- Email: [email protected]
Three-dimensional microstructure analysis of a polymer electrolyte membrane water electrolyzer anode
Journal of Power Sources 393 (2018) 62–66
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
Three-dimensional microstructure analysis of a polymer electrolyte membrane water electrolyzer anode
T
Friedemann Heggea, Riko Moronia, Patrick Trinkeb, Boris Bensmannb, Richard Hanke-Rauschenbachb, Simon Thielea,c,d,e,f, Severin Vierratha,c,∗ a
Laboratory for MEMS Applications, IMTEK Department of Microsystems Engineering, University of Freiburg, Georges-Koehler-Allee 103, 79110, Freiburg, Germany Institute of Electric Power Systems, Leibniz Universität Hannover, Appelstr. 9A, 30167, Hannover, Germany c Hahn-Schickard, Georges-Koehler-Allee 103, 79110, Freiburg, Germany d Forschungszentrum Jülich GmbH, Helmholtz-Institute Erlangen-Nürnberg for Renewable Energy (IEK-11), Forschungszentrum Jülich, Egerlandstr. 3, 91058, Erlangen, Germany e Department of Chemical and Biological Engineering, Friedrich-Alexander-Universität Erlangen-Nürnberg, Egerlandstr. 3, 91058, Erlangen, Germany f FIT - Freiburg Center for Interactive Materials and Bioinspired Technologies, Georges-Köhler-Allee 105, 79110, Freiburg, Germany b
H I GH L IG H T S
FIB-SEM tomography of a PEM water electrolyzer anode catalyst layer is presented. • APores are contrasted versus catalyst particles by silicone infiltration. • Modeling ionomer contents reveals its influence on transport parameters. • The study different confirms an ideal ionomer content of 40–50% of the pore volume. •
A R T I C LE I N FO
A B S T R A C T
Keywords: PEM water electrolyzers FIB-SEM tomography Catalyst layers Microstructure Ionomer modeling
The anode catalyst layer of a PEM water electrolyzer is reconstructed using a combination of FIB-SEM tomography and ionomer modeling. The pore space is infiltrated with silicone, enabling good discrimination between pores and IrRuOx catalyst material, while the ionomer cannot be imaged. The reconstructed volume of 29 μm × 24 μm x 7 μm contains catalyst particles with a median size of 0.5 μm and has a porosity of 55%. By modeling different ionomer contents inside the pore space, the impact on microstructural and transport parameters is investigated. At an ionomer content of 40–50% of the pore volume, all transport parameters are in a reasonable range, confirming experimental results from literature. At an ionomer content of 48% the catalyst layer has a porosity of 29%, a median pore size of 0.94 μm, a permeability of the pore space of 1.8 mD and a mean ionomer film thickness of 0.4 μm . The tortuosities of the ionomer and the pore space are calculated to 3.5 and 6.7 at the corresponding phase fractions of 26% and 29% respectively. The electrochemically active surface area estimated from the tomography (1.0 m2 g−1) is considerably lower than literature values, indicating a roughness below FIB-SEM resolution.
1. Introduction Hydrogen is not only a medium for energy storage with high energy density, but also an important precursor for the production of different chemical products such as ammonia or methanol [1]. Hence, water electrolysis represents a seminal technology, which can link the electricity sector with e.g. the mobility sector or the chemical industry [2]. This enables a flexible utilization of growing intermediate
overproduction by renewable energy sources. Polymer electrolyte membrane water electrolyzers (PEMWEs) are still too expensive to achieve high market penetration [3]. Besides saving resources and lowering operational expenditures, higher conversion efficiencies (high current densities at low voltages) result in a more efficient material use and thus reduced capital expenditures. Therefore, optimizing the design with the aim of achieving the highest possible conversion efficiency plays a key role for the future of PEMWE technology.
∗ Corresponding author. Laboratory for MEMS Applications, IMTEK Department of Microsystems Engineering, University of Freiburg, Georges-Koehler-Allee 103, 79110, Freiburg, Germany. E-mail address: [email protected] (S. Vierrath).
https://doi.org/10.1016/j.jpowsour.2018.04.089 Received 10 January 2018; Received in revised form 25 April 2018; Accepted 26 April 2018 0378-7753/ © 2018 Elsevier B.V. All rights reserved.
Journal of Power Sources 393 (2018) 62–66
F. Hegge et al.
imaging, a gold layer was sputtered on top of the sample.
PEMWEs comprise an anode at which the oxygen evolution reaction (OER) takes place, a polymer electrolyte membrane (PEM) for proton conduction as well as selective inhibition of gas or electron cross-over and a cathode at which the hydrogen evolution reaction (HER) occurs. Porous transport layers (PTL) enable two phase transport of water and produced oxygen at the anode and hydrogen at the cathode. These are located between the so called flow fields and the catalyst layers. At the cathode usually a carbon supported platinum catalyst is used, enabling fast HER kinetics. In contrast, the reaction kinetics of the OER are sluggish and account for a major part of the conversion losses [4]. Therefore, much effort has been put into optimizing the material composition of the anode in order to improve the catalytic activity [5–7]. State-of-the-art anodes are porous structures comprising Iridium (Ir), Iridiumoxide (IrOx) or Iridumrutheniumoxide (IrRuOx) as catalyst, which are sometimes supported by Titaniumoxide (TiOx) and typically bound by an ionomer [4]. The metallic catalyst material has to provide both active sites and pathways for the electrons. Pores serve as pathways for water access to reaction sites and as escape channels for produced oxygen. Water filled pore space furthermore enables proton conduction over short distances [8]. The proton conduction is increased by the ionomer binder. Thus a complex interplay of different interfaces and transport phases occurs in a working catalyst layer. Consequently, besides the material composition, the microstructure of the anode has a significant impact on the electrolyzer's performance as it determines catalyst accessibility and species transport [9]. Hence, optimizing the anode microstructure is crucial for obtaining higher efficiency and reducing the catalyst loading. Investigating the microstructure by tomographic methods is an important step towards a successful optimization process. Tomographic methods have been well established in other electrochemical cell technologies such as polymer electrolyte membrane fuel cells [10]. For PEMWEs a small number of publications exist, which investigate the structure of the PTL using X-ray tomography [11–13]. However, to our knowledge no work on a tomographic reconstruction of the anode microstructure of a PEMWE has been published yet. In this paper we analyze the anode microstructure of a commercially available catalyst coated membrane (CCM) of a PEMWE. Focused ion beam scanning electron microscope (FIB-SEM) tomography is used, as its resolution and field of view is suitable for the investigated anode microstructure [14]. Prior to the tomography the pore space was infiltrated with silicone in order to improve the contrast between pores and catalyst material. The ionomer phase was modeled inside the pore space, since it cannot be captured in the silicone-contrasted tomography. Based on the reconstruction and several modeled ionomer contents, we calculate pore size distributions, grain size distributions, transport parameters and the electrochemical active surface area of the anode.
