“Innocent” Hexafluorophosphate Salts Induce Capacity Fade in Nonaqueous Redox Flow Batteries

Abstract

The use of organic active materials in redox flow batteries (RFBs) presents a promising approach to sustainable large-scale energy storage. However, the stability of nonaqueous organic RFB electrolytes is generally limited by degradation reactions that cause capacity fade. These reactions are commonly thought to convert redox-active organics to products that are no longer electrochemically active. Here we uncover an additional pathway leading to capacity fade that involves the supporting electrolyte salt. Capacity fade in nonaqueous RFBs is studied in detail for the 1,2,4-benzotriazin-4-yl radical (1) as a model compound and extended to several other classes of representative redox-active organics. By using symmetrical batteries (1 ^0/– ∥1 ^0/+ ), we delineate that capacity fade occurs in a nonlinear (autocatalytic) fashion via acid-induced decomposition of the supporting salt anion PF₆ ^– in the posolyte solution. This is shown to be a universal degradation reaction in the posolyte of nonaqueous RFBs. Although the acidic degradation products are not detrimental to the posolyte, the crossover of acid to the opposite compartment leads to capacity-limiting protonation of the negolyte active material. Replacement of PF₆ ^– with other anions substantially improves the stability of these nonaqueous electrolytes, as demonstrated with a symmetrical RFB based on 0.38 M active material 1 that can be cycled for >69 days with very high capacity retention (fade rate of ≤0.1% per day). The improved understanding of factors determining the lifetime of nonaqueous electrolytes unlocks rational strategies to develop more durable electrochemical energy storage systems.

Introduction

The intermittent nature of renewable energy from wind and solar requires large-scale energy storage to regulate the stability of the grid. , Redox flow batteries (RFBs) are a promising technology in this regard, but the most mature RFBs use vanadium-based electrolytes that are expensive and cause environmental concern. Organic molecules are promising candidates to replace the critical active material in vanadium RFBs, as they do not contain scarce elements and can potentially be produced in a more cost-effective and sustainable manner. − In recent years, both posolyte and negolyte chemistries have been developed for nonaqueous electrolyte solutions that result in high cell voltages and thus high (theoretical) energy densities. − However, a major challenge is that the lifetime of organic redox-active materials (RAMs) is generally limited due to decomposition reactions that convert the active materials to degraded products that can no longer be charged/discharged. While active material degradation pathways have been established for select RFB chemistries, , it is difficult to generalize this knowledge to newly developed active materials. The complexity of interactions within the electrolyte solution (e.g., solvation, ion pairing, and reactions with cell materials such as membrane, electrode, etc.) may obscure the origin of the capacity fade, and nonaqueous RFB electrolyte stability remains poorly understood at a molecular level.

Nonaqueous RFBs often use tetraalkylammonium salts with noncoordinating anions to facilitate cell depolarization. In particular, the use of hexafluorophosphate (PF₆ ^–) is widespread due to the superior conductivity and wide electrochemical stability window of R₄NPF₆ salts in organic solvents. While several groups have reported that the stability of (charged) RAMs is markedly dependent on the nature of the supporting electrolyte salt, the reasons for these differences and the implications for battery durability are not well understood. − In this work, we study the lifetime of nonaqueous RFB electrolytes and reveal how the supporting electrolyte salt plays a major role in capacity fade. Our group previously reported that organic radical 1 is a promising candidate for symmetrical RFBs due to its reversible bipolar electrochemistry (Scheme ). Such bipolar materials provide the opportunity to construct symmetrical batteries with an identical posolyte and negolyte (in the discharged state), which mitigates crossover-induced capacity fade. , Here we exploit the bipolar character of 1 as a “reporter” for battery stability via the sensitivity of the 1 ^0/– redox couple to electrolyte degradation, providing new insight into the molecular processes underlying capacity fade. We demonstrate that decomposition of PF₆ ^–, a widely used supporting anion, is a critical contributor to electrolyte instability across various organic RFB chemistries and show how electrolyte design can extend the lifetime of symmetrical RFBs based on 1. These findings point to the broader relevance of electrolyte salt stability across nonaqueous electrochemistry, from energy storage to organic electrosynthesis and electrocatalysis.

1. Structure of Bipolar Radical 1 and Its Charged States (1⁺ and 1^– ), with Redox Potentials (vs Fc^0/+) of Both Half-Reactions Used in the Symmetrical Battery.

