Effects of Water Gas Shift Gases on Pd-Cu Alloy Membrane SurfaceMorphology and Separation Properties

Ames Kulprathipanja,† Go1khan O. Alptekin,‡ John L. Falconer,† andJ. Douglas Way*,§

Department of Chemical Engineering, University of Colorado, Boulder, Colorado 80309,TDA Research, Inc., Wheat Ridge, Colorado 80033, and Department of Chemical Engineering,Colorado School of Mines, Golden, Colorado 80401

Palladium-copper alloy membranes (2-4 µm thick), which were deposited on porous ceramictubular supports by electroless plating, separated H2 from a water gas shift (WGS) mixture at623-723 K. However, within 120 min at 623 K, the total permeance increased and membraneselectivity decreased. H2, N2, or He did not change the transport properties of the membranebefore WGS gas exposure. Annealing in CO2 and CO at 523-723 K increased the height of micron-scale conical hillocks and defect sizes on the membrane surface by at least a factor of 3. Annealingin H2 and He after CO and CO2 exposure decreased the hillock heights by half. The hillockswere clusters of grains with defined valleys. The membrane defects allowed gases other thanH2 to permeate through the membrane. Surface topology changes are partially due to the removalof C impurities by CO2 to form CO. Hillock heights on 25-µm-thick cast and rolled Pd-Cu alloyfoils, which had no C impurities, increased by a factor of 4 at 723 K in the presence of CO2. Thesurface of the electroless-plated films was a factor of 3 rougher than the foils. Decreasing themembrane surface roughness, increasing the membrane thickness, and minimizing C impuritiesdecreased membrane defect formation associated with surface rearrangement. Fewer vacanciesand lattice defects in the alloy lattice may also make the foil more resistant to atomrearrangement than the electroless-plated membranes. The extent of WGS, CO disproportion-ation, and methanation reactions on the membrane increased at higher Cu alloy concentrations.Exposure to CO and CO2 segregated Pd to the feed side of the membrane and changed themembrane alloy composition and phase structure. The change in phase structure from body-centered cubic to face-centered cubic decreased the H2 permeance through the membrane andmay increase surface rearrangement.

Introduction

Palladium alloys have been studied as H2-separatingmembranes in applications that include hydrogenationand dehydrogenation reactions, H2 recovery from plantstreams, and coal gasification.1-4 Many of these pro-cesses, including the water gas shift (WGS) reaction (eq1) utilize the H2-separating properties of Pd alloy

membranes in a membrane reactor to obtain conver-sions higher than thermodynamic equilibrium.2,5,6 Pdmembranes have been operated at temperatures as highas 1023 K. In addition, combining reaction and separa-tion into one unit can decrease the energy consumptionand capital costs. Because H2 is an essential feedstockin the refining and chemical industry and is needed forfuel cells, this type of membrane reactor can be used inmany applications.7-11

One disadvantage of Pd membranes is the high costof Pd. Alloys require less Pd and thus lower the cost,but they also have many advantages for separation andstability. Pd has been alloyed with Ag, Cu, Au, andRu.12-15 Pd-Cu alloys between 30 and 60 wt % Cu form

the body-centered cubic (bcc) phase, which has higherH2 permeability, is more resistant to sulfur compounds,and can operate at lower temperatures than the face-centered cubic (fcc) phase of pure Pd.13,15,16 Below 573K, pure Pd forms a palladium hydride, which embrittlesas a result of lattice expansion. Alloying Pd alsodecreases the problem of membrane embrittlementcaused by the pure Pd f palladium hydride phasetransition during temperature cycling.17 Pd-Cu alloyswith either the fcc or bcc phase structure do notembrittle at ambient temperature because they do notform a hydride phase.

Pd and Pd alloy membranes have been reported withthicknesses of less than 5 µm.1,12,18,19 These thinnermembranes require less Pd and are cheaper to produce.Though most Pd or Pd alloy membranes have been foilsmanufactured by casting and rolling, additional fabrica-tion techniques are being used to decrease the mem-brane thickness. Recently, membranes were preparedby sputtering, chemical vapor deposition, and electrolessplating. These membranes are usually supported onporous ceramic or stainless steel. Some membraneproperties from recent studies are shown in Table1.1,12,18,20 All of the membranes have similar H2 per-meances (∼10-6 mol/s/m2/Pa) but have different proper-ties due to composition, thickness, and the fabricationmethod. Uemiya et al.12 fabricated porous membranesby chemical vapor deposition, and the H2 transportthrough their membranes was hypothesized to besurface diffusion of H atoms. In contrast, H2 permeates

* To whom correspondence should be addressed. Tel.: (303)273-3519. Fax: (303) 273-3730. E-mail: dway@mines.edu.

† University of Colorado.‡ TDA Research, Inc.§ Colorado School of Mines.

CO + H2O f CO2 + H2 (1)

4188 Ind. Eng. Chem. Res. 2004, 43, 4188-4198

10.1021/ie030853a CCC: $27.50 © 2004 American Chemical SocietyPublished on Web 06/12/2004

through Pd and Pd alloy membranes fabricated byelectroless plating or casting and rolling by solid-statediffusion of H atoms through the metal lattice.21

For most Pd membranes, H2 dissociates on themembrane surface, and H atoms diffuse through themembrane and recombine on the permeate side. Solid-state diffusion in the Pd alloy is the rate-limiting stepfor most operating conditions, but other possible rate-limiting steps include external mass-transfer resistance,reduced sticking probability due to surface contamina-tion, and H2 desorption for membrane thicknesses ofless than 1 µm.22 A decline in H2 permeation due to COand H2O blocking H2 dissociation sites has been re-ported on Pd and Pd-Ag membranes.23,24 This studyexamined the effects of WGS reactants and products onthe separation and catalytic properties of Pd-Cu alloymembranes prepared by electroless plating.

Experimental Methods

Membrane Preparation. The electroless-plated Pd-Cu alloy membranes were fabricated using a proceduredescribed previously.25 The Pd-Cu alloy films wereprepared by successively depositing Pd and Cu ontoasymmetric R-alumina tubular supports with a 20- or50-nm-pore zirconia top layer (US Filter T1-70) andsymmetric R-alumina tubular supports with a 200-nm-pore top layer (CoorsTek GTC-998). The ceramic tubeswere cleaned and activated by impregnating the toplayer with a palladium acetate solution. The tubes wereheated in air to oxidize the acetate and then reduced inflowing H2 to leave dispersed Pd metal in the pores ofthe top layer. A Pd film was formed by nucleation andgrowth at the Pd metal sites to a desired thickness. Anosmotic pressure gradient was applied across the mem-brane by flowing an aqueous sucrose solution on thepermeate side of the support during plating to reducethe porosity and promote surface homogeneity.26 AfterCu was plated on top of the Pd film, intermetallicdiffusion of the Pd and Cu metals was induced byannealing in a 5% H2/He gas mixture at 623-723 K toproduce a uniform alloy film. The annealing process tookapproximately 5 days at 673 K for membranes of lessthan 5-µm thickness. Complete annealing was deter-mined by a steady-state H2 permeance through themembrane for 2 days. The integrity of the Pd-Cu filmwas tested by pressurizing the membrane with N2,immersing it in a 2-propanol/H2O mixture, and checkingit for gas leaks.

In addition to the Cu plating process previouslyused,25 Cu was also deposited using a commercialprocess (Technic-Electroless Copper by Technic Inc.).Before annealing, this process produced a more uniformCu layer, which increased selectivity but decreased H2permeation. At least 10 days was necessary to anneala 5-µm-thick membrane when the Technic Cu platingprocess was used.

