Performance Optimization of Differential pH Quinone-Bromide Redox Flow Battery

It is of paramount importance to operate redox flow batteries (RFB) at high power densities in order to minimize the stack and associated costs. With published data from our recent study [J. Mater. Chem. A, 5, 21875 (2017)], the cell potential of semi-organic, quinone-based RFB, was increased from 0.86 to 1.3 V by operating it in differential pH mode. The differential pH RFB uses bromine at pH ∼2 on the positive side and anthraquinone-2,7-disulfonate disodium (Na2AQDS) operated at pH ∼ 8 on the negative side. In the present work we have evaluated how the thicknesses of carbon paper and membrane, electrolyte flow rate, and redox species concentration affect the cell resistance and peak galvanic power density. The optimized cell delivered a maximum peak galvanic power density of 0.45 W/cm2 with an area resistance of 1 .cm2. The peak power density of differential pH battery was compared to those of acidic quinone-bromide and both are benchmarked against vanadium RFB tests in the same cell. It is shown that under identical conditions, the peak galvanic power density of the differential pH battery is 12% higher than acidic quinone-bromide, however lower than that of vanadium RFB tests. © The Author(s) 2018. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.0681816jes]

Currently one of the biggest challenges of further expansion of renewable PV and wind electricity is cost efficient storage. Here redox flow batteries (RFBs) are generally considered to have low-cost potential, because of independent scaling of power and capacity, long cycle life, and flexibility in configuration and operation. [1][2][3][4] Among the different RFB chemistries, all-vanadium RFBs are closest to a large commercial breakthrough due to its high reversibility, long life cycle, and minor cross-over issues. 1,2,[5][6][7][8] Nonetheless, further commercialization of vanadium RFBs is challenged by increasing cost of vanadium pentaoxide that has more than doubled during the past years, and nowadays it accounts for > 40% of the total capital costs of vanadium RFBs. 9,10 For this reason there is a need for investigations and development of more cost-effective (organic) redox species that decreases the capital costs of RFBs. 1,11,12 Various types of all-organic and semi-organic RFBs have been proposed during the past couple of years. Many of them suffer from low cell potential and low peak galvanic power density compared to the state-of-the-art vanadium redox flow battery. 1,[11][12][13] In general the cell power density scales with the cell potential squared and is inversely proportional to the internal cell resistance. The cell resistance is a sum of the contributions from electrical and membrane resistances, charge transfer resistance and mass transfer limitations, and the best performance in general is achieved with cation exchange membranes, protons as charge carriers and carbon papers as electrodes with high specific surface area. High current densities can be achieved even when electrochemical reaction kinetics are slow and charge transfer resistance is high by using porous carbon papers with a large specific surface. 14 One method to increase the cell potential in all-organic and semiorganic RFBs is by pH tuning of the whole cell or by differential pH in the two half cells. This method is applicable because the redox potential of organic redox pairs typically displays a strong pH dependence because the redox reactions involve protons, which typically is not the case for metals. The concept of differential pH batteries has earlier been applied. [15][16][17][18] Additionally, we recently demonstrated a semi-organic differential pH quinone-bromide battery based on bromide operated at pH∼2 on the positive side and anthraquinone-2,7disulfonate (AQDS) disodium operated at pH∼8 on the negative side that was successfully operated over a period of 30 days. 19 A cell potential as high as 1.3 V was achieved in that work, nonetheless, we obtained power densities that were significantly lower than the ones in the analogous acidic quinone-bromide battery. 