The effect of the chamber wall on fluorocarbonassisted atomic layer etching of SiO 2 using cyclic Ar/C 4 F 8 plasma

Transcription

1 The effect of the chamber wall on fluorocarbonassisted atomic layer etching of SiO 2 using cyclic Ar/C 4 F 8 plasma Running title: The effect of the chamber wall on FC assisted atomic layer etching of SiO 2 using cyclic Ar/C 4 F 8 plasma Running Authors: Masatoshi et al. Masatoshi Kawakami a) Electronic Device Systems Business Group, Hitachi High-Technologies Corporation, 794 Higashitoyoi, Kudamatsu, Yamaguchi , Japan Dominik Metzler a) Department of Material Science and Engineering, Institute for Research in Electronics and Applied Physics, University of Maryland, College Park, MD 20742, USA Chen Li a) Department of Physics, Institute for Research in Electronics and Applied Physics, University of Maryland, College Park, MD 20742, USA Gottlieb S. Oehrlein Department of Material Science and Engineering, Institute for Research in Electronics and Applied Physics, University of Maryland, College Park, MD 20742, USA a), b) a) American Vacuum Society member. (Please identify all AVS member authors.) b) Electronic mail: oehrlein@umd.edu 1

2 We studied the effect of the temperature and chemical state of the chamber wall on process performance for atomic layer etching of SiO 2 using a steady-state Ar plasma, periodic injection of a defined number of C 4 F 8 molecules, and synchronized plasmabased Ar + ion bombardment. To evaluate these effects, we measured the quartz coupling window temperature. The plasma gas phase chemistry was characterized using optical emission spectroscopy. It was found that although the thickness of the polymer film deposited in each cycle is constant, the etching behavior changed, which is likely related to a change in the plasma gas phase chemistry. We found that the main gas phase changes occur after C 4 F 8 injection. The C 4 F 8 and the quartz window react and generate SiF and CO. The emission intensity changes with wall surface state and temperature. Therefore, changes in the plasma gas species generation can lead to a shift in etching performance during processing. During initial cycles, minimal etching is observed, while etching gradually increases with cycle number. I. INTRODUCTION Following the trajectory of Moore s Law, semiconductor manufacturing is increasingly demanding atomic-scale process control to further decrease critical 2

3 dimensions. 1 High precision etching and material selectivity are indispensable. 2,3 In addition, device structures have become more complex, adding to the challenge of shrinking dimensions. 4 6 Atomic layer deposition was developed as early as 1977 by Suntola and Antson 7 and has been established as a common tool in device patterning However, the development of a corresponding atomic layer etching (ALE) method is still only beginning to enter production. 12 One of the key challenges for ALE is to overcome the long processing time that many approaches require, leading to an overall low wafer throughput. 13 However, ALE is expected to lower substrate damage and increase pattern fidelity. 14,15 Low-pressure, high-density fluorocarbon plasma processes, however, have been found to suffer from process drifts, as seen by a time dependent behavior of the etch rate. 17 At low processing pressures, plasma wall interactions are of significant importance in determining the discharge chemistry. Process drifts have been attributed to changes in the reactor wall conditions In previous work, although the thickness of the polymer film deposited in each cycle was constant, the etching behavior showed changes. Minimum etching was observed during initial cycles and the etching increased with cycle number, possibly related to changes in chamber wall interactions. 16 Chamber wall interactions may be crucial to the stability of such ALE processes. 3

4 In this article, the effect of the chamber wall conditions on the stability of ALE process performance is reported. Etching characteristics, wall surface condition dependence, wall temperature dependence, and plasma gas phase chemistry mechanisms are studied. Of special interest is the etching behavior during initial cycles. II. EXPERIMENTAL We used an inductively coupled plasma system excited at MHz. The plasma was confined within a 195-mm-diameter anodized Al confinement ring. A 125-mmdiameter Si substrate is located 150 mm below the top electrode on an electrostatic chuck and can be biased at 3.7 MHz. The base pressure achieved before processing was in the Torr range and the temperature of the samples (25 25 mm 2 ) was stabilized by substrate cooling (at 10 C) during plasma processing. The details of the plasma system have been described previously The materials used were 100 nm SiO 2 on Si substrate and were studied using in-situ ellipsometry 23 in real time. The plasma gas-phase chemistry was characterized using optical emission spectroscopy. The temperature of the quartz coupling window was measured with an infrared temperature sensor using a detection wavelength from 8 to 14 μm. Figure 1 shows the schematic of the experimental sequence for evaluating the chamber wall surface state. All plasma processes presented 4

