Polymer-zeolite composite membranes for direct methanol fuel cells.
Download Polymer-zeolite composite membranes for direct methanol fuel cells. ...
Polymer-Zeolite Composite Membranes for Direct Methanol Fuel Cells Brett Libby, W. H. Smyrl, and E. L. Cussler Dept. of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, MN 55455
Direct methanol fuel cells require membranes with the dual properties of high proton conducti®ity and low methanol crosso®er. Such membranes ha®e a high selecti®ity, that is a high ratio of proton conducti®ity to methanol permeability. This research reports such a membrane made of poly®inylalcohol and loaded with mordenite, a proton conducting, methanol impermeable zeolite. Protons tra®el directly through both the polymer and zeolite phases, but methanol has a more tortuous path around the zeolite particles. The composite membranes show up to twenty times higher selecti®ity than Nafion, the current benchmark. The impro®ed beha®ior, a result of the proper tailoring of the polymer and dispersion phase, is predicted using Maxwell’s theory for diffusion in composite media.
Introduction Fuel cells promise more power for less fuel than current thermal engines. This promise has caused a flurry of fuel cell research. Driven by a market that calls for reliable, inexpensive, and environmentally sound sources of energy, researchers are working to boost fuel cell efficiency and performance, while simultaneously reducing size, weight, and cost. Whereas fuel cells were once confined to expensive niche applications like the space program, fuel cells are now expected to broadly impact energy production. Applications of fuel cells are currently focused in three areas: electric vehicles, stationary electricity generation, and portable electronic devices. The most well publicized fuel cell research efforts are electric vehicles. Concept cars shine at major auto shows, while fuel cell powered buses roam the streets of Vancouver and Chicago. General Motors, Ford, Daimler Chrysler, Honda, and Toyota are all planning production ŽBusiness Week, 2000.. However, the high costs of current fuel cell powered automobiles will initially prevent their widespread use. Electricity generation of stationary fuel cells appears more immediate. Over the next 20 years, the electric utilities will need to add about 10 trillion kilowatt-hours of new capacity worldwide ŽBusiness Week, 2000.. Because of the deregula-
Correspondence concerning this article should be addressed to E. L. Cussler.
tion of the utility industry, fuel cells are widely investigated as on-site generators of electricity for homes and buildings. Portable electronic devices may also see widespread acceptance of fuel cells. Fuel cell technology can move into the evolving microelectronic market more easily than into other, more established industries ŽLibin, 2000.. Polymer electrolyte membrane fuel cells are especially attractive as portable battery replacement ŽJost, 2000.. Such cells, which use a solid polymer membrane as the electrolyte, offer several advantages. The immobilized electrolyte and absence of corrosive liquids simplifies sealing and minimizes corrosion ŽJacoby, 1999.. The low operating temperatures allow for fast startups and immediate responses to changes in power demand. Most importantly, such compact cells can deliver higher power densities. This results in reduced weight, cost, and volume, along with improved performance. A polymer electrolyte membrane fuel cell runs best on pure hydrogen as a fuel, so the cell produces only electricity and water. However, hydrogen is hard to store, whether as a compressed gas, as a cryogenic liquid, or in metal hydrides or carbon nanotubes ŽAppleby, 1999.. As a result, many try to utilize hydrogen produced by steam reforming a fuel like methanol, followed by hydrogen separation using a palladium alloy membrane to remove carbon monoxide, which poisons the cell’s platinum based electrode catalysts. Such processing dramatically increases the fuel cell’s weight and complexity.
April 2003 Vol. 49, No. 4
An alternative cell could directly use an organic fuel like methanol, rather than convert it to hydrogen. The oxidation potential of methanol ŽE o s1.21V. is comparable to that of hydrogen ŽE o s1.23V., and methanol has a 50% greater volumetric energy density ŽDyer, 1999.. However, the current membrane used in the cells, a perfluorosulfonic acid membrane like Dupont’s Nafion, leaks methanol from the anode to cathode. This methanol crossover has two effects. First, the chemical energy of the methanol is lost when it crosses through the membrane, severely lowering the efficiency of fuel utilization ŽRen et al., 1995.. Second, the two simultaneous electrochemical reactions at the cathode compete directly for cathodic catalyst sites, reducing the overall cell efficiency. Methanol crossover currently limits direct methanol fuel cells to a 0.5᎐1.0 M methanol fuel feed. Increasing the methanol feed concentration could potentially increase methanol oxidation kinetics ŽNarayanan et al., 1996.. This article seeks a new polymer electrolyte membrane with lower methanol crossover. In the following sections, we discuss our target, describe our experiments, and report our results towards this goal.
Theory The search for new membranes begins by establishing criteria for comparing candidates to the benchmark of Nafion. Transport of protons and methanol is described by the Nernst-Plank equations for the flux ji y ji s Di ⵜc i q z i c i F
where Di is the diffusion coefficient, c i is the concentration and z i is the charge of species ‘‘i’’; F is Faraday’s constant, R is the gas constant, T is the temperature, and ⌽ is the electrostatic potential. For the conductance of protons Žspecies ‘‘1’’ ., z1 is Žq1. and the concentration c1 is a constant, so for a membrane of thickness l y j1 s
D 1 c1 F RT
This result is more commonly written as a current density i isy Fj1 s Ž ⌬ rl .
where the conductivity is given by
s D 1 c1 F 2rRT
For methanol species Ž‘‘2’’., z 2 is zero and j2 s D 2 H2 Ž c 20 y0 . rl
where H2 is a partition coefficient, the product D 2 H2 is the methanol permeability P, and Ž c 20 y0. is the concentration difference of methanol across the membrane. This equation is exact for fuel cells only when electroosmosis does not contribute to the methanol flux. 992
We are interested in the selectivity of protons vs. methanol offered by different fuel cell membranes. For many membranes separations, selectivity is defined as the ratio of fluxes to driving forces ŽHo and Sirkar, 1992.