2.2. FIB-SEM reconstruction and ionomer modeling After sample preparation, a cubical volume was reconstructed using a Zeiss Neon 40 EsB FIB-SEM. Therefore, a trench of a few μm depth around a peninsula of app. 34 μm × 28 μm x 20 μm was milled into the sample with a FIB current of 2 nA and 30 kV accelerating voltage. From this, 650 FIB-SEM cross sections were generated with a SEM image resolution of 20 nm at 5 kV SEM accelerating voltage, using secondary electron (SE) and backscattered electron (BSE) detectors and a FIB cutting distance of 35 nm at 1 nA FIB current. From this image series a cubical volume of 29 μm × 24 μm x 7 μm was reconstructed in MATLAB and Fiji [17] following the steps: image registration, geometric correction, cropping and segmentation. SE images showed a high material contrast but poor image quality, while BSE images featured better image resolution, especially at the edges of the IrRuOx phase but a worse material contrast and distortions due to charging effects of the silicone. Therefore, images from both detectors were combined. First, a threshold was applied to the SE images to yield the regions with catalyst material. These regions were then dilated and served as a mask in the BSE images. The masked BSE images were then segmented using the trainable WEKA segmentation plugin of Fiji. With the reconstruction of the catalyst material and the pores as a basis, the ionomer was modeled inside the pore space with the “add binder” function of the ProcessGeo module of GeoDict [18]. The algorithm is based on the assumption that the binder behaves like a wetting fluid. Thus, small pores are filled first and larger pores are filled gradually with increasing ionomer content of the catalyst layer [19]. 2.3. Transport and geometric parameter calculation The mean anode layer thickness l was determined with MATLAB by averaging the cord length distribution from the membrane to the top of the reconstructed volume. The following modules of GeoDict were used for calculations: PoroDict for size distributions and surface area as described by Wiegmann and Glatt [20], FlowDict for permeabilities and ConductoDict for tortuosities. The mean ionomer thickness d was determined by calculating the shortest distance through the ionomer from pore each catalyst surface voxel to the open pore space d min or to the PTL PTL interface d min (see Fig. 3b). Taking the mean of all calculated distances yields d . For this calculation the reconstructed volume of 7 μm height was extended by mirroring to match the catalyst layer thickness. 2.4. Electrochemical measurements The polarization measurements of the FuelCellsEtc EZ CCM (based on a Nafion N115 membrane) were performed with a Scribner 857 Redox Flow potentiostat at atmospheric pressure conditions and 80 °C cell temperature. A 5 cm2 PEMWE electrolyzer cell was used with a 1 mm Mott sintered titanium PTL on anode side and a Freudenberg H23I2 PTL on cathode side. Anode and cathode deionized water feed was adjusted to 40 ml/min. The high frequency resistance (HFR) free voltage was obtained by measuring the HFR at each measurement point at 1 kHz.
2. Methods and experimental 2.1. PEMWE anode sample The investigated sample is the anode catalyst of the commercially available FuelCellsEtc EZ CCM with a loading of 3 mg cm−2 . The catalyst material is Ir0.5Ru0.5Ox and contains no supporting material, as determined by Energy-Dispersive X-ray Spectroscopy (EDX). In order to facilitate the segmentation of the microstructure reconstruction, the phase contrast between solid material and the pore space was enhanced by infiltrating the sample with silicone as proposed by Ender et al. [15]. Silicone yields a good contrast to the catalyst material. For good penetration of the pore space, a low viscosity silicone was used (Wacker Elastosil RT 604) following Liu et al. [16]. To further improve penetration, the sample was constantly kept under vacuum atmosphere, starting 1 h before infiltration until 30 min after. Excess silicone was removed and the sample was cured for 24 h at room temperature. In order to mitigate charging effects and sample drifting during SEM
3. Results 3.1. Microstructure of the catalyst layer From the catalyst layer a volume of 29 μm × 24 μm x 7 μm was reconstructed using FIB-SEM tomography. In order to enable discrimination between pores and the catalyst material, silicone was introduced into the pore space. However, the silicone has a low contrast to the ionomer. For this reason, the reconstruction contains only the 63
Journal of Power Sources 393 (2018) 62–66
F. Hegge et al.
Fig. 1. a) 3D-representation of the reconstructed volume with modeled ionomer (contents from 0 – 100% related to the pore volume). The z-direction is normal to the plane of the catalyst layer (through plane direction). b) Actual ionomer of the catalyst layer (highlighted blue in SEM cross section) resembles the modeled ionomer shown in a). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
unsupported IrRuOx catalyst structure and the pore space. Therefore, different ionomer contents were virtually added into the pore space as described in section 2.2. Fig. 1a shows the reconstructed catalyst layer comprising the IrRuOx structure and increasing ionomer contents, ranging from 0% to 100% with respect to the pore volume. The modeled ionomer structure is similar to that of the actual ionomer, as can be seen in a SEM cross section in Fig. 1b. The reconstructed volume contains catalyst particles with a median size of 0.5 μm and has a porosity of 55%. The virtual addition of ionomer affects the pore sizes and the sizes of the agglomerates of catalyst particles and ionomer. This can be seen in the porosity, in the pore size distribution and agglomerate size distribution as depicted in Fig. 2a and b and Table 1. The porosity decreases with the ionomer content due to the gradual substitution of pore space by the ionomer. The mean agglomerate size and the mean pore size increase due to the added ionomer, which fills small pores first according to its wetting behavior. A
Table 1 Morphological parameters of reconstructed volume. Ionomer content in pore space vol%
Overall ionomer fraction vol%
Porosity vol%
Median pore diameter μm
Median agglomerate diameter μm
0 25 48 73
0 14 26 40
55 41 29 15
0.53 0.71 0.94 1.39
0.50 0.89 1.74 4.08
reasonable amount of wet ionomer in the catalyst layer is approximately 50% of the pore space volume [21]. Due to the discrete nature of the ionomer modeling, a value of 48% was chosen as a reference, yielding a median agglomerate size of 1.74 μm, a median pore size of 0.94 μm and a porosity of εp = 0.29.
