Spatially resolved mass spectrometric sampling of inductively coupled plasmas using a movable sampling orifice

Transcription

1 Spatially resolved mass spectrometric sampling of inductively coupled plasmas using a movable sampling orifice Xi Li a),b) and Gottlieb S. Oehrlein a),c) Materials Science and Engineering and Institute for Research in Electronics and Applied Physics, University of Maryland, College Park, Maryland Marc Schaepkens a),d) Performance Coatings Lab, PSCT, K1-5A67 1 Research Circle, GE, Niskayuna, New York Robert E. Ellefson and Louis C. Frees Inficon Incorporated, East Syracuse, New York Received 16 April 2003; accepted 25 August 2003; published 15 October 2003 A quadrupole mass spectrometer equipped with a special sampling tube and positioned on a linear-motion table is shown to be capable of obtaining important information on the variation of the discharge properties with position. We measured signals obtained by post-ionization of species sampled from Ar and C 2 F 6 high-density plasmas using electron ionization energies of 30 and 16 ev. The variation of the ion signal with position of the sampling orifice inside the reactor provides evidence of the spatial nonuniformity of the discharge and the processes that control these variations. For Ar discharges we show evidence of gas heating which produces a nonuniform radial distribution of the neutral species. The neutral species are depleted in the center of the discharge due to plasma heating. The neutral gas temperature that can be extracted from the variation in density is consistent with temperature measurements using spectroscopic probes. We also monitored the intensity of different fluorocarbon ions as a function of sampling position. We find that the fluorocarbon mass spectrometer signals show qualitatively different dependencies on sampling position. For instance, the CF 2 signal increases towards the center of the discharge and decreases towards the wall of the chamber, whereas the CF 3 signal shows a maximum near the reactor wall. The CF 3 signal also grows as the wall temperature increases, indicating production at the wall. The application of this setup to spatially resolved monitoring of etching and deposition processes is also described American Vacuum Society. DOI: / I. INTRODUCTION For chemically reactive discharges, the monitoring of both ions and neutral is required to obtain information on the relative roles of these species. For instance, for fluorocarbon discharges used for etching of SiO 2 films, 1 4 both neutral and ionic species play an important role in determining the etching characteristics. The measurement of neutral species is therefore required for the understanding of plasma and surface processes. In addition, the discharges are typically spatially nonuniform, and information of the discharge parameters as a function of position is required. Spatial distributions of neutral and ion temperatures measured by laserinduced fluorescence in inductively coupled argon plasmas have previously been reported. 5 Radial profiles of the electron temperature, electron density, and neutral atom density have also been measured using Thompson and Rayleigh scattering of laser radiation in a radio-frequency inductively coupled argon plasma. 6 Measurements of spatial and temporal variations of CF and CF 2 radial densities using laserinduced fluorescence in high-density CF 4 and CHF 3 helicon a Work performed while the authors were with the Department of Physics, State University of Albany, Albany, New York b Electronic mail: lixi@glue.umd.edu c Electronic mail: oehrlein@glue.umd.edu d Electronic mail: schaepke@crd.ge.com wave discharges have been described. 