Blatter Radicals as Bipolar Materials for Symmetrical Redox-Flow Batteries

Redox-active organic molecules are promising charge-storage materials for redox-flow batteries (RFBs), but material crossover between the posolyte and negolyte and chemical degradation are limiting factors in the performance of all-organic RFBs. We demonstrate that the bipolar electrochemistry of 1,2,4-benzotriazin-4-yl (Blatter) radicals allows the construction of batteries with symmetrical electrolyte composition. Cyclic voltammetry shows that these radicals also retain reversible bipolar electrochemistry in the presence of water. The redox potentials of derivatives with a C(3)-CF3 substituent are the least affected by water, and moreover, these compounds show >90% capacity retention after charge/discharge cycling in a static H-cell for 7 days (ca. 100 cycles). Testing these materials in a flow regime at a 0.1 M concentration of the active material confirmed the high cycling stability under conditions relevant for RFB operation and demonstrated that polarity inversion in a symmetrical flow battery may be used to rebalance the cell. Chemical synthesis provides insight in the nature of the charged species by spectroscopy and (for the oxidized state) X-ray crystallography. The stability of these compounds in all three states of charge highlights their potential for application in symmetrical organic redox-flow batteries.


■ INTRODUCTION
The increased adoption of renewable energy sources such as wind and solar, which are inherently only available on an intermittent basis, necessitates cheap and scalable storage solutions in order to ensure a stable future energy supply. 1 Of the various battery technologies currently available, redox-flow batteries (RFBs) are attractive for large-scale stationary applications due to their scalability, decoupling of power and energy, and simplicity of design. 2 Current RFB technologies that are reaching commercial applications are based on the redox chemistry of transition-metal ions (e.g., vanadium or iron), 3 but widespread implementation is hampered, among others, by low volumetric charge densities, concerns regarding environmental impact, and high system cost. Electrolytes that are based on less harmful, earth-abundant materials are a prime target for next-generation RFBs. In recent years, there has been a surge in the development of organic charge-storage materials targeting both aqueous and nonaqueous electrolytes, but the long-term stability of organic compounds, during either battery cycling or storage, has proven difficult to achieve and material decomposition is an important factor in the capacity fade. 4 The vast majority of current organic RFB designs utilize different redox-active compounds on both the positive and negative sides of the battery. Such asymmetric electrolyte compositions impose stringent selectivity requirements on the separator membrane, as crossover of the electrolyte results in irreversible capacity fade. 5 Several ways to minimize active material crossover have been put forward: for example with improved membrane architectures, 6 with size exclusion, 7 or using immiscible electrolytes. 8 Conceptually, the use of symmetrical cell designs (with identical posolyte/negolyte composition in the discharged state) presents an additional strategy to mitigate the effect of membrane crossover. 9 Organic RFBs with symmetrical battery chemistries have been prepared by combining two redox-active units via a linker ("combimolecules"), 10 using bipolar small-molecule compounds with complementary redox units, 9a,11 or using helical carbenium ions. 12 In addition to closed-shell compounds, 13 several classes of stable organic radicals are known to possess bipolar electrochemistry and are thus of potential interest as battery materials. 14 Examples of bipolar organic radicals that have been explored to date include nitronyl nitroxides 15 and verdazyls, 16 but these systems show relatively rapid degradation 17 in the charged state and high capacity retention has yet to be achieved. Here we show that 1,2,4-benzotriazin-4-yl radicals (commonly known as Blatter radicals) 18,19 combine tunable redox potentials and high stability in all relevant redox states necessary for application as redox-active components in RFBs. The cell potentials of derivatives with a CF 3 substituent on the triazinyl ring are especially insensitive to water (up to 33 vol % of water in acetonitrile), which holds promise for this class of materials in energy storage systems based on earth-abundant elements.