j1rj2 ⌬ c1r⌬ c 2
Because the driving forces for protons and methanol are different here, the selectivity ␤ is better defined as ŽPivovar et al., 1999.
irj2 ⌬⌽r⌬ c 2
Substituting Eq. 3 and Eq. 5 into Eq. 7 yields
where is the proton conductivity and P is the methanol permeability. Note that this new electrochemical selectivity is independent of membrane thickness, but does have units. In this research, we seek a membrane with high selectivity ␤ . In the past, this goal was taken to imply a high proton conductivity. However, while the current standard fuel cell membrane Nafion 117 has a very high conductivity, its selectivity is compromised due to high methanol permeability. If another membrane had higher selectivity, but lower conductivity, the membrane conductance could be kept large by making the membrane thin. While Nafion 117 is about 200 m thick, membranes used in commercial separations can be 0.3 m thick ŽHo and Sirkar, 1992.. Thus, new membranes could be over a hundred times thinner than the present standard. Mechanical stability can be retained through the use of support layers that do not appreciably affect transport. To seek a membrane with higher selectivity, we investigated composites consisting of zeolite particles in a hydrophilic polymer. As shown in Figure 1, the components are chosen so that protons are transported through both the dispersed phase and the polymer matrix. If the dispersed material is impermeable to methanol, protons will have a more direct and shorter path compared to methanol. This membrane could enhance the proton selectivity, while reducing methanol crossover. We chose zeolites as the dispersed membrane phase because their regular pore size promises selective separations based on molecular size and shape ŽCoronas and Santamaria, 1999.. Zeolites can also separate components based on preferential adsorption, in which the strong adsorption of one component blocks or hinders the passage of the other components. However, membranes made of pure zeolites are plagued by defects such as cracks or gaps, exhibit poor mechanical properties such as brittleness and fragility ŽBerry et al., 2000., and are expensive to manufacture ŽCaro et al., 2000.. A polymer-zeolite composite membrane represents a compromise between the nonselective polymeric films and the brittle expensive zeolite film. The composite combines the
April 2003 Vol. 49, No. 4
Figure 1. Polymer-zeolite composite membrane. The membrane consists of highly selective zeolite particles embedded in a polymer matrix. Protons should travel in a direct path through both the polymer and zeolite phases. In contrast, methanol should travel around and between the methanol-impermeable zeolite particles.
highly selective solid-state proton conductor with the flexibility of a polymer matrix. Such membranes have been investigated for both liquid and gas separations, especially for pervaporation ŽSuer et al., 1994.. Applications include dehydration of azeotropic mixtures, separation of organics, and direct methanol fuel cells ŽJi and Sikdar, 1996; Rao et al., 1994.. Unfortunately, many of these composite membranes are plagued by the formation of voids at the polymerrzeolite interface ŽOkumus et al., 1994.. These voids must be eliminated for a composite membrane to function optimally. Even without nonselective, interfacial voids, a zeolite-polymer composite must fulfill additional transport requirements to function effectively. To see why, we consider a periodic array of spheres embedded in a continuous medium, as shown in Figure 2. We assume the spheres are impermeable to methanol, so methanol moves entirely in the continuous polymer phase. In contrast, we assume that protons move through both the dispersed and continuous phase. Whether or not the protons take advantage of the pathway through the spheres depends on the relative speed of proton transport in this dispersed phase vs. the continuous phase.
Figure 2. Potential transport scenarios in a composite membrane. In this idealized situation, the spheres are impermeable to methanol while protons can move through both the continuum and the spheres. In Case 1, protons travel around the spheres because the proton conductivity of the continuum is substantially greater than in the spheres. In Case 2, the spheres offer a highly conducting pathway compared to the continuum. The composite now offers a potential selectivity improvement.
Two limits for transport in this composite system are shown in Figure 2. In Case 1, proton conductivity in the spheres is much less than in the continuum. As a result, the protons move entirely in the continuous phase by traveling around and between the less conductive spheres. Because protons and methanol move by the same pathway through the continuum, the selectivity of the composite is equal to the selectivity of the continuum. Case 2 is the opposite of Case 1: the proton conductivity in the spheres is now substantially greater than in the continuum. In this scenario, protons quickly move through the spheres, while methanol travels around and between them.
April 2003 Vol. 49, No. 4
This separation of pathways for methanol and proton transport can result in the selectivity of the composite being greater than the continuum. In particular, when the membrane is filled with a periodic array of spheres Ds
ž / ž /
where D is the effective diffusion coefficient in the composite membrane, DP is the diffusion coefficient in the polymer continuum, DS is the diffusion coefficient through the spheres, and is the volume fraction of spheres ŽMaxwell, 1881.. We will use Eq. 9 in correlating our experimental data. We will be especially interested in the two limits of this equation given in Figure 2. Because the protons have a direct path through both phases, the proton conductivity Ž . given by an alternative form of Eq. 9
2q␥ y2 Ž 1y␥ . s
2q␥ q Ž 1y␥ .
where ⑀ 1Ž s rP . is the ratio of the proton conductivity in the composite vs. proton conductivity in the polymer phase, ␥ Ž s SrP . is the ratio of proton conductivity in the spheres to proton conductivity in the polymer, and is the volume fraction of spheres. Because methanol cannot go through the spheres, the methanol permeability P is described by Eq. 9 with DSrDP equal to zero
2 Ž 1y .
P s PP
where ⑀ 2 Ž s PrPP . is the ratio of the effective methanol permeability in the composite membrane relative to methanol permeability in the polymer phase PP . Dividing Eq. 10 by Eq. 11 yields the selectivity
␤ s ␤P
2q␥ y2 Ž 1y␥ .
2q␥ q Ž 1y␥ .
