Submitted 7 April 2015; Revised 15 July. This is the authors’ version of the work. It is posted here by permission of the AAAS for personal use, not for redistribution. The definitive version was published as Science 349, 1529 (2015), DOI: 10.1126/science.aab3033
Title: Alkaline Quinone Flow Battery
Authors: Kaixiang Lin1, Qing Chen2, Michael R. Gerhardt2, Liuchuan Tong1, Sang Bok Kim1, Louise Eisenach3, Alvaro W. Valle3, David Hardee1,
Roy G. Gordon1,2*, Michael J. Aziz2*, & Michael P. Marshak1,2*
Affiliations: 1 Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street, Cambridge, Massachusetts 02138, USA. 2 Harvard John A. Paulson School of Engineering and Applied Sciences, 29 Oxford Street, Cambridge, Massachusetts 02138, USA. 3 Harvard College, Cambridge, Massachusetts 02138, USA. *Correspondence and requests for materials should be addressed to MJA (aziz@seas.harvard.edu), RGG (gordon@chemistry.harvard.edu), or MPM (michael.marshak@colorado.edu).
Abstract:
Storage of photovoltaic and wind electricity in batteries could solve the mismatch problem
between the intermittent supply of these renewable resources and variable demand. Flow
batteries permit more economical long-duration discharge than solid-electrode batteries by using
liquid electrolytes stored outside of the battery. We report an alkaline flow battery based on
redox-active organic molecules that are composed entirely of earth-abundant elements and are
non-toxic, non-flammable, and safe for use in residential and commercial environments. The
battery operates efficiently with high power density near room temperature. These results
demonstrate the stability and performance of redox-active organic molecules in alkaline flow
batteries, potentially enabling cost-effective stationary storage of renewable energy.
Main Text:
The cost of photovoltaic (PV) and wind electricity has dropped so much that one of the largest
barriers to getting the vast majority of our electricity from these renewable sources is their
intermittency (1–3). Batteries provide a means to store electrical energy; however, traditional,

enclosed batteries maintain discharge at peak power for far too short a duration to adequately regulate wind or solar power output (1, 2). In contrast, flow batteries can independently scale the power and energy components of the system by storing the electro-active species outside the battery container itself (3–5). In a flow battery, the power is generated in a device resembling a fuel cell, which contains electrodes separated by an ion-permeable membrane. Liquid solutions of redox-active species are pumped into the cell where they can be charged and discharged, before being returned to storage in an external storage tank. Scaling the amount of energy to be stored thus involves simply making larger tanks (Fig 1A). Existing flow batteries are based on metal ions in acidic solution but there are challenges with corrosivity, hydrogen evolution, kinetics, material cost and abundance, and efficiency that thus far have prevented large-scale commercialization. The use of anthraquinones in an acidic aqueous flow battery can dramatically reduce battery costs (6, 7); however, the use of bromine in the other half of the system precludes deployment in residential communities due to toxicity concerns.
We demonstrate that quinone-based flow batteries can be adapted to alkaline solutions, where hydroxylated anthraquinones are highly soluble and bromine can be replaced with the non-toxic ferricyanide ion (8, 9) — a food additive (10). Functionalization of 9,10-anthraquinone (AQ) with electron-donating groups such as OH has been shown to lower the reduction potential and expand the battery voltage (6). In alkaline solution, these OH groups are deprotonated to provide solubility and greater electron donation capability, which results in an increase in the open circuit voltage of 47% over the previously reported system. Because functionalization away from the ketone group provides molecules with the highest solubility (11, 12), we initially targeted commercially available 2,6-dihydroxyanthraquinone (2,6-DHAQ), which we find exhibits a

room temperature solubility > 0.6 M in 1 M KOH. This system can achieve power densities in excess of 0.45 W cm−2 at room temperature and 0.7 W cm−2 at 45 °C.
The use of alkaline electrolyte exploits pH as a parameter to shift the thermodynamic potentials of proton-dependent reactions to more negative values. In acid solutions, AQ undergoes a two-electron two-proton reduction at a single potential, which shifts to more negative values as the pH increases (6). When the pH exceeds 12 the reduction potential of 2,6DHAQ becomes pH-independent because the reduced species is generated in its fullydeprotonated form (Fig. S1). In contrast with the pH-dependent electrochemical behavior of quinones (negative terminal), the ferro/ferricyanide redox couple (positive terminal) has a pHindependent redox potential. This contrasting pH-dependence can be exploited through the development of low reduction potential quinones at high pH. The cyclic voltammograms (CVs) of 2,6-DHAQ and ferro/ferricyanide predict an equilibrium cell potential of 1.2 V upon combination of these two half-reactions (Fig. 1B). A quantitative analysis of the CV of 2,6DHAQ at pH 14 (Fig. S2) revealed redox behavior consistent with two one-electron reductions at potentials separated by only 0.06 V with rapid kinetic rate similar to quinones in acid (6). This behavior raises interesting questions about the relationship between quinone redox and hydrogen bonding (13).
Cell testing was performed at 20 °C using solutions of 0.5 M 2,6-DHAQ dipotassium salt and 0.4 M K4Fe(CN)6 both dissolved in 1 M KOH. These solutions were pumped through a flow cell constructed from graphite flow plates and carbon paper electrodes, which were separated by a Nafion membrane. A charging current of 0.1 A cm−2 was applied to charge the cell, and polarization curves were measured at 10%, 50%, and 100% states of quinone charge (SOC). The open-circuit voltage (OCV) is 1.2 V at 50% SOC; its dependence on SOC is shown in Fig. 2A.

