Scientific papers

1 – Direct formic acid oxidation for liquid-fed PEM fuel cells

Published on February 2012 by Catherine Lepiller, PhD, for Pragma Industries

With the ever growing power demand of portable electronic appliances such as cell phones, MP3 players, laptop computers and similar consumer products in the 0.5-100 W range, micro fuel cells, and in particular low temperature PEM micro fuel cells, appear as the best bet for the future due to a higher energy density than batteries. However, the storage and transfer of volatile hydrogen gas in solid systems for mobile use is difficult because of the relatively low volumetric and weight energy density of metal hydride compounds. In such context, hydrogen chemically stored in a primary organic liquid for on-demand in situ release, at room or near-room temperature and without carbon monoxide contamination, from a disposable or recyclable cartridge, seems to be a very valuable option for portable applications.

Besides methane and methanol, renewable resources, such as (bio)ethanol and glycerol, have been considered as promising routes for hydrogen production. But all related reforming processes run at temperatures higher than 200°C, which renders their practical use problematic. This explains why the search for improved processes generating hydrogen from liquid fuels at higher reaction rates and under milder, well-controlled conditions, is an active area of interest in the fuel cell field.

Among possible liquid fuel candidates, formic acid (HCOOH) definitely shows a great potential as in situ source of hydrogen for fuel cells. Compared with methanol for example, it offers higher energy density and faster electro-oxidation kinetics at room temperature; and not to be overlooked as well, it is non toxic and can be handled in aqueous media and stored safely. Formic acid is even found on the US Food Drug Administration list of food additives!

The electro-oxidative decomposition of formic acid can follow two possible pathways, the first one yielding carbon monoxide and water (1), and the second one yielding carbon dioxide and hydrogen (2). Both reactions are dependent on the catalyst surface, formic acid concentration and temperature:

HCOOH = CO + H2O (1)
HCOOH = CO2 + H2 (2)

Formic acid/hydrogen cycle

Reaction (1) leading to CO has to be minimized by a proper choice of selective catalyst and experimental conditions, while the reversible reaction of CO2 hydrogenation (2) is a potential way to produce hydrogen.
Notably, formic acid has been widely used as hydrogen source in liquid-phase transfer hydrogenation reactions of carbon dioxide in the presence of base such as amines under ambient conditions (Fig. 1). Experimental conditions of these well-established processes could therefore serve as starting point for initial studies on acid formic electro-oxidation.
Figure 1. A CO2- neutral cycle for the storage of hydrogen in formic acid base adducts.
It is only in recent years that the feasibility of generating hydrogen on demand from mixtures of formic acid and amines at low temperatures < 100°C was demonstrated [1,2]. Excellent catalytic activities over homogeneous ruthenium-based compounds were reported: for example, Loges et al. [1] investigated the decomposition of formic acid in excess of amine in the presence of different catalysts at 40°C. A mixture of acid formic HCOOH and triethanolamine NEt3 at a formic acid to nitrogen atom ratio of 5:2 was chosen as base case owing to its prior use as hydrogen source in transfer hydrogenation reactions. The influence of different types of amines was then explored using the best performing Ru(II) catalyst at the same ratio 5:2, and additional experiments showed some improvement with higher amine content in the solution. Activities were only slightly lower at 26.5°C compared to 40°C. Loges et al. were eventually the first to be able to generate hydrogen from formic acid-amine adducts at high rates at ambient temperature with a commercially available Ru(II) complex [formula RuCl2(PPh3)3 with Ph = aromatic group C6H6]. An easy gas cleaning by charcoal is then applied to remove traces of amine. Compared with previous hydrogen generating systems based on organic liquids, their process can be operated at much lower temperatures and hydrogen is fed directly to the fuel cell.

In a close field, a Swiss patent in 2008 by the Ecole Polytechnique Fédérale de Lausanne [2] provided a method for converting formic acid to hydrogen continuously by a homogeneous catalytic reaction in aqueous media at high reaction rates in the low-temperature range 30-180°C. During the process, hydrogen is produced at partial pressures up to 600 bar convenient for direct on-demand feeding of a variety of hydrogen consuming devices such as fuel cells, and free of any CO by the reaction (2).

In the invention, the catalyst is a ruthenium-based complex (with general formula Ru(II)Ln) dissolved in acidic aqueous solution at high concentrations > 200 g/L. The solution is buffered at 2.5-4.5 by adding formate salt HCOO-,X+ [with X+ = inorganic cation: Na+, Ca2+, Li+, K+, NH4+], and in contrast to reference [1], there is no other acid or amine added to the reaction medium.

The two gaseous products H2 and CO2 are readily separated from the solution, since the pressurized gases just evaporate once generated. Hydrogen is then physically separated from co-product CO2 by exploiting properties such as melting temperature, volatility and/or diffusion coefficient that strongly differ between the two gases (the exact method is not described in detail). The Ru(II) catalyst is easily separated from the reaction products as well, due to the high solubility of the catalyst in the liquid phase and hardly any solubility in the gaseous phase. This way, it is entirely recycled and keeps effective for prolonged time without degradation.

The Swiss teamwork evaluated a range of highly water-soluble ligands L having at least one P atom (or a carbene group in certain embodiments), one aromatic group C6H6 and one hydrophilic group, which could satisfy to catalyst activity requirements toward formic acid oxidation under mild reaction conditions, and already at room temperature. Sulfonation of the aromatic group is an efficient way to increase water solubility of the Ru(II) complex. Like in reference [1], a commercial product was identified as sufficiently soluble and stable for producing hydrogen at proper rates and pressures: Ru(TPPTS)2X2 [with TPPTS = tris(3-sulfophenyl)phosphine; X = sulfonated anion: tosylate CF3SO2O− or triflate CH3C6H4SO2O−].

