Ion Chromatography and Combustion Ion Chromatography Analysis of Fuel Cell Effluent Water During Open Circuit Voltage

Sample fabrication

OCV testing

Ion chromatography (IC)

Two ion chromatograph methods were employed in this work. The first was routinely used to measure fluoride, sulfate, trifluoroacetate, and oxalate ions in the effluent water of each cell. These samples were analyzed without additional concentration using an ICS-2000 ion chromatography system (Dionex, Sunnyvale, CA). The second method was used to measure the total organic fluoride in a selected number of samples through combustion ion chromatography. Briefly, the total fluoride, which is comprised of both inorganic fluoride and total organic fluoride, was measured using a 3M Patented, Metrohm Inc. Combustion Ion Chromatography (CIC) custom set up. ³⁰ Initially the samples are passed through an in-line chloride removal module then combusted using a combustion chamber. The analytes are then separated by isocratic separation using sodium carbonate/sodium bicarbonate buffer and analyzed using a conductivity detector with a Metrohm IC system. The fluoride ion measured after combustion provides the total fluoride content in a sample. Inorganic fluoride is measured in parallel using a portion of the chloride free sample and a second Metrohm IC unit without combustion. The water soluble total organic fluoride (TOF) is obtained by subtracting measured inorganic fluoride from the total fluoride (TOF in a sample = TF_measured—IF_measured). In the instances where IF is present at significant levels compared to TOF, the TOF results are reported as <TOF-LOQ. The TOF-Limit of Quantitation (TOF-LOQ) is defined as the level of TOF that can be measured reliably above the level of IF in a sample. It is important to note that TOF-LOQ will depend on the amount of endogenous inorganic fluoride present in a sample. Based on the validation experiments conducted at 3M's Environmental Health and Safety (EHS) laboratory, TOF-LOQ is reported as <20%IF levels in the sample.

Since the same samples were evaluated for inorganic fluoride using traditional IC and CIC, the inorganic fluoride data was compared for consistency. Plotting the data from both methods against each other resulted in a high degree of correlation (R² = 0.99) with the CIC method resulting in slightly lower values of about 85% those of the traditional IC.

The ion release rate (IRR) was determined for fluoride (FRR), sulfate (SO₄RR), trifluoroacetate (TFARR), and oxalate (OARR) using the following formula where "I" represents the ion of interest, g_I·g_H2O ⁻¹ is the ion concentration, and g_H2O·h⁻¹ is the effluent water collection rate.

$$\begin{eqnarray}&&{\rm{IRR}}\left(\mu {g\cdot \mathrm{cm}}^{-{\rm{2}}}{\cdot d}^{-{\rm{1}}}\right)=\left(\mu {{\rm{g}}}_{{\rm{I}}}{{\cdot g}_{{\rm{H2O}}}}^{-{\rm{1}}}\right)\left({{\rm{g}}}_{{\rm{H2O}}}\cdot {{\rm{h}}}^{-{\rm{1}}}\right)\left({\rm{24}}\,{h\cdot d}^{-{\rm{1}}}\right)\left({{\rm{cm}}}^{-{\rm{2}}}\right)\end{eqnarray} \tag{ 1 }$$

A subscript is added when discussing data to designate traditional IC (IRR_IC) or CIC (IRR_CIC). The analyses in this study only include the data prior to the end-of-life to avoid complications that may arise due to direct crossover of gases after a membrane breach.

The OCV potential and ion release rate results of sixteen cells containing 3M 800EW ionomer membrane operated at OCV are shown in Fig. 1. All cell potentials and ion release rates are grouped by color for clarity. In addition to the expected inorganic fluoride and sulfate ions, several unidentified peaks are often observed in the ion chromatograms. These peaks are suspected to be small molecule organic ions associated with membrane degradation but typically cannot be identified due to a lack of known standards. Two notable exceptions are trifluoroacetate (TFA) and oxalate (OA) which are easily obtained allowing for quantification. Like the 3M Ionomer, eight Nafion™ NR211 membranes were run in the same way with the data shown in Fig. 2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The end of life for the OCV test is defined as the time at which the potential goes below 0.8 V. In most cases, this is immediately after an inflection in the OCV vs time curves and is a reliable indication of a breach in the membrane associated with end of life. Given the relatively large number of repeated tests for both membranes, the lifetime distribution can be fit using Minitab™ statistical software. Several distributions were evaluated for these data with normal, lognormal, and Weibull distributions all having high correlation coefficients. The best fit was the lognormal distribution with correlations of 0.987 and 0.984 for the 3M and NR211 samples, respectively. Figure 3 shows the overlaying histograms for each ionomer along with the distribution curves. The lifetime mean, median, and standard deviation are summarized in Table I. These distributions are determined to have different means as judged by a two sample T-Test P-value of 0.008, with the 3M Ionomer membranes lasting about 75 h longer than the NR211 counterparts. It should be noted that the lifetime is potentially influenced by the somewhat thicker 30 μm 3M Ionomer membrane compared to the 25 μm Nafion NR211 sample.

