Characterization of an Aqueous Flow Battery Utilizing a Hydroxylated Tetracationic Viologen and a Simple Cationic Ferrocene Derivative

Herein, a new semi‐organic aqueous flow battery based on a hydroxylated tetracationic viologen, 1,1′‐bis(3‐((2‐hydroxyethyl)dimethylammonio)propyl)‐[4,4′‐bipyridine]‐1,1′‐diium tetrachloride ([(DMAE‐Pr)2‐Vi]Cl4), and the ferrocene derivative, (ferrocenylmethyl)trimethylammonium chloride (FcNCl), is demonstrated. Efficiency, accessible capacity, and capacity retention of the battery are investigated at two concentrations of the active redox species: 0.1 and 0.5 mol dm−3 in 1 mol dm−3 KCl near‐neutral electrolytes. The implementation of the ferrocene‐derivative posolyte decreases the capacity fade rate by a factor of ≈4 with respect to previous work using 4‐hydroxy‐2,2,6,6‐tetramethylpiperidin‐1‐oxyl (TEMPOL) as posolyte. Capacity loss is driven by crossover of the positive redox couple into the negolyte, in particular at high concentrations, indicating the need for more selective membranes and less permeable ferrocene derivatives. Discussions are supported by conductivity measurements, cell resistance, and postmortem analysis of the electrolytes using cyclic voltammetry and nuclear magnetic resonance (NMR) spectroscopy. The characterization of [(DMAE‐Pr)2‐Vi]Cl4 is expanded as a high‐energy negolyte and future scaleup requirements for aqueous organic flow batteries are informed.

performance. [11]However, it was recently proved that BTMAP-Vi can still undergo dimerization, [13] though to a lesser extent than dicationic viologens.The latter, and thus MV, can form instead association complexes of higher order, for example, trimeric structures, complicating the chemistry of this negolyte. [13]everal modifications have been proposed to tetracationic viologens since the report of BTMAP-Vi aiming to improve flow battery performance.Recently, a PEGylated tetracationic viologen was proposed and characterized as negolyte by Peng et al. after pairing it with a simple ferrocene posolyte ((ferrocenylmethyl) trimethylammonium chloride or FcNCl). [14]It was concluded that the addition of a flexible polyethylene oxide trimethyl ammonium chain to the 4,4´-bipyridine core did not improve molecular stability, despite reducing its membrane permeability below detection limits. [14]Another recent modification involved the addition of a secondary alcohol moiety between the bipyridine core and the charged ammonium group in the so-called Dex-Vi, developed by Feng et al. [15] Dex-Vi showed improved stability and good solubility (%2.0 mol dm À3 ) and can be easily prepared from commercially available reagents. [15]n our previous work, [16] we evaluated the effect of a terminal primary alcohol unit on a tetracationic viologen by choosing dimethylaminoethanol (DMAE) as starting material for the permanently charged moiety, given that the addition of hydroxyl groups can alter both the electronic structure and the solubility of the molecule.The new viologen, 1,1 0 -bis(3-((2-hydroxyethyl) dimethylammonio)propyl)-[4,4 0 -bipyridine]-1,1 0 -diium tetrachloride or [(DMAE-Pr) 2 -Vi]Cl 4 (see Figure 1), showed improved solubility (2.7 AE 0.3 mol dm À3 ), [16] enhanced electrochemical reversibility and HOMO-LUMO levels in agreement with the presence of hydroxyl groups.However, the capacity retention of the [(DMAE-Pr) 2 -Vi]Cl 4 negolyte in a flow battery was severely limited by the choice of TEMPOL as a readily available posolyte.The TEMPOL molecule is known to undergo decomposition upon cell cycling and to have high membrane permeability due to its small size and absence of charge as a radical. [17]urthermore, it has been established that longer lifetimes and an overall better battery performance can be achieved with ferrocene-based posolytes. [17,18]e present work aims to continue the characterization of [(DMAE-Pr) 2 -Vi]Cl 4 as a negolyte along a more suitable posolyte, namely FcNCl, see Figure 1.The latter is a ferrocene derivative bearing a trimethylammonium group, which confers high water solubility to the ferrocene moiety (i.e., a reported solubility of 4.0 mol dm À3 in water or 3.0 mol dm À3 in 2.0 mol dm À3 NaCl). [19]FcNCl is used as a convenient posolyte mainly due to its ease of synthesis. [19]In contrast to previous work, flow battery tests for [(DMAE-Pr) 2 -Vi]Cl 4, were performed at the common concentration of 0.1 mol dm À3 but also at a more concentrated 0.5 mol dm À3 .The effect of the current density on battery metrics, such as efficiency and accessible capacity, was assessed in the range 20-120 mA cm À2 .Galvanostatic cell cycling with a potential hold at the end of each cycle was used to evaluate the capacity retention at both concentrations.Postmortem analysis was carried out by cyclic voltammetry and 1 H-nuclear magnetic resonance (NMR) spectroscopy to look for eventual decomposition and crossover.

