Ind. Eng. Chem. Res. 1993,32, 3006-3013
3006
MATERIALS AND INTERFACES Preparation and Characterization of a Composite Palladium-Ceramic Membrane John P.Collins+and J. Douglas Way' Department of Chemical Engineering, Oregon State University, Corvallis, Oregon 97331 -2702
Composite palladium-ceramic membranes with palladium films ranging from 11.4 to 20 pm were made by depositing palladium on the inside surface of asymmetric tubular ceramic membranes. Electroless plating was used to deposit the palladium film. Membranes were characterized by conducting permeability experiments with hydrogen, nitrogen, and helium at temperatures from 723 t o 913 K and feed pressures from 160 to 2445 kPa. The membranes had both a high hydrogen permeability and selectivity. The hydrogen permeability for a composite membrane with an 11.4mol-m/(m2.s.Pa0.602)at 823 K. Hydrogen/nitrogen selectivity for pm palladium film was 3.23 X this membrane was 380 a t 823 K a t a transmembrane pressure difference of 1500kPa. Improvemenb in membrane sealing should further increase the hydrogen selectivity. Results of this study demonstrate the potential for using composite metal-ceramic membranes i s membrane reactors including applications that require operation a t relatively high temperatures and transmembrane pressure differences.
Introduction There is increasing interest in inorganic membranes as a means of separating gas mixtures at high temperatures. One of their most promising applications is in membrane reactors where chemical conversion and product purification by separation take place in the same device. By selectively separating one or more reaction products through the membrane wall, it is possible to achieve significant enhancement over the equilibrium conversion of the reactor feed stream. Review articles by Hsieh (1991), Armor (1989),and Shuet al. (1991)discuss basic principles and potential applications for inorganic membrane reactors. Some of the many potential applications include steam reforming of methane, water gas shift, and dehydrogenation of various hydrocarbons including cyclohexane, ethylbenzene, ethane, propane, and butane. Another promising application is the decomposition of hydrogen sulfide and ammonia impurities in synthesis gas produced at coal gasification plants (Edlund, 1992;Cicero and Jarr, 1990; Collins et al., 1993). Since these reactions produce hydrogen, improvements in reactor performance are obtainable when using a membrane with high hydrogen permeability and selectivity. The two basic types of inorganic membranes are dense membranes and porous membranes. High-temperature microporous ceramic membranes are commercially available in a variety of design configurations including single tubes and multichannel monolithic membranes (Hsieh, 1991). A high permeability is achieved with an asymmetric design consisting of a thick macroporous support structure and a thin multilayer composite membrane. Pore diameters as low as 4 nm on the top (selective) layer are ~
~~~
* To whom
correspondence should be addressed. Current address: Department of Chemical Engineering and Petroleum Refining, Colorado School of Mines, Golden, CO 80401-1887. FAX (303)273-3730.E-mail:
[email protected]. + Current address: Department of Chemical Engineering, University of Colorado, Boulder, CO 80309-0424.