The polarization curves (Fig. 2B) show no sign of redox kinetic limitations and exhibit a peak galvanic power density exceeding 0.4 W cm−2.
The cell was cycled at a constant current density of ±0.1 A cm−2 for 100 cycles (Fig. 3A). The current efficiency exceeded 99%, with a stable round-trip energy efficiency of 84%. A 0.1% loss in capacity per cycle was observed during cycling, which appears to be a continuous loss of electrolyte over the 100 cycles. Three possible loss mechanisms were explored: chemical decomposition, electrolyte crossover through the membrane, and leakage from the pumping system. Chemical and electrochemical stability studies showed that the negative electrolyte is stable. 10 mM of 2,6-DHAQ was heated in 5 M KOH solution at 100 °C for 30 days and was characterized by proton NMR. Cycled negative electrolyte was also collected and characterized by the same method; both studies showed no degradation product at the sensitivity level of 1% (Fig S3). Membrane crossover contamination has been a common challenge in acid-based redox flow batteries where most electro-active molecules are either neutral or positive and tend to migrate through proton-conductive membranes (5). In this alkaline system, however, all the electro-active molecules remain negatively charged in all charge states, leading to a dramatic decrease in the degree of crossover during cell cycling. Cyclic voltammetry of the ferro/ferricyanide electrolyte collected at the end of cycling showed no evidence of the presence of 2,6-DHAQ. This observation places an upper limit on crossover of 0.8% of the DHAQ, implying a crossover current density less than 2.5 µA cm−2 (Fig. S4). Finally, hydraulic leakage was investigated because an apparent but unquantifiable small decrease in fluid levels was observed in the reservoirs. After cell cycling, the cell was washed with DI water until no coloration of eluent could be observed. The cell was then dissembled; coloration was found on the gaskets indicating the likely site of electrolyte leakage (Fig. S5). This source of capacity

loss—the equivalent to roughly 8 drops in our system—is expected to become negligible as system size is scaled up.
By increasing the temperature to 45 °C, the peak galvanic power density increases from 0.45 W cm−2 to approximately 0.7 W cm−2 (Fig. 2C), as the cell area-specific resistance (ASR) decreases from about 0.878 to 0.560 Ω cm2, estimated from the linear parts of the polarization curves in Fig. 2. The majority of this ASR decrease comes from a change in the high frequency area-specific resistance (rhf) measured by electrochemical impedance spectroscopy (Fig. S6). In both cases, the rhf contributes more than 70% of the ASR and is indeed the limiting factor to the cell current and power outputs. The rhf is dominated by the resistance of the membrane, which is an order of magnitude higher than the resistance of the same membrane in a pH 0 acid solution (14).
The sluggish kinetics of the hydrogen evolution reaction in alkaline solution on carbon electrodes results in a larger practical stability window in base rather than in acid (Fig. S7). Consequently, quinones with substantially more negative reduction potentials are feasible as negative electrolyte materials. Preliminary investigations into the synthesis of different hydroxysubstituted anthraquinones suggest further increases in cell potential are possible. Selfcondensation reactions of substituted benzene yield 2,3,6,7-tetrahydroxy-AQ (THAQ) (15) and 1,5-dimethyl-2,6-DHAQ (15-DMAQ) (Fig. S8 and S9). The CV’s of these species in 1 M KOH suggest cell potentials vs. ferri/ferrocyanide approaching 1.35 V (Fig. 4A and B), which exceeds that of many aqueous rechargeable batteries (Fig. 4C).
The cyanide ions in both ferro- and ferricyanide are bound too tightly to be released under most conditions. Consequently, it is non-toxic in both oxidized and reduced forms, and is even permitted for use as a food additive (10). The use of ferrocyanide offers significant advantages