They also found that the rate and conversion efficiency of the oxidation reaction could be controlled by varying the ratio of formic acid to formate ion in the solution, and identified an optimum value of 9:1. Reaction rate directly correlates with temperature, therefore it is slower at room temperature; yet, high conversion efficiencies are already achieved and high-pressure gases >> 1 bar are still produced under these conditions because pressure correlates with conversion yield. In the final system, hydrogen would be directed to the device through a buffer tank, and a valve would allow tuning the entry pressure to match the amount of hydrogen used by the device.

These works solved a number of drawbacks related to homogeneous catalytic reactions, like the separation of reactants and products, the use of organic solvents, ligands and additives that may lead to severe problems in scale-up. Nevertheless, the use of solid catalysts is becoming more like a better alternative than catalysts in solution: after a period where, despite the anticipated benefits in catalyst separation and processing, there was a consistent lack of such catalysts that were sufficiently active and/or selective for formic acid oxidation, research in the field finally took off.

Initial studies using solid catalysts were all carried out at temperatures higher than 50°C: such conditions raised issues related to volatility of the formic acid/water mixture and to heat management, which both were detrimental to miniaturization. Besides, although some catalysts exhibited a promising initial activity, rapid deactivation and concomitant CO production from parasitic side-reaction (1) at the levels of hundreds of ppm were generally reported.

Recently, palladium-based catalysts were identified to possess the highest activity toward formic acid oxidation among the transition metal series – including platinum –due to the optimum adsorption energy of hydrogen atoms on Pd surface. It was also demonstrated that the oxidation reaction mainly occurs through the direct pathway onto Pd instead of the dual pathway on Pt: thus, Pd-based catalysts are highly selective for reaction (2) and can overcome the CO poisoning effect. Following the « Pd line », better performing catalysts are currently being found, as exemplified below.

Formation mechanism of PD/C catalyst

In a first example, Liang et al. [3] improved the conventional precipitation-reduction procedure used to synthesize Pd/C nanoparticles from aqueous solutions by adding a hydrophilic stabilizer BDPA (1,4-butylenediphosphonic acid) and adjusting the pH to 9 (Fig.2). Under these basic conditions, Pd nanoparticles are in oxidized and colloidal state (i.e. hydrated), and BDPA is in its anionic form: one phosphonate group per BDPA molecule adsorbs onto the Pd-O surface while the second group extending outside has a strong electrostatic repulsion effect between nanoparticles.In the consecutive step, the nanoparticles are reduced by NaBH4; finally, the adsorbed phosphonate anion is easily removed by a simple rinsing with NaOH 0.1 M, which is made possible by the weak Pd-O-P bond at high pH.

Figure 2. Formation mechanism of PD/C catalyst by the improved precipitation-reduction process.

This procedure allows obtaining a high and stable dispersion of Pd/C nanoparticles with small particle size and very clean surface. Electrochemical performance confirmed that the activity of Pd/C nanoparticles prepared by the BDPA-based method was four times greater than that of Pd/C-NaBH4 prepared by the standard method. However, the initial decay of formic acid oxidation current at 0.15 V at room temperature is indicative of some kind of surface poisoning that must be looked for.

In a further effort, researchers at the University of Oxford prepared by wet chemical synthesis various engineered core-shell nanoparticles having an inner core of a metal element and an external shell of palladium [4]. As it has been widely adopted in research for fuel cell oxygen reduction catalysts, the strategy behind the bi-metallic core-shell structure is to retain the desirable Pd surface properties and play with the interacting effects of a particle core of a different metal. Moreover, this configuration, which spares most of the expensive noble metal Pd, is quite attractive from an economical point of view.

Rates of formic acid decomposition

Among all metals tested, the highest activity toward formic acid oxidation at room temperature was found for Ag@Pd nanoparticles (diameter 8 nm) with the thinnest continuous Pd shell (1-2 atomic layers)Turnover frequencies per surface Pd site were comparable to homogeneous catalysts: 125 h-1 at 20°C and 252 h-1 at 50°C.At 20°C, an equimolar mixture of hydrogen and CO2 was continuously produced without any trace of CO; on the other hand, CO was detected at temperatures higher than 50°C. In contrast to reference [3], no deactivation effects were observed during the time of experiments, and the rate closely followed the first order kinetics of formic acid decomposition by reaction (2)Furthermore, electrochemical experiments were in excellent agreement with theoretical calculations showing a strong correlation between the catalytic activity and the work function of the metal core: the largest net difference with the work function of the Pd shell will lead to the highest adsorption energy by charge transfer from the core to the shell, hence to the best possible activity of the resulting bi-metallic structure for formic acid electro-oxidation (Fig.3). The very short range of this so-called "ligand" electronic effect between the two metals explains why the highest performance is achieved for the thinnest Pd layer. Nanomaterials interface definitely plays a key role in catalysis.

Figure 3. Plot of rates of formic acid decomposition monitored by 13C NMR over different metal colloid catalysts. Correlation with the work function of the M core, where M = face-centred cubic (111) Ag, Rh, Au, Ru and Pt or hexagonal close-packed (0001) Ru, Ag, with the largest difference in work function in relation to Pd, gives the strongest electron promotion to the Pd shell.

In the light of previous results, engineered bi-metallic nanoparticles like the ones presented here appear as a very valuable option yielding convenient rates of pure hydrogen gas production from formic acid oxidation at room temperature. Based on the values of reference [4], it was roughly determined that 10-100 mg of catalyst would be enough to supply hydrogen for micro fuel cell operation. Would be current encouraging achievements confirmed in the future, they would no doubt boost efforts in the microfluidic fuel cell field… Coming next!