Results and Discussion

Long-Term Cycling Stability of Bipolar Radical 1

To investigate the long-term cycling stability of radical 1, we ran H-cell cycling experiments in MeCN using 6 mM 1 with 0.3 M Bu₄NPF₆ as the supporting salt. In agreement with our previous data, such a battery shows excellent capacity retention for at least 3 weeks. However, prolonged battery cycling results in the onset of capacity fade after ca. 25 days, which then rapidly increases in rate (Figure A). Analysis of these electrolyte solutions by cyclic voltammetry (CV) showed the disappearance of the redox wave at −1.03 vs Fc/Fc⁺, and new peaks appear at ∼0 V (Figure B). From this, we infer that loss of the negolyte half-reaction is responsible for the capacity fade. We attempted to identify the chemical decomposition products of 1 by mass spectrometry, but the only peaks detected in the positive ion mode were at m/z = 276 and 277, which are due to the ions 1 ⁺ and [1 + H] ⁺ that are also observed in the mass spectrum of pristine 1 (Figures S1 and S2). Similar results were obtained when using KPF₆ as the electrolyte salt, albeit the onset of capacity fade occurred already after a lag phase of ca. 4 days with full degradation after ca. 14 days (Figure A). The increasingly rapid capacity fade suggests that (auto)­catalysis may take place in both systems, but the onset is dependent on the nature of the cation of the supporting salt (K⁺ vs Bu₄N⁺; vide infra).

1.

(A) Capacity as a function of time for the H-cell cycling experiment in MeCN using 6 mM 1 with 0.3 M Bu₄NPF₆ or 0.3 M KPF₆ as the supporting salt. (B) CV of postcycling electrolyte solutions (Bu₄NPF₆ supporting salt). (C and D) CV of 6 mM 1 in 0.3 M Bu₄NPF₆/MeCN with added formic acid (FA) and p-toluenesulfonic acid (pTsOH), respectively.

The disappearance of the negolyte (1 ^0/– ) redox couple (E₂) in the CV of the degraded battery electrolytes was accompanied by an increase in the current for the posolyte half-reaction (E₁). This resembles the sequential one-electron electrochemistry of quinones (Q^2–/Q^•–/Q) in nonaqueous solvents: hydrogen bonding with protic additives results in a gradual shift of the negative redox wave, ultimately merging into a single [2e/2H⁺] process (H₂Q/Q). − Similarly, our postcycling CVs indicate that the formation of protic compounds affects the negative half-reaction of 1, resulting in the loss of its bipolar character and thus capacity fade. As shown in Figure C,D, control experiments with 1 in the presence of protic additives confirm the sensitivity of the 1 ^0/– couple, the extent of which depends on the pK _a of the additive. While a relatively weak acid (formic acid; pK _a ≈ 20.9 in MeCN) shifts and broadens the negative redox wave, the strong acid pTsOH (pK _a = 8.5 in MeCN) effectively eliminates the 1 ^0/– couple when the concentrations of pTsOH and 1 are identical, at which point a single two-electron [2e/H⁺] process is obtained.

In contrast to quinones (Q^2–/Q^•–/Q), the Blatter radical is positively charged in its most oxidized state (1 ⁺ ), and protonation to 1H ²⁺ does not occur as shown by the invariance of the posolyte redox reaction (E₁) with added acid. Thus, the electrochemistry of 1 is best described by the wedge scheme shown in Scheme . Deviations from the purely electron-transfer steps (black arrows) by hydrogen bonding (green) and/or protonation (red) will result in a loss of battery performance even when the active material remains otherwise intact, as demonstrated below.

2. Proposed Electrochemical Pathways for 1 in MeCN (Black; Redox Potentials E₁ and E₂), in the presence of Hydrogen Bond Donors (Green; E₁ and E₃), and with Strong Acids (Red; [2e/H⁺] Redox Reaction with E₄ ≈ E₁).

The formation of acid in the degraded electrolytes was also indicated by the appearance of a hydrogen evolution wave in the CV at potentials below −1.3 V (Figures S3 and S4) and confirmed by measuring the pH after dilution of an aliquot with deionized water. While a pristine electrolyte solution resulted in a pH of 8 using this procedure, the postcycling posolyte and negolyte from the cell with Bu₄NPF₆ have a pH of around 2. Moreover, the shift in E₂ due to the presence of acid is reversible, as shown with the addition of a stoichiometric amount of strong base (Bu₄NOH) to a solution of 1 with formic acid (Figure S5).

