Characterization of PEM Fuel Cell Electrocatalysts

Introduction

As the energy sector and transport industry seek to mitigate their environmental impact, green hydrogen will likely play a major role in achieving carbon neutrality. Green hydrogen is produced by the electrolysis of water using renewable energy sources, such as solar or wind power. Electrolyzers based on polymer electrolyte membrane (PEM) technology provide an efficient pathway for producing green hydrogen through water electrolysis. By inverting the process, PEM fuel cells (PEMFCs) can be used to generate electricity using hydrogen as fuel, in combination with oxygen (which acts as an oxidizing agent). Fuel cells have the potential to make hydrogen a viable replacement for fossil fuels in energy, transport, and other industrial sectors. However, to achieve this on a large scale, their performance and cost must be optimized.

The catalytic material used to produce PEMFC electrodes is the main component in determining both the performance and the cost of fuel cells. The catalytic ink is typically composed of a mixture of a catalytic active material, an ionomer, and a dispersion solvent. Within the ink, the catalyst – usually platinum-on-carbon (Pt/C) – is a composite of platinum (Pt)-metal group nanoparticles deposited on activated carbon, which serves as a support matrix. The catalyst’s activity and stability are the two key parameters that determine the PEMFC’s ultimate performance. Catalytic activity, in turn, is governed by the size, dispersion, and morphology of the Pt-metal group nanoparticles. Equally important are the structural, textural, and surface chemistry properties of the carbonaceous agglomerates during ink drying, upon deposition on the proton exchange membrane. The optimized pore structure of the C-support matrix can significantly reduce the amount of Pt needed, and the optimization of its distribution and maximization of the availability of the catalytic points for the oxygen reduction reaction (ORR) and hydrogen-oxidation reaction (HOR) in fuel cells reduces the overall cost of the process.

Catalytic activity and stability can be optimized in several ways: by developing novel Pt-alloy cathode materials, controlling the particle size for maximum mass activity, controlling the inter-crystallite distance, or ensuring uniform dispersion of Pt nanoparticles on the carbonaceous support.¹

Materials and Methods 

In this study, several techniques were used to analyze a set of catalyst powders made of Vulcan XC-72 carbon black, with three different Pt loadings: 60% (60 m2/g), 40% (70 m2/g), and 20% (100 m2/g). These catalytic powders were bought from Premetek, with the following specifications:

	Carbon	Surface area, total (m2/g)	Metal area, Pt (m2/g)	XRD XS size, Pt
40%Pt on carbon	Vulcan XC 72	140	70	3-4 nm
40%Pt on high-surface-area EC-300J	High-surface area EC-300J	120	-	<2 nm
40%Pt on Vulcan XC 72R	Vulcan XC 72R	-	70	3-4nm
20%Pt on Carbon	Vulcan XC 72	180	100	2-3 nm
60%Pt on Carbon	Vulcan XC 72	90	60	4-5 nm
40%Pt-Ru on Vulcan	Vulcan XC 72	-	70	3–4 nm
40%Pt-Co on Vulcan	Vulcan XC 72	-	70	3–4 nm

X-ray diffraction (XRD), X-ray fluorescence (XRF) spectroscopy, and laser diffraction were highly effective in analyzing the catalytic powders. Although highly useful in analyzing other aspects of catalytic inks, techniques such as morphological imaging and dynamic light scattering had limited application to the study of catalytic powders materials. Some Pt-alloy catalysts were also analyzed to see if the same techniques could also be applied to these materials.

Characterizing Pt/C catalysts can be difficult because its constituents span different size ranges:²

• 2–5 nm nanoparticles (Pt metal group catalyst)
• 20–40 nm primary particles (C support)
• 100–300 nm aggregates (C-support particles)
• 1,000–10,000 nm (i.e., 1-10 µm) agglomerates (Pt/C particles)

Furthermore, because the Pt particles are deposited on the C-support particles, they are difficult to size separately using large sampling statistic methods such as laser scattering. 

Characterization of “Pt on Carbon” Catalyst Powders

X-ray Diffraction Analysis

X-ray diffraction of Pt/C catalysts primarily provides information on the Pt-metal group nanocatalyst. Data can be collected using either a benchtop or a floor-standing diffractometer, configured with the conventional Bragg-Brentano powder diffraction geometry.

Figure 1 compares the diffraction patterns of the three samples with different Pt loadings on Vulcan XC-72 carbon black. These data points were collected using a 600W Aeris compact diffractometer, equipped with a Cu-anode X-ray tube, 1/4° divergence slit, and the PIXcel3D hybrid pixel detector. Small amounts of powder were prepared on zero background holders, using the dusting method. The X-ray diffraction patterns provide information about the size of coherently scattering domains (derived from the diffraction peak widths) and lattice parameters (derived from the diffraction peak positions).