3.2. Species transport In the catalyst layer, reaction species have to be transported to and from the active sites. The microstructure has a strong impact on the transport of each reactant. In the following, each species transport mechanism is discussed based on calculated transport parameters. To estimate water and oxygen transport through the pore space, the permeability and the relative effective diffusivity were determined, as both contribute to the transport. The ionomer, as a possible parallel pathway, was not taken into account due to its very low permeability [22] in comparison to the permeability of the pore network. The calculated permeability only depends on the morphology and is valid for a single phase fluid transported through the pore space (Darcy's law). This can be used as input for two phase transport calculations of water and oxygen through the pore space. However, this requires an additional relation between capillary pressure and saturation [23]. As displayed in Fig. 3a, the through-plane permeability decreases with the ionomer content up to a calculated percolation threshold of approx. 90% ionomer content in the pore space. Above this value, the only pathway for water and oxygen is through the ionomer. Below this value, it is a combination of transport through pore space and ionomer. At the reference ionomer content of 48%, the permeability is 1.8⋅10−15 m2 = 1.8 mD . For PEMWE anode catalyst layers no values of the saturated permeability have been reported in literature so far. The calculated value lies between typical values of PTLs or cathode gas diffusion layers in an order of 10−11 m2 and values of cathode catalyst layers, as assumed for fuel cell cathodes in simulation models, in an order of 10−16 m2 [13] [24,25]. The produced oxygen has to pass through the ionomer from the catalyst surface to the pore space and then through the pore space to the porous transport layer, while the water pathway is in the opposite direction. In order to quantify the transport resistance of the ionomer, the mean ionomer film thickness d around the IrRuOx phase was estimated (Fig. 3b) as described in section 2.3. At a pore space ionomer content of
Fig. 2. Pore (a) and agglomerate size (b) distributions for increasing ionomer contents related to the pore volume. 64
Journal of Power Sources 393 (2018) 62–66
F. Hegge et al.
Fig. 3. Through-plane permeability (a) and mean ionomer film thickness (b) vs. ionomer content in the pore space x . Fitting the permeability based on percolation law [26] yields k = kmax ((xp − x ) xp−1) b for x < x p and k = 0 m² otherwise, with kmax = 4.74⋅10−15m2 , xp = 0.92 , b = 1.241.
Fig. 4. Tortuosity of the ionomer (a) and tortuosity of the pore space (b) vs. ionomer content in the pore space and overall phase fraction. Fitting tortuosities based on percolation law for the transport parameters [26] and definition 1 ((ε − ε p)(0.553 − ε p)−1) b for ε > ε p and of tortuosity (see above) yields τ = ε α −max τ = 0 otherwise, with α = α 0α −eff1 and α max = 0.31, ε p = 0.072 , b = 1.60 (a) and α max = 0.31, ε p = 0 , b = 3.05 (b).
100%, the mean ionomer film thickness is more than half the anode layer thickness. At the reference ionomer content of 48%, the ionomer film thickness (d= 0.4 μm) is significantly lower. To describe relative effective diffusivities and conductivities the tortuosity τ= α 0 α−eff1 ε is commonly used [27]. Here α 0 is the intrinsic transport value for the conducting phase (electric conductivity of the IrRuOx, water or oxygen diffusivity of the two phase mixture in the pore space or ionic conductivity of the ionomer), α eff is the effective transport value for the composite and ε is the fraction of the transporting phase. For the electron conduction a through-plane tortuosity of the IrRuOx material phase of τIrRuOx = 4.88 was calculated (phase fraction of εIrRuOx = 45 %). The other tortuosities depend on the virtually added ionomer content, as the fractions and the morphology of the transporting phases change. Proton conductivity is described by the tortuosity of the ionomer, as plotted against the ionomer content in the pore space in Fig. 4a. The decreasing ionomer tortuosity with growing phase fraction corresponds to an increasing proton conductivity. At an overall ionomer fraction of approx. 7 vol%, the tortuosity runs into a percolation threshold of the fitted curve. At the reference ionomer content of 48% of the pore space (26% overall ionomer phase fraction) the throughplane ionomer tortuosity is 3.5. The oxygen diffusion in the pore space is affected by the pore space tortuosity, which is plotted against the ionomer content in the pore space and the porosity in Fig. 4b. The pore space tortuosity increases with the ionomer content. At the reference ionomer content of 48% of the pore space (29% overall porosity), its through-plane tortuosity is 6.7 . It is to note that the pore tortuosity values at high ionomer contents are not expected to be representative. Due to very few open pores at
high ionomer content, the tortuosity values are discretized and do not follow the percolation law applied for the fitted curve [26]. For this reason, a precise percolation threshold value cannot be provided. Also for this reason, data points below 10% porosity were excluded from the fit. As depicted in Fig. 4a, the ionomer tortuosity decreases strongly up to an ionomer content of 40–50% of the pore space volume. At the same time there is still a comparably low pore space tortuosity and ionomer film thickness (Figs. 4b and 3b). Furthermore, the permeability is in the medium range at this ionomer content (Fig. 3a). Without considering the precise influence of the single transport parameters in a full cell model, an ionomer content of 40–50% of the pore space volume seems to be a good range for an optimal species transport. This is also in agreement with the optimal ionomer content from experimental studies [21]. 3.3. Catalyst surface area Besides its impact on transport, the microstructure of the catalyst layer determines the accessibility of the catalyst. This is typically quantified in the electrochemically active surface area (ECSA). Since the investigated catalyst layer contains no support material, we assume that the entire surface area of the catalyst material is active, hence equals the ECSA. With this assumption the ECSA can be estimated from the volume specific surface area obtained from the tomographic dataset. The ECSA is calculated with the volume specific surface area of 65
Journal of Power Sources 393 (2018) 62–66
F. Hegge et al.
commercially available state-of-the-art catalyst layer, offer new possibilities for the extension of electrochemical models of PEM water electrolyzers. This will provide deeper insight into the relation between the structure of an electrolyzer anode and the electrochemical performance of the whole cell. Acknowledgements We acknowledge funding by the German Federal Ministry of Education (BMBF) within the POWER-MEE project (Grant No. 03SF0536A). The authors further thank Andrew Mondon and Brian Shanahan for their support. References Fig. 5. Polarization and HFR free curve of FuelCellsEtc EZ CCM, measured at cell temperature of 80 °C and atmospheric pressure conditions (mean HFR = 140 mΩ cm2 ).