Further evidence for the formation of acid was obtained by NMR analysis of the electrolytes using ¹⁹F NMR spectroscopy. The spectra of a fully faded H-cell battery with Bu₄NPF₆ as the supporting salt showed new fluorine-containing compounds in both the posolyte and the negolyte (Figure ). In addition to the PF₆ ^– signal, the ¹⁹F NMR spectrum shows an additional doublet at around δ −83 ppm with a coupling constant (J _PF) of ∼950 Hz attributed to the fluorophosphate anion PO₂F₂ ^– , and a signal at δ −151.7 ppm due to BF₄ ^–. The assignment of these species was confirmed by ³¹P and ¹¹B NMR spectroscopy, respectively (Figure S6). Both postcycling electrolytes also show resonances attributed to fluorosilicate (SiF₆ ^2–; δ −138 ppm), and HF (∼δ −180 ppm). Finally, the negolyte solution contains the leuco-form of the Blatter radical (1H; CF₃ group at δ −72.1 ppm), which is formed by protonation of the negolyte charged state, 1 ^– (Scheme ). From this, it is clear that decomposition of the supporting salt is involved in the capacity fade of 1. The same decomposition products were observed in the ¹⁹F NMR spectra of H-cell batteries with KPF₆ as the supporting salt (Figure S7). The only exception is SiF₆ ^2–, as its K⁺ salt is insoluble in organic solvents. , The KPF₆ electrolyte solution contains PO₃F^2– as a further degradation product.

2.

¹⁹F NMR spectra of negolyte (top) and posolyte (bottom) solutions taken from a fully faded H-cell battery of 6 mM 1 in 0.3 M Bu₄NPF₆/MeCN.

Decomposition of PF₆ ^– has been extensively investigated in the context of Li-ion battery electrolytes (i.e., LiPF₆ in organic carbonate solvents) and is thought to involve PF₅ generated in the equilibrium LiPF₆ ⇌ LiF + PF₅. While our work was in progress, Marbella et. al reported that PF₆ ^– stability in organic carbonates is correlated to the Lewis acidity of the cation. Based on this, it is expected that KPF₆ and Bu₄NPF₆ are stable due to the weak or nonacidic nature of K⁺ and Bu₄N⁺, respectively. This is in line with the fact that Bu₄NPF₆ is a common salt for nonaqueous electrochemistry and RFB research, and its role in electrolyte degradation and capacity fade has, to the best of our knowledge, not been previously considered. Despite its perceived innocence, our data indicate that acidic compounds are formed due to decomposition of the supporting electrolyte anion (PF₆ ^–) and that the negolyte becomes capacity-limiting due to protonation of 1 ^– . The appearance of boron- and silicon-fluorides further suggests that reactions with SiO _x and BO _x groups at the surface of the cell material (e.g., borosilicate) are involved in capacity fade.

Investigation of Cycling Stability for Each Half-Reaction

The stability of each half-reaction was investigated to identify on which battery side PF₆ ^– degradation originates. A solution containing a 1:1 mixture of 1 ^– and 1 (the negolyte half-reaction) was electrochemically generated as described in the Supporting Information and divided between the two compartments of an H-cell. Battery charge/discharge with this electrolyte composition allows the molecular stability of the 1 ^0/– couple to be studied. , H-cell cycling of this half-reaction in 0.3 M Bu₄NPF₆/MeCN showed a stable capacity for >140 days, and no electrolyte decomposition is observed by NMR spectroscopy. Also, the postcycling CV shows that the bipolar properties of 1 are retained (Figure S9). Similar stability was observed for 1 ^– in the KPF₆ electrolyte (Figure S10).