Two 25-µm-thick Pd60Cu40 foils from Oremet WahChang, Inc. (Albany, OR), and Idatech, Inc. (Bend, OR),were also used to compare to the electroless-platedmembranes. The Oremet Wah Chang foil was used formost of the characterization tests in this paper, unlessotherwise stated, because of its availability at the timethat this research was conducted.

Membrane Characterization. Scanning electronmicroscopy (SEM; JEOL 840) and energy-dispersiveX-ray spectroscopy (EDAX; Noran 5500) determined thefilm thickness and metal alloy composition of themembranes. The samples were sputtered with Au or Cto reduce sample charging caused by the aluminasupport. The membranes were characterized after WGSexposures. Atomic force microscopy (AFM; NanoscopeII by Digital Instruments, Inc.) and X-ray diffraction(XRD) determined changes in the surface topology andlattice structure. The maximum vertical distance of thesurface features measurable by AFM in contact modewas approximately 5 µm. The AFM images were ana-lyzed using WSxM Scanning Probe Microscopy software.Siemens D500 XRD used monochromatic Cu KR radia-tion with a wavelength of 0.154 06 nm. The averagegrain size of the membranes was determined by theScherrer equation from line broadening of the Pd-Cualloy (110), (200), and (211) peaks. A 15-µm-thick Pd65-Cu35 film was used for the AFM and XRD studies. Filmsof less than 4 µm thickness could not be removed fromthe tubular supports and handled without extensivedamage. Total C analysis (Leco C-200) determined thebulk C content in the Pd-Cu films and foils. Thesamples were oxidized at over 1773 K, and the amountof carbon oxidation products was measured. Thermo-gravimetric analysis (TGA) was also conducted in aShimadzu TGA-50 microbalance system, which alloweda continuous gas stream to be passed over a membranesample while the temperature of the system wasincreased and the weight change monitored.

Separation and Kinetic Measurements. The sepa-ration module consisted of a 2.5-cm-o.d. shell and a 0.6-cm-o.d. inner tube, both made from stainless steel 316tubing. The membrane was centered and attached tothe tube by Swagelok fittings and graphite seals. Thereactants entered through the tube, and gases perme-ated to the shell side. Heating tapes preheated the gasesto 623 K, and a tube furnace heated the membranemodule to 773 K. A digital flowmeter measured thepermeate flow rate. The system pressure was controlledwith a pressure control valve located downstream of themembrane module. Mass flow controllers fed H2, CO,CO2, He, and N2 to the system. Steam was introducedinto the system with a high-pressure liquid pump anda boiler built from a stainless steel tube, band heaters,and insulation. After passing through the separationmodule, steam was condensed and collected to preventcondensation in the system lines. The feed gas mixturewas directed either through the membrane module or

Table 1. Recent Studies on Pd and Pd Alloy Membranes for H2 Separation

Roa et al. Uemiya et al. Uemiya et al. Shu et al. Edlund

membrane composition Pd60Cu40 Pd Pt, Ru Pd Pd60Cu40support material Al2O3 Al2O3 Al2O3 Al2O3 nonemembrane thickness, µm 1.5 3 6, 3 2 15preparation technique electroless plating CVD CVD electroless plating chemically etched foiltemperature, Κ 623 773 773 673 673H2 permeance, µmol/s/m2/Pa 1.9 3.1 0.5, 2.0 1 5H2/N2 ideal selectivity 100 240 210, 120 N2 undetectable N2 undetectablereference 18 12 12 1 20

Ind. Eng. Chem. Res., Vol. 43, No. 15, 2004 4189

through a bypass loop. Permeate and retentate streamswere analyzed by a gas chromatograph equipped witha thermal conductivity detector and CO, CO2, and CH4gas analyzers.

H2/N2 ideal selectivity was determined by dividing thesingle gas fluxes at 623-723 K and 220-430 kPa. Thepermselectivity was calculated by dividing the per-meances of gases in a mixture separation. Permeanceis the flux divided by the transmembrane partialpressure difference. A linear pressure dependence wasused because of the presence of small defects thatallowed H2 permeation. Permeate and feed pressuresand mixture composition were used to calculate thepermselectivity.

Results

Membrane Characterization. (a) Compositionand Thickness. Membrane composition and thickness,the pore size of the support top layer, and how themembranes were used in this study are included inTable 2. EDAX and SEM measured the membranecompositions and thicknesses after various gas expo-sures except for membranes 1 and 8. Membrane 1 waselectrolessly plated onto a ceramic tubular support butdelaminated because of weak adhesion. The membranewas characterized before and after WGS exposure tocompare the changes in membrane properties due todifferent gas treatments. Additional Pd and Cu wereplated on membrane 8 after CO2/H2 exposure, so thealloy composition of the membrane after WGS exposurecould not be directly obtained. Therefore, by comparisonof the experimental conditions used to fabricate themembrane and earlier results of membranes made onGTC supports, its properties were approximated.25

Membranes 2-7 were exposed to WGS gases, broken,and then characterized. A SEM micrograph (Figure 1)of a Pd75Cu25 membrane (membrane 3) shows a crosssection of a membrane deposited on an alumina poroussupport with a zirconia top layer. The gap between thesupport top layer and membrane may have formedwhen the tubular support was broken. The voids in theelectroless-plated membrane probably formed duringthe nucleation and growth process.

Measurements using EDAX showed that the composi-tion of the top 1-µm surface region of the Wah Changfoil changed from 40 to 25 and 20 wt % Cu during CO2and CO exposure, respectively. Annealing in H2 afterCO exposure increased the Cu concentration of the foilto approximately 30 wt %. Therefore, the Cu concentra-tions in Table 2 for the feed side of the electroless-platedfilms exposed to carbon oxides may be lower than thosein the bulk. Pd segregation to the surface region mayexplain the low Cu concentration after CO and CO2

exposure because an alloy composition with 40 wt % Cuwas normally targeted during fabrication. In addition,cross-sectional analysis by EDAX of membranes exposedto carbon oxides showed higher concentrations of Cu(50-70 wt %) near the support than those on the feedside. Cu must segregate to the support side because Pdwas electrolessly deposited onto the support before Cu.The metals interdiffuse for the Pd-Cu alloy to form.After annealing in H2 but before exposure to CO andCO2, membrane 1 had a uniform alloy compositionacross the thickness of the membrane. The Cu concen-tration of membrane 5 was increased to 90 wt % todetermine the effects of a high Cu alloy composition onseparation and catalytic properties. Membrane 6 did notcontain any Cu and was used to compare a pure Pdmembrane with the Pd-Cu alloys.

(b) Electroless-Plated Film Surface Topology. A15-µm-thick Pd65Cu35 alloy membrane annealed in H2at 723 K and 220 kPa for 7 days exhibited hillocks,micron-scale conical features protruding from the sur-face, and valleys, micron-scale indentations on thesurface of the film. These are shown in the 5 × 5 µmAFM images of the feed side of the membrane in Figure2. The light spots in Figure 2a are the hillocks and thedarker spots the valleys (boundaries between separatehillocks). These conical hillocks and their heights areclearly seen in Figure 2b. The maximum hillock heightof the sample is shown on the z axis of the three-dimensional AFM image.