■ RESULTS AND DISCUSSION
The 1,2,4-benzotriazin-4-yl radicals described here were prepared on the basis of literature procedures. The parent Blatter radical (1; Scheme 1) was synthesized via the catalytic oxidation of N-phenyl amidrazone. 20 Derivatives with Nsubstitution at C(3) (2a,b) were synthesized by hydrolysis of Nitron, 21 whereas the Cu-mediated coupling route by Koutentis et al. 22 provided access to C(3)−CF 3 radicals that differ in the N(1) aromatic group (3a−h). Radical 3c was obtained in quantitative yield by Pd/C-catalyzed hydrogenation of the p-NO 2 group in 3f. The solid-state structures of 3c,h were determined by single-crystal X-ray diffraction. As expected for Blatter radicals, 23 all compounds (except 3g; vide infra) 24 showed cyclic voltammetry in anhydrous acetonitrile that is consistent with bipolar behavior: both oxidation and reduction to form the corresponding closed-shell cation and anion are reversible, indicating that the three different states of charge are chemically stable, at least on the time scale of voltammetric experiments. The redox potentials for the series of compounds in acetonitrile are reported in Table 1. A comparison of the redox potentials of compounds 1, 2a,b and 3a, which differ in the C(3) substituent, shows that both couples are sensitive to the electron-donating/-withdrawing properties of the group present at C(3), but not by the same amount. This allows tuning of the (calculated) cell potential, E cell = E 1/2 (x 0/+ ) − E 1/2 (x 0/− ), and leads to the highest value (E cell = 1.12 V) for the derivative with an electron-withdrawing CF 3 group at that position (3a). An analysis of diffusion coefficients (D) and standard rate constants (k 0 ) from the CV data at different scan rates (using Randles−Sevcik and Nicholson methods, respectively) gives similar values for both redox couples in this series of bipolar Blatter radicals (D ≈ 10 −5 cm 2 s −1 ; k 0 ≈ 10 −2 cm s −1 ; see Table S5). These values are comparable to those of other nonaqueous organic RFB systems. 25 Within the series of CF 3 Blatter radicals 3 with varying N(1)−Ar para substituents, the presence of electron-donating groups (3b, OMe; 3c, NH 2 ) shifts both redox potentials to more negative potentials, as expected, and the opposite is observed for electron-withdrawing groups (3d, F; 3e, CF 3 ; 3f, NO 2 ; see Table 1). Moreover, the introduction of 2,6-dichloro substituents on the N(1)−Ar ring provides an additional increase in the calculated cell potential of ca. 0.2 V to give an E cell value of 1.32 V for compound 3h.
Cyclic voltammetry with stepwise addition of up to 33 vol % (ca. 18.5 M) of H 2 O to the acetonitrile evaluated the effect of water on the redox chemistry of Blatter radicals. A comparison of the voltammetry of 1 and 3a shows that in both cases the oxidation of the radicals to the cations is only marginally affected by the addition of water (Figure 1), and this extends to the other compounds (see the Supporting Information). The reduction to the anions, E 1/2 (x 0/− ), on the other hand, shifts to more positive potentials, and the overall result is a decrease of 0.54 V in the maximum calculated cell potential for 1 (E cell = 0.51 V in a 2/1 MeCN/H 2 O mixture). For the CF 3substituted compounds the redox potentials of the 3 0/− couple, on the other hand, are substantially less sensitive to water and larger cell potentials are retained (up to 1.0 V for 3h; see the Supporting Information). We attribute the change in E 1/2 (x 0/− ) to the stabilizing effect of hydrogen-bonding interactions on the reduction products. 26 The observation that this shift depends on the nature of the C(3) substituent (and is smallest for the CF 3 -substituted radicals 3) suggests that the Brønsted basicity of the closed-shell anions 3 − is an important criterion. Recent work by Wang and co-workers also demonstrated the importance of pK a on reversibility in fluorenone-type RFB negolytes, 27 and it thus appears that    Table S3. b The reduction of 3g is irreversible, and the peak potential is given. molecular engineering can be used to further optimize the cell potential in bipolar Blatter radicals. Radical 3g, which has a carboxylic acid substituent, was sufficiently soluble to record its cyclic voltammogram in water alone (∼1 mM 3g; 1 M KCl as supporting electrolyte). Under those conditions the CV of 3g shows a somewhat broadened reduction wave, but both redox processes are (quasi-)reversible with E cell ≈ 0.60 V (Figure 2). Addition of Na 2 CO 3 to pH 10 increased the solubility of 3g (as the deprotonated carboxylate form) and a better defined, cathodically shifted 3g 0/− redox couple was obtained. The oxidation to the cation 3g + was irreversible, indicating that OH − has a detrimental influence on the stability of the oxidized state. This is similar to the instability observed for quinones, for which nucleophilic attack by H 2 O or OH − has been identified as one of the key decomposition pathways. 