In stark contrast, a similar H-cell cycling experiment with only the posolyte half-reaction (i.e., a 1:1 mixture of 1 and 1 ⁺ ) in KPF₆ electrolyte was stable during the first 30 days, but after that, the capacity started to increase (Figure A, black line). We attribute the capacity increase to the contribution of a [2e/H⁺] redox reaction due to the presence of a strong acid (vide supra). A CV taken after 33 days indeed confirms that acid is present in the solution: the 1 ^0/– couple is completely absent, and the 1 ^0/+ peak current has increased (Figure S11). ¹⁹F NMR spectra of the electrolyte solution after 43 days of cycling contain substantial amounts of PF₆ ^– decomposition products (fluorophosphates, fluoroborates, and HF). The subsequent rapid drop in capacity is due to a larger cell resistance of the H-cell (see the discharge voltage trace in Figure S11C), which we tentatively attribute to the precipitation of K₂SiF₆ onto the separator and/or the RVC electrode. As a control experiment, we deliberately added 1 mM pTsOH to a 1 ^0/+ half-cell that had been cycled stably for 15 days. This resulted in an instantaneous capacity increase due to the [2e/H⁺] process between 1 ⁺ and 1H (Figure A, red line). The lag phase observed previously, and its absence when substoichiometric pTsOH is added, together with the sigmoidal shape of the capacity profile, is consistent with an autocatalytic process.

3.

(A) Capacity changes for a 6 mM positive half-cell (1/1 ⁺ ) in 0.3 M KPF₆/MeCN, with and without the addition of pTsOH. (B) Plot of battery capacity and ratio of negolyte (1/1 ^– , i _P1) and posolyte (1/1 ⁺ , i _P2) peak currents (blue) from an in situ CV experiment monitoring both electrolyte solutions.

Next, we carried out in situ CV measurements during battery cycling to verify a direct link between acid concentration and overall battery capacity via the 1 ^0/– “reporter” unit. Charge/discharge cycling of a full battery (1 ^0/– ∥1 ^0/+ ), i.e., utilizing both half-reactions, showed that the peak current of the 1 ^0/– couple in the negolyte compartment remains constant during 25 days of cycling, in which time the capacity also does not appreciably change (Figure B; see Figures S12 and S13 for CV data). Conversely, the peak current of the 1 ^0/– couple on the posolyte side starts to drop already after ca. 10 days, i.e., it precedes the occurrence of battery capacity fade. This demonstrates that acidic compounds originate from PF₆ ^– decomposition in the posolyte, but this leads to capacity fade only upon crossover to the negolyte compartment. An observation that supports this notion is that the onset of capacity fade is dependent on the Coulombic efficiency of the H-cell used (Figure S14), which is determined by the properties (porosity and area) of the frit that separates both sides. Moreover, an experiment where the effect of pTsOH during H-cell cycling was investigated also showed that capacity fade is instantaneous when added to the negative side, but addition to the positive side is gradual due to the transfer of acid across the separator (Figure S15).

Cycling Stability in Redox Flow Battery Setup

The cycling studies were extended from H-cells to a small-scale flow battery to investigate the electrolyte stability in a glass-free setup. The electrolyte solutions were pumped from polypropylene containers by using a peristaltic pump via chemically resistant tubing (Masterflex Tygon A-60-G and Swagelok PFA) into a flow cell containing a polyethylene porous separator (Daramic 175). Achieving stable long-term cycling is more difficult in such a flow setup due to the gradual buildup of volumetric imbalances, as well as evaporation of the volatile MeCN solvent. In our symmetrical batteries, this can be mitigated by mixing and redistributing the solutions across both sides and/or adding fresh solvent. Long-term flow cycling with the same electrolyte composition (i.e., 6 mM 1 in 0.3 M Bu₄NPF₆ or KPF₆) showed a similar trend as in the H-cell experiments, with an induction period followed by rapid capacity fade, which happens earlier for KPF₆-containing electrolytes than with Bu₄NPF₆ (Figures , S16, and S17). Postcycling analysis of the solutions by CV and NMR spectroscopy indicated that the negolyte redox process is absent and similar PF₆ ^– degradation products are formed as before (PO₃F^–, PO₂F₂ ^–, and HF). Importantly, a significant amount of fluorosilicate (SiF₆ ^2–) is also observed, but fluoroborates are absent. Thus, PF₆ ^– degradation is not limited to glass cells, and it appears that silicon (SiO _x ) materials in the pump tubing, separator, or other cell components contribute to the observed electrolyte instability.

4.

Flow cell cycling of 6 mM compound 1 in 0.3 M Bu₄NPF₆/MeCN at a current of 2 mA/cm². (A) Plot of capacity change as a function of cycle time; arrow indicates mixing of both electrolyte solutions to rebalance the cell. (B) ¹⁹F NMR analysis of postcycling electrolyte (mixed).

