Ralph Ciotti, Seksan Dheandhanoo, James Yang, and David Yesenofski
Air Products and Chemicals, Inc.
7201 Hamilton Boulevard, Allentown, PA 18195-1501
BIOGRAPHY
Ralph Ciotti received his B.A. in chemistry from Cornell University and is a Research Technician in the Electronics Customer Applications group. Seksan Dheandhanoo received his Ph.D. in physics from the University of Pittsburgh and is a Senior Principal Research Scientist in the Electronics Customer Applications group. James Yang received his Ph.D. in chemical engineering from the University of Minnesota and is a Senior Research Engineer in the Electronics UHP Process Technology group. David Yesenofski is a Principal Research Technician in the Electronics Analytical Technology group.
ABSTRACT
Moisture is the most difficult impurity to remove from ultra-high purity gas distribution systems because of its strong adsorption onto piping surfaces. This paper compares experimental moisture drydown data from tubing and gas distribution systems with a model developed at Air Products. The model takes into account the fluid dynamics of the gas stream and incorporates two drydown regimes: (1) drydown when adsorbed moisture is physisorbed on top of other layers of moisture and is in equilibrium with the moisture in the boundary layer of the gas phase and (2) drydown when adsorbed moisture is chemisorbed on the piping surface and not in equilibrium with the moisture in the gas phase. Experimental drydown data from tubing and gas distribution systems as well as the model fits to the data are shown. The verified model can be applied as an engineering tool for designing a practical and effective gas distribution system. The model can also predict the effect of a moisture intrusion on the system. The longer the intrusion, the greater the maximum concentration in the system.
In addition, back diffusion of various contaminants such as H2O, H2, and O2 is briefly discussed. Both theoretical and experimental studies indicate that back diffusion is noticeable only at very low flow rates and at distances very close to the source of contamination.
DATA:
Drydown Experiments—The semiconductor industry utilizes gases containing very low amounts of moisture. The moisture requirement is so low that residual moisture in gas delivery systems must be eliminated. This moisture can be removed by purging the gas delivery system with an ultrahigh purity (UHP) gas such as nitrogen or argon.
Air Products has vast experience in drying down systems of various designs and materials. In most cases, regardless of the design of the piping system or materials of the piping system, drydown proceeds to the moisture level of the purge gas. In addition, advancements in filter technology have reduced drydown times dramatically (moisture loading in large surface area filters is usually the dominant source of moisture in a new gas distribution system). However, the drydown time, or response of the system to a moisture upset, still varies with changes in piping materials and piping design.
Air Products has simulated moisture removal behavior in gas delivery systems by developing the user-friendly DRYCOM™ model. The model incorporates the fluid dynamics of the gas stream, boundary layer resistances in the gas phase, and two drydown regimes [1]: (1) a regime where adsorbed moisture is physisorbed on top of other layers of moisture and is in equilibrium with the moisture in the boundary layer of the gas phase and (2) a regime where the adsorbed moisture is chemisorbed on the piping surface and not in equilibrium with the moisture in the gas phase. Further details have been discussed elsewhere [2]. DRYCOM can be run on a PC and can calculate drydown simulations of gas distribution systems in seconds. Such a model can easily evaluate current designs and generate optimum designs of gas distribution systems.
Experiments to confirm the model were conducted in our R&D facilities in Trexlertown, PA. The first experiment was drydown of 23 feet of 0.5-inch O.D., 10Ra electropolished 316 stainless steel tubing equilibrated with 115 ppb of moisture in N2. The drydown was accomplished by flowing 4.2 SCFH of UHP N2 (0.5 ppb moisture). The whole procedure was run at 87 psig.
The experimental apparatus was designed to switch the inlet gas from a known moisture standard to UHP N2 at the start of the experiment. Moisture standards were generated with a permeation source diluted with UHP N2. Atmospheric pressure ionization mass spectrometry (APIMS) was used to measure the moisture level at the outlet of the tube. The drydown data were compared with the drydown of the APIMS alone to insure that instrument drydown did not dominate the drydown profile of the tubing. Figure 1 shows that the experimental drydown profile at the end of the tube is well fitted by the DRYCOM model.
Figure 1. Comparison of DRYCOM model with experimental data for
0.5-in. O.D., 23-ft. tube of EP 316 stainless steel.
The second experiment used to confirm the model was drydown of a small gas distribution system. A diagram of this system that gives the flows through each of the legs is shown in Figure 2.
Figure 2. Diagram of drydown of 10Ra gas distribution system.
The system, comprised of 0.5-in. O.D., 10Ra electropolished 316 stainless steel tubing, was first equilibrated with 110 ppb of moisture in N2. The drydown was accomplished by flowing 4.9 SCFH of N2 at 0.5 ppb moisture into the system. Based on the flow splits in Figure 2, the middle 6 ft. of the main saw 4.4 SCFH of flow, while the last 12 ft. of the main was exposed to 3.5 SCFH of flow. The whole procedure was again run at 87 psig. DRYCOM can simulate this case in seconds. Figure 3 shows that the experimental drydown profile at the end of the main in the gas distribution system is also well fitted by the DRYCOM model, although the model lags the data somewhat at 20 minutes.
Figure 3. Comparison of model with experimental data for
0.5-in. O.D., EP 316 stainless steel gas distribution system.