[1] Controlled generation of hydrogen from formic acid amine adducts at room temperature and application in H2/O2 fuel cells. B. Loges, A. Boddien, H. Junge and M. Beller, Angew. Chem. Int. Ed., 47 (2008), 3962-3965.
[2] Hydrogen production from formic acid; patent WO 2008/047312 A1 ; Ecole Polytechnique Fédérale de Lausanne.
[3] Highly dispersed carbon-supported Pd nanoparticles catalyst synthesized by novel precipitation-reduction method for HCOOH electro-oxidation. Y. Liang, M. Zhu, J. Ma, Y. Tang, Y. Chen and T. Lu, Electrochim. Acta, 56 (2011), 4696-4702.
[4] Hydrogen production from formic acid decomposition at room temperature using a Ag-Pd core-shell nanocatalyst. K. Tedsree, T. Li, S. Jones, C.W.A. Chan, K.M.K. Yu, P.A.J. Bagot, E.A. Marquis, G.D.W. Smith and S.C. Edman Tsang, Nature Nanotechnol. (2011). DOI: 10.1038/NNANO.2011.42.

2 – Novel architectures for microporous layers without binding agent

Published on April 2011 by Catherine Lepiller, PhD, for Pragma Industries

My two previous notes were focused on porous layers in low temperature fuel cells, i.e. the gas diffusion layer (GDL), the microporous layer (MPL) and the catalyst layer (CL), which make up a crucial part of the fuel cell core and are involved in water management. In these three components with graded porosity and pore size distribution, water exists in liquid and vapor forms. For high performance and durable operation, water must be efficiently transported from the cathode CL, where it has been produced by the oxygen reduction reaction, to the external gas circuit. After years of contradictory interpretations of experimental results, the different interactions between porous layers have been eventually – almost? – elucidated. Meanwhile, the beneficial role of the MPL has been evidenced with help of both improved models and decisive experiments (check previous notes on water extraction here and mechanisms in microporous layers here).

Like other porous components of a fuel cell, the MPL is carbon-based due to high electronic conduction and chemical resistance requirements. Usual carbon sources include several types of carbon blacks (furnace, acetylene, etc.), which are all characterized by rounded nanoparticles in the ten to hundred nanometers range, a very porous structure and high surface area. Another requirement for the MPL is a high level of hydrophobicity to form a good capillary barrier to liquid water. Anti-flooding properties are typically obtained by mixing the carbon powder with a fluorinated polymer like PTFE, FEP or PVDF. In the conventional process, the hydrophobic polymer is introduced as a dispersion/solution and the resulting ink is deposited onto the gas diffusion media by a spraying or printing method.

However, a number of studies have started suggesting a stability issue with those polymers during long term operation, and especially with PTFE, which is the current standard. PTFE’s structural integrity in the bulk MPL (and GDL) is challenged under fuel cell’s warm and humid conditions, leading to a gradual loss of hydrophobicity as operating time increases. With that in mind, I picked up in recent literature a few recent experimental studies considering the architecture of the cell core without polymeric binding agent.

SEM images of GDL surface

After the seminal work of G.G. Park et al. [1] about the alternative use of carbon nanomaterials such as nanofibers and nanotubes as co-components with carbon black for replacement of PTFE in the MPL, Kannan et al. from the US Arizona State University [2] have further explored the idea and successfully grown in situ multi-wall carbon nanotubes (CNTs) as novel MPLs onto plain carbon paper substrate by an initial surface modification step followed by chemical vapor deposition (CVD) of a carbon source and Fe-based catalyst at 800°C. Unfortunately, the thickness of the carbon nanotube-based layer was not mentioned by the authors; but accounting for the surface deposition technique used, it was likely much thinner than standard sprayed or printed 20-100 µm-thick MPLs. Structural investigation clarified the composite nature of the MPL as an intimate mixture of entangled CNTs and oblong-shaped carbon nanoparticles homogeneously covering the carbon paper fibers and forming a porous but robust 3-D network without reduction of the initial GDL’s macroporosity in the 30 µm range (Figure 1). A complete characterization of the layers also demonstrated their intrinsic highly hydrophobic properties (the contact angle to water droplet was 145°) thereby ruling out the need of additional hydrophobic agent. When evaluated in a single fuel cell at 80°C with a standard Pt/C catalyst-coated Nafion? 212 membrane, these in situ grown CNT-based GDLs exhibited performances fairly insensitive to gas relative humidity conditions under oxygen and air feeding (Figure 2): the good resistance to water flooding was explained by the high hydrophobicity of the pristine CNTs and their uniform distribution onto the surface of the carbon fibers retaining large pore diameters, hence allowing quicker removal of product water.

Daniel Shua’s team at the National University of Singapore employed a similar Fe-catalyzed CVD technique at 750°C for synthesizing an in situ layer of carbon nanotubes on carbon paper from acetylene gas carbon source and further sputtered a pure platinum layer to make a Pt/CNT-based cathode [3]. They could achieve a dense carbon layer with nanotube diameters of 20-30 nm characterized by strong coiling and twisting of the curled nanotubes together. Consistently with [2], the 3-4 µm-thick nano carbon layer exhibited high surface roughness and porosity. It was also composed of a mixture of crystalline multi-wall carbon nanotubes and amorphous carbon, which was mostly located close to the surface. Interestingly, the disordered surface of the in situ as-grown CNTs is thought to provide a stable interface between the carbon nanotubes and the pure Pt particles due to their pristine high oxidation level. This permitted direct sputter deposition without additional surface oxidation step. The surface roughness of the layer could give rise to a scaly Pt layer with homogeneously dispersed nanoparticles 2-3 nm in diameter instead of the continuous layer usually achieved by the sputtering method (Figure 3).

A gas diffusion cathode with Pt loading controlled at 40 µg/cm² fabricated by this combined technique was implemented in a Nafion? 112-based fuel cell with a standard anode and run in hydrogen/oxygen configuration with fully humidified gases at 80°C and 2 atm pressure. Tang et al. did not vary the fuel cell testing conditions as in [1,2] but rather looked at catalyst and carbon support corrosion by in situ accelerated tests [3]: they found that compared with the conventional carbon black-based electrode, the Pt/CNT/carbon paper-based electrode shows a significant improvement in electrochemical activity and stability. The initial activity before oxidation was significant unlike Pt/carbon black electrode, and after 100 oxidation cycles, the activity loss was much less pronounced (Figure 4), which strongly suggests that the CNT support is more corrosion resistant than standard carbon black. The corrosion current measured under cyclic step potential between 0.6 and 1.8 V was also much lower, with a main part that may be readily attributed to the fraction of amorphous carbon at the surface.