5 here are at a source power of 200 W and a processing pressure of 10 mtorr with an Ar flow of 50 sccm. At the beginning of each cycle, a pulse of C 4 F 8 was injected for 1.6 s into continuous Ar plasma, and about 5 Å of fluorocarbon film is deposited. Twelve seconds after the C 4 F 8 pulse injection, a synchronized RF bias potential is applied to the substrate for 40 s to increase Ar + ion bombardment energies. A RF self-bias potential of 10 V creates maximum ion energies of about 25 ev. An empty pulse (no fluorocarbon gas injection) as a reference at every 4 th cycle is used to obtain information on the chamber contribution to deposition and etching at that time. To evaluate the effect of wall surface state and the reactor temperature, four different experimental conditions were used: the clean condition, with film condition, the cold condition, and the hot condition. The clean condition refers to a previously cleaned chamber by O 2 plasma. The with film condition refers to the walls covered with polymer film after one ALE process. The cold condition refers to the chamber being at room temperature at the start of the experiment. The hot condition refers to the chamber being heated to the saturated maximum temperature during processing. The conditions combining the wall condition and temperature condition for the experiment are as follows: (a) cold/clean; (b) cold/ with film ; (c) hot/clean and (d) hot/ with film. 5

6 FIG. 1. (Color online) Schematic of the experimental sequence for impact evaluation of the chamber condition. One ALE cycle includes a 12 s unbiased deposition step and a 40 s 25 ev ion bombardment etching step. A 1.6 s C 4 F 8 injection step was at the beginning of the deposition step for each injection cycle. After every three injections, one cycle without injecting C 4 F 8 was run to evaluate the chamber condition. III. RESULTS AND DISCUSSION The time dependence of the temperature of the quartz coupling window was measured by an infrared temperature sensor. The resulting data for the four conditions are plotted in Fig. 2. The temperatures for the cold/clean condition and the cold/ with film condition have a similar trend. The temperatures for the hot/clean condition and the hot/ with film condition were stable at 40 C during processing. The quartz coupling window temperature increased with increasing processing time for both cold conditions. However, the quartz coupling window temperature was constant after prior plasma based 6

7 heating. This is because the quartz coupling window was preheated up to the saturated temperature. FIG. 2 (Color online) Quartz temperature during processing for the four tested conditions. Figure 3 shows an example of measured CO (297 nm) and SiF (440 nm) optical emission intensities along with deposited FC film and etched SiO 2 film thickness for a cold/clean experiment. CO and SiF emission peaks were defined as the maximum intensity increase in one cycle. The deposition and etching thickness are defined as the thickness difference between the beginning and the end of a deposition step and between the beginning of a deposition step and the end of an etch step in each cycle. SiF and CO emission peaks appear immediately after C 4 F 8 injection. However, these peaks do not appear when no precursor is injected, indicating that after precursor injection, fluorocarbon reacts with the quartz window and generates SiF and CO. Additionally, the 7

8 impact of the chamber wall (and the quartz window) state could be evaluated in the no gas injection cycle. The deposited thickness without precursor injection shows redeposition of material (fluorocarbon) sputtered from the chamber wall. FIG.3. (Color online) An example of the CO and SiF optical emission intensity and thickness in the cold/clean condition. CO, SiF emission peaks are defined as the maximum intensity increase in one cycle. The deposition and etching thickness are defined as the thickness difference between the beginning and the end of a deposition and between the end of an etch step and the beginning of a deposition step in each cycle. 8

9 The bar chart in Fig. 4 shows the results of deposition and etching experiments for the four tested conditions. The thicknesses for cycles with and without injection are shown separately in the upper and lower part of each figure. Dotted lines highlight the initial cycles for each condition. FIG.4. (Color online) Deposition and etching thickness of four different conditions: (a) cold/clean condition; (b) cold/ with film condition; (c) hot/clean condition; and (d) hot/ with film condition. The thicknesses in the cycles with and without injection are shown separately in the upper and lower part of each figure. 9

10 Comparing the initial and the later cycles in the cold/clean condition [Fig. 4 (a)], initial cycles show minimum etching thickness in the injection cycles. A lower deposition than in the later cycles is seen in the no gas injection cycles. The hot/clean condition shows the same trend [Fig. 4 (c)].when comparing the initial cycles of the cold/clean condition and of the cold/ with film condition [Fig. 4 (a) vs. (b)], the cold/clean condition has a lower etching thickness than the cold/ with film condition in the injection cycles. Comparing initial no gas injection cycles, clean conditions show less deposition than with film condition for cold temperatures. The initial cycles of the hot/clean condition compared with the hot/ with film condition show the same trend [Fig. 4 (c) vs. (d)]. The amount of FC deposited during the no gas injection cycles provides a measure of the contribution of FC film coating on the chamber walls. In Fig. 4, the arrows indicate the change in deposition thickness during processing for cycles with no gas injection in each condition. The deposited FC film thickness increases with increasing number of cycles for the clean condition; however, there is no or minimal increase for the with film condition. The bar chart in Fig. 5 shows the results of SiF and CO optical emission peak intensities normalized to Ar (419nm) in gas injection cycles for the four tested conditions in order to confirm a relationship between wall condition and gas reaction. The dotted lines highlight the initial cycles for each condition 10