7,8 For the nonintrusive techniques, laser-induced fluorescence LIF 7 and Thompson and Rayleigh scattering of laser radiation, 6 the number of species that may be characterized is limited. Mass spectrometry MS enables the measurement of a broad set of relevant species in a plasma. By using a moving sampling orifice, spatially resolved measurements of the plasma species may be performed. For instance, axial distributions of CF 2 and CF 3 radical densities have been obtained using MS for a CF 4 plasma by Hikosaka et al. 9 A movable sampling tube with an orifice at the end has been used by Kastenmeier et al. 10 to probe the downstream etching environment for a microwave excited discharge. Employing a movable sampling tube for mass spectrometric sampling of process plasmas, it is possible to improve the signal and the relevance of the signal by probing close to a particular location. This approach is tested in this work for diagnostics of both Ar and C 2 F 6 plasmas. We described interferences with plasma sampling, e.g., heating of the sniffer tube due to plasma, and also describe insights that would be difficult to obtain without the possibility of moving the sampling orifice. Knowledge of the variation of a specific mass spectrometer signal with position provides important additional information that is helpful for the interpretation of the measurements. The results are compared with those reported previously J. Vac. Sci. Technol. A 21 6, NovÕDec Õ2003Õ21 6 Õ1971Õ7Õ$ American Vacuum Society 1971

2 1972 Li et al.: Spatially resolved mass spectrometric sampling 1972 FIG. 1. Schematic of the spatially resolved mass spectrometry apparatus employed in this work. II. EXPERIMENTAL APPARATUS AND PROCEDURES The inductively coupled plasma ICP system used for this work has been described by Standaert et al. 11 Briefly, a planar coil generates the plasma through a 19 mm thick, 230 mm diameter quartz coupling window. The coil is powered through a matching network by a MHz, W power supply. The distance from the coil to the wafer is 9 cm. A mechanical confinement ring is located below the quartz window to confine the plasma. The confinement is thermally isolated from the other parts of the reactor and heats up during the experiment. The inner diameter of the confinement ring is 20 cm. A 5 in. diameter Si wafer covered with either SiO 2 or Si and mounted on an electrostatic chuck was employed for the etching studies. In this case a rf bias 3.4 MHz was applied to the chuck. Figure 1 shows a schematic diagram of the mass spectrometer MS with the movable sampling tube used in this work. The sampling tube of the MS system consists of a 1/4 in. outside diameter ceramic tube length 19.5 in. mounted on the housing of the closed ion source mass spectrometer CISMS with a VCR connection. The CIS has been described by Li et al. 12 The orifice diameter is in. The distance between the tube axis and the wafer surface was kept constant 0.35 in. for all experiments. Information on surface and plasma processes could be obtained and distinguished by aiming the orifice either towards the wafer surface or the plasma, respectively. The linear motion table is controlled by a computer and moves approximately 10 in. from the reactor wall across the wafer. The sampling tube passes through a 1.6 in. high by 0.63 in. wide hole in the confinement ring when it moves from the reactor wall to the wafer. Spatially resolved measurements of the ion current density using a Langmuir probe have shown that the hole in the confinement has a negligible effect on the plasma uniformity. Figure 2a shows the sampling geometry of the reactor. The gas is injected through a gas injection ring not shown that is mounted inside and at the top of the mechanical confinement wall. In Fig. 2b we display the measured CF 3 intensity for a C 2 F 6 plasma at 16 ev electron energy as a function of position at 1400 W source power, 6 mtorr pressure, and 40 sccm gas flow. We notice that the CF 3 signal is peaked near the mechanical confinement wall MCW, but depleted near the center of the wafer. FIG. 2. Sample geometry of reactor structure for the spatially resolved mass spectrometry a and a measured intensity curve of the CF 3 signal by MS forac 2 F 6 plasma at a fixed 16 ev electron energy for 1400 W rf source power, 6 mtorr pressure, and 40 sccm total gas flow. J. Vac. Sci. Technol. A, Vol. 21, No. 6, NovÕDec 2003

