 |
 |
 |
 |
|
Comments? Suggestions? Send us your feedback.
MicroMagazine.com |
GREEN AND CLEANFacing the challenges of reducing PFC emissions in plasma chamber cleansJosep Arnó and Joseph D. Sweeney, EcoSys Perfluorocompounds (PFCs) are widely used in the semiconductor industry for plasma etching and chemical vapor deposition (CVD) chamber clean applications. The global warming potential of these gases has triggered a significant effort to minimize their release into the atmosphere. In December 1997, more than 160 countries negotiated the Kyoto Climate Protection Protocol to encourage efforts to reduce the emission of greenhouse gases. The international agreement specifically targeted six gases, including perfluorinated gases and sulfur hexafluoride (SF6). These six fluorine-saturated species are among the strongest greenhouse gases, with global warming potentials three and four orders of magnitude higher than carbon dioxide (CO2). These extremely stable molecules can remain in the atmosphere for thousands of years. Worldwide concerns regarding climate changes are likely to generate tight environmental regulatory constraints in the near future. Current approaches to reducing PFC emissions include us-ing alternative chemicals and recovery/ recycle systems, improving the use of PFCs within a tool, and applying abatement techniques. Tool optimization efforts have resulted in high PFC conversion into significant volumes of fluorine gas (F2) and, to a lesser extent, silicon tetrafluoride (SiF4) and hydrogen fluoride (HF). This article evaluates the performance of a point-of-use water scrubber challenged with changes in high concentrations of F2 and provides a continuous, in-line, quantitative characterization of the effluent gases. The conditions investigated represent or exceed concentrations released during PFC plasma cleans. Special attention was paid to improving abatement efficiency while minimizing the consumption of resources and eliminating the formation of unwanted by-products. The electronics industry uses PFCs in a number of plasma processes that generate highly reactive F2 and fluorine radicals. These in situ—generated species are produced to remove residue from process tools and also to etch thin films. The most common PFCs include carbon tetrafluoride (CF4), hexafluoroethane (C2F6), SF6, octafluoropropane (C3F8), and nitrogen trifluoride (NF3). Process tool chamber cleans after CVD processing account for 60—95% of the industry's PFC use.1 Ongoing research to reduce PFC emission levels falls into four categories: process optimization, alternative chemicals, recovery/recycle, and abatement. Updated overviews of recent developments in these areas were presented in April at the Global Semiconductor Industry Conference on Perfluorocarbon Emissions Control in Monterey, CA. Industry leaders preferred process optimization as the means to reduce PFC emissions. This approach involves adjusting the operating conditions inside the process reactor to achieve enhanced PFC conversion within the tool. Nonoptimized conditions result in PFC use that varies depending on the specific gas and process used. For instance, oxide etch, which uses a combination of CF4 and CHF3, ranks the lowest in PFC utilization with 15% efficiency. However, tungsten deposition processes reportedly use up to 68% NF3.1 Recent developments in optimized plasma clean technologies provide up to 99% NF3 utilization within the tool.1 High PFC conversions inevitably result in the formation of hazardous air pollutants (HAPs). During the conversion process, F2 and SiF4 gases are the most prevalent, while HF and carbonyl fluoride (COF2) are also present. The destruction of these fully fluorinated gases generates significant HAP yields compared to the initial PFC volumes delivered to the tool. For instance, assuming stoichiometric conversion of PFCs into F2, a 1 L/min flow rate of NF3 could potentially produce 1.5 L/min of F2. In a combined exhaust stream of four chambers, up to 6 std L/min of fluorine gas would be generated, resulting in a postpump effluent concentration of 3% F2 (using 50 std L/min N2 ballast per pump). These flows double with hexafluorinated PFCs (compared to NF3) and are likely to increase dramatically in the future with the projected throughputs of 300-mm wafer manufacturing. This represents a worst-case scenario and does not account for the short duration and periodic nature of processes using PFCs, the lower concentrations of F2 emissions during initial cleaning stages, and the reduced probability that two or more chambers run synchronized PFC cycles. The toxic and corrosive nature of fluorinated HAPs poses considerable health and environmental hazards as well as jeopardizes the integrity of exhaust systems. In particular, the oxidizing power of F2 is unmatched by any other compound, and fluorine gas is far more reactive than other halogens. The large volumes of F2 and other fluorinated hazardous inorganic gases released during optimized plasma processing will require point-of-use abatement devices in order to minimize potential dangers and to prolong