Cyclic Voltammetry
Cyclic voltammetry of the redox couples was carried out at diluted solutions of 0.002 mol dm À3 in 1 mol dm À3 KCl (see Figure 2) and at the concentration of flow battery electrolytes, namely 0.1 and 0.5 mol dm À3 in 1 mol dm À3 KCl (see Figure S3 and S4, Supporting Information).Based on previous experience, [16] only the first reduction event was accessed for [(DMAE-Pr) 2 -Vi]Cl 4 .As shown in Figure 2, the voltammograms for both redox couples in diluted conditions are fully electrochemically reversible, implying fast electrode kinetics.This was demonstrated during the characterization of [(DMAE-Pr) 2 -Vi]Cl 4 in previous work, [16] with the same being shown for FcNCl. [19]he separation between their formal potentials at open circuit is 0.96 V, which is slightly lower than the difference between the formal potentials of [(DMAE-Pr) 2 -Vi]Cl 4 and TEMPOL (1.15 V). [16] As shown in Figure S4, Supporting Information, the voltammograms for the redox species at a concentration of 0.5 mol dm À3 show the typical increase in peak separation and display no sign of impurities.The effect of the concentration on the peak height and overall shape of the voltammograms for [(DMAE-Pr) 2 -Vi]Cl 4 was also assessed as a preliminary indication of electrolyte stability.Increasing the concentration to 0.1 or 0.5 mol dm À3 did not affect the apparent stability for its first reduction during the cycling, see Figure S5 and S6, Supporting Information.Indeed, no strong concentration effects were detected in the voltammograms, which displayed only larger peak separation for each redox active species.Likewise, no significant concentration effects were recently reported for BTMAP-Vi, which was considered to prove that the stability of the generated radical cation limits the formation of adducts in solution. [20]A quantitative analysis of the dimerization of stable radicals like [(DMAE-Pr) 2 -Vi]Cl 4 will require more sophisticated methods, such as the in situ IR performed by Nolte et al. [13]

Conductivity of Electrolytes
For the [(DMAE-Pr) 2 -Vi]Cl 4 negolyte in the same 1 mol dm À3 KCl supporting electrolyte, the conductivity was 119.1 mS cm À1 at 0.1 mol dm À3 against 143.1 mS cm À1 at 0.5 mol dm À3 ; an increment of 20.1%.The values for the FcNCl posolyte increased only marginally from 105.9 mS cm À1 at 0.1 mol dm À3 to 109.6 mS cm À1 at 0.5 mol dm À3 in the same 1 mol dm À3 KCl supporting electrolyte, an increment of 3.5%.It is worth noting that the conductivity of 1 mol dm À3 KCl at 22 °C is 105.5 mS cm À1 , implying a low molar conductivity of the organic redox couples and particularly for the ferrocene derivative.In contrast, vanadium electrolytes have conductivities between approximately 300 and 460 mS cm À1 . [21]This highlights the need to eventually increase the concentration or the molar conductivity of supporting electrolytes, provided this is compatible with combined solubility limits.

Efficiency and Accessible Capacity
For the "low concentration" flow battery (0.1 mol dm À3 [(DMAE-Pr) 2 -Vi]Cl 4 ), %90% of the theoretical capacity was accessed at 20 mA cm À2 , with this value decreasing to 42.5% as the current density was increased to 120 mA cm À2 (Figure 3a).The EE ranged from 91.6% at 20 mA cm À2 to %60% at 120 mA cm À2 (Figure 3b).Meanwhile, the CE stayed over 99% for most of the current density values, except for 100 and 120 mA cm À2 , which showed values of 98.8% and 97.6%.The maximum CE of 99.6% was observed at 20 mA cm À2 .These values are in line with previous work employing the same membrane, [16] but higher when compared to other membranes such as Selemion DSV.For example, the EE for the PEGylated tetracationic viologen in a cell with this membrane was only 65.5% at 80 mA cm À2 , [14] and a similar value can be estimated for the flow battery using Dex-Vi. [15]In this work, the EE at the same current density is 71.5%, likely due to the thinner FAS-30 membrane.
