Reversible performance loss induced by sequential failed cold start of PEM fuel cells

Abstract

This study correlates the post start cell performance and impedance with the cold start process in the subzero environment. The sequential failed cold starts are deliberately conducted as well as the start at small current density. Here the failed cold start means the cell voltage drops to or below zero within very short time during the start process. It is found that there are reversible performance losses for the sequential failed cold starts, while not obvious degradation and no recovery happen for the start at small current density. Using the thin film and agglomerate model, it is confirmed that this is due to the water blocking effect. Comparing the results from different start processes, a model with respect to the shifting of reactive region within the catalyst layer is applied to explain that the reversible performance loss is associated with the amount of the generated water or ice and the water location or distribution during cold start. The relationship of the cold start performance at high current density and the pore volume in the catalyst layer is also discussed.

Introduction

Cold start of proton exchange membrane (PEM) fuel cells at subzero temperatures is a more than engineering subject. Recent experimental works and mathematical simulations have given insights into the durability issues related with the freeze/thaw (F/T) cycles and the transient electrochemical behavior during the cold start process [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16], [17], [18].

Gas purge after fuel cells shut down in subfreezing environment is widely accepted as an incontrovertible precondition for the successful cold start [19], [20], [21], [22], and also improves the freeze durability of materials and components [3], [4], [5], [6], [7]. As a matter of course, the freeze degradation of the key materials and components should be concerned under the subzero storage scenario. Although Nafion^® membrane itself shows no catastrophic damage at the experimental condition of either almost dry state for hundreds of F/T cycles [23] or fully hydrated state for several F/T cycles [1], the most pronounced effect seems from others such as the physical prick from the ice in the catalyst layer (CL) [9]. Also, water redistribution in the membrane or draining out of the membrane during the cooling process was observed and could decrease the membrane degradation [9], [19]. This dynamic phenomenon is mainly caused by the temperature gradient from the membrane to the endplate and the free space in the gas chamber. As for the electrode, there exist two representative and apparent opposite views about freeze degradation. For gas diffusion electrode (GDE), it was found evident performance degradation after several F/T cycles [1]. The performance of the fuel cell with another structure electrode, catalyst coated membrane (CCM) was reported no loss even after 100 F/T cycles from −40 °C [24]. The two contrary results come from the difference in the pore structure and hydrophobic net of electrode structures, and it has been confirmed that PTFE plays an important role in lowering the residual water content in the CL [25]. Although the membrane electrode assembly (MEA) is porous and has capacity of storing water, volume change due to the water and ice phase transition can still damage the structure and therefore cause performance degradation. As a result, it is necessary to understand the relationship of water amount with cell freeze degradation. Using electrochemical impedance, the cell electrochemical behavior after F/T cycles has been correlated with the residual water amount [26]. Based on the thin film and agglomerates model, the freezing of water in the pores among the agglomerates compresses the pores in agglomerates which decreases the electrochemical active area (ECA) and increases the oxygen diffusion in the agglomerates. The more the water resides in the cell the more significant this effect is. Furthermore, a thin water film covers the agglomerates and leads to the thin film diffusion effect when the residual water amount is high. The former induces the irreversible performance loss and the latter causes the reversible performance loss. This result was confirmed by the subsequent research [27]. The more serious situation for the MEA is the delamination which was also frequently reported [4], [9]. It is most likely that the water draining out of the PEM, ionomer in the CL and pores in the CL collects together and forms the ice lens in subfreezing temperature [9], [13]. Therefore, the CL with high Nafion^® or PTFE gradient especially the composite CL with the hydrophilic layer close to the PEM and the hydrophobic layer close to the diffusion layer (DL) should be easy to have such MEA delamination during F/T cycles. On the other hand, the ionic resistance profile of the ionomer across the whole CL is an indicator of the CL structure change. It was found that the ionic resistance close to the interface of the membrane and the catalyst layer decreased, while that away from the interface increased [28]. These results may further support the hypothesis of ice lens within the CL. As for the DL after freeze, the hydrophobicity on DL surface at the cathode side decreased [2], [10], and the change of the pore structure influenced the cell performance at high current density [3], [29]. The ex situ experiments showed that the stiff DL was better to prevent the MEA from the physical damage [9].

Cold start at subfreezing temperatures is a complex transient polarization process coupled with water freezing. From ohmic polarization point of view, the PEM with high conductivity especially at subzero temperatures is needed. However, the proton conductivity strongly depends on the water uptake of the membrane, and the water freezing in the membrane decreases the conductivity [30]. So, the water state, in detail, freezing water and nonfreezing water in the membrane and their contribution to the proton conductivity is crucial. Several kinds of aromatic based PEMs together with Nafion^® were investigated with respect to subzero conductivity and water state [31]. As with the kinetic polarization, the oxygen reduction reaction (ORR) shows intrinsic temperature dependence [32], [33]. From the activation energies in Ref. [33], it can be estimated that the exchange current density even at high Tafel region decreases less than ten to the power of 0.8 from 20 °C to −20 °C. Nevertheless, the ECA decrease ascribed to the freezing of generated water makes the apparent exchange current density decrease. Furthermore, the ice covers the catalyst sites, blocks the reactant gas access, causes very high mass transport polarization and finally the electrochemical reaction stops. This is most likely the reason that the fuel cell is hard to start at subzero temperatures. Generally, the cold start can be clarified into two processes. One is the non-isothermal start which is usual practical situation with the fuel cell stack. If the valid precondition and operational parameters are applied, the stack can self-start successfully at −10 °C without performance loss [34]. The other is the isothermal start which is often used to investigate the fundamentals in this process. Early mathematical modeling work based on the hypothesis that when the ECA was completely covered by the generated water the reaction is over [35]. This work gave accurate prediction that the minimum limit temperature at which the single cell could successfully start was −5 °C. Later, Wang’s group gave the detail description and deep understanding about the single cell cold start behavior [6], [7], [12], [14]. They defined more conditions such as the generated water in the CL firstly existed in gas form; when the saturated pressure was reached the water became ice; part of the generated water was taken out by the reactant gas; the heat released when the water became ice and etc. They considered that the MEA including the membrane, the ionomer and the pores in CL had the capacity of storing water which is most particular feature in their modeling work. The results indicated that the initial state of the membrane was an important parameter for the cold start process. The dry membrane could absorb the generated water in the CL, prolong the time the CL reaches the saturated state, and thus favor the cold start process. However, the start current density is smaller than 50 mA cm⁻² in the experimental and modeling work. When the start current density is high, the start process ends quickly before the CL is saturated which is not like what the model predicted. Based on the above consideration, the ohmic polarization can be excluded from the main contribution to this phenomenon. Consequently, it seems that the decreased ECA and increased mass transport polarization are the reasons. The distribution of the generated water or the reactive interface in the MEA during cold start process should be critical to clarify this problem.

In this study, to change the reactive interface, the sequential failed cold start was applied to the fuel cells as well as the start at very small start current density. The effects of these two different processes including sequential failed cold start and the start at very small current density on the fuel cell electrochemical response were investigated. The water blocking effect was found only for the sequential failed start. In particular, the corresponding reactive interface and the validity of gas purge were discussed.