Gas utilization in remote plasma cleaning and stripping applications

Gas utilization in remote plasma cleaning and stripping applications B. E. E. Kastenmeier IBM Semiconductor Research and Development Center, 2070 Rt. 52, Zip E40, Hopewell Junction, New York 12533 G. S. Oehrlein Department of Materials and Nuclear Engineering and Institute for Plasma Research, University of Maryland, College Park, Maryland 20742-2115 John G. Langan and William R. Entley Air Products and Chemicals, Inc., 7201 Hamilton Boulevard, Allentown, Pennsylvania 18195 Received 8 September 1999; accepted 26 May 2000 Nitrogen trifluoride (NF 3 ) is a likely candidate to replace perfluorocompounds PFCs in stripping and reactor cleaning applications. In this article, the performance of NF 3 for the etching of silicon, silicon dioxide (SiO 2 ), and silicon nitride (Si 3 N 4 ) is compared with that of CF 4,C 2 F 6, and C 3 F 8. The performance measures emphasized in this article are the dissociation efficiency of the parent molecule in the discharge, the etch rate, and the gas utilization. The destruction efficiency of NF 3 in the discharge as determined by mass spectrometry is typically 100%. The maximum destruction of the PFC gases for the parameters used in this investigation is approximately 75% for CF 4, and can approach 100% for C 2 F 6 and C 3 F 8. The removal rates for NF 3 obtained at optimum settings of O 2 addition and microwave power are significantly higher than those for PFC gases. The gas utilization, which describes the degree of conversion of the parent molecules into etch products and is defined in this article, is also higher for NF 3 than for the other gases investigated. 2000 American Vacuum Society. S0734-2101 00 05905-4 I. INTRODUCTION In semiconductor manufacturing and flat panel display production, processes for gasification of silicon, silicon dioxide, and silicon nitride are required. Examples are the cleaning of chambers used for plasma enhanced chemical vapor deposition PECVD, or the stripping of dielectric films from the wafer surface during IC manufacturing. Traditionally, these processes have been performed with perfluorocompounds PFCs, e.g., CF 4 or C 2 F 6. These gases, however, contribute significantly to the emission of gases relevant for the greenhouse effect due to their very high global warming potential and that of their dissociation products. 1 Semiconductor manufacturers have agreed in a Memorandum of Understanding with the U.S. Environmental Protection Agency to reduce the amount of annual PFC emissions. Since reactor cleaning and isotropic removal of dielectrics account for the major part of PFC emissions, alternative chemistries for these processes need to be investigated. The demands for these new process chemistries are a significantly reduced emission of greenhouse gases at equal or shorter run times. Nitrogen trifluoride (NF 3 ) is a likely candidate to replace PFCs for this application. In this article, its performance in stripping and cleaning applications is compared with that of CF 4,C 2 F 6, and C 3 F 8. The dissociation efficiency of these gases in the discharge is determined by mass spectrometry. Etch rates and etching yields of Si, SiO 2, and Si 3 N 4 for the four gases and their mixtures with O 2 and N 2 are presented. II. EXPERIMENT Figure 1 shows a schematic of the remote plasma system used for the experiments. The feed gases were excited using an Astex 2.45 GHz microwave applicator with a sapphire coupling tube. Nitrogen trifluoride, CF 4,C 2 F 6, and C 3 F 8 in combination with various amounts of O 2 and N 2 additives were examined. The flow of either NF 3,CF 4,C 2 F 6,orC 3 F 8 was kept constant at 300 sccm. The chamber pressure was 1000 mtorr for all experiments. The microwave power and the amount of oxygen addition were varied within the processing range typical for the respective feed gases. The microwave applicator was mounted directly on the reaction chamber for all experiments reported