Journal of Membrane Science 243 (2004) 243–251

Permeation behavior of a H2/CO gaseous mixture throughPt-included composite membranes

Dong-Wook Leea,b, Seung-Eun Nama, Bongkuk Seaa, Son-Ki Ihmb, Kew-Ho Leea,∗a Membrane and Separation Research Center, Korea Research Institute of Chemical Technology, P.O. Box 107, Yuseong, Daejeon 305-606, South Korea

b Department of Chemical and Biomolecular Engineering, National Research Laboratory for Environmental Catalysis,Korea Advanced Institute of Science and Technology, 373-1 Guseong-dong, Yuseong-gu, Daejeon 305-701, South Korea

Received 19 February 2004; received in revised form 25 June 2004; accepted 25 June 2004Available online 20 August 2004

Abstract

To exceed the limitation of intrinsic separation performance derived from the pore structure of the skin layer of composite membranes,� permeationt thep )mt Op . Therefore,a skin layer,r©

K

1

tcmbTcrswa

ount

lessthan

tinumn offeedsuch

anes,ift re-avewithrchPtce to, thehanol

0d

-Al 2O3/Pt-SiO2/stainless steel (SUS) membranes were prepared by the soaking-rolling method. The sweeping gas method as aest was used under atmospheric pressure in both feed and permeate side. While the�-Al 2O3/SiO2(100 nm)/SUS membranes fabricated forurpose of comparison showed the Knudsen-dominated transport characteristics for the CO removal, permeation results of a H2(99%)/CO(1%ixture for the�-Al 2O3/Pt-SiO2/SUS membranes presented a remarkable H2/CO separation factor of 5.22–7.03 between 25 and 200◦C due

o the lower CO permeance. Considering surface diffusion induced by Pt catalysts placed under�-Al 2O3 skin layer, the driving force of Cermeation is dominantly determined by the adsorbed phase of CO as well as the gas phase of CO in the lower side of skin layerdecrease in driving force of the CO permeation was induced by the CO confinement (poisoning) on Pt in the lower side of the

esulting in enhancement of CO removal efficiency, exceeding pure Knudsen system expected from the mesoporous�-Al 2O3 layer.2004 Elsevier B.V. All rights reserved.

eywords:Permeation behavior; Platinum; Composite membrane

. Introduction

In recent years, there have been major improvements inhe power density of proton exchange membrane (PEM) fuelell stacks such that they are now considered to be one of theost promising options as a substitute to the internal com-

ustion engine for a number of transportation applications.he drive systems for cars using PEMFC can lead to in-reased energy usage efficiency and, especially, substantiallyeduced emissions compared with the best foreseeable driveystem based on the internal combustion engine. Methanol,ith ready conversion of the methane in natural gas beingvailable to that, is the best primary fuel for such systems,

∗ Corresponding author. Tel.: +82 42 860 7240; fax: +82 42 861 4151.E-mail address:khlee@krict.re.kr (K.-H. Lee).

when taking availability, costs and safety aspects into acc[1–3].

However, the PEMFC anode feed gas should containthan 100 ppm carbon monoxide (CO) and preferably less20 ppm, since the anode catalyst is usually based on pla(Pt), which is easily poisoned by even low concentratioCO. In order to diminish CO concentration in the anodegas, several physical and chemical purification methods,as pressure swing adsorption (PSA), inorganic membrorganic membranes, solvent absorption, water–gas shaction, methanation and preferential oxidation (PROX), hbeen studied. In addition, new electrocatalyst materialslower affinity for CO was investigated by several reseagroups[4–6]. They showed that bimetallic systems, withas one of the components, can give substantial toleranthe presence of CO in fuel stream. In our previous workmembrane process before feeding the products of met

376-7388/$ – see front matter © 2004 Elsevier B.V. All rights reserved.oi:10.1016/j.memsci.2004.06.027

244 D.-W. Lee et al. / Journal of Membrane Science 243 (2004) 243–251

steam reforming into PEMFC system was introduced to re-duce the CO concentration in the feed gas[7]. The silica com-posite membranes, which is the most promising membranefor hydrogen purification, were used for the CO removal andfollowed molecular-sieve like permeation mechanism result-ing in the decrease in CO concentration to under 80 ppm.

The gas separation mechanisms of porous inorganic mem-branes such as the silica membranes mainly consist of Knud-sen diffusion, surface diffusion and molecular sieving. Ofthese three main mechanisms, only surface diffusion andmolecular sieving can provide selectivity high enough tobe commercially used. In order to achieve high selectivity,pinhole-free silica membranes with high microporosity mustbe synthesized. Raman and Brinker reported the use of anorganic templated approach for silica membrane preparation[8]. The permeance of helium, carbon dioxide and methanefor their membrane were 4.4× 10−8, 6.8 × 10−8 and 9.5× 10−10 mol m−2 s−1 Pa−1, respectively. The microporoussilica membranes with hydrogen permeance of (4.7–17.7)× 10−7 mol m−2 s−1 Pa−1 and carbon dioxide permeance of(2.0–3.2)× 10−7 mol m−2 s−1 Pa−1 were fabricated by deVos and Verweij[9]. Wu et al. reported the production ofsilica membranes with similar qualities made by the CVDmethod[10]. Kusakabe et al. synthesized organic templatedmembranes with TEOS and alkyltriethoxysilanes resulting inh −7 −2 −1 −1

bmW anesg em ther ly-s vef anes[ ithfi gherh minam otedb par-t lu-mHb werec erformP tiono

thiss m-b nismi oft O gap ionals thep intro-d the

intermediate layer of the composite membrane, and couldgive rise to the improvement in CO removal efficiency over-coming the limitation of separation through the intrinsic porestructure of the skin layer. In order to observe the effect of COpartial pressure in the permeate side excluding CO oxidation,all of the permeation tests were carried out under oxygen-freecondition. Study on the adsorption and desorption of CO andH2 for the catalysts loaded in the membrane was conducted toinvestigate the permeation mechanism through the compositemembrane.

