RequestLink
MICRO
Advertiser and
Product
Information

Buyer's Guide
Buyers Guide

tom
Chip Shots blog

Greatest Hits of 2005
Greatest Hits of 2005

Featured Series
Featured Series


Web Sightings

Media Kit

Comments? Suggestions? Send us your feedback.

 

MicroMagazine.com

Critical Materials

Selecting proper chamber-cleaning processes for installed-base CVD tools

Franco Blundo, Manfredi Alberghina, and Marcello Frazzica,
STMicroelectronics; Gary Loh, DuPont Electronic Gases; and
Michael Mocella, DuPont Electronic Technologies

Numerous studies have been performed on reducing perfluorocarbon (PFC) emissions from chemical vapor deposition (CVD) chamber-cleaning processes. In new-generation tools, nitrogen trifluoride, a low-emission chamber-cleaning gas, is widely used to reduce PFC emissions. In addition, the development of several new cleaning gases, such as F2 and COF2, has shown promising results.

However, fabs with 150- and 200-mm installed-base tools, which contribute significantly to PFC emissions, require other practical and economical solutions to reduce emissions. The goal of the project discussed in this article was to select a chamber-cleaning gas to replace C3F8 in Concept-1 tools from Novellus (San Jose). The alternative cleaning gas not only had to be effective at reducing PFC emissions, but it also had to be practical and economical to meet the following requirements:

• No increase in gas use and costs.

• No additional safety, health, and environmental (SHE) management costs.

• Low conversion costs.

• Low conversion effort.

This study presents a c-C4F8-based chamber-cleaning chemistry that offers both process and SHE benefits while involving minimal cost and process modifications. Based on a balanced approach, this work determined that the benefits of c-C4F8 compare to those of the C3F8 cleaning process. Several factors related to gas costs and SHE concerns were evaluated. Other issues investigated included gas delivery. Previously optimized c-C4F8 recipes were used to minimize c-C4F8 recipe development work. Fourier transform infrared (FTIR) and optical emission endpoint detectors were used to verify the completion of chamber cleaning and the emissions reductions. The results of the work described here demonstrated that the c-C4F8 cleaning process reduced PFC emissions by 57% over the C3F8 cleaning process. Also, reduced gas consumption from the use of c-C4F8 resulted in a 20% cost savings. This work demonstrates that a low-cost, low-effort approach can reduce PFC emissions in installed-base tools.

Comparing c-C4F8 and C3F8 gases

In chamber cleaning, a fluorine-containing cleaning gas is converted into atomic fluorine for removing deposition process by-products from CVD chambers. A high fluorine conversion rate enables high gas utilization. Cleaning gases that offer high gas utilization reduce the unreacted cleaning gas in tool exhaust, leading to lower PFC emissions.

Various studies indicate that of all CxFy chamber-cleaning gases, c-C4F8 has the highest gas utilization. Based on data from gas suppliers, c-C4F8 has been shown to attain higher gas utilization than CF4, C2F6, C3F8, or C4F8O.1 It has also been demonstrated that c-C4F8 requires a much lower gas concentration than C2F6 or C3F8 to accomplish the same etch rate.2 In addition, it has been shown that c-C4F8 has lower PFC emissions and gas use rates than C2F6, C3F8, and C4F8O.2,3 Many studies indicate that the higher gas utilization of c-C4F8 enables larger reductions in gas use and PFC emissions.

In the study presented in this article, cleaning-gas utilization was used as a theoretical basis for gas selection. Reductions in gas use and PFC emissions were estimated based on comparing the flow and utilization of c-C4F8 and C3F8,assuming that the cleaning time for both processes was the same. Based on that estimation, it was determined that c-C4F8 can reduce gas use by 45% and PFC emissions by 61%. Those estimated benefits met the goal of the project. Table I presents the estimated benefits of replacing C3F8 with c-C4F8.

Table I: Estimate of benefits from alternative chamber cleaning process.

At the developmental stage, some chamber-cleaning gases have had higher gas utilization rates than c-C4F8.3 While these gases can potentially reduce PFC emissions further, more developmental work is required for their implementation. In contrast, c-C4F8 provides a low-cost, low-effort alternative that can be implemented immediately in installed-base CVD tools to achieve PFC emissions reductions, as highlighted in Table II.

Data from fab tests show that c-C4F8 conversion can be accomplished with little cost and effort. For example, it has been demonstrated that the drop-in replacement of C3F8 with c-C4F8 in the Novellus Concept-2 tool significantly reduces PFC emissions and gas costs without having a negative impact on production and quality.4

This study evaluated SHE management of c-C4F8 in terms of health hazard, flammability, reactivity, and corrosivity, as shown in Table II. The gas is nontoxic and nonflammable. It is not reactive and is compatible with most metals, elastomers, and plastics under normal operating conditions. Based on that information, no change in the gas-delivery system was made to address the safety issues raised by shifting to the use of c-C4F8.