here. This resulted in a separation of the discharge from the sample position of approximately 40 cm. Silicon chips of 2.5 2.5 cm 2 were glued on a 125 mm carrier wafer, which was placed on an electrostatic chuck in the reaction chamber. The temperature of the sample was monitored with a fluoroptic probe which contacted the backside of the carrier wafer. It was kept constant at 10 C for all experiments. A pressure of 5 Torr of helium was maintained between the surface of the electrostatic chuck and the carrier wafer backside in order to obtain good heat conduction. Etch rates were measured in situ by monochromatic ellipsometry wavelength 632.8 nm. The very high etch rates of Si in NF 3 were determined by weight loss measurements. A quadrupole mass spectrometer Leybold Inficon Transpector is mounted on top of the reaction chamber with the distance from the orifice to the discharge being the same as that from the sample to the discharge. III. RESULTS A. Dissociation efficiencies Fluorocarbon gases and NF 3 behave in a significantly different manner in a plasma environment. The threshold for 2102 J. Vac. Sci. Technol. A 18 5, SepÕOct 2000 0734-2101Õ2000Õ18 5 Õ2102Õ6Õ$17.00 2000 American Vacuum Society 2102

2103 Kastenmeier et al.: Gas utilization in remote plasma cleaning 2103 FIG. 1. Schematic of the CDE tool. The gases are fed into the sapphire applicator, where a microwave discharge is ignited. The species effluent from the plasma travel trough tubing to the reaction chamber. The sample is placed in the center of an electrostatic chuck. A quadrupole mass spectrometer is mounted on the chamber on top of the sample, and monochromatic ellipsometry is used to determine etch rates. dissociative electron attachment of NF 3 is 0, 2,3 and approximately 2.5 ev for CF 4. 4 The zero threshold value for NF 3 is due to the fact that the bond strength of the NF 2 F bond is 2.5 ev, whereas the electron affinity of a F atom is 3.6 ev. Thus, the NF 3 molecule readily captures an electron and dissociates. The relative trends in the dissociation efficiencies of NF 3 and the fluorocarbon gases correspond to the relative magnitude of the dissociative electron attachment rate coefficients at the typical electron energies occurring in these discharges. The NF 2 radical is equally unstable in a plasma environment. Chemical reactions of the NF x, x 1 or2,to yield stable N 2 and F 2 molecules and free F atoms are energetically favored. This typically results in a very high degree of dissociation of the NF 3 parent molecules in high density discharges. In Fig. 2, typical mass spectra, sampled from the effluent of CF 4 and NF 3 based discharges, are shown. Oxygen is admixed to either CF 4 top image or NF 3 bottom image. The top image shows a strong peak at 69 amu FIG. 2. Typical mass spectra sampled downstream from the discharge. The gasflowiscf 4 /O 2 300/300 sccm top image and NF 3 /O 2 300/300 sccm bottom image. FIG. 3. Dissociation of CF 4.TheflowofCF 4 is 300 sccm, the chamber pressure 1000 mtorr. The flow of O 2 is varied from 0 to 300 sccm, the microwave power from 400 to 1000 W. (CF 3 ), and a smaller peak at 50 amu (CF 2 ), both of which are due to the dissociation of CF 4 in the ionization region of the mass spectrometer. Carbon oxides and carbon oxifluorides are detected in significant amounts. The bottom panel shows the complete destruction of NF 3 and the NF x. As products of the dissociation of NF 3 and O 2, only N 2,O 2, F 2, and NO are detected. The degree of dissociation of NF 3 and fluorocarbon gases (CF 4,C 2 F 6, and C 3 F 8 ) as a function of microwave power and O 2 addition was determined. The intensities of the CF 3 69 amu, C 2 F 5 119 amu, C 3 F 7 169 amu and NF 3 71 amu peaks were recorded for no discharge ignited, and with a discharge. The degree of dissociation can be determined from these intensities as (1 I plasma on /I plasma off ). In the case of the fluorocarbon gases, this value is a lower limit to the actual dissociation, because in the discharge-on state, radicals like CF 3, etc. and splitting products of larger molecules can potentially contribute to the intensity of the peaks. The error introduced by those species is estimated to be 10% or less. For NF 3, this method leads to accurate results within the error margin of the mass spectrometer, since the signal intensity of the NF 3 peak itself is used to determine the dissociation. Figures 3 6 show the dissociation of CF 4, C 2 F 6, C 3 F 8, and NF 3. The CF 4 dissociation depends strongly on the amount of O 2 added. 