Functionalized Carbon Molecular Sieve membranes containing Ag-nanoclusters

Journal of Membrane Science 219 (2003) 47–57

Functionalized Carbon Molecular Sieve membranescontaining Ag-nanoclusters

J.N. Barsemaa, J. Balstera, V. Jordanb, N.F.A. van der Vegta,∗, M. Wesslinga

a Membrane Technology Group, Faculty of Science and Technology, University of Twente,P.O. Box 217, 7500 AE, Enschede, The Netherlands

b Department of Chemical Engineering, Fachhochschule Münster, University of Applied Sciences, Steinfurt, Germany

Received 21 October 2002; received in revised form 20 March 2003; accepted 14 April 2003

Abstract

In Carbon Molecular Sieve (CMS) membranes, the separation of O2 and N2 is primarily based on the difference in sizebetween the gas molecules. To enhance the separation properties of these CMS membranes it is necessary to functionalizethe carbon matrix with materials that show a high affinity to one of the permeating gas species. Adding Ag-nanoclustersincreases the selectivity of O2 over N2 by a factor 1.6 compared to a non-functionalized CMS membrane prepared bythe same pyrolysis procedure. We have analyzed the structure of Ag-nanocluster (dcluster ≈ 50 nm) containing membranesproduced from different Ag sources, AgNO3 and AgAc, and with different Ag content (0, 6, 25, and 40 wt.%). By measuringthe pure gas permeabilities of He, CO2, O2, and N2 we have determined the effect of Ag-nanoclusters in the carbon matrix,concluding that in the case of pure gases, the Ag-nanoclusters act primarily as a spacer at pyrolysis end temperatures up to600◦C, increasing the O2 permeability by a factor of 2.4. However, they enhance the separation of O2 over N2 at higherpyrolysis end temperatures (700 and 800◦C). It was shown that the build up of an Ag layer on the surface of the membranereduces the permeability, but does not affect the selectivity.© 2003 Elsevier Science B.V. All rights reserved.

Keywords: Carbon Molecular Sieve; Ag; Nanocluster; Gas separation; Functionalized membrane

1. Introduction

Over the last decades, polymeric membranes haveproven to operate successfully in industrial gas sep-arations[1]. In certain separations, like the produc-tion of oxygen-enriched air, polymeric membranesstill lack sufficient selectivity to render their use eco-nomically attractive over conventional separations likecryogenic distillation or pressure swing adsorption.

∗ Corresponding author. Tel.:+31-53-4892962;fax: +31-53-4894611.E-mail address: [email protected](N.F.A. van der Vegt).

Carbon Molecular Sieves (CMSs) have shown perme-ation and separation properties significantly exceedingthose of their polymeric precursor membranes. In prin-ciple, two concurrently occurring separation mecha-nisms can be distinguished in CMS membranes. InCMS membranes with pores approaching the molec-ular diameters of the gases to be separated (<6 ×10−10 m (<6 Å)), the main mechanism is molecularsieving. The separation takes place on the basis ofsize exclusion and is therefore not dependent on porewall–gas molecule interactions nor on the feed pres-sure[2]. When the pore sizes of the CMS membraneare below 50×10−10 m (50 Å) a transition from Knud-sen to surface diffusion takes place[3]. Gas molecules

0376-7388/$ – see front matter © 2003 Elsevier Science B.V. All rights reserved.doi:10.1016/S0376-7388(03)00176-5

48 J.N. Barsema et al. / Journal of Membrane Science 219 (2003) 47–57

adsorb on the pore wall and travel through the poreby diffusion over the surface of the pore rather thanthrough the free volume of the pore itself. The mech-anism of surface diffusion becomes increasingly im-portant as the interaction between the pore wall andthe gas molecule increases. In the case of strongly ad-sorbing gases and small pores, pore blocking can oc-cur through selective adsorption of one of the species,which in certain cases can lead to pore condensation.