3 1973 Li et al.: Spatially resolved mass spectrometric sampling 1973 FIG. 3. Comparison of spectra for C 2 F 6 plasma at 16 and 30 ev electron energies with and without a plasma at 6 mtorr pressure, 1400 W rf power, and 40 sccm total gas flow. In this work we examined discharges produced using Ar, C 2 F 6, and C 4 F 8 /Ar feed gases. We employed gas flow rates of 40 sccm, and operating pressures of 6 30 mtorr. The inductive source power was varied between 600 and 1400 W. The distance from the reactor wall to the wafer center is 5.7 in., and the wafer center is chosen as the origin for all curves obtained with the MS. The starting point for all measurements was the reactor wall. The electron ionization energy of the MS was set at either 16 or 30 ev. This improves the contribution of neutral radical species to the overall signal by minimizing fragmentation effects. III. RESULTS AND DISCUSSION A. Fluorocarbon discharges 1. Data interpretation a. Effect of tube sampling on mass spectra. We initially investigated the potential influence of the sniffer tube on measured mass spectra of fluorocarbon discharges. Figure 3 shows mass spectra measured for C 2 F 6 with and without plasma excitation at 16 and 30 ev electron energies at 1400 W source power, 6 mtorr pressure, and 40 sccm gas flow. The sampling orifice was located at the wafer center and pointed upward. For operating conditions without a plasma, the CF 3,C 2 F 5,CF, and CF 2 and other signal intensities are higher for 30 ev electron energy than that for 16 ev electron energy. At 16 ev electron energy, the CF 3,CF 2, CF, and C 2 F 5 species signal intensities are higher with a plasma than without a plasma. These differences can be explained by both strong plasma-induced dissociation of the gas and enhanced formation of these species by electron impact at 30 ev relative to 16 ev in the mass spectrometer. The plasma also produces HF,Si /CO, COF 2, and similar FIG. 4. Experiment designed to illustrate MS signal intensity drifts due to probe heating in an exaggerated fashion at 1400 W rf inductive power, 6 mtorr pressure, and 40 sccm total C 2 F 6 gas flow. The electron energy of the MS was set to 16 ev. product species. For both 30 and 16 ev electron energies, CF 3 is the dominant species in the C 2 F 6 plasma. Although we attempted to obtain information on changes in neutral fluorocarbon species in the C 2 F 6 discharge by employing electron ionization energies of both 30 and 16 ev, this was only partially successful. When the electron energy is set at 30 ev the signal obtained with the MS is mainly produced by dissociative ionization of the parent gas see Fig. 3, no plasma data. The threshold energy for fragmentation of C 2 F 6 to produce CF and CF 2 in the quadrupole mass spectrometer QMS is about 17 ev Ref. 12. However, even when an electron ionization energy of 16 ev was used for C 2 F 6 without a plasma, CF and CF 2 were detected, possibly due to a spread in electron energies in our mass spectrometer. b. Sniffer tube heating effects. Initial experiments showed that drifts in the signal intensities of certain fluorocarbon species were observed when measuring radial profiles for C 2 F 6 discharges at slow scanning speeds. Figure 4 shows these signal changes in an exaggerated fashion for different sampling positions in the discharge. The sniffer orifice was positioned either at the wafer edge close to the wall close edge where the sniffer setup is mounted to the ICP, the center of the wafer, or the wafer edge far from the wall far edge. The ionization energy in the mass spectrometer ion source was set to 16 ev for this experiment and the orifice is facing upward into the C 2 F 6 plasma 40 sccm, 1400 W inductive power, and 6 mtorr operating pressure. Figure 4 JVST A - Vacuum, Surfaces, and Films

4 1974 Li et al.: Spatially resolved mass spectrometric sampling 1974 shows the intensities of masses 31 (CF ), 50(CF 2 ), and 69 (CF 3 ) as a function of time. When the sniffer orifice was positioned at the wafer center, the intensities of CF 2 and CF 3 ion signals showed a slow increase, while the CF ion signal intensity remained almost unchanged. At 85 s, the sniffer was quickly repositioned to the far edge and was in the new position at time 90 s. An increase of the signal intensities of all species was observed with time. At 240 s, the sniffer was again quickly moved to the center of the reactor. The CF and CF 2 ion signals decayed, whereas the CF 3 ion signal increased slightly and then decreased. After the signal intensities became stable, the sniffer was quickly positioned to the close edge. A slow increase in the CF and CF 2 intensities was observed. The signal intensities of CF and CF 2 ions decreased slowly, and the CF 3 ion signal intensity continued to