Abstract
The performance of a low temperature fuel cell is strongly correlated with parameters like the platinum particle size, platinum dispersion on the carbon support, and electronic and protonic conductivity in the catalyst layer as well as its porosity. These parameters can be controlled by a rational choice of the appropriate catalyst synthesis and carbon support. Only recently, particular attention has been given to the support morphology, as it plays an important role for the formation of the electrode structure. Due to their significantly different structure, mesoporous carbon microbeads (MCMBs) and multiwalled carbon nanotubes (MWCNTs) were used as supports and compared. Pt nanoparticles were decorated on these supports using the polyol method. Their size was varied by different heating times during the synthesis, and XRD, TEM, SEM, CV, and single cell tests used in their detailed characterization. A membrane-electrode assembly prepared with the MCMB did not show any activity in the fuel cell test, although the catalyst's electrochemical activity was almost similar to the MWCNT. This is assumed to be due to the very dense electrode structure formed by this support material, which does not allow for sufficient mass transport.
1. Introduction
One of the important issues in the development of highly efficient polymer exchange membrane fuel cells (PEMFCs) is the development of stable electrocatalysts for electrodes.
Supported catalysts have several advantages over unsupported catalysts: they have a higher stability than unsupported catalysts in terms of agglomeration under fuel cell operating conditions [1]; the porosity of the carbon black support allows for gas diffusion to the active sites; the good electronic conductivity of the carbon support provides electron transfer from catalytic sites to the conductive carbon electrodes and then to the external circuit; the small dimensions of nanoscaled catalyst particles dispersed on a carbon support maximizing the contact area between catalyst and reagents [2].
Electrical current is generated at Pt nanoparticles, which are randomly dispersed on a high-surface carbon matrix. From the catalyst powder an electrode has to be fabricated, which offers a porous structure and good transport properties for the transport of protons and electrons generated during the reaction. Standard processes for the fabrication of so-called membrane electrode assemblies (MEAs) are for instance airbrushing, DECAL, or sieve printing. During electrode fabrication, an intimate mixture of carbon/Pt and ionomer, the so-called ink, self-organizes into a phase-segregated composite with interpenetrating percolating phases for the transport of electrons, protons, and gases [3, 4]. As explored in [4], the process of microstructure formation depends on the type of the supported catalyst, the type and amount of ionomer added, the type of dispersion medium used during ink preparation, and the fabrication conditions.
Electrochemical reactions occur only at those Pt particles where the three phases meet. This region is called the three-phase boundary (TPB). In order to describe the diffusion processes taking place at the reactant-electrolyte-catalyst interface, several models have been developed [5]. The three-phase-boundary model assumes that the electrochemical reaction takes place only in a narrow region at the interface formed between the electrode, the electrolyte, and the reactant gas. This approach is based on the consideration that the solubilities and diffusivities of hydrogen and oxygen in aqueous electrolytes are low. Major constraints of this design are the statistical limitation of the Pt utilization due to the random three-phase morphology and the highly nonuniform reaction rate distributions. These occur when the thickness of the layer is large compared to the so-called “reaction penetration depth,” which is determined by the interplay of transport and reaction.
Another issue besides the TPB is the porosity. The micropores in the catalyst layer are necessary for gas transport to the catalytic sites. Fischer et al. [6] reported an original porosity of the catalyst layer after fabrication of 35%. But when pore formers were added, the porosity could become as high as 65% depending on the pore formers used. Unfortunately, cell performance is not proportional to catalyst layer porosity. In order to achieve maximum fuel cell performance, the catalyst layer should have an optimal porosity [6–8]: with higher catalyst layer porosity, the mass transfer rate increases, while the electron and proton transport rates decrease. Moreover, it has been observed that support corrosion can lead to a collapse of the porous electrode structure, which is even more severe in terms of fuel cell performance than the Pt dissolution [9]. Severe mass transport limitations will be the consequence. However, with the possibility to use self-assembly mechanisms, controlled support morphologies, and aligned growth of nanofibre electrodes, completely new opportunities will open up for the design of functional electrode [10].