0888-5885/93/2632-3006$04.00/0
commercially available. These membranes are thermally and mechanically stable, chemically resistant, and highly permeable, but their selectivity is low since gas transport through a 4-nm pore layer occurs primarily by Knudsen diffusion at high temperature. Developmental silica microporous membranes with higher gas selectivities were reported by Way and Roberb (1992). Metal oxide membranes and metal membranes are two types of dense membranes suitable for high-temperature applications. One recent development in the field of metal oxide membranes is the deposition of metal oxides in the pores of microporous membranes. Tsapatis et al. (1991) prepared hydrogen selective metal oxide membranes by chemical vapor deposition of Si02, TiO2, A1203, and B203 layers within the pores of Vycor tubes. These membranes had hydrogen to nitrogen selectivities of up to 1000-5O00, but a relatively low hydrogen permeability compared to hydrogen permeable metal membranes. Palladium and its alloys, nickel, platinum and the metals in groups 3-5 of the peroidic table are all permeable to hydrogen (Barrer, 1941). Hydrogen-permeable metal membranes made of palladium and ita alloys are the most widely studied due to their high hydrogen permeability, chemicalcompatibility with many hydrocarbon-containing gas streams, and infinite hydrogen selectivity. Several metals in groups 3-5 of the periodic table have a higher hydrogen permeability than palladium, but they are not compatible with many feed streams due to their high chemical reactivity (Edlund, 1992). Commercially available metal membranes are either thick films or thick-walled tubes of palladium or palladium alloys. Recent research efforts are focused on developing composite metal membranes consisting of relatively thin palladium or platinum coatings on hydrogen-permeable base metals (Edlund, 1992;Buxbaum and Marker, 1993),and composite metalmicroporous membranes (Uemiya et al., 1988). The composite metal membranes consist of thin (2-25 pm) metal coatings of palladium or platinum on a base metal such as vanadium, niobium, or tantalum (Edlund, 0 1993 American Chemical Society
Ind. Eng. Chem. Res., Vol. 32, No. 12, 1993 3007 1992; Buxbaum and Marker, 1993). The metal coatings are chosen for their chemical compatibility with feed gas constituents, and the base metal is chosen for its high hydrogen permeability, mechanical stability, and relatively low cost. Buxbaum and Marker (1993)tested membranes with 1-2-pm palladium coatings on 2-mm-thick disks of niobium and tantalum and 0.25-mm-thick niobium tubes at temperatures up to 698 K. The hydrogen permeabilities of these membranes were between 1 and 2 orders of magnitude higher than those for palladium membranes of the same total thickness. The durability of these membranes at high temperatures is a problem since intermetallic diffusion between the different metal layers reduces membrane permeability. Buxbaum estimates that membrane durability is a manageable issue for membrane operation below about 823 K (Buxbaum and Marker, 1993; Hsu and Buxbaum, 1986). Edlund and co-workers are developing a composite metal membrane suitable for use at higher temperatures. Membrane durability at high temperature is obtained by placing a barrier to intermetallic diffusion between the metal coatings and the base metal. A constant hydrogen flux was obtained over a 7-day thermal stability test at 973 K (Edlund, 1992). In contrast, the flux from a composite metal membrane with no diffusion barrier was essentially eliminated after 2 days at 973 K (Edlund, 1992). A membrane with a 25-pm platinum coating showed no degradation in performance after 8 h exposure to 793 kPa of pure hydrogen sulfide at 973 K (Edlund, 1992). The platinum-coated membrane was used in a membrane reactor for thermolysis of hydrogen sulfide (Edlund and Pledger, 1993). Another way to obtain a membrane with high hydrogen permeability and selectivity is to coat the selective layer of a microporous membrane with a thin film of hydrogen permeable metal. The result is a composite metalmicroporous membrane. The microporous membrane provides the mechanical support required for the thin metal film. Uemiya and co-workersused electrolessplating to deposit palladium films ranging from 13 to 20 pm on the outside surface of porous glass tubes with 0.3-pm pores (Uemiya et al., 1988, 1991a-c). They report an infinite hydrogen selectivity, which means a defect-free palladium layer was deposited. These membranes were tested in membrane reactor experiments for steam reforming of methane and the water gas shift reaction. Uemiya and co-workers (Uemiya et al., 1990) also made a composite palladium-ceramic membrane for use in a membrane reactor for aromatization of propane by depositing an 8.6pm layer of palladium on the outside surface of a porous alumina cylinder. The hydrogen selectivity for this membrane was not reported. The objective of this paper is to discuss the preparation and characterization of a composite palladium-microporous ceramic membrane tube. Ceramic membranes are better supports than porous glass membranes due to their superior thermal and mechanical properties. This allows for operation at higher temperatures and transmembrane pressure differences. The composite membrane is prepared by plating a thin palladium film on the selective layer of a commercially available ceramic membrane tube. The palladium film is deposited on the inside surface of the membrane tubes since the selective membrane layer resides on the inside surface. The membrane is characterized by conducting hydrogen permeation experiments at higher temperatures and over a much wider pressure range than the composite palladium-microporous glass membranes discussed above. This demonstrates the potential for using the composite metal-microporous
ceramic membrane design in membrane reactor applications involving high temperature and pressures such as those encountered at coal gasification plants.