2 Ž 1y .
where ␤P Ž s PrPP . is the selectivity in the continuous polymer phase. Equation 12 shows that potential selectivity improvement in a composite membrane is a function of two main factors: ␥ and . To determine the effect of ␥ on selectivity, we return to the two cases in Figure 2. In Case 1, when the spheres are impermeable to protons, ␥ is zero, and the bracketed terms of Eq. 12 reduce to one. The selectivity of the composite equals the inherent selectivity of the polymer Žthat is, ␤ s ␤ P .. In Case 2 of Figure 2, when proton transport through the spheres is very rapid, ␥ is infinity, and Eq. 12 becomes
␤ s ␤P 994
Selectivity improvement now depends on the inherent polymer selectivity ␤P and the volume fraction of spheres . When s 0.5, the selectivity of the composite is about 10 times better than in the pure polymer. We will return to these equations in discussing our results.
2 Ž 1y .
Experimental Studies Rationale for this model system For our studies, we chose mordenite as the zeolite phase and polyvinylalcohol ŽPVA. as the polymer. The hydrophilic nature of both PVA and mordenite prevents the formation of nonselective voids at the polymer-zeolite interface, while also exploiting the selectivity of the mordenite. Mordenite is one of the most stable zeolites in existence ŽBreck, 1974., and is one of the most highly conducting Ž10y3 srcm. and widely studied ŽKnudsen et al., 1988.. Polyvinylalcohol is also a logical choice, because it can be easily formed into membranes and it has a permeability which can be altered with heat. Mordenite is stable up to temperatures above 800⬚C ŽBreck, 1974., doesn’t decompose in the presence of acids, and has been studied before as a component in composite membranes for methanol fuel cells ŽKjaer et al., 1991; Rao et al., 1994.. Although these particular materials are not ideal for actual fuel cell use, they do possess the transport properties necessary for demonstrating improved selectivity in such a polymer-zeolite composite membrane. While we believe that the mordenite-PVA system is a good model for our research, we are concerned that the mordenite used is imperfectly defined. This leads to three more specific concerns: the zeolites SirAl ratio, its pore size, and its conductivity. Each merits discussion. The SirAl ratio of a zeolite dominates adsorption within the zeolite. In particular, increasing the SirAl is believed to change the selectivity from hydrophilic to hydrophobic. There is debate about where the change occurs. Typical mordenite has a SirAl ratio of 5, and is hydrophilic. A recent study demonstrated hydrophilic behavior using a zeolite membrane with SirAl ratio of 15 ŽNoack et al., 2000.. This study utilizes a dealuminated form of mordenite, with a SirAl ratio of 40, but dealuminated mordenite still preferentially adsorbs water over methanol ŽMeninghaus and Prins, 2000.. However, while adsorption of either species will affect unsteady uptake, it will usually not affect steady-state transport ŽCussler, 1997.. As we will show below, we see no evidence of adsorption in our experiments. We are also concerned about pore size. The mordenite used in the present study has most likely been dealuminated through steam treatment at elevated temperatures. In addition to changing the SirAl ratio, this treatment can increase the pore size. Typical mordenite has a main channel pore size of 0.7 nm, whereas dealuminated mordenite contains pores as large as 20 nm ŽMeninghaus and Prins, 2000.. The larger pore size will substantially reduce the effects of ionic repulsion between mordenite’s negatively charged framework and the sulfate anion. Because we always used the same mordenite, we have not investigated this effect. Finally, we are concerned about the inherently low proton conductivity of 10y3 Srcm in mordenite. While typical mordenite is estimated to have an internal molarity of 6M ŽRabo, 1976., the acid site concentration of dealuminated mordenite
April 2003 Vol. 49, No. 4
is reduced to about one-sixth of the normal value ŽHibino et al., 1993.. In this study, we increased the zeolite’s effective conductivity by soaking the polymer-zeolite membranes in sulfuric acid solutions. The sulfuric acid solutions should readily absorb into mordenite, similar to the uptake of sulfate species such as 0.5 M ammonium sulfate ŽLei et al., 1999. and tin sulfate ŽKnudsen et al., 1988.. Additionally, mordenite impregnated with other ionic species such as antimony ŽV. chloride ŽHix et al., 1994. and phosphotungstic acid ŽArico et al., 1995. also shows increased conductivity. Nonetheless, we recognize that soaking the composite membrane in sulfuric acid is a major disadvantage for fuel cells. Membrane fabrication. Composite membranes of mordenite particles embedded in a polyvinylalcohol ŽPVA. matrix were fabricated using commercially available materials. The mordenite powder ŽHM40, Sud-Chemie Inc., Louisville, KY. consisted of 2᎐4 m particles, with a SirAl ratio of 40. Because stability increases and hydrophilicity decreases as the SirAl ratio increases, this particular value represents a compromise. This mordenite is stable in 1M-H 2 SO4 , required for conductivity. The polyvinylalcohol ŽPVA. ŽElvanol 71-30, DuPont, Wilmington, DE. was a medium viscosity, fully hydrolyzed grade with a molecular weight of ;140,000. It is soluble in hot water. The polymer showed excellent adhesion to porous, water-absorbent surfaces. Such adhesion is important in forming a composite membrane with mordenite. Because of its ability to both disperse mordenite and dissolve PVA, water was used as the solvent in solution casting of PVArmordenite membranes. Membrane fabrication began by stirring overnight 1 g mordenite powder in 5 g water. A sonicating probe ŽUltrasonic Homogenizer-4710 Series, Cole-Parmer Instrument Co., Chicago, IL. further dispersed the mordenite for 5 min at 50% output, followed by 5 min of stirring, and repeated for 1h. Concurrently, a covered 5-wt. % solution of PVA in water was heated at approximately 100⬚C for 1 h. The PVA solution and the mordenite suspension were then mixed together for 4h at 60⬚C. Membranes were prepared by casting this solution in a polystyrene dish. After the wet membranes were allowed to air dry for one week, they were made insoluble through heat treatment at 150⬚C for one to 27 h. Membranes were then hydrated for at least 24 h. Fully hydrated membranes varied in thickness from ;100᎐300 m. The procedure for making a pure PVA membrane was the same, but without mordenite. The color and mechanical properties of the membranes varied with mordenite content and heat treatment. Prior to heat treatment, pure PVA membranes were transparent and flexible. Pure PVA heat-treated membranes retained their flexibility, but turned slightly yellow in color. PVA membranes containing mordenite were beige after drying and were considerably stiffer than pure PVA membranes. After a 1h heat treatment, the membranes became more fragile and the color changed from beige to black. Membranes heat-treated for 27 h were still more fragile but were brown.