The cost of converting to the use of c-C4F8 was evaluated based on the required modification of the gas-delivery system, the tool, and the abatement system. The only change to the delivery system was the installation of a low-pressure regulator for the c-C4F8 cylinder. Since c-C4F8 has a lower vapor pressure than C3F8, the pressure drop across the delivery system was evaluated. The vapor pressure of c-C4F8 under ambient conditions (21°C) is 25 psig. To ensure at least a 10-psig supply pressure to the mass-flow controller (MFC), the pressure drop at the required c-C4F8 flow rate should be less than 15 psi. Experimental and modeling studies of c-C4F8 under various conditions (pipeline length and diameter, the number of bends in the pipe, and c-C4F8 flow rate) provided the basis for studying the pressure drop.5 According to those studies, the pressure drop for c-C4F8 at a flow rate of 1 std L/min across a 1000 feet of 1¼-in. stainless-steel pipe is less than 1.5 psi. In the Concept-1, which has a c-C4F8 flow rate of 0.4 std L/min, the pressure drop is negligible.

Using c-C4F8 vapor-pressure data from DuPont (Wilmington, DE), the effect of temperature on pressure was also evaluated. Based on that evaluation, c-C4F8 cylinder pressure was adjusted to maintain 15 psi of pressure at the MFC. That pressure level ensured that adequate pressure would be supplied to the MFC and that the pressure of the gas-delivery system would be maintained at well below the c-C4F8 vapor pressure of 25 psig at 21°C, leaving enough room for unexpected reductions in ambient temperature. A low-pressure regulator on the c-C4F8 gas cylinder enabled pressure adjustment and control in the low pressure range.

Since c-C4F8 requires a much lower flow rate than C3F8, the MFC that was used to control C3F8 flow with a full flow range of 10 std L/min was replaced. A calibration factor of 0.16 was used for the c-C4F8 MFC, which was calibrated based on N2.

Because of limited tool availability, c-C4F8 recipe development work was minimized. Hence, to minimize the number of test runs required, pressure and power were kept unchanged. Instead, the study focused on the c-C4F8 flow rate and the ratio of c-C4F8 to O2, which have the greatest impact on PFC emissions, gas use, and cleaning time, according to published data.6

Experimental Procedure

Three C3F8 chamber cleans were performed to gather baseline data. The cleans were performed according to the following standards:

• Clean 1: 770 std cm3/min of C3F8, 960 std cm3/min of O2, 3.2 Torr, 2650 W, fixed time of 1500 seconds.

• Clean 2: 770 std cm3/min of C3F8, 960 std cm3/min of O2, 0.7 Torr, 2500 W, fixed time of 750 seconds.

The recipes for three c-C4F8 cleans were chosen based on DuPont’s previous c-C4F8 process optimization work on Concept-1 tools. The direct application of the optimized recipes reduced the efforts required to qualify the new gas. For c-C4F8 cleans, the following test ranges were used:

• Clean 1: 400–450 std cm3/min of c-C4F8, 1800–2000 std cm3/min O2, 3.2 Torr, 2650 W, fixed time of 1500 seconds.

• Clean 2: 450 std cm3/min of c-C4F8, 1800 std cm3/min of O2, 0.7 Torr, 2500 W, fixed time of 750 seconds.

Before each experimental cleaning, 25 wafers were processed using the standard nitride recipe. Deposition (including undercoating and precoating) took 90 minutes to control the film thickness at 10 k Å. Since film thickness was kept constant, comparisons between c-C4F8 and C3F8 cleaning performance were based on films of the same material and thickness.

The c-C4F8 chamber-cleaning endpoint was based on the equivalent C3F8 chamber-cleaning endpoint. The volumetric SiF4 emitted during a standard C3F8 clean was used as the baseline for an acceptably clean chamber. When the amount of SiF4 emitted from the chamber after c-C4F8 cleaning matched the amount emitted after C3F8 cleaning, it was assumed that both gases had achieved the same level of chamber cleanliness. The endpoint was also verified visually by observing the appearance of a standing-wave plasma pattern in the chamber during the cleaning process. When there was no longer any SiO2 film in the chamber for the fluorine-containing plasma to consume, horizontal stripes in the plasma appeared under the showerhead, indicating that chamber cleaning was completed.

