Cylinder Package effects on the purity of the electronic specialty gases
06/01/1997
Cylinder package effects on the purity of electronic specialty gases
Philip B. Henderson, Ronald M. Pearlstein, Air Products and Chemicals Inc., Allentown, Pennsylvania
The composition of electronic specialty gases (ESGs) will remain unchanged for over two years in a well-prepared and passivated cylinder. Nevertheless, the cylinder package can affect the purity of an ESG if improper materials of construction, preparation, or handling techniques are employed. This report discusses these effects and how they can be avoided.
As device geometries of ICs continue to shrink below 0.25 ?m, more stringent control of individual processing steps used in manufacturing is necessary. Semiconductor manufacturers are demanding higher-purity ESGs and consistent purity of an ESG over time, both when a gas cylinder is in storage and in use. One concern of the process engineer is the degrading purity of the ESG from the time of original analysis, which may be several weeks or months before the ESG is used. In previous studies [1], we have found that more than two years after cylinder fill, the composition of ESGs remains unchanged, provided the cylinder is well prepared and passivated, and the analytical samples are withdrawn under controlled conditions using proper handling techniques. This is true not only for nonreactive gases such as nitrogen, but for reactive (e.g., SiH4) and corrosive gases (e.g., HCl, HBr) also. This report discusses the deleterious effects of ESG cylinder packages and how they can be avoided.
Materials of construction
Proper choice of wetted materials in storage containers and distribution systems used with specialty gases ensures reliable contamination-free delivery to the point of use. Storage and distribution sometimes have different requirements and should be considered separately. Table 1 summarizes the materials compatibility for storage of common ESGs. There are often several acceptable materials for a given gas; thus, the choice for a preferred package may be based on other considerations as well.
Many of the specialty gases used in microelectronics processes are stored under pressure in a liquefied form. Since the solvation power of liquids generally exceeds that of gases, the choice of material for a compressed gas cylinder may sometimes differ from that of the gas-phase distribution system. For example, liquid tungsten hexafluoride (WF6) readily attacks the chromium in stainless steel (SS), but gaseous WF6 can be satisfactorily delivered to a chemical vapor deposition (CVD) tool through properly passivated, electropolished type-316L SS tubing [2]. This highly reactive, liquefied specialty gas is therefore typically stored in nickel cylinders to preclude the dissolution of metallic contaminants.
Chrome-moly steel. The Cr/Mo low-alloy (chrome-moly) steel compressed gas cylinder is the time-tested package for most specialty gases. Properly prepared chrome-moly cylinders maintain the high purity of reactive materials such as anhydrous HBr [1], BCl3 [1], and, to a limited extent, WF6 [3]. In some cases, the potential reactivity of an ESG requires packages other than chrome-moly steel. Carbon monoxide (CO), for example, forms toxic metal carbonyl compounds when in long-term contact with steel or nickel alloys and should be packaged in aluminum cylinders with SS cylinder valves. Furthermore, CO distribution systems should not use nickel-face seal gaskets. Since they can be exposed to high CO pressures, even nickel-plated cylinder valve relief devices can be problematic [4].
Nickel/aluminum. Some specialty gases have been sold in aluminum- or nickel-coated steel cylinders. Though these cylinders have been labeled as higher-purity packages than chrome-moly steel, users must establish the long-term stability of the product in contact with them. For instance, nickel and its alloys, including, to a lesser degree, 316L SS, can hasten the decomposition of many hydride gases [5-7]. Ohmi and coworkers showed that the temperature at which gases such as SiH4, PH3, AsH3, and B2H6 decomposed, probably by catalytic dehydrogenation, was correlated with the concentration of nickel at the surface of the metal tubing. For the decomposition of silane, they found catalytic activity decreasing in the following order:
Pure Ni > Hastelloy C-22 > EP316L SS = Cr2O3-passivated superferritic SS
The reaction is hypothesized to be a surface-catalyzed decomposition; it depends more on the surface area of the tubing and less on the concentration of the hydride. While these surface effects were amplified by testing the hydride gases at extremely high dilution in Ar, and may be much less significant at higher concentrations, the trends observed provide guidance toward the selection of noncatalytically active materials.
Stainless steel. Electropolished type 316L SS is widely used for delivery of compressed gases in semiconductor manufacturing facilities. This material performs extremely well in maintaining the purity of inert, corrosive, and reactive gases and is resistant to corrosion under normal ambient conditions. However, 316L SS is subject to corrosion by halide ions (e.g., Cl, Br) in the presence of moisture. In a dry environment, 316L SS and low-alloy steel are both corrosion resistant.