[1] International Energy Agency, Technology Roadmap Energy and GHG Reductions in the Chemical Industry via Catalytic Processes, (2013). [2] B. Bensmann, R. Hanke-Rauschenbach, I.K. Peña Arias, K. Sundmacher, Electrochim. Acta 110 (2013) 570–580. [3] L. Bertuccioli, A. Chan, D. Hart, F. Lehner, B. Madden, E. Standen, Study on Development of Water Electrolysis in the EU, (2014). [4] M. Carmo, D.L. Fritz, J. Mergel, D. Stolten, Int. J. Hydrogen Energy 38 (2013) 4901–4934. [5] R.S. Yeo, J. Orehotsky, W. Visscher, S. Srinivasan, J. Electrochem. Soc. 128 (1981) 1900–1904. [6] J. Ahn, R. Holze, J. Appl. Electrochem. 22 (1992) 1167–1174. [7] L. Wang, F. Song, G. Ozouf, D. Geiger, T. Morawietz, M. Handl, P. Gazdzicki, C. Beauger, U. Kaiser, R. Hiesgen, A.S. Gago, K.A. Friedrich, J. Mater. Chem. 5 (2017) 3172–3178. [8] N. Agmon, Chem. Phys. Lett. 244 (1995) 456–462. [9] R. García-Valverde, N. Espinosa, A. Urbina, Int. J. Hydrogen Energy 37 (2012) 1927–1938. [10] M.H. Shojaeefard, G.R. Molaeimanesh, M. Nazemian, M.R. Moqaddari, Int. J. Hydrogen Energy 41 (2016) 20276–20293. [11] M. Suermann, K. Takanohashi, A. Lamibrac, T.J. Schmidt, F.N. Büchi, J. Electrochem. Soc. 164 (2017) F973–F980. [12] F. Arbabi, A. Kalantarian, R. Abouatallah, R. Wang, J.S. Wallace, A. Bazylak, J. Power Sources 258 (2014) 142–149. [13] L. Zielke, A. Fallisch, N. Paust, R. Zengerle, S. Thiele, RSC Adv. 4 (2014) 58888–58894. [14] M. Cantoni, L. Holzer, MRS Bull. 39 (2014) 354–360. [15] M. Ender, J. Joos, T. Carraro, E. Ivers-Tiffée, Electrochem. Commun. 13 (2011) 166–168. [16] Z. Liu, Y.-c.K. Chen-Wiegart, J. Wang, S.A. Barnett, K.T. Faber, Microsc. Microanal. 22 (2016) 140–148. [17] Fiji, https://fiji.sc, accessed 18 April 2018. [18] GeoDict, https://www.geodict.com, accessed 18 April 2018. [19] A. Pfrang, D. Veyret, G. Tsotridis, Computation of Thermal Conductivity of Gas Diffusion Layers of PEM Fuel CellsConvection and Conduction Heat Transfer, (2011). [20] A. Wiegmann, E. Glatt, PoroDict, (2017). [21] M. Bernt, H.A. Gasteiger, J. Electrochem. Soc. 163 (2016) F3179–F3189. [22] Dawn M. Bernardi, Mark W. Verbrugge, J. Electrochem. Soc. 139 (1992). [23] B. Han, J. Mo, Z. Kang, F.-Y. Zhang, Electrochim. Acta 188 (2016) 317–326. [24] I.V. Zenyuk, P.K. Das, A.Z. Weber, J. Electrochem. Soc. 163 (2016) F691–F703. [25] X. Wang, T. van Nguyen, J. Electrochem. Soc. 155 (2008) B1085. [26] S. Kirkpatrick, Rev. Mod. Phys. 45 (1973) 574–588. [27] M. Doyle, J. Newman, A.S. Gozdz, C.N. Schmutz, J. Tarascon, J. Electrochem. Soc. 143 (1996) 1890–1903. [28] T. Audichon, E. Mayousse, S. Morisset, C. Morais, C. Comminges, T.W. Napporn, K.B. Kokoh, Int. J. Hydrogen Energy 39 (2014) 16785–16796. [29] J. Mo, Z. Kang, G. Yang, S.T. Retterer, D.A. Cullen, T.J. Toops, J.B. Green, F.Y. Zhang, Appl. Energy 177 (2016) 817–822. [30] A. Marshall, S. Sunde, M. Tsypkin, R. Tunold, Int. J. Hydrogen Energy 32 (2007) 2320–2324. [31] J. Cheng, H. Zhang, G. Chen, Y. Zhang, Electrochim. Acta 54 (2009) 6250–6256.
a= 3.1⋅106 m−1, the catalyst layer thickness of l= 10 μm and the catalyst loading of L cat = 3 mg⋅cm−2 as ECSA = a⋅l⋅L−cat1 = 1.03 m2g−1. This is considerably lower compared to values of IrRuOx catalysts determined by BET. Audichon et al. [28]. measured an ECSA of 47 m2g−1 for a Ir0.5Ru0.5Ox catalyst. However, they used IrRuOx nanoparticles with a particle size between 20 nm and 200 nm, which is much smaller than the 0.5 μm median particle size of our sample and is likely to be the main reason for the higher ECSA. This difference in particle size can maximally explain a factor of 25 assuming perfect spheres. Therefore, the rest of the difference could stem from a surface roughness below the FIB-SEM resolution and inaccuracies of the BET measurements of Audichon et al. Another indicator for a sub-resolution roughness is the decent electrochemical performance in the kinetic region (Fig. 5), which lies in the range of literature values [21,29–31]. The performance in the kinetic region is mainly governed by the ECSA of the anode and its catalyst activity [21]. Assuming a similar activity leads to the conclusion that the ECSA has to be comparable to the literature values and therefore higher than estimated from the reconstruction. 4. Conclusion We present the 3D microstructure of the anode catalyst of a PEM water electrolyzer obtained from a combination of FIB-SEM tomography and ionomer modeling. From the reconstructed volume (29 μm x 24 μm x 7 μm) size distributions, pore permeability, mean ionomer film thickness and phase tortuosities were calculated. The modeled ionomer phase allowed investigating the effect of ionomer content on these structural and transport parameters. This showed that at an ionomer content of 40–50% all important transport parameters are in a reasonable range confirming experimental results from literature. Further, the electrochemical active surface area (ECSA) was estimated from the reconstructed catalyst layer to be 1.0 m2g−1. This is considerably lower than experimental values presented in literature. One possible cause for this discrepancy is the limited resolution of the FIB-SEM tomography, which might substantially underestimate surface roughness of the catalyst particles. The presented microstructure parameters, which characterize a
66
Contents lists available at ScienceDirect
Journal of Power Sources journal homepage: www.elsevier.com/locate/jpowsour
Three-dimensional microstructure analysis of a polymer electrolyte membrane water electrolyzer anode
T
Friedemann Heggea, Riko Moronia, Patrick Trinkeb, Boris Bensmannb, Richard Hanke-Rauschenbachb, Simon Thielea,c,d,e,f, Severin Vierratha,c,∗ a
Laboratory for MEMS Applications, IMTEK Department of Microsystems Engineering, University of Freiburg, Georges-Koehler-Allee 103, 79110, Freiburg, Germany Institute of Electric Power Systems, Leibniz Universität Hannover, Appelstr. 9A, 30167, Hannover, Germany c Hahn-Schickard, Georges-Koehler-Allee 103, 79110, Freiburg, Germany d Forschungszentrum Jülich GmbH, Helmholtz-Institute Erlangen-Nürnberg for Renewable Energy (IEK-11), Forschungszentrum Jülich, Egerlandstr. 3, 91058, Erlangen, Germany e Department of Chemical and Biological Engineering, Friedrich-Alexander-Universität Erlangen-Nürnberg, Egerlandstr. 3, 91058, Erlangen, Germany f FIT - Freiburg Center for Interactive Materials and Bioinspired Technologies, Georges-Köhler-Allee 105, 79110, Freiburg, Germany b