Experimental Section Experiments are divided into two parts: membrane preparation and membrane characterization by means of permeability studies. Membrane Preparation. The ceramic membrane supports were Membralox T-170 alumina membranes obtained from U.S. Filter Corporation in Warrendale, PA. The alumina membranes consist of a macroporous support tube with an inner surface covered by a thin multiple layer microporous membrane. Whenever the pore diameter is mentioned in this paper, we are referring to the pore diameter of the top layer (selective layer). Membranes with pore diameters of 10-200 nm were used in the study. The inside and outside diameters of the membrane tubes were 0.7 and 1 cm, respectively. Electroless plating was used to deposit palladium on the selective membrane layer. The membranes went through several pretreatment steps and a surface activation step prior to electroless plating. Ceramic membrane tubes obtained in lengths of 25 cm were first cut to the desired length of 6 cm using a diamond saw. The 6-cm length was convenient for both the electroless plating operation and the permeation experiments. The outside diameter was reduced by gentle sanding so the membrane would fit into the compression fittings used for gas sealing in the permeation experiments. Tubes were then cleaned by ultrasonic rinsing in an alkaline solution, deionized water, and isopropyl alcohol to remove sanding grit and any contaminants present from the membrane fabrication process. Finally, a high-temperature sealant (Aremco617 from Aremco Products, Inc., Ossining, NY) was used to seal the ends of the membrane. These end seals were needed to prevent bypassing of gas through the porous support at the membrane inlet and outlet. The purpose of the surface activation step is to seed the inner membrane surface with finely divided palladium nuclei which initiate the plating process. Surface activation is a key step because defects in the palladium film usually occurred when the membrane surface was not uniformly activated prior to electroless plating. The activation procedure consisted of a two-step immersion sequence in an acidic SnC12/SnC& (sensitizing) bath followed by an acidic PdC12 bath with gentle rinsing in deionized water between baths. The immersion sequence was generally repeated 3-7 times to obtain a uniformly activated membrane surface. The membrane surface was activated when it had a grayish brown appearance. Teflon tape was wrapped around the membrane tube to prevent activation of the outside membrane surface. The sensitizing bath was the key step in the surface activation process. We used a bath recipe developed by Feldstein and Weiner (1972). The electroless plating bath consisted of a palladiumammine complex, the disodium salt of ethylenediaminetetraacetic acid (disodium salt of EDTA) as a stabilizer, and hydrazine as the reducing agent. The bath formulation is based on a recipe developed by Rhoda (1959) except a higher concentration of EDTA was used to improve bath stability. The resulting palladium film is pure and not alloyed (Rhoda, 1959). Table I shows typical plating bath composition and conditions. The membrane was wrapped with Teflon tape before plating to protect the end seals. The bath container was a 30-cm3glass vial. No agitation other than bubbles produced from the plating reaction
3008 Ind. Eng. Chem. Res., Vol. 32, No. 12, 1993 Table I. Typical Electroless Plating Bath Composition
component palladium chloride concn ammonium hydroxide (28 76 ) concn disodium EDTA concn hydrazine (1 M solution) concn PH temperature plating surface area
5.4 glL 390 mL/L 70 g/L 10 mL/L 11 343-353 K 527 cm2/L