persed throughout the membranes. The hydrophilicity of the mordenite was confirmed by the defect free boundary formed at the PVArmordenite interface. Although mordenite has a higher density than PVA, the viscosity of the PVA solution apparently prevents the mordenite particles from settling. Heat treatment caused no cracking at the PVArmordenite interface. The chemical effects of heat treatment were studied with infrared spectroscopy, collected using a Nicolet Magna-IR 550 spectrometer. Results for pure PVA membranes showed that the spectrum of a heat-treated membrane differed from that for an unheated membrane only in an increase in one peak in the C-O stretch region, at ;1,200᎐1100 cmy1. This peak increase is consistent with the formation of carboxyl groups. Outside of a slight yellowing, he heat treatment of pure PVA does not appreciably affect the chemical or physical nature of the polymer. Results for PVA membranes containing mordenite also showed spectra before and after a 1h heat treatment at 150⬚C which were virtually identical. An example is given in Figure 3a. The only noticeable difference was a reduction in the H-
Figure 3. Infrared absorption spectra of a PVAr r r50by volume) membrane: (a) mordenite (50r before heat treatment and (b) after a 27 h heat treatment.
Scanning electron microscopy ŽJEOL 840-II. revealed that the 2᎐4 m mordenite particles were homogeneously disAIChE Journal
April 2003 Vol. 49, No. 4
The heat treatment results in a reduction in the O-H stretch peak at ; 3,300 cm y1 , while also producing a carbonyl ŽC s 0 . peak at ; 1,700 cm y1 and an alkene ŽC s C . peak at ; 1,600 cmy1 .
O-H bending peak from residual water, a direct result of removing water from the membrane. In contrast, the spectra after a 27h heat treatment were clearly different, as shown in Figure 3b. The heat treatment again resulted in a reduction in the O-H stretch peak at ; 3,300 cmy1, but also produced new peaks at ;1,700 cmy1 and ;1,600 cmy1. Similar changes have been observed during the dehydration of PVA in the presence of sulfuric acid ŽImmelman et al., 1993. and of sulfated zirconia ŽKim et al., 2001.. Partial dehydration of PVA is known to cause the brown discoloration seen here. The acidic hydroxyl groups on the mordenite surface may act as catalysts for the PVA dehydration. The resulting structure contains both carbonyl ŽC⫽O. and alkene ŽC⫽C. groups, consistent with the new peaks in Figure 3b at ;1,700 cmy1 and ;1,600 cmy1 ŽPritchard, 1970.. To verify the proposed acid-catalyzed dehydration of PVA, membranes were fabricated by dissolving PVA in a weak 0.01M-H 2 SO4 solution instead of water. After solution casting and air drying, the membranes were heated in a 150⬚C oven for 1h. The final membranes displayed the same dark brown color and IR spectrum characteristics as the PVArmordenite membranes heated for 27 h, which indicates dehydration of the PVA ŽLibby, 2001.. Thus, mordenite and sulfuric acid both facilitate the dehydration of PVA during heating.
Transport properties Membranes were tested for both methanol permeability and proton conductivity using methods employed previously by Pivovar et al. Ž1999.. Methanol permeability was measured in a diaphragm cell connected to a differential refractometer. The diaphragm cell had two glass compartments, each about 15 cm3 in volume, separated by the membrane being tested and stirred at 60 rpm. The membrane was clamped vertically between the adjacent compartments, with a silicone rubber gasket placed on each side of the membrane to ensure a tight seal. The downstream compartment was filled with pure water, while the upstream side was typically filled with a 1 M aqueous solution of methanol. The downstream compartment was connected through the pump to the refractometer ŽWaters Model 410, Framinghorn, MA. to allow concentration measurements. Typical experiments involved solution sampling every second and data recording every minute. Proton conductivity of test membranes was determined using the two-point probe technique in which two electrodes on opposite sides of the membrane hold the sample in an acid solution ŽLibby, 2001.. The membranes were soaked in this acid solution for at least 48 h prior to testing. The small gold electrodes, 1.5 mm in diameter, reduced capacitance effects. A spring linked to one of the electrodes kept the membrane under slight but constant pressure, thereby providing good contact between the electrodes and the membrane. The conductivity cell was connected to a Schlumberger Solartron 1286 impedancergain-phase analyzer, which measured the impedance as a function of frequency. The frequency range was typically 1 MHz to 100 Hz. This unit was interfaced with a Labview 4.0 program on a Macintosh computer for data recording. The apparatus determined the conductivity in the traverse direction of the membrane, which is equivalent to the direction of proton transport in fuel cell membranes. 996
Figure 4. Methanol permeability (cm 2rs) vs. mordenite concentration. For a short heat treatment Žthe triangles ., permeability demonstrates a 2.5-fold reduction over the mordenite concentration range. In contrast, the long heat treatment Žthe diamonds . causes permeability to decrease 25 times. Lines representing Maxwell predictions agree well with the short heat treatment data but fail to predict the long heat treatment data.
Results The goal of this research is the development of composite membranes which are more selective for protons than for methanol. Such membranes may find application in direct methanol fuel cells, which currently use membranes of the perfluorinated, sulfonated polymer Nafion. This research will be successful if the new composite membrane shows a better proton to methanol selectivity than Nafion. The results of the research are most easily organized as measurements of methanol permeability, determinations of proton conductivity, and values of selectivity. The selectivities are, of course, the ratios of conductivity and permeability. In reporting these data, we will commonly refer to membranes which are heat treated for 1h or for 27 h. Some heat treatment is important to insure that the membranes are insoluble in water. The longer heat treatment retards transport in the PVA component of the membrane and increases selectivity.