H I GH L IG H T S
FIB-SEM tomography of a PEM water electrolyzer anode catalyst layer is presented. • APores are contrasted versus catalyst particles by silicone infiltration. • Modeling ionomer contents reveals its influence on transport parameters. • The study different confirms an ideal ionomer content of 40–50% of the pore volume. •
A R T I C LE I N FO
A B S T R A C T
Keywords: PEM water electrolyzers FIB-SEM tomography Catalyst layers Microstructure Ionomer modeling
The anode catalyst layer of a PEM water electrolyzer is reconstructed using a combination of FIB-SEM tomography and ionomer modeling. The pore space is infiltrated with silicone, enabling good discrimination between pores and IrRuOx catalyst material, while the ionomer cannot be imaged. The reconstructed volume of 29 μm × 24 μm x 7 μm contains catalyst particles with a median size of 0.5 μm and has a porosity of 55%. By modeling different ionomer contents inside the pore space, the impact on microstructural and transport parameters is investigated. At an ionomer content of 40–50% of the pore volume, all transport parameters are in a reasonable range, confirming experimental results from literature. At an ionomer content of 48% the catalyst layer has a porosity of 29%, a median pore size of 0.94 μm, a permeability of the pore space of 1.8 mD and a mean ionomer film thickness of 0.4 μm . The tortuosities of the ionomer and the pore space are calculated to 3.5 and 6.7 at the corresponding phase fractions of 26% and 29% respectively. The electrochemically active surface area estimated from the tomography (1.0 m2 g−1) is considerably lower than literature values, indicating a roughness below FIB-SEM resolution.
1. Introduction Hydrogen is not only a medium for energy storage with high energy density, but also an important precursor for the production of different chemical products such as ammonia or methanol [1]. Hence, water electrolysis represents a seminal technology, which can link the electricity sector with e.g. the mobility sector or the chemical industry [2]. This enables a flexible utilization of growing intermediate
overproduction by renewable energy sources. Polymer electrolyte membrane water electrolyzers (PEMWEs) are still too expensive to achieve high market penetration [3]. Besides saving resources and lowering operational expenditures, higher conversion efficiencies (high current densities at low voltages) result in a more efficient material use and thus reduced capital expenditures. Therefore, optimizing the design with the aim of achieving the highest possible conversion efficiency plays a key role for the future of PEMWE technology.
∗ Corresponding author. Laboratory for MEMS Applications, IMTEK Department of Microsystems Engineering, University of Freiburg, Georges-Koehler-Allee 103, 79110, Freiburg, Germany. E-mail address: [email protected] (S. Vierrath).
https://doi.org/10.1016/j.jpowsour.2018.04.089 Received 10 January 2018; Received in revised form 25 April 2018; Accepted 26 April 2018 0378-7753/ © 2018 Elsevier B.V. All rights reserved.
Journal of Power Sources 393 (2018) 62–66
F. Hegge et al.
imaging, a gold layer was sputtered on top of the sample.
PEMWEs comprise an anode at which the oxygen evolution reaction (OER) takes place, a polymer electrolyte membrane (PEM) for proton conduction as well as selective inhibition of gas or electron cross-over and a cathode at which the hydrogen evolution reaction (HER) occurs. Porous transport layers (PTL) enable two phase transport of water and produced oxygen at the anode and hydrogen at the cathode. These are located between the so called flow fields and the catalyst layers. At the cathode usually a carbon supported platinum catalyst is used, enabling fast HER kinetics. In contrast, the reaction kinetics of the OER are sluggish and account for a major part of the conversion losses [4]. Therefore, much effort has been put into optimizing the material composition of the anode in order to improve the catalytic activity [5–7]. State-of-the-art anodes are porous structures comprising Iridium (Ir), Iridiumoxide (IrOx) or Iridumrutheniumoxide (IrRuOx) as catalyst, which are sometimes supported by Titaniumoxide (TiOx) and typically bound by an ionomer [4]. The metallic catalyst material has to provide both active sites and pathways for the electrons. Pores serve as pathways for water access to reaction sites and as escape channels for produced oxygen. Water filled pore space furthermore enables proton conduction over short distances [8]. The proton conduction is increased by the ionomer binder. Thus a complex interplay of different interfaces and transport phases occurs in a working catalyst layer. Consequently, besides the material composition, the microstructure of the anode has a significant impact on the electrolyzer's performance as it determines catalyst accessibility and species transport [9]. Hence, optimizing the anode microstructure is crucial for obtaining higher efficiency and reducing the catalyst loading. Investigating the microstructure by tomographic methods is an important step towards a successful optimization process. Tomographic methods have been well established in other electrochemical cell technologies such as polymer electrolyte membrane fuel cells [10]. For PEMWEs a small number of publications exist, which investigate the structure of the PTL using X-ray tomography [11–13]. However, to our knowledge no work on a tomographic reconstruction of the anode microstructure of a PEMWE has been published yet. In this paper we analyze the anode microstructure of a commercially available catalyst coated membrane (CCM) of a PEMWE. Focused ion beam scanning electron microscope (FIB-SEM) tomography is used, as its resolution and field of view is suitable for the investigated anode microstructure [14]. Prior to the tomography the pore space was infiltrated with silicone in order to improve the contrast between pores and catalyst material. The ionomer phase was modeled inside the pore space, since it cannot be captured in the silicone-contrasted tomography. Based on the reconstruction and several modeled ionomer contents, we calculate pore size distributions, grain size distributions, transport parameters and the electrochemical active surface area of the anode.