Austenitic SS, including 316L, is susceptible to cracking when placed under tensile stress in the presence of halide ions [8]. This so-called stress-corrosion cracking greatly reduces mechanical strength, potentially leading to catastrophic failure. While this typically happens at elevated temperatures, stress-corrosion cracking of austenitic SS can also occur at near-ambient temperatures in the presence of traces of aqueous chloride ions. As a result, SS cylinders pose the potential risk of containment failure for the storage of compressed halide gases (e.g., HCl, HBr). Ferritic and duplex (ferritic/austenitic) SS alloys are more resistant to cracking in the presence of halide, and have been considered for use in the handling of corrosive gases.
Polymers. The polymeric materials often used for sealing surfaces in specialty gas service have more complex compatibility issues. While some fluorinated elastomers are chemically resistant, there is still no single material that is compatible with all specialty gases. The choice of elastomeric materials used on specialty gas cylinders is based on experience and specific exposure tests. For example, swelling of the valve seat on Cl2 gas cylinders hinders delivery of the gas at an acceptable rate [9]. We have found that a proprietary class of elastomers, however, eliminates this swelling problem.
Polyimide parts (e.g., Vespel) are widely used in certain high-purity valves due to the good machinability and lubricity of this polymer. While this material resists swelling in many gases, it can react chemically with gaseous bases such as anhydrous ammonia [10]. This reaction could weaken the polymer and potentially lead to particle shedding. All surfaces wetted by the product must be carefully evaluated both for physical and for chemical compatibility.
Cylinder preparation
Numerous reports and presentations have discussed the necessity of good cylinder preparation prior to electronic specialty gas service [11]. The major goal is to lower the residual moisture and particle concentrations within the cylinder by reducing the surface area, either mechanically or chemically, followed by baking at reduced pressures and cycle purging with a clean inert gas. Inadequate cylinder preparation procedures lead to moisture contamination of the ESG, which is detrimental to many fabrication processes. While moisture itself is a key impurity in many ESGs, several ESGs can also react with water vapor to make other contaminant species. For example, HF and mixed tungsten oxyfluorides are observed immediately upon introducing WF6 into a wet environment. Thus, HF and not water is the key potential impurity found in WF6, even though high levels of HF in WF6 are usually the result of a system`s moisture contamination. Table 2 lists examples of other key gaseous contaminants produced from reactions of ESGs involving water.
Cylinder handling
Without introducing any extraneous impurities, the purity level of an ESG can change while in use. Depending on the physical states of the ESG delivered, different procedures should be followed.
 |
Figure 1. Simulation of water vapor concentration in the gas withdrawn from a high-purity HBr cylinder that illustrates the dramatic increase in impurity concentration predicted as the liquid nears depletion. Note the logarithmic moisture scale.
Liquefied ESG. In a liquefied ESG, impurities in the cylinder can exist in both liquid and gaseous phases. Raoult`s Law for ideal solutions states that the total vapor pressure at any temperature is the sum of the vapor pressure of each component multiplied by the mole fraction of that component. The concentration of an impurity in the gas phase relative to that in the liquid phase will depend on the vapor pressures of the impurity and the ESG. When the impurity is less volatile (has a lower vapor pressure) than the ESG, it will tend to have more molecules in the liquid phase and increase in liquid concentration as the gas mixture is withdrawn from the cylinder. This is the case for many impurities found in liquefied ESGs. The liquid/vapor partitioning reduces the amount of contaminants delivered by the gas to the microelectronics fabrication process. As more gas is withdrawn, these impurities become more concentrated in the liquid "heel," while raising their relative gas-phase concentration proportionally with respect to the ESG concentration.
As soon as the liquefied ESG is depleted in the cylinder, however, the partitioning process ceases. The gas pressure in the cylinder continuously decreases as more product is removed. The impurity level in the withdrawn gas is now determined by the ratio of the impurity`s vapor pressure to the gas mixture pressure in the cylinder, until all of the impurities are vaporized. After this point, however, the relative concentrations of the species in the compressed gas mixture will remain constant as the pressure is reduced.
Since these partitioning effects can cause rapidly increasing impurity levels in the product withdrawn from an ESG cylinder that has run liquid-dry, most ESG cylinders should be replaced while some liquid heel remains. Figure 1 simulates vapor-phase moisture concentration as product is withdrawn from a cylinder starting with 65 lbs of HBr at a typical initial vapor-phase moisture concentration (250 ppbv).
The arguments above work in reverse, of course, for impurities that are more volatile than the ESG. These impurities, however, tend to be less of a concern. The partitioning processes obey the simple relationships of Raoult`s Law only if the impurities in question have a chemistry and molecular size similar to those of the ESG. In practice, most mixtures are imperfect. Nevertheless, the results of vapor/liquid partitioning will be qualitatively the same as those described for an ideal mixture.