and occasional shaking of the bath container was used. After each hour, the membrane was removed from the bath, rinsed with deionized water, and wrapped with fresh Teflon tape. It was then added to a fresh bath and the plating process was continued until a desired palladium film thickness was obtained. The 1-h plating period was used because the plating rate was significantly reduced after 1h. This is attributed to catalytic decomposition of hydrazine by palladium (Rhoda, 1959). Rinsing the membrane between plating baths as well as adding the EDTA a few hours before each bath was used helped maintain bath stability at the relatively high plating temperatures of 343-353 K. Palladium was plated at a rate of 2-2.5 pm/h under these conditions. Membranes were rinsed in deionized water and dried at 383-393 K for several hours after plating. The thickness of the palladium film was calculated by dividing the weight difference in the unplated and plated membranes by the plated surface area and the palladium density. Plating thickness for a membrane with a 20-pm palladium film was independently determined by weighing the film after it was removed from the membrane, and by taking a scanning electron microscope (SEM) micrograph of the film. The three measurements agreed within 5 5%. Membrane Characterization. Membranes were characterized by conducting permeability experiments with hydrogen, nitrogen, and helium. Figure 1shows a schematic of the experimental apparatus. A more detailed drawing of the membrane module which consisted of a shell and tube design is shown in Figure 2. Feed gas flowed down the inside of the membrane tube (tube side), and permeated gases were collected on the shell side. A sweep gas could also be used on the shell side, if desired. The membranes were connected to inlet and outlet tubes using stainless steel Swaglok compression fittings with seals made from Grafoil tape (graphite-based packing material). These Grafoil seals were used inside the heated zone of the furnace. Brooks 58503mass flow controllers were used to regulate the inlet gas flows and compositions. An Omega CN9000 temperature controller was used to regulate the membrane temperature. Type K thermocouples placed at the membrane inlet and outlet measured the temperature. The measured inlet temperature was higher than the outlet temperature due to a nonuniform temperature distribution in the furnace. The maximum difference between the inlet and outlet temperatures was 20 K. Membrane temperature was taken as the average of the inlet and outlet temperatures. Membrane pressure was controlled by a back-pressure regulator installed on the residue line. Pressure measurements were made with pressure gages. Residue and permeate stream flows were measured with bubble flow meters. A Hewlett Packard 5890 Series I1 gas chromatograph with thermal conductivity detectors measured residue and permeate gas compositions in experilaenta involving gas mixtures. Permeability experiments were conducted over a range of temperatures and pressures. Experimental temperatures ranged from 723 to 913 K. The upper temperature limit was set by the maximum operating temperature of
the Aremco 617 end seals. Feed pressures ranged from 156to 2445 kPa, and permeate pressures ranged from 101 to 140kPa. Transmembrane pressure differences ranged from about 40 to 2330 kPa. Operation at even higher transmembrane pressure differences appears possible with the ceramic membrane usedto support the palladium film. Wu and co-workers (1993) reported using a 4-nm Membralox membrane at transmembrane pressure differences up to 3575 kPa. During startup, membranes were heated to 723 K under a helium atmosphere to avoid hydrogen embrittlement and possible pinhole formation in the palladium film due to heating in a hydrogen atmosphere at temperatures below the critical temperature of the palladium-hydrogen system (about 573 K). Hydrogen and helium permeabilities were determined by flowing pure gas through the membrane at various pressures and measuring the permeate flow rate with the bubble flow meters. The permeate side pressure was slightly above atmospheric in all experiments. No inlet sweep gas was used in the hydrogen and helium permeation experiments or in experiments conducted with gas mixtures of hydrogen, nitrogen, and helium. A helium sweep gas was generally used in the nitrogen permeability experiments. Nitrogen permeabilities were determined by measuring the effluent sweep gas flow rate with the bubble flow meters and composition on the gas chromatograph. Theoretically, membranes with an infinite hydrogen selectivity are possible when a defect-free palladium film is deposited. In practice, tiny leaks in the end seals or Grafoil seals or defects in the palladium film will reduce membrane selectivity. Results and Discussion Figure 3 