The flow battery was then evaluated at "high concentration" with both redox couples at 0.5 mol dm À3 .In these conditions, the accessible capacity was as high as 86% at 20 mA cm À2 , reaching 70% at 120 mA cm À2 (Figure 4a).Regarding the efficiencies, the CE stayed above 99% for all the screened current densities (with a peak of 99.60% at 80 mA cm À2 ) while the EE ranged from 94% at 20 mA cm À2 to 72% at 120 mA cm À2 (Figure 4b).The improved accessible capacities and efficiencies recorded at the higher concentration can be ascribed to a combination of eased mass transfer overpotentials for the electrode reactions and to a lower cell resistance due to a more conductive negolyte.Yet, the conductivity had likely a lesser effect, given that only the value for the negolyte increased meaningfully by 20.1% in respect to the 0.1 mol dm À3 negolyte, as discussed in the previous section.This is corroborated by electrochemical impedance spectroscopy (EIS), which revealed an overall cell area-specific resistance (ASR) of %1.01 Ω cm 2 for this [(DMAE-Pr) 2 -Vi]Cl 4 /FcNCl flow battery for both electrolyte concentrations, as measured before and after cell cycling.Indeed, the ASR of the membrane-free flow cell used in this work was 0.24 Ω cm 2 as measured by EIS with circulating 1.0 mol dm À3 KCl solution, suggesting a predominant contribution of the membrane, which would then represent an ASR of 0.77 Ω cm 2 in these conditions.The value for the [(DMAE-Pr) 2 -Vi]Cl 4 /FcNCl flow battery is less than half of the one reported from Beh et al. [11] for the BTMAP-Vi/BTMAP-Fc cell separated by a Selemion DSV (2.5 Ω cm 2 ), and almost four times lower than the value reported by Hu et al. [22] for a BTMAP-Vi/N Me -TEMPO cell separated by a Selemion AMV (4.48 Ω cm 2 ).This is a consequence of the FAS-30 membrane having a thickness of 30 μm, with Selemion DSV and Selemion AMV being 90 and 110 μm thick, respectively. [23]Implications on crossover are commented later.The value of %1.01 Ω cm 2 for this work is also roughly one third of that reported by Hagemann et al. [24] for a TEMPO-copolymer/dimethyl-viologen-dichloride flow battery with the same FAS-30 membrane (3.3Ω cm 2 ).This is due to differing cell-resistance components, partly a less conductive 1 mol dm À3 NaCl-supporting electrolyte.The literature confirms that the ASR of this and the cited flow batteries is dominated by the resistance of the anion exchange membrane (AEM). [11]In general, the relatively high ASR values explain why current densities of up to 120 mA cm À2 could hardly be applied to this and similar viologen-based flow batteries, whereas vanadium systems can reach over 400 mA cm À2 in laboratory cells with similar design. [25]Typical ASR values for vanadium cells with felt electrodes and more conductive perfluorinated proton exchange membranes are 0.85-0.51Ω cm 2 . [26]

Capacity Fade Rate at "Low Concentration"
The performance of the [(DMAE-Pr) 2 -Vi]Cl 4 /FcNCl organicorganometallic flow battery was demonstrated under a galvanostatic/potentiostatic regime. [8]A current density value of 40 mA cm À2 was chosen from the previous section as a tradeoff between battery performance and cycle duration and to enable comparison versus previous flow batteries, including a [(DMAE-Pr) 2 -Vi]Cl 4 /TEMPOL system with a typical concentration of 0.1 mol dm À3 . [16]As shown in Figure 3c, the battery sustained 500 cycles for a total of %256 h (10.7 days), during which the volumetric capacity decreased from 96.2% to 87.5% with respect to the theoretical capacity.The initial accessible capacity was higher than in Figure 3b (86.1% average at 40 mA cm À2 ), as in that experiment the potential hold was omitted to permit electrode overpotentials and cell resistance to reach cell voltage cutoff values.Average CE of 99.8% and EE of 83.3% were recorded during the 500 cycles, improving over the average EE of 81.1% for a flow battery employing the same negolyte, but a nitroxyl-radical-based posolyte, that is, TEMPOL posolyte, which endured only 300 cycles in previous work. [16]he capacity fade rate of the flow battery was %0.02% cycle À1 or %0.04% h À1 (%1% day À1 ) as calculated from the steady slope between cycle 210 and 500, see Figure 3c,d.This translates into a capacity retention rate of 99.98%.The capacity fade rate of the [(DMAE-Pr) 2 -Vi]Cl 4 /FcNCl flow battery represents an improvement by a factor of %4 with respect to the previous battery using a TEMPOL posolyte. [16]This is consistent with the longer lifetimes reported