Indolo[2,3-b]quinoxaline as a Low Reduction Potential and High Stability Anolyte Scaffold for Nonaqueous Redox Flow Batteries

Redox flow batteries (RFBs) are a promising stationary energy storage technology for leveling power supply from intermittent renewable energy sources with demand. A central objective for the development of practical, scalable RFBs is to identify affordable and high-performance redox-active molecules as storage materials. Herein, we report the design, synthesis, and evaluation of a new organic scaffold, indolo[2,3-b]quinoxaline, for highly stable, low-reduction potential, and high-solubility anolytes for nonaqueous redox flow batteries (NARFBs). The mixture of 2- and 3-(tert-butyl)-6-(2-methoxyethyl)-6H-indolo[2,3-b]quinoxaline exhibits a low reduction potential (−2.01 V vs Fc/Fc+), high solubility (>2.7 M in acetonitrile), and remarkable stability (99.86% capacity retention over 49.5 h (202 cycles) of H-cell cycling). This anolyte was paired with N-(2-(2-methoxyethoxy)-ethyl)phenothiazine (MEEPT) to achieve a 2.3 V all-organic NARFB exhibiting 95.8% capacity retention over 75.1 h (120 cycles) of cycling.


I. General Information
All reactions were carried out in flame-dried glassware sealed with rubber septa, under a nitrogen atmosphere with dry solvents under anhydrous conditions, unless otherwise noted. Dimethylformamide (DMF) used for synthesis was obtained by passing these previously degassed solvents through activated alumina columns under argon. Reagents were purchased at the highest commercial quality and used without further purification unless otherwise stated. Solvents for chromatography were purchased from Sigma-Aldrich. Reactions were monitored by thin layer chromatography (TLC) carried out on S-2 0.25 mm E. Merck silica gel plates (60F-254) using UV light for visualizing and aqueous ammonium cerium nitrate or basic aqueous potassium permanganate as developing agent. E. Merck silica gel (60, particle size 0.040-0.063 mm) was used for flash column chromatography. All NMR spectra were recorded at the University of California, Berkeley NMR facility. NMR spectra were recorded on Bruker AV-400 and AV-600 instruments. The spectra were calibrated by using residual undeuterated solvents (for 1 H NMR) and deuterated solvents (for 13 C NMR) as internal references: chloroform (δH = 7.26 ppm), and CDCl3 (δC = 77.16 ppm); acetonitrile (δH = 1.94 ppm) and acetonitrile-d3 (δC = 1.32 ppm); acetone (δH = 2.05 ppm) and acetone-d6 (δC = 29.84 ppm). The following abbreviations are used to designate multiplicities: s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, quint = quintet, br = broad. IR spectra were recorded on a Bruker Vertex 80 FTIR spectrometer. High-resolution mass spectra (HRMS) were recorded on a Perkin Elmer UHPLC-TOF (ESI) at the Lawrence Berkeley National Laboratory Catalysis Laboratory located in the Department of Chemistry at the University of California, Berkeley.

3-(6H-indolo[2,3-b]quinoxalin-6-yl)-N,N,N-trimethylpropan-1-aminium hexafluorophosphate(V) (5b):
To a stirred mixture of SI1 (1.10 g, 5.00 mmol) in DMF (10.0 mL) was slowly added sodium hydride (240 mg, 60% wt, 6.00 mmol) at 0 °C. (Caution: Adding sodium hydride slowly, hydrogen is generated). The resultant mixture was allowed to stir at that temperature for 10 mins before 3-bromo-N,N,Ntrimethylpropan-1-aminium bromide (1.33 g, 6.00 mmol) was added. The mixture was warmed to 22 °C and stirred for 5 h before quenched with water. The resultant mixture was subject to a vacuum to remove the volatiles at 40 °C before water (10 mL) was added to the resultant mixture. NH4PF6 (1.63 g, 10.0 mmol) in 15.0 mL water was added to the resultant mixture at 22 °C. The resultant mixture was allowed to stir at that temperature for 1 h before extracted with CH2Cl2 (3 × 50 mL). The combined organic phases were washed with brine (100 mL), dried over anhydrous Na2SO4, and filtered. The volatiles were removed under vacuum, and the residue was purified by recrystallization with acetone and ethyl ether (1:10) to give 5b (2.14 g, 92%) as a yellow solid.