The exhaust gas during chamber cleaning was monitored using FTIR. In each experiment, chamber-cleaning time, gas use, cleaning efficiency, and PFC emissions were calculated from the FTIR measurements. Gas use for C3F8 and c-C4F8 was calculated based on the cleaning time and the MFC setting. The amount of SiF4 emitted during cleans 1 and 2 was measured to evaluate cleaning efficiency. Using the following equation, the researchers were able to integrate the total volumetric output of CxFy and CF4 during those cleans 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 in kilograms and GWP is the global warming potential (100-year integrated time horizon [ITH]). The GWP values for C3F8, CF4, and c-C4F8 are 7000, 6500, and 8700, respectively.

Results and Discussion

Table III lists the relative values of C3F8 versus c-C4F8 for cleaning time, gas use, cleaning efficiency (amount of SiF4 emitted), and PFC emissions (kg CE). All three c-C4F8 recipes consumed less gas and resulted in lower PFC emissions than the C3F8 gas. With the same cleaning time and cleaning efficiency as the other c-C4F8 recipes, c-C4F8 recipe 1 (with the lowest flow rate of 400 std cm3/min) consumed the least amount of gas and had the lowest level of PFC emissions. The measured reductions in gas use and PFC emissions were within 5% of the estimated reductions.

Figures 1 and 2 present FTIR concentration profiles from the tool exhaust streams recorded during the C3F8 and c-C4F8 chamber-cleaning processes. The figures demonstrate that the c-C4F8 concentration measured in the exhaust stream was much lower than that of C3F8, confirming that the high gas utilization of c-C4F8 leads to lower PFC emissions. Flow control of the c-C4F8 process was stable, as indicated by the stable c-C4F8 concentration shown in Figure 2. During cleans 1 and 2, the plasma was also stable. Figure 3 compares the unreacted cleaning gas (CxFy) that was emitted between cleanings for c-C4F8 and C3F8.

Figure 1: FTIR exhaust analysis showing chamber-cleaning performance of C3F8 and other gases. (The C3F8 flow rate was 770 std cm3/min and the gas mixture contained 55% O2.)
Figure 2: FTIR exhaust analysis showing chamber-cleaning performance of c-C4F8 and other gases. (The c-C4F8 flow rate was 400 std cm3/min and the gas mixture contained 83% O2.)
Figure 3: Comparison of cumulative unreacted CxFy from C3F8 and c-C4F8 cleans.

The c-C4F8 cleaning process also generated much less CF4 than the C3F8 process. The higher percentage of O2 during c-C4F8 cleaning may convert some CF4 to COF2. Figure 4 compares the cumulative CF4 generated per clean during C3F8 and c-C4F8 chamber-cleaning processes.

Figure 4: Comparison of cumulative CF4 from C3F8 and c-C4F8 cleans.

At this point, a dual-wavelength endpoint detector was installed to confirm the completion of chamber cleaning. The optical emissions of fluorine (704 nm) and CO (483 nm) were used to determine the endpoint. The endpoint-detection software scripts used for the C3F8 cleaning process were also used for c-C4F8 process, and the same threshold value of 0.85 was used for both gases. The traces observed during the C3F8 and c-C4F8 cleaning processes were almost identical, and the times required to reach the threshold ratio were the same. The endpoints for both C3F8 and c-C4F8 were equivalent, confirming chamber-cleaning efficiency as determined by the FTIR-based detection of SiF4 emissions from the chamber.

Figure 5 compares the clean 1 performance of C3F8 and c-C4F8 after the endpoint detector was installed. The figure also demonstrates that cleaning time as determined by the endpoint detector is affected by the ratio of c-C4F8 to O2. Reducing the O2 concentration by 4.4% through lowering the O2 flow from 2000 to 1500 std cm3/min at a c-C4F8 flow rate of 400 std cm3/min increased cleaning time by nearly 12%. However, a 13% increase in the c-C4F8 flow rate (from 400 to 450 std cm3/min) did not affect cleaning time. To achieve the optimum c-C4F8:O2 flow ratio, fine-tuning may be necessary to compensate for small MFC errors.

Figure 5: Comparison between cleaning time for C3F8 and c-C4F8 gases and impact of c-C4F8:O2 ratio on cleaning time.

Figure 6 compares the cost per wafer of performing C3F8 versus c-C4F8 chamber cleaning. The 42% reduction in gas consumption achieved by shifting to the use of c-C4F8 resulted in a 20% reduction in cleaning-gas costs.

Figure 6: Chamber-cleaning cost comparison between C3F8 and c-C4F8 gases.

Figure 7 compares the PFC emissions between C3F8 and
c-C4F8 chamber cleaning. As the figure shows, c-C4F8 reduced PFC emissions by 57%.

Figure 7: PFC emissions comparison between C3F8 and
c-C4F8 gases.