20,21 Despite similar reaction rate constants for the redox reactions, it was speculated z E-mail: bentien@eng.au.dk whether the poorer performance was related to a non-optimized cell design, or higher electrolyte resistance because the AQDS side was operated close to neutral pH. To answer these questions we have in the present made a comparative study between an acidic and a differential pH quinone-bromide battery under semi-optimized conditions where the internal battery resistance is minimized with respect to redox solution composition, flow rate, electrode and membrane thickness. The comparison between the two battery chemistries are done in terms of battery polarization curves and electrochemical impedance spectroscopy (EIS). Additionally, the performance of the differential pH and acidic batteries were compared to standard vanadium RFB in the same cell. flow rate of redox solutions, and concentration of redox species. The number of carbon papers was varied from 3 to 10 on each side of the cell; the performance of two proton exchange membranes with similar ion exchange capacity but different thickness (Nafion 117 and 212) was compared; the flow rate of positive and negative electrolytes was adjusted to 20, 50, 80 and 120 ml/min; and lastly, the effect of concentration of Na 2 AQDS and on cell resistance was studied. Electrochemical impedance spectroscopy (EIS) and polarization curve measurements were used to examine the influence of the studied parameters on the cell resistance. EIS was performed with a CH Instruments 660E potentiostat under different experimental conditions in a frequency range between 0.01 Hz and 10 MHz with a perturbation amplitude of 5 mV. For comparison to EIS data, polarization curves were measured by charging and discharging the battery at constant current density. All polarization curves data were collected using battery testing system (Neware BTS-5V3A). During the test, data points in the polarization curve were recorded symmetrically by first setting a discharge current at fixed value and the record the cell voltage for 60 s, followed by a short resting period, and then charging with the same current for 60 s. Subsequent data points at other currents were recorded similarly. The potential in the polarization curve was the average over the 60 s interval, but did not change more than a few percent. The symmetric test procedure ensured that the state-of-charge (SOC) did not change significantly during the recording of the whole polarization curve at a specific (SOC). Only data for the discharge test are shown in the polarization curves.
Battery tests in optimal configuration.-Charge/discharge tests were performed in the cell depicted in Fig. S1. It consists of two Poco graphite plates with serpentine flow field provided from Fuel Cell Technologies (FCT). Thermally treated carbon papers with nominal thickness of 230 μm and geometric area of 4 cm 2 were used as electrodes. Five layers of carbon papers were used in each half-cell separated by a Nafion 212 membrane (thickness 50 μm in dry conditions). The cell was sealed with two Viton gaskets with the thickness of 0.5 mm and 1 mm when using three and five electrodes on each side of the cell, respectively. In the case of 10 electrodes on each side of the cell two Viton gaskets with thickness of 1 mm were used. The flow cell was assembled using two 8 mm × 12 cm × 12 cm steel endplates and seven bolts tightened with a torque of 4.5 N m. The redox solutions volumes were 5 ml and 25 ml on the negative and positive side, respectively. During all RFB tests, the negative side was under the nitrogen atmosphere. All battery tests were performed at room temperature except for the measurements of the differential pH battery which was measured at both room temperature and 40 • C. A two-channel Intelligent BT600L peristaltic pump was used to circulate the redox solutions at 120 ml/min. Charge/discharge measurements were carried out by using battery testing system (Neware BTS-5V3A) in a four-wire configuration. The differential pH quinone-bromide cell was charged at different current densities using upper cutoff voltage of 1.5 V and lower cutoff voltage of 0.4 V.