Zoom In Zoom Out Reset image size

The ion release rates for the quantifiable ions are summarized in Table II. Effluent water is collected for about one hour at intervals of one to three days for each cell. Only the data from samples prior to end of life are included in the averages. Interestingly, the NR211 samples showed statistically significant lower FRRs compared to the longer-lasting 3M ionomer samples (P-value 0.00). The sulfate release and oxalate release rates are very low and effectively the same for both membranes. In the sulfate case, trace amounts are often detected in periodic blanks, however, the OCV samples were typically about 10 times higher than those values (see Table SI (available online at stacks.iop.org/JES/169/034526/mmedia)). The trifluoroacetate ions are notably higher for the NR211 membranes owing to fragments that can be traced back to the -CF₃ group in the side chain. ³¹

Next, selected samples of effluent water were analyzed for total organic fluoride (TOF) using combustion ion chromatography. This technique is based on the difference between the inorganic fluoride (F⁻) routinely measured by IC and the total fluoride of a combusted sample. Since the fluoride analysis was replicated as part of the TOF measurement, both values are reported. Figure 4 shows these data along with the TOF (open squares) for three representative 3M ionomer membranes. The inorganic FRR of both methods agree very well with the values from the traditional IC being slightly higher than those measured as using the CIC method.

Zoom In Zoom Out Reset image size

Similarly, two Nafion NR211 samples were selected for the same analysis (Fig. 5). As with the 3M samples, there is good agreement between both IC methods with the IC method being slightly higher than the CIC method.

Zoom In Zoom Out Reset image size

Open circuit voltage testing has been a longtime standard in accelerated life testing of fuel cell membranes. Coupled with fluoride release rates, this method has been an invaluable tool in developing modern membranes comprising mechanical support and radical scavenging additives. This study, however, was conducted without the use of support or additives to allow for maximum membrane degradation rates. The choice of catalyst also has an impact on membrane degradation, ³² for this reason simple platinum on carbon electrodes were chosen rather than alloys that are typically used in advance cathodes. These factors have the potential to alter the relative rates of membrane degradation reactions and caution is advised when generalizing the results of this study to other materials sets or operating conditions.

In general, there are four proposed membrane degradation pathways: (1) polymer chain end unzipping, (2) peroxide attack at the sulfonic acid side chain, (3) abstraction of the backbone tertiary fluorine, and (4) radical attack at the sidechain ether linkage. Since the products from the last two pathways are expected to be the same, and there is more experimental evidence suggesting t-fluorine abstraction occurs in a fuel cell, ^6,27,33 the subsequent discussions will refer to the t-fluorine path while recognizing cleavage of the ether linkage is also a possibility. Details of these mechanisms for PFSA membranes are outlined in many excellent papers and reviews. ^6–10,12 These reactions have been demonstrated in various controlled conditions: missing is an assessment of the relative importance of each pathway in operating fuel cells. For this discussion, the specific reactions steps will not be explicitly written out but rather only the expected products at the conclusion of the process.

Perhaps the most frequently invoked degradation pathway is polymer chain end unzipping. This mechanism was proposed by Curtin and later refined by Coms and others. ^6,17,18 It is also a well-documented reaction for the oxidative degradation of small molecule perfluorocarboxylic acids. ^34–36 Schemes 1a and 1b show the expected outcome of this backbone unzipping reaction for 3M Ionomer and NR211 membranes. The number of tetrafluoroethylene backbone repeat units is defined by the average equivalent weight of 800 for the 3M Ionomer (n = 4.2) and 1100 g mol⁻¹ for NR211 (n = 6.6). For the 3M Ionomer, in the absence of unzipping of the side chain fragment (or any other reaction), one would expect 21.8 inorganic fluoride ions and 6 organic fluorines (Scheme 1a). Likewise, the NR211 membrane would be expected to liberate 31.4 fluoride ions and 8 organic fluorines from the full sidechain (Scheme 1b). For this reaction, the ratio of organic to inorganic fluorine atoms is 0.275 for the 3M ionomer and 0.255 for NR211. Note that this ratio is the same on both mole and mass bases.