The separation of N2 and O2 takes place predom-inantly based on the first mentioned mechanism ofmolecular sieving. Both gas species posses little to noaffinity for the carbon matrix. Precise tailoring of thepore size has resulted in high separation factors forthese two gases[4,5], however, to achieve even higherseparation factors, it is necessary to go beyond thetailoring of pore sizes. By functionalizing the carbonmatrix with Ag-nanoclusters, we will show that it ispossible to increase the selectivity factor for the sep-aration of N2 and O2 compared to their CMS coun-terparts that were not functionalized. Metallic Ag iswell known for its interactions with O2 and is usedin composite inorganic membrane separations at ele-vated temperatures[6], in thin film Pd/Ag alloy densemetal membranes[7] and as catalyst in many oxida-tion reactions[8]. In contrary to the above-mentionedwork, we will perform our separations at room tem-perature. In the temperature range from 0 to 100◦C,the interactions between Ag and O2 can be classifiedas[9]:

• physical adsorption;• chemical adsorption;• solution (absorption);• bulk compound formation.

In physical and chemical adsorption processes,the reaction is confined to the surface of the metal.

Fig. 1. Structure of P84 co-polyimide.

They are distinguished by the nature of the surfaceforces. Between temperatures as low as−195◦C andtemperatures of 400◦C, O2 readily chemisorbs ontoAg-surfaces[10]. This implies that in the tempera-ture range between 0 and 100◦C both chemical andphysical adsorption take place. Generally, desorptionof chemically adsorbed O2 below 100◦C is negligi-ble [11], however, Czanderna[11] showed that evenat low temperatures, small amounts (<15%) of theO2 (i.e. chemically and physically adsorbed) eas-ily desorb simply by evacuation. The nature of theoxygen species adsorbed on the Ag-surface is gen-erally believed to be partly dissociative and partlynon-dissociative, e.g. O−, O2−, O2, O2

−, and O22−

[12]. These oxygen species form activated complexeswith a considerable higher mobility compared toadsorbed oxygen molecules[12].

In solution and bulk compound processes the gaspenetrates into the metal lattice. Often it can be veryimportant to distinguish between these four processes,but such a distinction is not always possible.

In this paper, we will present a method forthe preparation of CMS membranes containingAg-nanoclusters. These membranes are characterizedand their permeability and selectivity are comparedto non-functionalized CMS membranes.

2. Experimental

2.1. Materials

Both CMS and Ag-containing CMS (AgCMS)membranes were prepared using the commercial, ther-mally stable, co-polyimide (BTDA-TDI/MDI) P84from Lenzing as polymer precursor. The structure ofthe polymer is shown inFig. 1. TheTg of this polymer

J.N. Barsema et al. / Journal of Membrane Science 219 (2003) 47–57 49

Fig. 2. A schematic overview of the furnace set-up.

was determined with Differential Scanning Calorime-try (second run and 30◦C/min) and found to be315◦C. The theoretical carbon yield of this polymeris 54.2 wt.%, equal to the aromatic carbon content.

N-methyl-pyrrolidone (NMP) (Merck, 99%) wasused as the solvent. As additive, TriFluoroacetic Acid(TFA) (Merck–Schurchardt, 99%) was used. As Agsource, two different Ag salts were used: AgNO3(Merck, 99.8%) and AgAc (Merck, 99%).

All gases used for pyrolysis and gas permeationexperiments had a purity of at least 99.5%.

2.2. Precursor preparation

All precursor membranes were prepared based ona solution of 13 wt.% P84 in NMP. For the AgCMSprecursors, the Ag-containing salts were dissolved inthe NMP before the P84 was added. In the case ofAgAc, an equivalent of 1.2 mol TFA/mol of AgAcwas added to the solution to facilitate the solvation ofAgAc, forming AgTFA[13]. All solutions were madeand kept in the absence of light to prevent reductionof the Ag salt by UV radiation. The solutions wereallowed to dissolve overnight and were subsequentlyfiltered over a 25�m metal filter.

The solutions were cast on glass plates using a200�m casting knife and placed in a dry N2 box for3 days, then placed in a N2 oven for a further 3 daysat 50◦C to remove most of the solvent. The relativelow temperature during the removal of the solventwas chosen to prevent any thermal decomposition ofthe Ag salts prior to the pyrolysis process. The ob-tained precursor film had a thickness of approximately20�m.

