Reducing cost and PFC effluent with optimized C2F6 chamber cleaning
03/01/2007
Two chamber cleaning gases, CF4 and C2F6, dominated the cleaning gas market for old-generation CVD tools. While these gases are still widely used, others, such as C3F8, c-C4F8, and NF3 [1], have been introduced to meet updated process specifications and manufacturing efficiencies. Among these, remote plasma NF3, a low-emission chamber-cleaning gas, has emerged as one of the most environmentally friendly and popular; however, it is not suitable for old-generation CVD tools because of expensive hardware retrofit costs or technical limitations [2]. This presents a challenge for manufacturers needing to simultaneously minimize the impact of PFC emissions on the environment and improve manufacturing costs [3].
In this study, we demonstrate a process for converting inefficient CF4 cleaning to C2F6 for various SiH4-based CVD processes. Using systematic design of experiment (DOE) methodology, the optimized C2F6 cleaning demonstrates superior performance in terms of reduced PFC emissions, as well as higher gas utilization, chamber-cleaning efficiency, and throughput. After a pilot run of 120 production lots-for both SiH4-based silicon dioxide as well as silicon nitride processes-our findings confirm that the C2F6 cleaning recipe is an environmentally preferable and cost-effective replacement for CF4.
Experiment
The experiment was conducted in an AMAT P-5000 CVD tool using the standard CF4 cleaning recipe as a baseline reference. The same deposition recipe was used to keep film composition and thickness constant while varying chamber pressure, cleaning gas flow, and percent N2O in the C2F6 cleaning recipe. To ensure no residual film remained after each experimental run, a time-based over-clean was applied after each different DOE cleaning.
The exhaust gases emitted during the cleaning processes were monitored using Fourier transform infrared spectroscopy (FTIR). The exhaust gas was extracted via a sampling pump from a sample port downstream of the process pump and directed through a MIDAC FTIR analyzer to quantify the species of interest (CxFy, SiF4, COF2, and CF4). The sample was then returned to the exhaust line downstream of the sample port.
The chamber cleaning endpoint was monitored and verified using an optical emission spectoscopy (OES) tool mounted on the chamber. During the production tests, film properties (thickness, refraction index, stress), product electrical performance, and yield performance, as well as tool and parts appearance, were evaluated.
 Figure 1. The trend charts for PFC emissions versus pressure and C2F6 flow for a) silicon dioxide chamber cleaning and b) silicon nitride chamber cleaning. |
The chamber cleaning time, gas usage, cleaning efficiency, and PFC emissions were calculated from FTIR and OES measurement results for each experimental condition. The gas usage was calculated based on the cleaning time from FTIR and the mass-flow controller (MFC) flow setting of each clean condition. The amount of SiF4 emitted was integrated to evaluate the cleaning efficiency. Using the following equation, the researchers were able to integrate the total volumetric output of C2F6 and CF4 during those cleanings to calculate PFC emissions in the form of kilogram carbon equivalents (kg CE):
In this equation, for every species that contributes to global warming, Q is the amount of gas (kg), and GWP is the global warming potential in terms of a 100-year integrated time horizon (ITH). The 100-year ITH values used for C2F6 and CF4 are 9200 and 8700, respectively.
Results
Figure 1a shows the DOE result for PFC emissions in cleaning silicon dioxide film using C2F6 from the MINITAB calculation through FTIR data loading. The lowest gas usage and PFC emission rates occurred at 62.5% of the N2O flow ratio. The PFC emissions rates were reduced with a lower C2F6 flow rate. No obvious pressure impact was found in this DOE design range.
 |
The cleaning efficiency-defined as the ratio of SiF4 emission using C2F6 versus CF4-is around 0.9-1.1 for silicon dioxide (Table 1). After a first set of 15 runs, additional fine-tuning of process conditions resulted in cleaning efficiency that is consistently above 1.2, as shown in the rows labeled A1-A3 in Table 1. Maintaining efficiency >1 for both silicon oxide and silicon nitride films indicates sufficient chamber cleaning capability for the intentionally narrowed range under which this CVD tool must operate.
 |
Table 2 shows gas usage, PFC emissions, and cleaning time for silicon dioxide films cleaned using the optimized A3 condition. The result is an 82% reduction in gas usage, an 80% reduction in PFC emissions, and around a 30% reduction in cleaning time.
Similar methodology was employed for a silicon nitride film cleaning optimization based on the DOE results detailed in Fig. 1b. We saw significant improvement in the overall results, including an 81% reduction in gas usage, a 78% reduction in PFC emissions, and around a 20% reduction in cleaning time (Table 2).
A marathon run for both silicon oxide and silicon nitride cleanings was carried out based on the optimized condition labeled A3 from Table 1. A total of 475 test wafers were processed in one month. Film thickness, refraction index, stress, and chamber particle count were measured at 25-wafer intervals. All these critical process parameters were well within expected levels during this marathon run (Fig. 2).
The optimized C2F6 cleaning recipe (DOE condition A3) was then released for a pilot production run over a six-month period, and 120 production lots were processed using new cleaning conditions. In addition to the above process parameters that were monitored during the test wafer marathon run, product lot electrical and yield performance were also evaluated. It was concluded that, statistically, there is no significant difference in electrical and yield performance between lots processed with CF4 cleaning and optimized C2F6 cleaning.
Conclusion
Converting CF4 cleaning gases to C2F6 in legacy CVD tools contributes to reduced PFC emissions for better environmental protection. An optimized cleaning process with C2F6 significantly reduces gas usage with comparable cleaning efficiency and better throughput. Production tests over 120 product lots showed comparable performance in key film properties, as well as product yield.
Acknowledgments
The authors would like to thank Jin-Song Xu of Chartered for his work and discussions.
Reference
- Reduction of Perfluorocompound (PFC) Emission: 2005 State-of-the-Technology Report, Technology Transfer # 05104693A-ENG, International Sematech Manufacturing Initiative, 2005.
- Allen Evans, Lance Nevala, Charles Allgood, “Advances in Reducing PFC Emissions and Gas Costs,” SEMICON Southwest, 2002.
- Gary Loh, Michael Mocella, Terry T. Lee, “Cost and Emission Reduction in Existing (PE)CVD Tools,” SEMICON China, 2005.
Ai-Guo Jiang received his MS in chemical engineering from Shanghai Jiao Tong University. He is a senior CVD process engineer at Chartered Semiconductor Manufacturing Ltd, 60 Woodlands Industrial Park D, Street Two, Singapore 738406; ph 65/6360-4007, e-mail [email protected].
Eng-Hwa Wong received his BS from Singapore SIM U. He is a senior principal technology development engineer at Chartered Semiconductor Manufacturing Ltd.
Koh-Ping Chai> received his bachelors of engineering in electronic and electrical engineering from Loughborough U. of Technology, UK. He is CVD process manager at Chartered Semiconductor Manufacturing Ltd.
Fang-Hong Gn received her MS from the U. of Manchester Institute of Science and Technology, UK, in material science and device technology. She is technology development director at Chartered Semiconductor Manufacturing Ltd.
Terry T. Lee received his MS degree in chemical engineering from Tsing-Hua U. He is the technical marketing & service manager for DuPont Fluoroproducts, serving the Asia Pacific region.
Gary Loh received his MS in chemical engineering from Georgia Institute of Technology and his MBA from James Madison U. He is a technical program manager for DuPont Electronic Gases.