High-Performance Aqueous Organic Flow Battery with Quinone-Based Redox Couples at Both Electrodes

All experiments for the electrochemical characterization of the individual redox molecules were conducted in a standard three-electrode cell consisting of a rotating glassy-carbon disk working electrode, a platinum-wire counter electrode, and a mercury/mercuric sulfate reference electrode (E° = +0.65 V). The quinones, in either the fully-reduced or fully-oxidized form, were dissolved in 1 M sulfuric acid to a concentration of 1 mM. The solutions were de-aerated and kept under a blanket of argon gas throughout all the experiments. Linear-sweep voltammetry experiments were conducted at a scan rate of 5 mV s⁻¹ over a range of rotation rates from 500 rpm to 3000 rpm, using a Versastat 300 potentiostat and rotating electrode equipment from Pine Instruments.

To build a flow cell, we adapted fuel cell test hardware (Electrochem Inc.) with an electrode active area of 25 cm² and densified graphite flow field plates (Figure 2). The reactants were circulated using centrifugal pumps (March Pump Model #: BC-2CP-MD 12 V DC) at flow rates in the range of 0.5–1.0 liter min⁻¹. Two flow-field designs that incorporated "flow-through" and "flow-by" features were tested. The "flow-through" configuration refers to a fluid path that includes flow that is perpendicular and parallel to the plane of the electrode; thus, the entire volume of the electrode has the opportunity to encounter the flow of solution. In a "flow-by" design, the fluid flow is largely parallel to the plane of the electrode. The relative utilization of the area of the electrodes in these two configurations is dependent on the flow rate and the pressure drop between the inlet and the outlet. We prepared the membrane electrode assemblies (MEAs) needed for the cell using procedures that were developed in-house, similar to those used in direct methanol fuel cells.²⁶ Specifically, two sheets of carbon paper (Toray 030-non-teflonized, each with an area of 25 cm²) were coated with an ink made from 0.1 g of carbon black (Vulcan XC-72) and 0.3 g of Nafion with 0.5 ml of water and 0.5 ml of isopropanol. The equivalent weight of the Nafion ionomer was 1100 g/mole of protons. The coated electrodes were then hot pressed on to a Nafion 117 or 212 membrane to form the MEA. In some of the experiments, graphite felt (SIGRACELL GFD 4.6 from SGL Group) was used as electrodes, either as-received or after surface modification with Nafion and carbon black. This type of felt is widely used in vanadium-redox flow batteries.²⁷ To increase their wettability, the felts were immersed in a dilute solution of Nafion and then dried at 140°C for an hour. The Nafion uptake was 10% of the weight of the felt. The felts were then boiled in water for 2 hours, to remove any excess or unbound Nafion and to hydrate the Nafion that was retained on the felt. In another modification, the graphite felt was coated with varying amounts of the carbon black.

Zoom In Zoom Out Reset image size

All full cell experiments were carried out with solutions of redox couples dissolved in 1 M sulfuric acid at 23°C. Two glass containers served as reservoirs for the solutions of the redox couples (Figure 2). An argon flow was maintained at all times above these solutions to avoid reaction of the reduced form of the redox couples with oxygen. We have found that the exclusion of oxygen is critical to maintaining stable cell capacity. The current-voltage characteristics of the cells were measured at various states of charge. Charge/discharge studies were carried out under constant current conditions either using a potentiostat (Versastat 300) or a battery cycler (Maccor 4200). Typical currents ranged from 0.5 A to 5 A for an electrode area of 25 cm².

We used double-layer capacitance measurements to determine the electrochemically-active area of the electrodes. Electrochemical Impedance Spectroscopy (EIS) measurements were carried out over the frequency range of 10 mHz to 10 kHz with a sinusoidal excitation of amplitude ±2 mV peak-to-peak (Versastat 300 potentiostat). For these measurements, we circulated solutions containing solely 1 M sulfuric acid past both electrodes. Solutions were also carefully de-aerated to avoid oxygen. In the absence of faradaic reactions, the electrode/electrolyte interface is considered "ideally polarizable". Under these conditions, the imaginary component of the impedance, Z_imag is related to the series equivalent capacitive reactance by the following equation:

Where f is the excitation frequency and C_s is the value of the series-equivalent capacitance. The series-equivalent capacitance is also the double-layer capacitance because the interface in this experiment is ideally polarizable. Thus, the plot of Z_imag vs. 1/(2πf) is expected to yield a line with a slope of −(1/C_s). The surface area is then calculated by assuming a specific capacitance of 20 microfarad /cm² for a carbon surface and a geometric area of 25 cm².

