-980- Journal of the Korean Physical Society, Vol. 51, No. 3, September 2007 Fig. 3. (a) Experimental impedance analysis setup and (b) equivalent circuit model. sheath potential from the impedance of a plasma by measuring the amplitude of the voltage (V RF ) and the current (I RF ) and the phase difference (θ) between the voltage and the current. The impedance of the plasma was determined by measuring the voltage waveform with a high-voltage probe (TektronixR P6015) and by measuring current waveform with a current probe (Ion Physics CompanyR, CM-100-M). Both the RF voltage and current were measured between the RF matching network and the top electrode. The waveforms were collected by using a digital oscilloscope (LeCroyR 9400). The impedance was interpreted with the equivalent circuit model shown in Figure 3(b). The plasma sheath was modeled with a capacitance and a plasma bulk with a resistance. The electron density was determined from the resistance of the bulk plasma, and the ion energy was determined from the voltage drop across the capacitor. The phase difference also depends on the cable length in the radio-frequency signal, and the phase difference was calibrated with the no-plasma condition (100 % reflected) as a reference. That is, the phase angle difference for the no-plasma condition was set to 90. The following relations were used to determine the ion energy and the plasma density: Z = A + Bi (3) A = V RF I RF cos θ (4) B = V RF sin θ I RF (5) A R b = CpA 2 2 ω 2 + CpB 2 2 ω 2 2CpBω 2 2 + 1 (6) C b = 2 C2 pa 2 ω 2 + C 2 pb 2 ω 2 2C 2 pbω 2 + 1 C 2 pa 2 ω 2 + C 2 pb 2 ω 2 B (7) V S = I RF ωc S (8) d n e = qsr b µ e (9) Fig. 4. Relation between SiO 2 removal rate and relative atomic F concentration at 385 C. (a) C 2F 6/O 2 Plasma (b) PEVE/O 2. where A is the real part of the impedance (Z), B is the imaginary part of the impedance, R b is the resistance of the bulk plasma, C s is the sheath capacitance, C p is the parasitic capacitance, V s is the sheath voltage, n e is the electron density, d is the gap spacing of the electrodes, S is the electrode area, µ e is the electron mobility in a plasma. The parasitic capacitance (C p ) was determined from the no-plasma condition of a 90 phase difference with 100 % reflected power at low pressure ( 0.01 Torr). C p was 22 pf for the system. The electron mobility was estimated to be about 30 m 2 /V-s from the experimental data of Naidu and Prasad [11]. The resistance (Eq. (6)) and the capacitance (Eq. (7)) were determined by using the symbolic computation software MapleR. III. RESULTS & DISCUSSION Figure 4 shows the correlation between the SiO 2 removal rate and the relative F concentration determined by using the intensity ratio of the F peak (7037 Å) to the Ar peak (7504 Å). Figure 4 shows that the removal rate is linearly increasing with the relative F concentra-
Plasma Kinetic Study of Silicon-Dioxide Removal with Heeyeop Chae and Hebert H. Sawin -981- Fig. 5. Arrhenius plots and activation energies determined from SiO 2 removal rates. Process conditions: (a) C 2F 6/O 2 with O 2= 50 %, pressure = 1.5 Torr, power = 60 Watt and total flow rate = 30 sccm (b) PEVE/O 2 with O 2= 70 %, pressure = 2.5 Torr, power = 60 Watt, and total flow rate = 30 sccm. (a) C 2F 6/O 2 Plasma (b) PEVE/O 2 tion. This F concentration dependence is not typically observed in silicon-dioxide removal at temperatures below 100 C, where the removal rate strongly depends on the ion flux and the ion energy [12]. The linear correlation between the silicon-dioxide removal rate and the F atom concentration suggests that atomic fluorine is the dominant reactant and that the ion flux and the composition are relatively unimportant at an elevated temperature of 385 C, the PECVD chamber cleaning condition. Different species other than atomic F between those two plasmas is believed to be responsible for the difference in the slopes. Figure 5 shows apparent activation energies determined by using Arrhenius plots. To determine the activation energy from the measured removal rate, we reflected the temperature effect for the relative atomic F concentration. The removal rates were adjusted for the flux, which is a function