2. Experimental

2.1. Sol preparation

In order to modify the pore structure of porous stain-less steel supports, colloidal silica sols and a boehmitesol as materials for intermediate layer were prepared. Thecolloidal silica sol with particle size of 100 nm was syn-thesized from base-catalyzed hydrolysis-condensation re-action of tetraethyl orthosilicate (TEOS) purchased fromAldrich. Prior to the addition of the NH3/H2O mixture, theTEOS/ethanol mixture was stirred vigorously in an oil bathof 50◦C. The addition of NH3/H2O was carried out dropwisef ngr solwm itionv olarr sei)w itrica eac-t thev

2

thep facea ne-p ur-s ientw rates t4 zedb am-m ndp II)n rop-w oilb h( heP , andc stsw

ydrogen permeance of 1.0× 10 mol m s Pa , car-on dioxide permeance of 8.0× 10−9 mol m−2 s−1 Pa−1 andethane permeance of 2.8× 10−9 mol m−2 s−1 Pa−1 [11].est et al. prepared the sol–gel derived silica membr

overned by surface diffusion[12]. Costa et al. reported tholecular sieve silica membranes with the pore size in

egion of 3A resulting from the two-step catalyzed hydrois sol–gel process[13]. A number of research groups haocused on the sol–gel derived silica composite membr14–30]. Chai et al. found that the alumina membrane wne metal-dispersed alumina skin layer shows much hiydrogen separation factor than the conventional aluembrane, because the hydrogen permeability is promy surface diffusion induced by the dispersed fine metal

icles [31]. Uhlhorn et al. reported that Ag-containing aina and silica membrane exhibits the promoted CO2 and2 permeability due to surface diffusion[32,33]. A num-er of studies on organic-templated silica membranesonducted and showed enhancement of the membrane pance comparing with silica composite membranes[34,35].rimary objective of such studies is to exceed the limitaf intrinsic properties of the silica micropores.

A different conceptual membrane investigated intudy follows typical Knudsen diffusion mechanism coined with the concentration-driven separation mecha

nduced by intentionally controlling the partial pressurehe permeated gas, expected to be rejected, such as Croduced from the methanol steam reforming. The additeparation through the control of CO partial pressure inermeate side of the membrane could be achieved byuction of catalysts, which preferentially adsorb CO, into

-

s

ollowed by refluxing the mixture for 3 h at vigorous stirriesulting in the stable colloidal silica sol. The boehmiteas produced as suggested by Kusakabe et al.[36]. Poly-eric silica sol was prepared under acid-catalyzed cond

ia hydrolysis of TEOS and condensation reactions. A matio of TEOS, water and nitric acid (purchased from Junas 0.096/0.56/0.008. A mixture of TEOS, water and ncid was stirred at room temperature for 20 min. The r

ion mixture was diluted with additional water to adjustolume to 500 mL, and refluxed for 8 h at 80◦C [37].

.2. Preparation of Pt catalysts

Two platinum catalysts (C1 and C2) supported onrepared 100 nm silica xerogels, of which specific surrea is about 84 m2/g, were synthesized with tetraammilatinum(II) nitrate (purchased from Aldrich) as a precor. 3 wt.% Pt/silica catalysts C1 were prepared by incipetness impregnation with tetraammineplatinum(II) nitolution, followed by drying at 70◦C and calcination a00◦C. Pt/silica catalysts C2 (2 wt.%) were synthesiy means of sol–gel method with TEOS and tetraineplatinum(II) nitrate providing the source of silica alatinum. NH3 was mixed with the tetraammineplatinum(itrate and water mixture, and the solution was added dise to TEOS/ethanol mixture stirred vigorously in anath of 50◦C, followed by refluxing the mixture for 3TEOS:NH3:H2O:EtOH mole ratio = 1:5.4:53.6:6.1). Tt/silica catalysts C2 was dried with a rotary evaporatoralcined at 400◦C. Dispersion of platinum in the catalyas determined by chemisorption using CO and H2 as a

D.-W. Lee et al. / Journal of Membrane Science 243 (2004) 243–251 245

probe gas. In order to study on the desorption behavior ofPt/silica catalysts for CO and H2 pure gas, the temperature-programmed desorption (TPD) was carried out with gas chro-matography (GC-14B, SHIMADZU).

2.3. Preparation of composite membranes including Ptcatalysts

Disks of 316L porous stainless steel (SUS) purchased fromMatt Metallurgical has a thickness of 1 mm, a surface areaof 5 cm2, and an average pore size of 0.5�m. The porousSUS substrate was modified by 100 nm silica xerogels orPt/silica catalysts. The silica xerogels or Pt/silica catalystswere pressed into the macropores of one side of the stainlesssteel support by a press under 10 MPa, followed by calcina-tion at 650◦C. The whole procedure of the first modificationwas repeated twice. The second modification of the stainlesssteel support was conducted by the soaking-rolling methodwith a boehmite sol[38]. The back-side of the support loadedon theo-ring-sealed cell was vacuumed by a rotary vacuumpump. The boehmite sol was poured onto the front side ofthe support, followed by maintaining vacuum of the backside of the support for 3 min so that the boehmite sol couldpenetrate into inner pores of the support. After the soakingprocess, concentrated gel layer formed on the front side oft Them da enta ft thes mites da sings ps).

2

)a 5 and2 cmT d bya ely.T sideo molef werem niza-t the

TV d

Mc

in

MMMM

extremely low concentration of CO. The parameter to de-scribe the separation efficiency for a binary mixture is theseparation factorα, which is a measurement of the enrich-ment of a gas component after it has passed the membrane.Separation factorα is defined as:

α = y

1 − y· 1 − x

x

wherex andy are the mole fraction of the feed and per-meate, respectively.