The most recent study published by the Singaporean team deals more specifically with the influence of CVD synthesis conditions on the microstructure of the CNT layers [4]. They show that sufficiently high acetylene gas flow rates during CNT growth can provide layers at a weight ratio of 1-1.5% on carbon paper with both high density and porosity due to enhanced lateral growth, thus yielding refined gas diffusion channels and larger surface area for Pt deposition. Fuel cell tests performed under identical conditions as in [3] exhibit a significant improvement in the whole current density range for the integrated Pt-sputtered/in situ grown CNT cathode-based MEA compared to those of the Pt-sputtered/carbon black and Pt-sputtered/commercial CNT cathode-based MEAs prepared by the conventional ink fabrication process. These results further demonstrate the better catalyst utilization and mass transport properties achieved by this novel dense while porous morphology allowing larger penetration depth of sputtered Pt nanoparticles.

Above studies based on the chemical vapor deposition method, are quite convincing about the relevance of replacing carbon black by carbon nanotubes as integrated catalyst support/microporous layer. Nevertheless, an important question left to be raised is about cost and viable mass production for such a high temperature process. Some scientists have followed this pragmatic rationale and considered other fabrication techniques for carbon nanomaterial layers. For example, a feasibility study was conducted at the Dongshua University in China [5]. It was related to the development of carbon nanofibers using a combination of electrospinning of polyacrylonitrile (PAN) solution under high voltage and subsequent thermal treatments. Electrospinning is described as a versatile technique to process solutions or melts into a 3-D web of fibers with diameters in the nanometer to micrometer range, and stated as being both easy and cheap to implement at the industrial scale, which is undoubtedly a good starting point!

Figure 1. SEM images of GDL surface fabricated by using in situ CVD on carbon paper at: (a) low and (b) high magnifications.
FC polarization data at 80°C
Figure 2. Fuel cell polarization data at 80°C using H2/air at various RH conditions. (a) For cathode using Pt deposited on the in situ CNT/carbon paper GDL. (b) For cathode using Pt deposited on the standard wire-rod-coated GDL.
High resolution TEM image showing Pt nanoparticles
Figure 3. High Resolution TEM image showing Pt nanoparticles on a carbon nanotube tip.
Cyclic voltammograms of Pt/CNT catalysts
Figure 4. Cyclic voltammograms of Pt/CNT catalysts before and after 10 and 100 oxidation cycles between 0.1 and 1.2 V.
According to the process, carbon nanofibers 400-700 nm in diameter were produced as self-standing sheets after stabilisation at 250°C in air and carbonisation in nitrogen at 900°C of the room temperature as-spun PAN nanofiber sheets. These carbon nanofiber sheets (CNFSs) were then directly transferred onto plain carbon fiber paper by hot-pressing to serve as MPL with a uniform thickness of about 25 µm. Like previous CNT-based MPLs grown by chemical vapor deposition [2-4], the present nonwoven CNFSs have a porous morphology with open 3-D interconnected structure that clearly enhances reactant gases transport. Gas permeability of the GDL having a nanofiber sheet-based MPL is about twice that of a GDL with a conventional MPL made from slurry of carbon black powder and FEP suspension, in agreement with Park’s team results [1]. CNFSs are also highly hydrophobic with an average water contact angle of 136°, which allows using them as single component of an MPL. Fuel cell performances were also tested at 70°C under 100% RH hydrogen and air with the same type of Nafion? 212-based catalyst coated membrane as in [2] including a CNFS layer at the cathode side: a promising 23% power improvement was achieved as compared to the conventional structure. Yet, what is absent there is some durability testing to check the good initial behaviour of the carbon nanofiber sheets as operation goes on…

All previous results are clearly consistent about the great potential of using carbon nanotubes/nanofibers as free-standing MPLs without hydrophobic binding agent, and even as integrated catalyst and diffusion layer for low temperature fuel cell applications. These new micro-layer architectures are intrinsically hydrophobic and do not require mixing with a polymer like PTFE to achieve anti-flooding properties. They form also both highly porous and mechanically robust structures thanks to the three-dimensional network of nanotubes/nanofibers. Finally, despite high porosity, electric conductivity is comparable to standard MPLs with reported values about 60 S/cm [5]. Such favorable characteristics should push further the questioning of current hydrophobic polymer-based microporous layers and their insufficient reliability. As already noted above, an important point still to be addressed by ongoing research is about the proper fabrication method of these materials that would be compatible with future mass production and commercialization.

[1] G.G. Park, Y.-J. Sun, S.-D. Yim, T.H. Yang, Y.-G. Yoon, W.-Y. Lee, K. Eguchi and C.-S. Kim, J. Power Sources, 163 (2006), 113.
[2] A.M. Kannan, P. Kanagala and V. Veedu, J. Power Sources, 192 (2009), 297.
[3] Z. Tang, H.Y. Ng, J. Lin, A.T.S. Wee and D.C. Chua, J. Electrochem. Soc., 157 (2010), B245.
[4] Z. Tang, C.K. Poh, Z. Tian, J. Lin, H.Y. Ng and D.C. Chua, Electrochim. Acta, 56 (2011), 4327.
[5] Q. Duan, B. Wang, J. Wang, H. Wang and Y. Lu, J. Power Sources, 195 (2010), 8189-8193.

3 – How is liquid water generated at the cathode transported away during fuel cell operation?

