A PEM Electrolysis Cell for In Operando NMR and MRI Investigations of MEA Degradation

EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17244135 A1118 Page 1/9 A1118 A PEM Electrolysis Cell for In Operando NMR and MRI Investigations of MEA Degradation Michael Schatz* (1), Sven Jovanovic (1), Julian Borowec (1,2), Rüdiger-A. Eichel (1,2,3), Florian Hausen (1,2), Josef Granwehr (1,4) (1) Forschungszentrum Jülich GmbH, IET-1, Jülich/Germany; (2) RWTH Aachen University, IPC, Aachen/Germany; (3) RWTH Aachen University, Faculty of Mechanical Engineering, Aachen/Germany; (4) RWTH Aachen University, ITMC, Aachen/Germany; *Contact corresponding authors: www.EFCF.com/ContactRequest Abstract Proton exchange membrane (PEM) electrolysis is a promising process for sustainable hydrogen production, but its commercialization is delayed by high costs and elusive degradation of membrane electrode assemblies (MEAs) [1]. In operando Nuclear Magnetic Resonance (NMR) and Magnetic Resonance Imaging (MRI) offer the potential to investigate degradation mechanisms during electrolysis, and thus, provide highly relevant insights for enhanced performance [2,3]. In a first part of this contribution, a custom-designed miniature PEM electrolysis cell is presented, fitting the spatial constraints of a 1H coil of a commercially available imaging probe. In contrast to tailor-made probes [2,3], this approach allows for a broader range of NMR experiments – including not only 1H spectroscopy and T1 and T2 relaxometry, but also the first MRI and diffusion measurements on operating PEM electrolysis cells. The key design feature was a sealing concept without screws, utilizing O-rings in combination with precise compression geometry. Uniform electrical contacting minimizing metal content in the NMR-sensitive volume was validated via microelectrode voltage mapping. The inlet water temperature was controlled between 60 and 80 °C using a non-magnetic heat tube. The functionality of the newly developed NMR cell is demonstrated by electrochemical and NMR experiments in the second part of the contribution. The 1H signal-to-noise ratio and resolution allowed chemical shift analysis, while T1/T2 contrast enabled differentiation between MEA and water signals. MRI revealed water and gas bubble distribution during operation. Impedance spectroscopy and cyclic voltammetry results were consistent with labscale PEM electrolysis. This novel in operando NMR cell provides an effective method for investigating degradation phenomena during long-term PEM electrolysis experiments, leveraging the wide variety of experiments available with commercial probes. References [1] Q. Feng et al., Journal of Power Sources 366, 33 (2017) [2] A. S. Cattaneo et al., Energy & Environmental Science 8, 2383 (2015) [3] C. Mrad et al., Journal of Membrane Science 688, 122111 (2023) EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17244135 A1118 Page 2/9 Introduction The production of hydrogen through electrolysis has long been hindered by the high cost of electricity. However, with the increasing availability of renewable energy sources, electricity prices are expected to decrease, making electrolysis a more economically viable option for hydrogen production. Moreover, electrolysis can help mitigate the intermittency of renewable energy sources by providing a solution for energy storage [1-3]. Proton Exchange Membrane (PEM) electrolysis is a promising technology for hydrogen production, offering the advantage of flexible operation at varying current loads. Despite its potential, the commercialization of PEM electrolysis is slowed down by the high costs associated with the degradation of membrane electrode assemblies (MEAs) [4]. Previous studies have demonstrated post-test Nuclear Magnetic Resonance (NMR) investigations to be capable of identifying degradation phenomena. NMR signals of water and functional groups in the membrane are highly sensitive to their chemical environment, providing valuable insights into degradation mechanisms [5,6]. However, in operando NMR investigations, which can provide information on these degradation processes, are scarce. Most existing studies demonstrated customized probes tailored to specific geometries of PEM electrolysis cells, primarily focusing on water content in the membrane and water distribution within the cell [7-12]. In contrast, this study presents a novel approach, where the electrochemical cell is designed to fit a standard Magnetic Resonance Imaging (MRI) probe, as inspired by our previous studies [13-15]. This design enables the use of a broader range of standard NMR and MRI experiments, including 1H spectroscopy, T1 and T2 relaxometry, MRI, and diffusion measurements. By leveraging the capabilities of commercial MRI probes, this study aims to provide a new method for investigating degradation mechanisms of MEAs in PEM electrolysis. 