The average grain size of the Pd65Cu35 film wasapproximately 100 nm before and after exposure tovarious WGS gases, as determined by the Scherrerequation. Though applicability of the Scherrer equationdecreases above 80 nm, no change in the grain size was

Table 2. Electroless-Plated Pd and Pd-Cu MembraneComposition, Thickness, Pore Size of the Support TopLayer, and Usage

membraneCu

wt %thickness,

µm

pore size of thesupport toplayer, nm usage

1 35 ( 4 15 ( 1 50 AFM, XRD2 30 ( 5 3 ( 1 20 C analysis3 25 ( 3 3 ( 1 50 binaries, SEM4 20 ( 3 3 ( 1 50 WGS mixture5 90 ( 2 3 ( 1 20 catalytic activity6 0 2 ( 1 20 N2/CO2/H27 25 ( 3 4 ( 1 50 CO2/CO/He8 20 ( 4 12 ( 1 200 CO2/H2

Figure 1. SEM micrograph of a Pd75Cu25 alloy membrane.

Figure 2. AFM images of a Pd65Cu35 alloy film membraneannealed in H2 at 723 K and 220 kPa for 7 days: (a) two-dimensional view; (b) three-dimensional view.

4190 Ind. Eng. Chem. Res., Vol. 43, No. 15, 2004

seen before and after WGS gas exposure. Therefore, thehillocks are not single grains (crystallites with the samebulk structure) because they are on the order of 400 nmin width. Electroless plating nucleates and grows grainsformed from Pd seeds impregnated onto the supportduring activation. The voids, seen in Figure 1, may formthe valleys between the hillocks seen in the AFMimages. The valleys, if enlarged, can lead to membranedefect formation.

A summary of the surface roughness and maximumhillock heights obtained from the three-dimensionalAFM images for a 5 × 5 µm membrane sample areaafter various gas exposures at 523 and 723 K and 220kPa are shown in Tables 3 and 4. Different sections ofmembrane 1 were used for each gas exposure thoughthe samples were all first exposed to H2 at 723 K and220 kPa for 7 days. The films were peeled off thesupport, and both sides of the membrane were exposedto the various gases. The surface roughness and maxi-mum hillock heights in Tables 3 and 4 are on theretentate side and not the support side. The maximumhillock heights describe the vertical growth of the conicalhillocks that protrude from the surface. The surfaceroughness is calculated by the root mean square (rms)of the differences between the mean surface height andindividual hillock or valley heights using the AFManalytical software. The surface roughness is an averageof the height differences and does not describe thevariation in hillock heights of a particular sample.Therefore, a rougher surface has larger differencesbetween the mean surface height and individual hillockor valley heights. In addition, because the hillocks growhorizontally and vertically, there are fewer hillocks perunit area. The tables include the results for the Pd65-Cu35 electroless-plated film and the Pd60Cu40 WahChang foil. The data have a reproducibility of ap-proximately (10% for a 5-8 quantity sample set. TheIdatech foils were also exposed to H2 and CO2 and had

((10%) roughness and hillock heights similar to thoseof the Wah Chang foils.

Exposing the delaminated electroless-plated film topure CO2 at 723 K for 7 days coalesced the hillocks,which are seen as micron-size light spots, and producedlarger hillock heights and deeper valleys, seen as thedark spots (Figure 3; note the difference in the z-axisscale from Figure 2). The CO2-exposed film is rougherthan the original film because the hillocks increased inheight and width. The maximum height of the hillocksincreased from 0.43 to 2.5 µm. Interestingly, annealingthe membrane at 723 K in He or H2 partially reversedchanges caused by CO2. The maximum hillock heightsdecreased from 2.5 to 1.2 µm after He exposure and to0.93 µm after H2 exposure. A second CO2 exposure hadno effect on the hillock heights and surface roughnessof the electroless-plated membrane. When a separatesection of the original H2-annealed membrane wasexposed to CO at 723 K for 7 days, the maximum hillockheight increased from 0.43 nm to 1.1 µm, less than halfthe height of the CO2 exposure. The surface topology ofthe CO-exposed membrane was similar to that of theCO2-exposed membrane after H2 and He exposure (∼1µm). H2 at 723 K did not reverse the surface changescaused by the CO. The film’s surface topology changedless when exposed to CO2 and CO for 7 days at 523 Krather than 723 K.

(c) Cast and Rolled Foil Surface Topology. Thecast and rolled foils were smoother than the electroless-plated films (Tables 3 and 4). The Pd60Cu40 foil surfaceroughness and hillock heights were less than 40% ofthose of the Pd65Cu35 film before CO2 exposure. Theinitial Wah Chang foil surface roughness and maximumhillock height were 19 nm and 0.18 µm, respectively.The Idatech foil had an initial surface roughness andhillock height of 31 nm and 0.19 µm, respectively. Theinitial film surface roughness and maximum hillocksheights were 67 nm and 0.43 µm, respectively.

CO2 also changed the surface topology of the Pd60-Cu40 foils (Figures 5a and 5b). Though Figure 5b of thefoil after CO2 exposure may seem rougher than Figure3b of the film after CO2 exposure, there is more than afactor of 3 difference in the z-axis scales. Exposing theWah Chang foil to CO2 at 723 K increased the surfaceroughness from 19 to 148 nm. After CO2 exposure, thesurface roughness and hillock height of the Idatech foilincreased to 155 nm and 0.85 µm, respectively. Inaddition, unlike the electroless-plated film, the foils’surface topology did not change following CO exposure.He and H2 partially reversed the effects of CO2 on thefoils. Furthermore, unlike the electroless-plated film, asecond CO2 exposure increased the hillock height and

Table 3. Surface Roughness after Gas Exposures at 220kPa for 7 days; 5 × 5 µm Sample Area

surface roughness(rms), nm

gasPd65Cu35

filmPd60Cu40

foilexposuretemp, K

H2 67 ( 7 19 ( 4 723CO2 415 ( 43 148 ( 18 723He after CO2 201 ( 17 108 ( 12 723H2 after CO2 175 ( 25 78 ( 10 723CO2 after CO2 and H2 143 ( 15 123 ( 30 723CO 170 ( 10 14 ( 3 723H2 after CO 190 ( 7 12 ( 2 723CO2 160 ( 10 24 ( 4 523CO 150 ( 10 18 ( 2 523

Table 4. Maximum Hillock Height after Gas Exposuresat 220 kPa for 7 days; 5 × 5 µm Sample Area

maximum hillockheight, µm

gasPd65Cu35

filmPd60Cu40

foilexposuretemp, K

H2 0.4 ( 0.1 0.18 ( 0.04 723CO2 2.5 ( 0.3 0.77 ( 0.13 723He after CO2 1.2 ( 0.1 0.76 ( 0.10 723H2 after CO2 0.9 ( 0.2 0.46 ( 0.10 723CO2 after CO2 and H2 1.0 ( 0.1 0.79 ( 0.20 723CO 1.1 ( 0.1 0.10 ( 0.02 723H2 after CO 1.3 ( 0.3 0.09 ( 0.02 723CO2 1.1 ( 0.2 0.17 ( 0.05 523CO 0.7 ( 0.2 0.12 ( 0.05 523

Figure 3. AFM images of a Pd65Cu35 alloy film membraneexposed to CO2 at 723 K and 220 kPa for 7 days: (a) two-dimensional view; (b) three-dimensional view.

Ind. Eng. Chem. Res., Vol. 43, No. 15, 2004 4191

surface roughness to values similar to those of the firstCO2 exposure (Tables 3 and 4). The foils’ surfacetopology did not change at 523 K.

(d) Surface Area. The surface area of the films andfoils, calculated using the AFM software, also cor-responded well with the trends in surface roughness.However, the magnitudes of the change in the surfacearea were small because of the nanometer-scale surfacefeatures. The surface area of the foils and films annealedin H2 were 25 and 31 µm2, respectively, for a 5 × 5 µmsample area. Surface area and roughness both increasedor decreased with different gas exposures except for theCO2- and CO-exposed film at 523 K. The lower temper-ature, which increased the surface area but decreasedthe surface roughness, produced more, but shorterhillocks, whereas the higher temperature producedfewer, but taller hillocks as described earlier.