Platinum supported on a porous carbon support (Pt/C), for example, Vulcan or Ketjen black, is the state-of-the-art electrocatalyst for the anodic and cathodic reactions in PEM fuel cells. The type of carbon material utilized in the preparation of the catalyst critically affects the performance and durability of the cell. In particular, graphitized carbon and MWCNT have been reported to be more stable in fuel cell operation conditions. In this work, two distinctively different carbon support materials have been applied, which in contrast to the standard Vulcan support exhibit a severely decreased inherent porosity: multi-walled carbon nanotubes (MWCNTs) and mesocarbon microbeads (MCMBs). Moreover, while the mesoporous carbon microbeads appear as large-sized grains, but still comparable to Vulcan, the multi walled carbon nanotubes (MWCNTs) have a 1D fibrous morphology. The different morphologies are expected to lead to the formation of different electrode architectures with different thickness, three-phase boundary, and porosity, thereby affecting mass transport and the final performance of the fuel cell.
In addition to the support material also the nanoparticle size was varied in this study. An effective way for the controlled synthesis of nanosized Pt colloidal particles in solution is the use of ethylene glycol, as reported recently [11, 12]. Using nanostructured carbon supports with higher surface areas increases the number of nucleation sites for platinum deposition and leads to catalysts with smaller particles. The size of the active phase was adjusted by different heating times yielding particle sizes of 5.1 nm and 2.8 nm for the Pt/MWCNT and 9.3 nm and 4.4 nm for the Pt/MCMB, respectively.
2. Experimental
For the preparation of Pt-loaded catalysts with different particle size, two different supports were used, which show significant differences in morphology and porosity between each other and compared to the standard carbon supports, for example, Vulcan XC-72. The decreased internal porosity in MWCNT and MCMB should help to reduce the Pt loading by increasing the catalyst utilization and to improve the catalyst activity. Being able to synthesize fuel cell catalysts in the 1 to 5 nm range, that is, the size range of interest in practical fuel cell catalysis, in a controlled manner, is very important. In the case of the reduction of O2 (ORR), Pt particles in the 3.5 to 4 nm size range are believed to be the most active catalysts [13, 14]. The synthesis of high surface area catalysts in ethylene glycol, referred to as the polyol method, has been widely applied to this purpose, as it allows for a good control of particle size and dispersion and does not require inert gas atmosphere [15].
2.1. Chemicals
Multi-walled carbon nanotubes (purity 70–90%) were purchased from Thomas Swan Co, England. Mesoporous carbon microbeads (MCMBs) were obtained from Professor X.-P. Qiu’s group in Beijing, China, whereas hexachloroplatinic acid (H2PtCl6) was from Aldrich and ethylene glycol (EG) from VWR. As a reference and for the counter electrode in the membrane-electrode assembly (MEA), Pt/C (platinum, nominally 20% on carbon Black (CB) Hispec 3000) was obtained from Alfa Aesar. Nafion 117 membranes made by Dupont were purchased from Ion Power, USA. Nafion ionomer resin, 5 wt.% solution was obtained from Aldrich.
2.2. Catalyst Synthesis
Before deposition, the catalyst supports have to be functionalized by concentrated acids to obtain active centers at the outer surface of the MWCNTs and MCMBs. The oxidative pretreatment was done in an equal ratio mixture of HNO3 (69%) and H2SO4 (95%) under sonication. The treatment generates functional groups, such as OH, COOH, CO on the surfaces, which are believed to act as metal-anchoring sites and to facilitate metal nuclei formation [16]. After oxidation, the MWCNT or MCMB acid mixture was diluted with H2O and filtered, flushed 3 times with H2O and ethanol. Hexachloroplatinic acid (H2PtCl6 6 H2O) was used as platinum precursor and ethylene glycol (EG) as solvent and soft reducing agent simultaneously. To deposit particles with larger particle sizes, a mixture of 100 ml EG and 50 ml deionized water (DI) was used as solvent in the reaction. For smaller particle sizes, the solvent was just EG. The oxidized MWCNT and MCMB were dispersed in the EG solution, and the Pt salt solved in EG was added dropwise under stirring. The mixture was heated to 160°C with different heating times applied. For the deposition of smaller particles, the solution was heated for 3 hours; for larger particles it was heated for 16 hours in a three-necked round bottom flask under reflux and magnetic agitation. The products were filtrated with ethanol and kept in wet condition for the Pt/MWCNT. Pt/MCMB was dried at 80°C in a drying furnace. Table 1 shows details of the catalyst synthesis.