Results and Discussion
Cell optimization.-Thickness of the carbon electrode.-EIS measurements were performed with 3, 5 and 10 carbon papers on each side of the cell with equivalent compressions of 27%, 13% and 13%, respectively. The flow rate and concentration of redox species were kept at maximum (120 ml/min and 0.5 M Na 2 AQDS and 0.3 M Br 2 in 2 M NH 4 Br) with a Nafion 212 cation exchange membrane as separator. Fig. 1a shows the Nyquist plots for different number of the carbon paper electrodes. All data reported in the Nyquist plots (Figs. 1a to 1d) have been described with the equivalent circuit presented in Fig. 1e (see Eq. (S1) in Supplementary Information, section S2 23 ) using MEISP 3.0 software and the results of the fits are presented in section 2 of Supplementary information. In the Nyquist plots we interpret the first intercept (typically 10 5 -10 6 Hz) with the real axis (Z = 0) as the pure ohmic electrical and ion resistance (R ei = R e + R i ). This consists of contributions from electron transport in the cell (R e ), i.e. current collectors, graphite blocks, carbon paper electrodes and contact resistances between the carbon paper electrodes. 24,25 Here it is safe to assume that only the contact resistances have significant contributions. The ion resistance (R i ) has combined contributions from the membrane and electrolyte. Finally, we interpret the second and third intercept (Hz range) with the real axis as charge-transfer resistance (R ct = R a + R c ) from the anode (R a ) and cathode (R c ) sides. Attribution of second and third resistance to anode and cathode, respectively, is based on the electrochemical reaction rate of Na 2 AQDS and Br 2 /Br − redox couples. Based on standard heterogeneous electron transfer rate constant (k 0 ) of Na 2 AQDS at pH = 8 19 and Br 2 /Br − redox couple, 4,26 the charge transfer resistance of Br 2 /Br − is ∼122 times greater than Na 2 AQDS (Supplementary Information, section S3). However, we did not observe this ratio in our experiments (Supplementary Information, section S2) where the R c /R a ratios are of the order 1-2 depending on the experimental conditions. This indicates that effective thicknesses of the electrodes that are involved in the electrochemical reactions are larger for the Br 2 /Br − side. Furthermore, it also indicates that only a small fraction of the electrodes close to the membrane surface are involved in the electrochemical reactions. As can be seen from Fig.  1a and Table II, the number of carbon paper electrodes affects both R ei and R ct . R ei only varies between 0.59 .cm 2 and 0.68 .cm 2 for 5 and 10 carbon papers, respectively. A constant R ei is expected if it is dominated by the membrane resistance. Nonetheless, we attribute the small variation in R ei mainly to the contact resistance of electrodes and compression of membrane. By decreasing the number of the electrodes from 10 to 5, the porosity/compression remains constant but the contact resistance is decreased as seen by the relatively small decrease of R ei . [27][28][29][30][31][32][33] Further reduction to 3 carbon papers, increases R ei slightly which is somewhat counterintuitive. However the increase was very small (0.03 cm 2 ) and we anticipate that the higher compression leads to a lower solution uptake in the membrane and hence a lower ion conductivity. [34][35][36] Moving the attention to R ct , it is seen that R a decreases slightly with the number of carbon that R ct increases from 0.43 .cm 2 to 0.63 .cm 2 for 5 and 10 papers, respectively, while it is in between with 0.58 cm 2 for 3 papers. The interpretation of these results is not straightforward. On one hand it is expected that increasing the number of carbon papers will lead to higher surface area and thus reduce R ct . [30][31][32][33] However, this does not account for the changes in the fluid dynamic; indeed the local velocity close to the membrane-electrode interface will decrease as the number of carbon papers is increased because of the higher hydraulic resistance. For this reason, it is anticipated that the minimum R ct observed for 5 carbon papers is a trade-off between high electrode surface area and optimized electrolyte circulation through carbon papers. Based on these results five carbon papers are found the optimal value of the number of electrodes among the experimental conditions tested in the present work. Furthermore, battery polarization curves at different thicknesses of electrodes (Fig. S4) provide independent measurements of R tot which are in excellent agreement with the EIS measurements as seen in Table II. Membrane thickness effect on cell resistance.