1. Scientific Approach The design of the in operando NMR cell was tailored to the Bruker MIC-WB imaging probe, which features a 25 mm 1H coil that defines the radial spatial constraint. The probe's sensitive volume extended approximately 30 mm in the axial direction. Tubing and electrical connectors were positioned outside of this region. To facilitate easier access and minimize the risk of probe damage, all connections were implemented from the top of the probe. The primary challenge in designing the cell was finding a balance between maximizing the active MEA area and achieving a leak-proof sealing and uniform electrical contacting. Conventional lab-scale electrolysers typically employ teflon (PTFE) sheets pressed together by screws for sealing, with electrical contacting implemented via stainless steel housing and a porous conductive structure (porous transport layer; PTL) that is perfused by water and in contact with the active MEA layer. However, these approaches are not feasible for an in operando NMR cell due to the limited construction space and the need to minimize conductive material content in the NMR active region [16]. To address these challenges, a novel sealing concept was developed that eliminates the need for screws. Instead, a precise compression mechanism is used to apply force on O-rings, ensuring a reliable seal. The cell design is depicted in Figure 1. Two half cells, cf. Figure 1a, with grooves for O-rings, a flow field, and bores for water inlet and outlet are placed on either side of a MEA. This assembly is then pressed into a cell holder, cf. Figure 1b, that is manufactured from polylactic acid (PLA) filament using a Fused Deposition Modeling (FDM) 3D printer (Prusa i3 MK2). The half cells are designed to be mechanically manufacturable from polyether ether ketone (PEEK), while rapid iteration steps were facilitated by using a stereolithography (SLA) resin printer (Stratasys Connex 350) for prototyping. EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17244135 A1118 Page 3/9 For electrical contacting, a Cu wire with 0.24 mm diameter was placed in an additional groove in the half cell, surrounding the flow field. The uniformity of the contacting was evaluated by mapping the potentials across the electrode using microelectrodes [17]. Figure 1c shows the tubing for water inlet and outlet, as well as coaxial SMA connectors for contacting the anode and cathode. The assembled cell can be inserted into the probe, cf. Figure 1d, connected to tubing and wires, and then inserted into the magnet. Temperature control is achieved using a heating tube positioned along approximately 2 m of the inlet tubing, just prior to the cell inlet, enabling precise regulation of the cell's operating conditions. Figure 1: CAD drawing of a) 3D-printed half cell with flow field and grooves for O-ring (red) and contacting wire (green); b) assembled cell, consisting of two half cells, inserted into holder. Photo of c) cell assembly with tubing and electrical connectors; d) cell inserted in NMR probe. 2. Experiments The experimental setup for in operando NMR experiments is illustrated in Figure 2. A peristaltic pump was employed to circulate deionized (DI) water at room temperature from a reservoir located approximately 2 m below the upper opening of the magnet bore. The water was transported through two separate 1/16” perfluoroalkoxy alkane (PFA) tubes, one supplying the anode and the other the cathode sides of the PEM electrolysis cell. The tubes were routed into the magnet from above, requiring the water to be pumped against a height difference of approximately 2 m. Additional tubes for the water outlet lead out of the top of the magnet, allowing for continuous circulation of water during operation. To control the temperature of the water entering the cell, the inlet tubes were inserted into a self-regulating heating hose over the last 2 m before reaching the cell inlet. The heating hose maintained a temperature of 90 °C, ensuring that the water was preheated to a consistent temperature before entering the cell. This setup enabled precise control over the operating conditions of the PEM electrolysis cell, allowing for in-depth investigations of the degradation mechanisms under various conditions. EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17244135 A1118 Page 4/9 Figure 2: Experimental setup of in operando NMR experiments including tubes for inlet and outlet, as well as electrical wiring with filters, shielding and grounding. To ensure the cell's leak-proofness and test the inlet temperature, the electrolysis was initially performed outside the magnet. Any leakage of water could be easily detected through an opening at the bottom of the cell holder, allowing for prompt identification of any issues. Additionally, the open design of the probe initially designed for small