2.2. FIB-SEM reconstruction and ionomer modeling After sample preparation, a cubical volume was reconstructed using a Zeiss Neon 40 EsB FIB-SEM. Therefore, a trench of a few μm depth around a peninsula of app. 34 μm × 28 μm x 20 μm was milled into the sample with a FIB current of 2 nA and 30 kV accelerating voltage. From this, 650 FIB-SEM cross sections were generated with a SEM image resolution of 20 nm at 5 kV SEM accelerating voltage, using secondary electron (SE) and backscattered electron (BSE) detectors and a FIB cutting distance of 35 nm at 1 nA FIB current. From this image series a cubical volume of 29 μm × 24 μm x 7 μm was reconstructed in MATLAB and Fiji [17] following the steps: image registration, geometric correction, cropping and segmentation. SE images showed a high material contrast but poor image quality, while BSE images featured better image resolution, especially at the edges of the IrRuOx phase but a worse material contrast and distortions due to charging effects of the silicone. Therefore, images from both detectors were combined. First, a threshold was applied to the SE images to yield the regions with catalyst material. These regions were then dilated and served as a mask in the BSE images. The masked BSE images were then segmented using the trainable WEKA segmentation plugin of Fiji. With the reconstruction of the catalyst material and the pores as a basis, the ionomer was modeled inside the pore space with the “add binder” function of the ProcessGeo module of GeoDict [18]. The algorithm is based on the assumption that the binder behaves like a wetting fluid. Thus, small pores are filled first and larger pores are filled gradually with increasing ionomer content of the catalyst layer [19]. 2.3. Transport and geometric parameter calculation The mean anode layer thickness l was determined with MATLAB by averaging the cord length distribution from the membrane to the top of the reconstructed volume. The following modules of GeoDict were used for calculations: PoroDict for size distributions and surface area as described by Wiegmann and Glatt [20], FlowDict for permeabilities and ConductoDict for tortuosities. The mean ionomer thickness d was determined by calculating the shortest distance through the ionomer from pore each catalyst surface voxel to the open pore space d min or to the PTL PTL interface d min (see Fig. 3b). Taking the mean of all calculated distances yields d . For this calculation the reconstructed volume of 7 μm height was extended by mirroring to match the catalyst layer thickness. 2.4. Electrochemical measurements The polarization measurements of the FuelCellsEtc EZ CCM (based on a Nafion N115 membrane) were performed with a Scribner 857 Redox Flow potentiostat at atmospheric pressure conditions and 80 °C cell temperature. A 5 cm2 PEMWE electrolyzer cell was used with a 1 mm Mott sintered titanium PTL on anode side and a Freudenberg H23I2 PTL on cathode side. Anode and cathode deionized water feed was adjusted to 40 ml/min. The high frequency resistance (HFR) free voltage was obtained by measuring the HFR at each measurement point at 1 kHz.
2. Methods and experimental 2.1. PEMWE anode sample The investigated sample is the anode catalyst of the commercially available FuelCellsEtc EZ CCM with a loading of 3 mg cm−2 . The catalyst material is Ir0.5Ru0.5Ox and contains no supporting material, as determined by Energy-Dispersive X-ray Spectroscopy (EDX). In order to facilitate the segmentation of the microstructure reconstruction, the phase contrast between solid material and the pore space was enhanced by infiltrating the sample with silicone as proposed by Ender et al. [15]. Silicone yields a good contrast to the catalyst material. For good penetration of the pore space, a low viscosity silicone was used (Wacker Elastosil RT 604) following Liu et al. [16]. To further improve penetration, the sample was constantly kept under vacuum atmosphere, starting 1 h before infiltration until 30 min after. Excess silicone was removed and the sample was cured for 24 h at room temperature. In order to mitigate charging effects and sample drifting during SEM
3. Results 3.1. Microstructure of the catalyst layer From the catalyst layer a volume of 29 μm × 24 μm x 7 μm was reconstructed using FIB-SEM tomography. In order to enable discrimination between pores and the catalyst material, silicone was introduced into the pore space. However, the silicone has a low contrast to the ionomer. For this reason, the reconstruction contains only the 63
Journal of Power Sources 393 (2018) 62–66
F. Hegge et al.
Fig. 1. a) 3D-representation of the reconstructed volume with modeled ionomer (contents from 0 – 100% related to the pore volume). The z-direction is normal to the plane of the catalyst layer (through plane direction). b) Actual ionomer of the catalyst layer (highlighted blue in SEM cross section) resembles the modeled ionomer shown in a). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
unsupported IrRuOx catalyst structure and the pore space. Therefore, different ionomer contents were virtually added into the pore space as described in section 2.2. Fig. 1a shows the reconstructed catalyst layer comprising the IrRuOx structure and increasing ionomer contents, ranging from 0% to 100% with respect to the pore volume. The modeled ionomer structure is similar to that of the actual ionomer, as can be seen in a SEM cross section in Fig. 1b. The reconstructed volume contains catalyst particles with a median size of 0.5 μm and has a porosity of 55%. The virtual addition of ionomer affects the pore sizes and the sizes of the agglomerates of catalyst particles and ionomer. This can be seen in the porosity, in the pore size distribution and agglomerate size distribution as depicted in Fig. 2a and b and Table 1. The porosity decreases with the ionomer content due to the gradual substitution of pore space by the ionomer. The mean agglomerate size and the mean pore size increase due to the added ionomer, which fills small pores first according to its wetting behavior. A
Table 1 Morphological parameters of reconstructed volume. Ionomer content in pore space vol%
Overall ionomer fraction vol%
Porosity vol%
Median pore diameter μm
Median agglomerate diameter μm
0 25 48 73
0 14 26 40
55 41 29 15
0.53 0.71 0.94 1.39
0.50 0.89 1.74 4.08
reasonable amount of wet ionomer in the catalyst layer is approximately 50% of the pore space volume [21]. Due to the discrete nature of the ionomer modeling, a value of 48% was chosen as a reference, yielding a median agglomerate size of 1.74 μm, a median pore size of 0.94 μm and a porosity of εp = 0.29.
