Experimental characterization and numerical simulation of differential PEM fuel cells

Abstract

To accomplish a global market entry with polymer electrolyte membrane fuel cells, several challenges have to be tackled. One of them is finding the right compromise between functionality, lifetime and costs. For this purpose, understanding the influence of design, materials and operating conditions on the fuel cell behavior is essential. In this work, the focus is set on the operating conditions. The first aim is to address the lack of consistent datasets for electrochemical parameters of state-of-the-art materials in the literature by using in situ characterization techniques. The second aim is to propose a new, simple but complete parameterized performance model. When a fuel cell is operated, several loss mechanisms can be observed. Close attention is paid to the oxygen reduction reaction (ORR) since it represents the highest contribution, even though other mechanisms are also characterized in order to deconvolute the performance signatures. In order to ensure high data quality and minimize unwanted in-plane effects, single cells under differential conditions are used. The base of this work consists of a comprehensive dataset containing polarization, electrochemical impedance spectroscopy (EIS), limiting current and voltammetry data obtained from systematic parameter variations in H2/O2, H2/N2 and H2/H2 configurations. In the first publication, the hydrogen crossover and the protonic resistances of both the cathode catalyst layer (CCL) and the membrane are characterized. First, the equivalency of cyclic voltammetry (CV) and linear sweep voltammetry (LSV) to determine the hydrogen crossover is demonstrated and validated with an online gas analysis. Then, an H2/N2 measurement campaign with a full factorial variation of the relative humidity RH and the temperature T is carried out. Based on the CV data, a model for the hydrogen permeation coefficient is parameterized and with the EIS spectra the parameters of a specific transmission line model (TLM) for blocking electrodes are fitted. Especially the membrane and cathodic ionomer resistances R_PEM and R_p are thereby fitted as functions of RH and T and then used to parameterize models of the ionomer conductivities. Although the values that are measured are comparable to values reported in the literature, distinct deviations and trends can be observed which highlights the need for specific characterization of new materials. Finally, the change of the ionomer conductivity caused by modified water management under load (H2/O2) is investigated and an approach to estimate the local CCL relative humidity is discussed. The deviations of the ionomer resistances under load from those in the H2/N2 state reveal that a correction of polarization data for ohmic contributions based on H2/N2 data is not recommended. In the second publication, proton pump (H2/H2) measurements provide a quantification of the anode loss contributions which are strongly dependent on RH and the hydrogen partial pressure p_H2. Within these O2-free experiments, carbon monoxide (CO) poisoning on the catalysts from trace impurities in the gas feed is observed. Thus, a new recovery procedure to counter CO poisoning is presented which allows for a reliable parameterization of the hydrogen oxidation reaction (HOR) based on linearized Butler-Volmer (BV) kinetics. Then, a dataset containing polarization curves for a full factorial variation of T, RH and the oxygen partial pressure p_O2 is created and corrected by the ionomer resistances and hydrogen crossover. This dataset is used to parameterize the ORR reaction based on a Tafel law and allows to prove that RH has no significant influence on the ORR performance. Subsequently, O2 mass transport contributions are discussed based on limiting current techniques and heliox measurements and a global loss analysis is performed. The latter shows that up to 0.1 A/cm2, anode and oxygen transport contributions do not interfere with the parameterization of the ORR. Moreover, the analysis reveals that the simple Tafel law with one intrinsic slope of -70 mV/dec is sufficient to capture the ORR kinetics at half-cell potentials above 0.8 V. Below 0.8 V, the data deviates from the Tafel line. Furthermore, the EIS spectra under load show trends that are not covered by the Tafel law even at very small current densities. Deviations between Tafel slopes obtained from polarization data and EIS charge transfer resistances hint at more complex kinetics, which is investigated further in the third publication. In the third publication, the low-frequency inductive loop in EIS and its responsibility for the discrepancies in the Tafel slopes between polarization curves (-70 mV/dec) and the charge transfer resistance (<= -100 mV/dec) is demonstrated. Based on the H2/O2 test run, the low-frequency contributions are deciphered by subtracting the local slopes of the polarization curves (calculated by numerical derivation) from the low-frequency real-axis-intercepts of the capacitive EIS that occur at approximately 1 Hz. Based on this knowledge and on the ionomer resistances under load, the slow humidification effects leading to inductivities can be separated from the platinum oxidation contributions. Therewith, a new simple representation of the cathode kinetics containing the ORR and platinum poisoning effects that result in a low-frequency inductive loop is parameterized. This reaction kinetics is integrated into a transient 1D through-plane membrane electrode assembly (MEA) model which is used for an extensive parameter study. This model yields consistent results and both polarization curves and EIS match the experiments well at high humidity conditions, thus reconciling steady-state and dynamic performance signatures of PEMFCs. In Chapter 5, the link between the publications and their significance in the literature context are discussed. Additionally, the MEA model is extended by ionomer hydration dynamics. The parameterizations of R_PEM and R_p depending on RH and the current density j is obtained by fitting polynomials to selected data of the H2/O2 test run. This final MEA model is now able to simulate all humidity conditions meaningfully and even slightly enhances the predictions at high RH compared to the model from the third publication. Finally, an extensive loss analysis containing both overpotentials and differential resistances is carried out and depicts the most important loss mechanisms in state-of-the-art PEMFCs depending on the operating parameters.