11 and arrow marks show the evolution of the emission peak during processing for each condition. The initial cycles of both clean conditions show higher SiF and CO emission intensity than the initial cycles of both with film conditions. During processing under the clean conditions, the emission peaks of SiF and CO decrease. However, for with film conditions there was no or minimal decrease in emission peak intensities. FIG.5. (Color online) SiF and CO optical emission peak intensities normalized to Ar with injection cycles for four conditions: (a) cold/clean condition; (b) cold/ with film condition; (c) hot/clean condition; and (d) hot/ with film condition. These trends are indicated with arrows in Fig. 5. This result is consistent with the assumption that the deposited FC film thickness measured for the no injection cycles is 11

12 an indication of FC wall coverage. For the clean wall condition, C 4 F 8 reacts with the quartz window and generates SiF and CO after injection. When the surface of the quartz coupling window is covered by a fluorocarbon film, the overall amounts of CO and SiF decreased, which enables greater substrate etching. Therefore, minimum etching occurs at the beginning of the clean condition. We conclude that fluorocarbon film on the quartz window can reduce CO and SiF generation and the fluorocarbon accumulation in the chamber leads to a shift in etching. We confirmed that CO and SiF have the same trend, which shows a quartz window and injection gas reaction. With a clean condition, minimal etching is observed because of a lack of film on the quartz window. When comparing the cold/clean condition and the hot/clean condition of the initial cycles, the hot/clean condition results in a lower SiO 2 etching thickness than the cold/clean condition during the injection cycles (see Fig. 4). When comparing the cold/clean condition and the hot/clean condition of the initial cycles, the hot/clean condition has a higher normalized intensity of SiF and CO than the clean/cold condition in the injection cycles as shown in Fig. 5. The etching during injection cycles and the emission intensities have an opposite trend. A higher quartz temperature accelerates the precursor-window reaction and generates a higher amount of SiF and CO. There is less SiO 2 etching at the beginning of the hot/clean condition than in the cold/clean condition. 12

13 A hot quartz window can increase CO and SiF generation and less etching is initially observed because of a greater reaction at the quartz window. Table 1 shows a summary of Fig. 4. Table 1: Summary of Fig. 4. Cycle (a) cold/clean (b) cold/"with film" minimal etching is observed Injection etching is observed in initial cycle in initial cycle increase in FC deposition no increase in FC deposition from No Injection from chamber wall chamber wall Cycle (c) hot/clean (d) hot/"with film" Injection No Injection less etching is observed in initial cycle increase in FC deposition from chamber wall etching is observed in initial cycle minimal increase in FC deposition from chamber wall From the results of this study we conclude that one key requirement for achieving stable ALE processes in a plasma reactor is the need for more precise chamber wall chemical state and chamber temperature stability. Ideally, chamber wall state (FC composition and/or coverage) and temperature need to be same at the beginning of each ALE cycle during one or multiple ALE processes in order to achieve stable and reproducible etching control. 13

14 IV. SUMMARY AND CONCLUSIONS We studied the effect of the quartz coupling window temperature and chamber wall chemical state on process performance for atomic layer etching of SiO 2 using a steady-state Ar plasma, periodic injection of a defined number of C 4 F 8 molecules, and synchronized plasma-based Ar+ ion bombardment. We found that C 4 F 8 and the quartz window react and generate SiF and CO emission after C 4 F 8 injection. The emission intensity changes with the wall surface state and temperature. This leads to a shift in etching performance during processing. The fluorocarbon film on the quartz window can reduce CO and SiF generation and the fluorocarbon accumulation in the chamber leads to a shift in etching. With a clean condition, there was minimal etching because there was a lack of film in the quartz window. A hot quartz window can increase CO and SiF generation and causes less etching during initial process cycles because of a greater quartz window reaction. In this study, we focused on etching behavior during initial cycles. To decide optimum conditions, further investigation is needed. In future work, we will investigate chamber wall effect in each later cycle. The process chamber material and plasma chemistry investigated is commonly used in conventional etch processes, however, we will evaluate other materials and chemistries in future work, for instance a ceramic window alternatively to a quartz window, and CHF 3 alternatively to C 4 F 8. 14