Methanol Permeability The methanol permeability is plotted in Figure 4 vs. the volume fraction of mordenite for both the short heat treatment Ž1 HR HEAT. and the long heat treatment Ž27 HR HEAT.. The 1 h treated membranes were tested in 1 Mmethanol solutions, whereas composite membranes heated for 27 h were tested in 12.3 M methanol Ž50% methanol by volume.. Additional experiments reported elsewhere ŽLibby, 2001. prove that composite membranes heat-treated for 27 h have methanol permeabilities independent of methanol concentration. Inspection of the 1 HR HEAT data shows a slight drop in permeability as the amount of mordenite increases. In particular, a membrane containing 60% mordenite by volume has about a 2.5 fold decrease in permeability over a pure PVA membrane. The decrease supports the expectation that methanol preferentially moves through the polymer phase, while passing around and between the mordenite particles. The methanol permeability of a pure Nafion 117 membrane
April 2003 Vol. 49, No. 4
is included in Figure 4 as the dashed line spanning the plot. Although the plot may appear to show Nafion permeability as a function of mordenite volume fraction, this is not the case: the line is only for comparing a pure Nafion membrane which does not contain mordenite with the PVArmordenite composites. For example, a membrane containing 60% mordenite by volume is about 6 times less permeable than Nafion. As shown in Figure 4, the pure PVA membranes from the 27 HR HEAT data are a little over 3 times less permeable than their 1 HR HEAT counterparts. This is probably due to the increase in crystallinity through removal of residual water, and to the formation of new hydrogen bonds between the macromolecules ŽSakurada, 1985.. Thus, as heat treatment time increases, molecular packing in the PVA becomes tighter, leading to the lower permeability. The 27 HR HEAT data fall below the 1 HR HEAT data across the entire mordenite fraction range. The most striking feature of the plot is the steep decline in permeability with increasing mordenite. For example, the permeability of PVArmordenite membrane Ž50r50 by volume. is ; 25 times smaller than the pure PVA membrane. The explanation behind this steep decrease lies in the IR analysis. The acidic mordenite catalyses further dehydration of the PVA structure, which further increases molecular packing, resulting in reduced free volume in the polymer phase. This combination of increased dehydration and increased methanol path length around the mordenite leads to the steep drop shown.
Proton conducti©ity The proton conductivities reported in Figure 5 are for membranes equilibrated in 1 M sulfuric acid solutions. As shown, a pure PVA membrane heated for 1 h has a conductivity of ; 0.06 Srcm, not far below Nafion Ž ; 0.11 Srcm.. As the mordenite content increases, proton conductivity drops slightly to ; 0.04 Srcm for 60% mordenite despite the 1 M sulfuric acid in the mordenite’s pores. The data imply that because the conductivity in the polymer phase is high, the mordenite pores contribute no significant transport. Most
protons stay in the polymer phase and travel around the mordenite particles, thus leading to a lower conductivity as the mordenite concentration increases. The conductivity in membranes heated for 27 h also decreases as mordenite content increases. The pure PVA membrane has a conductivity of ; 0.03 Srcm, whereas a membrane with 50% mordenite has a conductivity of 0.01᎐0.015 Srcm. This small decrease, primarily due to changes in the PVA, is much less than the change shown in Figure 4, where methanol permeability dropped by a factor of 25 over the entire mordenite range. The high conductivity of the acid doped mordenite counteracts the decrease in polymer conductivity here, resulting in only a slight decrease in the conductivity of the composite. Thus, the 1 HR HEAT and 27 HR HEAT treatments demonstrate the importance of the ratio of proton conductivity in the mordenite vs. that in the polymer. For the 1 HR HEAT data, the mordenite is not substantially more conductive than the polymer, so protons are content moving around the mordenite and through the polymer phase. In contrast, for the 27 HR HEAT data, transport in the polymer phase is more hindered, so protons take full advantage of the highly conducting mordenite pathway. This demonstrates the decoupling of transport of protons vs. transport of methanol.
Electrochemical selecti©ity Electrochemical selectivity, defined in Eq. 8 as the proton conductivity divided by the methanol permeability, allows a comparison of the new membranes with each other and with the Nafion benchmark. As Figure 6 shows, the 1 HR HEAT and 27 HR HEAT treatments show different behavior as the mordenite content increases. For 1 HR HEAT treatment, the slight decrease in methanol permeability is accompanied by an even smaller decrease in conductivity, which results in a slight improvement over pure PVA. The composite’s data are also similar to Nafion’s selectivity. In the best case, a PVA membrane containing 60% mordenite by volume demon-
r cm) vs. mordenite Figure 5. Proton conductivity (Sr concentration.
Figure 6. Selectivity relative to Nafion vs. mordenite concentration.
For both the short heat treatment Žthe triangles. and the long heat treatment Žthe diamonds., the conductivity drops as the mordenite content increases. Maxwell predictions Žthe solid lines. agree well with both the short and long heat treatment data.
For a short heat treatment Žthe triangles ., the mordenite changes the selectivity less than two times. For the longer heat treatment Žthe diamonds . the selectivity increases 20 times. Maxwell predictions Žthe solid line . support the experimental data.
April 2003 Vol. 49, No. 4
Figure 7. Selectivity of the 27 HR HEAT PVAr rmordenite (50r r50 by volume) membranes equilibrated in various sulfuric and phosphoric acid solutions. Only a large sulfuric acid concentration Žthe diamonds . provides a conductivity value necessary for selectivity improvement. The lower conductivity of phosphoric acid, data Žthe triangles . yields little.
strates a mere two-fold increase in selectivity over Nafion. These 1 h heat treatments are not a major improvement. On the other hand, the 27 HR HEAT data demonstrate a dramatic selectivity improvement as mordenite content increases. The substantial decrease in methanol permeability is not offset by the decrease in proton conductivity, yielding better selectivity. The most successful result by far is a PVA membrane containing 50% mordenite by volume. Compared to Nafion, this membrane demonstrates a 200-fold decrease in methanol permeability, with only an order of magnitude decrease in proton conductivity, which results in ; 20-fold increase in selectivity. The polyvinylalcohol membranes studied here have no fixed charge, and so must be doped with acid to be effective in a direct methanol fuel cell. This is a disadvantage, for acid doping can cause corrosion or reduce catalyst activity. As a result, we explored broadly the effect of acid type and concentration, with results exemplified by those in Figure 7. These data, for membranes containing 50% mordenite by volume heat treated for 27 h, demonstrate a linear increase in selectivity over the acid concentration range. The results are probably due to different partitioning of acid into the membranes.