of speed and gas density. The gas temperature was assumed to be T gas T surface + 200 K. Both plots show lower apparent activation energy below 100 C (373 K) and higher apparent activation energy Fig. 6. (a) Sheath potential and (b) electron density at different temperatures. above 300 C (473 K). Flamm and coworkers etched SiO 2 downstream of the F 2 discharge and measured an activation energy of 0.163 ev at a temperature between 20 C and 90 C [13]. Poulsen etched SiO 2 in a barrel etcher and measured an activation energy of 0.20 ev [14]. Reinberg obtained 0.16 ev for SiO 2 removal using a shielded discharge in a radial flow reactor [15]. In those studies, the removal process was attributed to the reaction of F with the surface, under little or no ion bombardment. Yin and coworkers measured the activation energy for plasma removal of silicon dioxide as 0.02 ev near room temperature in reactive ion etching (RIE) [16]. Butterbaugh measured an activation energy of 0.03 ev for SiO 2 removal in a similar reactor [17]. The low activation energy was attributed to an ion-enhanced removal mechanism. The activation energy measurements in this work indicate that ion-enhanced removal is the dominant mechanism below 100 C, but spontaneous removal with the F atoms is dominant above 300 C. This result clearly shows that the atomic fluorine concentration is the most dominant factor in silicon dioxide cleaning for surfaces at the process temperature. The sheath potential and the electron density determined from the impedance analysis are shown in Fig-
-982- Journal of the Korean Physical Society, Vol. 51, No. 3, September 2007 ure 6. Here the ion energy is represented by the sheath potential, and the ion density is assumed to be equal to the electron density. Overall C 2 F 6 /O 2 plasmas show a lower sheath potential and a higher electron density than PEVE/O 2 plasmas. The higher electron density of C 2 F 6 /O 2 plasmas can also explain the higher SiO 2 removal rate for C 2 F 6 /O 2 plasmas than for PEVE/O 2 plasmas. A model for the SiO 2 removal rate is suggested by combining the spontaneous removal rate and the ionenhanced removal rate as follows: R total = R spontaneous + R ion enhanced, (10) where R total is the total removal rate, R spontaneous is the removal rate by spontaneous chemical removal, and R ion enhanced is the removal rate due to ion-enhanced removal mechanism. Flamm and coworkers developed the following rate expression from oxide removal experiments without any ion bombardment by using the downstream F 2 plasma [13]: R spontaneous = 8.29 10 13 exp( 0.163 ev/kt)t 1/2 [F], (11) where k = 8.55 10 5 ev/k (Boltzmann constant), and [F] is the atomic fluorine concentration in units of cm 3. Note that the activation energy is close to the activation energy obtained in this work at high temperatures. This is also consistent with other results [18]. For ion-enhanced removal, the model developed by Gray et al. is adopted [19]. They studied ion-enhanced removal of an oxide with ion and neutral beams and modeled the ion-enhanced removal rate by using the ion flux, the ion energy, and the surface fluorination: R ion enhanced = βiθ F /ρ (12) β = b(e 1/2 4 1/2 ), (13) θ = sq/(sq + 4βI), (14) where β is the ion-enhanced removal yield, θ F is the surface fluorination, b is the energy dependence parameter, s is the sticking coefficient, Q is the neutral flux, I is the ion flux, and ρ is the density of SiO 2 (2.27 g/cm 3 ). The surface fluorination (θ F ) can be approximated to 1 because the neutral flux (Q) is typically more than 3 orders of magnitude higher than the ion flux (I) in a typical capacitively-coupled plasma. The ion-enhanced removal can be summarized as R ion enhanced = b(e 1/2 4 1/2 )I/ρ. (15) In this work, the ion energy (E) was estimated by using the sheath potential V s as shown in Figure 8. The ion flux was estimated from the electron density by using the following Bohm sheath criteria [20]: I = 0.6n e (kt e /m i ) 1/2, (16) where n e is the plasma density, T e is the electron temperature, and m i is the ion mass. CF + 3 is assumed to be the Fig. 7. Modeling results for combined spontaneous and ion-induced SiO 2 removal. (a) In C 