3.2. Species transport In the catalyst layer, reaction species have to be transported to and from the active sites. The microstructure has a strong impact on the transport of each reactant. In the following, each species transport mechanism is discussed based on calculated transport parameters. To estimate water and oxygen transport through the pore space, the permeability and the relative effective diffusivity were determined, as both contribute to the transport. The ionomer, as a possible parallel pathway, was not taken into account due to its very low permeability [22] in comparison to the permeability of the pore network. The calculated permeability only depends on the morphology and is valid for a single phase fluid transported through the pore space (Darcy's law). This can be used as input for two phase transport calculations of water and oxygen through the pore space. However, this requires an additional relation between capillary pressure and saturation [23]. As displayed in Fig. 3a, the through-plane permeability decreases with the ionomer content up to a calculated percolation threshold of approx. 90% ionomer content in the pore space. Above this value, the only pathway for water and oxygen is through the ionomer. Below this value, it is a combination of transport through pore space and ionomer. At the reference ionomer content of 48%, the permeability is 1.8⋅10−15 m2 = 1.8 mD . For PEMWE anode catalyst layers no values of the saturated permeability have been reported in literature so far. The calculated value lies between typical values of PTLs or cathode gas diffusion layers in an order of 10−11 m2 and values of cathode catalyst layers, as assumed for fuel cell cathodes in simulation models, in an order of 10−16 m2 [13] [24,25]. The produced oxygen has to pass through the ionomer from the catalyst surface to the pore space and then through the pore space to the porous transport layer, while the water pathway is in the opposite direction. In order to quantify the transport resistance of the ionomer, the mean ionomer film thickness d around the IrRuOx phase was estimated (Fig. 3b) as described in section 2.3. At a pore space ionomer content of
Fig. 2. Pore (a) and agglomerate size (b) distributions for increasing ionomer contents related to the pore volume. 64
Journal of Power Sources 393 (2018) 62–66
F. Hegge et al.
Fig. 3. Through-plane permeability (a) and mean ionomer film thickness (b) vs. ionomer content in the pore space x . Fitting the permeability based on percolation law [26] yields k = kmax ((xp − x ) xp−1) b for x < x p and k = 0 m² otherwise, with kmax = 4.74⋅10−15m2 , xp = 0.92 , b = 1.241.
Fig. 4. Tortuosity of the ionomer (a) and tortuosity of the pore space (b) vs. ionomer content in the pore space and overall phase fraction. Fitting tortuosities based on percolation law for the transport parameters [26] and definition 1 ((ε − ε p)(0.553 − ε p)−1) b for ε > ε p and of tortuosity (see above) yields τ = ε α −max τ = 0 otherwise, with α = α 0α −eff1 and α max = 0.31, ε p = 0.072 , b = 1.60 (a) and α max = 0.31, ε p = 0 , b = 3.05 (b).
100%, the mean ionomer film thickness is more than half the anode layer thickness. At the reference ionomer content of 48%, the ionomer film thickness (d= 0.4 μm) is significantly lower. To describe relative effective diffusivities and conductivities the tortuosity τ= α 0 α−eff1 ε is commonly used [27]. Here α 0 is the intrinsic transport value for the conducting phase (electric conductivity of the IrRuOx, water or oxygen diffusivity of the two phase mixture in the pore space or ionic conductivity of the ionomer), α eff is the effective transport value for the composite and ε is the fraction of the transporting phase. For the electron conduction a through-plane tortuosity of the IrRuOx material phase of τIrRuOx = 4.88 was calculated (phase fraction of εIrRuOx = 45 %). The other tortuosities depend on the virtually added ionomer content, as the fractions and the morphology of the transporting phases change. Proton conductivity is described by the tortuosity of the ionomer, as plotted against the ionomer content in the pore space in Fig. 4a. The decreasing ionomer tortuosity with growing phase fraction corresponds to an increasing proton conductivity. At an overall ionomer fraction of approx. 7 vol%, the tortuosity runs into a percolation threshold of the fitted curve. At the reference ionomer content of 48% of the pore space (26% overall ionomer phase fraction) the throughplane ionomer tortuosity is 3.5. The oxygen diffusion in the pore space is affected by the pore space tortuosity, which is plotted against the ionomer content in the pore space and the porosity in Fig. 4b. The pore space tortuosity increases with the ionomer content. At the reference ionomer content of 48% of the pore space (29% overall porosity), its through-plane tortuosity is 6.7 . It is to note that the pore tortuosity values at high ionomer contents are not expected to be representative. Due to very few open pores at
high ionomer content, the tortuosity values are discretized and do not follow the percolation law applied for the fitted curve [26]. For this reason, a precise percolation threshold value cannot be provided. Also for this reason, data points below 10% porosity were excluded from the fit. As depicted in Fig. 4a, the ionomer tortuosity decreases strongly up to an ionomer content of 40–50% of the pore space volume. At the same time there is still a comparably low pore space tortuosity and ionomer film thickness (Figs. 4b and 3b). Furthermore, the permeability is in the medium range at this ionomer content (Fig. 3a). Without considering the precise influence of the single transport parameters in a full cell model, an ionomer content of 40–50% of the pore space volume seems to be a good range for an optimal species transport. This is also in agreement with the optimal ionomer content from experimental studies [21]. 3.3. Catalyst surface area Besides its impact on transport, the microstructure of the catalyst layer determines the accessibility of the catalyst. This is typically quantified in the electrochemically active surface area (ECSA). Since the investigated catalyst layer contains no support material, we assume that the entire surface area of the catalyst material is active, hence equals the ECSA. With this assumption the ECSA can be estimated from the volume specific surface area obtained from the tomographic dataset. The ECSA is calculated with the volume specific surface area of 65
Journal of Power Sources 393 (2018) 62–66
F. Hegge et al.
commercially available state-of-the-art catalyst layer, offer new possibilities for the extension of electrochemical models of PEM water electrolyzers. This will provide deeper insight into the relation between the structure of an electrolyzer anode and the electrochemical performance of the whole cell. Acknowledgements We acknowledge funding by the German Federal Ministry of Education (BMBF) within the POWER-MEE project (Grant No. 03SF0536A). The authors further thank Andrew Mondon and Brian Shanahan for their support. References Fig. 5. Polarization and HFR free curve of FuelCellsEtc EZ CCM, measured at cell temperature of 80 °C and atmospheric pressure conditions (mean HFR = 140 mΩ cm2 ).