28 The battery performance of Blatter radicals was evaluated by cycling in a stirred H-cell with a porous glass frit. The volume of electrolyte used on each side was 5 mL, and the solutions were magnetically stirred to ensure efficient mixing (see the Supporting Information for details). A symmetrical ("poleless") battery was constructed using 1 (12 mM) as the chargestorage material in anhydrous MeCN with 0.3 M [Bu 4 N][PF 6 ] as the supporting electrolyte, which corresponds to a theoretical capacity of 1.6 mAh. The battery was charged/ discharged (current ±1.6 mA, cutoff voltages of 1.8 and 0.1 V) in cycles of 1 h, and the changes in capacity were monitored over the course of several days ( Figure 3A). Although the coulombic efficiency remains high throughout (>99%), a gradual decrease in capacity is observed such that after approximately 10 days (157 cycles) the remaining battery capacity is ca. 80% of the initial value (average discharge capacity fade of 2.0% per day). When a similar battery containing 2a (9 mM, current ±1.22 mA) as the electroactive component is cycled, the capacity drops rapidly in the first cycles to ca. 60% and then continues to fade at a rate that is still appreciably higher than that for 1 (see Figure S57). Visual inspection of the H-cell after the first charge/discharge cycle shows the posolyte and negolyte to be different, suggesting that the system does not return to a symmetrical composition in the discharged state. UV−vis absorption spectra of both sides (in the discharged state) after cycling for 44 h (27 cycles) shows that the posolyte remains unchanged, but the starting material 2a in the negolyte solution is fully consumed. The broad absorbance maximum of 2a (centered around 550 nm) is replaced by a new band at λ max = 596 nm, which is indicative of the formation of 2b on the basis of a comparison with an authentic sample (see Figure S58). Although the conversion of 2a to 2b under strongly alkaline conditions has been reported, 21 these results suggest that, despite the stability of 2a on the CV time scale, reduction to the closed-shell anion 2a − results in its decomposition to 2b and perhaps other, not yet identified, products. Starting with pure compound 2b (12 mM, current ±1.6 mA, cutoff voltages of 1.7 and 0.1 V) in both electrolyte compartments leads to a discharge capacity fade rate that is gradual but somewhat larger than in 1 (3.6% per day).
The CF 3 -substituted Blatter radical 3a (12 mM in MeCN) was cycled in an H-cell for a total of 20 days (277 cycles; current ±1.6 mA, cutoff voltages of 1.8 and 0.1 V). This resulted in a high capacity retention of >94%, corresponding to an average fade rate of only 0.3% per day, or 0.02% per charge/ discharge cycle ( Figure 3A). Similar results were obtained for the other derivatives 3, highlighting the remarkable stability of this class of Blatter radicals. The only exception is 3h, which loses almost 40% of the capacity in the first week of cycling and then shows accelerated decomposition to reach 0% after ca. 12 days (see Figure S68). The p-NH 2 -substituted derivative 3c shows even higher stability in comparison to 3a, and virtually no decay is discernible after 111 charge/discharge cycles (7 days; see Figure S64). The "poleless" nature of a battery using 3a (6 mM in MeCN) as the active material was demonstrated by carrying out a cycling experiment in which polarity inversion was applied every 5 cycles, showing 98.5% capacity retention after 25 cycles (37.6 h) ( Figure 3C,D and Figure  S69). Furthermore, the H-cell experiment was repeated at a higher concentration of active material (∼4-fold increase, 50 mM of 3a), which also resulted in good capacity retention, indicating that higher concentrations of the active material do not result in a significantly faster capacity fade ( Figure S70).