Compressed gases. In an ESG that has a critical temperature below the ambient temperature, no liquid/vapor partitioning can occur. For example, BF3, B2H6, CF4, CO, NF3, SiF4, and SiH4 are all delivered as compressed but not liquefied gases. The pressure in these cylinders drops continuously as the product is removed. In general, sufficiently volatile impurities will be uniformly distributed; thus, the impurity levels should remain relatively constant. Due to the high-purity levels required for these gases, however, the effects of impurities desorbing from the walls of the cylinder often become significant. For instance, if adsorbed moisture is not adequately removed from the cylinder walls before filling, water vapor will desorb in the vapor phase until the partial pressure in the cylinder reaches equilibrium with the walls. The ratio of this equilibrium partial pressure of water vapor to the pressure of the ESG in the cylinder then determines the moisture impurity of the withdrawn gas. As the product is used and the total pressure in the cylinder decreases, the relative concentration of moisture desorbed from the walls steadily increases.
An interesting example of an intermediate case is C2F6, which has a critical temperature and pressure of approximately 67?F and 432 psia. A cylinder filled with C2F6 may thus have either one or two phases present, depending on the ambient temperature. A small change in the storage temperature can dramatically affect whether trace impurities are delivered with the gas or retained in the cylinder. Supercritical phases can have dissolving powers that resemble those of a liquid more than a gas [12]. Consequently, gas delivery will include transport of species that are commonly present in the cylinder and other gas components, but not normally considered as possible impurities, such as lubricants or high-molecular-weight oils. Other commonly used specialty gases with readily accessible supercritical phases include CHF3 (78?F, 697 psia), N2O (98?F, 1050 psia), SF6 (114?F, 545 psia) and HCl (124?F, 1200 psia).
 |
Figure 2. Measured water concentration in flowing HCl with time from two cylinders filled from the same HCl source. One cylinder had a clean valve, while the other had a noticeably corroded valve.
Effects of moisture on corrosion. The moisture impurity level of an ESG can increase if any part of a corrosive gas system, such as the cylinder valve or pigtail, that is routinely exposed to ambient air should become corroded. The products from the corrosion of steel by HCl, e.g., FeCl3, will adsorb copious quantities of moisture when in equilibrium with ambient humid air. When used to deliver the relatively dry corrosive gas, these hydrated, corroded components will slowly desorb detectable amounts of moisture into the gas stream (Fig. 2) [13].
In this experiment, two HCl cylinders with SS valves were filled from the same HCl source. One cylinder had a clean valve, while the other had a noticeably corroded valve. The moisture concentration of the HCl streams from the two cylinders was tracked with time. The HCl from the clean valve contained <1 ppm (the limit of detection of the analysis) moisture after 10 min of flow, whereas the HCl from the corroded valve dried down more slowly, containing over 3 ppm of moisture after 4 hr of gas flow.
Since avoiding corrosion within the cylinder valve and the gas distribution system is critical to the consistent delivery of pure gas to the point of use, proper procedures are essential to purge moisture from the system thoroughly before delivering corrosive ESG. Also, corrosive gases must be purged before opening the system to the atmosphere for maintenance. Failure to do so may lead to mixtures of the corrosive gas with water vapor that can condense as a liquid film and initiate corrosion. While the exact purging procedure depends on the pressure ranges, the layout of the gas distribution system, and the identity of the ESG, multiple cycles of evacuation and repressurization with a dry, inert gas are recommended.
The amount of time needed to purge a gas distribution system may be quite significant and should thus be minimized. We found that the dry-down to a given moisture level in the gas distribution system for a plasma-etch tool could be optimized by increasing the number of evacuation
epressurization cycles at the expense of shorter evacuation time [14].
Flow restrictions in a gas distribution system may inhibit moisture removal by cycles of evacuation
epressurization purging. One extreme example involves the flow-restricting orifices that are sometimes incorporated into corrosive gas systems as a safety measure to limit the consequences of an uncontrolled release. Even orifices with diameters as small as 0.006 in. (0.15 mm) installed in high-purity cylinder valves do not significantly limit the ability to remove trace moisture contamination by cycle purging. This result probably indicates that desorption from the wetted surfaces is dominating the moisture removal rate.
Thermally unstable diborane. Diborane (B2H6) is an exceptional ESG that does not degrade with time in a properly specified and handled cylinder. It is thermally unstable and decomposes slowly at room temperature, forming higher-order boron hydrides, nonvolatile solids that can plug the regulator or other small orifices of the gas distribution system.
Studies of diborane decomposition kinetics [15] showed that the decomposition rate constant was independent of the initial concentration of diborane and the diluent in diborane mixtures. Using the reported decomposition rate, a typical 5% diborane in nitrogen or hydrogen mixture would contain 4.1% diborane (a loss of 18%) after storage at 20?C for 1 year, whereas storage for 1 year at 4?C would retain a diborane concentration of 4.85%, a loss of just 2.9%. Under practical handling and storage conditions, a shelf life of 6 months for diborane should insure less than 2% decomposition before expiration.