Zoom In Zoom Out Reset image size

Another proposed degradation mechanism is the attack of hydroxyl radicals (HO·) or other species on the sulfonate or sulfonic acid. ^6,9,37–39 Once the carbon sulfur bond of the sulfonate group is cleaved, the remaining unstable radical intermediate is expected to undergo a series of reactions to ultimately form a carboxylic acid. Assuming this carboxylic acid is less stable than the original sulfonic acid, it should proceed to unzip up the sidechain, cleave the backbone, and continue with the unzipping mechanism. In essence, the entire polymer would degrade to simple inorganic ions with no fluorinated organic compounds. Alternatively, backbone unzipping, followed by continued unzipping of the side chain fragment would also result in only simple ions. For the 3M Ionomer, one sulfate ion would accompany 27.8 fluoride ions resulting in a mole ratio of 0.037 or mass ratio of 0.182 (Scheme 2a). In the NR211 case, the mole ratio is 0.024 and mass ratio 0.128 (Scheme 2b).

Zoom In Zoom Out Reset image size

The last mechanisms of interest are abstraction of a tertiary fluorine (t-fluorine) by a hydrogen atom (H·) ^6,33 or cleavage of a carbon-oxygen ether bond from peroxide attack. ^9,40 In either case, the fragments would be expected to be identical. In this scenario, only the side chain would be liberated, and new chain ends formed due to backbone cleavage. Inorganic fluoride ions accompany this process even in the absence of the traditional unzipping mechanism. For 3M Ionomer shown in Scheme 3a, 6 organic fluorines are expected and 5 inorganic fluoride ions resulting in a 1.20 ratio. For NR211, there are two similar t-fluorine/ether groups, one on the backbone and one on the side chain. If only the backbone location reacted as shown in Scheme 3b, 8 organic fluorines and 5 fluoride ions are expected (1.60 ratio). In the most extreme case (Scheme 3c), both groups would react to give 5 organic fluorines and 8 fluoride ions (0.625 ratio). This same result would also be expected for degradation initiating at the side chain t-fluorine since the reactive intermediate after the side chain is cleaved should undergo subsequent reactions that lead to backbone cleavage even if the backbone t-fluorine or oxygen do not participate in the initiation of this process.

Zoom In Zoom Out Reset image size

Throughout this discussion, the reaction products are represented as carboxylic acids, however, the partially hydrogenated counterparts of these fragments have been reported where the –CO₂H group is a reduced to a –CH₃ group and may be the predominant species in some reactions. ^24,27,28 Nevertheless, the same organic to inorganic fluoride ratios are expected for Schemes 2 and 3 where further unzipping via the carboxylic acid is not considered. The organic fluorine, fluoride, and sulfate ions expected for each scheme along with the mole or mass ratios are summarized in Table III.

Before discussing the experimental data, it should be noted that this analysis is predicated on the assumption that all ions and fragments exit the cell in the relative amounts in which they were generated. It is possible that some more volatile fragments may be lost to evaporation, or certain ions or fragments may be preferentially retained in the membrane or other components of the MEA. The conclusions of this analysis must, therefore, be viewed with the understanding that the collection efficiency for each ion or fragment was not quantitatively determined.

Considering the sulfate ion data first, a plot of the sulfate release rate vs fluoride ion release rate is shown in Fig. 6. Accompanying each plot is a line representing the expected mass ratio of sulfate to fluoride for the case where the ionomer degrades completely to simple ions depicted in Scheme 2. The correlation between the measured sulfate and fluoride release rates are rather poor with R² values of 0.25 for 3M and 0.03 for NR211. This could be a consequence of the very low sulfate concentrations in the effluent water where quantification is difficult. Even so, the amount of sulfate measured is significantly lower than expected for reactions that initiate at the sulfonic acid site on the side chain. These data suggest that the reaction outlined in Scheme 2 is effectively nonexistent or occurs at a much slower rate than other reactions at OCV.