2/3-(tert-butyl)-6-(2-(2-methoxyethoxy)ethyl)-6H-indolo[2,3-b]quinoxaline (5i):
To a stirred mixture of SI5 (275 mg, 1.00 mmol) in DMF (1.0 mL) was added sodium hydride (44.0 mg, 60% wt, 1.10 mmol) at 0 °C. The resultant mixture was allowed to stir at that temperature for 5 mins before 1-bromo-2-(2-methoxyethoxy)ethane (162 μL, 1.20 mmol) was added. The mixture was warmed to 22 °C and stirred for 5 h before it was quenched with saturated aq. NaHCO3 (5.0 mL), then diluted with EtOAc (10 mL) and brine (10 mL). The resultant mixture was extracted with EtOAc (3 × 10 mL). The combined organic phases were washed with brine (50 mL  General methods and materials: Acetonitrile (MeCN) (99.9%, extra dry over molecular sieves) was obtained from Thermo Scientific. Tetrabutylammonium hexafluorophosphate (TBAPF6; >99%, for electrochemical analysis) was obtained from MilliporeSigma TM , dried under high vacuum for 24 h at 80 °C and transferred to a N2-filled glovebox for storage and use. All electrochemical experiments were performed in a N2 filled glove box with an atmosphere <0.1 ppm oxygen and <0.1 ppm water. Electrolyte solutions were prepared in the glovebox by first drying the acetonitrile over freshly activated 3 Å molecular sieves for at least 24 h. Supporting electrolyte was then added and the solvent/electrolyte mixture was further dried for another 24 h before use. The resulting solvent/electrolyte mixtures were stored over the 3 Å molecular sieves in the glovebox. All potentials are reported relative to the ferrocene/ferrocenium couple (Fc/Fc + ), and this adjustment is made for each sample through the addition of a ferrocene reference at the end of each set of CV experiments.

Cyclic voltammetry: Cyclic voltammetry (CV) experiments were performed with a CH Instruments 760
Bipotentiostation with a three-electrode electrochemical cell. A glassy carbon electrode (BASi, 0.071 cm 2 ) was used as a working electrode, a Ag/Ag + electrode (10 mM AgBF4 in 0.5 M TBAPF6/MeCN) sealed with a Coralpor frit was used as a non-aqueous quasi-reference electrode (BASi), and platinum mesh was used as a counter electrode. The glassy carbon electrode was polished outside the glovebox using alumina (MicroPolish II, Buehler) in Milli-Q ® water before being dried with acetone and brought into the glovebox. Unless otherwise indicated, all CV measurements were performed by dissolving the compound in stock 0.5 M TBAPF6 in acetonitrile to give a concentration of 5 mM.

Linear sweep voltammetry: Linear sweep voltammetry (LSV) experiments were performed with a CH
Instruments 760 Bipotentiostation with a three-electrode electrochemical cell. A rotating disk electrode (RDE) (glassy carbon, 5 mm in diameter, PINE) as the working electrode, a Ag/Ag + electrode (10 mM AgBF4 in 0.5 M TBAPF6/MeCN) sealed with a Coralpor frit was used as a non-aqueous quasi-reference electrode (BASi), and graphite electrode was used as a counter electrode.
H-cell cycling: Bulk charge/discharge measurements were carried out in a nitrogen-filled glovebox with a CH Instruments 760 Bipotentiostat in a custom H-cell (pictured below) with a fritted glass separator (P5). The working and counter electrodes were carbon (Duocel® RVC Foam, 100 PPI, 3% relative density). A Ag/Ag + quasi-reference electrode (described above) was used on the working side of the Hcell. The active compound was dissolved in 0.5 M TBAPF6 in acetonitrile to give a redox-active material concentration of 5 mM. The working chamber of the H-cell was first loaded with 5 mL of the electrolyte/ROM solution while the counter chamber was loaded with 5 mL of only 0.5 M TBAPF6 in acetonitrile. One charging event of the working chamber was completed at which point the solution was removed from the counter chamber and replaced with 5 mL of the electrolyte/ROM solution. A discharge event was then conducted followed by 201 more charge-discharge cycles. Charging and discharging were all conducted with a current of 5 mA and both chambers of the H-cell were continuously stirred with magnetic stir bars. 5 mL of 5 mM solution of electrolyte in 0.5 M TBAPF 6 /MeCN is placed in the working side with 5 mL of 0.5 M TBAPF 6 /MeCN in the counter side. Electrolyte is reduced/oxidized at a constant rate (−5/+5 mA) to a voltage limit of 350 mV lower/higher than the E 1/2 (determined with CV at a scan rate of 100 mV/s) of electrolyte to maximize the state-of-charge, and then the current was reversed to regenerate neutral electrolyte. This cycle was repeated iteratively to evaluate the cycling stability of electrolyte. Results of H-cell bulk electrolysis are displayed here in graphs of normalized discharge capacity (normalized relative to theoretical capacity) vs cycle number for 202 charge-discharge cycles. Each data point in these figures represents one cycle. The capacity fade is based on the maximum discharge capacity unless otherwise indicated.

S35
Picture of custom H-cell

S51
Electrochemical impedance spectroscopy (EIS) on the flow cell before cycling and after 50th, and 75 th cycles.

NMR analysis of post-cycling mixture:
The solvent was removed under vacuum from a mixture solution containing 0.1 M 5a and 0.1 M 5h.
Subsequently, 30 mL of Et2O and 10 mL of hexane were added to the mixture, which was vigorously stirred for 10 minutes before filtering. The filter cake was washed three times with a mixture of 8.0 mL Et2O and hexane in a ratio of 3:1. The volatiles from the combined filtrate were removed under vacuum, and the resulting residue was directly subjected to 1 H NMR analysis. Please refer to S32 and S33 for 1 H NMR details.