over bromine because it is non-volatile and non-corrosive, allowing simpler and less expensive
materials of construction. In addition, these tri- and tetra-anionic organometallic molecules
exhibit low crossover rates through cation-exchange membranes.
The results reported herein highlight the ability of hydroxy-substituted anthraquinone and
ferrocyanide to function as stable flow battery electrolytes in alkaline solution. The use of
organic and organometallic coordination complexes in base, rather than aqueous metal ions in
acid, resolves serious cost, corrosion, and safety concerns of previous flow battery chemistries.
Alkaline flow batteries can compensate for higher membrane resistance with higher voltage,
leading to performance similar to their acidic counterparts. In addition, quinone-ferrocyanide
alkaline chemistry avoids the membrane crossover, corrosivity, toxicity and regulations
associated with bromine. This reduced corrosivity can lead to a substantially lower materials cost
because many components can be made of inexpensive polyolefin or PVC plastics.
References and Notes: 1. B. Dunn, H. Kamath, J.-M. Tarascon, Electrical Energy Storage for the Grid: A Battery of
Choices. Science. 334, 928–935 (2011).
2. Z. Yang et al., Electrochemical Energy Storage for Green Grid. Chem. Rev. 111, 3577– 3613 (2011).
3. T. Nguyen, R. F. Savinell, Flow Batteries. Electrochem. Soc. Interface. 19, 54–56 (2010).
4. D. Biello, Solar Wars. Sci. Am. 311, 66–71 (2014).
5. M. Skyllas-Kazacos, M. H. Chakrabarti, S. A. Hajimolana, F. S. Mjalli, M. Saleem, Progress in Flow Battery Research and Development. J. Electrochem. Soc. 158, R55–R79 (2011).
6. B. Huskinson, M.P. Marshak, C. Suh, S. Er, M.R. Gerhardt, C.J. Galvin, X. Chen, A. Aspuru-Guzik, R.G. Gordon and M.J. Aziz, A metal-free organic–inorganic aqueous flow battery. Nature. 505, 195–198 (2014).
7. R. M. Darling, K. G. Gallagher, J. A. Kowalski, S. Ha, F. R. Brushett, Pathways to low-cost electrochemical energy storage: a comparison of aqueous and nonaqueous flow batteries. Energy Env. Sci. 7, 3459–3477 (2014).

8. G. G. I. Joseph, A. J. Gotcher, G. Sikha, G. J. Wilson, High performance flow battery (2011), (www.google.com/patents/US20110244277).
9. J. R. Goldstein, Novel flow battery and usage thereof (2014; www.google.com/patents/US20150048777).
10. “Seventeenth Report of the Joint FAO/WHO Expert Committee on Food Additives. Report No. 539,” Wld Hlth Org. techn. Rep. Ser. (539, World Health Organization, 1974).
11. H. Pal, T. Mukherjee, J. P. Mittal, Pulse radiolytic one-electron reduction of 2-hydroxy-and 2, 6-dihydroxy-9, 10-anthraquinones. J Chem Soc Faraday Trans. 90, 711–716 (1994).
12. S. Er, C. Suh, M. P. Marshak, A. Aspuru-Guzik, Computational design of molecules for an all-quinone redox flow battery. Chem. Sci. (2014), doi:10.1039/C4SC03030C.
13. M. Quan, D. Sanchez, M. F. Wasylkiw, D. K. Smith, Voltammetry of Quinones in Unbuffered Aqueous Solution:  Reassessing the Roles of Proton Transfer and Hydrogen Bonding in the Aqueous Electrochemistry of Quinones. J. Am. Chem. Soc. 129, 12847– 12856 (2007).
14. Q. Chen, M. R. Gerhardt, L. Hartle, M. J. Aziz, A Quinone-Bromide Flow Battery with 1 W/cm2 Power Density. J. Electrochem. Soc. 163, A5010–A5013 (2015).
15. T. S. Balaban, A. Eichhöfer, M. J. Krische, J.-M. Lehn, Hierarchic Supramolecular Interactions within Assemblies in Solution and in the Crystal of 2, 3, 6, 7-Tetrasubstituted 5, 5′-(Anthracene-9, 10-diyl) bis [pyrimidin-2-amines]. Helv. Chim. Acta. 89, 333–351 (2006).
16. D. Linden, T. B. Reddy, Handbook of batteries (McGraw-Hill, 2002).
Acknowledgments: This work was funded by the U.S. DOE ARPA-E award # DE-AR0000348. Methods, along with any additional Extended Data display items and Source Data, are available in the online version of the paper; references unique to these sections appear only in the online paper. M.P.M., K.L., R.G.G. and M.J.A. formulated the project. K.L., S.B.K., and D.H. synthesized, analyzed and purified the compounds. K.L., L.T. and S.B.K. collected and analysed the NMR and MS data. K.L. A.V. L.E. measured solubility. K.L., Q.C. L.E. and M.R.G. collected and analysed the electrochemical data. K.L., Q.C., M.R.G., M.P.M., R.G.G. and M.J.A. wrote the paper, and all authors contributed to revising the paper.
Supplementary Materials:
Materials and Methods
Supplementary Text
Fig. S1 to S9
References (17-22)