5 At the highest microwave power level, the dissociation increases by a factor of 3.5 as the O 2 flow is increased from 0 to 150 sccm. The dissociation decreases for higher flows of O 2. The dependence on the microwave power is weaker than that on O 2 addition. At an O 2 flow of 150 sccm, the dissociation increases by a factor of 1.4 when the power is increased from 400 to 1000 W. The maximum dissociation of CF 4 at 1000 mtorr and a CF 4 flow rate of 300 sccm is around 75%. Similarly, the dissociation of C 2 F 6 see Fig. 4 increases with the amount of O 2 added, but also a strong dependence on the microwave power is observed. In Fig. 5, the dissociation of JVST A - Vacuum, Surfaces, and Films

2104 Kastenmeier et al.: Gas utilization in remote plasma cleaning 2104 FIG. 4. Dissociation of C 2 F 6. The same parameters are used in Fig. 3. C 3 F 8 is shown as a function of O 2 addition and microwave power. The degree of dissociation does not depend on the O 2 flow, but strongly on the microwave power. For C 2 F 6 and C 3 F 8, dissociation values of almost 100% can be obtained. Figure 6 demonstrates complete dissociation of NF 3 at a microwave power level of 800 W and higher. The dependence of the degree of dissociation on the percentage of O 2 added to NF 3 is weak. Oxygen increases the fluorocarbon dissociation in a twofold way. First, O atoms oxidize CF x species, setting free F atoms, and forming carbon oxides and carbon oxyfluorides. 5 This new channel for CF x destruction increases the dissociation. Second, the CO x and COF 2 molecules are less likely to recombine into the mother species because of their higher stability as compared to the CF x radicals. At high amounts of O 2 addition, the dissociation tends to decrease due to power limitation. For the longer-chain fluorocarbon molecules FIG. 6. Dissociation of NF 3. The same parameters are used in Fig. 3, with the exception of the microwave power. In the experiments reported here, it was not possible to ignite a NF 3 discharge with less than 600 W of microwave power. The dissociation of NF 3 in the high-density microwave discharge is typically 100%. C 2 F 6 and C 3 F 8 ), the dissociation is less dependent on the O 2 addition than for CF 4. Instead, the dissociation is limited by the microwave power. This is due to the fact that at a given flow rate of the fluorocarbon gas, the available power is distributed over 1.75 and 2.5 times as many bonds for C 2 F 6 and C 3 F 8, respectively, than for CF 4. It is important to note that the dissociation as calculated here from the mass spectrometric plasma-on and plasma-off data describes the destruction of the parent molecule. The discharges of fluorocarbon gases produce significant amounts of molecules and radicals that do not contribute to etching reactions and even can cause fluorocarbon deposition, such as COF 2,CF 2, and larger carbon chain molecules like C 2 F 6. The production rate of these species must be taken into account if one wants to derive a measure for etch efficiencies or for the emission of greenhouse gases from mass spectrometric data alone. Therefore, in the following sections, etch efficiencies are calculated from known fluxes of gases and measured etch rates instead. Silicon etch rates of the different feed gas mixtures are compared in the next section. Removal of this material from deposition reactors is important in IC, flat panel display, and solar