Three different techniques in the HighScore Plus software package were used to calculate the average coherently scattering domain (known as “crystallite”) size of the Pt nanoparticles: Scherrer, Williamson-Hall, and whole pattern-fitting (in this case, using the Pawley method). The Scherrer method estimates crystallite size using the width of a single diffraction peak – in this case, the (111) peak at ~39.8° 2θ. The Williamson-Hall method estimates the volume average crystallite size by evaluating how the width of multiple peaks varies in relation to the Bragg angle. This method not only separates the contributions of crystallite size and microstrain to the peak width, but it also provides a better estimate of the precision by using linear regression statistics. Finally, the whole pattern-fitting methods model the entire diffraction pattern, including peak width and unit cell lattice parameters. 

The results of all three methods are shown in Table 1. All three produced consistent results, showing that the average crystallite size continuously increases with Pt loading, within a range of 2–5 nm. This comparison indicates that any of the three methods could be used to evaluate the coherent domain size of the Pt nanocatalyst in these samples. Each has its advantages and disadvantages, namely:

• The Scherrer method is quick and easy because it only requires the measurement of a single diffraction peak. In fact, with modern position-sensitive detectors, which observe >3° 2θ simultaneously, data for Scherrer analysis could be collected with a stationary detector for very fast monitoring in a production environment. However, the Scherrer method is not as precise as other methods that use multiple peaks.
• The Pawley whole pattern-fitting method has the advantage of determining lattice parameters and K-factor in addition to the crystallite size. As shown in Table II, there was a slight variation in the lattice parameters of Pt in the three different catalytic powders. This is consistent with previous findings showing that lattice parameters may vary with crystallite size, though the precise relationship depends on the nanoparticle processing.³

K-factor⁴  is a method of quantifying differences in the absolute diffraction pattern intensity, which is the most obvious difference between diffraction patterns observed in Figure 1. While most often used to quantify amorphous content in specimens, the K-factor with Pt loading is shown in a linear variation in Figure 2. However, a K-factor approach would be an imprecise way to monitor loading in a production environment, because it requires precise duplication of the sample preparation. If XRD data must be collected for other reasons— for example, to monitor crystallite sizes – then the data could also be used to semi-quantitatively monitor Pt loading at no additional expense. However, if the Pt loading itself is a critical factor to monitor (and thus requires high precision), XRF is a better technique. 

Figure 1: X-ray powder diffraction data collected with an Aeris benchtop diffractometer from samples of Pt /C catalytic powder with three different Pt-loading levels (20%, 40%, 60%) on Vulcan XC-72 carbon black support particles.

Sample	Coherent crystal domain (Crystallite) size (nm)			Lattice Parameter (A)	K-Factor
	Scherrer	Williamson-Hall	Pawley
60% Pt on XC-72	5	4.5 (1)	5.53 (1)	3.9157 (1)	3171
40% Pt on XC-72	3	3.2 (2)	3.17 (1)	3.9270 (5)	1400
20% Pt on XC-72	2	2.6 (3)	2.69 (2)	3.934 (2)	612

Table 1: Pt-catalyst parameters calculated from the position, width, breadth, and intensity of X-ray diffraction peaks using HighScore Plus analysis software.

Figure 2: Variation of K-factor with Pt loading in Pt/C catalytic powders, which could be used for the approximate monitoring of Pt loading amount.

X-ray fluorescence Analysis

X-ray fluorescence (XRF) spectroscopy of Pt/C catalysts yields information about the elemental composition of Pt-metal group nano-catalyst. Data can be collected using either a benchtop or floor-standing spectrometer, based on either energy-dispersive or wavelength-dispersive methods. The XRF spectrum, example in figure 3, can immediately provide information about the Pt loading, composition of Pt alloys, or purity of the catalyst.  

Figure 3: XRF spectra showing elements present in a Pt/C catalyst obtained by measuring 40% Pt/C samples with an Epsilon 1.

Malvern Panalytical’s Omnian package provides a straightforward and simple method to measure the composition of Pt/C catalysts with XRF is using a standardless method. This solution doesn’t need in-type standards but is able to correct for differences in matrix and sample representation. This allows Omnian to be used for powder, solid and liquid samples.

Table III shows the determined Pt loading for the three commercial Pt/C test samples. The samples were prepared as loose powder in a plastic cup, and an Epsilon 1 benchtop EDXRF spectrometer was used for the measurement. The evaluation shows the 60 %- and 40 %-Pt-loaded powders are produced close to the listed specification, while the 20 %-Pt-loaded powder shows slightly % less loading than expected, which can mean the catalyst is off-spec.  