2.3. Pyrolysis procedure

The pyrolysis is preformed using a Carbolite® TZF12/100 High Temperature Tube Furnace, mountedwith a Eurotherm 2408 CP temperature controller.Fig. 2 gives a schematic overview of the furnaceset-up. The precursor is placed in a quartz glass tube,using a stainless steel grid as sledge. The atmospherein the quartz tube during pyrolysis was N2, with aflow rate of 10 cm3/min.

Fig. 3 shows the pyrolysis trajectory for the prepa-ration of the CMS and AgCMS membranes. Afterreaching the end temperature, the membranes wereimmediately quenched into N2 to room temperaturein an external stainless steel double-hulled cooler. The

Fig. 3. A schematic overview of the pyrolysis trajectory and thetemperatures at which the CMS membranes are quenched.


membranes are cooled by flowing tap water throughthe outer hull of the cooler.

2.4. Analysis

The carbon yield, respectively, carbon Ag yield,defined as the weight relative to the initial precursorweight, was determined using Thermo GravimetricAnalysis (TGA). The TGA experiments were pre-formed using a Perkin-Elmer TGA 7 with a N2 atmo-sphere and flow rates of 20 cm3/min. Heating routeswere based on pyrolysis procedures.

To investigate the distribution of Ag and Ag-nano-clusters in the AgCMS structure, Scanning ElectronMicroscopy (SEM) was used to look at both thecross-section and surface of the membranes. The mi-crographs were obtained using a JEOL JSM T220ASEM for magnifications up to 25,000× and a JEOLJSM 5600 LV SEM for higher magnifications. Theprecursor polymeric materials were coated with plat-inum in an argon atmosphere using a JEOL JFC1300 (20 mV, 90 s) before they were placed in themicroscope. To obtain a more detailed picture of thesurface of the produced membranes, Atomic ForceMicroscopy—Nanoscope 3A from Digital Instru-ments—was used.

To elucidate the CMS and AgCMS microstructure,pure gas permeation experiments were conducted.The membranes were glued into metal discs, using anAraldit two component adhesive, to prevent crackingof the membranes by the rubber ring seals. The gaspermeation experiments were done, using a variablepermeate pressure set-up. In the experimental set-up,a vacuum was applied on the permeate side, whereasthe other side was brought into contact with the feedgases. All measurements were taken after the fluxes

Table 1Composition of precursor solution, precursor film, and AgCMS membranes

Name Ag source P84 in solution(wt.%)

Ag in solution(wt.%)a

Ag in precursor(wt.%)a

Ag in AgCMS(wt.%)a,b

MP1 – 13.04 – – –MAg1 AgNO3 12.94 0.49 3.59 6.21MAg2 AgNO3 12.56 2.36 14.5 24.64MAg3 AgNO3 12.11 4.56 23.65 39.59MAg4 AgAc 12.86 0.50 3.63 6.28

a Based only on the Ag content of the silver source.b Based on the theoretical carbon yield.

became constant in time. A more extensive descrip-tion of the determination of both permeability andselectivity can be found elsewhere[2]. The reportedresults are based on measurements performed on twoor more membranes.

Besides N2 and O2, CO2 and He were used forthe gas permeation experiments. Of these gases, He(2.6×10−10 m (2.6 Å)), O2 (3.46×10−10 m (3.46 Å)),and N2 (3.64× 10−10 m (3.64 Å)) have a relative lowaffinity to the carbon matrix compared to CO2 (3.3×10−10 m (3.3 Å)) [14].

To show the effect of functionalizing the CMS mem-brane by adding Ag-nanoclusters the results are partlypresented as enhancement factors (Ef), which are de-fined in analogy with the facilitation factor in facil-itated transport[15], e.g. liquid membranes. The Efgives the ratio of permeabilities of a specific gas forthe functionalized AgCMS over the CMS (Eq. (1)).