decrease after the sniffer had been positioned at the close edge. The above observations can be explained by heating of the sniffer tube in the plasma. The sniffer tube has poor thermal contact to the rest of the reactor and will heat up once it is exposed to the plasma. In an earlier study Ref. 13, it was found that fluorocarbon deposition on a surface decreased significantly with increasing temperature and in extreme cases, deposited fluorocarbon material can leave the surface at elevated temperatures. The same effects are expected to occur inside the sniffer tube. As the temperature of the sniffer tube increases when the sniffer is advanced into the plasma, less fluorocarbon species are lost to the sniffer tube walls by deposition processes. For a hot sniffer tube, these species will reach the mass spectrometer ionization source and be detected. For a low temperature sniffer tube these species would be lost in part to the internal surfaces of the sniffer tube. In extreme cases, fluorocarbon material deposited earlier in the sniffer tube could leave the tube wall and be transported toward the mass spectrometer ionization source and be detected. The data of Fig. 4 show that the CF 2 and CF signals appear most affected by the sniffer temperature, since they show a much greater relative change than the CF 3 signal. This may indicate that the CF 3 is not significantly lost in deposition reactions. In order to interpret experimental results obtained with the sniffer, it is necessary to understand the sniffer heating mechanism. The heating of the sniffer is mainly due to the impact of ions that are accelerated over the sheath between the plasma and the electrically floating sniffer surface, see Fig. 5. For our conditions, the sheath voltage is about 15 V and the ion flux toward the sniffer is about 15 ma/cm 2. The heat dissipated by the ions needs to be transported through the total sniffer tube towards the mass spectrometer flange/ rest of the reactor, which acts as a heat sink. In the above heating mechanism, it is assumed that in the temperature range of interest radiation plays no significant role in the heat loss from the sniffer tube. The mechanism can be mathematically described by a simple differential equation: FIG. 5. Schematic diagram of the sniffer heating mechanism and heat transport in the sniffer tube. pc T 2 x t Tgx,t, 1 where p is the density, c is the heat capacity, is the thermal conductivity of the sniffer tube material, and g(x,t) is the sniffer heat source: gx,t ICDx,tV shx,t2r, 2 A cr where ICD is the ion current density, V sh is the sheath potential, R is the sniffer tube radius, and A cr is the crosssectional area through which the heat conduction towards the heat sink needs to occur, i.e., the shaded area in Fig. 5. The boundary conditions that are needed to numerically solve the differential equation are given by 1 that the temperature at the flange/heat sink stays at the reference temperature, i.e., room temperature, and 2 that both the initial sniffer temperature profile, and the temperature gradient at the end of the sniffer located in the plasma, are zero. When calculating maximum temperatures and temperature profiles in the sniffer tube, using the above mathematical model, it was found that the variation in temperature could be greatly reduced if the sniffer tube was already at an elevated temperature at the beginning of the experiments due to the slow cooling rate of the tube. The rest of the fluorocarbon plasma mass spectrometry data described in this article was therefore obtained using a preheated probe. Apre- heated probe minimizes intensity variations and drifts in the fluorocarbon species densities due to changes in surface fluorocarbon deposition otherwise introduced by probe heating during the experiment. This is the preferred way of using this tool for fluorocarbon plasmas Radial profiles In the following we report radial profiles of mass spectrometer signals obtained by pointing the sampling orifice either towards the plasma or the wafer surface and setting the electron ionization energy of the MS to 16 ev. We reduced the electron ionization energy to this low value to reduce the importance of dissociative ionization of C 2 F 6 relative to direct ionization of CF, CF 2, and CF 3 species. However, our primary interest is in the spatial variations of these signals. Figure 6 shows the radial distributions of the CF 2 and CF 3 ion signals obtained with the C 2 F 6 discharge at 600 and 1400 W rf powers, 6 mtorr pressure, and 40 sccm gas flow. J. Vac. Sci. Technol. A, Vol. 21, No. 6, NovÕDec 2003