Chemical Stability of PF₆ ^– Salts

The role of electrolyte degradation products on battery performance and longevity has been well-recognized for the solid/solution interphases in Li-ion batteries, − but their influence on electrolyte stability in nonaqueous RFBs remains virtually unexplored. To shed light on the chemical stability of PF₆ ^– in MeCN solution, we prepared 0.3 M solutions of PF₆ ^– with Li⁺, Na⁺, K⁺, and Bu₄N⁺ counterions in anhydrous MeCN in a glovebox and monitored the solution composition in the absence of an applied voltage by NMR spectroscopy. As expected, both Li⁺ and Na⁺ salts immediately contain PF₆ ^– decomposition products, which increase upon standing at room temperature (PO₂F₂ ^–, HF, and BF₄ ^–; see Figure S18). The corresponding salts with the less polarizing cations K⁺ and Bu₄N⁺ are not prone to fluoride abstraction, , and we find no indication of PF₆ ^– degradation even when 50 mM water is added and/or the solutions are heated to 70 °C for several days (Figure S19). Similarly, solutions containing 0.3 M KPF₆ or Bu₄NPF₆ in MeCN with 6 mM solutions of either neutral 1 or anion 1 ^– kept at 70 °C did not show signs of decomposition by NMR spectroscopy (Figure S20). In marked contrast, when a 0.3 M Bu₄NPF₆ solution was monitored in an NMR tube at 70 °C in the presence of 6 mM 1 ⁺ (generated electrochemically), extensive PF₆ ^– decomposition was observed with substantial amounts of PO₂F₂ ^– and HF formed already after 12 h. After 3 days, about half of the PF₆ ^– was converted to PO₂F₂ ^– (∼145 mM) as the major P-containing product, and other products in solution are fluorosilicates (71 mM) and fluoroborates (27 mM). The extent of decomposition in the corresponding KPF₆ solution is more difficult to quantify due to the formation of insoluble K₂SiF₆, but it is clear that also in this case, the cation 1 ⁺ (but not 1 or 1 ^– ) induces PF₆ ^– degradation after 12 h at 70 °C, with concentrations of PO₂F₂ ^– and HF that are in the same range as that found for Bu₄NPF₆. Moreover, both solutions contain ¹H NMR resonances indicative of acidic groups (broad resonance at δ > 8 ppm; Figure S22).

To evaluate the role of silicon-containing materials, we performed a similar electrolyte stability study using 1 ⁺ /Bu₄NPF₆/MeCN that was stored at 70 °C in a Teflon-lined NMR tube (no contact with the SiO _x surface) or an identical sample to which a small amount of powdered borosilicate (ground P5 frit) was added. Whereas the sample without SiO _x shows only a trace of degradation products (HF and fluorophosphates), the PF₆ ^– in the sample that is exposed to a SiO _x surface is extensively decomposed based on NMR analysis, including the formation of acidic groups (14.2 ppm in the ¹H NMR spectrum; see Figure S23). The phosphate products of PF₆ ^– decomposition incorporate O atoms that could originate from H₂O (i.e., hydrolysis, overall reaction (4) in Scheme ). , This appears unlikely, however, since the residual water content in our Bu₄NPF₆ and KPF₆ electrolyte solutions was determined by Karl Fischer titration to be <10 ppm. Instead, it seems that HF-initiated etching of borosilicate glass (SiO₂/B₂O₃) or other SiO _x -containing cell materials is the source of O atoms (reactions (5)–(8), Scheme ), ,, accounting also for the formation of SiF₆ ^2– and BF₄ ^–. It should be noted that complete removal of water from SiO _x surfaces is challenging, and it is likely that our drying procedures (e.g., glassware is placed in a 150 °C oven overnight) will retain at least some of these reactive O-groups.

3. Overview of Reactions That Could Contribute to PF₆ ^– Decomposition through Hydrolysis (1)–(4); Reactions with Cell Materials (SiO _x or B₂O₃) That Generate H₂O (5)–(8).