Ef = PAgCMS

PCMS(1)

3. Results and discussion

In Table 1, an overview is given of the membranesprepared for this research. The membranes contain-ing no Ag are coded MP, whereas the Ag-containingmembranes are denominated by MAg. From the TGAexperiments, it is possible to derive the onset of boththe reduction of the Ag-ions as the pyrolysis of thepolymer.Fig. 4agives the TGA results for all the dif-ferent precursor compositions and pure P84 powder inthe temperature range of 50–200◦C. The Ag-ions startreducing at approximately 155◦C. Simultaneously tothe process of reduction the residual solvent evaporatesfrom the sample, as can be seen from the differences


Fig. 4. TGA micrograph of the reduction process (a) and the totalpyrolysis (b) for all precursor membrane compositions.

in the curves for precursors P84 powder and MP1.This evaporation conceals the weight loss caused bythe reduction of Ag-ions in precursor MAg1, becauseof the small amount of Ag-ions present.

Fig. 4bshows the weight loss of all the precursorsover a temperature range from 150 to 700◦C. TheTGA curves were corrected for water and solvent lossbelow 150◦C. The P84 polymer starts to decomposerapidly at temperatures above 500◦C. Although theAg content in the different precursors differs from 0to 23 wt.%, the final weight loss is in the same rangefor all samples. A simple mass balance over the pyrol-

ysis can explain this, showing that the weight loss ofpyrolysis products and Ag salt counter-ions coincide.

As the heat treatment progresses, the physicalproperties of the membranes change depending ontemperature and Ag content.Table 2 displays thedifferent physical properties of the membranes. TheMAg3 membranes with a high Ag content are verybrittle. This is contributed to the high amount of Agdistributed in the carbon matrix and on the surface ofthe membrane.

In Table 1, the composition of the prepared AgCMSis given for different Ag sources before and after py-rolysis. Although the content of Ag is high, it is stillbelow the percolation threshold, Shen et al.[16] de-termined the percolation threshold to be between 33and 36 wt.% for Ag in Bi–Y–O–Ag composites. Al-thoughTable 1contains values up to 40 wt.% of Ag,the percolation threshold is not reached as a large partof the Ag migrates to the surface of the membrane.The source of the Ag-ions can play an importantrole in the formation of the functionalized mem-brane. Southward and Thompson[13] reported thatthe formation of conductive Ag layers on polyimidesurfaces is affected by the ligands. Depending on thetype of ligand, more or less Ag-ions are directed tothe surface. They found that using AgAc in combi-nation with TFA produced better conductive surfaces,whereas the use of AgNO3 gave brittle and degradedfilms. Although the use of AgAc as Ag source inCarbon Molecular Sieves has been reported[17], weobserved the best results using AgNO3. The observedbrittleness of the films produced with AgNO3 reportedby Southward and Thompson[13] is most likelycaused by the formation of gaseous products duringthe reduction of AgNO3. By using thin precursorfilms and limiting the AgNO3 content, we obtainedflexible films with mechanical properties comparableto their non-functionalized counterparts. The differentbehavior of these two Ag sources can be observed bytaking a more detailed look at the cross-section andsurface of the functionalized membrane.Figs. 5 and 6show, respectively, the SEM (cross-section) and AFM(surface) micrographs of functionalized membranes(MAg1 and MAg4) pyrolyzed at 700◦C. We shouldemphasize that SEM and AFM micrographs show nostructure at these resolutions for samples without Ag.

In the case of functionalized membranes pre-pared with AgNO3 (Fig. 5a), a fine distribution of


Table 2Physical properties of the produced CMS and AgCMS membranes

Name 200◦C 350◦C 500◦C 600–800◦C

MP1 Polymer Polymer Intermediate CarbonYellow Yellow Black BlackTransparent Transparent Brittle ReflectingFlexible Flexible Brittle

MAg1 Polymer Polymer Intermediate CarbonYellow/brown Brown Golden GoldenTransparent/little Mirror Mirror MirrorReflecting flexible Flexible Flexible Brittle

MAg2 Polymer Polymer Intermediate CarbonBrown Silver Golden SilverTransparent/reflecting Mirror Mirror MirrorFlexible Flexible Flexible Brittle

MAg3 Polymer Polymer Intermediate CarbonSilver Silver Silver SilverMirror Mat Mat MirrorBrittle Brittle Brittle Very brittle

MAg4 Polymer CarbonYellow/brown SilverTransparent/little MirrorReflecting flexible Brittle

Ag-nanoclusters, seen as white dots, is observed inthe black carbon cross-section. This, in contrary tothe AgAc-containing samples (Fig. 5b), where virtu-ally no Ag-clusters can be observed. InFig. 6a and b,the black regions represent the carbon matrix and thewhite regions the Ag-clusters as obtained by AFMtapping mode. It is evident from the surface picturesthat in the case of AgAc (Fig. 6b), a significant part ofthe Ag diffuses to the surface, resulting in a denselypacked layer. This is in agreement with the findingsof Southward and Thompson[13], as they aimed fora layer of Ag on the surface, whereas we aim fora containment of the Ag-nanoclusters in the carbonmatrix. For this purpose the AgNO3 is more suitable.