(e) Phase Structure. The Pd65Cu35 film and bothfoils were annealed in H2 for 7 days at 723 K andexposed to pure CO2 and CO at 723 K for 7 days. TheXRD pattern showed that the film had a Pd-Cu bcclattice parameter of 0.298 nm (Figure 6). The latticeparameter is the length of the unit cell of the cubicstructure. A small amount of the fcc phase was alsopresent with a lattice parameter of 0.377 nm. The bcclattice parameter calculated from the (110) and (211)crystal planes parallel to the surface did not changeafter exposure to CO2 or CO; no indications of bulkcarbide or oxide formation were detected. However, astronger fcc peak for the film was detected after CO2,CO, and a second H2 exposure (Figure 7). The unan-nealed Wah Chang and Idatech Pd60Cu40 foils were bothfcc but changed to bcc after annealing in H2. The Idatechfoil crystal plane parallel to the surface was similar tothe electroless-plated film; both have a larger bcc (110)orientation, whereas the Wah Chang foil was mainlybcc with a (211) orientation. Both of the foils’ bulk

crystal structures also saw a small increase in the fccstructure after exposure to CO2 or CO.

(f) Carbon (C) Content. Total C analysis on a Pd70-Cu30 film measured 8 wt % C in the membrane afterH2 annealing, 5 wt % after CO annealing for 7 days at723 K, and below the detection limit of 0.01 wt % afterCO2 annealing for 7 days at 723 K. No C was detectedin either foils after H2, CO2, or CO exposures. The Ccontents of three 10-mg film and foil samples weremeasured after each gas exposure and were within 1wt % or less of each other. Furthermore, TGA deter-mined that flowing CO2 over electroless-plated films andcast and rolled foils resulted in sample weight losses ofapproximately 6 and 0.2%, respectively. Sample sizesof approximately 6 mg were used and heated to 673 K.

Membrane Transport Data. (a) Effect of CO andCO2. H2/CO2 permselectivity versus temperature andtime for a 50% CO2/H2 mixture through a Pd75Cu25membrane is shown in Figure 8. As stated earlier, thealloy compositions for the electroless-plated membraneswere measured after CO2 exposure and had a high Pdconcentration at the feed side surface region because ofPd segregation. Moreover, the 25 wt % Cu concentrationis similar to the alloy composition measured by EDAXfor the Pd60Cu40 Wah Chang foil exposed to CO2. BothH2 and CO2 permeate flows increased with temperature.The H2/CO2 permselectivity reached a maximum of 55at 150 min and then decreased to 28 after 1200 min.Before introduction of CO2, the H2 permeance at 623 Kwas 2.5 × 10-7 mol/s/m2/Pa. The initial H2 permeance

Figure 4. AFM images of a Pd60Cu40 alloy foil membrane exposedto H2 at 723 K and 220 kPa for 7 days: (a) two-dimensional view;(b) three-dimensional view.

Figure 5. AFM images of a Pd60Cu40 alloy foil membrane exposedto CO2 at 723 K and 220 kPa for 7 days: (a) two-dimensional view;(b) three-dimensional view.

Figure 6. XRD patterns of a Pd65Cu35 alloy membrane annealedin H2 at 723 K and 220 kPa for 7 days, a Wah Chang Pd60Cu40foil unannealed and annealed in H2 at 723 K for 7 days, and aIdatech Pd60Cu40 foil annealed in H2 at 723 K for 7 days.

Figure 7. XRD patterns of a Pd65Cu35 alloy membrane annealedin H2, CO2, and CO at 723 K and 220 kPa for 7 days.

4192 Ind. Eng. Chem. Res., Vol. 43, No. 15, 2004

was 1 order of magnitude lower than that for mem-branes fabricated by Roa et al. (Table 1). A similarelectroless plating technique was used to fabricate themembrane except for the use of the Technic Cu platingbath. Additional differences were membrane thicknessand alloy composition. The Pd60Cu40 membrane fabri-cated by Roa et al. was 1.5 µm thick compared to 3 µmthick for the Pd75Cu25 membrane. A 40 wt % Cu alloycomposition has the highest H2 permeability so anydeviation in the alloy composition would decreasepermeation.18

Introducing CO2 to the H2 feed at 20 min and 623 Kdecreased the H2 permeance by 2 orders of magnitude(Figure 9). After 200 min, the H2 permeance increasedwith temperature to 5.0 × 10-7 mol/s/m2/Pa at 723 K.An increase in the H2 permeance with temperatureinitially increased the H2/CO2 permselectivity, but theCO2 permeance increased with time and this decreasedthe permselectivity. After CO2 exposure, the membranewas exposed to CO and then H2O for approximately 20h each. The H2 permeance increased to 1 × 10-6 mol/s/m2/Pa during a 5% CO/H2 exposure at 723 K. A 7%H2O/H2 mixture decreased the H2 permeance throughthe membrane (Figure 10), but a 10% H2O/He mixturedelaminated the film from the support, causing themembrane to fail within 60 min.

The H2 and N2 single gas permeances and idealselectivities for the same Pd75Cu25 membrane weredetermined between the different gas exposures de-scribed in the previous paragraph (Table 5). The mem-

brane was exposed to the gas(es) listed in the lastcolumn of the table for at least 1 day before measure-ment of the H2/N2 ideal selectivity. After annealing ofthe membrane in He and a H2/N2 mixture at 723 K for6 days, the N2 permeance was below the detection limitof 1.5 × 10-9 mol/s/m2/Pa. Binary combinations of CO2/H2 and CO/H2 increased H2 and N2 permeances anddecreased the ideal selectivity to 32 and 11, respectively.He flow after CO2 exposure increased the ideal selectiv-ity to 63; both of the H2 and N2 single gas permeancesdecreased. The addition of a H2O/H2 mixture after COexposure formed CO2 and CH4 (Figure 11) and alsoincreased the ideal selectivity to 51. The decrease in theformation of the C products coincides with the reductionof C being removed by the H2O/H2 mixture on themembrane surface.

Figure 8. H2/CO2 permselectivity and temperature versus timefor a Pd75Cu25 alloy membrane: total transmembrane pressure )420 kPa; 50% CO2/H2 feed composition.

Figure 9. H2 permeance and temperature versus time for a Pd75-Cu25 alloy membrane: total transmembrane pressure ) 420 kPa;50% CO2/H2 feed composition.

Figure 10. H2 permeance and temperature versus time for a Pd75-Cu25 alloy membrane: total transmembrane pressure ) 390 kPa;7% H2O/H2 feed composition.

Figure 11. CO2 and CH4 concentration in the retentate after COand during H2O/H2 exposure for a Pd75Cu25 alloy membrane: totaltransmembrane pressure ) 380 kPa; 7% H2O/H2 feed composition.