PF₆ ^– Degradation in RFB Electrolytes Is Universal

Having established that 1 ⁺ is involved in the decomposition of PF₆ ^–, we investigated whether other common organic posolyte RAMs would behave similarly. As a starting point, we cycled a ferrocene (Fc) half-cell, i.e., with both compartments containing a 6 mM 1:1 mixture of Fc and Fc⁺, , for ca. 30 days in an H-cell with 0.3 M KPF₆/MeCN. While the capacity remains constant and the CV of the solution does not change (Figure S24), the NMR spectra clearly show that PF₆ ^– decomposition occurs with the formation of acidic products as indicated by broad signals in the ¹H NMR spectrum at >8 ppm. Similar results were obtained for an H-cell cycling experiment using the phenothiazine 2 (Figure S25). Thus, both of these RAMs, previously reported to be highly stable in nonaqueous electrolytes, initiate PF₆ ^– decomposition during battery cycling. It appears that this pathway has previously escaped detection because their redox chemistry and UV–vis spectroscopic signature are insensitive to acidic degradation products from supporting salt breakdown. Additionally, nonaqueous battery cycling is rarely carried out for periods longer than a week, − and these studies may have been stopped before the onset of (autocatalytic) PF₆ ^– degradation was noticeable.

The generality of these results was extended to a series of representative RFB chemistries shown in Scheme . To rapidly evaluate the stability of electrolytes with these active materials, we prepared solutions (ca. 40 mM RAM) in anhydrous MeCN with ∼0.1 M Bu₄NPF₆, which were monitored by ¹⁹F NMR spectroscopy. As expected, solutions of the parent compounds remain unchanged after several days at temperatures of up to 70 °C. However, when the same solutions were chemically oxidized by the addition of 0.8 equiv of [NO]­[PF₆], i.e., to 80% state-of-charge, in all cases the formation of PO₂F₂ ^–, SiF₆ ^2–, BF₄ ^–, and HF was clear from the NMR spectra (Figures S26–S30). Prolonged heating to 70 °C caused the appearance of additional PF-containing compounds, which could be di- or triphosphate condensation products, but these were not further characterized. Thus, PF₆ ^– degradation is a general phenomenon for the oxidized (charged) state of typical nonaqueous RFB posolytes. Importantly, this decomposition reaction is observed regardless of whether the electrolyte solutions are prepared chemically, via bulk electrolysis, or during battery charge/discharge cycling.

4. Posolyte RAMs Tested to Establish Generality of PF₆ ^– Degradation.

Early work in the literature that examined the oxidative stability of PF₆ ^– salts in organic solvents reported the onset of anodic oxidation at >3 V vs Ag/Ag⁺, − well beyond the potentials accessed in our work; this was confirmed computationally. − These data rule out direct (electro)­chemical oxidation of PF₆ ^– as the cause of its degradation. Although our current data do not allow detailed mechanistic interpretation, it seems plausible that the oxidized (charged) state of RAMs undergoes electron transfer with SiO _x -adsorbed H₂O molecules or surface Si–OH groups to generate H⁺: this initiates fluoride abstraction from PF₆ ^– to produce HF and PF₅, the latter of which reacts with trace water to the PO₂F₂ ^– ion. Once the formation of HF and other acidic products is initiated, it leads to PF₆ ^– degradation via an autocatalytic pathway. This reactivity is not limited to MeCN solvent, as an H-cell cycling experiment carried out in DME showed similar PF₆ ^– degradation products (Figure S31). To what extent the acid-induced decomposition of PF₆ ^– will limit cycling stability in scaled-up (technical) RFB systems remains to be evaluated, but it is clear that materials selection for cell components such as pumps, tubing, and membranes will have an important influence on PF₆ ^– electrolyte stability. In this context, it should be noted that chemical compatibility of cell materials (including toward HF) is also crucial for the cycle life of Li-ion batteries. ,

Cycling Stability with Other Supporting Electrolyte Anions

To improve the long-term cycling stability of 1 in lab-scale experiments, we replaced PF₆ ^– with other anions that are known to provide good electrochemical stability and conductivity. H-cell batteries were constructed with 6 mM 1 in MeCN and tetrabutylammonium salts of tetrafluoroborate (BF₄ ^–), trifluoromethanesulfonate (OTf^–), and bis­(tri­fluoro­methane­sulfon­yl)­imide (TFSI^–) anions. , Gratifyingly, all these electrolyte compositions indeed performed much better (Figure A), without indication of the “catastrophic” capacity fade observed with PF₆ ^–. Although the initial fade rate for an H-cell of 1 with BF₄ ^– is higher than that with PF₆ ^–, it does not show autocatalytic degradation and thus has an improved overall lifetime. The most stable electrolyte was that using Bu₄NTFSI as the supporting salt, which retained 87.5% of its initial capacity after 220 days (3500 cycles). Although solvent evaporation/replenishment and mass transfer limitations (variable efficiency of magnetic stirring) cause fluctuations in the discharge capacity during long-term cycling, CV analysis indicated minimal changes to the electrolyte composition (Figure S32). The fade rate is only 0.044% per day based on a linear fit of capacity vs time data, which testifies to the remarkable stability of this nonaqueous RFB electrolyte.