Conclusion
While the purity of an ESG does not change with time when stored in a properly prepared cylinder, with the exception of the thermally unstable B2H6, improper materials of construction, preparation, or handling techniques can add contaminants. It is essential for gas suppliers to provide the ESG in a properly prepared cylinder made of the best material to preserve the purity of the ESG. At the same time, knowledge of the underlying chemistries can assist the gas user in delivery system designs to preserve the purity of the ESG to the point of use. Proper gas handling techniques are important to prevent degradation of corrosive gases. Avoiding system corrosion requires removal of the corrosive gas prior to exposing the system to the atmosphere and removal of residual atmospheric moisture prior to delivery of the corrosive gas.n
References
1. S.M. Fine, P.B. Henderson, "What Is the Shelf-Life of Electronics Specialty Gases?" Proceedings for the 42nd Annual Technical Meeting of the Institute of Environmental Sciences, Orlando, FL, May 12-16, 1996.
2. D.A. Bohling, M.A. George, "Controlling Contamination in Tungsten Hexafluoride Applications," Semiconductor International, Vol. 14, pp.104-106, 1991.
3. V. Houlding, R. Doane, G. Johnson, B. Streusand, "Production and Analysis of High-Purity Tungsten Hexafluoride for Microelectronics,"Nippon Sanso Giho, Vol. 14, pp.77-82, 1995.
4. Air Products and Chemicals Safety Bulletin Update, February 8, 1994.
5. T. Ohmi, "Very Stable Gas Supply System without Corrosion and Catalytic Behavior," SEMI Technical Programs Present: Gas Distribution Systems Workshop, SEMI, Mountain View, CA, pp. 6-49, 1996.
6. T. Ohmi, "Corrosion-free Cr2O3 Passivated Gas Tubing System for Specialty Gases," Solid State Technology, pp. S18-22, October 1995.
7. Y. Shirai, S-K. Lee, S. Miyoshi, T. Ohmi, "The Evaluation of Thermal Decomposition Characteristics of Active Specialty Gases on Various Metal Surfaces Using FT-IR Method," Proc. - Inst. Env. Sci., 41st (contamination control), pp. 17-21, 1995.
8. B.E. Wilde, "Stress-Corrosion Cracking," ASM Handbook, Vol. 11, pp. 203-224, 1986.
9. S. Lau, R. Torres, "Performance of Diaphragm Valves in Chlorine," CleanRooms `96 West, Session 602, pp. 1-5, 1996.
10. L. Iler, W.J. Koros, H.B. Hopfenberg, "Analysis of Coupled Diffusion and Chemical Reaction of Ammonia in Kapton Polyimide at 30?C," Polym. Prepr., Vol. 24, pp. 102-103, 1983.
11. J. Hart, A. Paterson, "Evaluating the Particle and Outgassing Performance of High-Purity, Electronic-Grade Specialty Gas Cylinders," Microcontamination, pp. 63-67, July 1994.
12. M.L. Lee, K.E. Markides, eds., "Analytical Supercritical Fluid Chromatography and Extraction," Chromatography Conferences Inc., Provo, Utah, 1990.
13. P.B. Henderson, "HCl Manufacture, Purification and Analysis," SEMICON/West 1994: Forum on Process Gases - Silicon Epitaxial Deposition, SEMI, Mountain View, CA, pp. 37-53, 1994.
14. S. M. Fine et al., "Optimizing the UHP Gas Distribution System for a Plasma Etch Tool," Solid State Technology, pp. 71-81, March 1996.
15. E.T. Flaherty et al., Contamination Control and Defect Reduction in Semiconductor Manufacturing III, ed. D. N. Schmidt, The Electrochemical Society, Pennington, NJ, 94-9, p. 371, 1994.
PHILIP B. HENDERSON received his undergraduate degree in chemistry from the University of California, Berkeley, in 1983, and his PhD degree
in organic chemistry from the University of
Southern California in 1987. He is a lead research chemist in the Fluorine Technology Center at Air Products and Chemicals Inc., 7201 Hamilton Blvd., Allentown, PA 18195-1501; ph 610/481-6262, fax 610/481-6517.
RONALD M. PEARLSTEIN received his undergraduate degrees in chemical engineering and chemistry from the University of Pennsylvania in 1984, and his PhD degree in inorganic chemistry from
the Massachusetts Institute of Technology in 1988. He is a principal research chemist for the Electronics Division of Air Products and Chemicals Inc., where he conducts research on materials compatibility and microcontamination control for specialty gases.