5 1975 Li et al.: Spatially resolved mass spectrometric sampling 1975 FIG. 6. Radial distributions of a CF 2 and CF 3 signal produced in the C 2 F 6 plasmas orifice up and down at 600 and 1400 W rf powers, 6 mtorr pressure, and 40 sccm total gas flow. Signals obtained with the orifice pointing up or down are displayed in the figure. The CF 2 and CF 3 MS signal intensities are higher when the sampling orifice is located in the plasma bulk rather than at the reactor wall. The CF 2 intensity increases with inductive power when the orifice is pointing upward, but is nearly the same when the orifice points towards the wafer. The CF 3 radial distribution is strongly peaked near the wall at high inductive power, but does not show this peak at 600 W, or when the orifice is pointing towards the wafer. The fact that both the CF 2 and CF 3 signals increase with inductive power indicates that a significant portion of these signals is due to the ionization of CF 2 and CF 3, rather than the dissociative ionization of C 2 F 6, where a decrease of the CF 2 and CF 3 signals with inductive power is expected. B. Wall temperature effect Schaepkens et al. 13 showed that the temperature of the mechanical confinement wall MCW is an important parameter in etching processes using inductively coupled fluorocarbon plasmas. The fluorocarbon deposition rate onto the wall was found to be strongly dependent on the MCW temperature. In this work the dependence of fluorocarbon signals on wall temperature was investigated by moving the orifice of the sniffer tube next to the confinement wall. Figure 7 shows the CF, CF 2, and CF 3 signals as a function of time for a 1400 W source power, 6 mtorr pressure, and 40 sccm gas flow C 2 F 6 plasma. Initially, the confinement wall was at room temperature. Plasma operation heated the thermally isolated confinement wall. After 300 s the confinement wall reached a temperature of 200 C. After the plasma was turned on, the intensities of CF and CF 2 ion signals decreased at first slightly see Figs. 7a and 7b, possibly related to a change in fluorocarbon film deposition at the wall. After the MCW started to heat up at about 75 s, the intensities of the CF, CF 2, and CF 3 signals FIG. 7. Dependence of the CF x (x1 3) signals on the MCW temperature at 1400 W rf source power, 6 mtorr pressure, and 40 sccm total gas flow. The MCW was initially at room temperature and reached about 200 C as measured with a thermal couple after 300 s of plasma operation. increased. This increase may be attributed to the release of fluorocarbon deposition precursors from the hot wall. This result is consistent with results from Schaepkens et al., 13 who reported that fluorocarbon species, that are precursors for fluorocarbon film deposition on low temperature surfaces, do not deposit on hot surfaces and remain in the plasma. Chinzei et al. 14 reported that no fluorocarbon deposition occurs on surfaces at temperatures higher than 200 C inac 4 F 8 discharge, and the density of CF and CF 2 radicals as measured by mass spectrometry is one order of magnitude higher in a heated reactor at 200 C in comparison to a reactor at 30 C. Consistent with the observation shown in Fig. 7c, Hikosaka et al. 15 reported an increase of the CF 3 mass spectrometric signal as the temperature of the quartz wall in their reactor increased. C. Effect of mass spectrometer orifice location on the time dependence in changes of the measured signals We compared the time dependence of the mass spectrometer signals obtained for a fluorocarbon etching process on the location of the sampling orifice. A C 4 F 8 /40%Ar plasma process was used for etching of a SiO 2 film over a Si substrate. The C 4 F 8 /40%Ar discharge was produced using 600 W rf source power, 40 sccm total gas flow, and 20 mtorr pressure. A self-bias voltage of 100 V was applied to the wafer. For this test, three mass spectrometers were mounted on the same process chamber. The first mass spectrometer was mounted downstream and sampled the effluent in the pumping line before the turbopump. The second mass spectrometer was mounted on the reactor wall and sampled the JVST A - Vacuum, Surfaces, and Films