A comparison of the cycling data for 3a−f at 12 mM concentration during a 7-day experiment shows that the nature of the N(1)−Ar para substituent has no (detrimental) influence on the stability, suggesting that this is an ideal position to engineer desirable properties such as (aqueous)  Daramic membrane is a nonselective porous separator that has low ohmic resistance but allows crossover of both charged and neutral species relatively readily. At the current density used (35 mA/cm 2 ), we were able to access charge capacities close to the theoretical value and ca. 90% was recovered upon discharge, in line with the relatively low coulombic efficiencies due to crossover-induced self-discharge when a microporous separator was used. 30 To evaluate whether the decrease in capacity observed in the first 50 cycles could be the result of volume and/or concentration imbalances between the two electrolytes, we inverted the polarity of the battery at that point and ran another 50 charge/discharge cycles. As shown in Figure 4, this indeed results in capacity recovery (to 99.5% of the initial capacity utilization at cycle 75) and demonstrates that polarity inversion in these bipolar systems can be used to rebalance the cell in a greatly simplified manner in comparison to physical rebalancing methods that require battery disassembly. 5b, 30 It should be noted that, in contrast to "compositionally symmetrical cells" prepared by premixing of (different) posolyte and anolyte materials, 31 the use of intrinsically bipolar materials such as 3a allows the utilization of all electroactive material in solution. Potentiostatic electrochemical impedance spectroscopy (PEIS) measurements of the flow cell before and after cycling did not indicate a significant increase in resistance. With an overall capacity retention of 98% after 100 cycles (0.02% capacity fade per cycle, 1.5% capacity fade per day) and average coulombic, voltaic, and energy efficiencies of 89%, 67%, and 60%, respectively, the flow cell studies demonstrate the excellent stability of 3a in all three states of charge. The battery performance is similar to that of other symmetrical RFBs reported in the literature; however, the battery outperforms most RFBs with regard to cycling stability (Table S6).
While neutral Blatter radicals have been studied in a variety of scientific fields (e.g., molecular electronics, magnetochemistry, biochemistry), 19 the chemical stability of these compounds in the oxidized and reduced state, which is of key importance in energy storage applications, has received scant attention in the literature. This aspect was explored by treating radical 3a on the NMR scale (CD 3 CN solution) 4 ] suitable for X-ray diffraction ( Figure 5). A comparison of the metrical data (see Table S3) of radical 3a 22 and the cation 3a + indicates that, while both have a planar heterocyclic core, oxidation is accompanied by a pronounced decrease in the N1−N2 bond length from 1.370(2) Å in 3 to 1.318(1) Å in 3a + , and also the C−N bonds to the annulated ring shorten to 1.359(2) Å (C7−N1) and 1.341(2) Å (C2−N3) in 3a + (in 3: 1.390(2) and 1.387(3) Å, respectively). Together with the long/short bond alternation in the annulated C 6 ring, these metrical parameters Journal of the American Chemical Society pubs.acs.org/JACS Article validate that 3a + has 10π aromatic character analogous to that of naphthalene and its derivatives. 32 In a similar manner, the anion 3a − (formally a 12π antiaromatic compound) was obtained in quantitative yield on the NMR scale by reduction with Na/Hg in CD 3 CN (Scheme 2). As expected for a more electron rich compound, the 1 H NMR resonances were shifted upfield by ca. 1.5 ppm to δ 6.1−7.3 ppm ( 19 F NMR: −72.3 ppm). When they are kept in a Teflon-sealed NMR tube under an N 2 atmosphere, both of these charged species are stable and show no sign of decomposition over the course of 7 days. In agreement with the battery cycling data, the NMR spectra also remain unchanged upon addition of H 2 O to either 3a + or 3a − . Compound 3a + is also stable under ambient conditions, whereas 3 − is quickly oxidized in air. UV−vis absorption spectra of the series 3a −/0/+ show that the broad longwavelength bands in 3a that give rise to the characteristic red color (λ max = 479 nm, shoulder extending to ca. 600 nm) are blue-shifted upon oxidation/reduction to λ max = 420 and 402 nm for 3a + and 3a − , respectively. The changes in the electronic absorption spectra make this system suitable for online battery monitoring, which is useful going forward to obtain molecular insight into performance characteristics in more extensive flow studies.

