Membrane Screening for Iron–Chrome Redox Flow Batteries

Since the electrolyte in an iron chrome redox flow battery (ICRFB) is inexpensive, the cost of the separator can contribute up to 38% of the CapEx cost of an ICRFB. Since the membrane also influences the RFB performance, it is the aim of this study to screen various commercial ion‐exchange membranes (IEMs) and a microporous separator (MPS) in an ICRFB to identify possible alternative membranes to the currently used Nafion. The suitability of six cation (CEMs) and two anion exchange membranes (AEMs), as well as one MPS, is investigated. No discharge curves are attained with either of the AEMs, which probably result from the formation of an anionic FeCl4− species at the elevated operating temperatures (65 °C) used, confirming literature on the unsuitability of AEMs for ICRFBs. Similarly, although the MPS is stable in the ICRFB electrolyte, it yields a high capacity decay ascribed to excessive crossover. Whereas all six CEMs yield similar CE, VE, and EE values, the fumatech FS‐950 yields a comparable capacity decay but higher EE and capacity discharge than the currently used Nafion 117 counterpart. Due to the significant cost reduction, modified or customized MPS should be further evaluated.


Introduction
[3] For example, Zeng et al. showed in 2015 that the use of the perfluorinated Nafion 212 membrane as separator contributed to 19% of the CAPEX for the VRFB and 38% for the ICRFB. [1]While a fluorinated membrane is required for the VRFB due to the presence of the highly oxidative VO 2 þ species, the ICRFB operates in a less-corrosive environment creating the impetus to consider the use of nonfluorinated and possibly less expensive membranes. [4]e three groups of membranes that could theoretically be used in ICRFBs include nonporous ion exchange membranes (IEMs) such as cation or anion exchange membranes (CEMs and AEMs) and microporous separators (MPSs). [3]n general, CEMs deliver higher voltage efficiency (VEs) due to the high mobility of protons improving conductivity, while AEMs deliver higher current efficiencies (CEs) due to an improved selectivity due to the Donnan exclusion principle. [5]Consequently, a trade-off exists between selectivity and conductivity.
During the NASA redox storage system development project, the initial focus was placed on identifying AEMs that could deliver a high selectivity as well as a low resistivity. [6]However, the shift toward higher operational temperatures, to help push the equilibrium toward the electrochemically active chromium-species, resulted in a significant decrease in AEM selectivity. [6,7]This in turn leads to the use of mixed electrolytes, which removed the requirement for highly selective membranes.This then led to a shift toward the use of CEMs due to their ease of fabrication and lower cost. [6,7]Since then, the CEMs that have been most widely used in aqueous acidic flow batteries are Nafion derivatives.
10][11] Cho et al., [8] for example, compared the VRFB performance of a commercial Nafion212 (CEM) to a novel AEM consisting of a quaternary ammonium-functionalized poly (pentafluorostyrene) blended with a F6-PBI polymer. [8]Not only were they able to show that the CE of the VRFB could be improved from 95.5% (Nafion 212) to %100% when using the novel AEM (containing 40% F6-PBI for improved mechanical stability), but more importantly, no capacity decay occurred over the first 100 cycles when using the novel AEM versus a capacity loss of 90% after 92 cycles when using Nafion 212, which was ascribed to a significant crossover of active species between the positive and negative electrolyte. [8]n recent years there have also been a few studies that focused on the investigation of alternative CEMs for ICRFBs. [9,12,13]hen investigating the effect of CEM thickness (178-50 μm) using a series of commercial available Nafion membranes, Sun and Zhang found that even though thicker Nafion membranes (such as Nafion 117-178 μm) were able to reduce active specie crossover, thinner CEMs, such as Nafion 212 (50 μm), were Since the electrolyte in an iron chrome redox flow battery (ICRFB) is inexpensive, the cost of the separator can contribute up to 38% of the CapEx cost of an ICRFB.Since the membrane also influences the RFB performance, it is the aim of this study to screen