tool operation. Abatement Alternatives At high concentrations, fluorine reacts exothermically with all elements except oxygen, nitrogen, and noble gases. Consequently, a reasonable method of F2 abatement is to remove this highly active gas using naturally occurring reactions without adding energy to the system. Heat dissipation and the formation of acceptable by-products present the main challenges to this approach. Thermal abatement units combine reactive materials and F2 inside a reactor that is heated using fuel or electrical energy. The by-products generated by the thermal abatement of F2 typically include hot acids that necessitate the use of a postwater scrubber. The containment of hot, concentrated acids requires expensive construction materials to prevent a temperature-enhanced corrosion attack on ducting and other downstream systems. The removal efficiencies in these postscrubber beds are often compromised since the scrubbing efficiency of most acid gases decreases as a function of temperature. An alternative would be to flow the fluorine gas stream through a dry bed filled with a reactive material. Suitable dry chemicals would convert F2 into innocuous solids or benign gases without generating excessive heat. However, this solution would have limiting factors, especially when exposed to large volumes of F2. Fluorine gas does react quickly and efficiently with water. The main by-products of this reaction were shown to be HF, O2, and H2O2.2 There have been objections to using water scrubbers to abate fluorine because of concerns about the lack of applicable experimental data, the formation of unwanted oxygen difluoride (OF2), and thermal considerations regarding exothermic reactions. This article addresses these objections and other concerns by providing relevant experimental data about the abatement of F2 using a point-of-use water scrubber. Scrubbing efficiencies, temperature profiles, and complete by-product formation was measured continuously in situ and in real time using state-of-the-art analytical equipment. Abatement efficiencies are reported as a function of experimental parameters that used considerable amounts of F2 gas (up to 6% F2). Foremost attention was paid to minimizing water use and to discerning and inhibiting the mechanisms responsible for the formation of OF2. The study reported here resulted in the development of innovative methods and led to enhanced fluorine abatement without the formation of unwanted OF2. Aqueous F2 Chemistry The nature and concentration of by-products formed by the reaction between water and F2 depend on a combination of competing reactions and physical conditions. Reported by-products of this reaction include HF, O2, and H2O2, along with small concentrations of OF2. Fundamental studies have identified HOF as a reaction intermediate3 and determined the mechanisms causing the formation of oxygen difluoride,4 which is a stable, colorless, poisonous gas that can produce high yields under any of the following three exceptional conditions. • Adding electric energy (electrolysis) to an aqueous HF solution:5 2 HF(aq) + 6 H2O(l) — OF2(g) + O3(g) + 7 H2(g)
• Passing fluorine through an aqueous alkaline solution (60% maximum yield using 0.5 to 1-molar alkali concentrations):2 2 F2(g) + 2 H-(aq) — OF2(g) + 2 F-(aq) + 7 H2O(l) • Passing F2 over cooled ice.4 (Reactions took place at the surface of ice and did not occur in liquid water): F2(g) + H2O(s) — HF(g) + OF2(g) Oxygen difluoride was also reported to be formed when fluorine gas was passed through 60% HClO4, H5IO6, and hydrated alkali fluorides. Under normal scrubber operating conditions, without the use of caustic injection, the concentration of OF2 generated from the reaction of F2 and water was insignificant. Water Scrubber Technology The experiments described in this article were performed in the applications laboratory of EcoSys (Danbury, CT). Fluorine gas mixtures delivered to the scrubber were generated using a passivated gas delivery manifold equipped with mass-flow controllers from MKS Instruments (Andover, MA). Fluorine gas was sourced from 5% certified-grade and 100% technical-grade cylinders (Air Products and Chemicals, Allentown, PA). Every water and gas flowmeter used in this study was calibrated to provide accurate experimental parameters. The point-of-use abatement tool consisted of an enhanced Vector-100 water scrubber (EcoSys) as shown in Figure 1. In this study, the detailed description of the tool will be limited to features that are relevant to the analytical method and results. The standard scrubber operates using a vertical cocurrent flow of water and contaminated gas stream. Water-active species are hydrolyzed as they interact with water in a high-surface-area-packed region, and the resulting liquid falls to a water reservoir or sump. The resulting scrubber gas stream exits the scrubber through a vertical 4-in. duct connected to a blower. The water dynamics within the scrubber include fresh or makeup water flowing into the system, wastewater draining out, and continuous recirculation of water stored in the sump.