Sample	Measured Pt concentration (%)
60% Pt on XC-72	59.678
40% Pt on XC-72	40.087
20% Pt on XC-72	19.086

Table III: Pt concentrations measured using an Epsilon 1 XRF spectrometer with Omnian standardless

For further improved quality control requiring better accuracies and precision, users could opt to develop empirical calibration with in-type calibration standards. With calibration, the XRF spectrum can easily quantify the Pt loading with 0.1 - 0.3 % precision. Additional method robustness can be obtained, moving to more precise sample preparation methods, such as pressed pellets or fused beads. 

Laser diffraction

Laser diffraction gives insights into the particle size distribution of the carbon-support particles rather than the Pt catalysts. Laser diffraction is fast, non-destructive, and suited for both laboratory and continuous in-line measurements.  The wide measurement range of laser diffraction, from 10 nm to 3500 µm, is well suited to cover the coarse and fine agglomerates that may be present in a catalytic powder.  While measurement of powder particles in a dry dispersion is possible, it is more common to measure them dispersed in a solvent such as isopropyl alcohol (IPA). 

Figure 4 shows the particle size of Pt/C catalytic powders dispersed in IPA as determined using the Mastersizer 3000. Samples loaded with 60 % and 40 % Pt had similar size distributions. The powder containing 20 % Pt was different, with a significant population of smaller agglomerates in the 1 µm size range. 

The laser diffraction data reveal the most potentially significant difference between the three catalytic powders with different Pt loading levels. The difference in the initial particle size of the C agglomerates could change the way the catalytic powder disperses in the ink, which is known to have a significant effect on the deposition process and thus the performance. It is unlikely that differences in Pt loading would affect the particle size distribution of the C support particles, suggesting that another process parameter may be changing during the production of the catalytic powders. 

Figure 4: Particle size measured with a Mastersizer 3000 laser diffraction instrument from samples of Pt/C catalytic powder with three different Pt-loading levels (20%, 40%, 60%) on Vulcan XC-72 carbon black support particles.

Morphological image analysis

Automated morphological image analysis can also be used to analyze agglomerates of the C-support particles. Systems such as the Morphologi 4 image individual particles and produce a particle size distribution based on discrete particle counting. The resolution for Pt/C catalytic powders is limited because the agglomerate size is near, and below the limit of, the size range of this technique, which is 1 to >1,000 µm. Particle size determined with the Morphologi 4 may thus underrepresent the smaller agglomerates. As shown in Figure 5, the three Pt/C catalytic powders have similar particle size distributions when measured by the Morphologi 4 as obtained with laser diffraction. However, the small agglomerate population measured by laser diffraction for the 20 % Pt loaded powder is not represented in this analysis. Also, imaging techniques suffer from inferior particle statistics compared to laser diffraction but provide additional information on the shape and other particle morphology parameters.

Figure 5: Particle size by volume as measured with a Morphologi 4 from samples of Pt/C catalytic powder with three different Pt-loading levels (20%, 40%, 60%) on Vulcan XC-72 carbon black support particles.

Particle imaging with the Morphologi 4 is advantageous because images of individual agglomerates can be retrieved, compared, and evaluated by parameters such as circularity, convexity, and roughness. Figure 6 shows a snapshot of the particles imaged in the 40 % Pt loaded catalytic powder. In this case, the agglomerates are nearly spherical not exhibiting any significant anisotropy to affect the ink processability. However, the roughness of the agglomerates could affect the ease of dispersion in the ink and the formation of porosity during deposition and drying. 

Figure 6: Partial catalog of the Pt/C catalyst powders imaged by the Morphologi 4. 

Characterization of alloy catalysts

In the section above, catalytic powders using pure Pt were analyzed. These techniques can also be used to measure catalytic alloys. Some alloys, such as platinum-cobalt (PtCo), are being studied with the aim of reducing their costs by lowering the amount of precious metal required to produce PEMFCs. Others, such as platinum-ruthenium (PtRu), are being investigated with the aim of enabling reversible PEMFCs that can be used as electrolyzers for H2 production and as fuel cells to generate electricity from H2. 