6 1976 Li et al.: Spatially resolved mass spectrometric sampling 1976 FIG. 8. Comparison of real-time monitoring and response times for three different mass spectrometer sampling configurations for a fluorocarbon SiO 2 etching process. One of the mass spectrometers sampled from the pumping line, a second mass spectrometer sampled at the reactor wall, and in the third configuration the sampling orifice of the sniffer tube pointed towards the wafer. A C 4 F 8 /40%Ar plasma at 600 W rf power, 40 sccm gas flow, 20 mtorr pressure, and 100 V self-bias voltage was used for the etching process. FIG. 9. Typical radial profiles of Ar species signal intensities measured with the MS system at 1400 W inductive power. a The operating pressure was 6, 10, 20, and 30 mtorr, respectively. b Radial distribution of the normalized Ar species signal intensity. The signal intensity at the wafer center was set to 1. The electron energy of the MS was set to 30 ev. plasma directly. The third mass spectrometer employed the sniffer tube with the orifice pointing towards the etching sample. We selected mass 28 (CO or Si ) since it is a good choice for monitoring the SiO 2 etching process. The mass 28 time evolution is shown in Fig. 8. During the process initially fluorocarbon etching took place fluorocarbon films deposited for plasma operating conditions without rf bias were etched back, followed by etching of the SiO 2 film down to the Si underlayer. After overetching, the plasma was then extinguished. For the mass spectrometer employing the orifice located in the pumping line a large increase is seen when the plasma is ignited, which likely corresponds to CO produced by capacitive coupling of the coil to the plasma and erosion of the quartz coupling window. However, the SiO 2 etch end point is not apparent in this case as a change of the CO signal. The CO signal obtained with the sniffer configuration shows a clearly defined etching end point also detected simultaneously by ellipsometry. The response time is shortest in this case relative to the other sampling locations. D. Gas heating in an argon discharge In this section we report on how the radial dependence of the mass spectrometer signals measured with the sniffer configurations can be used to obtain information on the gas temperature. Radial temperature distributions of neutral and ionic Ar species have been reported for a high-density plasma by Hebner. 5 In that work it was shown that the neutral gas temperature had a maximum in the center of the discharge and decreased towards the edge of the discharge. The ion temperature increased at the edge of the discharge. Measurements of electron and neutral atom densities in highdensity Ar plasma indicated that the electron density always peaked in the plasma center, and that the neutral density near the discharge center became depleted for certain plasma conditions. 6 Typical radial distributions of Ar species obtained in the current work with Ar high-density plasmas maintained at 1400 W inductive power are shown in Fig. 9a. Without a plasma the Ar mass spectrometer species signal is uniform across the wafer. With a plasma the distribution shows a minimum at the center and increases towards the edge. When the pressure was increased from 6 to 30 mtorr, the Ar species signal intensity increased see Fig. 9a, but the relative Ar species signal intensity decreased at the center see Fig. 9b. This phenomenon can be explained by neutral atom heating due to collisions with the charged particles electron-neutral and ion-neutral. As shown in Fig. 9, the neutral Ar density exhibits a hollow-shape profile for these conditions. In our experiment, the Ar MS signal is proportional to the gas density. According to the ideal gas law pvnkt, with V and k constant, where n is the particle density and T is the gas temperature. When the gas temperature T is constant without plasma, the particle density n is proportional to pressure p. Our experimental results verified this expectation. If both pressure and gas temperature are independently varying, then the particle density change, dn should satisfy the following relation: 3 J. Vac. Sci. Technol. A, Vol. 21, No. 6, NovÕDec 2003