Fig. 1 Cyclic Voltammetry of Electrolyte and Cell Schematic. (A) Cyclic voltammogram of 2 mM 2,6-DHAQ (dark cyan curve) and ferrocyanide (gold curve) scanned at 100 mV/s on glassy carbon electrode; arrows indicate scan direction. Dotted line represents cyclic voltammogram of 1 M KOH background scanned at 100 mV/s on graphite foil electrode. (B) Schematic of cell in charge mode. Cartoon on top of the cell represents sources of electrical energy from wind and solar. Curved arrows indicate direction of electron flow and white arrows indicate electrolyte solution flow. Blue arrow indicates migration of cations across the membrane. Essential components of electrochemical cells are labeled with color-coded lines and text. The molecular structure of oxidized and reduced species are shown on corresponding reservoirs.
Fig. 2 Cell Performance. (A) Cell OCP vs. SOC. All potentials were taken when cell voltage stabilized to within ± 1 mV. 100% SOC was reached by potentiostatic holding at 1.5 V until the current decreases to below 20 mA/cm2. (B) and (C), Cell voltage & power density vs. current density at 20 oC and 45 oC, respectively, at 10%, 50%, and ~100% SOC. Electrolyte composition: At 20 °C, 0.5 M 2,6-DHAQ and 0.4 M ferrocyanide were used in negative electrolyte and positive electrolyte, respectively. At 45 °C, both concentrations were doubled. In both cases, potassium hydroxide content started at 1 M for both sides in the fully discharged state.
Fig. 3 Cell Cycling Performance. (A) Representative voltage vs. time curves during 100 charge-discharge cycles at 0.1 A/cm2, recorded between the 10th and 19th cycles. (B) Capacity retention, current efficiency and energy efficiency values of 100 cycles. Normalized capacity is evaluated based on the capacity of the first charge and discharge cycle.
Fig. 4 Molecular Structure and Cyclic Voltammetry of 2,6-DHAQ Derivatives. (A) and (B) Molecular structures and cyclic voltammogram of 2,3,6,7-THAQ (pink) and 1,5-DMAQ (olive), respectively, plotted along with ferrocyanide (orange curve) scanned at 100 mV/s on glassy carbon electrode. Both 2,6-DHAQ derivatives/ferrocyanide couples showed higher equilibrium potential than 2,6-DHAQ/ferrocyanide. (C), Selected aqueous secondary batteries showing voltage and flow status. Literature data from (3) and (16).

A

Negative Electrolyte Reservoir

2e-

O O

O O

Cations

O Pump
O
O O

B

2e-

Positive Electrolyte Reservoir

N C
NC

N 3-

C

N C

FeIII C CN

N

Pump

N C C
N

N 4-

C

N C

FeII C CN

N

Ion-selective Membrane Porous Carbon Electrodes Current Collector & Flow Field

E = 1.20 V
O O
O O
2,6-DHAQ

Fe(CN)63-

Current Density (mA cm-2)

O O
O O
2,6-reDHAQ

Fe(CN)64-

Potential (V vs. SHE)

AB

State of Charge

Current Density (A cm-2)

Cell Open Circuit Potential (V)
Cell Voltage (V)
Power Density (W cm-2) Cell Voltage (V)
Power Density (W cm-2 )

C

45
Current Density (A cm-2)

Cell Voltage (V)

A Cycle No. 10 11 12 13 14 15 16 17 18 19
Time (hr) B
Current Efficiency at 0.1 A/cm2 Energy Efficiency at 0.1 A/cm2 Normalized Capacity: Charging Discharging
Cycle Number

Efficiency or Capacity (%)

A

Current Density (mA cm-2)

O HO OH
HO OH O
2,3,6,7-tetrahydroxyanthraquinone
Potential (V vs. SHE) B

Current Density (mA cm-2)

O OH
HO O
1,5-dimethyl-2,6-dihydroxyanthraquinone

C
2.0

Potential (V vs. SHE)
Solid-state battery Previously reported RFBs 2,6-DHAQ/Fe(CN)46- (26DHAQ-FeCy) Prospective RFBs reported in this work

Voltage (V)

1.5 1.4

1.2 1.2 1.2

1.2 1.2

1.3 1.3 1.2

0.8

Acid Acid Alkaline Alkaline Alkaline

Alkaline

Acid

Alkaline

Alkaline

Alkaline

Acid

23126567DDTMHHAAAQQQ---FFFeeeCCCyyy

Supplementary Materials for Alkaline Quinone Flow Battery
Kaixiang Lin, Qing Chen, Michael R. Gerhardt, Liuchuan Tong, Sang Bok Kim, Louise Eisenach, Alvaro Valle, David Hardee, Roy G. Gordon*, Michael J. Aziz*, & Michael P.
Marshak* *Correspondence to: aziz@seas.harvard.edu (M.J.A.); gordon@chemistry.harvard.edu
(R.G.G.); michael.marshak@colorado.edu (MPM) This PDF file includes: Materials and Methods Supplementary Text Figs. S1 to S9
1