3. Results and discussion

3.1. Permeation behavior of a composite membranewithout Pt catalysts

In this study, additional separation of CO through the con-trol of CO partial pressure in the permeate side of the mem-brane was induced by introduction of Pt/silica catalysts intothe intermediate layer of the composite membrane. The var-ious �-Al2O3 membranes shown inTable 1were used toinvestigate the effect of Pt/silica catalysts on the CO removalefficiency and permeation behavior through the compositemembrane containing the catalysts, because when the mem-b of COc o rel-abb poresw em-b s aref

( inetP nceo aledt1 )1C . Thet l per-m gion[ Oc erme-a n-t wasd d theH sep-a thep eredt send owt med

he support was rolled out with an urethane rolling pin.odified support was dried overnight at 25◦C and calcinet 650◦C. The various modified supports including differmount of platinum are listed inTable 1. The preparation o

op layer using polymeric silica sol was carried out byame manner as the support modification with the boehol, and the silica top layer was dried at 25◦C and calcinet 500◦C. The membrane morphology was observed ucanning electron microscopy (SEM; XL 30S FEG, Phili

.4. Permeation measurement

Permeation tests were conducted with H2(99%)/CO(1%nd H2(99%)/N2(1%) mixtures as a feed gas between 200◦C. The permeation area of the membrane was 4.522.he flow rates of the feed and sweeping gas controllemass flow controller were 40 and 80 ml/min, respectivhe sweeping gas of helium was fed to the permeatef the membrane and diluted the permeated gas. The

ractions of the permeated gases through the membraneeasured using gas chromatography with a corona io

ion detector (CID), which make it feasible to measure

able 1arious composite membranes prepared by the soaking-rolling metho

embraneode

Composition of the layers Weight of platinumthe membrane (mg)

1 �-Al2O3/SiO2(100 nm)/SUS –2 �-Al2O3/C1/SUS 0.263 �-Al2O3/C1+SiO2(100 nm)/SUS 0.0244 �-Al2O3/C2/SUS 0.18

ranes in the Knudsen region are used, measurementoncentration and removal efficiency become easy due ttively high CO permeance of the mesoporous�-Al2O3 mem-rane. Nitrogen adsorption of the unsupported�-Al2O3 filmy the BET (ASAP 2400) method showed the averageize of 4–5 nm and the specific surface area of 240 m2/g, byhich the Knudsen diffusion mechanism through the mrane can be speculated if no serious pinholes or crack

ormed on the�-Al2O3 layer.A mixture gas permeation test of�-Al2O3/SiO2

100 nm)/SUS membrane M1 was carried out to determhe separation performance of the�-Al2O3 layer withoutt/silica catalysts.Table 2shows the temperature dependef H2 and CO permeances for the membrane M1. It is reve

hat the membrane M1 with H2 permeances of (3.6–4.4)×0−7 mol m−2 s−1 Pa−1 and CO permeances of (1.1–1.3×0−7 mol m−2 s−1 Pa−1 showed a decrease in the H2 andO permeances with increasing permeation temperature

emperature dependence of the membrane M1 is typicaeation behavior appearing in the Knudsen diffusion re

39–43]. Table 3shows the H2/CO selectivity and the Concentration in the permeate side measured by the ption test of a H2(99%)/CO(1%) mixture. The CO conce

ration of 10,000 ppm in the feed side of the membraneiminished to 3000–3030 ppm in the permeate side, an2/CO selectivity is about 3.4 approaching to the idealration factor for pure Knudsen diffusion system. Fromermeation results of the membrane M1, it can be consid

hat the membrane M1 predominantly follows the Knudiffusion mechanism with a little contribution of viscous fl

hrough pinholes or cracks which can be occasionally for

246 D.-W. Lee et al. / Journal of Membrane Science 243 (2004) 243–251

Table 2H2 and CO permeances of the various composite membranes for permeation test of the H2(99%)/CO(1%) mixture with different permeation temperatures

Membrane code H2 permeances with permeation temperature(mol m−2 s−1 Pa−1)

CO permeances with permeation temperature(mol m−2 s−1 Pa−1)

25◦C 100◦C 200◦C 25◦C 100◦C 200◦C

M1 4.38× 10−7 3.98× 10−7 3.58× 10−7 1.32× 10−7 1.19× 10−7 1.07× 10−7

M2 4.45× 10−7 4.31× 10−7 4.54× 10−7 8.51× 10−8 6.97× 10−8 6.45× 10−8

M3 4.46× 10−7 4.48× 10−7 4.46× 10−7 1.03× 10−7 9.03× 10−8 8.51× 10−8

M4 4.35× 10−7 4.41× 10−7 4.38× 10−7 8.05× 10−8 6.70× 10−8 5.80× 10−8

Table 3Permeation results of the H2(99%)/CO(1%) mixture for the various composite membranes with permeation temperature

Membrane code CO concentration in the permeate side withpermeation temperature (ppm)

H2/CO selectivity with permeationtemperature (−)

25◦C 100◦C 200◦C 25◦C 100◦C 200◦C

M1 3030 3000 3010 3.33 3.34 3.35M2 1930 1630 1435 5.22 6.19 7.03M3 2326 2030 1924 4.33 4.96 5.24M4 1865 1532 1335 5.40 6.58 7.55

on the membrane depending on preparation conditions of themembrane. As presumed from the structural properties ofthe unsupported�-Al2O3 layer, the separation factor for themembrane M1, containing no materials with the chemicalaffinity for permeated gases such as H2 and CO, cannot ex-ceed the ideal separation factor for pure Knudsen diffusionsystem, even though crack-free�-Al2O3 layer is successfullyobtained.