Table 5. H2 and N2 Permeances and H2/N2 IdealSelectivity versus Time at 723 K and 430 kPaTransmembrane Pressure for a Pd75Cu25 AlloyMembrane after Various Gas Exposures

time,days

H2 permeance× 107,

mol/s/m2/Pa

N2 permeance× 109,

mol/s/m2/Pa

idealselectivity

H2/N2 gas

5 2.3 ( 0.1 <1.5 >150 He6 2.3 ( 0.1 <1.5 >150 N2/H27 2.8 ( 0.1 8.9 ( 0.4 32 ( 3 CO2/H2

11 6.5 ( 0.3 20 ( 1 32 ( 3 CO2/H212 5.4 ( 0.3 8.6 ( 0.4 63 ( 6 He13 11 ( 0.6 98 ( 5 11 ( 1 CO/H214 8.6 ( 0.4 17 ( 0.9 51 ( 5 H2O/H2

Ind. Eng. Chem. Res., Vol. 43, No. 15, 2004 4193

He permeance versus time for several binary mixturesthrough a Pd75Cu25 membrane showed that CO2 and COexposure changed the membrane performance and notbyproducts from reactions with H2 (Table 6). A 2% H2/He mixture and pure He did not alter the total per-meance through the membrane, but the addition of CO2and CO increased the permeation as shown earlier withthe H2 binaries. The He permeance in a 50% CO2/Hebinary reached a steady state, but the He permeancein a 10% CO/He mixture did not. After exposure of themembrane to CO2 and CO for a total of 1350 min, theHe permeance was 2.5 times its original value.

CO2 did not decrease the membrane selectivity orincrease the H2 permeance of a 12-µm-thick electroless-plated Pd80Cu20 membrane. Though the initial H2permeance of the membrane was only 9.8 × 10-9 mol/s/m2/Pa because of the thickness and alloy composition,the membrane had a H2/N2 ideal selectivity of at least150 and it did not change during exposure to a 50% CO2/H2 mixture for 1800 min at 723 K and 1120 kPa. Thepermselectivity of the 3-µm-thick Pd75Cu25 membraneexposed to the same conditions except at 420 kPatransmembrane pressure decreased after 60 min at 723K (Figure 8). The permeability (permeance × membranethickness) of the Pd80Cu20 12-µm-thick membrane (1.2× 10-13 mol/s/m/Pa) was similar to that of the Pd80Cu203-µm-thick membrane (1.5 × 10-13 mol/s/m/Pa).

A pure Pd membrane exposed to a tertiary mixtureof H2, N2, and CO2 showed transport characteristicssimilar to those of the Pd-Cu alloys. The H2 permeancewas 1 order of magnitude higher than the N2 and CO2permeances, and all three gas permeances increasedwith temperature (Figures 12 and 13). At 723 K, boththe H2/N2 and H2/CO2 permselectivities reached maximaof approximately 15 but decreased to 8 after 2500 min.The H2 permeance increased to a maximum of 3.4 ×10-6 mol/s/m2/Pa at 1600 min, whereas the N2 and CO2permeances reached maxima of approximately 2.5 ×10-7 mol/s/m2/Pa at 2200 min. The N2 permeance wasslightly higher than the CO2 permeance at 723 K.

(b) Effect of the Alloy Composition. The H2permeance of the pure Pd membrane was close to 2orders of magnitude higher than that of the Pd80Cu20membrane used for the WGS exposure because of thealloy composition and thickness. The large differencein H2 permeability is due to the variation in thediffusivity and solubility of the different Pd-Cu alloycompositions. Piper reported that diffusion coefficientsfor Pd-Cu alloys span more than 2 orders of magni-tude.16 The increase in the diffusion coefficients is dueto the change from the fcc to bcc phase structure as theCu concentration is increased. However, increasing theCu concentration in the Pd-Cu alloy also decreases theH2 solubility. Roa et al. have reported that the H2permeability for an electroless-plated Pd membrane canbe at least 4 times higher than that for a Pd80Cu20membrane.18

(c) Effect of the WGS Mixture. A Pd80Cu20 mem-brane exposed to a WGS mixture (5% CO, 20% CO2, 35%H2O, and 40% H2) at 623 K (Figure 14) had an initialH2/N2 permselectivity of 45 and a total permeance of 5× 10-8 mol/s/m2/Pa. The H2/CO and H2/CO2 permselec-tivities were initially 23 and 15, respectively. Thepermselectivities increased slightly with time, but thenboth decreased to approximately 15 as the total per-meance increased to 1.8 × 10-7 mol/s/m2/Pa. The experi-ment was stopped before reaching steady state so thatthe membrane could be characterized.

Membrane Catalytic Activity. The Pd-Cu mem-brane catalyzes the WGS (eq 1), CO disproportionation

Figure 12. H2 permeance and temperature versus time for a purePd membrane: total transmembrane pressure ) 300 kPa; 25%N2, 25% CO2, and 50% H2 feed composition.

Table 6. He Permeance versus Time for a Pd75Cu25 AlloyMembrane Exposed to H2, CO2, and CO at 723 K

time,min gas

He permeance× 108, mol/s/m2/Pa

1500 2% H2/He 3.0 ( 0.21750 He 3.3 ( 0.22750 50% CO2/He 5.1 ( 0.33100 10% CO/He 7.8 ( 0.43400 He 7.6 +/-0.4

Figure 13. Nitrogen and CO2 permeance and temperature versustime for a pure Pd membrane: total transmembrane pressure )300 kPa; 25% N2, 25% CO2, and 50% H2 feed composition.

Figure 14. Total permeance and H2/N2, H2/CO, and H2/CO2permselectivities versus time for a Pd80Cu20 alloy membrane:temperature ) 623 K; total transmembrane pressure ) 280 kPa;5% CO, 20%, CO2, 35% H2O, and 40% H2 feed composition.

4194 Ind. Eng. Chem. Res., Vol. 43, No. 15, 2004

(eq 2), and methanation (eq 3) reactions.

At 723 K and 340 kPa feed pressure, a Pd10Cu90membrane converted 16% of the CO2 in a 50% CO2/H2mixture to CO by the reverse WGS reaction, whereas aPd75Cu25 membrane converted only 6%. A pure Pdmembrane converted less than 0.5% of the CO2 in thefeed.

Discussion

Effect of CO and CO2. (a) H2 Separation. Unlikethe behavior reported for other pure Pd and Pd-Agalloy membranes,5,6,23,24,27,28 introducing CO2 or CO toPd and Pd-Cu alloy membranes prepared by electrolessplating decreased the selectivity and increased thepermeation over time. The conical hillocks on the Pd65-Cu35 electroless-plated membrane grew to 2.5-6 timestheir original size during exposure to CO and CO2(Figures 2 and 3). CO and CO2 enlarged valleys andformed membrane defects, which increased transportof molecules other than H2 through the membrane(Table 5). In membrane 3, the H2/CO2 permselectivitydecreased with time as CO2 permeated through mem-brane defects (Figure 8). The initial increase in thepermselectivity was due to the increase in the H2permeance with temperature. H2 permeation by solutiondiffusion increased faster with temperature than theincrease in the membrane defect size, which increasedCO2 permeation. The formation of membrane defectsduring CO and CO2 exposure also explains the higherH2/N2 permselectivity before introduction of carbonoxides compared to the H2/CO2 and H2/CO permselec-tivities in Figure 14.

Permeation through Knudsen size defects depends onthe molecular weights of the individual molecules.Therefore, if permeating only by Knudsen diffusion, N2should diffuse 1.3 times faster than CO2. The similarN2 and CO2 permeances at 623 K in Figure 13 may bebecause CO2 has a higher adsorption energy than N2on the Pd surface. CO2 and N2 may transport throughmembrane defects by an additional transport mecha-nism such as surface diffusion. At 723 K, CO2 per-meance was lower than N2, possibly because of thehigher rate of CO2 desorption at higher temperatures.Unlike the results shown in Figure 13, Knudsen diffu-sion also predicts that CO2 and N2 permeance shoulddecrease with increasing temperature. However, in-creasing temperature increased the size of the hillocksand membrane defects (Table 4), which increased thepermeation of CO2 and N2.