for organometallic redox couples (e.g., Fc) with respect to fully organic posolytes. [17]Since oxygen levels are negligible, [27] the observed behavior can be ascribed to the higher permeability of TEMPOL across the AEM toward the limiting negolyte, caused by the lack of positively charged groups, and to the subsequent redox and degradation reactions within the negolyte (e.g., protonation, ring-opening reactions). [17]In contrast, FcNCl possess a positively charged group that increases Coulombic repulsion with the AEM.However, the small size of the substituent and the presence of only a single positive charge still make FcNCl susceptible to membrane crossover. [28,29]ts stability was recently compared to new analogues and will be discussed later. [30]eturning to Figure 3c, it must be noted that the capacity profile versus cycle number shows three different regions, two appearing before the steady slope.The usual "first day" fast capacity fade rate was observed during the first %20 cycles (%0.07%cycle À1 or %0.13% h À1 or %3.2% day À1 ).This "first day" behavior has been reported not only for other laboratory AOFBs, for example, ref. [31] or ref. [32], but also in molecular stability studies using both the unbalanced, compositionally symmetric flow cell and the steady-state amperometric state of health techniques. [33]Following the 20th cycle, a second capacity fade slope appears until cycle %210 with a fade rate of %0.007% cycle À1 (%0.013 h À1 or %0.3% day À1 ).The "first day" effect and the second capacity fade rate are due to time-dependent membrane-resistance changes resulting from electrolyte uptake and reaction of the redox species within the membrane. [10,34]ndeed, representative values are to be measured afterward from a steady slope, particularly in the week range. [33]ese results can be put in perspective by comparing them to similar AOFBs.For instance, the pyrrolidinium-based extended viologen recently reported by Pan et al. [35] shows a capacity fade rate of 0.04% cycle À1 over 500 cycles (4.56% day À1 ) when used at 0.2 mol dm À3 in combination with a pyrrolidinium-based TEMPO posolyte.The battery here reported represents an improvement of a factor of %5 on the fade rate per day.In another work, a new PEGylated viologen shows a capacity fade rate of %0.11% cycle À1 (the cell retained 82.8% of the initial capacity after 160 cycles) when used as two-electrons-storage negolyte in combination with FcNCl as posolyte.However, it should be underlined that the capacity fade rate obtained when the latter viologen was used as one electron-storage negolyte was not reported by the authors, while they report a capacity retention of 99.996%. [14]This battery displayed a capacity fade rate per cycle 6.5 times higher than the one in this work.
Regarding molecular degradation, the viologen here used demonstrated remarkable stability as proved by the postmortem analysis, that is, cyclic voltammetry and 1 H-NMR spectroscopy, see Figure 5 and S10 and S11, Supporting Information.This stability can be attributed to a combination of two factors.First, the high dihedral angle between the two pyridinium rings in the cationic radical (see Figure S2, Supporting Information) which, as hypothesized by Liu et al., [36] could make the dimerization less favorable.[(DMAE-Pr) 2 -Vi]Cl 4 shows in fact the highest dihedral angle (%4.8°) among several derivatives, including BTMAP-Vi (%0.8°); second, a possible interaction between the OH groups and the radical cation. [16]In contrast, the PEGylated viologen shows decomposition in similar experiments, as indicated by the presence of unexpected signals in the 1 H-NMR of the spent negolyte, suggesting that side reactions occurred during operation. [14]Indeed, the only difference between the voltammograms and the proton spectrum of the uncharged [(DMAE-Pr) 2 -Vi]Cl 4 negolyte and the spent one in Figure 5a and S9, Supporting Information, is the presence of the FcNCl posolyte (estimated to be %20% of the viologen, based on the H-NMR integration).This confirms that the membrane permeability of [(DMAE-Pr) 2 -Vi]Cl 4 across the AEM is negligible in these conditions, contrary to the active species in the posolyte.In contrast, the 1 H-NMR of the spent posolyte is quite different from the fresh FcNCl as shown in Figure S12, Supporting Information.We ascribe this to a combination of low concentration (due to decomposition, crossover, and further dilution with D 2 O to get a stable solvent lock) and closeness of the saturated solvent peak.