3.2. Permeation behavior of a composite membrane withPt catalysts

To investigate the improvement of H2/CO separation fac-tor by the introduction of Pt/silica catalysts with the chemicalaffinity for H2 and CO gases, the�-Al2O3/C1/SUS mem-brane M2 with Pt catalysts was prepared by the soaking-rolling method. 3 wt.% Pt catalysts C1 supported on the silicaxerogel with particle size of 100 nm were used to modify theporous stainless steel support instead of the silica xerogelof 100 nm. Pt/silica catalysts C1 were pressed into the poresof the substrate under 10 MPa, followed by calcination at650◦C. Fig. 1 shows the surface morphology of the well-modified stainless steel support with the catalysts C1. Thepore size of the stainless steel substrate with rough surface anda wide pore size distribution was successfully diminished byt tiono lingm sup-pm raneM ant 1t ilicac

The permeation test of a H2(99%)/CO(1%) mixture be-tween 25 and 200◦C was conducted to observe the effect ofPt catalysts included in the membrane on CO removal effi-ciency and permeation behavior for H2 and CO. To excludethe CO removal via CO oxidation on the Pt/silica catalysts,all of the permeation tests were carried out under oxygen-freecondition using a sweeping gas.Table 2shows a change of H2and CO permeances by the Pt/silica catalysts C1, and the tem-perature dependence of the permeances of the membrane M2.Comparing to the permeation results of the membrane M1,the membrane M2 showed a little change of the H2 perme-ance. However, the H2 permeance of the membrane M2 wasunexpectedly independent of the permeation temperature de-spite Knudsen-dominated pore structure of the�-Al2O3 layer,while the membrane M1 showed a decrease in the H2 per-meance with an increase in permeation temperature.Table 3shows the H2/CO separation factor of the membrane M2 and

F silicac

he well-packed Pt/silica catalysts. The second modificaf the support with the boehmite sol via the soaking-rolethod was carried out to decrease the pore size of theort into the region of Knudsen diffusion.Fig. 2shows SEMicrographs of surface and cross-section of the memb2. �-Al2O3 layer formed several micrometer lower th

he surface of the stainless steel substrate was about�mhick and was partly overlapped with the layer of the Pt/satalysts C1.

ig. 1. A SEM image of surface of the SUS substrate modified by the Pt/atalysts.

D.-W. Lee et al. / Journal of Membrane Science 243 (2004) 243–251 247

Fig. 2. SEM images of the membrane M2: (a) surface; (b) cross-section.

the CO concentration in the permeate side. The CO concen-tration of 10000 ppm in the feed side of the membrane wasreduced to 1435–1930 ppm in the permeate side which ismuch lower than that for the membrane M1, and a remark-able H2/CO separation factor of 5.22–7.03 was obtained inthe range of permeation temperature from 25 to 200◦C. Inter-estingly, the permeation results of the membrane M2 contain-ing the Pt/silica catalysts C1 exceeded the limit of separationperformance expected from the mesoporous�-Al2O3 layer.To investigate permeation behavior under CO-free condition,a H2(99%)/N2(1%) mixture gas permeation test was alsoconducted between 25 and 200◦C. Fig. 3 shows N2 and H2permeances of the membrane M2 for the H2(99%)/N2(1%)mixture gas permeation test. Comparing to permeation re-sults of the H2(99%)/CO(1%) mixture, the H2 permeancewas decreased, and the N2 permeance was higher than theCO permeance. In addition, a H2/N2 selectivity was lowerthan that of the pure Knudsen diffusion system. From theresults, it is considered that the H2/CO separation factor ex-ceeding pure Knudsen diffusion system must be attributedto additional separation behavior induced by interaction be-tween CO and Pt/silica catalysts, combined with Knudsendiffusion mechanism.

Fig. 3. H2 and N2 permeances of the membrane M2 for H2(99%)/N2(1%)mixture gas permeation test with permeation temperature.

3.3. Study on permeation mechanism of the compositemembrane with Pt catalysts

Chemisorption and temperature-programmed desorption(TPD) of the Pt/silica catalysts for CO and H2 were carriedout to study the additional separation mechanism resulting inthe improvement of CO removal efficiency. The TPD curvesat coverage of saturation for CO and H2 from the 3 wt.%Pt/silica catalysts C1 are shown inFig. 4. The TPD spec-tra were for adsorption at 25◦C. As shown inFig. 4(a), twodesorption peaks of H2 from the catalysts C1 are seen at 60and 80◦C. The CO TPD result seen inFig. 4(b) shows that asingle peak appeared at 240◦C, and the starting temperature

F ositem

ig. 4. TPD results of the Pt/silica catalyst C1 included in the compembrane: (a) for H2 and (b) for CO.

248 D.-W. Lee et al. / Journal of Membrane Science 243 (2004) 243–251

Table 4Chemisorption results of the Pt/silica catalysts for CO and H2 (analysistemperature = 35◦C)

Catalyst code Dispersion (%) Adsorbed volume perweight of Pt (cm3/g)

CO H2 CO H2

C1 51.8 19.1 57.3 21.3C2 84.6 34.1 97.0 39.2

of CO desorption was about 170◦C. Chemisorption resultsof the Pt/silica catalysts C1 and C2 for H2 and CO are shownin Table 4. Dispersion and the adsorbed gas volume on thePt/silica catalysts for CO were much higher than those for H2.Table 5shows the adsorbed gas volume on the Pt/silica cat-alysts loaded in the membranes, which was calculated fromthe results appeared inTables 1 and 4. The adsorbed vol-ume of CO on the catalysts included in the membrane M2was about three times larger than the adsorbed volume of H2.FromTables 4 and 5, it can be considered that active sites ofthe Pt/silica catalysts for CO are dominant over those for H2.