various commercial ion-exchange membranes (IEMs) and a microporous separator (MPS) in an ICRFB to identify possible alternative membranes to the currently used Nafion.The suitability of six cation (CEMs) and two anion exchange membranes (AEMs), as well as one MPS, is investigated.No discharge curves are attained with either of the AEMs, which probably result from the formation of an anionic FeCl 4 À species at the elevated operating temperatures (65 °C) used, confirming literature on the unsuitability of AEMs for ICRFBs.Similarly, although the MPS is stable in the ICRFB electrolyte, it yields a high capacity decay ascribed to excessive crossover.Whereas all six CEMs yield similar CE, VE, and EE values, the fumatech FS-950 yields a comparable capacity decay but higher EE and capacity discharge than the currently used Nafion 117 counterpart.Due to the significant cost reduction, modified or customized MPS should be further evaluated.
preferable for optimal ICRFRB performance due to higher energy efficiencies (EEs). [12]When investigating different types of membranes, Sun et al. showed that it is possible to replace the expensive Nafion membranes with a more cost-effective sulfonated poly(ether ether ketone))-based CEM, which would allow for a significant cost reduction while increasing both the CE and, consequently, the EE of an ICRFB. [9,13,14]part from CEMs and AEMs, another possible separator to consider would be a MPS. [3]The obvious advantage associated with MPSs is the cost reduction (IEMs generally cost around 500 $ m À2 , while the cost of MPSs are around 2 $ m À2 ), while the obvious disadvantage is the lack of a physical barrier which could result in higher crossover and, consequently, lead to increased self-discharge and capacity loss. [5,15,16]However, while the identification of alternative CEMs for the ICRFB has received some attention in recent years, little research has been done on identifying alternative AEMs and MPSs for ICRFBs.In a recent review, novel aluminum-containing zeolite membrane were mentioned as an alternative separator that yielded high proton transport selectivities. [17]rom the literature it is thus evident that both performance improvements and/or cost reductions are possible using alternative or less expensive membrane options.Consequently, it was the aim of this study to screen various commercial CEMs, AEMs, and one MPS to identify possible alternatives to the expensive Nafion.The commercial membrane samples were subjected to ten charge/discharge cycles from which important performance indicators such as CE, VE, EE, Cap Dis , capacity decay rate, and ASR were calculated and used to evaluate membrane performance.
A potentiostat/galvanonstat (Gamry 5000 E) together with a generic lab-scale RFB (NWU instrument makers) was used to perform the galvanostatic charge/discharge cycles.The generic lab-scale flowthrough single RFB cell with an active area of 28 cm 2 was used.The cell consisted of the applicable membrane separator sandwiched between two cell frames (Teflon), two carbon felt electrodes (GFA6 EA, SIGRACELL), two bipolar plates (TF6, SIGRACELL), and two copper current collectors; all held together by two aluminum endplates.Charge/discharge cycles were performed at a constant current density of 40 mA cm À2 between 1.25 and 0.75 V at (65 AE 2.5 °C).A volume of 45 mL of electrolyte was cycled through each half cell at a flow rate of 50 mL min À1 using a dual-channel peristaltic pump (Watson-Marlow 323S).

Membranes
A list of the various commercial membranes/separators that were tested in this study with their measured thicknesses, as well as known properties, is given in Table 1.

Scanning Electron Microscopy and Electron Diffraction Spectroscopy
Scanning electron microscopy (SEM) was performed using a FEI Quanta 250 FEG with ESEM capabilities to investigate the active surface of membrane samples (before and after cycling) where necessary.In addition, energy-dispersive X-ray spectroscopy (EDS) analysis with the Oxford system using INCA software, coupled with the SEM, was used to provide an elemental analysis of membrane samples where elemental analysis was deemed necessary (before and after cycling).Before SEM, the samples were coated with an ultrathin coating of an electrically conducting gold/palladium alloy, using an EM Scope.