7 1977 Li et al.: Spatially resolved mass spectrometric sampling 1977 FIG. 10. Gas heating in Ar vs radial position. The edge of the sample at 6 mtorr process pressure was 420 K as measured with a thermal couple. dn n n dp p T dt V kt TC dp pv 1 k T dt. 4 2 pc Using this relationship, we can interpret Fig. 9 in terms of a temperature variation. Figure 10 shows the neutral temperature distribution in the plasma bulk calculated using the MS data of Fig. 9. For the 6 mtorr experiment we find that the gas temperature is flat. The wall temperature of the liner in thermal equilibrium with the gas was measured using a thermocouple and found to be 420 K in this case. The neutral temperature increases with process pressure from 6 to 30 mtorr. The gas temperature is highest in the center of the discharge. The maximum gas temperature is reached at 30 mtorr and corresponds to a temperature increase of about 450 K above room temperature. The temperature difference between the center of the sample and the edge also increases with process pressure. A comparison of the present results with data obtained by Hebner et al. 5 for Ar ICP discharges employing LIF measurements shows consistency between the two sets of data. IV. CONCLUSIONS We modified a mass spectrometer system to enable spatially resolved sampling of Ar, C 2 F 6, and C 4 F 8 /Ar highdensity discharges. From the variation of the signal with position of the sampling orifice, certain conclusions on the discharge can be made. For instance, the radial variation of the Ar signal for Ar plasmas can be interpreted by gas heating, with the maximum gas temperature at the center of the discharge. The gas temperature that is extracted from these measurements is consistent with data obtained by LIF methods. For fluorocarbon discharges we show that plasma induced heating of the sampling tube leads to a time dependence of the detected signal. This is avoided by employing a preheated sampling tube. We also employed a low electron energy of 16 ev for ionization to maximize the contribution of direct ionization of discharge produced radicals to the detected signal. A comparison of the spatial variation of CF 2 and CF 3 signals shows distinct differences, which provide information on the generation and loss processes of these species. We also showed a comparison of the kinetics of etch product detection for a mass spectrometer mounted in the exhaust line versus mass spectrometric sampling of etch products at the wall of the reactor, and etch product detection employing the sampling tube with an orifice in close proximity to the etching surface. As expected, the signal obtained with an orifice in a movable sampling tube in close proximity to the etching surface shows the most distinct transitions as a function of time and the shortest response times. ACKNOWLEDGMENT The authors would like to acknowledge Inficon Corporation for financial support of this work. 1 J. Hopwood, Plasma Sources Sci. Technol. 1, J. Hopwood, C. R. Guarnieri, S. J. Whitehair, and J. J. Cuomo, J. Vac. Sci. Technol. A 11, J. H. Keller, J. C. Forster, and M. S. Barnes, J. Vac. Sci. Technol. A 11, M. Schaepkens and G. S. Oehrlein, Appl. Phys. Lett. 72, G. A. Hebner, J. Appl. Phys. 80, T. Hori, M. D. Bowden, K. Uchino, K. Muraoka, and M. Maeda, J. Vac. Sci. Technol. A 14, C. Suzuki, K. Sasaki, and K. Kadota, J. Appl. Phys. 82, C. Suzuki, K. Sasaki, and K. Kadota, J. Vac. Sci. Technol. A 16, Y. Hikosaka, H. Toyoda, and H. Sugai, Jpn. J. Appl. Phys., Part 2 32, L B. E. E. Kastenmeier, P. J. Matsuo, G. S. Oehrlein, R. E. Ellefson, and L. C. Frees, J. Vac. Sci. Technol. A 19, T. E. F. M. Standaert, M. Schaepkens, N. R. Rueger, P. G. M. Sebel, G. S. Oehrlein, and J. M. Cook, J. Vac. Sci. Technol. A 16, X. Li, M. Schaepkens, G. S. Oehrlein, R. E. Ellefson, L. C. Frees, N. Mueller, and N. Korner, J. Vac. Sci. Technol. A 17, M. Schaepkens, R. C. M. Bosch, T. E. F. M. Standaert, and G. S. Oehrlein, J. Vac. Sci. Technol. A 16, Y. Chinzei, T. Ichiki, R. Kurosaki, J. Kikuchi, N. Ikegami, T. Fukazawa, H. Shindo, and Y. Horiike, Jpn. J. Appl. Phys., Part 1 35, Y. Hikosaka, M. Nakamura, and H. Sugai, Jpn. J. Appl. Phys., Part 1 33, JVST A - Vacuum, Surfaces, and Films