2.3. Catalyst and MEA Characterization
Transmission electron microscopy (TEM) investigations were carried out with a Jeol JEM-3010 transmission electron microscope equipped with LaB6 cathode operating at 300 kV acceleration voltage. Powder samples were prepared by dispersing the catalyst powders in methanol and depositing a drop of the dispersion onto a standard holey carbon film-covered copper grid. Average particle sizes and particle size distributions were obtained for all samples using the semiautomatic software LINCE by Lucato [17]. At least 200 particles of each image were counted to obtain particle size distributions. Using the respective histograms, we calculated the linear diameter and the volume-surface diameter [18]: where is the diameter of Pt particles, is the corresponding number of particles. Then, the Pt particle dispersion on the surface was calculated using the following equation:where is the area per Pt atom on the surface, ; is the Pt atomic volume, , and.
The specific Pt surface area for a sample, (m2/g Pt), can be approximated using the following equation: where g/cm3.
X-ray powder diffraction (XRD) measurements were performed with a STOE STADI-P diffractometer with germanium monochromized Cu radiation and a position-sensitive detector with a 40° aperture. The measured diffraction patterns were refined by the Rietveld method using the FULLPROF software suite [19] to determine the average size of the deposited platinum particles. The average particle size was then extracted from the fitting parameter according to the following equation: with the wavelength of the X-ray beam and the Lorenzian isotropic size parameter.
Scanning electron microscopy (SEM) investigations were carried out with an FEI Quanta 200 FEG scanning electron microscope equipped with a field emission gun and operating at 20 kV acceleration voltage. Sample preparation for the SEM measurements was carried out by cutting slices off the membrane electrode assemblies in order to obtain cross sections.
Electrochemical catalyst characterization was carried out by cyclic voltammetry (CV) at room temperature in a three-electrode cell. 0.1 M HClO4 was used as electrolyte for all measurements. The measurements were in the range against the Ag/AgCl electrode, which was separated from the working electrode compartment by an electrolyte bridge in the style of a Haber-Luggin capillary to avoid chloride contamination [20]. For the measurements a scan rate of 30 mV/s and a step size of 2 mV were chosen. The potentiostat used was a Reference 600 from Gamry Instruments. Only the data obtained after five cycles was taken into account for the analysis. When 1 M methanol solution is added, the hydrogen adsorption region decreases due to CO poisoning of catalytically active sites [21]. For the ink preparation 5 mg of catalyst was dispersed in 1 ml MilliQ water and 167 Nafion solution (20% in aliphatic alcohols, Aldrich) by ultrasonic treatment for 60 seconds. For each sample 5 of the dispersion was readily applied to a glassy carbon disk electrode (diameter 3 mm) and dried in air at 80°C for approximately 10 minutes.
2.4. Fuel Cell Tests
The preparation of the membrane-electrode assembly (MEA) was carried out using a Nafion 117 membrane from Dupont. For both electrodes, an ink solution was prepared using a method slightly modified from the one reported by Wilson and Gottesfeld. [22]. MEAs with an active electrode area of 25 cm2 were fabricated by airbrushing the catalyst ink onto one side of the Nafion membrane, heated to and kept at 120°C. For the cathode, 100 mg of the catalyst were dispersed in 0.5 ml ultrapure water, isopropanol, and 1 ml Nafion 5% solution under sonication. The dispersion was stirred with a high shear mixer at 7000 rpm. For the anode, 200 mg of Pt on carbon (20 wt% Pt, Alfa Aesar HISPEC 3000) were dispersed in 4 ml of H2O and 2 ml isopropanol. 1.2 ml of Nafion 5% solution was added, and the solution was dispersed by sonication and stirred with a high shear mixer. The inks were filled into an airbrush pistol (Evolution by Harder & Steenbeck) and sprayed successively onto the heated membrane surface, allowing each layer to dry for 10 seconds. Fuel cell tests were performed in a home-made single cell test bench with a Quick Connect fixture from Baltic Fuel Cells GmbH in hydrogen/oxygen operation. The flow rates of both gases were adjusted to H2/O2 55/25 and 83/38, respectively, and the cell temperatures varied between 25°C, 50°C, and 65°C. Hydrogen was loaded with water in a humidifier (25°C) and fed into the anode.