As the pyrolysis end temperature is increased, theAg layer on the surface of the MAg1 membranechanges from single Ag-clusters to a percolated struc-ture at temperatures above 500◦C. This can be seenby comparingFig. 6a with Fig. 7, which shows thesurface of a MAg1 membrane prepared at 350◦C.As the pyrolysis temperature is increased or the endtemperature soak is extended, the concentration ofAg on the surface of the membranes increases. Thiscontinues until an almost dense Ag layer is formed.

By reducing the concentration of AgNO3 in the pre-cursor, less Ag diffuses to the surface, resulting in thin-ner surface layers. The average cluster size reducesfrom 65 to 50 nm for, respectively, MAg2 and MAg1,with increasing number of Ag-clusters.

Fig. 8shows the effect of cluster size on the numberof clusters and total cluster surface area calculated for1 mmol Ag, assuming perfectly spherical clusters withdensityρ(Ag) = 10.5 g/cm3 (293 K). As the diameterdecreases, the number of particles and their accumu-lated surface area increase rapidly. It is clear that weshould work towards small cluster diameters, to ob-tain the highest functional area in the functionalizedmembrane.

The effect of functionalizing the carbon matrix withAg-nanoclusters becomes evident when the pure gaspermeabilities are determined. Numerical values aregiven inTable 3and the results are interpreted belowby plotting the data in various ways.Fig. 9 displaysthe selectivity of O2 over N2 versus the O2 perme-ability. For reference, we have added the Robeson up-per limit of 1991[18]. It is clear that the carbonizedmembranes (MP1 600 and MP1 700) show permeabil-ity and separation properties exceeding the properties


Fig. 5. SEM micrographs of the AgCMS membrane cross-section,using AgNO3 (a) or AgAc (b) as Ag source, respectively (pyrolysisend temperature 700◦C).

of their polymeric precursors (MP1 350 and MP1500). The MP1 800 membranes have a selectivity ofO2 over N2, which approaches 1. This phenomenonwill be discussed in more detail in a later paragraph.The addition of Ag-nanoclusters to the carbon matrixhas a significant effect on the permeability and se-lectivity of the membranes. To quantify the effect ofAg-nanoclusters in the carbon matrix,Fig. 10 showsthe enhancement factor for O2 and N2, as defined byEq. (1). The line at the angle of 45◦ indicates that O2

Fig. 6. AFM micrographs of the AgCMS membrane surface ofMAg1 and MAg4, using AgNO3 (a) or AgAc (b) as Ag source,respectively (pyrolysis end temperature 700◦C).

and N2 are equally enhanced. Values below the line in-dicate a preferential N2 enhancement, whereas valuesabove the line indicate a preferential O2 enhancement.Considering that the value of the enhancement factoris a measure for the increase of the permeability of agas, we can derive fromFig. 10that below 600◦C theAg acts primarily as a spacer within the carbon ma-trix, thereby significantly increasing the permeabilityof the membrane for both gases and simultaneouslydecreasing the size selectivity. However, at 700◦C the


Fig. 7. AFM micrograph of the AgCMS membrane surface ofMAg1, using AgNO3 as Ag source (pyrolysis end temperature350◦C).

oxygen permeability is enhanced with a factor of 4,compared to an enhancement of 3 for N2. The spacereffect of the Ag, i.e. adding additional volume for per-meation to the carbon is shown inFig. 11, where theselectivity of He over CO2, two gases with low affinityto the functionalized carbon matrix, is plotted versusthe He permeability. The arrows indicate membranes

Fig. 8. Number of clusters (�) and total surface area (�) vs.the cluster diameter, calculated for 1 mmol Ag assuming perfectlyspherical clusters with densityρ(Ag) = 10.50 g/cm3 (293 K).