is an SEM micrograph of a composite palladium-ceramic membrane. The ceramic membrane had a 10-nmpore top layer. The palladium film is shown on the lower horizontal band of the SEM micrograph. The slight delamination of the palladium film occurred when the membrane was cut with a jeweler’s saw to take the SEM micrograph. Palladium film thickness on this membrane was about 1.5 pm. Membranes with palladium films ranging from 11.4 to 20 pm were tested in the high-temperature permeability tests. Ambient temperature leak tests were conducted by immersing membranes pressurized with nitrogen to about 240 kPa in water. Membranes with palladium films less than 10pm leaked nitrogen, so they were not tested in the high-temperature experiments. Tiny leaks from the Grafoil sealswere observed on all membranes. Seal leakage was reduced but not eliminated with practice and experience. We attribute the major portion of the measured nitrogen and helium permeabilities to leaks in the Grafoil seals rather than defects in the palladium film. Therefore, even higher hydrogen selectivities than those measured inthis work are possible with improvements in membrane sealing. Hydrogen must permeate through both the palladium metal film and the ceramic membraneaupport. Hydrogen fluxes for a 4-nm Membralox membrane were estimated using permeability data reported by Wu and co-workers (Wu et al., 1993). The estimated hydrogen fluxes at transmembrane pressure differences from 100to 2000 kPa are 23-44 times higher than the fluxes measured for the compositepalladium ceramic membranes. Therefore, we assume the mass transfer resistance of the ceramic support is minimal and the measured hydrogen permeabilities are essentially the permeabilities of the palladium film.
Ind. Eng. Chem. Res., Vol. 32, No. 12,1993 SO09
MFC Mu,now CDnMla (normal flow capacity)
ME*
w
now Controlla (blah flow up.snly) UP"
(hl0ff"alVc Nccdlevalvc
$(
Solaaidvalve
w
Relidvalve
y
Rc=g.ge
a
>wqv.lVc
00
Rcm*"lllor
Figure 1. Gaa flow system for permeability experiments.
comprmion mi
1.54 c n OD .lW"i"l
.:: -..".
I-
Figure 3. SEM micrograph of a palladium film deposited on an asymmetrieceramiemembranewith 10-nmtopporelayer. The third horizontal hand from the top of the image is the palladium film.
bulk layers of the metal, (3)diffusion of atomic hydrogen through the bulk metal. Steps 1and 2 take place on both surfaces of the metal. The equation for the hydrogen flux is written in terms of Fick's first law as follows (Lewis, 1967; Uemiya et al., 1991a):
J = (PH/t,)(PHt - Pw")
iiy Temperature Controller
Figure 2. Membrane module for permeation experiments.
Hydrogen permeates through metals by a multistep process, which involves the following steps (Barrer, 1941; Shuet al.,1991): (1)reversible dissociativechemisorption of molecular hydrogen on the membrane surface; (2) reversible dissolution of surface atomic hydrogen in the
(1)
When diffusion through the hulk metal is the rate-limiting step and hydrogen atoms form an ideal solution in the metal (Sievert's law hydrogen solubility dependence), n is equal to 0.5 (Hurlbert and Konecny, 1961). The hydrogen flux (J)is inverselyproportional to the palladium film thickness (t.) when bulk diffusion is the rate-limiting step. Avalueof ngreaterthan0.5mayresultwhensurface processes influence the permeation rate or when Sievert's lawisnotfollowed. Dependenceofthe hydrogendiffusion coefficienton the concentration of dissolved hydrogen has also been proposed as the reason for n values greater than 0.5 (Uemiya et al., 1991a). Leakage of hydrogen through
3010 Ind. Eng. Chem. Res., Vol. 32, No. 12, 1993 Table 11. Hydrogen Permeabilities at Specific Temperatures for Composite PalladiumCeramic Membranes membrane description 823 1.43 X 108 20-pm palladium film on ceramic membrane with 10-nm pore layer 723 2.34 X 17-pmpalladium film on 4.04 X le9 ceramic membrane with 773 823 6.82 X 1P 200-nm pore layer 873 9.96 X 1O-g 3.23 X 10-8 11.4-pm palladium film on 823 5.84 x 10-9 ceramic membrane with 873 200-nm pore layer ~~~
. 1
~
~~~
0.526
T
0
T r=7773 73K
A
T = 823 K
I
0.622 0.595 0.568 0.552 0.602 0.566
Hydrogen permeabilities calculated from nonlinear regression analysis of eq 1using permeation data collected at each temperature.