Materials and Methods
Materials All chemicals were purchased from Sigma Aldrich and used as received unless
stated otherwise. 2,6-dihydroxyanthraquinone (2,6-DHAQ) (90% purity) was purchased from AK Scientific Inc. and was recrystallized from DMF-water mixture. 2-methyl-3hydroxybenzoic acid (98% purity) was purchased from AK Scientific Inc. and was used as received. Nafion 212 and Sigracet SGL 10AA porous carbon paper were purchased from Ion Power and both pretreated (see method below). The cell hardware and Poco pyrosealed graphite flow plates were purchased from Fuel Cell Technologies.
Chemical characterization and preparation Proton nuclear magnetic resonance (1H NMR) spectra were recorded using Varian
INOVA 500 (500 MHz) NMR spectrometers at 23 °C. Proton chemical shifts are expressed in parts per million (ppm, δ scale) and are referenced to residual protium in the NMR solvent (CDCl3, δ 7.26 ppm; (CD3)2CO, δ 2.05 ppm; (CD3)2SO, δ 2.50 ppm; CD3OD, δ 3.31 ppm). Carbon nuclear magnetic resonance (13C NMR) spectra were recorded using Varian INOVA 500 (125 MHz) NMR spectrometers at 23 °C. Carbon chemical shifts are expressed in parts per million (ppm, δ scale) and are referenced to the carbon resonances of the NMR solvent (CDCl3, δ 77.16 ppm; (CD3)2CO, 29.84 ppm; (CD3)2SO, 39.50 ppm; CD3OD, 49.00 ppm).
Mass spectrum was collected using a maXis impact UHR time-of-flight mass spectrometer system (Bruker Daltonics Inc, Billerica, MA) equipped with an electrospray ionization (ESI) source. The sample was dissolved in acetonitrile (5 uM) and analyzed in negative mode.
2,3,6,7-tetrahydroxyanthraquinone (2,3,6,7-THAQ) was prepared using a previously reported synthetic route (15). The overall yield was 29% (1H and 13C NMR spectra shown in Fig. S8). 1,5-dimethyl-2,6-dihydroxyanthraquinone (1,5-DMAQ) was prepared by the following method: AlCl3 (10 g) and NaCl (1 g) were mixed and heated to 200 oC to afford a molten salt. 2-methyl-3-hydroxybenzoic acid (2 g) was slowly added in four portions into the AlCl3-NaCl melt. The mixture was continuously heated at 200 oC for 1 hour. After cool down, distilled water (100 mL) was slowly added to this solution, which was then stirred at room temperature overnight. The dark yellow precipitate was collected on frit glass filter and extracted with ethyl acetate (100 mL). Distilled water (50 mL) was then added to this solution. Ethyl acetate layer was separated from aqueous layer by separatory funnel and dried on magnesium sulfate. The organic phase was concentrated (to approximately 25 mL) and cooled in an ice bath. The resulting precipitates were filtered on celite and washed with cold ethyl acetate (50 mL). This filter cake was then dissolved in acetone (50 mL) where ethyl acetate (50 mL) was added. This mixture was gently heated to remove acetone. Precipitates from this resulting solution were removed on celite. After concentrating the ethyl acetate filtrate, hexane was added to precipitate out desired product which was then collected on frit glass filter and dried under vacuum to afford pale yellow solid. The yield of this product was 12% (1H, 13C NMR spectra and mass spectroscopy results are shown in Fig. S9).
Solubility measurement
2

Room temperature solubility in 1 M KOH solution was determined by measuring the absorbance at 413 nm and comparing to an absorbance-vs.-concentration calibration curve determined by preparing known concentrations of 2,6-DHAQ. UV-Vis spectrophotometry measurements were performed using an Agilent Cary 60 spectrophotometer equipped with a Quantum Northwest T2 temperature regulator. Appropriate aliquots of 2,6-DHAQ stock solution were added to 1 M KOH blank solution and their UV-Vis absorbance spectra measured. A saturated solution of 2,6-DHAQ in 1 M KOH was prepared by adding 2,6-DHAQ potassium salt into 1 M KOH solution (10 mL) until a thin layer of precipitate formed, the remaining solution was diluted by known proportions, and the absorbance of the resulting solution was compared to the calibration curve.
Electrochemical characterization A glassy carbon electrode was used for three-electrode cyclic voltammetry tests,
except in the electrochemical window tests where graphite foil was used to mimic the condition of porous carbon paper. Rotating-disk electrode (RDE) experiments of 2,6DHAQ (1 mM) in 1 M KOH solution were performed using a BASi RDE-2 rotating-disk electrode system (Fig. S2A). All tests were carried out using a Gamry Reference 3000 potentiostat, with a Pt counter electrode and a Ag/AgCl reference electrode (equilibrated with 3M NaCl, 213 mV vs. standard hydrogen electrode). A Levich plot was constructed from the RDE data by plotting the mass-transport limited current vs. the square root of the rotation rate (Fig. S2B). The diffusion coefficient of 2,6-DHAQ was calculated for each of three runs from the slope of the line fit to the Levich equation (17). The kinematic viscosity was taken to be 1.08 ×10−6 m2 s−1 (18). The resulting value of the diffusion coefficient is 4.8(2) ×10−6 cm2 s−1.
Membrane and electrode pretreatments Pretreatment of Nafion ® 212 membrane was performed by first heating in 80 oC
de-ionized water for 20 minutes and then soaking in 5% hydrogen peroxide solution for 35 minutes. These pre-treated membranes were stored in 0.1 M KOH solution at room temperature.
Sigracet ® SGL 10AA (~ 400 µm thickness) porous carbon paper was pretreated by baking in air at 400 °C for 24 hours.
Full cell measurement 20 oC cycling study
The positive electrolyte was prepared by dissolving potassium ferrocyanide trihydrate (2.56 g) in 1 M KOH solution (15 mL) to afford a 0.4 M ferrocyanide and 2.6 M potassium electrolyte solution. The negative electrolyte was prepared by dissolving 2,6-DHAQ (0.48 g) in 2 M KOH solution (4 mL) resulting a 0.5 M 2,6-DHAQ and 1 M potassium electrolyte solution. 20 oC cell performance study
The positive electrolyte was prepared by dissolving potassium ferrocyanide trihydrate (2.56 g) in 1 M KOH solution (15 mL) to afford a 0.4 M ferrocyanide and 2.6 M potassium electrolyte solution. The negative electrolyte was prepared by dissolving
3