Indeed, FcNCl has been shown to display both membrane crossover issues [28,29] and poor capacity retention in long-term experiments. [30]While evaluating new two electrons negolyte materials, Liu et al. reported FcNCl crossover problems with membrane as thick as 130 μm, yet to a less extent than the crossover recorded here with a thinner membrane (i.e., 30 μm) thus confirming that the membrane plays a key role in the crossover of the redox active species. [28]Additionally, when relating degradation rates to three chain lengths of FcNCl analogues (namely C 1 , C 2 , and C 3 ) between the trimethylammonium group and the cyclopentadienyl ring (Cp), Luo et al. reported that the C 1 turned out to be the most unstable, despite being the easiest derivative to prepare. [30]In fact, the C 1 FcNCl derivative showed increased O 2 sensitivity in addition to lower thermal-and photostability at both the charged and discharged state. [30]The literature thus confirms that the capacity loss of this type of AOFB based on viologen and ferrocene derivatives is mainly driven by the posolyte. [8,30]pecifically, and since in this case the posolyte is non-limiting, capacity loss originates not in the instability of the redox species, but in its crossover and subsequent interaction with the negolyte species. [35]The recent work from Feng et al. on the Dex-Vi negolyte further confirms this point. [15]There, the spent negolyte did not show any sign of the ferrocene derivative, namely bis-trimethylammoniopropyl-ferrocene or BTMAP-Fc. [15]This ferrocene is recognized as more stable than FcNCl due to a combination of its chain length and substitution on both Cp rings. [11,30,37]In contrast, its synthesis implies a multistep procedure requiring a catalyst and purification by column chromatography for the starting material complicating its scaleup. [11]In these conditions, Dex-Vi showed negligible capacity loss after 1 month of cell cycling at 1.5 mol dm À3 as limiting negolyte, aided by both its stability, owing to the addition of a secondary alcohol moiety, and its negligible permeability through the AEM.The work from Beh et al. came to similar results for the battery with BTMAP-Vi as limiting negolyte and BTMAP-Fc as nonlimiting posolyte, which showed a very low capacity fade rate (11% year À1 , with a contribution of %6% due to electrolyte crossover). [11]These authors also concluded that an optimal ratio of redox-active reactants could minimize water crossover. [11]ollowing this approach, it may be expected that using a larger volume of more diluted [(DMAE-Pr) 2 -Vi]Cl 4 negolyte would decrease the water crossover from the FcNCl posolyte.We also propose to adjust the concentration of Cl À in the supporting electrolyte to deter water osmosis.For instance, KCl could be added to the posolyte to have the same initial concentration of Cl À at both sides of the membrane.Taking everything into account, overall capacity retention is a result of combining intrinsic stability of redox couples and negligible membrane crossover.

Capacity Fade Rate at "High Concentration"
The performance of the [(DMAE-Pr) 2 -Vi]Cl 4 /FcNCl battery at concentrations of 0.5 mol dm À3 in 1 mol dm À3 KCl was evaluated at the same current density used for the low concentration experiment, that is, 40 mA cm À2 (Figure 4c,d).The battery sustained 120 cycles or %297 h of operation (%12.4 days) with a capacity fade rate of %0.18% cycle À1 or %0.073% h À1 (%1.7% day À1 ), calculated from cycle 10 to 120.The accessible volumetric capacity decreased from 97.0% to 76.3% of the theoretical capacity.Regarding the efficiencies, the CE stayed above 99% for the whole experiment yielding an average of %99.2%, while the averaged EE was %86.4%.The EE decreased from %88.3% to %83.7% over the 120 cycles.This was likely due to the observed crossover and water transport as discussed later.Despite the increase in viologen concentration, the battery was not negatively affected in terms of voltage profile, as seen in Figure 4d.In addition, and in agreement with the literature, [15] no pumping problems due to viscosity were encountered.Indeed, the viscosity of the uncharged negolyte at 0.5 mol dm À3 was 1.84 cSt, which is near double of the one reported for the 0.1 mol dm À3 solution, 1.01 cSt. [16]The viscosity of the charged negolyte was 1.86 cSt, an insignificant difference from the uncharged state.The capacity fade rate recorded in this experiment can be compared to other flow batteries using different tetracationic viologen at the same concentration.For instance, a value 0.17% cycle À1 was reported for a π-conjugated viologen by Tang et al. [32] this value was obtained at 80 mA cm À2 with an EE of %75% and it should be underlined that the conjugated nature of the viologen implies accessing the second e À reduction.A value of 0.005% cycle À1 (calculated from the reported loss of 2.5% capacity over 500 cycles) was reported for BTMAP-Vi by Hu et al. when paired with a TEMPO-based posolyte at 60 mA cm À2 (EE of %57%). [22]The latter value was further lowered to 0.006% cycle À1 when switching to BTMAP-Fc as the posolyte, as reported by Beh et al. [11] for an argon-filled glovebox.
The higher capacity fade rate obtained at 0.5 mol dm À3 , compared to the low concentration experiment, cannot be explained by the well-known bimolecular degradation mechanisms of viologens. [11]In fact, as evidenced by the postmortem analysis of the electrolytes, crossover of FcNCl into the negolyte can be held accountable as the main capacity loss mechanism.Indeed, as seen in Figure 6a, the postmortem voltammogram does not show any sign of degradation (the 1 H-NMR spectrum only shows minor changes, see Figure S9 and S10, Supporting Information, while clearly showing the presence of FcNCl).