Table 3Pure gas permeabilities for the prepared CMS and AgCMS

Membrane Tpyrolysis

(◦C)PHe

(barrer)PCO2

(barrer)PO2

(barrer)PN2

(barrer)

MP1 350 7.9 1.5 0.35 0.04MP1 500 26.5 17.5 3.9 0.73MP1 600 447 276 72.2 7.8MP1 700 166 64.1 16.8 2.1MP1 700120

a 37.0 2.4 0.46 0.14MP1 800 58.5 3.3 3.0 2.8MAg1 350 6.5 1.1 0.26 0.05MAg1 500 42.1 27.5 6.1 1.8MAg1 600 668 619 170 19.9MAg1 700 589 290 81.3 6.7MAg1 700120

a 366 83.6 30.8 2.5MAg1 800 169 12.4 3.0 0.19MAg2 700 720 284 77.0 8.4MAg4 700 298 148 39.4 3.9

a The membrane was allowed to soak for 120 min at 700◦C,before quench.

treated at the same temperature. The He permeabilityincreases by adding Ag-nanoclusters, whereas the se-lectivity over CO2 is lowered, when the same pyrolysisend temperature is considered. The sample inFig. 11coded 700120 min was allowed to soak at 700◦C for120 min before the quench.

To explain the permeation behavior, we should lookmore closely to the mechanisms for diffusion within

Fig. 9. Selectivity of O2 over N2 vs. the O2 permeability fordifferent pyrolysis end temperature, Ag content, and source (298 K,2 bar). (�) MP1, (�) MAg1, (�) MAg2, (�) MAg4.


Fig. 10. Enhancement factors for O2 and N2 for MAg1 comparedto MP1 (298 K, 2 bar).

our functionalized carbon structure.Fig. 12 shows aschematic representation of the functionalized carbonstructure. Three transport routes (Fig. 12a-c) can beproposed for a gas molecule to diffuse through a func-tional part of the membrane. Independent of the diffu-sion route through the membrane, the gas moleculesmust first diffuse through the metal layer on the surface

Fig. 11. Selectivity of He over CO2 vs. the He permeability forMAg1 (�) and MP1 (�) (298 K, 2 bar). The arrows indicatemembranes treated at the same temperature.

Fig. 12. Proposed diffusion routes through a functional part ofan AgCMS membrane. (a) Diffusion through carbon matrix; (b)diffusion through Ag-cluster; (c) bypass of Ag-cluster diffusionthrough free volume.

of the membrane before entering or leaving the mem-brane. In the case of the proposed route a where nometal cluster is present, the molecules diffuse throughthe carbon matrix by micropore diffusion. Follow-ing route b, the gas molecule diffuses through themetal cluster. From Hwang and Kammermeyer[19],the permeation rate through dense Ag-clusters at atemperature of 25◦C can be estimated to be small(<10−7 barrer). Finally, considering route c, the per-meating gas molecule can bypass the metal cluster,because of a poor adhesion between the metal clus-ter and the carbon matrix. Depending on the size ofthe bypass volume, this occurs by macropore diffusionand Knudsen diffusion (<150× 10−10 m (<150 Å))or alternatively by surface diffusion (<50× 10−10 m(<50 Å)) in the case of interaction between the gasand the metal cluster, i.e. in our case O2. As the vir-gin Ag-nanocluster is exposed to O2 it will chemisorbmost of the O2 irreversibly onto the surface, part ofthe O2 molecules are weakly bound to the surface byboth physical and chemical adsorption and can easilydesorb[11]. As the O2 adsorbs, it can dissociate intoatoms allowing it to pass through even smaller bypassareas[12]. The total static adsorbed amount of oxygenon the Ag particles in the carbon matrix, as determinedfrom the sorption isotherm, can give an indication forthe presence of active Ag sites in the carbon matrix,however as most of the O2 is chemically adsorbed, itdoes not contribute to the permeation process. This im-plies that the translation of these values to the dynamicadsorption—surface diffusion—desorption process isnot trivial. Therefore, we will use the increase in se-lectivity of O2 over N2 as a measure for this process.