The ohmic resistance of the cells was determined from the real component of the complex impedance measured at 10 kHz. This value of ohmic resistance was used wherever the reported cell voltage was corrected for the voltage drop across the ohmic resistance of the electrodes, electrolyte and membrane. Such a correction allowed us to compare the mass transport and other kinetic limitations, ignoring variations in ohmic resistance caused by changes to the contact resistance resulting from changes to the flow field and compression of the electrodes. The correction was determined by multiplying the current with the ohmic resistance value, and subtracting this value from the cell voltage during charge or adding to the cell voltage value during discharge.

Samples withdrawn from cells were analyzed by proton-NMR to determine the transformations of the molecular structure of the quinone compounds. Typically, 100 μL of electrolyte sample was diluted with 0.5 mL of deuterated water (D₂O) and ¹H NMR spectra were obtained on Varian 400 MHz or 500 MHz NMR instruments with 128 scans. The chemical shifts were determined relative to the water signal (4.79 ppm).

Our previous paper emphasized that high values of solubility of the redox couples is necessary for achieving high current densities and high efficiencies in ORBAT.⁵ We had used the sodium sulfonate salts of BQDS and AQDS (Sigma-Aldrich) "as received" in all our previous experiments. These salts exhibited solubility values ranging from 0.3 to 1.0 M in 1 M sulfuric acid at 23°C. When the sodium sulfonate salts of the quinones were transformed to the free acid form (sulfonic acid) by passing the solution through an ion-exchange column (Amberlyst 15(H) ion exchange resin), the solution concentrations of AQDS and BQDS could be increased to as high as 1.5 M and 4 M, respectively. For fully-dissociated electrolytes resulting in solvated ions in solution, solubility is governed by the difference between the crystal lattice energy and the solvation energy.²⁸ The higher solubility observed for the acid form compared to the sodium salt is consistent with the solvation energy of a sodium ion being 162.7 kcal/mole less than that of a hydronium ion.²⁹

We observed a five-fold increase in current density with the free sulfonic acid form over the sodium salt form (Figure 3); this increase in performance can only be attributed to the increase in concentration of the reactants and improved ionic conductivity of the membrane. The higher solubility of the acid form avoided the abrupt drop in cell voltage resulting from concentration polarization caused by inadequate mass transport, especially at high current densities and low states-of-charge. The resistance of the cell at 10 kHz with 1 M sulfuric acid was 0.30 Ohm cm², and the addition of 0.5 M solution of the sodium salt of AQDS raised this value of cell resistance to 0.58 Ohm cm².Thus, the absence of sodium ions in the electrolyte increased the ionic conductivity of the Nafion membrane, as protons have a higher mobility in the membrane compared to the sodium ions.^30,31

Zoom In Zoom Out Reset image size

Solutions of 1 M AQDS and 2 M BQDS (both in the acid form) were circulated past the electrodes of the flow cell. In a typical experiment, 600 ml of the AQDS solution and 100 ml of the BQDS solution were used to give the mole ratio of 3:1 for AQDS to BQDS. We tried to use as high a concentration as possible to improve mass transport. However, since the solubility values for BQDS are substantially higher than that of AQDS even in the acid form, we had to use different concentrations for the positive and negative electrolyte solutions. The cell was charged and discharged multiple times to determine:

• (a)  
  the chemical/electrochemical transformations that the redox couples undergo during charge and discharge (as determined by proton-NMR and linear sweep voltammetry), and
• (b)  
  the propensity of the electroactive organic molecules to diffuse through the proton-conducting membrane, referred to hereafter as "crossover".

We found that stable capacity could be reached after about five cycles of charge and discharge (Figure 4). The charge acceptance in the first cycle was about three times that of the discharge capacity output following this charge. Therefore, the coulombic efficiency in the first cycle was about 33%. However, in subsequent cycles the coulombic efficiency gradually converged to 100% (Figure 4a). In the initial cycles, the charging curves showed a single voltage plateau that developed into two voltage plateaus (Figure 4b), suggesting the formation of at least two types of redox active species that underwent transformations. The initial value of the charging voltage of the cell before the transformations was 0.9 V (Figure 4b), and this reduced to 0.7 V by the end of 100 cycles, suggesting that chemical transformations of the redox active species had occurred during the cycling.