2F 6/O 2 Plasma (b) In PEVE/O 2 Plasma dominant ion. Electron temperature was assumed to be 4 ev. These two expressions were combined to explain the temperature dependence of the removal rate. The ion energy and the flux were determined by using voltage and current measurements before the top electrode: R total = k 0 exp( E a /kt )T 1/2 [F ]+b(e 1/2 4 1/2 )I/ρ(17) Figure 7 shows modeling results with the F density and the ion-enhanced removal yield as the adjusted parameters. The atomic fluorine density was set to 9.0 10 15 cm 3 for the C 2 F 6 /O 2 plasma and to 5.5 10 15 cm 3 for the PEVE/O 2 plasma. The higher atomic fluorine density of the C 2 F 6 /O 2 plasma can also be explained by the higher ion density measured by using the impedance analysis. The yield parameter b was set to 0.15 for the C 2 F 6 /O 2 plasma and 0.025 for the PEVE/O 2 plasma in ion-enhanced removal. The difference in the ion-enhanced removal yield coefficients is possibly due to the differing compositions of reactive ions. CF + 3 is known to be the dominant ion in a C 2 F 6 plasma while other heavier ions could be responsible in a PEVE/O 2 plasma. This model shows good agreement with the experimental results. The modeling and experimental results show that ion-enhanced removal is dominant at low
Plasma Kinetic Study of Silicon-Dioxide Removal with Heeyeop Chae and Hebert H. Sawin -983- temperatures and that chemical removal is dominant at high temperatures. IV. CONCLUSION In this work, the plasma kinetics of SiO 2 removal is studied over a wide range of temperatures from room temperature to 400 C. SiO 2 removal rate model was suggested to explain the temperature dependence. This kinetic study of this research shows that SiO 2 removal by spontaneous chemical removal is dominant at high temperatures of 300 C 400 C, while SiO 2 removal by ion-enhanced removal is dominant below 100 C. This result suggests that atomic F concentration is the most important factor in SiO 2 cleaning process. Therefore, an alternative chemical for PFCs in the SiO 2 cleaning application should be able to produce a high concentration of atomic F. ACKNOWLEDGMENTS The authors would like to acknowledge Dr. Michael T. Mocella for the valuable discussion and DuPont Fluoroproducts for the financial support. REFERENCES [1] M. T. Mocella, SRC Technology Transfer Course, (Cambridge, MA, 1995). [2] J.-H. Kim, K.-W. Chung and Y.-S. Yoo, J. Korean Phys. Soc. 47, 249 (2005). [3] J. H. Park, N. -E. Lee, J. Lee, J. S. Park and H. D. Park, J. Korean Phys. Soc. 47, S422 (2005). [4] B. Kim, J. Kim, S. H. Lee, J. Park and B. T. Lee, J. Korean Phys. Soc. 47, 712 (2005). [5] V. Mohindra, H. Chae, H. H. Sawin and M. T. Mocella, IEEE Trans. on Semiconductor Manufacturing 10, 399, (1997). [6] A. R. Ravishankara, S. Solomon, A. A. Turnipseed and R. F. Warren, Science 259, 194 (1993). [7] J. W. Coburn and M. Chen, J. Appl. Phys. 51, 3134 (1980). [8] R. d Agostino, F. Cramarossa, S. De Benedictis and G. Ferraro, J. Appl. Phys. 52, 1259 (1981). [9] L. D. B. Kiss, J-P. Nicolai, W. T. Conner and H. H. Sawin, J. Appl. Phys. 71, 3186 (1992). [10] J. W. Butterbaugh, L. D. Baston and H. H. Sawin, J. Vac. Sci. Technol. A 8, 916 (1990). [11] M. S. Naidu and A. N. Prasad, J. Phys. D: Appl. Phys. 5, 983 (1972). [12] J. Ding, J.-S. Jenq, G.-H. Kim, H. K. Marynard, J. S. Hamers, N. Hershkowitz and J. W. Taylor, J. Vac. Sci. Technol. A 11, 1283 (1993). [13] D. F. Flamm, C. J. Mogab and E. R. Sklaver, J. Appl. Phys. 50, 6211 (1979). [14] R. G. Poulsen, J. Vac. Sci. Technol. 14, 266 (1977). [15] A.R. Reinberg, in Etching for Pattern Definition, edited by H.G. Hughes and M. J. Rand (Electrochem. Soc., Princeton, New Jersey, 1976), p. 91. [16] G. Z. Yin, M. Ben-Dor, M. S. Chang and T. O. Yep, J. Vac. Sci. Technol. A 7, 691 (1989). [17] J. W. Butterbaugh, Ph.D. Thesis, Massachusetts Institute of Technology, 1990. [18] C. M. Melliar-Smith and C. J. Mogab, in Thin Film Processes, edited by J. L. Vossen and W. Kern (Academic Press, New York, 1978), p. 497. [19] D. C. Gray, I. Tempermeister and H. H. Sawin, J. Vac. Sci. Technol. B 11, 1243 (1993). [20] B. Chapman, Glow Discharge Processes (John Wiley & Sons, New York, 1980), p. 69.