The increase in the H2 permeance (Figure 9) duringCO2/H2 mixture exposure was not due to surface rear-rangement caused by byproducts from side reactionsbecause the He permeance also increased with time fora CO2/He mixture (Table 6). However, CO formed whenboth H2 and CO2 were exposed to the membrane becausethe Pd-Cu alloy promotes the reverse WGS reaction.CO, similar to CO2, increased the membrane defect sizesand decreased the membrane selectivity (Table 5).

(b) Surface Roughness and Thickness. The elec-troless-plated film surface was 3 times rougher than thefoil initially and after exposure to the WGS gases. Themaximum hillock height of the foil after exposure to CO2

was 0.77 µm compared to 2.5 µm for the film. The higherhillock heights of the electroless-plated films partiallyexplain the lower resistance of the films to WGS gasesthan the foils. Because hillocks that formed from elec-troless plating coalesce and grow vertically from thesupport, a minimum membrane thickness may benecessary to reduce the negative effects of valleys ormembrane defects. Hillock heights of the Pd-Cu alloyfilms increased from 0.4 to 2.5 µm after exposure to CO2at 723 K for 7 days. If the original film thickness wasonly 3 µm, valleys would penetrate through the entiremembrane, form membrane defects, and decrease themembrane selectivity, as seen in the transport results.In contrast, a 12-µm-thick Pd80Cu20 membrane fabri-cated using the same electroless plating procedure didnot show a decrease in the selectivity during CO2 ex-posure to the same temperature and more than 2 timeshigher feed pressure. The thicker membrane allowed thehillocks to grow without forming valleys that penetratedthrough the entire membrane. The permeance did notincrease with surface rearrangement, because the over-all change in the membrane thickness was averaged byboth hillock and valley formation. Therefore, smootherand thicker membranes resist membrane defect forma-tion due to surface rearrangement better than rougheror thinner membranes. Pd and Pd-Ag alloy membranesin other studies were more than 15 µm thick and werefabricated by electroless plating or obtained as foils. ThePd-Cu alloy membranes that deteriorated during COand CO2 exposure were less than 4 µm thick and weremade by electroless plating.

(c) Alloy Composition. The low Cu concentrationdetected by EDAX at the top 1-µm surface region wasdue to exposure of carbon oxides. The target compositionduring membrane fabrication was 40 wt % Cu for all ofthe membranes except membranes 5 and 6. However,the membrane composition in most cases could not beverified until after breaking the membrane for charac-terization. The Pd65Cu35 film phase structure trans-formed from mainly bcc to some fcc in the presence ofCO2 and CO (Figure 7). The phase structure changedbecause of Pd segregation to the surface region of themembrane. The fcc phase structure has a lower diffusioncoefficient and is present at less than 30 wt % Cu andgreater than 60 wt % Cu.29 The fcc phase formed notonly at the feed side because of Pd segregation but alsocloser to the support side because of enrichment of Cu.The surface morphology rearrangement did not dependon the specific bcc phase orientation (110 vs 211)because both Wah Chang and Idatech foils were affectedby CO2 exposure. The presence of Cu in the membraneis also not responsible for the decline in the membraneperformance because the effect was also seen with purePd membranes that were prepared using our fabricationtechnique.

(d) C Impurities. Removing C from the membranestructure may be partially responsible for the changesin the surface topology and selectivity. C was detectedin a Pd70Cu30 membrane before CO2 and CO exposure,probably from the activation procedure and/or platingprocess. After CO2 exposure for 7 days at 723 K, C wasnot detected in the membrane, probably because it wasoxidized to CO (eq 4). TGA completed with CO2 on both

foils and films in this study supports the removal of Cvia eq 4. Moreover, exposure to CO deposited C and

2CO f C(s) + CO2 (2)

CO2 + 4H2 f CH4 + 2H2O (3)

C(s) + CO2 f 2CO (4)

Ind. Eng. Chem. Res., Vol. 43, No. 15, 2004 4195

formed CO2 on the Pd-Cu alloy membrane by thereverse reaction (eq 2). This would explain the small Ccontent detected using total C analysis after CO but notCO2 exposure. C was also removed in the form of CO2and CH4 when a H2O/H2 mixture was introduced to amembrane after CO exposure (Figure 11). The shortdecline in CO2 and CH4 formation at 200 min may beevidence of two types of C on the membrane surface ortwo different mechanisms of C removal. Most of the Cfrom the plating process, but maybe not all, had alreadybeen removed from the membrane by CO2 before COwas introduced. Galuszka et al. reported that COexposure formed filamentous C, which led to swellingand membrane failure in an electroless-plated Pdmembrane.30 Membrane swelling by C may explain whythe total permeance during 10% CO/He exposure didnot reach a steady state after 350 min at 723 K (Table6).

Removing C with CO2 exposure increased the surfaceroughness (Table 3), hillock height (Table 4), and defectdiameter, and it decreased the H2 selectivity (Figure 8).A second CO2 exposure did not change the film’s surfacetopology after H2 exposure possibly because CO2 hadremoved all C during the first exposure (Tables 3 and4). Similarly, the foils did not have any detectable Ccontent, and they exhibited a much lower degree ofsurface rearrangement than the electroless-plated films.CO2 increased the hillock heights on the foils to 0.77µm compared to 2.5 µm for the films. However, unlikethe film, a second CO2 exposure changed the foil’ssurface topology after H2 exposure. Therefore, CO2 musthave an additional effect on the Pd-Cu membranesbesides removing C. The film’s surface may not havechanged significantly after the second CO2 exposurebecause of the initially rougher surface than the foil.Aspects that may induce topology changes due to COand CO2 exposure exposure are lattice stress relaxation,phase structure, grain coalescence, and metal sinteringas discussed below.

Surface Topology Rearrangement. (a) LatticeStress Relaxation. Hillocks form on metal filmsgenerally as a result of mismatch stresses between thesubstrate and the film, because of thermal expansion.31

This is not the reason for the surface rearrangement ofthe Pd-Cu alloy foils and films because both wereunsupported. Furthermore, because H2/N2 mixtures at623 and 723 K did not decrease the membrane selectiv-ity of supported films (Table 5 and Figure 14), thermalexpansion probably did not form hillocks on the mem-brane. However, because these metal films were poly-crystalline, stress relaxation at grain boundaries andmembrane defects is possible. Intrinsic stresses formbecause of lattice defects or vacancies in a particularlattice structure. Lattice defects may include substitu-tional or interstitial impurities in the lattice. Theannealing process did not remove the C impurities inthe electroless-plated films as seen using total C analy-sis. In addition, atom stoichiometry is controlled duringmembrane fabrication and not the annealing process.An ordered metal structure in equilibrium has thehighest atomic density and is much less likely torearrange. In Pd60Cu40 alloys, the bcc phase has a higheratom density than the fcc phase.16 Because the filmshave a rougher surface (more grain boundaries andmembrane defects), slightly higher surface area (sitesfor adsorption), and more impurities (from the electro-less plating process) and undergo larger surface topology

changes than the foils during different gas exposures,the films may have a less ordered structure and undergostress relaxation more readily. Fewer vacancies andlattice defects in the alloy lattice may make the foilsmore resistant to atom rearrangement than the films.