15 ACKNOWLEDGMENTS The authors gratefully acknowledge the financial support of this work by the National Science Foundation (CBET ), the U.S. Department of Energy (DE-SC ), and Hitachi High-Technologies Corporation. The authors also thank Dr. E. Bartis, A. Knoll, P. Luan, and A. Pranda for contributions to this project. 1 G. E. Moore, Proc. IEEE 86, 82 (1998). 2 C. G. N. Lee, K. J. Kanarik, and R. A. Gottscho, J. Phys. D Appl. Phys.47, (2014). 3 V. M. Donnelly and A. Kornblit, J. Vac. Sci. Technol. A 31, (2013) 4 K. J. Kanarik, G. Kamarthy, and R. A. Gottscho, Solid State Technol. 55, 15 (2012). 5 M. Guillorn et al., International Electronic Devices Meeting (2009), pp S. Bangsaruntip et al., International Electronic Devices Meeting (2009),pp T. Suntola and J. Antson, U.S. patent 4,058,430 A (14 November 1977). 8 S. M. George, Chem. Rev. 110, 111 (2010). 9.R. L. Puurunen, J. Appl. Phys. 97, (2005). 10 M. Leskelä and M. Ritala, Angew. Chem. Int. Ed. 42, 5548 (2003). 15

16 11 M. Leskelä and M. Ritala, Thin Solid Films 409, 138 (2002). 12 K. J. Kanarik, S. Tan, J. Holland, A. Eppler, V. Vahedi, J. Marks, and R.A. Gottscho, Solid State Technol. 56, 14 (2013). 13 A. Agarwal and M. J. Kushner, J. Vac. Sci. Technol. A 27, 37 (2009). 14 G. S. Oehrlein, D. Metzler, and C. Li, ECS J. Solid State Sci. Technol. 4,N5041 (2015). 15 K. J. Kanarik, T. Lill, E. A. Hudson, S. Sriraman, S. Tan, J. Marks, V. Vahedi, and R. A. Gottscho, J. Vac. Sci. Technol. A 33, (2015). 16 D. Metzler, R. Bruce, S. Engelmann, E. A. Joseph, and G. S. Oehrlein, J. Vac. Sci. Technol. A 32, (2014). 17 M. F. Doemling, N. R. Rueger, J. M. Cook, and G. S. Oehrlein, J. Vac.Sci. Technol. B16, 1998 (1998). 18 S. C. McNevin, K. V. Guinn, and J. Ashley Taylor, J. Vac. Sci. Technol.B 15, 214 (1997). 19 J. A. O Neill and J. Singh, J. Appl. Phys. 77, 497 (1995). 20 M. Schaepkens, R. C. M. Bosch, T. E. F. M. Standaert, G. S. Oehrlein and J. M. Cook J. Vac.Sci. Technol. A16, 2099 (1998). 21 S. Engelmann et al., J. Vac. Sci. Technol. B 25, 1353 (2007). 16

17 22 T. E. F. M. Standaert, P. J. Matsuo, S. D. Allen, G. S. Oehrlein, and T. J.Dalton, J. Vac. Sci. Technol. A 17, 741 (1999). 23 H. G. Tompkins, A User s Guide To Ellipsometry (Dover, Mineola, 1993). 17

18 Table: Table 1: Summary of Fig4. Cycle (a) cold/clean (b) cold/"with film" Injection No Injection minimal etching is observed in initial cycle increase in FC deposition from chamber wall etching is observed in initial cycle no increase in FC deposition from chamber wall Cycle (c) hot/clean (d) hot/"with film" Injection No Injection less etching is observed in initial cycle increase in FC deposition from chamber wall etching is observed in initial cycle minimal increase in FC deposition from chamber wall 18

19 Figure captions FIG. 1. (Color online) Schematic of the experimental sequence for impact evaluation of the chamber condition. One ALE cycle includes a 12 s unbiased deposition step and a 40 s 25 ev ion bombardment etching step. A 1.6 s C 4 F 8 injection step was at the beginning of the deposition step for each injection cycle. After every three injections, one cycle without injecting C 4 F 8 was run to evaluate the chamber condition. FIG. 2 (Color online) Quartz temperature during processing for the four tested conditions. FIG.3. (Color online) An example of the CO and SiF optical emission intensity and thickness in the cold/clean condition. CO, SiF emission peaks are defined as the maximum intensity increase in one cycle. The deposition and etching thickness are defined as the thickness difference between the beginning and the end of a deposition and between the end of an etch step and the beginning of a deposition step in each cycle. FIG.4. (Color online) Deposition and etching thickness of four different conditions: (a) cold/clean condition; (b) cold/ with film condition; (c) hot/clean condition; and (d) hot/ with film condition. The thicknesses in the cycles with and without injection are shown separately in the upper and lower part of each figure. FIG.5. (Color online) SiF and CO optical emission peak intensities normalized to Ar with injection cycles for four conditions: (a) cold/clean condition; (b) cold/ with film condition; (c) hot/clean condition; and (d) hot/ with film condition. 19