Reducing PFC Emissions Using C3F8-based PECVD Clean
Cleaning PECVD chambers with perfluoropropane (C3F8) instead of hexafluoroethane (C2F6) reduces PFC emissions by 64% and decreases consumable costs by 47%.
Staff -- Semiconductor International, 2/1/1998
 |
At a Glance | | |
|
| ||
| |||
While C3F8 is also a PFC, previous work showed that its use for PECVD chamber cleaning in place of C2F6 will reduce chemical usage as well as emissions of PFCs, based on greater utilization of C3F8 gas.1 Further, it was indicated that a faster etch rate with C3F8 could be achieved, reducing chamber cleaning time and increasing equipment uptime. Thus, until suitable non-PFC chemical alternatives can be identified, a switch of existing C2F6 usage to C3F8 meets demands for re-ducing PFC usage and emissions, with no re-quirement for hardware changes. C3F8 use might also be extended to other semiconductor processes using C2F6.
Project goals and procedures
The project optimized and qualified the plasma clean process based on C3F8/O2 chemistry, while estimating gas consumption savings and measuring PFC emissions. Relative cost-of-ownership (COO) differences were determined, and the impact of the new cleaning process on device yield, TEOS film thickness film stress and particle performance was evaluated.
The project, conducted between June and December of 1996, was performed in six phases. In the first phase, a screening experiment determined the possible operating window for the C3F8/O2 chemistry. A full-factorial design of experiment (DOE) in the second phase optimized the process parameters and results. The third phase focused on measuring process emissions during C3F8 and C2F6 chamber cleans.
This was followed by a manufacturing capability study involving an extended wafer run over a period of six weeks to check for TEOS process repeatability, plasma stability and other hardware impacts resulting from the use of C3F8 clean process. Using product device lots, the fifth phase provided process stability verification and qualification of the process for production. Finally, project results were documented in a comprehensive final report.
PFC reduction programs
The National Technology Roadmap for Semiconductors, first published in 1994, calls for proactive reduction of emissions that may cause a global climate change. PFCs are strong infrared radiation absorbers and long-lived in the atmosphere, making them greenhouse gases with high global warming potentials. Only a fraction of these gases, on the average, are consumed in PECVD cleaning processes; the rest are often emitted to the atmosphere.
Semiconductor firms participate in voluntary PFC emissions reduction programs that allow participants the freedom to design individualized emissions reduction strategies for CF4, C2F6, SF6, NF3 and C3F8 all of which are widely used in etch and PECVD chamber cleans. The SIA and the EPA developed the PFC Emission Reduction Partnership for the Semiconductor Industry, with program details described in a Memorandum of Understanding (MOU) that was sent to all companies with U.S.-based manufacturing operations. These programs are based on the Climate Change Action Plan, a document released by President Clinton in 1993, outlining the U.S. program for managing and limiting greenhouse gas emissions. This plan developed as a result of the Framework Convention on Climate Change, an international agreement signed by representatives from the United States and 160 other nations at the Earth Summit in 1992.
High-pressure plasma clean optimization
The purpose of the C3F8 experiment was to develop a C3F8-based plasma etch process that minimized C3F8 and O2 gas consumption, while not increasing the etch time when compared to the existing C2F6 based process.
The PECVD Concept Two Sequel tool, configured in the Dual Frequency Capacitance Ground mode, deposited a known amount of TEOS oxide (10 min fixed time deposition). The time required for the plasma to remove the TEOS oxide from the heaterblock was determined. An optical emission endpoint detector operating at 704 nm was used. All depositions occurred with the heaterblock set at 400°C.
A full factorial central composite response surface design was used to model the effects the etchant gases and the pressure had on the etch time. C3F8 flow rate was varied between 800-1200 sccm, O2 flow was varied between 1200-1600 sccm and pressure range was 3.4-3.8 Torr. The etch power was fixed for all the experiments. Results over 16 trials showed that etch time ranged from 2.75 to 3.48 min. Using the same procedure, the existing C2F6 process has an etch time of 3.20 min.