cell manufacturing. FIG. 5. Dissociation of C 3 F 8. The same parameters are used in Fig. 3. B. Silicon etch rates Figures 7 9 show the etch rates of silicon for different feed gas mixtures as a function of microwave power. The flow of CF 4,C 2 F 6,C 3 F 8, and NF 3 was kept constant at 300 sccm. The pressure was 1000 mtorr for all experiments. The left ordinates of the figures show the etch rates, and the feed gas utilization in percent, defined below, is shown on the right ordinates. The different images in each figure represent different amounts of O 2 added to the discharge. In Fig. 7, the J. Vac. Sci. Technol. A, Vol. 18, No. 5, SepÕOct 2000

2105 Kastenmeier et al.: Gas utilization in remote plasma cleaning 2105 FIG. 7. Etch rate of silicon left-hand side and gas utilization right-hand side as a function of microwave power. The flow of the CF 4 is 300 sccm, the pressure 1000 mtorr for all experiments. The top image shows etch rates for no oxygen added to the discharge. The flow of O 2 was 50 sccm for the center panel, and 300 sccm for the bottom panel. etch rates for CF 4 and CF 4 /N 2 feed gas mixtures are shown. No oxygen was added to the discharge for the data shown in the top image. For the data shown in the center panel, 50 sccm of O 2 was added to the discharge. At this O 2 /CF 4 ratio, the generation of atomic F from CF 4 is known to be at its maximum, 5,6 and the Si etch rates are higher by approximately a factor of 10 as compared to the top panel. Finally, the bottom image shows etch rates for 300 sccm of O 2 added to the discharge. For this O 2 flow, the Si etch rates are limited by surface oxidation and are suppressed for low microwave powers. The etch rates of silicon in the afterglows of NF 3 discharges are presented in Fig. 8 in the same fashion as in Fig. FIG. 9. Si etch rates and gas utilization for C 2 F 6 and C 3 F 8.HighflowsofO 2 were added to the discharges in order to avoid fluorocarbon deposition. Data shown are for 300 sccm top image and 600 sccm bottom image of O 2. 7. Etch rates of Si for C 2 F 6 and C 3 F 8 gas mixtures are shown in Fig. 9. These gases deposit fluorocarbon material at low O 2 additions. In the present experiments, the transition from the fluorocarbon deposition to the etching regime occurred at 70 sccm of O 2 addition for C 2 F 6, and at 200 sccm for C 3 F 8. The etch rates increase for increasing O 2 additions, and assume a maximum at around 300 sccm of O 2 for C 2 F 6, and at around 600 sccm for C 3 F 8. Therefore, the etch rate data shown are measured at 300 sccm top image and 600 sccm bottom image of O 2 addition. The Si etch rates for NF 3 reported in Fig. 8 are consistently higher than those for CF 4,C 2 F 6, and C 3 F 8. For instance, the silicon etch rate is 50 times higher for pure NF 3 than it is for pure CF 4 see Figs. 7 and 8, top image. No etching at all occurs for C 2 F 6 and C 3 F 8 at this gas mixture. When 50 sccm of O 2 is added, the Si etch rate for CF 4 increases by a factor of 5 10, while the etch rate in NF 3 remains unchanged center image, and still no etching occurs for C 2 F 6 and C 3 F 8. The etch rate in NF 3 is still higher by a factor of 5 10. At 300 sccm of O 2, the etch rate in NF 3 is 10 times as high as that in CF 4 bottom image. For this amount of O 2 addition, Si is etched using C 2 F 6 at a rate that is approximately 60% that of NF 3 compare Figs. 8 and 9 The Si etch rates for C 3 F 8 at 300 sccm of O 2 addition are still limited by the competing deposition of fluorocarbon material and do not exceed 150 nm/min. The Si etch rate with respect to the flow of O 2 reaches its maximum at 600 sccm of O 2 addition. The rate is about 900 nm/min at 1000 W of microwave power, which corresponds to 60% of the comparable NF 3 etch rate. FIG. 8. Etch rates and gas utilization during removal of Si for NF 3, presented in the same fashion as Fig. 7. C. Etch rates of SiO 2 and Si 3 N 4 and gas utilization Etch rate data are presented for SiO 2 in Figs. 10 and 11, and for Si 3 N 4 in Figs. 12 and 13. The same experimental conditions were used as above for the Si data. In particular, JVST A - Vacuum, Surfaces, and Films