-R ei includes a relatively large contribution from the membrane (Supplementary Information, section S5) and is expected to be approximately linear with its thickness. 37,38 In the present work, we compared the cell performance using untreated Nafion 117 (183 μm) and untreated Nafion 212 (53 μm). In both cases, five carbon papers were used on each side of the cell and other operational conditions were kept as in the previous set of experiments with a flow rate of 120 ml/min, 0.5 M Na 2 AQDS and 0.3 M Br 2 in 2 M NH 4 Br. As can be seen in Fig. 1b and Table  S2b, both R ei and R tot are significantly decreased by switching Nafion 117 to Nafion 212. These results are in agreement with polarization curve results (Fig. S6) and with previous similar study on vanadium redox flow battery. 39 It is worth to notice that these two membranes have very similar characteristics i.e. ion exchange capacity (0.9 and 0.91 meq/g) [40][41][42] and solution uptake (∼20 and 21 H 2 O per SO 3 − ) of Nafion 117 and 212, respectively, 43-46 even though our experiments showed a different out-of-plane swelling. The morphological differences are expected to have only minor impact on the membrane resistivity and the resistance should scale with wet membrane thickness ratio which is about 3.5. The relative change of R ei is only about 2.2 and the difference can be attributed to electrical resistance in the cell (carbon paper, contact resistance, etc) and solution ionic resistance. If it is assumed that the intrinsic ionic conductivity of Nafion 117 and 212 is the same, the non-membrane R ei (R ei,nm ) is about 0.3 cm, 2 whereby the membrane R ei is about 1 cm 2 and 0.3 cm, 2 for Nafion 117 and Nafion 212 respectively. This is in good agreement with estimates from literature conductivity values of Nafion ( Supplementary Information, section S5). Furthermore, if it assumed that all the non-membrane R ei (R ei,nm ) is due to ionic resistance in the solution it is possible from the ion conductivity of the solution (σ = 206 mS cm −1 , Supplementary Information, section S12) to calculate that the total effective electrode thickness to be σ R ei,nm ∼ 0.6 mm, which is significantly lower than the total electrode thicknesses of approximately 2 mm. It is emphasized this is a maximum value since R ei,nm has significant contributions from the graphite cell, carbon paper electrodes and contact resistances in between. Furthermore, the model is very crude but corroborates with the above analysis of the R c /R a ratios which also indicated an effective electrode thickness significantly smaller than the geometric ones.
Flow rate and concentration effect on cell resistance.-Increased solution flow rate in cells with flow field design plays an important role in reducing concentration polarization and enhancement of convective mass transfer, which consequently leads to lower cell resistance. 24,47,48 Fig. 1c shows the EIS results (results from fitting in Table S2c) with flow rates from 20 ml/min to 120 ml/min. In agreement with previous studies 33,49,50 , the R tot decreases with the increase of flow rate due to the reduced mass transfer resistance. Based on Table S2c, as the flow rate increases, R ct decreases significantly but R ei does not experience a big change which is in good agreement with literature. 33 With respect to concentration effect, it was investigated with 0.1 M Br 2 in 2 M NH 4 Br and 0.05 M Na 2 AQDS in 2 M NH 4 Br (low concentration) and 0.3 M Br 2 in 2 M NH 4 Br and 0.5 M Na 2 AQDS in 2 M NH 4 Br (high concentration). As seen from Fig. 1d (Table S2d), R ct is more than halved in the high concentration compared to the low concentration experiment as expected. 51,52 Again, this is because of decreasing mass transfer resistance with increasing concentration of redox species which ultimately leads to lower R ct . Furthermore, it is observed that increased concentration of active material reduces conductivity of both positive and negative electrolytes about 5-10% (Table S7). This is in contradiction with the results from the EIS, where it is seen that R ei decreases about 10% which is explained by a lower membrane resistance. This is because at high concentration of the external phase the total ion uptake within the membrane markedly increases due to a reduced Donnan exclusion i.e. membrane is less selective to counter ions. This ultimately results in a higher ion conductivity. 53 The optimum values of the variables for the minimum cell resistance and highest output power density are shown in Table III.  