3. Result and Discussion
3.1. Characterization of Catalyst by XRD
As described in Table 1, four different catalysts were synthesized and characterized using X-ray diffraction. The respective powder patterns are shown in Figure 1 in comparison. Only the face-centered cubic (fcc) phase of metallic platinum is visible in the patterns with no indications of a crystalline Pt oxide phase. The difference in the used supports shows at low diffraction angles; while the MWCNTs show only a small and rather broad (002) reflection from the graphite in good agreement with the literature [23], the pattern of the MCMB support exhibits strong and rather sharp reflections of a graphitic phase not only at low, but also at higher diffraction angles. These also had to be implemented in the Rietveld refinement. However, the crystallites of the support appear to be slightly distorted, as the peak intensities do not match the database pattern for graphite perfectly. As intended, the synthesis with the longer heating times led to larger Pt nanoparticles with approximately twice the size of the catalysts heated for only 3 hours. Moreover, the MCMB-supported catalyst nanoparticles were larger than the ones decorated on MWCNT pointing towards fewer anchoring sites on these materials. This is not unexpected, as it is well known that the more graphitic a support, the less nucleation sites are present at its surface [24]. The average particle sizes determined by XRD can be found in Table 2.
3.2. Characterization of Catalyst by TEM
Figure 2 shows representative TEM images of the different catalysts at various magnifications and Figure 3 the corresponding size histograms. The Pt/MWCNT catalysts show spherical Pt nanoparticles with a relatively narrow size distribution uniformly distributed in high dispersion on the carbon nanotubes. Agglomeration of the particles appears to be scarce and seems to occur only in the nooks of the nanotubes (also seen in the histogram). In high resolution images, parallel lattice planes become visible underlining the metallic nature of the particles. For the MCMB catalyst (samples B, D), a lesser amount of decoration with metallic nanoparticles is observed than in the case of the MWCNT. Moreover, the particle distribution is not homogeneous for all support grains, and, in contrast to the MWCNT, the platinum particles tend to agglomerate (also apparent from Figure 3). At higher magnification, the support material has a layered appearance with a step width of the visible layers of about 10 nm, which has not been found for the other standard carbon supports.
(a)
(b)
(c)
(d)
(a)
(b)
(c)
(d)
3.3. Cyclic Voltammetry (CV)
Cyclic voltammetry was measured to obtain a first indication of the catalysts’ electrocatalytic activity for the methanol oxidation reaction (MOR). Pt nanoparticles supported either on multi walled carbon nanotubes or on mesoporous carbon microbeads are both active showing an oxidation activity with peak potentials for the first oxidation peak at around 0.61 V versus Ag/AgCl (ca. 0.84 V versus RHE). The onset of the oxidation activity at approximately 0.25 V versus Ag/AgCl and the maximum current density of about 5 mA/cm2 were also very similar. A comparison of the base voltammograms and the respective oxidation activity of 2 samples (C and D) is shown in Figure 4. All results are summarized in Table 3. The region is more pronounced for the Pt/MWCNT than for the MCMB-supported sample. When 1 M methanol solution is added, the hydrogen adsorption region decreases due to CO poisoning of catalytically active sites [5].