animal experiments prevents major damage in the case of small leakage during flow cell operation. The temperature at the inlet was measured using a miniature temperature sensor (Honeywell HEL-705-U-0-12-00) installed into a three-way tube fitting, providing the temperature before cell entry. For in operando measurements, the cell was inserted into the probe, cf. Figure 1d. The probe was then carefully positioned underneath the magnet, and the tubing and cables were inserted through the top of the magnet and connected to the cell. At this point, the cell's leak-proofness was reassessed. The probe was then inserted into the magnet, while a second person carefully pulled the tubes and cables from the top to minimize forces acting on probe and cell, without pulling the cell out of the probe. The electrochemical performance of the PEM electrolysis cell was characterized using electrochemical impedance spectroscopy (EIS), cyclic voltammetry (CV), and chronoamperometry (CA) at both room temperature and elevated temperature to assess the functionality of temperature control. To evaluate the uniformity of the electrical contacting, potential mapping was performed using microelectrodes. The cell was modified to include a window that allows access to a microelectrode robot. A hydrated MEA sample (Ion Power HYDRion) was inserted between a regular half cell (cf. Figure 1a) and the modified half cell with a window (cf. Figure 3a). The half cell without window was filled with DI water to keep the MEA hydrated during the experiment. The anode side of the MEA was positioned towards the opening of the window. A bipotentiostat was used to apply a constant potential of 1.8 V across the MEA using the wire contacting, while the potential was measured between the counter electrode and the microelectrode at various positions on the working electrode. Two different types of contacting were tested: a wire completely surrounding the active area, and contacting only from the top side of the cell, cf. Figure 3b and 3c, respectively. The potential was mapped along two lines in the two dimensions of the active layer plane, cf. Figure 3a, providing a comprehensive understanding of the electrical contacting uniformity. EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17244135 A1118 Page 5/9 Figure 3: a) Adjusted cell design for potential mapping. One half cell includes a window through which the MEA is accessible by microelectrodes. The direction of xand yaxis are defined here; half cell with b) surrounding contacting and c) one-sided contacting; the indicated wire segments were connected at the backside of the half cell, forming a single continuous wire path routed to the top of the cell. d) Sketch of microelectrode potential measurement; the arrow indicates the microelectrode tip that can be positioned freely within the window of the anode. The suitability of NMR measurements for distinguishing between the different components in the PEM electrolysis cell was evaluated by assessing whether they provided sufficient contrast. This is a crucial requirement for in-depth analysis of the cell's behaviour and degradation mechanisms. Firstly, 1H 90° pulse experiments were performed to distinguish between the MEA signal, background signals, and water signal on the chemical shift axis. Next, T1 measurements were carried out using saturation recovery and inversion recovery sequences [18], as well as the Carr-Purcell-Meiboom-Gill (CPMG) sequence [19] for T2 evaluation. Magnetic Resonance Imaging (MRI) measurements were performed using the FLASH pulse sequence [20] to visualize the water distribution within the cell. Finally, the first in operando experiments were conducted to test the long-term leak-proofness of the cell and evaluate if NMR signals were compromised by external noise or interference. 3. Results The potential mapping experiments are presented in Figure 4. A comparison of the two contacting methods revealed significant differences in the potential distribution across the MEA active layer, even though the same potential of 1.8 V was applied. Along the x-axis, the surrounding contacting method ensured a more stable potential, with values ranging from 0.65 V to 0.7 V, whereas the one-sided contacting method resulted in a significantly lower potential that decreased with increasing distance to the contacting wires from 0.13 V to 0.08 V. Along the y-axis, the potential decreased towards the edges in the surrounding contact case, with a range of 0.5 V to 0.68 V and stayed constant at ca. 0.085 V in the onesided contacting case. Notably, the potential distribution in both cases suggests that the stress on the MEA would not be uniform, as the potential can locally be lower than the activation potential. This is also expressed in the comparison of CVs for the two scenarios in Figure 4c, where the onset of OER is at lower