■ CONCLUSIONS
In conclusion, we have shown that 1,2,4-benzotriazin-4-yl radicals combine bipolar electrochemistry with high chemical stability in all states of charge, which renders these compounds promising energy-storage materials for symmetric redox-flow batteries. In particular, the derivatives with a C(3)−CF 3 substituent (compounds 3) show minimal capacity fade during battery cycling; the stability of the charged states (3a + and 3a − ) was confirmed through an independent chemical synthesis. Cyclic voltammetry and cell cycling studies in aqueous MeCN solution indicate that the electrochemical reversibility and chemical stability is not affected by the presence of water, but the reduction potential is anodically shifted to decrease the (theoretical) cell potential. The data presented here indicate that this is due to hydrogen-bond stabilization of the anionic reduction products by H 2 O. The E cell drop can be (partially) mitigated by decreasing the Hbond acceptor character of the anions, providing acceptable (∼1 V) cell potentials also in aqueous electrolytes. In this regard, it is noted that a high cycling stability is retained regardless of the para substituent on the N(1)−Ar ring, suggesting that this position is a prime target to further engineer desirable properties, including charged groups for water solubility. Testing a representative radical (3a) in a flow regime at 0.1 M concentration confirmed the high cycling stability and demonstrated that polarity inversion in a symmetrical flow battery may be used to rebalance the cell. While further research is needed to investigate the long-term stability, our work provides a proof of concept for the use of Blatter radicals in electrochemical energy-storage applications and a stepping stone toward innovative all-organic, symmetrical battery chemistries.

■ EXPERIMENTAL SECTION
Cyclic Voltammetry. Cyclic voltammetry (CV) was performed using a three-electrode configuration comprising of a Pt-wire counter electrode, a Ag/Ag + (0.01 M AgPF 6 in 0.1 M [Bu 4 N][PF 6 ]/ acetonitrile) junction reference electrode, and a GC-disk working electrode (CHI104, CH Instruments, diameter 3 mm). The GC working electrode was polished before the experiment using an alumina slurry (0.03 μm), rinsed with distilled water, and subjected to brief ultrasonication to remove any adhered alumina microparticles. The CV data were referenced to ferrocene in acetonitrile. For measurements in H 2 O, an SCE reference electrode was used.
H-Cell Cycling. Charge−discharge tests of the Blatter radicals were performed in a custom H-cell with high surface area to volume ratio. The cell consisted of two electrolyte chambers separated by a glass frit (porosity 5, ∼1.6 cm 2 ) to minimize crossover. Reticulated vitreous carbon (Duocel, 45 ppi) was used as electrodes with an interelectrode distance of about 20 mm. Generally, for the battery tests, the electrolyte chambers were loaded with 5 mL of 12 mM active species in 0.3 M [Bu 4 N][PF 6 ]/MeCN and stirred continuously at 1400 rpm. The current was set such that theoretical charging and discharging times were 1 h each (1C rate). Current densities were estimated using the membrane area size. An ohmic resistance of about 250 Ω was measured in all tests.
Flow-Cell Cycling. Measurements under flow conditions were carried out using a zero-gap flow cell. 29 The oven-dried battery was assembled outside the glovebox. A combination of a graphite chargecollecting plate and two layers of a nonwoven carbon paper electrode with an area of 2.55 cm 2 (Sigracet 29AA) was put on either side of the flow cell. A ±10% compression of the felt was achieved by the use of Gore-tex ePTFE gaskets. The two half-cells were separated by a Daramic 175 porous membrane. The gasket window provided for an exposed area of the membrane which was used as the active area of the flow cell. The cell was connected to a peristaltic pump (Cole-Parmer) by Masterflex C-flex Ultra pump tubing using a flow rate of 20 mL/min. The catholyte and anolyte reservoirs were filled with 6 mL of a solution of redox materials and electrolyte salt in the reported concentrations. Before the measurement was started, the cell was pretreated by flowing the solution through the cell for 30 min. Once the membrane was fully wetted as evidenced by impedance measurements, the cycling was started. Galvanostatic charge/ discharge cycling was performed using currents of ±89.25 mA (±35 mA/cm 2 ) for the batteries shown in Figure 4 and Figure S71 with potential cutoffs at +1.7 and 0 V. Potentiostatic electrochemical impedance spectroscopy (PEIS) measurements were performed at various stages of charge from 500 kHz to 50 Hz using a 10 mV sine perturbation. A polarization measurement was collected at full SOC and ranging from −12.75 to −255 mA. Energy efficiencies (EEs) were determined by the ratio of the time-integrated output and input power density during discharging and charging over each cycle. The voltage efficiency (VE) was then determined from the EE by dividing by CE.
Additional experimental details and characterization data (PDF)