For the sweeping gas method used in this study, perme-ation and separation through the membranes are determinedby the partial pressure difference of each gas, as a drivingforce of the gas permeation, between the upper and lower sideof the active skin layer. The permeation behavior of the Pt-included composite membrane could be described in terms ofdriving force or in terms of average residence time. However,based on the CO poisoning shown in the CO TPD result forPt catalysts, CO permeation mechanism is difficult to be de-scribed in terms of CO residence time through the membrane,because CO starts to desorb at about 170◦C. Considering thecontribution of surface diffusion induced by the Pt catalystsplaced under�-Al2O3 skin layer, it is noted that the drivingforce of CO permeation is dominantly determined by the ad-s in thel hav-i vingf oft d bys ationes

2w e them h

TT them

M

MMMM

increasing permeation temperature, which is the typical tem-perature dependence of pure Knudsen system. The temper-ature independence of H2 permeance for the membrane M2is attributed to an increase of H2 permeance resulting froman increase in desorption rate of H2 from the Pt catalystswith permeation temperature, giving rise to the compensa-tion effect for the decrease tendency of H2 permeance dueto Knudsen diffusion through the�-Al2O3 layer. In otherwords, surface diffusion induced after H2 permeated throughthe �-Al2O3 skin layer contributed to the temperature in-dependence of H2 permeance. However, comparing to theH2 permeances of the membrane M1, H2 permeances of themembrane M2 were not significantly increased despite thesurface diffusion, because the surface diffusion was inducedafter H2 permeated through the�-Al2O3 skin layer.

In case of CO permeances shown inTable 2, both of themembranes M1 and M2 showed the decrease tendency of COpermeance with permeation temperature, however, CO per-meances of the membrane M2 between 25 and 200◦C wasmuch lower than those of the membrane M1. Differently fromthe temperature independence of H2 permeance for the mem-brane M2, the CO permeance was decreased with permeationtemperature, because CO, which never desorbs from the Ptcatalysts at low temperature as shown inFig. 4(b), does notshow the increase in desorption rate of CO with permeationt per-m at is,i n thec COp -c 00i orbf ncs sys-t thant M1,t aneM overH ceo ithK par-t re itc n int ase ind enti

ttta of theH COi ivingf se int f the

orbed phase of CO as well as the gas phase of COower side of skin layer. Therefore, CO permeation beor should be described in terms of partial pressure (driorce) of CO in the lower side of the skin layer. In casehe H2 permeation, permeation behavior can be explaineimultaneously considering both residence time (retardffect) and partial pressure (driving force) of H2 in the loweride of the skin layer.

As shown inTable 2, H2 permeance of the membrane Mas nearly constant with permeation temperature, whilembrane M1 showed the decrease in H2 permeance wit

able 5he volumes of adsorbed H2 and CO on the Pt/silica catalysts loaded inembrane (calculated fromTables 1 and 4)

embrane code Adsorbed volumes of each gas on Ptincluded in the membranes (cm3)

CO H2

1 – –2 1.49× 10−2 0.55× 10−2

3 0.14× 10−2 0.05× 10−2

4 1.75× 10−2 0.71× 10−2

emperature. As a result, only decrease tendency of COeance with permeation temperature was observed. Th

n the case of CO, the compensation effect as shown iase of the H2 permeance was not observed due to theoisoning on the Pt catalysts below 170◦C. However, the derease gap of CO permeance for the membrane M2 at 2◦Cs smaller than that at 100◦C, because CO starts to desrom the Pt/silica catalysts at 170◦C and the CO desorptioontributes to an increase in the CO permeance. The H2/COelectivity of the membrane M2 exceeding pure Knudsenem is dominantly attributed to the lower CO permeancehat of the membrane M1. Comparing to the membranehe ratio of H2 to CO in the permeate side of the membr2 was decreased due to dominant adsorption of CO2 presented inTables 4 and 5. The decrease in driving forf the CO permeation via surface diffusion combined wnudsen diffusion was induced by the increase of CO

ial pressure in the lower side of the skin layer. Therefoan be concluded that the confinement of CO populatiohe permeate side of the membrane resulted in the decreriving force of CO permeation and allowed the improvem

n CO removal efficiency.For the permeation test of the H2(99%)/N2(1%) mix-

ure for the membrane M2 shown inFig. 3, comparing tohe H2(99%)/CO(1%) permeation results shown inTable 2,he H2 permeances were decreased, and the N2 perme-nces were higher than the CO permeances. In case2(99%)/CO(1%) permeation, dominant adsorption of

n the lower side of skin layer leads to the decrease in drorce of the CO permeation, resulting in a relative increahat of H2 permeation. However, when the permeation o

D.-W. Lee et al. / Journal of Membrane Science 243 (2004) 243–251 249

H2(99%)/N2(1%) mixture is conducted under CO-free con-dition, partial pressure of the permeated gases in the lowerside of skin layer is significantly different from that of theH2(99%)/CO(1%) permeation for the membrane M2. That is,the permeation of the H2/N2 mixture for the membrane M2leads to dominant adsorption of H2 in the lower side of skinlayer, resulting in an increase in average residence time ofH2. Therefore, transport characteristics of the membrane M2exceeding the Knudsen diffusion system cannot be expectedin case of the H2/N2 binary mixture, although the Pt catalystsare included in the intermediate layer of the composite mem-brane. In addition, comparing with the permeation results ofthe membrane M1 without Pt catalysts shown inTable 2, theH2 permeances were also decreased due to the retardation ofthe H2 permeation induced by adsorption and desorption ofH2 on Pt catalysts under the�-Al2O3 skin layer. The tempera-ture independence of the H2 permeance for the membrane M2was attributed to the increase in desorption rate of H2 fromthe Pt catalysts with permeation temperature. Interestingly,the decrease in H2 permeances derived from the retardationeffect is observed under CO-free permeation condition, whenPt catalysts are placed below the active skin layer. While thehydrogen-selective catalysts contained in and on the activeskin layer lead to the increase in H2 permeance due to thesurface diffusion[32–35], the catalysts below the active layercm (sur-f erg inc 2,a mt ayer,H aringt f theH se ind theP