5.

(A) Comparison of capacity fade for H-cell cycling experiments with different supporting salt anions (6 mM 1 in MeCN, 0.3 M Bu₄NX; X = PF₆ ^–, BF₄ ^–, TfO^–, and TFSI^–). (B) Discharge capacity and Coulombic efficiency of a high-concentration flow cell (0.38 M 1 in MeCN, 1.0 M Bu₄NTFSI, 60 mA/cm²); data are shown for each battery polarity separately, and theoretical capacity (10.2 Ah/L) is shown as a dashed line. Between days 16 and 18 (yellow), the cell was cycled at different current densities. Inset shows representative voltage traces highlighting the inversion of polarity after every cycle. Discontinuities in the CE are due to the addition of MeCN (at 0% SOC) to compensate for solvent evaporation.

Finally, we used this insight to construct a high-concentration symmetric RFB with an electrolyte composed of 0.38 M 1 in 1.0 M Bu₄NTFSI/MeCN with a theoretical capacity of 10.2 Ah/L. To minimize the buildup of volumetric imbalances, the battery was run with polarity inversion after every charge/discharge cycle (i.e., with cutoff voltages of +1.9 and −1.9 V; current density of 60 mA/cm²). This allowed stable cycling for over 2750 cycles (69 days) with the periodic addition of fresh MeCN in both electrolyte reservoirs to counteract solvent evaporation. After 16 days of cycling, we ran the battery at different current densities (20–80 mA/cm²) which all showed high material utilization (>85%). The use of a porous Daramic membrane results in low ohmic resistance (1.5 Ω from PEIS) but also relatively rapid active material crossover, which manifests as poor Coulombic efficiency at lower current densities (CE = 79% at 20 mA/cm²; energy efficiency (EE) = 65%). Importantly, the cycling stability of this symmetrical RFB is not negatively affected by crossover, and prolonged cycling at 60 mA/cm² (days 20–69) shows that the discharge capacity is very stable with a fade rate of 0.0025% per cycle or 0.1% per day (averaged over both cycling polarities). These performance metrics compare favorably to current state-of-the-art nonaqueous RFB electrolytes and even approach the stability of well-established aqueous systems. Thus, to the best of our knowledge, the electrolyte based on 1/Bu₄NTFSI provides the first symmetrical RFB that demonstrates excellent battery longevity at high active material concentration. Whether the stability gain obtained with Bu₄NTFSI outweighs its additional cost requires further technoeconomic evaluation. Nevertheless, it is clear that “electrolyte engineering” holds considerable promise to improve the lifetime of nonaqueous RFB electrolytes.

Conclusion

Our results show that the long-term cycling stability of nonaqueous RFBs is influenced by decomposition of the commonly used supporting salt anion, PF₆ ^–. By using the 1 ^0/– redox couple of bipolar radical 1 as a “reporter” of the formation of protic degradation products, we were able to delineate that PF₆ ^– is decomposed in the posolyte solution via an autocatalytic pathway, which is induced by the oxidized (charged) state 1 ⁺ . Capacity limitation is only incurred upon the crossover of these products to the negative side of the battery. These findings are general for a range of representative nonaqueous posolyte chemistries based on organic/organometallic active materials. This has important implications for the interpretation of stability studies that focus only on a single half-reaction because degradation on one side can have a major capacity-limiting effect on the opposing battery side. Among the several lifetime extension strategies that these findings enable, we show here that replacement of PF₆ ^– with alternative supporting anions significantly increases the stability of nonaqueous electrolytes. This enables long-term cycling of a high-capacity symmetrical flow battery by using Bu₄NTFSI as the supporting salt. With this optimized electrolyte, we achieved a capacity fade rate of ≤0.1% per day in both H-cell and flow battery cycling studies. The insights gained highlight the necessity of considering the reactivity of organic RFB electrolytes as a whole, i.e., beyond the active material, thereby establishing a pathway for the development of robust nonaqueous RFBs.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsaem.5c03070.

• Experimental details, characterization data, and supplementary figures, including a schematic of the H-cell cycling test setup, MS and NMR spectra, and additional electrochemical data for H-cells and flow cells (PDF)

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript.

The authors declare no competing financial interest.