2106 Kastenmeier et al.: Gas utilization in remote plasma cleaning 2106 FIG. 10. Etch rate of SiO 2, presented in the same fashion and for the same parameters as Fig. 7. Rates for the fluorocarbon gases do not exceed 15 nm/min, those for NF 3 can be as high as 40 nm/min. FIG. 12. Etch rate of Si 3 N 4, presented in the same fashion and for the same parameters as Fig. 7. Silicon nitride is etched effectively with fluorocarbon gases only when small amounts of N 2 are added to the discharge compare the curves for CF 4 and CF 4 /N 2 in the center image. The etch rates for the fluorocarbon gases can reach about 100 nm/min, but those for NF 3 are higher by approximately a factor of 2. the flows of O 2 are 0, 50, and 300 sccm. The etch rates for C 3 F 8 were measured for 300 and 600 sccm of O 2 due to the high degree of fluorocarbon deposition of this gas at low O 2 flows. For these materials, NF 3 exhibits etch rates 2 10 times faster than those of CF 4 and C 2 F 6. No significant etching of SiO 2 occurs when CF 4 with no O 2 addition is used as feed gas Fig. 10, top image. The SiO 2 etch rate for CF 4 gas mixtures remains below 6 nm/min, that for C 2 F 6 gas mixtures between 10 and 15 nm/min. Comparable rates for NF 3 can be as high as 40 nm/min Fig. 10, bottom image. Figure 11 shows SiO 2 etch rates for C 3 F 8. At 300 sccm of O 2 addition, the etch rates are suppressed by competing fluorocarbon deposition. The etch rates can be as high as 30 nm/min at an O 2 flow of 600 sccm, the setting at which the etch rate maximum occurs. Silicon nitride shows no significant etching for pure CF 4, the rate being lower than 1.6 nm/min for all power levels Fig. 12, top image. Also, the etch rate for CF 4 with 20 sccm of N 2 added is on the order of 1 nm/min. 7 The Si 3 N 4 etch rate for pure NF 3 is approximately 40 nm/min. At 50 sccm of O 2 added center image, the etch rate using NF 3 is twice as high as that using CF 4 with N 2, and at 300 sccm of O 2 bottom image, it is about 2.5 times as high. Etch rates of Si 3 N 4 for C 3 F 8 are shown in Fig. 13. They are approximately 50% of the comparable NF 3 etch rates for this material. The FIG. 11. SiO 2 etch rates for C 3 F 8, presented for the same parameters as Fig. 9. The highest SiO 2 etch rates for C 3 F 8 are about 75% of those obtained with NF 3. FIG. 13. Si 3 N 4 etch rates for C 3 F 8, presented as Fig. 9. Etch rates do not exceed 100 nm/min. J. Vac. Sci. Technol. A, Vol. 18, No. 5, SepÕOct 2000

2107 Kastenmeier et al.: Gas utilization in remote plasma cleaning 2107 TABLE I. Comparison of etch rates. The data are taken from Figs. 7 to 13. The values are the maximum rates for the respective feed gases and materials, and therefore correspond to different experimental conditions. Etch rate nm/min CF 4 C 2 F 6 C 3 F 8 NF 3 Si 220 900 900 1500 SiO 2 4 15 30 40 Si 3 N 4 75 100 100 200 TABLE II. Comparison of gas utilization. Again, the values correspond to the maximum etch rate values from Figs. 7 to 13. Utilization % CF 4 C 2 F 6 C 3 F 8 NF 3 Si 1.8 8.0 8.0 13 SiO 2 0.015 0.055 0.11 0.15 Si 3 N 4 0.56 0.75 0.75 1.5 etch rates and mechanisms of all three materials with respect to either CF 4 or NF 3 have been discussed in detail in previous publications. 6 9 The big differences in the etch rates between NF 3 and the fluorocarbon gases lead to big differences in gas utilization. For the right ordinates of Figs. 7 13, the etch rates were converted into gas utilization in percent, which is defined as the number of removed Si atoms, N Si Rem, divided by the number of CF 4,C 2 F 6,C 3 F 8,orNF 3 molecules going into the discharge, N XF, Utilization N Si Rem 100%. 