3C

dedi , them yt andC -a r Ptd Pt int M4,t braneM er Ptd f them cat-a ingi them cat-

alyst C1.Table 5shows that adsorbed volume of each gas onthe Pt catalysts in the membranes is proportional to amount ofthe exposed surface Pt.Tables 2 and 3present the permeationresults of the membrane M3 with permeation temperature.Temperature dependence of the H2 and CO permeance forthe membrane M3 is almost in agreement with that for themembrane M2. However, CO permeance of the membraneM3 was increased comparing to the membrane M2, resultingin a decrease in H2/CO separation factor and an increase inCO concentration in the permeate side. The results indicatethat the decrease of the Pt amount loaded in the membranegives rise to the decrease of CO removal efficiency. As theamount of Pt included in the membrane is decreased, theportion of permeated gas interacting with the Pt catalysts isdiminished. The decrease of CO adsorbed on the Pt catalystscauses a decrease in CO partial pressure in the lower side ofthe skin layer, resulting in an increase of driving force for theCO permeation via surface diffusion combined with Knudsendiffusion. On the other hand, as the Pt amount contained inthe membrane is decreased, H2 partial pressure in the lowerside of the skin layer is relatively increased, and the retarda-tion phenomenon of H2 permeation induced by adsorptionand desorption is decreased. The decrease of H2 permeancedue to the increase in H2 partial pressure in the lower sideof the active layer was compensated for the increase of H2p n ef-f Hp t witht

M4s Ha istentw them n theC -t ease

F il-i eationt

ontribute to the decrease in H2 permeance except the H2/COixture permeation, because adsorption and desorption

ace diffusion) of H2 after permeating through the skin layive rise to the retardation of H2 permeation. However,ase of H2(99%)/CO(1%) permeation for the membrane Mlthough the decrease in the H2 permeance was derived fro

he retardation effect by the Pt catalysts under the skin l2 permeaces were not substantially decreased comp

o those for the membrane M1. Because an increase o2 permeance was also induced by the relative increariving force of H2 permeation due to CO poisoning ont catalysts.

.4. Effect of Pt amount in the composite membrane onO removal efficiency

In order to investigate the effect of Pt amount inclun the composite membrane on CO removal efficiency

embranes M3 and M4 shown inTable 1were prepared bhe soaking-rolling method with different catalysts C12, respectively. As presented inTable 4, the Pt/silica catlyst C2 fabricated by sol–gel method exhibited higheispersion than the catalyst C1. Although the amount of

he membrane M2 was more than that in the membranehe amount of exposed Pt, as an active site, in the mem2 was less than that in the membrane M4 due to highispersion of the catalyst C2. The intermediate layer oembrane M3 was prepared by the modification with thelyst C1 following that with 100 nm silica xerogel, result

n less amount of Pt included in the membrane M3 thanembrane M2 fabricated by twice modification with the

ermeance derived from the decrease in the retardatioect of H2 permeation. The compensation effect allowed2ermeance of the membrane M3 to be almost consisten

hat of the membrane M2.According to the permeation results of the membrane

hown inTables 2 and 3, temperature dependence of the2nd CO permeance for the membrane M4 is also consith that for the membrane M2. However, comparing withembrane M2, the membrane M4 exhibit a decrease iO permeance and an increase in the H2/CO separation fac

or. Contrary to the case of the membrane M3, the incr

ig. 5. Permeation results of a H2(99%)/CO(1%) mixture gas for the sca composite membrane supported on the membrane M4 with permemperature.

250 D.-W. Lee et al. / Journal of Membrane Science 243 (2004) 243–251

in surface Pt included in the membrane led to an increaseof CO adsorbed on the Pt catalysts, resulting in the lowerCO permeance due to the decrease in driving force of COpermeation. Consequently, it can be concluded that the COpermeance is inversely proportional to the amount of surfacePt loaded in the membrane, although the H2 permeance doesnot show the significant dependence upon the Pt amount inthe membrane.Fig. 5 presents the permeation results of thesilica composite membrane supported on M4 exceeding theKnudsen-dominated transport characteristics for the CO re-moval. CO was not detected in the permeate side of the mem-brane by the corona ionization detector (CID). The results areattributed to additional separation behavior induced by the Ptcatalysts loaded in the composite membrane, combined withmolecular sieving derived from the micropores of the silicatop layer.

4. Conclusions

The Pt-included composite membranes (M2-4) exceededthe limit of H2/CO separation performance expected fromthe mesoporous�-Al2O3 layer at permeation temperature be-tween 25 and 200◦C, while the membrane M1 without the Ptcatalysts showed the Knudsen-dominated transport charac-t rd-i Ht de-s int e Ptc byb sure(

f inde-p send Hpa by theP r them ew e ofH be-l sd f theP hoseo forceo iths rtialp

ureK easei artialp rds,t e oft CO

permeation and allowed the improvement of CO removal effi-ciency. In addition, the CO permeance was inversely propor-tional to the amount of surface Pt included in the membrane,while H2 permeance was not dependent on the Pt amount inthe membrane.

References

[1] B. Emonts, J.B. Hansen, S.L. Jørgensen, B. Hohlein, R. Peters, Com-pact methanol reformer test for fuel-cell powered light-duty vehicles,J. Power Sources 71 (1998) 288–293.

[2] F. Panik, Fuel cells for vehicle applications in cars—bringing thefuture closer, J. Power Sources 71 (1998) 36–38.

[3] J.C. Amphlett, R.F. Mann, B.A. Peppley, On board hydrogen pu-rification for steam reformation/PEM fuel cell vehicle power plants,Int. J. Hydrogen Energy 21 (1996) 673–678.