2,6-DHAQ (0.6 g) in 2 M KOH solution (5 mL) resulting a 0.5 M 2,6-DHAQ and 1 M potassium electrolyte solution. 45 oC cell performance study
The positive electrolyte was prepared by heat dissolving potassium ferrocyanide trihydrate (5.1 g) in 1 M KOH solution (15 mL) at 45 oC to afford a 0.8 M ferrocyanide and 4.2 M potassium electrolyte solution. The negative electrolyte was prepared by heat dissolving 2,6-DHAQ (1.2 g) in 3 M KOH solution (5 mL) at 45 oC resulting a 1 M 2,6DHAQ and 3 M potassium electrolyte solution.
For all full cell studies, the electrolytes were assembled in the fully discharged state. Cell hardware from Fuel Cell Tech. (Albuquerque, NM) was used to assemble a zero-gap flow cell configuration, similar to previous reports (19). Serpertine flow pattern pyrosealed POCO graphite flow plates were used for both sides. A 5 cm2 geometric surface area electrode comprised a stack of three sheets of Sigracet SGL 10AA porous carbon paper. A sheet of pretreated Nafion 212 membrane served as the ion-selective membrane. The rest of the space between the plates was gasketed by Teflon sheets. The electrolytes were fed into the cell through PFA tubing, at a rate of 60 mL/min controlled by Cole-Parmer Micropump gear pumps. Before and during the tests, the electrolytes were purged with ultra-high purity argon to ensure deaeration. Galvanostatic cycling was performed at ±0.1 A/cm2, with voltage limits of 0.6 and 1.7 V, controlled by a Gamry 30K Booster potentiostat. To obtain the polarization curves, the cell was first charged to the desired SOC, and then polarized via linear sweep voltammetry at a rate of 100 mV/s. This method was found to yield polarization curves very close to point-by-point galvanostatic holds, yet to impose minimal perturbation to the SOC of the small-electrolyte-volume cell. EIS was performed at 50% SOC, open-circuit potential, and 10 mA/cm2 AC current density, with frequency ranging from 1 to 100,000 Hz.
Supplementary Text
Cyclic voltammogram modeling Computation was performed using Mathematica 10.0.1.0 according to algorithms by
Oldham and Myland (20). The diffusivity of 2,6-DHAQ and both the one-electron and two-electron reduction products were assumed to be 4.8 ×10−6 cm2 s−1, based on our RDE studies. The temperature was 293 K. Time was discretized into 40 ms increments, and the current at each time was computed based on the sum of previous currents as described by Oldham (20).
To model the two-electron reduction process (Fig. S2C), the Butler-Volmer equation with the number of electrons n = 2 was used in place of Oldham equation (4:5), and the right-hand-side of equation (8:3) was multiplied by 2 to account for a two-electron reduction (17, 20). The reduction potential was assumed to be equal to the voltage equidistant from the two peaks on the measured cyclic voltammogram (−0.684 V vs SHE) and the charge transfer coefficient α was assumed to be 0.5. The electrochemical rate constant k0 was then varied from 10−3 cm s−1 to 10−5 cm s−1.
To model the two successive one-electron reductions (Fig. S2D), Oldham equations (12:13) and (12:14) were used (20). The charge transfer coefficients α1 and α2 were assumed to be equal to 0.5, and the electrochemical rate constants for the two reductions were assumed equal. The values of the single rate constant and the first and second
4