Discussion The value of composite membranes developed here and summarized in Table 1 is best judged by replotting the selectivity as shown in Figure 8. The ordinate in this figure is the proton conductivity; the abscissa is the reciprocal of methanol permeability, equivalent to a methanol resistivity. A perfect membrane, with both high proton conductivity and methanol resistivity, would be located in the upper righthand corner of this figure. To compare alternative membranes with Nafion, we plot a line of constant selectivity with a slope of y1 through the value for Nafion. Any membranes falling above this line are more selective than Nafion. Any falling below this line are not. 998
Figure 8. Selectivity plot-proton conductivity of membranes equilibrated in 1M-H 2 SO 4 vs. methanol resistance. Points labeled 0.60 Ž1 HR . and 0.50 Ž1 HR . are polyvinyl alcohol membranes containing 60% and 50% mordenite by volume, respectively, heat treated for the time noted. Points labeled PVA are pure polyvinyl alcohol membranes. Reference data are from Pivovar et al. Ž1999 ..
The composite membranes studied here include pure polyvinyl alcohol membranes heated for 1 h and 27 h, respectively, labeled PVA Ž1HR. and PVA Ž27HR.. The data also include polyvinyl alcohol membranes containing 60% and 50% mordenite by volume, heated for the time noted in parentheses and labeled 0.60Ž1HR. and 0.50Ž27HR.. All other data, with the exception of Nafion, are from earlier work ŽPivovar et al., 1999.. The figure illustrates the dramatically different behavior of membranes heat treated for 1 h vs. 27 h. A pure PVA membrane heated for 1 h has essentially the same selectivity as Nafion. The presence of 60% mordenite by volume increases selectivity slightly. In contrast, a membrane heated for 27 h and containing 50% mordenite shows a huge selectivity gain. Not only is its selectivity ; 20 times better than Nafion, but it is better than PBI, another membrane developed for direct methanol fuel cells. At this point, we may wonder, ‘‘Why does the 27 HR HEAT treatment produce better selectivity than the 1 HR HEAT treatment?’’ The answer to this question depends on the conductivity in the mordenite vs. that in the polymer. For the 1 HR HEAT treatment, the mordenite phase isn’t substantially more conductive than the polymer, so the majority of protons are content moving through the polymer phase. In this scenario, shown as Case 1 of Figure 2, both methanol and protons travel around and between the mordenite particles, staying in the polymer. Since both species move by the same pathway through the polymer phase, the composite selectivity is roughly equivalent to the pure PVA selectivity. No selectivity gain is realized. On the other hand, for the 27 HR HEAT treatment, transport in the polymer phase becomes drastically reduced by
April 2003 Vol. 49, No. 4
mordenite-induced dehydration. The conductivity of the mordenite is now substantially greater than that of the polymer. As shown as Case 2 of Figure 2, protons take full advantage of the highly conducting mordenite pathways, whereas methanol is still forced around the mordenite particles. This separation of transport mechanisms results in the substantial selectivity improvement. To place this argument on a more quantitative basis, we return to the Maxwell theory for transport through a periodic array of spheres. For 1 HR HEAT treatment, we assume that methanol cannot penetrate the mordenite, so the Maxwell theory reduces to Eq. 11. With the experimental value of PP s1.0=10y6 cm2rs from the data in Figure 4, we can use Eq. 11 to predict the composite methanol permeability for various mordenite volume fractions. The results, shown in Figure 4 as the solid line, show that the Maxwell prediction agrees reasonably with the experimental data. The experimental data are slightly higher than predicted, especially at the large mordenite concentration. The discrepancy may be a result of using a theory that is strictly applicable only to dilute suspensions of periodic particles, which is not true here. Nevertheless, this theory has proved applicable for higher concentrations in polymerrzeolite gas separation membranes ŽZimmerman et al., 1997.. Here, the Maxwell theory provides a reasonable prediction for transport at various mordenite concentrations in the composite membranes. For the 1 HR HEAT treatment data, the conductivity decreases with increasing mordenite content, just as methanol permeability does. Equation 10 predicts the decrease, as shown in Figure 5, especially at the lower volume fractions where the Maxwell theory is more reliable. Thus, protons and methanol move in a similar manner by transport primarily in the polymer phase, which results in the lack of selectivity improvement seen in Figure 6. The Maxwell predictions with both the methanol and proton conductivity data explain this selectivity. The 27 HR HEAT treatment gives different results. As shown by the lower solid line in Figure 4, use of Eq. 11 systematically overpredicts the methanol permeability of the composite. The main difference is the mordenite-induced dehydration of PVA during this treatment. Because the extent of the dehydration depends on the amount of mordenite present, the transport properties through the polymer phase became a function of the mordenite content. As a result, the predictions for