[1] International Energy Agency, Technology Roadmap Energy and GHG Reductions in the Chemical Industry via Catalytic Processes, (2013). [2] B. Bensmann, R. Hanke-Rauschenbach, I.K. Peña Arias, K. Sundmacher, Electrochim. Acta 110 (2013) 570–580. [3] L. Bertuccioli, A. Chan, D. Hart, F. Lehner, B. Madden, E. Standen, Study on Development of Water Electrolysis in the EU, (2014). [4] M. Carmo, D.L. Fritz, J. Mergel, D. Stolten, Int. J. Hydrogen Energy 38 (2013) 4901–4934. [5] R.S. Yeo, J. Orehotsky, W. Visscher, S. Srinivasan, J. Electrochem. Soc. 128 (1981) 1900–1904. [6] J. Ahn, R. Holze, J. Appl. Electrochem. 22 (1992) 1167–1174. [7] L. Wang, F. Song, G. Ozouf, D. Geiger, T. Morawietz, M. Handl, P. Gazdzicki, C. Beauger, U. Kaiser, R. Hiesgen, A.S. Gago, K.A. Friedrich, J. Mater. Chem. 5 (2017) 3172–3178. [8] N. Agmon, Chem. Phys. Lett. 244 (1995) 456–462. [9] R. García-Valverde, N. Espinosa, A. Urbina, Int. J. Hydrogen Energy 37 (2012) 1927–1938. [10] M.H. Shojaeefard, G.R. Molaeimanesh, M. Nazemian, M.R. Moqaddari, Int. J. Hydrogen Energy 41 (2016) 20276–20293. [11] M. Suermann, K. Takanohashi, A. Lamibrac, T.J. Schmidt, F.N. Büchi, J. Electrochem. Soc. 164 (2017) F973–F980. [12] F. Arbabi, A. Kalantarian, R. Abouatallah, R. Wang, J.S. Wallace, A. Bazylak, J. Power Sources 258 (2014) 142–149. [13] L. Zielke, A. Fallisch, N. Paust, R. Zengerle, S. Thiele, RSC Adv. 4 (2014) 58888–58894. [14] M. Cantoni, L. Holzer, MRS Bull. 39 (2014) 354–360. [15] M. Ender, J. Joos, T. Carraro, E. Ivers-Tiffée, Electrochem. Commun. 13 (2011) 166–168. [16] Z. Liu, Y.-c.K. Chen-Wiegart, J. Wang, S.A. Barnett, K.T. Faber, Microsc. Microanal. 22 (2016) 140–148. [17] Fiji, https://fiji.sc, accessed 18 April 2018. [18] GeoDict, https://www.geodict.com, accessed 18 April 2018. [19] A. Pfrang, D. Veyret, G. Tsotridis, Computation of Thermal Conductivity of Gas Diffusion Layers of PEM Fuel CellsConvection and Conduction Heat Transfer, (2011). [20] A. Wiegmann, E. Glatt, PoroDict, (2017). [21] M. Bernt, H.A. Gasteiger, J. Electrochem. Soc. 163 (2016) F3179–F3189. [22] Dawn M. Bernardi, Mark W. Verbrugge, J. Electrochem. Soc. 139 (1992). [23] B. Han, J. Mo, Z. Kang, F.-Y. Zhang, Electrochim. Acta 188 (2016) 317–326. [24] I.V. Zenyuk, P.K. Das, A.Z. Weber, J. Electrochem. Soc. 163 (2016) F691–F703. [25] X. Wang, T. van Nguyen, J. Electrochem. Soc. 155 (2008) B1085. [26] S. Kirkpatrick, Rev. Mod. Phys. 45 (1973) 574–588. [27] M. Doyle, J. Newman, A.S. Gozdz, C.N. Schmutz, J. Tarascon, J. Electrochem. Soc. 143 (1996) 1890–1903. [28] T. Audichon, E. Mayousse, S. Morisset, C. Morais, C. Comminges, T.W. Napporn, K.B. Kokoh, Int. J. Hydrogen Energy 39 (2014) 16785–16796. [29] J. Mo, Z. Kang, G. Yang, S.T. Retterer, D.A. Cullen, T.J. Toops, J.B. Green, F.Y. Zhang, Appl. Energy 177 (2016) 817–822. [30] A. Marshall, S. Sunde, M. Tsypkin, R. Tunold, Int. J. Hydrogen Energy 32 (2007) 2320–2324. [31] J. Cheng, H. Zhang, G. Chen, Y. Zhang, Electrochim. Acta 54 (2009) 6250–6256.
a= 3.1⋅106 m−1, the catalyst layer thickness of l= 10 μm and the catalyst loading of L cat = 3 mg⋅cm−2 as ECSA = a⋅l⋅L−cat1 = 1.03 m2g−1. This is considerably lower compared to values of IrRuOx catalysts determined by BET. Audichon et al. [28]. measured an ECSA of 47 m2g−1 for a Ir0.5Ru0.5Ox catalyst. However, they used IrRuOx nanoparticles with a particle size between 20 nm and 200 nm, which is much smaller than the 0.5 μm median particle size of our sample and is likely to be the main reason for the higher ECSA. This difference in particle size can maximally explain a factor of 25 assuming perfect spheres. Therefore, the rest of the difference could stem from a surface roughness below the FIB-SEM resolution and inaccuracies of the BET measurements of Audichon et al. Another indicator for a sub-resolution roughness is the decent electrochemical performance in the kinetic region (Fig. 5), which lies in the range of literature values [21,29–31]. The performance in the kinetic region is mainly governed by the ECSA of the anode and its catalyst activity [21]. Assuming a similar activity leads to the conclusion that the ECSA has to be comparable to the literature values and therefore higher than estimated from the reconstruction. 4. Conclusion We present the 3D microstructure of the anode catalyst of a PEM water electrolyzer obtained from a combination of FIB-SEM tomography and ionomer modeling. From the reconstructed volume (29 μm x 24 μm x 7 μm) size distributions, pore permeability, mean ionomer film thickness and phase tortuosities were calculated. The modeled ionomer phase allowed investigating the effect of ionomer content on these structural and transport parameters. This showed that at an ionomer content of 40–50% all important transport parameters are in a reasonable range confirming experimental results from literature. Further, the electrochemical active surface area (ECSA) was estimated from the reconstructed catalyst layer to be 1.0 m2g−1. This is considerably lower than experimental values presented in literature. One possible cause for this discrepancy is the limited resolution of the FIB-SEM tomography, which might substantially underestimate surface roughness of the catalyst particles. The presented microstructure parameters, which characterize a
66