From the increased permeation rates compared tothe non-functionalized carbon matrix and the lowpermeation rates through dense metal clusters, weconclude that route c is the mechanism contributingmost to the enhanced permeability. At higher py-rolysis end temperatures, the bypass volume aroundthe Ag-clusters as described in route c is reduced,thereby increasing the effect of surface diffusion overthe Ag-clusters.

The point inFig. 10coded MAg1 700∗/MP1 600 isdisplayed because these specific membranes show anEf of 1 for N2. This indicates that the reported CMSmembrane prepared at 600◦C has the same perme-ability for N2 as the AgCMS prepared at 700◦C, thusindicating that both have the same pore size. The ef-fect of the Ag-nanoclusters now becomes evident, asthe O2 permeability shows an Ef of 1.3.

The build up of an Ag layer on the surface of themembrane has a strong effect on the permeability ofthe membrane. To investigate the properties of thisAg layer we analyzed three middle Ag-content MAg2membranes. Between reaching the pyrolysis end tem-perature and start of the quench procedure, the mem-branes were brought into contact with air, resulting inan oxidation of the membrane surface. The extent ofoxidation was varied.Fig. 13shows the effect of theoxidation step on the O2 over N2 selectivity and theO2 permeability. With prolonged oxidation, the perme-

Fig. 13. Selectivity of O2 over N2 vs. the O2 permeability fordifferent extent of oxidization of an MAg2 membrane (298 K,2 bar).

ability of O2 increases by approximately 700% with-out showing any loss of selectivity. This implies that byoxidizing the membrane we have reduced the Ag layeron the membrane surface, however we have not alteredthe carbon matrix as suggested by Shusen et al.[20],who increased the mean pore size by post-oxidationof a carbon membrane.

The properties of the above-mentioned CMS mem-branes prepared at 800◦C differ from the CMS mem-branes prepared at lower pyrolysis temperatures. At800◦C, residual N atoms are removed from the car-bon matrix[21], this leads to the formation of a bi-modal pore size distribution. This has been reportedpreviously[2]. The He over CO2 selectivity remainshigh whereas the He over N2 decreases by two or-ders of magnitudes. This, together with the observedO2 over N2 selectivity (1.1), can only be explained bythe existence of two pore sizes having radii ofrHe <

r1 < rCO2 andr2 > rN2. This phenomenon will be thetreated in a subsequent research.

4. Conclusions

We have shown that it is possible to functional-ize CMS membranes with Ag. Moreover, we haveshown that the addition of Ag-nanoclusters to CMSmembranes increases the pure gas O2 over N2 se-lectivity by a factor of 1.6, when compared to itsnon-functionalized CMS counterpart. By introducingthe enhancement factor Ef we can clearly show the ef-fect of the functionalization. The permeability of themembranes increases significantly (240%, 600◦C) forpyrolysis end temperatures below 700◦C, where theAg-nanoclusters can act as spacers. Above this tem-perature, an Ag layer develops on the surface of themembrane. Oxidation of this layer has proven benefi-cial to the permeability, without affecting the selectiv-ity. We have suggested two possible diffusion routesthrough the AgCMS, opting for the route where thegas molecules pass through the free volume betweenAg-nanocluster and the carbon matrix.

We analyzed several Ag contents (0, 6, 25, and40 wt.%) of two different Ag sources (AgNO3 and Ag-Ac), determining that the lower Ag content (6 wt.%)with AgNO3 as the source resulted in the most sta-ble AgCMS membrane. This membrane containedAg-nanoclusters with a diameter of approximately


50 nm and a fine distribution throughout the bulk ofthe carbon matrix. Using a higher Ag content affectsthe mechanical stability of the membrane, while bothpermeability and selectivity are negatively influenced.The addition of AgAc as the Ag source led to theformation of Ag-surface layers with an increasedcoverage of the total surface as compared to AgNO3.

References

[1] R.W. Baker, Future directions of membrane gas separationtechnology, Ind. Eng. Chem. Res. 41 (2002) 1393–1411.