In addition to the analysis related to crystallite size and lattice parameter as discussed above, XRD can be used to confirm the phase of catalytic alloys. Figure 7 compares the diffraction patterns of PtRu, PtCo, and Pt catalysts loaded at 40% on Vulcan XC-72 carbon black support particles. The diffraction pattern of PtRu indicates that it had an FCC crystal structure, like that of Pt, whereas the diffraction pattern of PtCo revealed a primitive tetragonal crystal structure. This is an important observation, since PtCo can be produced in FCC, face-centered tetragonal, and primitive tetragonal polymorphs. The crystal structure significantly affects the catalytic activity as it determines the type of exposed facet, which strongly differ from one another in terms of surface energy.⁵

Rietveld refinement of the PtCo/C diffraction pattern, shown in Figure 8, revealed that modeling the catalyst as a purely tetragonal PtCo does not satisfactorily reproduce the experimental data. The fundamental reflections – those shared between the cubic alloy phase and the tetragonal intermetallic phase – have undercalculated intensities. A satisfactory fit of the experimental XRD data is only produced when the sample is modeled as a mixture of 69% tetragonal PtCo intermetallic and 31% cubic PtCo alloy. This phase mixture has significant implications for the performance of this catalytic powder.

The volume average crystallite size of the PtCo alloy was 5 nm, larger than that of the pure Pt (3 nm) and Pt-Ru (2 nm).  

XRF analysis of loose powders, performed using standardless quantification with Omnian analysis software, determined that the mass ratio of the PtCo alloy was 0.28, close to the nominal value of 0.3. Likewise, the calculated mass ratio of the PtRu alloy was 0.47, close to the nominal value of 0.52. 

Laser diffraction analysis revealed that the C-support particle agglomerates had similar sizes for catalytic powders made with all three alloys, as shown in Figure 9.

Figure 7: X-ray powder diffraction data collected with an Aeris benchtop X-ray diffractometer from samples of 40%-Pt-loaded catalytic powder with three different Pt alloys (Pt, PtRu, PtCo) on Vulcan XC-72 carbon black support particles.

Figure 8: Comparison of Rietveld modeling of 40%-loaded PtCo on C-support catalytic powder, when modeled as pure tetragonal PtCo (left) and as a mixture of the tetragonal intermetallic and cubic alloy (right).

Figure 9: Laser diffraction data measured with a Mastersizer 3000 for catalytic powders consisting of 40% loading of PtRu on C, Pt on C, and PtCu on C.

Conclusion

The studies outlined above show that XRD, XRF, laser diffraction, and automated image analysis can all be effective techniques for analyzing various properties of pure and alloyed Pt nanoparticles supported on a C matrix. 

XRD is particularly ideal for analyzing crystallite size, strain, lattice parameters, and phase composition. XRF measures Pt loading more precisely, as well as providing effective analysis of Pt catalyst purity and Pt alloy composition. Laser diffraction is ideal for analyzing the particle size of C-support that influences the Pt dispersion and porosity in the catalytic active compound. Finally, while morphological imaging provides limited insight into the particle size, it is valuable in providing images of individual agglomerates to enable shape analysis. Together, these techniques enable manufacturers to optimize both the cost and performance of systems based on PEM technology by maximizing catalyst efficiency, reducing the amount of Pt required, or developing Pt-based alloy nanoparticles. 

References

1. Popov, Branko N., and Weidner, John. Development of Ultra-Low Platinum Alloy Cathode Catalysts for PEM Fuel Cells. (2016) United States. https://doi.org/10.2172/1330985
   
2. M Wang, JH Park, S Kabir, KC Neyerlin, NN Kariuki, H Lv, VR Stamenkovic, DJ Myers, M Ulsh, and SA Mauger, “Impact of Catalyst Ink Dispersing Methodology on Fuel Cell Performance using in-situ X-ray Scattering”, ACS Appl. Energy Mater. 2019, 2, 9, 6417-6427. https://doi.org/10.1021/acsaem.9b01037 
   
3. Leontyev, I. N. and Kuriganova, A. B. and Leontyev, N. G. and Hennet, L. and Rakhmatullin, A. and Smirnova, N. V. and Dmitriev, V., “Size dependence of the lattice parameters of carbon-supported platinum nanoparticles: X-ray diffraction analysis and theoretical considerations” RSC Adv., 2014, 4, 35959-35965, https://dx.doi.org/10.1039/C4RA04809A 
   
4. O'Connor, B., & Raven, M. (1988). Application of the Rietveld Refinement Procedure in Assaying Powdered Mixtures. Powder Diffraction, 3(1), 2-6. https://doi.org/10.1017/S0885715600013026
   
5. Chen, Liang & Bock, Christina & MacDougall, Barry & Mercier, Patrick. (2011). “Characteristics of Cubic and Tetragonal Pt”. ECS Transactions. 35. 69-80. https://doi.org/10.1149/1.3654203