In contrast, inspection of the posolyte's postmortem voltammogram revealed a small broad cathodic peak around À0.5 V, see Figure 6b.The identity of this peak is unclear.Therefore, a voltammogram of a fresh diluted solution of posolyte plus a small amount of [(DMAE-Pr) 2 -Vi]Cl 4 (30% by mol) was recorded for comparison, see Figure S7, Supporting Information.The typical two reversible reduction events of the viologen here used could be clearly recognized, despite the low concentration.This finding allows us to exclude the presence of pristine viologen in the posolyte.Thus, the peak in Figure 6b at À0.5 V might belong to a small amount of the product of decomposed negolyte after crossing the membrane.As discussed by Beh et al., [11] typical decomposition products of viologens come from either a dealkylation reaction or from dimerization, for example, disproportionation or formation of stable dimers.However, monoalkylated bipyridine usually appears around À1.0 V, [38] while the neutral bipyridine is insoluble in water and would likely crush out from the solution.It should be mentioned that the pH of the posolyte turns acidic during cell cycling and the eventual negolyte decomposition product could be further protonated.This decomposition is further supported by small signals in the aromatic region and consistent with the bipyridine moiety in the 1 H-NMR spectrum, see Figure S11, Supporting Information.The implementation of mass spectroscopy on the spent negolyte could possibly help in the identification of this minor decomposition pathways.Regarding the 1 H-NMR of the posolyte, as shown in Figure S12, Supporting Information, only broad signals can be seen, probably due to a combination of the closeness of the solvent peak and to the decomposition of FcNCl itself.
It should also be noted that the volume ratio between the negolyte and posolyte was %1:1 after the experiment, in contrast with the initial 1:2, which means that more water crossover took place in comparison to the low concentration experiment.This is reasonably expected, as higher concentration means more diffusion and migration, with the associated water transport.As mentioned earlier, an improvement strategy would involve a tailored ratio of chloride redox-active salts among the two half-cells to minimize solvent crossover. [11]Again, we speculate that the tuning of the supporting electrolyte concentration could also help to exclude the imbalance in Cl À concentration between half-cells, which favors water crossover due to osmosis.Perhaps the development of charge-reinforced ion-selective (anion) membranes for viologen-based flow batteries could significantly cut crossover from the posolyte redox species, as shown for cation exchange membranes. [39]

Conclusion
A new organic/organometallic flow battery based on the highly soluble [(DMAE-Pr) 2 -Vi]Cl 4 as negolyte was demonstrated by implementing a FcNCl posolyte as a significant improvement over the TEMPOL posolyte used in previous work.The performance of the flow battery was discussed in terms of accessible capacity and efficiency over current density as well as capacity fade rate in extended cell cycling for two concentrations of the redox couples.While [(DMAE-Pr) 2 -Vi]Cl 4 showed an acceptable performance as negolyte under the chosen conditions, postmortem analysis of the electrolytes showed that the overall capacity retention was dominated by FcNCl crossover into the limiting negolyte.Crossover mechanisms, including that of water, were exacerbated at higher reactant concentrations, confirming the role of FcNCl diffusion, in particular for a thin membrane such as the FAS-30.The severe capacity loss observed at higher concentrations of the posolyte could be ameliorated by using a ferrocene derivate less prone to crossover or a thicker membrane.However, from a general viewpoint, this behavior highlights the pressing need for more selective membranes possessing low resistivity.Indeed, presently capacity retention in laboratory-scale AOFBs depends on the trade-off among membrane ASR, thickness, and permeability.ASR of AOFB cells must be reduced to approach the operational current densities of vanadium systems.The low conductivity of organic, near-neutral electrolytes is an important factor to be addressed.Future work will explore new moieties for the development posolytes according with the guidelines here derived, in addition to improved membrane selection.