Moisture Intrusion—The DRYCOM model can also predict the effect of moisture intrusions on the system. Figure 4 diagrams a hypothetical gas distribution system comprised of 1.0-in. O.D., 10Ra electropolished stainless steel tubing equilibrated at 0.5 ppb moisture. In this simulation, a moisture intrusion of 100 ppb moisture in N2 was then introduced to the system for 3 different time periods: 10 seconds, 60 seconds, and 300 seconds. The simulation began when the 100 ppb moisture was introduced. After each intrusion, 0.5 ppb moisture in N2 was again fed to the system.
Figure 4. Diagram of hypothetical 10Ra gas distribution system devised for
moisture intrusion simulations.
Figure 5 shows the predicted effect of the various moisture intrusions on the moisture concentration at the outlet of the main. The longer the intrusion, the greater the maximum concentration at the end of the main. The short intrusion of 10 seconds barely affects the moisture concentration at the end of the main, while the longest intrusion of 300 seconds dramatically increases the moisture concentration at the end of the main.
Figure 5. DRYCOM simulations of moisture intrusions at various times for
the gas distribution system described in Figure 4.
Back Diffusion—Besides contamination from residual or intruding moisture, back diffusion of contaminants such as H2O, H2, and O2 is an industry concern. Work at Air Products previously investigated moisture back diffusion for the following case [3]:
Tubing: 0.25-in. O.D. of EP 316 stainless
Moisture level at tube end: 0.13 to 200 ppm
Flow rate: 500 to 5500 scc/min
Pressure: 15 to 40 psig
Temperature: ambient
Moisture was monitored with a DuPont 5700 moisture analyzer (limit of detection = 100 ppb) 30 cm upstream of the moisture source. After steady state was reached, no moisture was seen in the gas phase, implying no gas phase back diffusion. To see if moisture would migrate along the tube surface, the experiment was extended for 2 weeks; however, no moisture was seen 30 cm upstream of the source for the duration of the experiment. This results support the hypothesis that back diffusion of low concentrations of moisture along tubing surfaces is small.
Because of its high gas diffusivity, H2 is the most likely species to contaminate upstream lines because of back diffusion. Assuming no surface diffusion of H2, the concentration of H2 in a steady, one-dimensional flow of gas having velocity U is described by
U(dC/dx) = D(d2C/dx2) (1)
where D is the diffusion coefficient, which is constant for a given condition, and x is the distance upstream of the H2 contaminant. This equation can be integrated to yield:
(C/C0) = exp(-Ux/D) (2)
where C0 is the H2 concentration at the source.
In one of Verma's experiments [4], back diffusion of O2 in a 0.5-in. O.D. tube of electropolished 316 stainless steel was monitored 15 cm upstream of an O2 source. The system pressure was 3 psig. Figure 6 includes Verma's O2 experimental results obtained for various N2 flow rates and our theoretical prediction for H2 back diffusion under the same conditions. The ratio of contaminant concentration 15 cm upstream of the source to the contaminant concentration at the source is plotted versus the nitrogen flow rate.
Figure 6. Experimental O2 back diffusion data from Verma, et al.
and prediction of H2 back diffusion 15 cm upstream of a contaminant source
for 0.5-in. O.D., EP 316 stainless steel at P=3 psig. Also shown is H2 back diffusion
under same conditions 100 cm upstream of source.
Both contaminants do exhibit back diffusion even when the flowrate is greater than zero. As expected, H2 is more likely to contaminate upstream tubing than O2. However, the graph indicates that a flow rate as small as 20 scc/min in 0.5-in O.D. tubing reduces the H2 contaminant concentration by 4 orders of magnitude 15 cm upstream of the source. At higher flow rates and for distances further upstream, the effect of the contaminant will be even less pronounced, as shown by the dashed line in the figure, a plot of the ratio of the H2 concentration 100 cm upstream of the source to the H2 source concentration as a function of flow rate.
CONCLUSION:
Air Products has simulated moisture removal behavior in gas delivery systems by developing the user-friendly DRYCOM model. The model was shown to fit the drydown data for single tubing and a small gas distribution system. The model can also predict the effect of a moisture intrusion on the system. The longer the intrusion, the greater the maximum concentration in the system. It was also shown that back diffusion of H2O, H2, and O2 can be minimized with sufficient flow through a system.
REFERENCES:
- R. Desai, M. Hussain, and D. M. Ruthven, "Adsorption of Water Vapour on Activated Alumina. I - Equilibrium Behavior," The Canadian Journal of Chemical Engineering, Vol. 70, 699-706, 1992.
- D. G. Coronell, A. D. Johnson, M. S. K. Chen, S. N. Ketkar, D. A. Zatko, and J. V. Martinez, "An Integrated Approach to Understanding Moisture Behavior in UHP Gas Delivery Systems: Component Testing and Computer Simulations," IES Annual Technical Meeting Proceedings, 237-245.
- W. T. McDermott, internal memo, 14 March 1988.
- N. K. Verma, A .M. Haider, and F. Shadman, "Computer Simulation of Gaseous Impurity Distribution due to Back-diffusion in Ultra-pure Systems," 1993 Microcontamination Conference Proceedings, 201-211.
ACKNOWLEDGMENTS:
The authors thank Steve Fluharty and Scott Tomkinson for their development of the interface and Alex Schwarz and Andy Homyak for their technical advice. In addition, we thank Professor Farhang Shadman for permission to use the O2 back diffusion data in Reference [4]. |
|