defects in the metal film or membrane seals would also increase the value of n. The small transport resistance of the ceramic substrate may also slightly increase the value of n. Hydrogen permeation experiments conducted with palladium have yielded n values of 0.68 (Hurlbert and Konecny, 1961), 0.76 (Uemiya et al., 1991a), and 0.8 (DeRosset, 1960). These experiments were conducted at temperatures ranging from 616 to 727 K and hydrogen feed pressures ranging from 101 to 4926 kPa. The permeability coefficient (PH) depends on temperature in the following manner (Barrer, 1941):
P H = PHo exp(-EIRT)
723 K
0
(2)
This equation assumes the value of n in eq 1does not vary with temperature. In fact, n may also depend on temperature since it is influenced by solubility and the relative rates of surface processes and bulk diffusion, which all depend on temperature. Results of Permeability Experiments. Table I1 summarizes values of PH and n obtained for three membranes evaluated in this study. The pH and n values were determined from nonlinear regression analysis of eq 1. The 'nvalues decreased with increasing temperature. The membrane with the 17-pm palladium film had significantly higher nitrogen fluxes than the other two membranes, which means that significantlymore hydrogen permeated through leaks in the Grafoil seals or defects in the palladium film. The hydrogen leakage increased the values of n obtained for this membrane. The membrane with the 20-pm palladium film membrane delaminated before hydrogen permeability experiments could be conducted at other temperatures. This was probably caused by a pressure surge on the sweep side which pushed the palladium film away from the ceramic membrane surface. The ceramic membrane pore diameter for this composite membrane was 10nm. In general, we found that composite membranes made using ceramic membranes with 200-nm pores were more durable than those made using 10-nm pores. The adhesion strength between the metal film and ceramic surface depends on mechanical bonding and anchoring effects (Honma and Kanemitsu, 1987). The palladium film was anchored more strongly to the 200-nm pore surface than the 10-nm pore surface. Therefore, palladium film delamination was not a problem with composite palladium-ceramic membranes made with the 200-nm ceramic membranes. Figure 4 summarizes results of hydrogen permeability experiments performed with the membrane with the 17pm palladium film. The membrane was operated for over 200 h at temperatures of 723 K and above. The apparent activation energy of 14450 J/mol was determined by nonlinear regression analysis of eqs 1 and 2 using the
0
2000 3000 p 0.573 p 0.573 (pa0.573)
1000
Ht
.
4 4000
HE
Figure 4. Hydrogen permeation data for composite palladiumceramic membrane with 17-pm palladium f i b . The solid lines represent predicted hydrogen fluxes from parameters obtained by nonlinear regression analysis of eqs 1 and 2 using the combined permeation data from the four temperatures. Calculated parameters E = 14 450 J/mol, and n are &O = 5.29 X 108 mol~m/(m2~s~Pa0~673), = 0.573.