Self-Discharge
Where deemed necessary, self-discharge tests were performed.Initially, an electrolyte consisting of 1.3 M FeCl 2 .4H 2 O (99.0%, Sigma-Aldrich), 1.4 M CrCl 3 .6H 2 O (98.0%, Sigma-Aldrich), 5 mM Bi 2 O 3 (99.9%,Aldrich), and 1.0 M HCl (32%, Labchem) was charged to 50% state of charge (SOC).The charge cycle was followed by an open-circuit voltage period (i.e., no current is applied and the voltage was monitored as a function of time) until a voltage of 0.75 V was achieved.Charge was performed at a constant current density of 40 mA cm À2 and electrolyte volume of 45 mL of electrolyte was used at a flow rate of 50 mL min À1 using a dual-channel peristaltic pump (Watson-Marlow 323S), as discussed in Section 4.2.1.

Results and Discussion
Since the aim of this feasibility study was to screen various membranes, only ten charge and discharge cycles were used as this provided adequate information to distinguish between these membranes.From the results, it is evident that there was virtually no difference between the first charge/discharge cycle of the F-1050, FS-940, and FS-950, which was virtually identical Nafion 212.Also evident from Figure 1 is that the two thicker CEMs (Nafion 117 and F-10150-PFsee Table 1) had considerably higher charge and lower discharge voltages resulting in lower charge/discharge times which can be directly be attributed to the increased thickness of these CEMs. [5]rom the first ten charge/discharge cycles, numerous performance indicators were calculated and the results are presented in Table 2. From these results, it is evident that the use of thicker CEMs (fumasep F-10150-PF of 145 μm and the Nafion 117 of 183 μm) resulted in lower VEs and, as a consequence, significantly lower capacity utilization and ultimately lower discharge capacities.These results are in line with the findings made by Sun et al. [12] as well as Liu et al., [17] who showed that thinner perfluorinated CEMs improved the ICRFB performance.[12,17] When comparing the performance of the two thicker membranes, it is apparent though that the fumasep F-10150-PF did outperform the Nafion 117 on all indicators except the decay where the Nafion 117 yielded a 50% reduced decay compared to the fumasep F-10150-PF membrane.In fact, its decay (0.32% h À1 ) was the lowest obtained for all the membranes which can probably be ascribed to the decreased crossover due to the increased thickness.
Comparing the thinner fumasep CEMs of similar thicknesses to that of the commonly used Nafion 212 with a similar thickness (50 μm), it was evident that all of the fumasep CEMs options deliver comparable performance to that of Nafion 212 with the overall efficiencies (EE) of the fumasep membranes being %2% below that of Nafion 212.Nafion's slightly higher overall efficiency, resulting from a slightly lower CE and VE, is probably the result of Nafion's high conductivity. [5]A similar pattern was observed when comparing the capacity discharge and ASR where the Nafion212 performed slightly better than the three Fumasep membranes.Interesting to note however that the decay when using the Nafion 212 membrane was higher than both the fumasep FS-940 and FS-950, with the fumasep FS-950 CEM yielding a capacity decay rate that was 46% lower than that of Nafion 212.The observed reduction in capacity decay could be a result of improved selectivity, that is, less crossover. [5]While beyond the scope of this study, ion selectivity experiments could be  conducted to verify whether the improved decrease in capacity decay rate was due to the improvement of the ion selectivity.
From these results, it is evident that various commercial alternatives to Nafion 212 exist.It is furthermore apparent that thinner CEMs, such as the fumasep FS-950 or Nafion 212 membranes, yielded an improved overall ICRFB performance confirming the work presented by Sun & Zhang who studied the effect of various Nafion thicknesses on ICRFB performance. [12]espite having identified alternative CEM options, it is evident that although an improvement in capacity retention could be obtained using the fumasep FS-950 CEM, for example, no significant improvement in terms of CE or VE was attained.Consequently the next section will focus on screening two commercial AEMs to investigate whether the CE could possibly be improved via this route.