Experimental Section
All substances were purchased from Fluka, TCI, BLD pharm, or Sigma-Aldrich, stored under nitrogen when required and used as such.Amberlite IRA-900Cl was purchased from Thermo Fisher Scientific.Experiments were performed under a nitrogen atmosphere unless otherwise stated.High-performance liquid chromatography (HPLC)-grade water (Carl Roth) with a conductivity of 18.2 MΩ cm was used in the preparation of the solutions.The pH of solutions was measured with a pH-Check 5040-0301 probe (Dostmann Electronic).Conductivity measurements were carried out by triplicate with a SevenGo Duo conductivity/pH meter (Mettler Toledo) at 21 °C.Spectral NMR data were collected with a Bruker spectrometer, either an AVANCE NEO 400 MHz or an AVANCE III 600 MHz.High resolution electrospray ionization mass spectroscopy (HRESI-MS) spectra were acquired in an liquid chromatography/mass selective detector (LC/MSD) system Serie HP1100 atmospheric pressure ionization-electrospray ionization (API-ES) spectrometer (Agilent).The viscosity of uncharged and charged negolyte was measured by triplicate at 25 °C under air using an Ubbelohde viscosimeter type 501 (SI Analytics) after filtrating the solutions.For this measurement, the charged state of [(DMAE-Pr) 2 -Vi]Cl 4 was accessed by chemical reduction with zinc powder (1 eq).The temperature was controlled within AE0.1 °C using a recirculating water bath (Huber).The procedure was performed as described earlier. [16]However, the synthesis was successfully demonstrated on a larger scale (17 g, 23.0 mmol), without any detrimental effects on the final isolated yield of 88%.The molecular characterization was consistent with the previous work.
Synthesis of 1,1'-Bis(3-((2-Hydroxyethyl)dimethylammonio)propyl)-[4,4'-Bipyridine]-1,1'-Diium Tetrachloride, [(DMAE-Pr) 2 -Vi]Cl 4 : Amberlite IRA-900Cl was packed in a column and used to exchange bromine with the desired chlorine ions.In a typical procedure, [(DMAE-Pr) 2 -Vi]Br 4 (15 g, 20.3 mmol) was dissolved in water and was passed through an excess of resin until the solution coming out from the column turned colorless.The water was then removed under reduced pressure and the resulting solid dried under high vacuum (99% yield, 11.26 g, 20.1 mmol).The characterization of the chloride salt was consistent with previous work. [16]ynthesis of Ferrocenylmethyl)trimethylammonium Iodide (FcNI): The procedure was replicated from an earlier description. [40]Spectroscopic data were consistent with the literature. [40]ynthesis of Ferrocenylmethyl)trimethylammonium Chloride (FcNCl): Amberlite IRA-900Cl was employed to exchange iodine to produce the desired chloride salt.In a typical experiment, FcNI (18.4 g, 47.8 mmol) was dissolved in water/methanol (2:1, v/v) and the resulting solution was passed through an excess of resin in a packed column until the solution became colorless.The solution was then concentrated under reduced pressure and dried over high vacuum (98.4% yield, 13.8 g, 47 mmol). 1 H-NMR was superimposable to the iodide salt and consistent with the literature. [19]yclic Voltammetry: All the experiments were performed with an Autolab PGSTAT204 potentiostat/galvanostat (Metrohm).A threeelectrode cell configuration was used.Before each measurement, the glassy carbon working electrode (d = 2 mm) was polished with a 0.03 μm Al 2 O 3 slurry and then rinsed carefully with deionized water.A platinum sheet was used as counter electrode and the reference was an Ag/AgCl (3.0 mol dm À3 KCl) electrode.All the measurements were carried out using either 0.1 or 1 mol dm À3 KCl as the aqueous supporting electrolyte at a temperature of 22 °C under a nitrogen atmosphere after purging the solutions from molecular oxygen for 30 min.Postmortem analysis of the spent electrolytes was recorded after the necessary dilution to reach a concentration of 0.002 mol dm À3 of redox-active species.
Flow Battery: A rectangular channel flow cell (PinFlow) with a projected active area of 4 cm 2 was employed in all the experiments.Two pieces of Sigracell GDF 4.6 EA carbon felt (SGL Carbon) with a thickness of 4.6 mm were used as electrodes after being activated for 30 h at 400 °C in air.The felts reached a nominal thickness of 4.0 mm (compressed to approximately 87% of their original thickness) after assembling the flow cell and tightening its six tie bolts with a torque of 8 N m each.The two half-cells were separated by a Fumasep FAS-30 AEM (Fumatech) with a nominal thickness of 30 μm, which was pretreated in the supporting electrolyte for no less than 24 h.A Masterflex Drive L/S 600R peristaltic pump (Cole-Parmer) employing Viton and fluorinated ethylene propylene (FEP) tubing formed the ancillary flow system.The electrolyte reservoirs consisted of glass vials placed inside 100 cm 3 amber glass bottles with HPLC caps (Bola) sealed with Viton gaskets.All the tubbing was protected from ambient light.Experiments were performed in a nitrogen-filled box developed in-house (Figure S1, Supporting Information) and equipped with a calibrated TO2-1x oxygen sensor (Southland Sensing).Oxygen concentration within the box fluctuated between 160 and 200 ppm during the experiments, that is, oxygen levels were always below 0.02%.All experiments were performed at a temperature of 22 AE 2 °C, as indicated by a thermometer placed inside the inert-gas box.