Published on March 2011 by Catherine Lepiller, PhD, for Pragma Industries

After my September’s science note about the beneficial role of the microporous layer in low temperature fuel cell’s water management, further highlights in the following will expand the subject to all porous components adjacent to the membrane. The catalyst layer and the gas diffusion layer – with or without a microporous layer – interplay and contribute to mass transport losses in different ways, especially on the cathode side.

How is liquid water generated at the cathode transported away during fuel cell operation? This question has proved difficult to address experimentally; nevertheless, smart ideas and clever lab setups, in conjunction with the ongoing development of more realistic non isothermal models, are improving our comprehension of the actual fuel cell physics for better component’s material design.

In recent years, significant work has been devoted to experimental and modelling studies of the water transport from the cathode catalyst layer (CCL) to the bipolar plate channels in liquid form by the capillary-fingering mechanism. As reported in my previous note, pore network models have clearly evidenced that the microporous layer (MPL) acts like a physical capillary barrier: it reduces the number and size of liquid water droplets formed in the cathode nano/micro interfacial pores to be directed towards the macroporous gas diffusion layer (GDL) and finally blown away with the exit air [1,2]. This process decreases the overall saturation in the GDL appreciably, and hence helps mitigate flooding issues. In short, the cathode microporous layer controls both the quantity – lowered by more water back diffusion to the anode through the membrane – and the morphology of liquid water removed through the diffusion media [2].

In his macroscopic model, Weber has examined limiting-case scenarios of water and gas transport and established numerically that liquid water saturation is not capable to affect gas phase diffusion in the bulk of standard commercial diffusion media noticeably under normal fuel cell operating conditions. Meanwhile, the model confirms that the MPL dominates the overall response of the porous medium, and interfacial capillary interactions between the MPL and the GDL, depending on both material properties, are more important than bulk GDL properties alone to control water management [3].

Still, beside capillary-dominated liquid water flow, vapor phase diffusion must not be overlooked. The main heat source in the fuel cell core is also located at the CCL where the irreversible oxygen reduction reaction occurs: the temperature is highest at the CCL and decreases towards the gas channels. Temperature gradients therefore do exist between the different porous components leading to evaporation/condensation of water. The exact location of phase change will depend on the operating conditions (temperature, relative humidity of inlet gases, current density) but also on the fuel cell material’s thermal properties. There is a vapor flux from the cathode to the bipolar plate through the MPL and then the GDL along the water concentration gradient.

Modelling work by a Russo-American team has focused on assessing these thermal effects with aim at clarifying the importance of an intermediate porous layer to aid vapor diffusive transport. A water-saturated transport plate as developed by UTC Power was simulated as bipolar plate [4]. It was found that water transport depends on the combination of water vapor diffusion coefficients and thermal conductivities of the GDL components, and for composite GDLs (with MPL) the fraction of water removed in vapor phase is 90% and higher, depending on the MPL thickness and porosity. Assuming saturated conditions at both GDL interfaces (but no evaporation/condensation reactions in the porous layers), the main conclusion was that a composite GDL structure with low thermal conductivity near the cathode-GDL interface helps to sustain under-saturated conditions in the GDL bulk.

GDL cross section micro graph

These model predictions have been confirmed lately. In a recent invited lecture at the CARISMA conference in La Grande Motte (France), September 20th 2010, General Motors has presented the neutron imaging technique used as in situ water management tool to identify the key GDL and flow field parameters that impact liquid water accumulation. The starting point is that air outlet conditions as close as possible to saturation are desirable for maximizing water transport in vapor state – rather than liquid. It was shown that in a fuel cell operating at steady state, the exit air relative humidity levels off to a value that is to critical saturation in the GDL. The important point is that the critical saturation is strongly related to the GDL’s thermal conductivity: higher thermal conductivity will lead to higher critical saturation due to higher temperature gradient between the MEA and the GDL. Therefore, low- thermal conductive GDLs should be preferred in most cases. More specific tuning using for example a high-thermal GDL at the air inlet could also help to decrease MEA temperature and increase water condensation in this –generally water-deficient – region. General Motors and Rochester Institute’s teams have further examined the dominant water discharge mechanism at a variety of normal (dry and wet) operating conditions in additional experimental work [5]. Several sets of experiments with various types of porous fuel cell components and different locations were designed to study vapor- and liquid-driven mechanisms independently (e.g., in Fig. 1 = Fig. 4 in [5]). Results again make it clear that liquid water accumulation against the CCL results in large mass transport losses: an MPL placed against the CL minimizes this phenomenon whereas an MPL within the GDL away from the CL exacerbates it. In the meantime the primary role of gas-phase water transport is confirmed, and GDL’s thermal properties are found preponderant in controlling the thermal and RH gradients driving water from the CCL. The authors suggest that the flux of water vapor could alone be sufficient to remove product water at high current densities even with saturated inlet gases. They propose that the primary role of the cathode MPL is to prevent liquid water condensed in the large pores of the GDL from pooling at the CCL interface rather than improving capillary-driven flow from the nano/micropores of the CCL. According to this hypothesis, water crosses the MPL in a vapor state and the presence of liquid water in the GDL is the result of subsequent condensation.

This is a sort of “paradigm shift” when considering capillary transport-based previous discussions. As often, reality is probably in-between: thermal conductivities of the porous layers (GDL and MPL) have no doubt a great influence on global water management in establishing the thermal gradient that will promote – or hinder – vapor-phase transport out of the fuel cell. Diffusive vapor flux is dominant especially at high temperatures and with low-thermal conductive porous layers. But the MPL is also a capillary transport component that controls liquid water migrating towards gas channels and limits saturation in the GDL. This role is more evident at low temperatures, when the thermal conductivity of the GDL substrate has much smaller influence on vapor transport, since the introduction of an MPL significantly improves performance regardless of substrate’s thermal conductivity (Fig. 2 = Fig. 12 in[5]).

