Analysis of the cyclic voltammograms

A pronounced hysteresis effect of the current-voltage characteristic is predicted by the modeling results and observed in the potential sweep experiment. In the following section, an analysis of this hysteresis effect is given, based on the dy-namic fuel cell model. Fig. 4.13 shows the fuel cell sandwich, which consists of the gas diffusion layers, the catalyst layers, and the membrane. In the gas dif-fusion layers and the catalyst layers, the volume saturation of liquid water s_l is indicated by solid lines. In the membrane, the mean value of the fraction of ex-panded channels S_mean is indicated. The numbers refer to Fig. 4.6, where specific points in the current voltage characteristic are marked.

At the beginning of the sweep experiment (1), the distribution of the liquid water is almost uniform across the cell. In the membrane, the fraction of expanded chan-nels is small at the beginning of the first cycle (1). The highest current density is reached at a cell voltage of 0.4 V (2). According to the results of the model, the liquid water saturation is still low at this point. While an increase in the liquid water saturation is observed at the cathode, the anode side shows a small decrease of s_l. This is due to the influence of the electro-osmotic drag, which is high in the high current density regime. The fraction of expanded channels in the membrane is still small at point (2). In contrast to the electro-osmotic drag, the accumulation of liquid water shows a delay with respect to the time evolution of the current density. This is indicated by the relatively strong increase in the liquid saturation between the highest current density (2) and the end of the first forward sweep (3).

The saturation of liquid water shows a further increase during the first backward sweep, where the relative maximum is reached between 0.4 V and 0.5 V cell volt-age (4). This leads to the observed hysteresis behavior. The absolute maximum liquid water saturation is seen during the backward sweep of the second cycle (5).

Figure 4.12: Comparison of two different assumptions regarding the generation of the product water at the cathode. In the base case the water is assumed to be generated as water vapor. case B shows the simulated IV-curves if the water is assumed to be generated in the liquid phase. case B shows an increased mass transfer resistance compared to the base case.

Anode

CL CL

5

1,3

∂Ω₃

∂Ω₆

∂Ω₁ ∂Ω₄

Figure 4.13: This graph shows the volume saturation of liquid water slin the gas diffusion layers (GDL), and the catalyst layers (CL). In the membrane, the average fraction of expanded channels Smean is shown. During the experiment, the liquid water saturation and the fraction of expanded channels in the membrane increase. On the anode side, the saturation of liquid water remains low due to the influence of the electro-osmotic drag of water from the anode to the cathode. The numbers refer to Fig. 4.6 and have the following meaning: (1) start of first cycle, (2) maximum current density, (3) minimum cell voltage, (4) maximum liquid water saturation during first cycle, (5) maximum of liquid water saturation during second and subsequent cycles. Normalized coordinates corresponding to Fig. 4.1 are used.

The time-span between the highest current density and the maximum value of the liquid water saturation in the cathode is between 70 s and 80 s. The liquid water accumulation on the anode side is generally weaker than on the cathode side due to the electro-osmotic drag. The fraction of expanded channels S in the mem-brane influences the water transport through the memmem-brane in accordance with Eq. (4.34). The liquid water saturation at the interfaces between the catalyst lay-ers and the membrane,∂Ω3and∂Ω4, and the fraction of expanded channels S are coupled. The higher the liquid water saturation, the smaller is the critical radius in Eq. (4.40). This is the smallest possible radius of pores that are expanded. Cor-respondingly, the fraction of expanded channels, and, hence, the amount of water that is driven through the membrane by a pressure gradient, increases during the experiment. This is shown in Fig. 4.13 in the membrane subdomain at the points (3),(4), and (5).

In Fig. 4.14, the normalized charge generation rate Q^C_norm=

¯¯

¯Q_c/a,C/Q^mean_c/a

¯¯

¯ is shown. This allows one to identify those regions of the catalyst layers that are most active during the course of the experiment. The influence of the liquid water accumulation on the electrochemical reactions can be seen. Fig. 4.14 shows the catalyst layers of the anode side and the cathode side that are separated by the membrane. The numbers refer to different points of the current-voltage charac-teristic that is shown in Fig. 4.6. On the anode side, the distribution of the charge generation rate is close to stationary during the sweep experiment. The catalyst layer of the anode side is most active close to the membrane, where the difference between the the electronic potential φe and the protonic potentialφp is highest.

There is no mass transfer limitation for hydrogen. On the cathode side, the be-havior is different. At the beginning of the experiment (1) and during the first forward sweep (2), the charge generation is distributed more uniformly over the catalyst layer than on the anode side. This is due to the slower reaction kinetics of the oxygen reduction reaction compared to the hydrogen oxidation reaction.

When liquid water accumulates on the cathode side, the distribution of the charge generation changes dramatically. During the first backward sweep between 0.1 V and 0.9 V (3,4) and during the second cycle (5), the catalyst layer is only active close to the gas diffusion layer of the cathode side. The reason is the strong mass transport limitation for oxygen on the cathode side due to the accumulation of liq-uid water. A comparison with the behavior of a self-breathing fuel cell highlights the importance of efficient water-management. Fig. 3.18 shows that the catalyst layer is most active close to the membrane in case of a partially hydrated ionomer.

The utilization of the platinum catalyst particles is most efficient if the ionomer is fully hydrated. However, flooding of the catalyst layer must be avoided.

Fig. 4.15 compares measured cyclic voltammograms with different sweep rates.

The base case with a sweep rate of 10 mV/s used in Fig. 4.10 is compared to a

Catalyst layer Anode

2

4

Figure 4.14: The normalized charge generation rate Q^C_norm is shown on a logarithmic scale. Q^C_normis defined by

¯¯

¯Q_c/a,C/Q^mean_c/a

¯¯

¯, where Q^mean_c/a is the average charge generation rate in the catalyst layer of the cathode and the anode, respectively. The results show the influence of the liquid water saturation on the activity of the cathode. Normalized coordinates corresponding to Fig. 4.1 are used.

Figure 4.15: Cyclic voltammograms with different sweep rates are compared. The hys-teresis effect is found in both cases. The limiting current density is smaller in case of the smaller sweep rate.

measurement with a sweep rate of 5 mV/s. While hysteresis is observed in both cases, the mass transport resistance in the high current density region is larger in case of the lower sweep rate. This is consistent with the explanation of the hysteresis due to the accumulation of liquid water at the cathode. The modeling results show the same trend. Fig. 4.16 compares simulation and measurement for two different sweep rates.