Stress relaxation can be initiated by migration ofatoms in the bulk material or creep processes alonggrain boundaries.32,33 Atom diffusion was detected in thePd-Cu membranes through changes in the alloy com-position and phase structure after exposure to differentgases. Measurements using EDAX showed that thecomposition of the top 1-µm surface region of the foilchanged from 40 to 25 wt % Cu after CO2 exposure.Stress relaxation may also change the macroscopicshape of the metal film by diffusional transport ofatoms, as seen in the hillock formation with the Pd-Cu alloys. Diffusion of atoms in metals has beenobserved through thermal relaxation31 but may also bepromoted by specific gas exposures. Bonding interactionbetween the carbon oxides and metal surfaces seen inthis study may promote atom mobility and thereforesurface rearrangement. Introducing H2 and He after COand CO2 exposure to the membranes also caused atommobility and reduced hillock heights (Table 4).

Another site of stress relaxation may be boundariesat sections of nonuniform alloy composition. The elec-troless-plated films were deposited as two distinct layersof Pd and Cu, which interdiffused during annealing.Because the metals were not mixed first and then rolledlike foils, the alloy composition may vary spatially by 5wt % Pd or Cu, as seen using EDAX, because of thenonuniformity in thickness of the deposition process.The differences in alloy composition within the filmcould result in more lattice defects and be partlyresponsible for the higher degree of surface topologychanges in the films than the foils, which have a moreuniform alloy composition.

(b) Phase Structure. The phase transition from abcc to fcc structure can also change the surface mor-phology and bulk structure integrity because of rear-rangement of the lattice structure (Figure 7). The fccstructure of Pd60Cu40 metal has a 1.3 times bigger latticeparameter and a lower atomic density than the bccstructure. The structure changed because of Pd segrega-tion to the feed side of the membrane. Hillocks alsoformed as ordered nanostructures on Pd films annealedin O2 at 1173 K.34 Palladium oxide hillocks approxi-mately 1 µm high formed from 200-nm-thick Pd films.A 38% volume increase accompanying oxidation of Pdmetal formed hillocks to relax the large compressivestresses due to the difference in volume between the Pdand PdO2 structures. As seen using XRD, bulk Pd-Cuoxide did not form on the membrane with CO2 or COexposure.

(c) Grain Coalescence. As seen in the AFM imagesof the Pd-Cu alloy membranes, hillocks and valleysformed during the electroless plating process. Theelectroless plating process produces grains, which coa-lesce to form hillocks. Membrane microstructure suchas the grain size has been shown to affect the membraneperformance and may also affect hillock formation. Anincrease in the Pd-Ag alloy membrane grain size from18 to 62 nm increased the H2 permeance from 3.8 to 18× 10-8 mol/s/m2/Pa and the H2/He permselectivity from23 to 45.19 The grains were fabricated by magnetronsputter deposition and grew during a heat treatmentat 773 K in He. The enlarged grain size appeared to

4196 Ind. Eng. Chem. Res., Vol. 43, No. 15, 2004

increase the H2 permeation in the bulk. In contrast,Varma et al. reported that pure Pd membranes fabri-cated with 0.5-µm grain cluster diameters had 1.5 timeshigher H2 permeability than membranes with 7-µmdiameters.35 The finer-grained cluster microstructurewas hypothesized to increase H2 permeation through thegrain boundaries. Their membranes were made byelectroless plating using different osmotic conditions tovary the cluster sizes.

(d) Metal Sintering. Metal sintering also rearrangedthe surface topology of Pd surfaces. Heemeier et al.reported rearrangement of Co, Rh, and Pd particlesdeposited on thin alumina films grown on a NiAlsubstrate at 603-903 K.36 The metal particles sinteredand migrated into the substrate, causing the surfacerearrangement. Pd nanoparticles also grew on CeO2 andTiO2 supports during CH3OH synthesis from CO andH2 at 473-873 K.37,38 Coke did not deposit significantlyon Pd/CeO2, and particle growth was also explained bysintering. The electroless-plated Pd-Cu hillocks in thisstudy enlarged at temperatures as low as 523 K, andmetal sintering may increase the grain size duringintermetallic diffusion of Pd and Cu. Before annealing,grain sizes of electroless-plated Pd-Cu membraneswere approximately 20-50 nm.39 Grain sizes of ap-proximately 100 nm were detected after 7 days in H2at 723 K. No increase in the grain size was seen afterCO2 and CO exposure. Furthermore, the reversibilityof the hillock growth after H2 or He exposure makesmetal sintering unlikely to be the cause for the hillockenlargement due to exposure of carbon oxides.

Catalytic Activity. Cu-based alloys are used com-mercially as WGS and CO oxidation catalysts. Thereverse WGS conversion was 10 times higher on a Pd75-Cu25 membrane than on a pure Pd membrane. There-fore, Pd segregation to the membrane feed side duringCO2 and CO exposure decreased the membrane catalyticactivity. The catalytic reactivity of the Pd-Cu mem-brane may also explain the decrease in H2 permeancewhen CO2 was first introduced to H2 at low tempera-tures (Figure 9). CO2 may react with the dissociated Hand C atoms on the membrane surface. The reaction ofH atoms on the surface decreases the amount of Havailable to permeate through the membrane. In con-trast, removing C from the surface through the produc-tion of CO (eq 4) increased the H2 permeation byincreasing hillock heights and defect diameters. Intro-ducing CO2 also segregated Pd to the membrane surface,decreased the catalytic activity, and increased the H2solubility over time.

Conclusions

Pd-Cu alloy membranes separated H2 from a mixtureof CO2, CO, and H2O. Over time in the presence of COand CO2, valleys on the membrane surface expandedinto membrane defects because of coalescence of hill-ocks, and molecules other than H2 then transportedthrough the membrane, which decreased the selectivity.CO2 increased the hillock heights by a factor of 6 onPd-Cu films and a factor of 2.5 on Pd-Cu foils. Thewidths of the valleys were reduced and the selectivityincreased with membrane exposure to He and H2.Decreasing the membrane surface roughness and in-creasing the membrane thickness decreased the mem-brane defect formation associated with surface rear-rangement. Surface topology changes are partially dueto the removal of C impurities by CO2 to form CO.

Furthermore, the hillocks and valleys in the films andfoils may also form because of stress relaxation by atomdiffusion as seen in Pd segregation to the feed side ofthe membrane during exposure to CO and CO2. Fewervacancies and lattice defects in the alloy lattice maymake the foil more resistant to atom rearrangementthan the electroless-plated films. Pd segregation in thepresence of different gases altered the surface regionand bulk alloy composition of the membranes. Changingthe phase structure from bcc to fcc decreased the H2permeability. The fcc structure also has a 1.3 timesbigger lattice parameter and a lower atomic densitythan the bcc structure. The lattice rearrangement maychange the surface morphology of the membrane. Theuse of Pd-Cu alloys instead of pure Pd increased sidereaction conversion.

Acknowledgment

We gratefully acknowledge support from the CO2Capture Project Consortium cofunded by the U.S.Department of Energy (U.S. DOE), the European Union,BP, and Norwegian Klimatek Agencies under ContractDE-FC26-01NT41145, Grant DE-FG26-03NT41792 fromthe DOE University Coal Research program, and GrantDE-FG03-99ER14363 from the U.S. DOE Office of BasicEnergy Sciences to J.D.W. We also thank Dr. DavidEdlund of Idatech, Inc., and Oremet Wah Chang, Inc.,for donating the Pd-Cu foils and Michael Block and Dr.Fernando Roa of the Colorado School of Mines forvaluable discussions.

Literature Cited

(1) Shu, J.; Grandjean, B. P. A.; Kaliaguine, S.; Ciavarella, P.;Giroir-Fendler, A.; Dalmon, J. A. Gas permeation and isobutanedehydrogenation over very thin Pd/ceramic membranes. Can. J.Chem. Eng. 1997, 75, 712.