AEMs
Initially the same electrolyte used for the CEM study, that is, 1.3 M FeCl 2 .4H 2 O, 1.4 M CrCl 3 .6H 2 O, and 5 mM Bi 2 O 3 dissolved in 1 M HCl, was used to evaluate the performance of two commercial AEMs, which were FAP 450 from Fumatech and an undisclosed AEM used by the German RFB company Volterion in their commercial RFB stacks (developed for VRFB application).The performance of these membranes was again assessed at 65 °C with a current density of 40 mA cm À2 and a flow rate of 50 mL min À1 .From the first attempted charge/discharge curve, shown in Figure 2, it is evident that both ICRFBs containing either FAP 450 or the Volterion AEM as membrane were able to charge to 1.25 V, but were not able to discharge even when the voltage limit was set as low as 0.0 V.In addition, as could have been expected when considering the membrane thickness (Table 1), the thinner Volterion AEM (38 μm) charged at a lower voltage than the FAP 450 (50 μm).
After cleaning and disassembly of the ICRFB cell, black coating was observed on the surface area of both AEMs that was tested as shown in the photographs presented in Figure 3. Upon further inspection, it seemed that the solids might have been denser on the membrane surface that was placed on the negative side of the ICRFB.
To obtain further clarification, both the positive and negative sides of the membrane surfaces were analyzed using SEM and compared to SEM images of the surfaces before cycling.Since the solid appeared denser on the surface that was at the negative side of the ICRFB after cycling, both sides (positive and negative) of the AEMs were analyzed.The SEM images prior to (a and b) and after cycling for both the positive (c and d) and negative (e and f ) side are shown in Figure 4 (FAP 450 -left and Volterion AEM -right).From the results, it is evident that some substance had crystallized on both the positive and negative sides of the surfaces of both AEMs after cycling in the ICRFB.In addition, by comparing the images of the positive (c and d) and negative (e and f ) sides of both AEMs, it is also evident that the crystals were significantly denser on the membrane surface that was on the negative side of the ICRFB.
To obtain a qualitative composition (elemental analysis) of the crystals/solids observed on the membrane surfaces after cycling (Figure 4c-f ), the samples were subjected to EDS analysis.The results including the specific points or areas used for the EDS analysis for both the positive and negative sides of both the membranes after cycling are given in Table 3.As shown in Table 3, the EDS analysis (spectra) was performed on points or areas where the solids were clearly visible, and then compared to areas (where possible) where the solids were not, or less, visible.For the negative side of the FAP 450 the Volterion AEM, only two spectra were recorded as only two regions were identified with significantly different compositions.
From the comparison, it became evident that the amount (%) of Bi significantly decreased when moving from an area where the solids were visible to an area where the solids were not.For example, when one compares the EDS results (positive side of the FAP 450) of spectrum 3 (no solids visible) with the results of spectrum 2 (solids clearly visible), it is evident that spectrum 3 (no solids) did not contain any Bi while spectrum 2 contained 13.3%.Similarly, when analyzing the large solid chunk visible on the SEM image of the positive side of the Volterion AEM (spectrum 2), it was evident that this solid mass consisted of %80.0%Bi, while the empty area (spectrum 3) again contained no Bi.
From the data it did seem that the solid particles most likely were Bi deposits from the Bi 2 O 3 added to the electrolyte.To verify these results, the ICRFB testing was repeated using in this case only one membrane (FAP 450 AEM) and an electrolyte where Bi had not been added.The battery performance was again assessed using cycling conditions of 40 mA cm À2 , 65 °C, at a flow rate of  50 mL min À1 .The results of the first charge cycle are shown in Figure 5.It is evident that the exclusion of Bi did not resolve the absence of the discharging experienced previously.It did, however, significantly increase the capacity utilization during the charge cycle.In the presence of Bi (Figure 2), the charging capacity (up to 1.25 V) was only 0.62 Ah for the FAP 450 AEM.In its absence (Figure 5), the charge capacity increased to 1.37 Ah, which corresponds to a capacity utilization increase of 55% when using the FAP 450 membrane.
This experiment however did resolve the deposition issue seen previously (Figure 3), as shown in Figure 6.The lack of visible solid formation on both the positive and negative AEM surfaces supports the findings made from the EDS analysis, that is., that the observed solids were directly caused by the Bi within the electrolyte.Consequently, for all further AEM tests, Bi was excluded from the electrolyte.
In a final attempt to resolve the absence of a discharge curve when using AEMs in an ICRFB, the effect of ment (as is commonly performed in with IEMs) was investigated.For this purpose, the FAP-450 AEM was pretreated by soaking it in 3 M HCl for 1) 0 h, 2) 24 h, and 3) 14 days before evaluating the ICRFB performance using an electrolyte consisting of 1.  no pretreatment and the 24 h pretreatment, it is clear that the 14 day pretreatment reduced the charging voltage of the FAP-450 AEM by an average of 10 mV.Since the charging voltage is directly related to the internal resistance of the battery, it is evident that the prolonged pretreatment significantly reduced the internal resistance during charging.The magnification of Table 3. EDS analysis of positive and negative side surfaces of FAP 450 and the Volterion AEM after charge.