Figure 1: Scrubber features and setup. The scrubber performance was enhanced using an innovative polishing countercurrent packed bed installed at the gas exhaust (see Figure 1). This improvement alone accounted for an estimated 60% increase in F2 removal efficiency compared to the standard configuration. A Type-9 entry was used to minimize solid deposits. The metal portion of this system was chemically treated to protect it from corrosion. Gas and water temperatures within the scrubber were measured at nine selected points (T1—T9), as shown in Figure 1. Throughout these experiments, temperatures were logged into a computer every 30 seconds to characterize the thermal profiles during the abatement processes. On-Line Analytical Methods Figure 2 shows the analytical setup used to monitor gas-phase species at the exhaust of the water scrubber. Infrared active gases were sampled in an I-2000 Fourier transform infrared (FTIR) spectrophotometer (MIDAC, Irvine, CA) for quantitative analyses. A three-way valve installed at the sampling port permitted the collection of either dry nitrogen or process gas into the analyzer. Nitrogen was used to purge the sampling lines and to collect the background spectra before each experiment. The unit was equipped with a nickel-coated, 10-m path-length gas cell with zinc selenide windows and liquid nitrogen—cooled mercury cadmium telluride detector. Full spectra collected every 30 seconds provided continuous, real-time information about the nature and concentration of species of interest.
Figure 2: Schematic of analytical setup and sampling system. Accurate quantitative analyses were achieved by calibrating the analyzer in situ using known concentrations of HF. Oxygen difluoride absorbances were converted into concentrations using a quantitative spectral library provided by the FTIR supplier. The spectral regions chosen for analyses are shown in Figure 3. These regions were selected to minimize water interference and to optimize the signal-to-noise ratios.
Figure 3: FTIR spectra and selected spectral regions used for quantitative analyses of species of interest. Fluorine gas was analyzed in a continuous mode using an F2-specific gas sensor cell (PureAire Monitoring Systems, Rolling Meadows, IL). This electrochemical sensor uses a combination of gas membrane and galvanic cell technology to monitor low concentrations of the hazardous gas. The sensor is designed specifically for in situ monitoring of F2 under water vapor—saturated conditions. This feature was an important consideration that led to this analytical method being chosen over standard mass spectrometric techniques. A recent study reported important limitations using mass spectrometers to measure acid gases in moisture-rich streams.6 In order to provide continuous analyses within the detection limit of the monitoring device (3-ppm F2), the known flow rates of scrubber gas exhaust were diluted with metered nitrogen flows. The combined stream was introduced into a mixing chamber equipped with the F2 sensor (see Figure 2). The responses to changing F2 concentrations were logged into a computer every 30 seconds. By calibrating the sensor against known concentrations of F2, accurate quantitative results were achieved. Experimental Results The removal efficiencies of fluorine challenges ranging between 0.5 and 5 std L/min were investigated as a function of selected experimental parameters. These gas streams were diluted with 50 std L/min of nitrogen, resulting in concentrations between 1 and 6% F2. The effects of residence time within the scrubber were also studied by increasing the nitrogen flow rate to 200 std L/min. The scrubber performance was tested using standard-flow (1.2 gal/min) and low-flow (0.75 gal/min) makeup water rates. A proprietary chemical was injected into the scrubber during high fluorine gas challenges to improve fluorine removal and to eliminate the formation of any OF2 by-products. Previous unreported tests performed in EcoSys laboratories confirmed that caustic injection was not a suitable option. Although using caustic injection enhanced fluorine abatement, it generated unacceptable levels of OF2. In contrast, a nontoxic, nonalkali, readily available, inexpensive compound that was used in combination with the water scrubber did not produce OF2 or other unwanted reaction products.