Zoom In Zoom Out Reset image size

The relationship between organic fluoride and fluoride ions offers more insight into the degradation process. The fact that organic fluoride is detected suggests that Scheme 2 is not the sole degradation pathway. Figure 7 shows the total organic fluoride vs fluoride release rates for both membranes along with the predicted ratios for Schemes 1 and 3 in dashed lines. In the 3M Ionomer case, the experimental data slope of 0.267 nearly overlaps with the expected ratio of 0.275 for Scheme 1a. It is difficult to determine if the slight disagreement between these values is due to experimental error or a genuine difference. If real, one explanation could be a small degree of continued degradation such as carboxylate unzipping of fragments after the side chain is liberated from the backbone. While Scheme 3a may be occurring at some slow rate, the implications of these data are that backbone unzipping as outlined in Scheme 1a is the dominant degradation mechanism. Moving to the NR211 membranes, a similar analysis shows a slope of 0.312 for the organic to inorganic fluoride compared to the expected ratio of 0.255 for the unzipping reaction from Scheme 1b. Again, these values are quite close, but the difference in slope is larger than the 3M Ionomer and deviates to the higher side of the ratio. In other words, there is more organic fluoride than can be explained by the unzipping reaction in Scheme 1b alone. At the same time, the organic fluoride ratios for Schemes 3b and 3c are substantially higher than the experimental data suggesting that, while not the dominant source of organic fluoride, they are likely occurring to some degree. Taking this analysis one step further, the excess fluoride ratio from NR211 in Fig. 7 suggests about 24% more organic fluoride than can be accounted for through unzipping alone. One or both NR211-specific Scheme 3 mechanisms could explain the additional organic fluoride detected beyond the side chain fragments from unzipping only.

Zoom In Zoom Out Reset image size

The NR211 effluent water offers additional analysis possibilities through the trifluoroacetate (TFA) release rate. Figure 8 shows the organic fluoride attributed to the measured amount of TFA. In other words, this plot does not represent the actual TFA release rate but rather the organic fluoride attributed to the TFA allowing for a consistent basis of comparison to previous data. Like the total organic fluoride, there is a reasonable correlation between TFA release rates and FRR (R² = 0.71 and 0.87 for 3M and NR211). The most likely source of TFA is from the reaction outlined in Scheme 3c where the t-fluorine bond on the sidechain is broken leading to both TFA and, presumably, the two-carbon carboxylate/sulfonate fragment (HO₂CCF₂SO₃H). It is also conceivable that the TFA is released during the unzipping of a larger NR211 carboxylate terminated sidechain fragment; however, the continued oxidation of TFA to CO₂ and fluoride would also be expected in such a scenario. Perhaps TFA can escape the fuel cell environment prior to oxidation that other, less mobile, carboxylates experience, but that cannot be determined from these data. Comparing the slope of the NR211 line to the expected fluoride ratio from Scheme 3b, one can see the experimental data is far below the predicted organic fluoride for this reaction alone. Yet, unlike the sulfate release rates, these data contribute a meaningful amount of organic fluoride to the effluent water. Furthermore, for every TFA that is quantified, the companion two-carbon carboxylate/sulfonate fragment can be assumed to be present but not directly measured. Adjusting the slope from Fig. 8 (right) by 5/3 to account for all the organic fluoride shown in Scheme 3b, the more likely organic to inorganic fluoride ratio is 0.077. This represents approximately 12% of the fluoride ratio possible from Scheme 3b. These data, therefore, support the conclusion from the previous analysis that more reactions are occurring than backbone unzipping alone.

Zoom In Zoom Out Reset image size

The level of TFA detected in the 3M Ionomer samples is substantially less than that of NR211 as shown in Fig. 8 (left). Interestingly, despite the lack of -CF₃ functionality in the 3M ionomer, there is still a moderate correlation between the overall fluoride release rate and the TFA values (R² = 0.71). The origin of TFA in these samples is unclear and will be the subject of future investigations.

The conclusion from the fluoride, sulfate, TFA, and organic fluoride data in total suggests that the dominant degradation mechanism for both the 3M ionomer and NR211 is backbone unzipping that liberates small molecule side chains (Scheme 1). Cleavage of the C-S bond of the sulfate group, while above baseline values, does not appear to be an important mechanism under the conditions of this experiment. Abstraction of the t-fluorine (Scheme 3) is likely occurring in these ionomers especially in the case of NR211 where more organic fluorine is detected than can be explained by unzipping only. Interestingly, the slope of the organic fluoride to inorganic fluoride line in the 3M case might suggest that t-fluorine abstraction is not occurring. This seems unlikely given the similarities between both ionomers. Perhaps the rate of t-fluorine abstraction is sufficiently slow that the fluoride ion from unzipping dominates the signal. This would be anticipated in the case where the 3M ionomer has more carboxylate end groups at beginning of life and the formation of additional end groups through Scheme 3 is not large enough to alter the organic to inorganic ratio. The somewhat lower fluoride release rate of the NR211 samples early in the test that increases over time may suggest a lower level of end groups at the beginning of life for this membrane. Due to the different ionomer manufacturers, it is reasonable to assume the original number of carboxylate end groups is unlikely to be the same between membranes.