Sample area of EDS analysis EDS analysis results
Element Spectrum [%]  the first 0.2 h in Figure 7 shows the small discharge observed after 24 h (red line) as well as the discharge after 24 days (green line); as there was no discharge after 0 h, there is no black line.It is evident from this that although some discharge was attained after pretreatment, which increased with pre-treatment duration, the discharge did not complete (compared to the case of CEMs, e.g., see Figure 1).While the exact cause of the discharge issue when employing AEMs in the ICRFB is unknown, it might be because of the formation of an anionic FeCl 4 À aqueous species at elevated temperature.Recently, Gammons et al. [18] showed that FeCl 4 À quickly becomes the dominating aqueous species with increasing temperature and increasing Cl À concentration. [18]In the ICRFB, during the charge process (when an AEM is employed), Fe 2þ is oxidized to Fe 3þ , resulting in electrons flowing (via the external electrical circuit) from the positive to the negative electrolyte compartment where Cr 3þ is then reduced to Cr 2þ .The flow of electrons from the positive to the negative side results in an equivalent flow of Cl À ions from the negative to the positive electrolyte (across the membrane) to maintain balance of charge.Consequently, it may very well be that the resulting increase in [Cl À ] in the positive electrolyte combined with the operating temperature (65 °C) results in FeCl 4À being the dominant Fe 3þspecies in solution.This, in turn, could result in FeCl 4 À becoming the dominant charge-carrying species during discharge (since the Cl À anions are now presumed to be coordinated with FeCl 3 ).Considering the increase in size (FeCl 4 À vs Cl À ) of the charge-carrying species, it is logical that this could result in a significant ASR increases during discharge which could be the cause for discharge failure.This would result in the cutoff voltage being reached prematurely, which would in turn result in a low CE value.Whilst proving this mechanism is beyond the scope of this study, it is worthwhile to consider during future work.That would then also provide an opportunity to investigate the cause of the stepping observed in the 14 day curve after 0.4 h.