Method A (through Cyclic Voltammetry Experiments):
Diffusion coefficients (D) for the neutral species were estimated using the Randles-Ševčík equation (eq. 1) by varying the scan rate of CV measurements between 25 and 500 mV/s. Plotting the cathodic and anodic current versus the square-root of the scan rate results in a linear relationship providing evidence for a diffusion limited chemically reversible process. The slope of this linear relation was used to estimate the diffusion coefficient for the neutral molecule and radical anion.
In the Randles-Ševčík equation 2 , ip is the peak current in amps, n is the number of electrons transferred,   The heterogeneous electron transfer rate constant (k0) was calculated according to the Nicholson method 3 as modified by Magno and coworkers. 4 At each scan rate, the separation between the potential of the anodic and cathodic peaks, ΔEp, can be converted to a dimensionless number using a "working curve" of the Nicholson paper (eq. 2). 3 The data is given in Table S1, S2. Plotting the resulting values of versus the inverse square root of the scan rate ( Figure S18, S19) gave a relationship from which the slope was used to determine the heterogeneous electron transfer rate constant k0 according to eq 3.

Method B (through Linear Sweep Voltammetry Experiments):
A solution of 2 mM 5a or 5h in 0.
where ilim is limiting current, n is the number of electrons transferred (n = 1), F is the Faraday constant (96485 C mol−1), A is the surface area of the working electrode (0.19625 cm 2 ), C is molar concentration in 2 × 10 −6 mol cm −3 , ν is the kinetic viscosity in cm 2 s −1 (0.00442 for 0.1 M TBAPF6/MeCN) 6 and ω is the routing angular velocity in rad s −1 . The Koutecky-Levich plots at different overpotentials were extrapolated to get the kinetic current ik according to the Koutecky-Levich equation (eq.5): The exchange current (i0) can be obtained by fitting ik to the Tafel plot at the overpotential of zero, from which the reaction rate constant (k0) was determined according to the Butler-Volmer equation (eq.6): i0 = nFCk0 (eq. 6)

VIII. Density Functional Theory Calculations
If not stated otherwise, the Q-Chem 20 program suite (version 6.0) was employed for all density functional theory calculations. Initial geometries for the neutral species were relaxed with the semiempirical method GFN2-xTB 21  reduced species are assumed, leading to the half-potential being formally identical to the standard redox potential obtained from the difference in Gibbs energies. 48 Population analysis based on natural bond orbitals 49 (NBOs) was performed by calling the NBO 50 package (version 5) incorporated in Q-Chem. In addition, Mulliken 51 charges and condensed spin densities, and CHELPG 52,53 charges were calculated. Figure SI27. Chemical structure of all computationally investigated species (5a, 8, 9, 11, 12, 2-tert-butyl-5h (5h_2), and 3-tert-butyl-5h (5h_3)).
To identify an appropriate density functional for predictive calculations, we first benchmarked the reduction potentials of 8, 9, and 5a against the experimental half potentials (Table SI5). Multiple density functionals such as TPSSh-D4, PW6B95-D4, MPW1B95-D3(BJ), and the range-separated hybrid functionals adequately reproduce the experimental redox potentials. We chose to use the hybrid meta-GGA functional TPSSh-D4 in the following, given that TPSSh only contains one empirical parameter in the form of the admixture of exact exchange 37 . Table SI5. Calculated first reduction potential (E 1/2 ) of 8, 9, and 5a versus the calculated Fc/Fc + redox couple in V for different density functionals. The deviation from the experimental reference (ΔE 1/2 ) is also given. MAD is the mean absolute deviation ( !" ∑ |∆ "/$, ' | ( ')" ) and AMAX refers to the absolute maximal deviation observed.
Density Functional E 1/2 (8) ΔE 1/2 (8) E 1/2 (9) ΔE 1/2 (9) E 1/2 (5a) ΔE 1/2 (5a) MAD AMAX B97-D3(BJ) -2.17 -0. Having demonstrated that the computational protocol outlined above can predict reduction potentials to a very high accuracy vs experiment, we consider two new compounds, 11 and 12 ( Figure SI27). The calculated potentials are -2.13 V and -2.38 V vs Fc/Fc + , respectively, and are shown relative to the other structures in Table SI4. As outlined in the main text, extending the π-system, e.g., going from 8 to 9 or S70 from 11 to 5a/5h, makes the reduction potential more positive due to the stabilization of the reduced species. On the other hand, introducing an electron-donating group such as a trivalent nitrogen atom into the π-system leads to a much lower potential (8 to 12) despite extending the π-system due to hyperconjugative effects and the lone pair situated at the nitrogen atom (vide infra). Interestingly, this shift is significantly larger going from 9 to 11, presumably, since the size of the π-system is almost maintained.

Figure SI 28.
Charge and spin density changes upon reduction from 5a to 5a •-. Both NBO and CHELPG charges, as well as condensed NBO and Mulliken spin densities, yield a consistent picture where the excess electron is more localized on N5, N11, and C13. Figure SI 29. HOMO, LUMO, and SUMO for 5a, 5h_2 and 5h_3. ρspin is the spin density, the difference of spin-up electron density and spin-down electron density (positive: blue, negative: red). It shows that the excess electron is delocalized within the π-system.