Zoom In Zoom Out Reset image size

While the analysis of the negative electrode solution by proton NMR indicated no change of molecular composition of AQDS upon cycling, significant changes were observed in the solution at the positive electrode. Samples from the positive side of the cell drawn at different stages of the extended cycling at 100 mA/cm² were analyzed by NMR (Figure 5). We determined that BQDS underwent a slow chemical transformation in addition to the expected electro-oxidation and electro-reduction during charge and discharge. The two doublets at 7.34 ppm (J = 2.2 Hz, 1H) and 7.10 ppm (J = 2.2 Hz, 1H) in the proton NMR at the start of the cycling corresponded to 4,5-dihydroxybenzene-1,3-disulfonic acid (BQDS), with protons on the aromatic ring present in two distinct environments. After the first charge, the ¹H NMR spectrum showed a new "aromatic" proton signal at 7.14 ppm (singlet) that was attributed to 1,2,4-trihydroxybenzene-3,5-sulfonic acid. After extended cycling (120 cycles), the ratio of 1,2,4-trihydroxybenzene-3,5-sulfonic acid to BQDS had increased. After 400 cycles all the "aromatic" proton signals of BQDS and other reaction products disappeared, indicating the occurrence of yet another transformation that produced compounds with no peaks on the proton NMR spectrum. The lack of NMR-active "aromatic" protons on the aromatic ring suggested the formation of 1,2,4,6- tetrahydroxybenzene-3,5-sulfonic acid. Although the samples for NMR were obtained only after 120 cycles and 400 cycles, the electrochemical data indicated that changes were probably complete even in the earlier stages of cycling. The changes to the charge-discharge curves and the NMR spectra during cycling were consistent with the stepwise transformation of 4,5-dihydroxybenzene-1,3-disulfonic acid to 2,4,5-trihydroxybenzene-1,3-disulfonic acid and ultimately to 1,2,4,6-tetrahydroxybenzene-3,5-disulfonic acid (Figure 6). These transformations involve electrochemical oxidation of the hydroxyl group to the quinone, followed by the addition of water to form reduced products. The addition of nucleophiles such as water to the α,β-unsaturated carbonyl compounds in a 1,4-fashion is called the Michael reaction. In our study, the nucleophilic addition of water is accompanied by re-aromatization and exchange of the proton.³² The Michael reaction necessitated additional charge input for the re-oxidation of the product during charging.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The limiting current at a rotating disk electrode during a linear sweep voltammetry experiment was used to analyze the progress of conversion of the materials from the discharged to the charged state. The value of the limiting current indicated the concentration of the reduced and oxidized species during charge and discharge. When AQDS was in its fully discharged form, there was a reduction current indicating presence of the oxidized form (Figure 7a). Similarly, when the solution on the negative side was in the fully-charged state there was no longer any reduction current − only an oxidation current − indicating the full conversion of the reactants.

Zoom In Zoom Out Reset image size

However, the linear sweep voltammetry studies on BQDS indicated changes to the composition of the solution during the charging process. The persistent presence of the reduced state of the positive side material (Figure 7b) was consistent with the chemical transformations occurring during charging. The electrochemical studies confirmed that BQDS undergoes three redox transformations during charging that includes two Michael addition steps followed by re-aromatization with intermediates in the "reduced" state (Figure 6). Following these transformations, the 1,2,4,6-tetrahydroxybenzene-3,5-disulfonic acid became the positive electrode material that was repeatedly cycled against AQDS at the negative electrode.

The transformation of BQDS during the initial stages of cycling presented some challenges with the overall cell operation. First, at least three mole equivalents of AQDS were required compared to BQDS, in order to convert all the BQDS completely to its final form. After BQDS had reached its fully charged form, 66% of the AQDS remained unused during the subsequent cycles. Secondly, the products formed after the Michael reactions were expected to reduce the cell voltage, as additional hydroxyl substituents were added to BQDS. A reduction of about 150 mV in the charge voltage is noted after the first few cycles (Figure 4). This reduction is slightly lower than the 100 mV/ hydroxyl group expected from the study of substituent effects.⁶ We are in the process of synthesizing the transformed additive in the pure form and determining its standard reduction potential that is needed for understanding the energy efficiency values. Placing a fully-transformed BQDS (1,2,4,6-tetrahydroxybenzene -3,5-disulfonic acid) molecule in the cell at the start of cycling would avoid the need for excess AQDS. Synthesis of new molecules and electrochemical studies are currently underway to address these issues.

We confirmed from the NMR analysis of the solutions that even after 400 cycles that neither compound diffused through the Nafion membrane. These results were also consistent with the ex-situ studies− even after 2 months of holding solutions of 0.2 M AQDS or BQDS dissolved in 1 M sulfuric acid on one side of the membrane, and just 1 M sulfuric acid on the other side of the membrane, no crossover was detected. The absence of crossover confirmed the low permeability of the Nafion membrane to the anionic form of the redox molecules.