MPS
Replacing ion-exchange membranes (IEMs) with MPSs would significantly reduce the cost of storage associated with RFBs as IEMs cost %500 $ m À2 , while MPSs are generally below 3 $ m À2 . [13,15]To assess the suitability of using an MPS for an ICRFB, a MPS sample from Entek (EW 200) was obtained and its applicability as an ICRFB separator was evaluated.The MPS was largely composed of ultrahigh-molecular-weight polyethylene and precipitated silica providing an ultrafine interconnected pore structure.When measured, the Entek MPS had a Gurley number of 1135, which equates to a permeance of 0.0112 m 3 m À2 s À1 .
First, simple stability test was done by placing the EW 200 MPS in an ICRFB electrolyte (1.3 M FeCl 2 , 1.4 M CrCl 3 , and 5 mM Bi 2 O 3 in 1.0 M HCl) at 65 °C for 24 h.Having observed no physical damage or changes in weight, the ICRFB performance was tested using the EW 200 MPS.For this purpose of the discussion, the MPS results were compared to that of the best-preforming CEM from Section 4.3.1 (Fumasep FS-950) and not the benchmark Nafion 212 used initially.The results of the first charge/discharge cycle are shown in Figure 8.
It is clear from Figure 8 that the use of the MPS resulted in an increase in charge voltage and a decrease in discharge voltage.Additionally, it can be seen that, whilst the charge time significantly increased when the EW200 was employed, a decrease in discharge time was observed.The efficiencies (CE, VE, and EE) and the discharge capacities (Wh/L) over ten cycles for the EW 200 and the fumasep FS-950 (b) are given in Figure 9a,b respectively.
Considering the microporous nature of the MPS versus the dense CEM, the reduced barrier properties of the MPS, it is understandable that the EW 200 delivered a considerably lower CE value (70%) compared to the 92% obtained when using the FS-950.Interestingly, however both the EW 200 and the FS-950 delivered similar VEs of 76% and 77%, respectively.Due to the aforementioned lack of physical barrier associated with the use of an MPS, one would have expected the ASR of the MPS-based RFB to be considerably lower, resulting in higher VE values.Whilst the use of MPSs in ICRFBs has not been reported in literature, Wei et al. [16] compared the performance of Deramic polyethylene-based MPSs (thicknesses between 200 and 500 μm were tested) to that of Nafion in an iron-vanadium RFB. [16]They also showed that the use of MPS does not improve the VE of the RFB which they attributed to the increased thickness of the MPS combined with the fact that the proton transport in an MPS is not facilitated by the proton surface hopping mechanism found in CEMs. [16]It hence seems that similar results were obtained in both the iron-vanadium RFB [16] and the ICRFB when using an MPS.Because of the low CE, the EW 200 MPS delivered an average EE of only 54% compared to the 77% obtained when using the FS-950 membrane.Finally, and in addition to the low efficiencies, the MPS also resulted in a substantially high decay rate (compared to that of the FS-950), as shown in Figure 9.
To investigate whether the obtained low CE and high capacity decay rate of the EW 200 was due to self-discharge across the interconnected pores, an experiment was conducted where the self-discharge of the EW 200 MPS was compared to that of the FS-950.This was achieved by charging an ICRFB to 50% SOC (one containing the MPS EW 200 and another one containing the CEM FS-950 as separator) and allowing an open-circuit voltage period for self-discharge.The resulting voltage decrease as a function of time would therefore be representative of the self-discharge across the membrane surface.Both self-discharge tests were conducted at 65 °C at an electrolyte flow rate of 50 mL min À1 .As shown in Figure 10, the EW 200 MPS had a considerably higher self-discharge rate compared to the of the fumasep FS-950 CEM, where the OCV of the EW 200 MPS dropped to 0.75 V (%15% SOC) within 6 h, while it took more than 45 h to reach 0.75 V when using the CEM FS-950.Since identical electrolytes were used in both cases, it can be assumed that the significant difference in self-discharge rate resulted from the difference of the self-discharge across the membrane surfaces. [19]Since the use of a MPS as separator within the ICRFB leads to a significant reduction of a physical barrier between the positive and negative electrolyte, it is logical to assume that the high discharge rate may be attributed to self-discharge within the interconnected pores within the MPS.