reduction potentials were permitted to vary freely. A reasonable fit to the experimental data was observed when the reduction potential of the first reduction E1 = −0.657 V vs SHE, the reduction potential of the second reduction E2 = −0.717 V vs SHE, and the two electrochemical rate constants k0,1 and k0,2 both equaled 7 ×10−3 cm s−1, a value very close to that observed in other anthraquinone systems (6). The simulation was repeated for k0,1 and k0,2 equal to 7 ×10−4 cm s−1 to establish a lower bound on possible k0 values. Energy Density
The negative electrolyte of 0.5 M DHAQ, with two electrons per molecule, has a charge capacity of 27 Ah/L. The positive electrolyte of 0.4 M ferrocyanide has a charge capacity of 11 Ah/L. In the experimental cell used here, in which there is 50% more positive electrolyte than needed to balance the negative electrolyte, the energy density is 6.8 Wh/L. However, eliminating the positive electrolyte volume excess and using established methods of raising the ferrocyanide solubility to 1.4 M (21), the energy density would become 18.8 Wh/L.
5

-580

Potential (mV vs. SHE)

-600

Slope = -61 mV/pH

-620

-640

-660 -680

Slope = 0 mV/pH

-700
10 11 12 13 14
pH

Fig. S1. Pourbaix diagram of 2,6-DHAQ.
Above pH ~11.7, the equilibrium potential of 2,6-DHAQ is pH-independent, indicating that both oxidized and reduced form are fully deprotonated. The solid line is linear fit to the data between pH 10 and 12; the dashed line is not a fit, but zero slope.

6

AB
CD
Fig. S2 Study and simulation of the electrochemistry of 2,6-DHAQ. (A) Rotating disk electrode study of the reduction of a 1 mM solution of 2,6-DHAQ in 1 M KOH on a glassy carbon electrode at various rotation rates. (B) Levich plot of 1 mM 2,6-DHAQ in 1 M KOH. Data are an average of the current at −1.187 V vs SHE for each of three runs; error bars indicate the standard deviation. (C) Cyclic voltammogram of 1 mM 2,6-dihydroxyanthraquinone in 1 M KOH (solid line). Dashed lines represent simulated cyclic voltammograms of a two-electron reduction of varying electrochemical rate constant k0 and a reduction potential E0 of −0.684 V vs SHE. The simulations assumed α = 0.5. The scan rate in all cases was 25 mV s−1. (D) Cyclic voltammogram of 1 mM 2,6-dihydroxyanthraquinone in 1 M KOH (solid black line). Dashed red lines represent the simulated components of two successive one-electron reductions with reduction potentials of −0.657 V vs SHE and −0.717 V vs SHE. The solid red line represents the simulated total current arising from such a reaction. Each simulated reduction has a rate constant k0 = 7 ×10−3 cm s−1. The same simulation was performed assuming each reduction has a rate constant k0 = 7 ×10−4 cm, and the total current plotted as a solid green line. The slower rate constants do not adequately describe the observed cyclic voltammogram, setting a lower bound on the value of the rate constant. The simulations assumed α = 0.5.
7

A
B B C Fig. S3 Chemical and electrochemical stability of 2,6-DHAQ. 1H NMR (500 MHz, DMSO-d6) spectra. (A) 2,6-DHAQ δ: 8.04 (d, 2H, J = 8.3 Hz, 2 x ArCH), 7.47 (d, 2H, J = 2.5 Hz, 2 x ArCH), 7.19 (dd, 2H, J = 8.3, 2.5 Hz, 2 x ArCH). (B) 2,6-DHAQ, after 30 days heating in 5 M KOH solution at 100 oC. (C) 2,6-DHAQ, after 100 charge-discharge cycles.
8

Current Density (mA/cm2)

A
80 60 40 20
0 -20 -40 -60
-1.5
B
4
B
2
0

2,6-DHAQ

ferri/ferrocyanide
-1.0 -0.5 0.0 0.5 1.0
Potential (V vs. SHE)

1.5

7.6 mM 2,6-DHAQ added

Current Density (mA/cm2)

-2 3.9 mM 2,6-DHAQ

Cycled Posolyte

added

-4

-6

-8 -1.0

-0.8 -0.6
Potential (V vs. SHE)

-0.4

Fig. S4 Test of electrolyte permeation through membrane.
(A) Full and (B) Zoomed in cyclic voltammograms of 0.4 M ferro-ferricyanide electrolyte after 100 charge-discharge cycles (black trace). Same posolyte to which we added 3.9 mM 2,6-DHAQ (orange trace), and 7.6 mM 2,6-DHAQ (pink trace).