Table I: Summary of experimental conditions and fluorine abatement results. Scrubbing efficiencies, selected experimental conditions, and by-product formation are summarized in Table I. The percentage of destruction and removal efficiency (% DRE) was determined using the standard expression:
in which Inlet F2 represents fluorine inlet concentration in ppm, and Outlet F2 Equiv. is defined as:
The study confirmed that fluorine gas reacts with water quickly and effectively. Achieving more than 99% destruction and removal efficiencies of fluorine was possible under all the conditions investigated. As expected, the makeup water flow rate affected scrubbing efficiency and was a limiting factor under high fluorine challenges (see Figure 4). Even without chemical enhancement, and with 50 std L/min N2 ballast, OF2 concentrations remained below 10 ppm for initial F2 concentrations of <3% and between 0.75 and 1.2 gal/min of makeup water.
Figure 4: Flourine abatement efficiency as a function of makeup water, without chemical injection, and with 50 std L/min N2 ballast and water temperature at 20°C.
Figure 5: Formation of gas-phase HF and OF2 by-products as a function of inlet fluorine concentration, without chemical injection. Figure 5 depicts breakthroughs of OF2 and HF as a function of initial F2 concentration using reduced water flow rates. HF breakthroughs were probably the result of vapor-liquid equilibrium effects within the concentrated aqueous HF formed in the water scrubber. Tests 8 to 10 (see Table I) demonstrated that chemical injection inhibited OF2 formation as well as enhanced scrubbing efficiency. These improvements were instrumental in the destruction of up to 5 std L/min of fluorine gas under the stringent conditions of reduced residence times (200 std L/min N2). Comparing the data from tests 6 and 8 showed that outlet concentrations of HF and F2 were reduced by a factor of 10, and OF2 concentration fell below detection limits when chemical injection was used. The continuous, on-line concentrations of the scrubber effluent gases measured during tests 8—10 are shown in Figure 6.
Figure 6: Continuous concentration of gases measured on-line at the outlet of the scrubber during chemical injection experiments.
Analyses of the temperature data indicated that any exothermic effects were effectively dissipated within the scrubber. The only measurable temperature changes were recorded inside the Type-9 entry where the first contact occurred between incoming gas and water vapor. A maximum temperature increase from 17° to 25°C ( Process Applications The removal efficiencies reported in this article are likely to describe worst-case scenarios when compared to the scrubbing efficiencies of the effluent gases released during plasma chamber cleans. First, the reported outlet concentrations represent the equilibrium values reached after extended and continuous delivery of fluorine gas into the scrubber. This steady state was typically reached between 10 and 30 minutes after the start of tests, depending on the initial F2 concentration. The duration of chamber cleans is often a fraction of the time necessary to reach that equilibrium. Second, the time-dependent concentration of F2 released during typical chamber cleans is not constant. During the initial stages, most F2 produced in the chamber is used to react with SiO2, releasing SiF4 gas. It is only after SiO2 is depleted that excess F2 is discharged by the tool in significant amounts.7 Regardless of whether or not a chemical was injected into the scrubber, the drained water contained high concentrations of dissolved fluoride in a low-pH solution. Some fabs are already equipped with end-of-pipe water treatment systems to recover HF or fluoride from their effluent streams. Other water-handling techniques may include postscrubber water neutralization to produce benign fluoride salts. Conclusion The fluorine gas abatement efficiency of a water scrubber was studied as a function of initial F2 concentration, water flow rate, and overall gas flow rate conditions that represent or exceed the F2 levels released during PFC plasma chamber cleans. A 99% destruction and removal efficiency of fluorine was achieved under all the conditions investigated. In addition, OF2 formation was virtually