potential and overall activity is increased for the surrounding contacting in comparison with one-sided contacting. In the case of one-sided contacting, the potential was found to be significantly lower across the entire active area, indicating that no electrolysis would occur. For the surrounding contacting case, the applied potential of 1.8 V is at the EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17244135 A1118 Page 6/9 edge of OER activation. The measured local potential differences of ca. 150 mV imply that only parts of the cell actively participated in the electrolysis process. Furthermore, the results for surrounding contacting indicate that the stress would be highest in the middle of the active area, compared to the edges. For in operando measurements, two key objectives were identified: 1) ensuring a potential that is high enough to activate the entire active area, and 2) minimizing the potential difference between middle and edges of the cell. In our case, the potential difference along the y-axis on the order of 100 mV was considered low, but its implications should be taken into account when evaluating long-term measurements and local degradation patterns. The potential difference along x-axis could however be neglected for the surrounding contact scenario. Given the superior performance of the surrounding contacting method, only this approach was considered for further investigation. Figure 4: Potential measured vs. tip position of microelectrode along a) x-axis and b) yaxis for the cases of surrounding (blue) and one-sided contact (green). c) Cyclic voltammogram for surrounding (blue) and one-sided contact (green). The results of EIS and CV showed comparable performance to lab-scale electrolysers [1,21,22], though higher cell potential had to be applied, indicating that the cell design and construction did not compromise its electrochemical behaviour, but increased the cell resistance. In CV experiments, 43 mA/cm² was measured for a cell potential of 4 V. CA experiments at lower applied potential of 1.8 V showed an activity of approximately 0.68 mA/cm². A comparison of the cell performance at room temperature versus elevated temperature of 60 °C showed a significant increase in activity for the temperature-regulated experiments, cf. Figure 5. This enhancement in activity at elevated temperature is consistent with expectations [21,22] and confirms the functionality of temperature regulation. EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17244135 A1118 Page 7/9 Figure 5: a) CV and b) EIS measurements on the cell operated with water at 20°C (blue) and at elevated temperature of 60°C (green). A comparison of the 1H NMR signals of the various components is presented in Figure 6. The background signal, which consists of signals from probe, cell housing and holder, was significant, resulting in a broad signal, but the MEA signal could be differentiated from the rest due to its increased chemical shift by ca. 5 ppm relative to the background resonance, as well as a linewidth that is approximately ten times narrower with ca. 400 Hz. However, upon addition of water, the spectral resolution is compromised, making it challenging to distinguish between the different components solely based on their chemical shifts. As an additional source of contrast, T1 and T2 may be utilized to differentiate between the signal component in the presence of water. As a reference, the T1 and T2 relaxation times of the MEA signal were measured in a sample without background from cell housing and holder, i.e. in a 25 mm NMR tube. The resulting values were 0.01 s and 0.005 s for T1 and T2, respectively. In contrast, the T1 values of the background components were found to be in the range of 1-2 s, providing good contrast and ensuring separability of the signals. However, the T2 values of the background components were in the same range as the MEA signals, albeit slightly lower on the order of 0.001 s. In the assembled cell without water, the T1 values of the MEA could be easily distinguished from the background using Inverse Laplace Transform (ILT) evaluation [23], cf. Figure 6b. The T2 values of the MEA and background could also be distinguished using ILT, cf. Figure 6c, although some components could not be assigned to a specific cell component or species in the MEA. Upon addition of water to the flow field of the cell, the strong signal of water introduced new challenges to the ILT evaluation, such as a dominant signal overlapping with a much weaker signal, or exchange between the different water reservoirs. Mitigation strategies allowing a separation of the different contributions are currently under development. Figure 6: a) 1H NMR spectrum of full cell with hydrated MEA. Background signal can be distinguished from MEA signal. b) 1H NMR spectrum of background components. The probe background