9

Fig. S5 Leakage of electrolyte into gaskets. Image showing the Teflon gaskets and graphite flow plates (A) before and (B) after cell cycling. Orange colored area indicates leakage of the negative electrolyte (top plate: negative side; bottom plate: positive side) leading to capacity fade.
10

-Zimg( cm2)

0.3

45 oC, rhf = 0.419 cm2
0.2

20

oC,

r hf

=

0.625

cm2

1 Hz

1 kHz
0.1

1 Hz

1 kHz

0.0
10 Hz
-0.1

10 Hz

-0.2 0.4

0.5 0.6
Zreal( cm2)

0.7

0.8

Fig. S6 Electrochemical impedance spectroscopy (Nyquist plot).
Frequency decreases from left to right. The high frequency area-specific resistance (rhf) values discussed in the main text are obtained by fitting the high frequency parts to a resistor (rhf) in series with an inductance component. The inductance came from thick cables and current collectors that connect the cell to the potentiostat. The rhf includes contributions from the Nafion membrane, carbon electrodes, and electrical leads between the cell and the potentiostat. The rhf values at two different temperatures are indicated within the figure.

11

Current Density (A cm-2)

40

Eo = -0.83 V

Eo = 0.40 V

30 1 M KOH

20

10

0 -10 1 M KOH -20

1 M H SO 24

-30

-40

Eo = 0 V

Eo = 1.23 V

-1.5 -1.0 -0.5 0.0 0.5 1.0 Potential (V vs. SHE)

1.5

Fig. S7 Background cyclic voltammograms.
Cyclic voltammograms of 1 M KOH and 1 M H2SO4 background scanned at 100 mV/s using graphite foil electrode. Dashed lines indicate commonly-reported equilibrium potentials of water splitting reactions. This illustrates the practical stability window for aqueous flow batteries.

12

A
B B
C
Fig. S8 Synthetic scheme and NMR characterization of 2,3,6,7tetrahydroxyanthraquinone (2,3,6,7-THAQ). (A) Synthetic scheme of 2,3,6,7-THAQ from inexpensive, commodity chemicals. i. condensation with acetaldehyde to afford 2,3,6,7-tetramethoxydimethylanthracene (MeCHO, H2SO4), ii. oxidation to afford 2,3,6,7-tetramethoxyanthraquinone (Na2Cr2O7 in acetic acid), iii. hydrolysis of methoxy (HBr, reflux) We carried out all three steps of reaction in this work according to literature reference (15). (B) 1HNMR (500 MHz, DMSO) spectrum of 2,3,6,7-THAQ δ 10.42 (br, 4H, 4 x Ar-OH), 7.43 (s, 4H, 4 x ArCH). (C) 13CNMR (125 MHz) spectrum δ 181.50, 151.23, 127.46, 113.31, 113.28. Solvent peaks are labeled with asterisks.
13

A A
B
C
Fig. S9 Synthetic scheme and NMR characterization of 1,5-dimethyl-2,6dihydroxyanthraquinone (1,5-DMAQ). (A) Synthetic scheme of 1,5-DMAQ from inexpensive, commodity chemicals. i. sulfonation of naphthalene to afford 1,3,5-nanphthalenetrisulfonic acid (H2SO4/SO3), ii. preparation of sodium salt to afford trisodium naphthalene-1,3,5-trisulfonate (NaOH), iii. reaction of trisodium naphthalene-1,3,5-trisulfonate with alkali (NaOH in autoclave at high temperature (280-310 °C)), iv. Dimerization to afford 1,5-DMAQ (AlCl3-NaCl melt). We carried out step iv in this work. Steps i-iii were reported in a patent application (22). (B) 1HNMR (500 MHz, DMSO-d6) spectra of 1,5-DMAQ δ: 10.78 (br, 2H, 2 x ArOH), 7.93 (d, 2H, J = 8.5 Hz, 2 x ArCH), 7.21 (d, 2H, J = 8.5 Hz, 2 x ArCH), 2.55 (s, 6H, 2 x Ar-CH3). (C) 13CNMR (125 MHz, DMSO-d6) δ 185.63, 161.87, 133.10, 128.94, 128.04, 126.74, 119.83, 14.03. Solvent peaks are labeled by asterisk. Mass spectroscopy: C6H12O4 M-H calculated 267.0657 found 267.0659
14

References, Supplementary Materials
17. A. J. Bard, L. R. Faulkner, Electrochemical Methods: Fundamentals and Applications, 2nd ed. (John Wiley & Sons, New York, 2001)
18. L. B. Hitchcock, J. S. McIlhenny, Viscosity and Density of Pure Alkaline Solutions and Their Mixtures. Ind. Eng. Chem. 27, 461–466 (1935).
19. Q. H. Liu et al., High Performance Vanadium Redox Flow Batteries with Optimized Electrode Configuration and Membrane Selection. J. Electrochem. Soc. 159, A1246–A1252 (2012).
20. K. B. Oldham, J. C. Myland, Modelling cyclic voltammetry without digital simulation. Electrochimica Acta. 56, 10612–10625 (2011).
21. A. J. Esswein, J. Goeltz, D. Amadeo, High solubility iron hexacyanides (2014); www.google.com/patents/US20140051003
22. H. Behre, F. Mueller-Hauck, J. Scherer, G. Schroeder, Method for producing 3-hydroxy-2methylbenzoic acid (2004); www.google.com/patents/WO2003080542A3