Figure 1: Cross section micrograph of the Ketjen Black EV300J (KB, mean pore diameter 20 nm) on Pureblack SCD 205-110 (SG, mean pore diameter 80 nm) MPL deposited on carbon fiber paper GDL substrate
Polarisation performance with and without MPL at 40 and 80°C
Figure 2: Polarisation performance with and without MPL at 40 and 80°C. (a) MPL coated on Toray carbon paper TGP-H-060 (high thermal conductivity k = 1.4 W/m K) and (b) MPL coated on Mitsubishi Rayon Co. MRC 105 (low thermal conductivity k = 0.3 W/m K)
Finally, neither liquid- or vapor-only water transport might be appropriate, especially when considering actual defects and cracks in the MPL. The network model by Medici et al. [2] has precisely demonstrated that the presence of small defects in an MPL has the combined effect of reducing the saturation level in the GDL while retaining a low injection pressure for the water transport toward the gas channels. This could enable water to cross the MPL in the liquid state and induce capillary-fingering in the GDL with low saturation, as experimentally observed!

Obviously, the water transport mechanism from the cathode catalyst layer is deeply connected to porous layer’s morphological and thermal properties, and there is not a single truth…

[1] J.H. Nam, K.-J. Lee, G.-S. Hwang, C.-J. Kim and M. Kaviany, Int. J. Heat Mass Transfer, 52 (2009), 2779-2791.
[2] E.F. Medici and J.S. Allen, J. Electrochem. Soc., 157 (2010), B1505-B1514.
[3] A.Z. Weber, J. Power Sources, 195 (2010), 5292-5304.
[4] S.F. Burlatsky, V.V. Atrazhev, M. Gummalla, D.A. Condit and F. Liu, J. Power Sources, 190 (2009), 485-492.
[5] J.P. Owejan, J.E. Owejan, W. Gu, T.A. Trabold, T.W. Tighe and M.F. Mathias, J. Electrochem. Soc., 157 (2010), B1456-B1464.

4 – Mechanism of microporous layer for PEMFC explained

Published on September 2010 by Catherine Lepiller, PhD, for Pragma Industries

After years of scientific debate due to contradictory results, the beneficial effect of the microporous layer (MPL), a thin porous layer made of carbon nanoparticles mixed with a hydrophobic agent that is usually inserted between the Pt/C catalyst layer (CL) and the carbon-based gas diffusion layer (GDL) in low temperature fuel cells, appears close to full elucidation now. Both experimental studies [1,2] and fuel cell models [3,4] have recently demonstrated that one main function of the microporous layer is to reduce the number of injection sites for liquid water from the catalyst layer to the gas diffusion layer, which in turn reduces the overall saturation. Less liquid water in the GDL means enhanced gas transport to and from the reaction sites at the CL, and leads to the improved mass transfer noticeable especially under high current conditions when adding an MPL to the fuel cell sandwich.

First, by simultaneous measurement of the capillary pressure and liquid water saturation, Gostick et al. [1] have demonstrated a drastic reduction of the saturation for commercial GDLs in the presence of an MPL, i.e., from ca. 25% to ca. 5%, and concluded that most pores in the GDL are rendered inaccessible to water by the smaller pore sizes of the MPL; hence much lower water saturation is reached.

Second, searchers at the Seoul National University [2] have used a quite original approach called “similarity model experiment”, through which the liquid water transport in hydrophobic GDLs was indirectly investigated: the drainage process was analyzed by replacing the porous layer with transparent hydrogel spheres initially saturated with liquid water, and then slowly injecting a dyed non-miscible fluid having equal density for visual inspection of the internal transport process. By controlling the dimensionless hydraulic parameters in the setup to be close to those existing in actual GDLs, they were able to show that the liquid water drainage process is indeed a capillary-driven process governed by an invasion-percolation mechanism with dendrite-like penetration (also called capillary fingering). Morphological similarities between the non-wetting fluid distribution and predictions by a pore-network model were evidenced as well.

experimental apparatus

These experimental findings are highly consistent with a simulation work performed at the French University of Toulouse [3], which strongly suggests using pore network models rather than continuum models for PEMFCs. In a strict point of view, continuum models are not valid for fuel cell materials because the pore size (~50 µm) and thickness (170-400 µm) of a GDL are separated by less than a decade. Therefore, the condition for applying the continuum approach to a porous media, i.e., having an elementary volume of much smaller size than the porous domain (at least 10 times) for computation, is not met in a GDL. These length scale effects can explain both the inability of continuum models in predicting fuel cell performances accurately and the liquid-vapor phase distribution being in the form of a fractal capillary fingering regime [2,3]. The computational study of the water invasion process as a function of the injection boundary condition has confirmed that liquid water injection through discrete points is much more favourable to gas transport than injection through the total surface since the resulting saturation will be lower everywhere along the GDL thickness. As observed experimentally, such condition of one-point injection is achieved in the presence of an MPL [1]. Surface injection is instead related to a fuel cell configuration without MPL leading to the formation of large droplets and high water saturation level at the CL surface due to a large gap size [4].

Finally, the Korean team in [2] has further clarified the previous points and the exact role of the MPL [4]. They have developed a model for morphology control of liquid phase across multiple porous layers and proposed two roles of MPL in this aspect, in strong agreement with other selected papers [1,2,3]: the MPL inserted between the CL and the GDL reduces the number and the size of interfacial droplets formed on CL surface. Both effects reduce the saturation level and the overall liquid breakthrough toward the GDL. Smaller pores in MPL (vs. GDL) at the catalyst interface prevent water droplets from growing very large and clogging the CL surface. This mechanism seems to be the way the MPL effectively alters the liquid water distribution in the fuel cell for better management!