Changes to the flow-field and electrode structure resulted in at least a three-fold increase in power density (Figure 8b). In these tests, an adequate amount of AQDS was used at the start of cycling tests to ensure complete conversion of BQDS. The flow field was modified from a flow-by configuration to an interdigitated design that added "flow-through" characteristics to the original flow field. We still had a small component (∼20%) of flow-by due to features that allowed some by-passing of the flow interdigitated structure (see Figure S1 in the Supplementary Material). With this improved design, the solutions were forced to move through the electrode structure, instead of simply flowing past the electrode. Several studies confirm the benefits of the flow-through design for redox flow batteries.^23,33–36 By utilizing more of the surface area of the electrode and by improving mass transport to the electrode surface an increase in the cell performance was achieved. We note that the benefits of the flow-through design tended to increase with the flow rate. However, the flow-through design caused an increase in pressure drop for the flow of the solutions. The pressure drop was high when the carbon-coated Toray paper electrodes were used with a simple interdigitated flow field. With an open electrode structure, such as the SGL graphite felt, and with a combination of flow-through and flow-by designs, we could minimize the pressure drop and still maintain the high level of cell performance. We will continue pursuing the research on optimization of the flow-field design with open electrode structures for improving cell performance.

Zoom In Zoom Out Reset image size

Efforts on improving the electrode design focused on increasing the accessible surface area of the electrodes. It is known that attaching high surface area carbon to electrode surfaces will improve performance in redox flow batteries.³⁷ Double-layer capacitance measurements were used to determine the electrochemically active surface area. While the graphite felt treated with 2.5 wt% of Nafion appeared more wettable than the un-treated felt, both these samples had nearly the same double-layer capacitance (Figure 9, Table I). We expected that when the surface became more wettable more of the electrochemical active area would increase and consequently the double layer capacitance would increase. While Nafion increased the apparent wettability, it did not affect the accessible active area of the electrode to any significant degree. This observation suggests that the inaccessible area was in the micropores of the graphite fibers that were not affected by the addition of Nafion.

Zoom In Zoom Out Reset image size

The addition of Vulcan XC-72 carbon black into the graphite felt material greatly increased the double-layer capacitance (Table I). The incorporation of 30 weight % of the Vulcan XC-72 and 2.5% of Nafion in the graphite felt yielded the highest capacitance. Thus, the surface area of the electrode was substantially increased without any noticeable impact on the resistance to flow through the electrode structure. The graphite felt treated with 2.5 wt% of Nafion and 30% carbon black was employed in further optimization of ORBAT performance.

After making the improvements to the electrode structures and flow fields as described above, a power density as high as 100 mW/cm² could be achieved with the high concentration of reactants in the acid form (Figure 8). Continuous cycling at a power density of 55 mW/cm² at about 23°C could be sustained at a round-trip energy efficiency of 70% for at least a 100 cycles (Figure 10). We expect to achieve higher power density by replacing BQDS with a redox couple with a higher standard reduction potential that will avoid Michael reaction transformations. With further optimization of the flow field structure, reduction of the thickness of the electrodes, and small increases in operating temperature, we expect that a power density of 300 mW/cm², similar to the standard operating power density of commercial vanadium redox flow batteries, can be achieved.^24,25

Zoom In Zoom Out Reset image size

Maintaining 100% coulombic efficiency at all depths of discharge is essential for large-scale energy storage applications. The charge-discharge efficiency of 100% was maintained for at least 100 cycles after the transformations in the first 4 to 5 cycles (Figure 10). Capacities as large as 5 Ah were realized from the cells. While the initial charge-discharge efficiency was low due to the afore-mentioned chemical transformations of BQDS, the charge-discharge efficiency stayed close to 100% once the transformations were complete. The value of 100% coulombic efficiency over the 100 cycles is consistent with the absence of crossover as confirmed by the electrochemical measurements, NMR data, and independent ex-situ studies.

The round-trip energy efficiency of ORBAT when charged and discharged at 100 mA/cm² was about 70% over 100 cycles. In each cycle 35% of the total capacity was introduced during charge and the same quantity was withdrawn during discharge. The voltage cutoff limits were held at 1.0 V and 0.0 V during charge and discharge, although these were not reached during the 100 cycles of testing. The electrolyte path length in the graphite felt electrodes (thickness of 2.5–3.0 mm) contributed significantly to the ohmic resistance of the cell. This resistance was considerably less with Toray paper electrodes. Thus, the value of energy efficiency was determined from the area under the charge and discharge curves after correcting the cell voltage for losses from the ohmic resistance in the cell. The value of energy efficiency can be improved further by reducing the mass transport losses. Such optimization studies are currently underway.