[4] C. Lu, R.I. Masel, The effect of ruthenium on the binding of CO,H2, and H2O on Pt(1 1 0), J. Phys. Chem. B 105 (2001) 9793–9797.

[5] D.D. Papageorgopoulos, M. Keijzer, F.A. de Bruijn, The inclusionof Mo, Nb and Ta in Pt and PtRu carbon supported 3electrocatalystsin the quest for improved CO tolerant PEMFC anodes, Electrochim.Acta 48 (2002) 197–204.

[6] C. Lu, C. Rice, R.I. Masel, P.K. Babu, P. Waszczuk, H.S. Kim, E.Oldfield, A. Wieckowski, UHV, electrochemical NMR, and electro-chemical studies of platinum/ruthenium fuel cell catalysts, J. Phys.Chem. B 106 (2002) 9581–9589.

[7] D.-W. Lee, B.K. Sea, K.-Y. Lee, K.-H. Lee, Preparation and charac-gen

cular.nes

[ ar-em-

[ re ofoxysi-146.

[ is,mem-

[ ularion of. 198

[ aneMater.

[ no,siev-tivity,

[ ren-ilica

ir 18

[ m aated-998)

[ oka,

eristics for the H2(99%)/CO(1%) permeation test. Accong to TPD results of the Pt/silica catalysts for CO and2,he extraordinary behavior of CO permeation should becribed in terms of partial pressure (driving force) of COhe lower side of the skin layer due to CO poisoning on thatalysts, while H2 permeation behavior can be explainedoth residence time (retardation effect) and partial presdriving force) of H2 in the lower side of the skin layer.

The permeation results of the H2(99%)/CO(1%) mixtureor the Pt-included membranes showed the temperatureendence of the H2 permeance, being attributed to Knudiffusion combined with surface diffusion induced after2ermeated through the�-Al2O3 skin layer. The H2 perme-ces were not decreased despite the retardation effectt catalysts under the skin layer, comparing to those foembrane M1. Because an increase of the H2 permeancas also induced by the relative increase in driving forc2 permeation due to CO poisoning on the Pt catalysts

ow 170◦C, although the decrease in the H2 permeance waerived from the retardation effect. CO permeances ot-included membranes were substantially lower than tf the membrane M1, because the decrease in drivingf the CO permeation via Knudsen diffusion combined wurface diffusion was induced by the increase of CO paressure in the lower side of the skin layer.

Therefore the H2/CO separation factor exceeding the pnudsen diffusion system is dominantly due to the decr

n the CO permeance induced by the increase of CO pressure in the lower side of the skin layer. In other wo

he confinement of CO population in the permeate sidhe membrane led to the decrease in driving force of the

terization of SiO2 composite membranes for purification of hydrofor PEMFC, Ind. Eng. Chem. Res. 41 (2002) 3594–3600.

[8] N.K. Raman, C.J. Brinker, Organic template approach to molesieving silica membranes, J. Membr. Sci. 105 (1995) 273–279

[9] R.M. de Vos, H. Verweij, Improved performance of silica membrafor gas separation, J. Membr. Sci. 143 (1998) 37–51.

10] J.C.S. Wu, H. Sabol, G.W. Smith, D.L. Flowers, P.K.T. Lin, Chacterisation of hydrogen-permselective microporous ceramic mbranes, J. Membr. Sci. 96 (1994) 275–287.

11] K. Kusakabe, S. Sakamoto, T. Saie, S. Morooka, Pore structusilica membranes formed by a sol–gel technique using tetraethlane and alkyltri ethoxysilanes, Sep. Purif. Tech. 16 (1999) 139–

12] G.D. West, G.G. Diamond, D. Holland, M.E. Smith, M.H. LewGas transport mechanisms through sol–gel derived templatedbranes, J. Membr. Sci. 203 (2002) 53–69.

13] J.C.D. da Costa, G.Q. Lu, V. Rudolph, Y.S. Lin, Novel molecsieve silica (MSS) membranes: characterisation and permeatsingle-step and two-step sol–gel membranes, J. Membr. Sci(2002) 9–21.

14] Y.-S. Kim, K. Kusakabe, S.-M. Yang, Microporous silica membrsynthesized on an ordered mesoporous silica sublayer, Chem.15 (2003) 612–615.

15] B.-N. Nair, K. Keizer, H. Suematsu, Y. Suma, N. Kaneko, S. OT. Okubo, S.-I. Nakao, Synthesis of gas and vapor molecularing silica membranes and analysis of pore size and connecLangmuir 16 (2000) 4558–4562.

16] A.W. Verkerk, E.L.V. Goetheer, L.J.P. van den Broeke, J.T.F. Keutjes, Permeation of carbon dioxide through a microporous smembrane at subcritical and supercritical conditions, Langmu(2002) 6807–6812.

17] B.K. Sea, K. Kusakabe, S. Morooka, Hydrogen recovery froH2–H2O–HBr mixture utilizing silica-based membranes at elevtemperatures. Part. 2: Calculation of exergy losses in H2 separation using inorganic membranes, Ind. Eng. Chem. Res. 37 (12509–2515.

18] B.K. Sea, E. Soewito, M. Watanabe, K. Kusakabe, S. MoroS.S. Kim, Hydrogen recovery from a H2–H2O–HBr mixture utilizing

D.-W. Lee et al. / Journal of Membrane Science 243 (2004) 243–251 251

silica-based membranes at elevated temperatures. Part. 1: Preparationof H2O- and H2-selective membranes, Ind. Eng. Chem. Res. 37(1998) 2502–2508.

[19] S. Kim, G.R. Gavalas, Preparation of H2 permselective silica mem-branes by alternating reactant vapor deposition, Ind. Eng. Chem.Res. 34 (1995) 168–176.