permeation data collected at all temperatures evaluated for this membrane. Hurlbert and Konecny (1961)reported an apparent activation energy of 11925 J/mol for hydrogen permeation through palladium films ranging from 27 to 154 pm in thickness while Uemiya and co-workers (1988) reported an activation energy of 10700 J/mol for a composite palladium-porous glass membrane with a 13pm palladium film. The n value listed in Figure 4 is the value which best fits the combined temperature data. It is important to note the temperature history of the membrane. The chronological temperature history was 723, 773, 823, 873, 823, 773, and 723 K. Hydrogen permeability experiments were conducted at each of these temperatures. A significant increase in membrane permeability occurred when the membrane was heated from 773 to 823 K. When the membrane was cooled back down to 773 and 723 K, the hydrogen permeabilities were significantly higher than the initial permeabilities and the n value obtained from nonlinear regression was lower. At 723 K, the n value decreased from 0.73 to 0.62 and the hydrogen fluxes increased by 20-40%, depending on the transmembrane pressure difference. A possible explanation is that surface contaminants initially present on the membrane surface were removed after it was heated to 823 K. Operation at the higher temperatures may also have had an annealing effect on the palladium surface. The initial data from 723 and 773 K were not used to determine the parameters in Table 11,and are not included in Figure 4. Figure 5 summarizes results of hydrogen permeability experiments for the membrane with the 11.4-pmpalladium film. Hydrogenhitrogen and hydrogen/helium selectivities for this membrane are summarized in Figure 6 as a function of the transmembrane pressure difference. The selectivities are defined as the ratio of the hydrogen flux to the nitrogen or helium flux at the same transmembrane pressure difference. The hydrogenlnitrogen selectivity at 823 K ranged from 1170 at a transmembrane pressure difference of 124 kPa to 380 a t a transmembrane pressure difference of 1500 kPa. Therefore, leakage through membrane defects or seals was minimal. The hydrogen selectivity decreases with increasing transmembrane pressure difference because the permeation rate through membrane seals or palladium film defects is proportional to a higher power of pressure than permeation of hydrogen
Ind. Eng. Chem. Res., Vol. 32, No. 12, 1993 3011 >
3E
2
E
2-0E-05-
!
0
0
II
0
6
5
-
1.OE-05
oo
M
'5L 1
0
a
B ' 4 o
i-06:
so
; L
2" -0
500
1000
1500
2000
2500
3000
. m
0
0
0
o 0
20 pm palladium film
0
17 pm palladium film
0
11.4 pn palladium fllm
8
O.OE+OO . . . , . . . . , . . O.OE+OO 5.OE+05 1.OE+06
.
pH,0.S80 pHs0.580 cpa0.580)
Figure 5. Hydrogen permeation data for composite palladiumceramic membrane with 11.4-pm palladium film. The solid lines represent predicted hydrogen fluxes from parameters obtained by nonlinear regression analysis of eqs 1 and 2 using the combined permeation data from the two temperatures. Calculated parameters E = 8880 J/mol, and n are &O = 1.62 X 1 P mol.m/(m2.s.Pa0~680), = 0.680. HzINp selectivity, 823 K H d N z selectivity, 873 K H d H e seleetivily, 813 K
p
i
6001
B
0
4ol,,:,,,;,A,::,,!
a
,
200
0 O.OE+OO
5.OE+05
1.OE+06 PI - Ps
1.5E+06
,
,
,
1
2.OE+06
(Pa)
Figure 6. Hydrogen selectivity data for composite palladiumceramic membrane with 11.4-pm palladium film. Hydrogen selectivity defined as ratio of hydrogen flux to nitrogen or helium flux at the same transmembranepressure differencemeasured in permeation experiments conducted with pure gases.
through the palladium film. Correlation of the helium and nitrogen permeability data with eq 1 resulted in n values ranging from 1.37 to 1.51. The helium/nitrogen selectivity ranged from about 1.5 to 1.75 and decreased as the tranmembrane pressure difference increased. The helium/nitrogen selectivities were significantly lower than the Knudsen diffusion selectivity of 2.6, which indicates that viscous flow was significant in gas permeation through seal leaks or membrane defects. The hydrogen selectivity for the membrane with the 20-pm palladium film was similar to the selectivity of the membrane with the 11.4pm palladium film. The hydrogen selectivity of the membrane with the 17-pmpalladium film was significantly lower with hydrogenhitrogen selectivities at 873K ranging from 135 at a transmembrane pressure difference of 193 kPa to 42 at a transmembrane pressure difference of 1917 kPa. The hydrogen flux increased as the palladium film thickness decreased, but it was not proportional to the reciprocal of the palladium film thickness (l/tm), A comparison of hydrogen permeabilities for the membranes was made by dividing hydrogen fluxes by the palladium Normalized hydrogen fluxes should film thickness
t
1.5E-05
PHt
.