[1] J.T. Gostick, M.A. Ioannidis, M.W. Fowler, M.D. Pritzker, Electrochem. Commun., 11 (2009), 576.
[2] J.H. Kang, K.-J. Lee, J.H. Nam, C.-J. Kim, H.S. Park, S. Lee and I. Kwang, J. Power Sources, 195 (2010), 2608-2612.
[3] M. Rebai and M. Prat, J. Power Sources, 192 (2009), 534.
[4] J.H. Nam, K.-J. Lee, G.-S. Hwang, C.-J. Kim and M. Kaviany, Int. J. Heat Mass Transfer, 52 (2009), 2779-2791.

Pictures courtesy of the Seoul National University.

More information, contact

5 – Carbon foam for mitigating water management issues in low temperature fuel cells

Published on May 2010 by Catherine Lepiller, PhD, for Pragma Industries

In low temperature fuel cells, two-phase (liquid water and air) flow mal-distribution in the gas channels is further complicated by the generation of water at discrete spots and at different rates on the cathode catalyst layer. Uneven local current density may therefore deteriorate the flow distribution in channels, which in turn alters the power output.

Two papers recently published in the Journal of Power Sources stand out by their innovative approach for addressing this issue. Both propose the use of carbon foam material either as or in the bipolar plate of a PEM fuel cell. Their preliminary results are quite interesting.

Carbon foam Carbon foam Carbon foam

The first approach by J. Chen at the University of California (J. Chen, J. Power Sources, 195 (2010), 1122-1129) has been to fill the parallel channels of a conventional graphite end plate with porous carbon foam and use visualization images to compare the two-phase flow under different regimes representative of a real fuel cell. After verifying that the porous media inserts in the channel did not modify the homogeneity for two-phase flow, it was clearly observed that the mal-distribution was much less severe in the device with porous channels vs. standard hollow channels. This was attributed to the random arrangement of water droplets in numerous tiny pores (~ 100 µm) of the carbon foam network. Despite overall low porosity (3%) the carbon foam inserts featured self-adjustment capacity to the amount of water in flow channels: under steady-state operation, over the whole volume of porous media, there would be a number of pores retaining liquid water whatever the air flow velocity (down to ~ 10% at high values), but the rest would remain free anytime for air to pass. Although the pressure drop with this configuration is 4-fold higher than in hollow parallel channels, the benefits are more significant, and particularly the mitigation of flooding/drying phenomena thanks to the water self-adjustment allowed by porous carbon foam inserts.

The Canadian team at the Queen’s Royal Military College Fuel Cell Research Centre (J. Kim and N. Cunningham, J. Power Sources, 195 (2010), 2291-2300) has also selected carbon foam for its feasibility study. Unlike in the previous work, such porous material was considered as whole cathodic bipolar plate, and not merely as insert in conventional channels. They used Reticulated Vitreous Carbon foam with different pore sizes, and compared the single cell performance of their new design (3.5 mm thick carbon foam at the cathode and serpentine flow-field at the anode) with that of a conventional fuel cell with serpentine channels on both sides. Consistently with J. Chen, they found that the RVC foam fuel cell offers advantages over the conventional design especially at low operating temperatures and fully humidified gases. Moreover, the parametric study of geometrical parameters of the carbon foam (thickness, average pore size) under various operating conditions has confirmed that those with small pore sizes in the range of 100 µm are best suited for water management because they provide higher electrical conductivity to the material. Lower gas permeability also enhances forced convection in addition to diffusion as well, thereby increasing air penetration and residence time in the gas diffusion layer for a more efficient electrochemical reaction. Fuel cell tests show comparable performances in terms of power density and long-term (250 h) stability.These two complementary studies agree to point to a new possible design of fuel cell bipolar plates including carbon foam, at least at the cathode side: both low porosity and isotropic characteristics of this material favor 3-D mass transfer and extend the effective GDL area accessible by the reactant flow through the porous medium to the catalyst layer. Under flooding conditions, the two-phase flow is made more uniform in a sustainable way. Maybe a first step towards good water management at low temperatures?

6 – A metal oxide alternative to carbon as catalyst support in low-temperature fuel cells

Published on April 2010 by Catherine Lepiller, PhD, for Pragma Industries

Following the general trend observed in the low-temperature fuel cell research to replace Pt/C catalysts by less costly and more durable compounds, as already exemplified in Pragma’s September Science Note, promising new results point to titanium dioxide.

Current polymer electrolyte fuel cells use platinum and platinum-based alloys supported on nanoporous carbon as electrodes. However, during the duty cycles of repeated start-ups and shut-downs, the fuel cell undergoes high potentials that lead to carbon and Pt degradation processes. In order to maximize catalyst utilization in the electrodes, Pt nanoparticles have been downsized to 2-3 nm. Thermodynamic size effects make them less stable than bulk Pt, which causes the dissolution/sintering into bigger agglomerates in order to restore stability. Meanwhile, the agglomeration process is accelerated by carbon corrosion in oxidative conditions: As a consequence, Pt particles are detached from their support and tend to gather together.

Unfortunately, degradation is fairly rapid under typical fuel cell conditions. Fuel cell performance and lifetime are greatly affected by these concomitant phenomena. Computational modeling shows that degradation can be significantly mitigated by increasing the particle size from 2 to 4-5 nm. Nevertheless, platinum metal is expensive and not very abundant. If current loadings are maintained in fuel cells, catalyst cost and supply will be an overwhelming obstacle for their widespread commercialization. There is an urgent need for both more robust non-platinum catalysts and non-carbon catalyst supports such as metal oxides.

Several studies have recently converged to promote titanium dioxide as alternative catalyst support in a fuel cell. Titanium dioxide is an already widely used semiconductor material with further potential applications in solar cells, biotechnology, photocatalysis, and gas sensors. Despite a high mechanical resistance and stability in acidic and oxidative environments that have raised interest for TiO2 in fuel cells for a while now, its low electronic conductivity has prevented so far any application. But a first milestone seems to have been reached lately, as shown below. This should at least address the stability issue with carbon-supported Pt catalysts. At the Universities of Erlangen-Nürnberg (Germany) and Turku (Finnland), researchers have successfully imparted semimetal conductivity to TiO2 nanotubes through carbonization in acetylene gas atmosphere at 850°C. While carbonization forms a new carbon-containing titanium oxy-carbide compound, the nanotube structure is hardly altered. The compound has been identified as a solid solution between TiCx and TiOx rather than C-doped TiO2.