The posolyte solution was supplied in 2Â volumetric excess of the capacity-limiting negolyte keeping the same concentration of the latter: 30 versus 15 cm 3 .A flow battery with 3.5Â volumetric excess of posolyte was tested to confirm that this electrolyte was actually non-limiting over the duration of the experiments.Electrolytes were pumped at a rate of 43.7 cm 3 min À1 (50 rpm), which corresponded to an average velocity of %1.0 cm s À1 through each porous electrode.For an initial concentration of 0.1 mol dm À3 of active species and this cell, these flow rates equaled a stoichiometric factor of 43.9 at a constant current density of 40 mA cm À2 .For experiments with an initial concentration of 0.5 mol dm À3 , the stoichiometric factor was 219.6 in the same conditions.Both electrolytes were purged with humidified nitrogen for 1 h before initiating the experiments.
Charge-Discharge Cycling Protocols: Cell cycling was performed in all cases with an Interface 5000E potentiostat-galvanostat (Gamry Instruments) in two-electrode mode.Flow battery characterizations were carried out first at "low concentration" then at "high concentration" to reveal effects closer to industrially relevant conditions. [41]"Low concentration" experiments were conducted by coupling 15 cm 3 of 0.1 mol dm À3 [(DMAE-Pr) 2 -Vi]Cl 4 in 1 mol dm À3 KCl as the limiting negolyte and 30 cm 3 of 0.1 mol dm À3 FcNCl in 1 mol dm À3 KCl as the non-limiting posolyte.The "high concentration" experiments were similar, except that the concentration of [(DMAE-Pr) 2 -Vi]Cl 4 was 0.5 mol dm À3 in 1 mol dm À3 KCl and the concentration of FcNCl was 0.5 mol dm À3 in 1 mol dm À3 KCl.
The effect of current density on the EE, Coulombic efficiency (CE), and accessible capacity of the flow battery was evaluated through galvanostatic charge-discharge cycling.A range of 20-120 mA cm À2 was considered in steps of 20 mA cm À2 between cell voltage limits of 0.6 and 1.2 V. Values EE, CE, and accessible capacity were the average of 10 charge/ discharge cycles.A cell potential hold was not implemented in this case, replicating common operation at large scale.However, 20 "stabilizing cycles" were added before the current density loop to avoid the effects of electrolyte and water uptake to the pristine membrane.The ASR of the cell was determined before and after this procedure from an EIS measurement performed over a frequency range from 0.01 Hz to 10 MHz and an amplitude of 10 mV.
The capacity fade rate of the flow battery in laboratory conditions was determined from a galvanostatic charge/discharge regime at 40 mA cm À2 implementing a constant potential step at the end of every charge or discharge step until the total cell current fell below 2 mA.Cell cycling was carried out between the same 0.6 and 1.2 V cell voltage cutoffs.The capacity over time and number of cycles were taken from the final slope resulting from the cell and membrane reaching quasi-steady state conditions.Energy and Coulombic efficiencies were reported for each full chargedischarge cycle and not for steady state of charge (SOC) values.In agreement with the recommended reporting practice, all the capacity fade rates are reported as a function of time. [42]ostmortem NMR Analysis: The spent negolyte was analyzed by 1 H-NMR spectroscopy.An aliquot of spent electrolyte was diluted to 500 μL with D 2 O, which helped the lock of the solvent peak.The measurements were carried out in a high-quality 528 NMR tube (Wilmad) using the 600 MHz spectrometer described earlier.

Figure 3 .
Figure 3. Performance of a [(DMAE-Pr) 2 -Vi]Cl 4 /FcNCl flow battery with a concentration of redox couples of 0.1 mol dm À3 .Effect of the current density on a) voltage and accessible volumetric capacity (1st and 10th cycle reported for each value), and b) efficiencies (Coulombic and energy) and accessible volumetric capacity (10 cycles for each value).Flow battery demonstration at 40 mA cm À2 plus potentiostatic hold c) accessible volumetric capacity versus time and cycle number, and d) voltage and volumetric capacity at selected cycles.

Figure 4 .
Figure 4. Performance of a [(DMAE-Pr) 2 -Vi]Cl 4 /FcNCl flow battery with a concentration of redox couples of 0.5 mol dm À3 .Effect of the current density on a) voltage and accessible volumetric capacity (1st and 10th cycle reported for each value), and b) efficiencies (Coulombic and energy) and accessible volumetric capacity (10 cycles for each value).Flow battery demonstration at 40 mA cm À2 plus potentiostatic hold c) accessible volumetric capacity versus time and cycle number, and d) voltage and volumetric capacity at selected cycles.