eliminated by the injection of an innovative, noncaustic chemical during high fluorine challenges. Throughout this study, there was no detected corrosion and temperature degradation of the point-of-use abatement tool. Recent unreported studies performed in the company laboratory have resulted in important breakthroughs in the area of silane abatement, also using a water scrubber. Silane is used in a number of CVD processes preceding plasma chamber cleans. By adding this capability to a conventional water scrubber, a single abatement device might be possible for all CVD processing involved. Although the semiconductor industry is focused on reducing PFC emissions through optimized tool solutions, the inevitable result is an increase in hazardous air pollutants. The high-PFC conversion process generates fully fluorinated by-products that are a significant health and environmental hazard. Using a point-of-use abatement solution will become increasingly important as the industry moves toward greater 300-mm wafer production, which has the potential to produce higher volumes of fluorinated hazardous inorganic gases. The point-of-use abatement solution provided here reduces the health and environmental hazards of by-products of the PFC conversion process and lessens the downstream corrosion effects of untreated fluorinated gases. Acknowledgments The authors would like to recognize Peter Dodson, Laura Hiscock, and Bruce Amico of EcoSys for their invaluable skills in facilitating the scrubber, building the hazardous gas delivery system, and assisting with data collection and effluent management. References 1. Langan J, Maroulis J, and Ridgeway R, "Strategies for Greenhouse Gas Reduction," Solid State Technology, 39(7):115—122, 1996. 2. Cady GH, "Reaction of Fluorine with Water and with Hydroxides," Journal of the American Chemical Society, 57(1):246—248, 1935. 3. Appelman EH, and Thompson RC, "Studies of the Aqueous Chemistry of Fluorine and Hypofluorous Acid," Journal of the American Chemical Society, 106(15):4167—4172, 1984. 4. Appelman EH, and Jache AW, "Concerning the Mechanism of Formation of Oxygen Difluoride," Journal of the American Chemical Society, 109(6):1754—1757, 1987. 5. Encyclopedia of Inorganic Chemistry, vol 3, King RB (ed), West Sussex, England, Wiley & Sons, p 1236, 1994. 6. Arnó J, "Continuous In-Line Monitoring of C12 and HCl in Water-Vapor-Saturated Gas Streams Using Mass Spectrometry," Spectroscopy, 13(2):54—62, 1998. 7. Ridgeway RG, Langan JG, Huling BA, et al., "Reduction in Emissions of Perfluorinated Compounds through Optimization of PECVD Chamber Cleans Using Nitrogen Trifluoride," in Proceedings of the 41st Annual Technical Meeting, Institute of Environmental Sciences, Mount Prospect, IL, IES, pp 480—486, 1995. Josep Arnó, PhD, is the chief technologist at EcoSys (Danbury, CT), an ATMI company. He is working on the development of innovative abatement technologies and effluent characterization. His research experience includes development plasma abatement technologies, molecular spectroscopy, and computational chemistry. He received a BS in chemistry and physics and a PhD in physical chemistry from Texas A&M University (College Station). (Arno can be reached at 203/792-1100 or jarno@atmi.com.) Joseph D. Sweeney has been an engineer at EcoSys for the past two years. He holds a BS in chemical engineering and a BS in chemistry from the University of Wisconsin (Madison). A member the American Institute of Chemical Engineers, Sweeney has submitted applications for more than four patents awaiting issuance. (Sweeney can be reached at 408/526-9400 or jsweeney@atmi.com.)
MicroHome |
Search | Current Issue | MicroArchives Questions/comments about MICRO Magazine? E-mail us at cheynman@gmail.com. © 2007 Tom Cheyney |
T = 9°C) was detected during the highest fluorine challenges (5 std L/min F2). The significant heat capacity of water combined with the unique high flow of recirculating water within the water scrubber provided the necessary cooling to remove the heat generated by the hydrolysis of F2. In addition, the scrubber was exposed to 3.2 lb (or the equivalent of 855 L) of fluorine gas, and no significant signs of corrosion or material deterioration were found within the scrubber after the tests were completed.