Conclusion
Due to the high cost associated with Nafion, the aim of this section was to evaluate the suitability and applicability of alternative membranes/separators (CEMs', AEMs and an MPS) in an ICRFB.Charge/discharge curves were used to calculate various performance indicators of the selected commercial membranes/ separators and compared against a Nafion 212 benchmark membrane.
Most of the CEMs tested (Fumatech's fumasep FS 940, fumasep FS 950, and fumasep F-1050) was near identical to Nafion 212.The fumasep FS-950 showed slightly lower capacity decay rate (0.33% h À1 ) while still delivering an acceptable discharge  capacity (24.22 Ah L À1 ).During this study it was found that the AEMs tested (FAP 450 from Fumatech and an undisclosed AEM used by Volterion in their commercial VRFB stacks) were not suitable in an ICRFBs when using an electrolyte composition of 1.3 M FeCl 2 , 1.4 M CrCl 3 , and 5 mM Bi 2 O 3 in 1.0 M operated at 65 °C.Whilst the charging performance of the ICRFB using AEMs could be improved significantly by excluding Bi from the electrolyte composition, we were unable to discharge the ICRFB when AEMs were used as separator.When comparing the performance of the EW 200 MPS to that of the fumasep FS-950, it became apparent that although the MPS showed excellent stability in the ICRFB, the selectivity was too low to obtain an acceptable CE.In addition, it was found that severe crossover resulted in significant self-discharge and ultimate capacity decay.Consequently, future work should focus on identifying MPSs with improved selectivity while still delivering an acceptable VE.

3. 1 .
CEMsDuring this study, three thin (%50 μm) and one thicker (145 μm) commercial CEMs from fumasep were screened and compared to Nafion 212 (50 μm) and Nafion 117 (183 μm).This allowed the comparison of the Fumatech's fumasep FS 940, fumasep FS 950, and fumasep F-1050 membranes to Nafion 212, while the Fumatech's fumasep F-10150-PF membrane was tested and compared to Nafion 117.All CEMs were tested using an electrolyte composition of 1.3 M FeCl 2 , 1.4 M CrCl 3 , and 5 mM Bi 2 O 3 in 1 M HCl between 0.75 and 1.25 V at 65 °C using a current density of 40 mA cm À2 .The voltage versus time graphs of the first charge/discharge curves for the 6 CEMs tested are shown in Figure 1.

Figure 1 .
Figure 1.Voltage as a function of time for the first charge/discharge cycle of various CEMs measured at 65 °C and 40 mA cm À2 .

Figure 2 .
Figure 2. First charge/discharge curve of two AEMs measured at 65 °C and 40 mA cm À2 .

Figure 3 .
Figure 3. Black solid formation on the active surface of a) FAP 450 and b) Volterion AEM after one charge cycle.
3 M FeCl 2 , 1.4 M CrCl 3 in 1.0 M HCl.The effect of pretreatment is shown in Figure 7.While there was little difference between

Figure 4 .
Figure 4. a,b) SEM images of both the FAP 450 (left) and the Volterion (right) AEMs prior to cycling as well as images of both the c,d) positive and e,f ) negative side after cycling.

Figure 5 .
Figure 5. First charge cycle where Bi was excluded from the electrolyte composition using the FAP 450 AEM in the ICRFB.

Figure 6 .Figure 7 .
Figure 6.FAP 450 AEM after charge cycle where no Bi was used in the electrolyte.

Figure 8 .
Figure 8.First charge discharge cycle measured at 65 °C and 50 mL min À1 .

Figure 9 .
Figure 9. CE, VE, EE (primary), and discharge capacity (secondary) of a) the MPS EW 200 and b) the CEM fumasep FS-950 as measured at 65 °C and 50 mL min À1 .

Figure 10 .
Figure 10.Self-discharge curves measured at 65 °C at a flow rate of 50 mL min À1 .

Table 1 .
Commercial CEMs, AEMs, and MPS tested during this study.
a) In 0.5 M H 2 SO 4 ; b) In 2 M H 2 SO 4 at t = 25 °C; c) Commercial AEM used in Volterion's commercial VRFB stacks.

Table 2 .
Performance indicators of various commercial CEMs measured at 65 °C and 40 mA cm À2 , where CE refers to current efficiency, VE to voltage efficiency, EE to energy efficiency, and Cap Dis to discharge capacity, Decay to capacity decay rate and ASR to area specific resistance.