. , . . . . 1.5E+06 2.OE+06
- P H ~(Pa)
Figure 7. Comparison of normalized hydrogen fluxes at 823 K for composite palladium-ceramic membranes.
be equal if diffusion through the palladium film is the rate determining step. Normalized fluxes at 823 K are plotted as a function of transmembrane pressure difference in Figure 7. Normalized hydrogen fluxes for the membrane with the 11.4-pm palladium film are significantly less than those for the membranes with 17- and 20-pm palladium films. This indicates that the hydrogen permeability of the 11.4-pm palladium film is lower than the permeability of the 17- and 20-pm palladium films. This observation is consistent with the data of Hurlbert and Konecny (1961), who also reported that the hydrogen permeability of an 11.4-pm palladium film was significantly lower than the permeability of films of 24 pm and above. Their conclusion was that hydrogen fluxes through palladium films approach a limiting value when the film thickness is reduced below 20 pm. On the other hand, Uemiya and co-workers (1991~)reported hydrogen fluxes for their composite palladium-porous glass membranes were inversely proportional to the palladium film thickness down to a film thickness of 13 pm. They concluded that diffusion of atomic hydrogen through the palladium film was still the rate-limiting step at this film thickness and that the hydrogen permeability of the 13-pm film was the same as for thicker films. Examination of the n values listed in Table I1 provides support for the theory that surface processes decrease the hydrogen permeability of the membrane with the 11.4pm palladium film. The n values for 823 K should be equal for all membranes if diffusion of atomic hydrogen through the palladium film is rate limiting. The higher n value for the membrane with the 11.4-pm palladium film may indicate that surface processes have become a resistance for hydrogen transport through the membrane. In addition, a lower n value for the membrane with the 17-pm palladium film is obtained when the hydrogen leakage through Grafoil seals or membrane defects is estimated and subtracted from the measured hydrogen fluxes. Assuming a hydrogen leakage equal to twice the nitrogen flux for a given transmembrane pressure difference results in a leak corrected n value of 0.538. The hydrogen leakage was estimated by multiplying the nitrogen flux by 2 instead of the Knudsen diffusion value of 3.74 since permeation through seal leaks or membrane defects occurs by both Knudsen diffusion and viscous flow. The viscous flow contribution reduces the selectivity. The estimated hydrogen leakage for the membrane with the 17-pm palladium film was less than 5% of the total hydrogen flux at all transmembrane pressure differences, but it was high enough to significantly change the value
3012 Ind. Eng. Chem. Res., Vol. 32, No. 12, 1993 Table 111. Comparison of Hydrogen Fluxes for Inorganic Membranes.
membrane description composite palladium-ceramic membrane (11.4-pm palladium film) composite palladium-porous glass membrane (13-pm palladium film) composite metal membrane (25-pm palladium film on 30-pm vanadium foil with 1-pm intermetallic diffusion barrier between palladium and vanadium) composite metal membrane (1-2-pm palladium film on 0.25-mm-thick niobium tube) metal oxide membrane (Si02deposited in pores of 4-nm Vycor glass membrane) ceramic membrane (asymmetric membrane with 4-nm pore top layer)
HdN2
823
hydrogen flux (mol/(m2.s)) 0.71
773
0.56b
m
Uemiya et al. (1988)
973
0.3OC
m
Edlund (1992)
698
0.4b
m
Buxbaum and Marker (1993)
723
0.015b
T (K)
811
selectivity 650
1000-5000
23b