1 N XF N XF is obtained from the flow of CF 4,C 2 F 6,C 3 F 8,or NF 3, which was 300 sccm for all measurements (1 sccm 2.42 10 19 particles/min. The number of removed Si atoms, N Si Rem, is obtained from the etch rate, which is assumed to be constant across the surface of the 125 mm wafer, and the density and stoichiometry of the material. The absolute utilization values calculated in this way are applicable to the remote plasma stripping of films from 125 mm wafers. For different areas-to-be-etched, the relative magnitude of the utilization values can be assumed to be conserved. The utilization is proportional to the etch rate, therefore the same discussion applies as that for the above etch rates. In the case of Si etching, the best utilization achieved in the experiments is 13% and was obtained with NF 3. The best values for Si etching with CF 4, C 2 F 6, and C 3 F 8 are 1.8%, 8.0%, and 8.0%, respectively. Also, for SiO 2 and Si 3 N 4 etching, the utilization of NF 3 is higher than that of CF 4, C 2 F 6, and C 3 F 8. Best values for the NF 3 utilization during SiO 2 etching are 0.15%, and during Si 3 N 4 etching 1.5%. We have shown that NF 3 is superior to PFC gases in both the removal rates obtained and the emission of gases relevant for the greenhouse effect. Process times for NF 3 based processes will be significantly lower than those for PFC processes, allowing for higher tool utilization or throughput. The fast removal by NF 3 can be attributed to a the high dissociation rate, which is typically 100%, b the high production rate of reactive species F, F 2, and, in the case of Si 3 N 4 etching, NO, and c the absence of polymerizing CF x species causing deposition of a fluorocarbon film. An estimation of the amount of emission of greenhouse gases for a typical CF 4 based process in million metric tons of carbon equivalent MMTCE can be obtained from the values of CF 4 dissociation and utilization listed above. This estimate takes into account only the contribution from undissociated CF 4 leaving the process chamber, and thus is a lower limit for the actual emission. The process was arbitrarily chosen as the removal of 1000 nm of Si from a 125 mm wafer, a volume of 1.23 10 4 cm 3. The emission for this process is 4.4 10 9 MMTCE or 4.4 kg of carbon equivalent when CF 4 is used at best process parameters. The corresponding value for using pure NF 3 is virtually zero, since no greenhouse gases other than NO are produced, which can be abated from the exhaust efficiently. ACKNOWLEDGMENTS The authors would like to thank Robert Ellefson and Louis Frees of Leybold Inficon for their help with mass spectrometry. This work was supported financially by Air Products and Chemicals. IV. CONCLUSIONS The performance of different feed gas mixtures in remote plasma stripping and reactor cleaning applications was investigated. The gases used were CF 4,C 2 F 6, C 3 F 8, and NF 3, and their mixtures with O 2 and N 2. Etched materials were Si, SiO 2, and Si 3 N 4. Etch rate values at best process conditions are typically higher by a factor of 3 10 for NF 3 gas mixtures than for CF 4 mixtures, and higher by a factor of 2 3 than for C 2 F 6 and C 3 F 8 mixtures. These etch rate values at best process conditions are summarized in Table I. In Table II, the best utilization for each gas mixture and material are listed. 1 T. Kawane, Semiconduct. Int. 66 1997. 2 R. M. Reese and V. H. Dibeler, J. Chem. Phys. 24, 1175 1956. 3 J. C. J. Thynne, J. Phys. Chem. 63, 1586 1969. 4 L. G. Christophorou, A. J. K. Olthoff, and M. V. V. S. Rao, J. Phys. Chem. Ref. Data 25, 1341 1996. 5 C. J. Mogab, A. C. Adams, and D. L. Flamm, J. Appl. Phys. 49, 3796 1978. 6 P. J. Matsuo, B. E. E. Kastenmeier, J. J. Beulens, and G. S. Oehrlein, J. Vac. Sci. Technol. A 15, 1801 1997. 7 B. E. E. Kastenmeier, P. J. Matsuo, J. J. Beulens, and G. S. Oehrlein, J. Vac. Sci. Technol. A 14, 2802 1996. 8 B. E. E. Kastenmeier, P. J. Matsuo, G. S. Oehrlein, and J. G. Langan, J. Vac. Sci. Technol. A 16, 2047 1998. 9 P. J. Matsuo, B. E. E. Kastenmeier, G. S. Oehrlein, and J. G. Langan, J. Vac. Sci. Technol. A 17, 2431 1999. JVST A - Vacuum, Surfaces, and Films