The etch time was modeled (R2=0.94) using Design-Expert v5 software. The pressure and interaction between the C3F8 and O2 had the largest effects on the plasma clean etch time. A perturbation plot of the main effects (Fig. 1) shows that the pressure (line C) caused the largest change in the etch time. One can also observe that the C3F8 flow (line A) has a significant quadratic effect on the etch time.
Fig. 1. Pressure had the largest effect on plasma etch time.
Interestingly, when holding the O2 flow at 1600 sccm, the etch time decreases with increasing the C3F8 flow from 800 to 1200 sccm (Fig. 2). However, if the O2 is held at 1200 sccm, the plasma clean etch time increases as the C3F8 flow is increased from 800 to 1200 sccm. This interaction can be explained. At higher C3F8 ratios (i.e., lower O2 flows), the efficiency of the plasma etching is reduced because of the competing reaction of depositing amorphous fluorinated carbon. Therefore, a low C3F8:O2 ratio (<1) is required for an optimized etch process.
Fig. 2. A low C3F8:O2 ratio is required for an optimized etch process.
In the center of the C3F8/O2 contour plot (Fig. 3), there is an optimum area where the gas consumption is minimized without increasing the etch time (when compared to the 3.20 min etch time for the existing C2F6 etch process).
Fig. 3. In the center, gas consumption is minimized without increasing etch time (pressure = 3.6 Torr).
Measuring emissions during cleans
FTIR spectroscopy was used to monitor process emissions and to compare etch gas utilization efficiencies and net PFC emissions. Results showed that the lower flow rate and higher utilization of etch gas during the C3F8 process contributes to a significant reduction in net emissions.
The equipment, materials and procedures that make up the protocol for FTIR measurements of fluorinated compounds in semiconductor process tool exhaust will be discussed in a future publication. Using this FTIR protocol, process emissions containing multicomponent gas mixtures of the MOU listed compounds (CHF3, CF4, C2F6, C3F8, SF6 and NF3) can be simultaneously identified, quantified and displayed on site, during testing of new CVD chamber cleaning processes.
Comparisons of C2F6 and C3F8 emissions during production runs are based on the mass of effluent emitted during cleaning. Mass balance verification compares the mass of fluorine going into the reactor as C2F6 or C3F8 (Fin) with the mass of fluorine exhausted as CF4, C2F6, C3F8, SiF4 and COF2 (Fout). Throughout this test, the mass balance was typically >90%, a reasonable average as the effluent profiles did not account for HF and F2 emissions.
Etch gas requirements
The alternative C3F8/O2 process substantially decreased etchant gas consumption from 2500 sccm to 1000 sccm, 60% less than the gas flow used in the existing C2F6 process. Considering the molecular weights of each gas, the C3F8 process flows 45% less mass of etch gas than the standard C2F6 process. The O2 flow for the optimized C3F8 etch process was 1400 sccm, 30% less than the O2 flow of 2000 sccm for the C2F6 process. Etch time for the TEOS oxide used in the optimized C3F8 process was slightly shorter than the C2F6 process.
Cleaning gas utilization
The utilization of the etch gas refers to the percentage of the etch gas that reacts during the plasma clean. Utilization was determined by measuring the etch gas concentration with the RF power off and on, as shown in the equation:
Etchant gas utilization doubles from 30-35% to 68-71% when cleaning TEOS films using C2F6 and C3F8, respectively. For the Novellus oxide, utilization increased from 24-28% to 65-70%. Gas utilization during a high-pressure clean (i.e., no film) increased from 33% to 63%. The utilization determined by the FTIR method in this study confirmed the improved utilization of C3F8 compared to C2F6.1,2
PFC emissions and COO
The primary PFC emissions from the PECVD tool exhaust for both cleaning processes are CF4 and unreacted C2F6 or C3F8. Concentrations of the emissions can change significantly during the chamber clean. Figure 4 shows the process emissions during C3F8 high-pressure cleaning, after deposition of TEOS film. The SiF4 concentration changes the most with time and serves as a good endpoint indicator.