[20] C.E. Megiris, J.H.E. Glezer, Synthesis of hydrogen-permselectivemembranes by modified chemical vapor deposition. Microstructureand permselectivity of silica/carbon/vycor membranes, Ind. Eng.Chem. Res. 31 (1992) 1293–1299.

[21] R.M. de Vos, H. Verweij, High-selectivity, high-flux silica mem-branes for gas separation, Science 279 (1998) 1710–1711.

[22] K. Kusakabe, F. Shibao, G. Zhao, K.-I. Sotowa, K. Watanabe, T.Saito, Surface modification of silica membranes in a tubular-typemodule, J. Membr. Sci. 215 (2003) 321–326.

[23] V. Richard, E. Favre, D. Tondeur, A. Nijmeijer, Experimental studyof hydrogen, carbon dioxide and nitrogen permeation through a mi-croporous silica membrane, Chem. Eng. J. (2001) 84.

[24] B.N. Nair, K. Keizer, W.J. Elferink, M.J. Gilde, H. Verweij, A.J.Burggraaf, Synthesis, characterization and gas permeation studieson microporous silica and alumina-silica membranes for separa-tion of propane and propylene, J. Membr. Sci. 116 (1996) 161–169.

[25] T. Yoshioka, E. Nakanishi, T. Tsuru, M. Asaeda, Experimental stud-ies of gas permeation through microporous silica membranes, AIChEJ. 47 (2001) 2052–2063.

[26] R.F.S. Lenza, W.L. Vasconcelos, Synthesis and properties of micro-porous sol–gel silica membranes, J. Non-Cryst. Solids 273 (2000)164–169.

[27] C.-Y. Tsai, S.-Y. Tam, Y. Lu, C.J. Brinker, Dual-layer asymmet-255–

[ m-.

[ .-I.es, J.

[ sisstudy178

[31] M. Chai, Y. Yamashita, M. Machida, K. Eguchi, H. Arai, Prepara-tion and characterization of metal-dispersed alumina membranes forselective separation of hydrogen, J. Membr. Sci. 97 (1994) 199–207.

[32] R.J.R. Uhlhorn, K. Keizer, A.J. Burggraaf, Gas and surface diffusionin modified �-alumina systems, J. Membr. Sci. 46 (1989) 225–241.

[33] R.J.R. Uhlhorn, M.H.B.J. Huis in’t Veld, K. Keizer, A.J. Burggraaf,High permselectivities of microporous silica-modified�-aluminamembranes, J. Mater. Sci. Lett. 8 (1989) 1135.

[34] G. Xomeritakis, S. Naik, C.M. Braunbarth, C.J. Cornelius, R. Pardey,C.J. Brinker, Organic-templated silica membranes: I. Gas and vaportransport properties, J. Membr. Sci. 215 (2003) 225–233.

[35] K. Kusakabe, S. Sakamoto, T. Saie, S. Morooka, Pore structure of sil-ica membranes formed by a sol–gel technique using tetraethoxysilaneand alkyltriethoxy silanes, Sep. Purif. Technol. 16 (1999) 139–146.

[36] K. Kusakabe, K. Ichiki, J.I. Hayashi, H. Maeda, S. Morooka, Prepa-ration and characterization of silica–polyimide composite membranescoated on porous tubes for CO2 separation, J. Membr. Sci. 115(1996) 65–75.

[37] M. Naito, K. Nakahira, Y. Fukuda, H. Mori, J. Tsubaky, Processconditions on the preparation of the supported microporous SiO2

membranes by sol–gel modification techniques, J. Membr. Sci. 129(1997) 263–269.

[38] D.-W. Lee, Y.-G. Lee, B.K. Sea, S.-K. Ihm, K.-H. Lee, Improvementin the thermal stability of stainless steel supported silica membranesby the soaking-rolling method, J. Membr. Sci. 236 (2004) 53–63.

[39] J.C.S. Wu, Feasibility of manufacturing hydrogen and styrenethrough the use of porous ceramic membranes, Ind. Eng. Chem.Res. 38 (1999) 4491–4495.

[40] K.-I. Okamoto, H. Kita, K. Horii, K. Tanaka, M. Kondo, ZeoliteNaA membrane: preparation, single-gas permeation, and pervapo-

Ind.

[ mble1999)

[ , Aacter-

[ ment485.

ric microporous silica membranes, J. Membr. Sci. 169 (2000)268.

28] R.M. de Vos, W.F. Maier, H.J. Verweij, Hydrophobic silica mebranes for gas separation, J. Membr. Sci. 158 (1999) 277–288

29] B.N. Nair, T. Yamaguchi, T. Okubo, H. Suematsu, K. Keizer, SNakao, Sol–gel synthesis of molecular sieving silica membranMembr. Sci. 135 (1997) 237–243.

30] B.N. Nair, W.J. Elferink, K. Keizer, H. Verweij, Sol–gel syntheand characterization of microporous silica membranes: SAXSon the growth of polymeric structures, J. Colloid Interf. Sci.(1996) 565–570.

ration and vapor permeation of water/organic liquid mixtures,Eng. Chem. Res. 40 (2001) 163–175.

41] D. Zhao, P. Yang, B.F. Chmelka, G.D. Stuky, Multiphase asseof mesoporous–macroporous membranes, Chem. Mater. 11 (1174.

42] N. Nishiyama, D.H. Park, A. Koide, Y. Egashira, K. Ueyamamesoporous silica (MCM-48) membrane: preparation and charization, J. Membr. Sci. 182 (2001) 235.

43] D.H. Park, N. Nishiyama, Y. Egashira, K. Ueyama, Enhanceof hydrothermal stability and hydrophobicity of a silica MCM-membranes by silylation, Ind. Eng. Chem. Res. 40 (2001) 610