Like the t-fluorine reaction process (Scheme 3), there is no reason to expect the unzipping reaction (Scheme 1) is fundamentally different between each ionomer structure. Nevertheless, even though Scheme 3 appears to be secondary to unzipping, the ramifications could prove more significant due to the formation of new carboxylate end groups. In the absence of reactions that break the backbone, unzipping should happen at a relatively slow rate defined by the original number of carboxylate chain ends. Since modern ionomers are treated to convert the chain ends to stable –CF₃ groups, the numbers of available carboxylates are low at the beginning of life. Backbone cleavage reactions, however, generate new carboxylate chain ends that can subsequently unzip. A relatively small number of chain cleavage events can have a large effect on overall membrane degradation rates. ⁴¹

So far, this analysis has only considered cases where the entire backbone degrades (Schemes 1 and 2) and cases where the only water-soluble fragments are attributed to side chain fragments (Schemes 1 and 3). However, a previous study identified perfluoro dicarboxylic acid fragments (HO₂CC_nF_2nCO₂H) that originate from the polymer backbone. ^27,28 Therefore, it is possible that some portion of the TOF is due to these fragments and not solely to the cleaved side chains. Even so, the framework of degradation schemes represents boundaries for the relative amount of organic and inorganic fluoride from each process and offers useful insights to the overall importance of each scheme.

Another issue to address is the apparent coexistence of hydroxyl radicals (HO·) and hydrogen radicals (H·) that are responsible for ionomer degradation. Clearly, if these two species where in proximity to each other they would preferentially combine to form water rather than react with the ionomer. One explanation for this apparent paradox might be the existence of different reactive species in different locations of the membrane or electrodes. Such a hypothesis is supported by reports of localized degradation in the membrane. ³³ Perhaps this could be a focus of future studies.

Given the measured fluoride and organic fluoride release rates, an inventory of the membrane lost at end of life can be calculated using Eq. 2. The lifetime of the cell is designated in hours (h), and ratio of the molecular mass of the parent ionomer fragment (frag.) to the fluoride ion (F) mass is g_frag·g_F ⁻¹.

$$\begin{eqnarray}&&\begin{array}{l}{\rm{Membrane}}\,{\rm{Loss}}\left( \% \right)=\left(\mu {{\rm{g}}}_{{\rm{F}}}{\cdot \mathrm{cm}}^{-{\rm{2}}}{\cdot d}^{-{\rm{1}}}\right)\\ \,\times \,\left({\rm{h}}\right)\left({\rm{24}}\,{h\cdot d}^{-{\rm{1}}}\right)\left({{\rm{g}}}_{{\rm{frag}}}{{\cdot g}_{{\rm{F}}}}^{-{\rm{1}}}\right){\left({\rm{6000}}\,\mu {{\rm{g}}}_{{\rm{mem}}}{\cdot \mathrm{cm}}^{-{\rm{2}}}\right)}^{-{\rm{1}}}\end{array}\end{eqnarray} \tag{ 2 }$$

In the case of the fluoride release rate, it is assumed that most of the F⁻ ions are coming from the –(CF₂)– backbone with a g_frag·g_F ⁻¹ ratio of 1.32. These assumptions are surely overly simple as fluoride ions are released in the absence of –(CF₂)– unzipping and do not account for side chain/backbone juncture points. For the total organic fluoride, only side chain fragments were considered with g_fraf·g_F ⁻¹ values of 2.30 for the 3M Ionomer and 2.16 for NR211. Again, this may be an oversimplification as some fragments may undergo degradation after cleavage from the backbone. Nevertheless, this analysis provides an approximate inventory of the membrane lost in this experiment. Individual lifetimes, FRR, and percent membrane loss calculated from traditional IC are shown in Table SII for the 3M Ionomer and S4 for NR211 membranes. Similarly, membrane loss calculated from FRR and TOF from CIC are shown in Tables IIIS and VS for each membrane. The averages of each data set are summarized in Table IV.