It exhibits high electronic conductivity similar to metals and a much superior mechanical hardness thanks to its titanium carbide content. Together with very good electrochemical properties, these new conductive titanium oxy-carbide nanotubes show great promise especially for DMFC applications: when introduced as support for Pt and Pt-Ru anode catalysts they are claimed to increase the activity for methanol oxidation by 700%. [P. Schmuki et al., Angewandte Chemie International Edition, 48 (2009), 7236-7239].

At the University of South Carolina, the structure of titanium dioxide and the “wet” synthesis method was quite different from above, but results proved promising as well. In this study, mesoporous TiO2 was prepared via a template-assisted route where the porosity was controlled by the hydrolysis reaction of the dissolved titanium precursor in presence of a surfactant. Colloidal Pt particles were synthesized separately, and the mixture Pt(coll)+TiO2 was finally carried out in presence of a reducing agent.

The novel Pt/TiO2 catalyst was tested as cathode in a PEMFC: it showed performances comparable to or even better than Pt/C at the same loading (0.4 mg/cm²), which was attributed to improved mass transport in the thinner cathode layer. In addition, the accelerated test protocol showed an extremely high stability toward oxidation conditions: zero decrease in performance was observed after a corrosion time of 200 h. Since the Pt particle size was initially larger when supported on TiO2 than on C and the corrosion resistance of titanium dioxide is better than carbon, the decay of the active catalyst surface is strongly reduced. A strong metal support interaction between the Pt particles and the TiO2 support is also assumed to inhibit the catalyst migration and agglomeration, therefore enhancing the overall stability of the cathode. Based on these results, mesoporous TiO2 should be considered as an alternative support for Pt in fuel cells [S.-Y. Huang et al., Journal of the American Chemical Society, 131 (2009), 13898-13899].

platinum powder

7 – Iron to replace platinum in PEM fuel cells anytime soon?

Published on April 2010 by Catherine Lepiller, PhD, for Pragma Industries

It is a real pleasure to introduce this scientific and technical capsule by highlighting the top-end results achieved recently at the Institut National de la Recherche Scientifique – Énergie, Matériaux et Télécommunications, and published earlier this year in Science under the title “Iron-based catalysts with improved oxygen reduction activity in polymer electrolyte fuel cells” (M. Lefèvre, E. Proietti, F. Jaouen and J.-P. Dodelet, Science 324 (2009), 71). The international research team, led by French Professor Jean-Pol Dodelet, has been involved in the development of non-platinum, iron-based catalysts for PEM fuel cells at INRS-EMT near Montreal, Canada, for many years. Further evidence is given today that consistent efforts and talent are paying back.

Michel Lefèvre, Frédéric Jaouen and their colleagues have successfully synthesized iron-nitrogen catalysts exhibiting an activity for the cathodic oxygen reduction reaction (ORR) that has proved quite comparable with that of commercial platinum supported on carbon black. As in the case of platinum and platinum alloys, the compound’s microstructure is of utmost importance regarding the final activity: therefore, work has focused toward both a fundamental understanding of the relation between microstructure and activity, and the optimization of the catalyst’s preparation based on this knowledge. The overall rate of ORR reaction, i.e. volumetric activity of the Fe-N compound was eventually improved more than 35 times compared with the previous best non-precious metal catalyst (and within 10% of the best Pt-based catalysts to date), by increasing the number of active sites for ORR per unit volume thanks to proper choice of the reactants, reaction method and subsequent thermal conditions. Results are very close to the Department of Energy’s 2010 technical target for PEM fuel cells’ non-precious metal catalyst activity.

platinum powder

Prior to this work, the team had achieved promising ORR activities with iron-based catalysts synthesized by impregnation of microporous carbon black with a soluble Fe precursor (iron acetate, for example) and nitrogen-containing complexing agents like phenanthroline. Hat treatment in ammonia vs. neutral atmosphere was also investigated. It was hypothesized that the coordinating bridges between Fe and the C-N-C bounds formed in the carbon black micropores during heat treatment in NH3 were responsible for the catalytic activity. Results would also show that only micropores created during the heat treatment in ammonia might host active sites, and not pristine micropores present in carbon black material. In the current work, researchers at INRS have filled the pores with a Fe-N containing material (including the acetate Fe(II) precursor and a suitable N-bearing complexing agent) by using planetary ball-milling instead of impregnation. Due to lower thermodynamic limitations, ball-milling allows filling more pores than does close-to-equilibrium impregnation while hardly affecting carbon’s microstructure. The modified carbon black support was then pyrolised in argon and/or ammonia atmosphere. The most active catalyst contained 1 wt% of Fe content. At 0.9 V, the current density of a cathode made with a catalyst loading of 5.3 mg/cm², i.e., 90 µg/cm² of Fe, is 30-40 mA/cm², which is equivalent to a commercial Pt-based cathode with a Pt loading of 400 µg/cm². Losses at current densities > 100 mA/cm² arise from excessive mass transport limitations due to electrode thickness at high Fe loadings.

Being able to replace platinum by iron at the cathode of a PEM fuel cell means that, since most of the platinum loading is currently required for the sluggish ORR and iron is the cheapest among metals the related cost of the catalyst could drop in a terrific way. Maybe a major technical breakthrough in the world of fuel cells! I hope to hear follow-on news from this high-profile research field very soon… Durability tests and further optimization of the density of Fe active sites in the carbon support are still necessary.

The complete set of results is available in the Science paper Science 324 (2009), 71.