(a)
(b)
This is more pronounced for Pt/MWCNT than for Pt/MCMB. The tolerance of the catalysts towards carbonaceous intermediates can be estimated by the ratio of the forward and backward current intensities and is also listed in Table 3. For all four samples the ratio is higher than one, which is an indication for the excellent oxidation of methanol [25]. Both Pt/MWCNT samples show higher ratios than the MCMB-supported samples, and also less suppression of the region is observed. This could be due to a support effect but for its confirmation additional analysis has to be done. It is, however, difficult to discuss a particle size effect for the methanol oxidation from the CVs, since all samples showed very similar onset and peak potentials shift despite the different particle sizes (Table 3).
3.4. MEA Characterization by SEM
Membrane-electrode assemblies (MEAs) were prepared for all four catalyst samples as explained in the experimental section. Figure 5 shows typical SEM micrographs of both a top view of the electrode surface and the respective cross-sections. The surface of the Pt/MWCNT MEA appears rather rough with a nonuniform electrode thickness. In the top view, single strands of carbon nanotubes are already visible at comparatively low magnification in the SEM giving the surface a “fluffy” appearance. In contrast, for the Pt/MCMB MEA a smooth surface with uniform electrode thickness is observed. A globular structure of the support aggregates is preserved despite the spraying process and is clearly visible in the cross-section. The top view in higher magnification reveals a very dense, furrowed structure with smeared features. These micrographs were taken after the fuel cell test has been done. Hence, the structure of the surfaces might also be dependent on the force used to mount the cell or the carefulness with which the MEAs were handled.
(a)
(b)
(c)
(d)
3.5. Fuel Cell Tests
Single cell fuel cell tests were performed with all four MEAs at different temperatures and with different gas flows (Table 4). Unfortunately, the MEAs of Pt/MCMB did not show any performance, and no current could be drawn during the fuel cell test. It is assumed that the very dense structure of the electrodes effectively hampers sufficient mass transport. Either the gases cannot penetrate the electrode due to path-blocking by ionomer layers (note that the carbon support is less porous than conventional materials and the ink preparation not adapted to the novel material) caused by the nonoptimized electrode preparation, or the three-phase boundary is only weakly developed, so that not many particles can participate in the electrocatalytic reaction. In contrast, the MEAs fabricated from the Pt/MWCNT performed well, and the polarization curves and power densities of samples A and C in different operation conditions can be found in Figure 6.
(a)
(b)
From the figure, it can be seen that both the particle size of the active phase and the operation temperature have a significant effect on the polarization curves. The curves of the catalyst with the larger particles at high temperature are almost identical to the curve of the catalyst with the small particle size at low working temperature. The best results were obtained for the sample with an average particle size of 2.8 nm at the highest operating temperature of 65°C. The gas flows appear to be less important and do not affect the results significantly.
4. Conclusion
In this work, two different carbon materials—mesoporous carbon microbeads (MCMBs) and multi walled carbon nanotubes (MWCNTs)—were studied and compared for their applicability as alternative support materials for electrocatalysts. These supports show structures and morphologies significantly different from the Vulcan carbon routinely applied. Adequate control of the particle size and distribution of Pt particles for both Pt/MWCNT and Pt/MCMB was obtained by the polyol method. Longer heating times (16 hours versus 3 hours) led to larger particle sizes, which were detrimental to the fuel cell performance of the Pt/MWCNT.
It was demonstrated that the carbon support materials affect not only the catalyst characteristics such as platinum particle size and dispersion, but also the electrode layer characteristics such as water management, gas transport, and electronic and ionic conductivities. MEAs of Pt/MCMB did not show any performance at all in the fuel cell test. This might be explained by the very dense structure of their electrodes, which did not allow for a good transport and an optimal water balance. However, by adapting the ink preparation parameters, for example, the ionomer-to-catalyst ratio, to the new support materials, an increase in the fuel cell performance might be feasible. Moreover, the particle density on the MCMB supports is comparatively low. A higher number of anchoring sites for Pt deposition might be obtained by a selective treatment of this support, also leading to higher activities.
Acknowledgment
G. Sevjidsuren gratefully acknowledges Professor H. Fuess for enabling her research to stay at the Institute for Materials Science at TU Darmstadt, Germany.