Fig. 4. Process emissions during C3F8 high-pressure clean of TEOS film change the most for SiF4 and CF4.
PFC emissions were quantified on the basis of mass in grams and million metric tons of carbon equivalence (MMTCE), as shown in Table 1. The chamber cleans followed deposition of 24 TEOS wafers. The mass of the PFCs emitted is calculated based upon the concentration of the effluent species in the exhaust, the exhaust flow rate and the duration of the chamber clean.
Table 1. PFC Emissions in Mass and the Million Metric Tons of Carbon Equivalence (MMTCE)
|
PFCs from C2F6 clean |
PFCs from C3F8 clean |
|||
|
PFC (global warming potential) |
Mass (g) |
MMTCE |
Mass (g) |
MMTCE |
|
CF4 (6500) |
17.6 |
3.11 x 10-8 |
21.8 |
3.87 x 10-8 |
|
C2F6 (9200) |
92.4 |
23.2 x 10-8 |
0 |
0 |
|
C3F8 (7000) |
0 |
0 |
29.9 |
5.71 x 10-8 |
|
Total |
110 |
26.3 x 10-8 |
51.7 |
9.58 x 10-8 |
|
Net emission reduction (%) |
-53 |
-64 |
||
PFC gases are greenhouse gases, as is CO2. Thus, emissions from the use of C2F6 and C3F8 for chamber cleanings can be converted to a CO2 equivalence (mass of CO2 causing the same cumulative warming). This calculation considers the mass and global warming potential (GWP) of each PFC emission. GWP is calculated by integrating radiative forcing, a measure of the ability of a greenhouse gas to trap infrared radiation, as the gas decays over time. By dividing this integral by an identical integral for an equivalent mass of a reference gas (generally CO2), GWP is determined.3 The GWP values used in this report CF4 = 6500, C2F6 = 9200 and C3F8 = 7000 are for a 100-year time horizon.4 The net PFC emission reduction is 64% using C3F8 instead of C2F6 for cleaning the AMD TEOS film.
A COO study investigated the use of these C3F8 gases at a total of 34 process steps in Fab 25. Results indicate that replacing the C2F6 gas with C3F8 gas will lower the consumable-related costs of ownership by 47%. [note: COO numbers do not consider tool-related purchase or installation costs.]
Process repeatability
Using a single Sequel chamber, an extended wafer run over a period of six weeks and involving 1300 test wafers was conducted to check TEOS process repeatability, plasma stability and other hardware impacts resulting from the new process. The frequency and activities of all maintenance operations for the tool chamber were also evaluated.
Film thickness and uniformity were measured using a Prometrix UV-1050 tool for each of the initial 200 wafers. The thickness range was ±300 Å below 1.5% for a target thickness of 21,300 Å. Stress level, measured using a Tencor Flexus tool, maintained an average of -1.4 x 109 dynes/cm2 (compressive) with a small variation of ±2 x 108 dynes/cm2. Clean prime wafers were used for both stress and particle measurements. Particle performance proved acceptable as the highest particle level was ~40 (>0.25 µm) during the entire period.
A maintenance technician performed routine weekly and monthly preventive maintenance activities during the period. The chamber was thoroughly checked to see if any degradation of heaterblock, showerhead or O-rings had occurred. No degradation in these was found.
Device split-lots
Three advanced logic device lots were used to qualify the use of C3F8. The lots were split two ways at each ILD TEOS step. Wafers 1-12 (Split A) were processed using C3F8 plasma clean gas, while wafers 13-24 (Split B) were processed using C2F6. The wafer electrical test (WET), SORT yield and wafer level reliability (WLR) tests were performed for the C3F8 process stability investigation. Five wafers per split received WLR VT-Fluence testing.
VT-Fluence tests did not indicate any additional Vt shift or difference in charge trapping between the C2F6 and C3F8 splits, setting aside earlier concerns of fluorine contamination with C3F8. The n-channel threshold voltage shift under its current-stress condition did not indicate any significant difference because of the cleans (Fig. 5). P-channel results were similar.