signal is present in all spectra. O-ring, cell housing and PLA holder were measured in a 25 mm NMR tube. Inverse Laplace Transform of b) T1 and c) T2 measurements, representing chemical shift resolved T1 and T2 relaxation times. MRI images of the water distribution within the cell, as presented in Figure 7, reveal the presence of gas bubbles and incomplete water filling of the cell. Both sagittal and axial crosssectional views exhibit consistent spatial patterns, confirming the reliability of the observed features. EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17244135 A1118 Page 8/9 Figure 7: a) One longitudinal and b) – c) two axial MRI sections through the PEM-NMR cell representing water distribution and gas bubble position; d) position of the two axial sectional views relative to the longitudinal section. In operando experiments could be conducted without issues related to external noise, and the electrochemical operation remained stable under constant potential. During a 2 h measurement, no significant changes in the T1 and T2 relaxation times were observed. These findings suggest stable measurement conditions; however, long-term testing will be necessary to assess potential temporal changes and degradation effects. 4. Conclusion The sealing and temperature control strategy of the space-confined PEM electrolysis cell was validated through ex situ testing. The iterative prototyping process ultimately provided a leak-proof PEM electrolysis cell design. Additionally, the temperature control system was found to be effective in maintaining a stable temperature of 60 °C at the inlet, when the temperature in the heating tube was controlled to 90 °C. Electrical contact via a surrounding wire was demonstrated to be effective; however, local potential variations across the MEA must be considered when evaluating long-term experiments. It is essential to apply a cell potential that ensures uniform electrochemical activity across the entire active area. The potential mapping method showed capable of determining the required applied cell potential. Potential interference in microelectrode measurements due to the presence of liquid water on the MEA surface was identified. To mitigate this, the use of an electrically isolating tip is recommended. The observed contrast in chemical shift, as well as in T1 and T2 relaxation times, appears sufficient for distinguishing relevant signals. Nonetheless, suppression of background signals, particularly the dominant water signal, remains a challenge. Approaches such as background subtraction or signal filtering by different pulse programs should be considered. In future in operando MRI measurements, the formation and dynamics of gas bubbles could be observable. This could enable correlating bubble formation with degradation phenomena occurring within the cell. Acknowledgements The authors gratefully acknowledge the financial support by the German Federal Ministry of Education and Research (BMBF) within the H2Giga project DERIEL (grant number 03HY122C). The authors thank Mike Haferkorn for 3D printing of the prototype cell, as well as Tobias Offermanns and Christoph König for assisting in cell design and manufacturing. EFCF 2025: Low-Temp. Fuel Cells, Electrolysers & H2 Processing 1 – 4 July 2025, Lucerne Switzerland https://doi.org/10.5281/zenodo.17244135 A1118 Page 9/9 References [1] Carmo, M. et al., Int. J. Hydrogen Energy 38, 4901 (2013) [2] Karl, A. et al., Electrochem. Sci. Adv. 5, e202400041 (2025) [3] Javed, A. et al., Int. J. Hydrogen Energy 98, 280 (2025) [4] Q. Feng et al., J. Power Sources 366, 33 (2017) [5] Ghassemzadeh, L. et al., J. Am. Chem. Soc. 135, 8181 (2013) [6] Tang, H. et al., J. Power Sources 170, 85 (2007) [7] A. S. Cattaneo et al., Energy Environ. Sci. 8, 2383 (2015) [8] C. Mrad et al., J. Membr. Sci. 688, 122111 (2023) [9] Luo, R. et al., J. Magn. Reson. 361, 107666 (2024) [10] Tsushima, S. et al., J. Electrochem. Soc. 157, B1814 (2010) [11] Wang, M. et al., J. Power Sources 195, 7316 (2010) [12] Zhang, Z. et al., J. Magn. Reson. 193, 259 (2008) [13] Schatz, M. et al., J. Phys. Chem. C 127, 18986 (2023) [14] Jovanovic, S. et al., Commun. Chem. 6, 268 (2023) [15] Niemöller, A. et al., J. Magn. Reson. 269, 157 (2016) [16] Schatz, M. et al., Magn. Reson. 5, 167 (2024) [17] Kishimoto, T. et al., ACS Omega 7 , 39437 (2022) [18] Li, Z. et al., Magn. Reson. Chem. 54, 1041 (2016) [19] Meiboom, S. et al., Rev. Sci. Instrum. 29, 688 (1958) [20] Haase, A. et al., J. Magn. Reson. 67, 258 (1986) [21] Noor Azam, A. et al., Polymers 15, (2023) [22] Maier, M. et al., Int. J. Hydrogen Energy 47, 30 (2022) [23] Granwehr, J. et al., J. Chem. Theory Comput. 8, 3473 (2012) Keywords: EFCF2025, H2, LowTemp. Fuel Cells & Electrolysers, PEM, MEA degradation, NMR, MRI, Operando Remark: This work is licensed under Creative Commons Attribution 4.0 International