Conclusion

The selection of a chamber-cleaning gas requires a balanced approach. For fabs with installed-base CVD tools that must reduce PFC emissions immediately at minimal cost and with little effort, c-C4F8 offers a practical and economical solution. The goal of the project presented in this article was to reduce PFC emissions inexpensively and easily. The data reported in this article demonstrate that the c-C4F8 chamber-cleaning process has accomplished that goal. As a result, c-C4F8 is being implemented in production.

Acknowledgment

This article is an edited version of a poster presentation at the IEEE/SEMI Advanced Semiconductor Manufacturing Conference, held April 11–12, 2005, in Munich, Germany.

References

1. OF Schedlbauer, “Cost Reduction Challenges in CVD Chamber Cleaning: Strategies to Reduce Gas Costs,” Future Fab International 13 (2002): 164–169.

2. CC Allgood and MT Mocella, “CVD Chamber Cleaning: A Critical Comparison of Processes and Gases,” DuPont technical paper [online] [cited 1 September 2005]; available from Internet: www.dupont.com/zyron/techinfo/recent/semiconw2000.html.

3. T Beppu, “Emission Reduction from Semiconductor Industry Chamber Cleaning” (paper presented at the ISESH Conference, Noordwijk, The Netherlands, June 29–July 3, 2003).

4. A Evans, “Advances in Reducing PFC Emissions and Gas Costs: Qualification and Implementation of c-C4F8 Chamber Clean Gas in a Novellus Concept-2 Chemical Vapor Deposition Tool” (paper presented at the Semicon Southwest Partnership for PFC Emission Reduction Workshop, Austin, TX, October 15–16, 2002).

5. “Pressure Drop Considerations in Gas Delivery Lines for Zyron 8020 (c-C4F8) CVD Chamber Clean Gas,” DuPont technical paper [online] [cited 1 September 2005]; available from Internet: www.dupont.com/zyron/techinfo/briefs/index.html.

6. U Schilling and P Heger, “Evaluation of c-C4F8 as Cleaning Gas in an Applied Materials Lamp Heated Universal Chamber on 200mm Wafers,” International Sematech Technology Transfer 01084152A-TR, 2001.


Franco Blundo is the environmental champion at STMicroelectronics’ 150-mm production fab in Catania, Italy, where he is also responsible for the company’s PFC reduction program. After gaining experience in the petrochemical industry, he joined STMicro in 1994 as a process engineer section head for CVD, PVD, and back-end process. He received a degree in industrial chemistry from the University of Catania in 1988. (Blundo can be reached at +39 095 7407274 or franco.blundo@st.com.)

Manfredi Alberghina is a process and equipment engineer at STMicroelectronics’ 150-mm front-end wafer fab in Catania, where he is responsible for APCVD/PECVD deposition tools. For more than five years, he has been involved in various aspects of front-end manufacturing, including development, characterization, and integration of deposition processes; statistical process control; overall equipment effectiveness; and benchmarking, training, and automation in the passivation area. He received a degree in physics from University of Catania in Italy. (Alberghina can be reached at +39 095 7407854 or manfredi.alberghina@st.com.)

Marcello Frazzica is a process engineer in the CVD, PVD, and back-end areas at STMicroelectronics’ 150-mm wafer fab in Catania. Previously, he worked at Wyeth Lederle. He received a degree in physics from Italy’s Messina University in 1997. (Frazzica can be reached at +39 095 7407672 or marcello.frazzica@st.com.)

Gary Loh is a technical program manager for the electronic gases group of DuPont (Wilmington, DE), where he is responsible for global electronic gas applications, technical support, and technical programs in the semiconductor industry. He has extensive professional experience with a variety of fluorocompound products and analysis techniques. He received an MS in chemical engineering from the Georgia Institute of Technology in Atlanta and an MBA from James Madison University in Harrisonburg, VA. (Loh can be reached at 302/999-4971 or gary.loh@usa.dupont.com.)

Michael Mocella, PhD, is a technology planning manager at DuPont Electronic Technologies in Wilmington, DE, where he is responsible for coordinating technical programs for fluorine-containing gas applications in the dry etch and chamber-cleaning areas. In addition, he works with advanced technology programs in new materials and processes for interconnect and lithography applications. For more than 20 years, Mocella has worked with semiconductor materials at DuPont, including thin-film polyimides, advanced photomasks, and polishing slurries. He has also focused on catalysis and chemical process optimization in the chemical industry. He received a BS in chemistry from Michigan State University in East Lansing and a PhD in inorganic chemistry from the University of Illinois in Urbana-Champaign. (Mocella can be reached at 302/992-3510 or michael.mocella@usa.dupont.com.)


MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit

Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com.

© 2007 Tom Cheyney
All rights reserved.