Fig. 5. Threshold voltage shift was not significant between C2F6 and C3F8 cleans.
In addition to the WLR data, several WET parameters were examined for possible differences relating to the splits. Analysis of gate-oxide capacitors and Long L transistors provided no indication of additional charge trapping between the two splits. Finally, the SORT yield data provided acceptable results with no statistically significant differences between standard C2F6 and C3F8 wafers (Fig. 6).
Fig. 6. Die yield did not change significantly between C2F6 and C3F8 cleans.
Conclusion
The CVD chamber clean using C3F8 chemistry has proved to be a viable alternative etch gas for chamber cleaning in Novellus Concept Two Sequel reactors. The results of the screening DOE indicated that a low C3F8 ratio
(<1) is required for an optimized C3F8 plasma etch process. The etch process developed with C3F8/O2 is a drop-in replacement to the existing C2F6 process. Differences between the existing C2F6 process and alternate C3F8 process include the following:
- Less gas use 60% less etch gas, 30% less oxygen and 46.7% less total gas flow;
- A 6% faster etch rate for the oxide and TEOS processes using C3F8;
- A 45% reduction in required pounds of C3F8;
- An improvement in etch efficiency from 24-35% for C2F6 to 63%-71% using C3F8;
- A 60-70% reduction in net PFC emissions;
- A 47% reduction in consumables-related COO.
The extended wafer run showed that the C3F8 clean did not impact the tool hardware for tests performed during the period. Electrical tests indicated that no process instability or adverse impact was found in WET, SORT yield and WLR results using C3F8 clean process.
Acknowledgments
The authors would like to thank Stephanie Grelle, Terry McDonough and Michele Smith of AMD for their great support and coordination of facility and environmental PFC aspects of this study. Thanks are also extended to Mike May, Jenna Latt and Kou-Yhi Young of AMD for their guidance, assistance and support. The authors also thank SEMATECH for its sponsorship of this work. The guidance, assistance and support of Paul Mahal and Doug Winandy from Novellus Systems Inc. was key to the success of this program. Many thanks to Dr. Jim Wolter and Lew Tousignant of the 3M Co. and Brian Wright of the Midac Corp. for an excellent effort in determination of C2F6 and C3F8 emissions and utilization at the AMD site. Special thanks to Lydia S. Zapata for her assistance in preparation of the manuscript.
References
1. L. Zazzera, W. Reagen and A. Cheng, J. Electrochem. Soc., vol. 144, no. 10, p. 3597, (1997).
2. J. Langan, P. Maroulis and R. Ridgeway, Solid State Technology, 39, 7, 115 (1996).
3. IPCC (Intergovernmental Panel on Climate Change) Climate Change, The Supplementary Report to the IPCC Scientific Assessment 1991, J. J. Houghton, B. A. Callander and S. K. Varney, editors, p. 205, Cambridge University Press, UK., (1992).
4. IPCC Climate Change, The Science of Climate Change 1995, J. J. Houghton, L. G. Meiro Philo, B. A. Callander, N. Harris, A. Kattenberg, and K. Maskell, editors, p. 16, Cambridge University Press, UK, (1996).
|
Phone: (512) 602-6072
Phone: (512) 602-5930
Phone: (612) 737-5462
Phone: (612) 778-6565 |
Sey-Ping Sun has a doctorate in physics and materials science and engineering. Currently, he is a member of the technical staff and project leader for multilevel interconnect dielec-trics development at Fab 25 of AMD.
David Bennett has a master's degree in chemical engineering and is currently the technology trans-fer manager for AMD. 
Larry Zazzera has a doctorate in materials chemistry and is presently a senior development engineer for the 3M Performance Chemicals and Fluids Division. 
William Reagen has a bachelor's degree in chemistry and a doctorate in inorganic chemistry. He is presently a specialist in the 3M Environmental Laboratory. Semiconductor International February 1998