the longer heat treatment data use a new procedure. Data for the specific example given below clarify how the predictions are made. The data, which appear in parentheses at the end of each step, are for a PVArmordenite Ž50r50 by volume. composite membrane, heated for 27 h, and equilibrated in 1M-H 2 SO4 . The procedure is as follows: Ž1. Start with the methanol permeability of the composite in question. Label this as P. ŽThe PVArmordenite composite membrane has a permeability of P s1.25=10y8 cm2rs . Ž2. Assuming methanol doesn’t permeate through the mordenite, calculate the methanol permeability of the polymer phase PP from Eq. 11. Ž Substitute s 0.5 for the ®olume fraction of mordenite and P s1.25=10y8 cm2rs to calculate PP s 3.13=10y8 cm2rs .. Ž3. Compare the methanol permeability of pure PVA Ž PPVA . with the polymer phase in the composite Ž PP . to find AIChE Journal
the permeability reduction Permeability Reduction s
Ž Substitute PPVA s 3.0=10y7 cm2rs and PP s 3.13=10y8 cm2rs to yield a Permeability Reductions9.6. Ž4. Because experiments show that PVA dehydrated in acid has the same selectivity as pure PVA heated in the absence of acid ŽLibby, 2001., the reduction in permeability must be accompanied by an equivalent reduction in conductivity Conductivity Reduction s
ŽThe conducti®ity reduction is 9.6. Ž5. Calculate P using the known values of proton conductivity in pure PVA Ž PVA . and the conductivity reduction factor. Ž In this case a pure PVA membrane heated for 27 h has a proton conducti®ity PVA s 3.0=10y2 Srcm. Thus, P s Ž3.0= 10y2 Srcm.r9.6 s 3.13=10 37 Srcm.. Ž6. Assume that the proton conductivity of the mordenite particles equilibrated in acid are equivalent to the proton conductivity of the solution itself ŽLibby, 2001.. Ž From the conducti®ity of a 1 M y H2 SO4 solution, M s 0.4 Srcm.. Ž7. Calculate the conductivity of the composite Ž . using the Maxwell theory in Eq. 10: Ž Substitute ␥ s MrP s Ž0.4 Srcm.Žr3.13=10y3 Srcm. and s 0.5 to calculate s 0.012 Srcm.. Ž8. Finally, compare the predicted composite conductivity value to the experimental value. ŽThe calculated ®alue for composite conducti®ity of s 0.12 Srcm is within the experimentally measured range of 0.010᎐0.015 Srcm.. This calculation provides quantitative support for qualitative picture in Figure 2. It not only predicts the result of the example given, but it also predicts the conductivity variations with mordenite concentration shown in Figure 5. Additionally, by combining the methanol permeability results with the conductivity predictions, the Maxwell theory successfully predicts the selectivity changes shown in Figure 6. We conclude that the Maxwell theory can be a reasonable guide to selectivity changes in these membranes. On this basis, we can ask how much more selective we can hope to make the composites. To explore this, we consider the limit of Eq. 12 as the mordenite conductivity becomes infinite, that is, as ␥ ™⬁. The result, given by Eq. 13, gives a maximum predicted selectivity improvement of 10 when s 0.5. From Figure 6, our experimental results appear to have reached this maximum increase, as the PVArmordenite Ž50r50 by volume. composite membrane heated for 27 h is ten times more selective than its pure PVA counterpart. Because the Maxwell theory is successful, we can use Eq. 13 to predict the relative selectivity Ž ␤r␤P . as a function of the composite’s conductivity ratio ␥ Ž s SrP .. The result, shown in Figure 9 for two values of mordenite volume fraction , suggests that this conductivity ratio must be well above ten to approach the maximum selectivity.
April 2003 Vol. 49, No. 4
Acknowledgments This work was primarily supported by the US Army Žgrant ARO DArDAAH04-95-1-0094-MURI. and the U.S. Air Force Žgrant F49620-01-1-0333.. Other support came from the Petroleum Research Fund of the American Chemical Society Žgrant 36528-AC9..
Figure 9. Achieving the largest possible selectivity. The highest selectivity Ž ␤r␤ P . requires a zeolite with a relative conductivity more than ten times the polymer continuum.
This key transport ratio criteria sheds light on the commonly asked question of ‘‘Why not put mordenite particles into a fuel cell ready membrane such as Nafion?’’ The answer is that mordenite Ž ;10y3 Srcm. or sulfuric acid filled mordenite Ž ; 0.4 Srcm. does not have a conductivity many times greater than Nafion, which is necessary to achieve maximum selectivity in a composite membrane. According to the present theory, the conductivity of the dispersed phase would have to be ;10 Srcm, since the typical room temperature conductivity value for Nafion is ; 0.1 Srcm. Unfortunately, such a solid proton conductor currently does not exist ŽNorby, 1999.. The results above do suggest that we have achieved the maximum selectivity increase possible with evenly dispersed particles of a proton conductor. However, this is definitely not the maximum selectivity increase which could be achieved. Such a maximum depends strongly on how the proton conducting material is arranged within the composite membrane. For example, if the particles were all on one surface of the membrane, then any selectivity increases would be minor. If they were arranged in conducting channels passing through the membrane, then the selectivity increase could be much larger. This remains a fruitful area for more research.