(2) Damle, A. S.; Gangwal, S. K.; Venkataraman, V. K. A simplemodel for a water gas shift membrane reactor. Gas Sep. Purif.1994, 8, 101.

(3) Kikuchi, E. Membrane reactor application to hydrogenproduction. Catal. Today 2000, 56, 97.

(4) Uemiya, S.; Sato, N.; Ando, H.; Matsuda, T.; Kikuchi, E.Steam reforming of methane in a hydrogen-permeable membranereactor. Appl. Catal. 1990, 67, 223.

(5) Basile, A.; Chiappetta, G.; Tosti, S.; Violante, V. Experi-mental and simulation of both Pd and Pd/Ag for a water gas shiftmembrane reactor. Sep. Purif. Technol. 2001, 25, 549.

(6) Uemiya, S.; Sato, N.; Ando, H.; Kikuchi, E. The water gasshift reaction assisted by a palladium membrane reactor. Ind. Eng.Chem. Res. 1991, 30, 585.

(7) Falconer, J. L.; Noble, R. D.; Sperry, D. P. In Catalyticmembrane reactors; Falconer, J. L., Noble, R. D., Sperry, D. P.,Eds.; Elsevier Science BV: Amsterdam, The Netherlands, 1995;p 669.

(8) Hsieh, H. P. In Inorganic membrane reactorssa review;Hsieh, H. P., Ed.; American Institute of Chemical Engineers: NewYork, 1989; Vol. 85, p 53.

(9) Berstein, L. A.; Lund, C. R. F. Membrane reactors forcatalytic series and series-parallel reactions. J. Membr. Sci. 1993,77, 155.

(10) Armor, J. N. Catalysis with permselective inorganic mem-branes. Appl. Catal. 1989, 49, 1.

(11) Itoh, N. In Simultaneous operation of reaction and separa-tion by a membrane reactor; Itoh, N., Ed.; Elsevier: New York,1990; Vol. 54, p 268.

(12) Uemiya, S.; Kato, W.; Uyama, A.; Kajiwara, M.; Kojima,T.; Kikuchi, E. Separation of hydrogen from gas mixtures usingsupported platinum-group metal membranes. Sep. Purif. Technol.2001, 22, 309.

(13) McKinley, D. L. Method for Hydrogen Separation andPurification. U.S. Patent 3,439,474, 1969.

Ind. Eng. Chem. Res., Vol. 43, No. 15, 2004 4197

(14) Amandusson, H.; Ekedahl, L. G.; Dannetun, H. Hydrogenpermeation through surface modified Pd and PdAg membranes.J. Membr. Sci. 2001, 193, 35.

(15) McKinley, D. L. Metal Alloy for Hydrogen Separation andPurification. U.S. Patent 3,350,845, 1967.

(16) Piper, J. Diffusion of hydrogen in copper-palladium alloys.J. Appl. Phys. 1966, 37, 715.

(17) Hunter, J. B. A new hydrogen purification process. Plati-num Met. Rev. 1960, 4, 130.

(18) Roa, F.; Block, M. J.; Way, D. The influence of alloycomposition on the H2 flux of composite Pd-Cu membranes.Desalination 2002, 147, 411.

(19) McCool, B. A.; Lin, Y. S. Nanostructured thin palladium-silver membranes: Effects of grain size on gas permeationproperties. J. Mater. Sci. 2001, 36, 3221.

(20) Edlund, D. J. Hydrogen-permeable metal membrane andmethod for producing the same. U.S. Patent 6,152,995, 2000.

(21) Paglieri, S. N.; Way, J. D. Innovations in palladiummembrane research. Sep. Purif. Methods 2002, 31, 1.

(22) Ward, T. L.; Dao, T. Model of hydrogen permeationbehavior in palladium membranes. J. Membr. Sci. 1999, 153, 211.

(23) Amandusson, H.; Ekedahl, L. G.; Dannetun, H. The effectof CO and O2 on hydrogen permeation through a palladiummembrane. Appl. Surf. Sci. 2000, 153, 259.

(24) Li, A.; Liang, W.; Hughes, R. The effect of carbon monoxideand steam on the hydrogen permeability of a Pd/stainless steelmembrane. J. Membr. Sci. 2000, 165, 135.

(25) Roa, F.; Way, D.; Paglieri, S. Preparation and character-ization of Pd-Cu membranes for hydrogen separation. Chem. Eng.J. 2003, 93, 11.

(26) Souleimanova, R.; Mukasyan, A. S.; Varma, A. Pd mem-branes formed by electroless plating with osmosis: H2 permeationstudies. AlChE J. 2002, 48, 262.

(27) Cheng, Y. S.; Pena, M. A.; Fierro, J. L.; Hui, D. C. W.;Yeung, K. L. Performance of alumina, zeolite, palladium, Pd-Agalloy membranes for hydrogen separation from towngas mixture.J. Membr. Sci. 2002, 5273, 1.

(28) Lin, Y.-M.; Rei, M.-H. Separation of hydrogen from the gasmixture out of catalytic reformer by using supported palladiummembrane. Sep. Purif. Technol. 2001, 25, 87.

(29) Subramanian, P. R.; Laughlin, D. E. In Binary Alloy PhaseDiagrams; Subramanian, P. R., Laughlin, D. E., Ed.; ASMInternational: Materials Park, Ohio, 1992; Vol. 3.

(30) Galuszka, J.; Pandey, R. N.; Ahmed, S. Methane conversionto syngas in a palladium membrane reactor. Catal. Today 1998,46, 83.

(31) d’Heurle, F. M. Metallurgical topics in silicon deviceinterconnections: thin film stresses. Int. Mater. Rev. 1989, 34, 53.

(32) Coble, R. L. A model for boundary diffusion controlled creepin polycrystalline materials. J. Appl. Phys. 1963, 34, 1967.

(33) Herring, C. Diffusional viscosity of a polycrystalline solid.J. Appl. Phys. 1950, 21, 437.

(34) Aggarwal, S.; Monga, A. P.; Perusse, S. R.; Ramesh, R.;Ballarotto, V.; Williams, E. D.; Chalamala, B. R.; Wei, Y.; Reuss,R. H. Spontaneous Ordering of Oxide Nanostructures. Science2000, 287, 2235.

(35) Varma, A.; Yeung, K. L.; Souleimanova, R. S.; Mukasyan,A. S. Novel approach for thin dense nanoscale-grained metal films.Ind. Eng. Chem. Res. 2002, 41, 6323.

(36) Heemeier, M.; Stempel, S.; Shaikhutdinov, S. K.; Libuda,J.; Baumer, M.; Oldman, R. J.; Jackson, S. D.; Freund, H. J. Onthe thermal stability of metal particles supported on a thinalumina film. Surf. Sci. 2003, 523, 103.

(37) Shen, W.-J.; Kobayashi, A.; Ichihashi, Y.; Matsumura, Y.;Haruta, M. Growth of Pd particles in methanol synthesis over Pd/CeO2. Catal. Lett. 2001, 73, 161.

(38) Bowker, M.; Stone, P.; Bennett, R.; Perskins, N. COadsorption on a Pd/TiO2 (110) model catalyst. Surf. Sci. 2002, 497,155.

(39) Paglieri, S. Palladium and Palladium-Copper CompositeMembranes for Hydrogen Separation. Ph.D. Dissertation, ColoradoSchool of Mines, Golden, CO, 1999.

Received for review December 4, 2003Revised manuscript received April 29, 2004

Accepted May 4, 2004

IE030853A

4198 Ind. Eng. Chem. Res., Vol. 43, No. 15, 2004