In this experiment, at least 12.6% of the 3M Ionomer mass is lost as determined by the FRR using either IC or CIC method. However, the more complete assessment including the TOF from the CIC adds another 6.6% mass loss attributed to water soluble organic fragments for a total mass loss of around 19%. A similar situation exists for the NR211 samples where fluoride ions account for 5.5%–7% of the membrane mass and organic fragments another 3.5% for a total estimated mass loss of at least 10%. It is unclear why the 3M Ionomer loses more membrane mass yet lasts longer than the NR211 membranes. This may be an artifact of the slightly thicker 3M Ionomer membranes (30 vs 25 μm) or may be influenced by mechanical properties of the membranes. It must be emphasized that these data are from the cathode effluent water only. It is reasonable to assume fluoride ions or water soluble fluorinated organic fluoride will diffuse within the membrane and exit in both the cathode and anode streams. ^42,43 Therefore, the values calculated almost certainly underestimate the true membrane loss, which could be up to twice these values if fluoride and fragments exit either side of the cell at the same rate.

Yet to be discussed is the detection of oxalate ions in the effluent water of both the 3M Ionomer and NR211 samples (see Table II). Even though oxalate is detected, its role in degradation reactions is unclear. The levels of this ion are also extremely low and show no correlation to FRR (R² = 0.017 and 0.003 for 3M and NR211). One possibility was suggested by Yu where a hydroxyl radical (HO·) attacks the sulfonate sidechain end to remove HSO₄ ⁻ leaving a carbon centered radical that further eliminates tetrafluoroethylene oxide (Scheme 4). ¹³ While not articulated by Yu, this oxide is expected to hydrolyze to the perfluoro di-alcohol which would rapidly hydrolyze to oxalic acid (Scheme 5). ³⁴ While an intriguing mechanism, more work needs to be done before this is confirmed or another source of oxalate discovered.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The lack of correlation of the sulfate and oxalate ions to the fluoride release rate remains an unresolved question. These ions are detected at very low levels (<0.05 ∼ 0.5 μg ml⁻¹) but are still above the deionized water blanks. The lack of correlation to fluoride release rates might suggest that whatever the mechanism, it is slow compared to primary degradation reactions.

Finally, there may be a desire to capture membrane decomposition products from fuel cells and other electrochemical cells. Strategies including filters with ion exchange resins or high surface area adsorbents such as activated carbon could be employed in commercial systems. Additionally, separating the liquid water from water vapor may aid in the efficient removal of water-soluble membrane fragments.

A series of 3M Ionomer and NR211 membranes without radical scavenging additives or mechanical support were evaluated for durability using the DOE OCV test. Lifetimes were fit to a lognormal distribution as is typical for durability testing, with 3M Ionomer membranes having longer lifetimes despite higher fluoride release rates. Ion chromatography and combustion ion chromatography analysis of cathode effluent water was used to infer relative significance of three proposed degradation schemes. Sulfate ion release rates were used to rule out meaningful degradation originating at the terminal end of the sulfonic acid side chain. Total organic fluoride methods were used to determine the amount of fluorine containing, water soluble, membrane fragments. The ratio of organic fluorine to inorganic fluoride is consistent with backbone unzipping as the primary degradation pathway. The NR211 membranes, however, have a higher ratio implying degradation at the backbone or side chain tertiary fluorine to be an important secondary degradation pathway. This conclusion is supported by the detection of trifluoro acetate in the NR211 membrane effluent water. Interestingly, low levels of oxalate are detected in both the 3M Ionomer and NR211 samples. The levels of this ion have no correlation to the fluoride, or other ion, release rates and no conclusions can be drawn from these data yet.

The combination of fluoride and organic fluorine values were used to estimate membrane mass loss for both types of membrane. The 3M Ionomer membranes lose at least 20% of their mass while NR211 is closer to 10%. These values represent inventories based on cathode water only and therefore are likely to be underestimates as degradation products are also apt to escape through the anode effluent stream.

Lastly, it is well known that the conditions used in this study are especially harsh favoring rapid membrane degradation. Cells under normal operating conditions have orders of magnitude lower degradation rates resulting in very low levels of fluoride ion release and organic fluorine compounds. ^21,24,44 The techniques described in this work can be applied to these non-accelerated operating conditions to better capture normal use degradation mechanisms and validate the OCV or other accelerated methods.

This work was supported by Department of Energy grant DE-EE000924.