Conclusions The PVArmordenite composite membranes have fulfilled their potential. In particular, a PVArmordenite membrane containing 50% mordenite by volume, heat-treated for 27 h to make it insoluble, demonstrates a 20-fold improvement in selectivity over Nafion. This successful result contrasts the lack of selectivity improvement seen for the short heat treatment composites, thus demonstrating the importance of tailoring the transport properties between the polymer and zeolite phases in order to achieve selectivity gains. Specifically, the zeolite conductivity must be at least 100 times greater than the polymer phase conductivity to achieve maximum selectivity. The polymer phase cannot solely be responsible for the selectivity gains. The experimental results are well correlated by Maxwell’s theory for conducting spheres in a continuum. 1000
c i sconcentration of species i Dsdiffusion coefficient of composite membrane Di sdiffusion coefficient of species i F sFaraday’s constant Hi spartition coefficient of species i iscurrent density ji sflux of species i l smembrane thickness P spermeability Di Hi Rsgas constant T stemperature z i schange of species i ␣ sgas constant ␤ selectrochemical membrane selectivity ␥ sratio of conductivity in mordenite to that in polymer ⑀ 1 , ⑀ 2 sratio in composite to that in polymer for conductivity or permeability, respectively sconductivity ⌽ selectrochemical potential svolume fraction
Literature Cited Appleby, A. J., ‘‘The Electrochemical Engine for Vehicles,’’ Scientific Amer., 74, ŽJuly, 1999.. Berry, M. B., B. E. Libby, K. Rose, K.-H. Haas, and R. W. Thompson, ‘‘Incorporation of Zeolites into Composite Matrices,’’ Microporous and Mesoporous Mat., 39, 205 Ž2000.. Breck, D. W., Zeolite Molecular Sie®es: Structure, Chemistry, and Use, Wiley, New York Ž1974.. Business Week, ‘‘A Clean Technology Powers Up,’’ Business Week, 3680, 102 Ž2000.. Caro, J., M. Noack, P. Kolsch, and R. Schafer, ‘‘Zeolite Membranes ᎏ State of their Development and Perspective,’’ Microporous and Mesoporous Mat., 38, 3 Ž2000.. Coronas, J., and J. Santamaria, ‘‘Separations Using Zeolite Membranes,’’ Sep. and Purif. Methods, 28, 127 Ž1999.. Cussler, E. L., Diffusion, Cambridge University Press, Cambridge Ž1997.. Cyer, C. K., ‘‘Replacing the Battery in Portable Electronics,’’ Scientific Amer., 88 ŽJuly, 1999.. Ho, W. S. W., and K. K. Sirkar, Membrane Handbook, van Nostrand Reinhold, New York Ž1992.. Immelman, E., R. D. Sanderson, E. P. Jacoba, and A. J. Van Reenen, ‘‘Poly ŽVinyl alcohol. Gel Sublayers for Reverse Osmosis Membranes: 1. Insolubilization by Acid-Catalyzed Dehydration,’’ J. of Appl. Poly. Sci., 50, 1013 Ž1993.. Jacoby, M., ‘‘Fuel Cells Heading for Sale,’’ Chem. and Eng. News, 31 Ž1999.. Ji, W., and S. K. Sikdar, ‘‘Pervaporation using Adsorbent-Filled Membranes,’’ Ind. Eng. Chem. Res., 35, 1124 Ž1996.. Jost, K., ‘‘Fuel-Cell Concepts and Technology,’’ Automoti®e Eng. Int., 108, 173 Ž2000.. Kim, K.-J., S.-H. Park, W.-W. So, and S.-J. Moon, ‘‘Pervaporation Separation of Aqueous Organic Mixtures through Sulfated Zirconia-PolyŽvinyl Alcohol. Membrane,’’ J. of Appl. Poly. Sci., 79, 1450 Ž2001.. Kjaer, J., J. Yale-Andersen, and N. A. Knudsen, ‘‘Solid State Electrolyte Membranes for Direct Methanol Fuel Cells,’’ Solid State Ionics, 46, 169 Ž1991.. Knudsen, N., E. K. Andersen, I. G. K. Andersen, and E. Skou, ‘‘TinŽIV. Oxide Containing Mordenite: Synthesis and Ionic Conductivity Measurements,’’ Solid State Ionics, 28᎐30, 627 Ž1988.. Libby, B., ‘‘Improving Selectivity in Methanol Fuel Cell Membranes: a Study of a Polymer-Zeolite Composite Membrane,’’ PhD Thesis, University of Minnesota Ž2001..
April 2003 Vol. 49, No. 4
Libin, K., ‘‘Power Play,’’ Canadian Business, 73, 92 Ž2000.. Maxwell, J. C., Treatise on Electricity and Magnetism, Claredon Press, London Ž1881.. Meininghaus, C. K. W., and R. Prins, ‘‘Sorption of Volatile Organic Compounds on Hydrophobic Zeolites,’’ Microporous and Mesoporous Mat., 35-36, 349 Ž2000.. Narayanan, S. R., A. Kindler, B. Jeffries-Nakamura, W. Chun, H. Frank, M. Smart, T. I. Valdez, S. Surampudi, and G. Halpert, ‘‘Recent Advances in PEM Liquid-Feed Direct Methanol Fuel Cells,’’ Proc. Battery Conf. on Ad®ances and Applications, Long Beach, CA Ž1996.. Noack, M., P. Kolsch, J. Caro, M. Schneider, P. Toussaint, and I. Sieber, ‘‘MFI Membranes of Different SirAl Ratios for Pervaporation and Steam Permeation,’’ Microporous and Mesoporous Mat., 35-36, 253 Ž2000.. Norhy, T., ‘‘Solid-State Protonic Conductors: Principles, Properties, Progress, and Prospects,’’ Solid State Ionics, 125, 1 Ž1999.. Okumus, E., T. Gurkan, and L. Yilmar, ‘‘Development of a MixedMatrix Membrane for Pervaporation,’’ Sep. Sci. and Technol., 29, 2451 Ž1994..
Pivovar, B. S., Y. Wang, and E. L. Cussler, ‘‘Pervaporation Membranes in Direct Methanol Fuel Cells,’’ J. Memb. Sci., 154, 155 Ž1999.. Pritchard, J. G., Poly(®inyl Alcohol): Basic Properties and Uses, Gordon and Breach Science Publishers, London Ž1970.. Rao, N., T. P. Andersen, and P. Ge, ‘‘Tin Mordenite Membranes for Direct Methanol Fuel Cells,’’ Solid State Ionics, 72, 334 Ž1994.. Ren, X., T. A. Zawodzinski, F. Unbe, H. Dai, and S. Gottensfeld, ‘‘Methanol Cross-over in Direct Methanol Fuel Cells,’’ Electrochemical Soc. Proc., 95᎐23, 284 Ž1995.. Sakaruda, I., Poly®inyl Alcohol Fibers, Marcel Dekker, New York Ž1985.. Suer, M. G., N. Bac, and L. Yilmar, ‘‘Gas Permeation Characteristics of Polymer-Zeolite Mixed Matrix Membranes,’’ J. Memb. Sci., 91, 77 Ž1994.. Zimmerman, C. M., A. Singh, and W. J. Koros, ‘‘Tailoring Mixed Matrix Composite Membranes for Gas Separations,’’ J. Memb. Sci., 137, 145 Ž1997.. Manuscript recei®ed Feb. 11, 2002, and re®ision recei®ed No®. 13, 2002.
April 2003 Vol. 49, No. 4