[2] J.N. Barsema, N.F.A. van der Vegt, G.H. Koops, M. Wessling,Carbon molecular sieve membranes prepared from porousfiber precursor, J. Membr. Sci. 205 (2002) 239–246.

[3] J.D. Way, D.L. Roberts, Hollow fiber inorganic membranesfor gas separations, Sep. Sci. Technol. 27 (1992) 29–41.

[4] M.B. Shiflett, J.F. Pedrick, S.R. McLean, S. Subramoney,H.C. Foley, Characterization of supported nanoporous carbonmembranes, Adv. Mater. 12 (2000) 21–25.

[5] H. Suda, K. Haraya, Gas permeation through microporesof carbon molecular sieve membranes derived from Kaptonpolyimide, J. Phys. Chem. B 101 (1997) 3988–3994.

[6] J. Kim, Y.S. Lin, Synthesis and oxygen permeation propertiesof ceramic-metal dual-phase membranes, J. Membr. Sci. 167(2000) 123–133.

[7] H. Amandusson, L.G. Ekedahl, H. Dannetun, Hydrogen per-meation through surface modified Pd and PdAg membranes,J. Membr. Sci. 193 (2001) 35–47.

[8] T. Furusawa, K. Seshan, J.A. Lercher, L. Lefferts, K. Aika,Selective reduction of NO to N-2 in the presence of oxygenover supported silver catalysts, Appl. Catal. B: Environ. 37(2002) 205–216.

[9] D.O. Hayward, Gas adsorption in chemisorption and reactionson metallic films, in: J.R. Anderson (Ed.), vol. 1, AcademicPress, London, 1971, p. 231.

[10] R. Keim, Gmelig handbuch der anorganischen chemie, Silber:verbindungen, vol. B1, eighth ed., Verlag Chemie GMBH,Weingeim, 1971, p. 32.

[11] A.W. Czanderna, The adsorption of oxygen on silver, J. Phys.Chem. 68 (1964) 2765–2772.

[12] X.E. Verykios, F.P. Stein, R.W. Coughlin, Oxidation ofethylene over silver: adsorption, kinetics, catalyst, Catal. Rev.Sci. Eng. 22 (1980) 197–234.

[13] R.E. Southward, D.W. Thompson, Inverse CVD: A novelsynthetic approach to metallized polymeric films, Adv. Mater.11 (1999) 1043–1047.

[14] M. Ogawa, Y. Nakano, Gas permeation through carbonizedhollow fiber membranes prepared by gel modification ofpolyamic acid, J. Membr. Sci. 162 (1999) 189–198.

[15] C.A. Koval, Z.E. Reyes, in: R.D. Noble, J.D. Way (Eds.),Liquid Membranes: Theory and Application, AmericanChemical Society, Washington, DC, 1987, p. 29.

[16] Y. Shen, M. Liu, D. Taylor, S. Bolagopal, A. Joshi, K. Krist,Mixed ionic-electronic conductors based on Bi–Y–O–Agmetallic-ceramic system, in: T.A. Ramanarayanan, W.L.Worrell, H.L. Tuller (Eds.), Second International Symposiumon Ionic and Mixed Conducting Ceramics, vol.94-12, The Electrochemical Society, Pennington, 1994,p. 574.

[17] H. Suda, K. Haraya, Oxygen/nitrogen permselectivity ofsilver/carbon nanocomposite membranes, in: Proceedings ofthe Sixth International Congress on Inorganic Membranes2000, Montpellier, 2000, p. 406.

[18] L.M. Robeson, Correlation of separation factor versuspermeability for polymeric membranes, J. Membr. Sci. 62(1991) 165–185.

[19] S.-T. Hwang, K. Kammermeyer, Membranes in Separations,first ed., Wiley-Interscience, New York, 1975, p. 94.

[20] W. Shusen, Z. Meiyun, W. Zhizhong, Carbon membranesfor gas separation, Sep. Sci. Technol. 31 (16) (1996) 2299–2306.

[21] H. Hatori, Y. Yamada, M. Shiraishi, M. Yoshihara, T. Kimura,The mechanism of polyimide pyrolysis in the early stage,Carbon 34 (1996) 201–208.

Documents

Functionalized Carbon Molecular Sieve membranes containing Ag-nanoclusters