Table III to evaluate the performance of optimized RFB. The theoretical capacity of the differential pH battery using 0.5 M of Na 2 AQDS and based on 1.5 electrons accessible per Na 2 AQDS molecule 54 is 101 mAh. The discharge capacity value at different current densities (Fig.  2a), varies from 99 mAh at 30 mA/cm 2 to 65.3 mAh at 250 mA/cm 2 . The decreasing capacity with higher current densities is due to the higher overvoltage because of the non-zero internal cell resistance. 55 This is also reflected in voltage efficiency which decreases from about 95% to less than 65% when current density is increased from 30 to 250 mA/cm 2 (Fig. S8). Given that energy efficiency is directly proportional to the voltage efficiency this also explains its 38% reduction at higher current densities (Fig. 2b). The current efficiency increases from 97.0% to 99.6% when the current density is increased from 30 mA/cm 2 to 250 mA/cm 2 (Fig. 2b). This is a consequence of shorter charging/discharging times at high currents and thus lower self-discharge, which mainly is caused by bromine crossover in the membrane. The charge/discharge curves shown in the Fig. 2c have a Nernstian behavior as reported previously. 19 With respect to the stability, the differential pH battery was operated for 100 charge/discharge cycles (Fig. 2d), and in our previous study for 30 days. 19 We partly ascribe the pH stability of negative electrolyte at pH∼8 and disequilibrium of negative and positive electrolytes after 100 cycles to the: i) presence of buffers NH + 4 /NH 3 and H 2 CO 3 /HCO − 3 /CO 2− 3 determined by the supporting electrolyte (NH 4 Br) and CO 2− 3 bound in Na 2 AQDS, respectively and ii) a less well-understood net/asymmetric transport of protons to the bromide side during battery cycling that balances the transport of Br 2 and HBr to the Na 2 AQDS side. This is discussed in detail in Ref. 19.

Differential pH quinone-bromide battery performance.-The battery tests were run for the optimal conditions presented in
Nonetheless, during one charge/discharge cycle, the pH of Na 2 AQDS side varies significantly (see e.g. Fig. S9a in Ref. 19) mainly due to the involvement of protons in reduction/oxidation of Na 2 AQDS. During battery cycling in the present study we did not observe any sudden drops in the cell potential during 100 cycles ('shoulders' in Fig. S9a in Ref. 19) due to pH values below ∼ 5 on the Na 2 AQDS side. This was mainly because the Na 2 AQDS side was well-sealed under N 2 atmosphere, whereby CO 2 could not escape the system, contrary to the tests in Fig. S9a in Ref. 19 where the Na 2 AQDS side was opened regularly for pH measurements and CO 2 could escape.
The average stable current efficiency of 99.5% was achieved after 100 cycles, which is a major improvement with respect to 95% found in our previous work. 19 Additionally, a capacity loss of 0.026% per cycle was observed during cycling which is ∼50% less than previous non-optimized cell. Lower capacity loss is most likely a consequence of shorter cycling times due to higher current densities. Two possible capacity loss mechanisms can be identified: Na 2 AQDS crossover through the membrane (from negative side to the positive side of the battery) and chemical degradation of Na 2 AQDS during cycling. A wide scan cyclic voltammogram recorded for the electrolyte on the positive side of battery after 100 cycles (Fig. S9) shows that there is no evidence of the presence of Na 2 AQDS on the positive, indicating that the capacity loss is related to chemical instability of Na 2 AQDS. 56

Power density of the differential pH quinone-bromide battery.-
The power density of differential pH quinone-bromide battery was obtained by running battery tests under the optimal conditions presented in Table III. Figs. 3a and 3b shows the polarization and power density curves at room temperature for differential pH quinone-bromide battery. The area resistance of around 1 .cm 2 was achieved under optimal conditions which is 3.5 times lower than previous work. 19 The peak galvanic power density of 0.4 W/cm 2 is found at 90% SOC and approximately 500 mA/cm 2 which is significantly higher than the previous work (Fig. S10). The power density of the differential pH battery can be boosted further up to 0.45 W/cm 2 by increasing the temperature to 40 • C as shown in Figs. 3c and 3d.   Performance comparison of acidic quinone-bromide, differential pH quinone-bromide and vanadium RFBs.- Table S14 provides the standard heterogeneous electron transfer rate constant (k 0 ) and diffusion coefficient of Na 2 AQDS in two different experimental conditions and vanadium redox species. In general, diffusion coefficient of redox couples affect the concentration gradient on the electrode surface meaning that the high diffusion coefficient leads to less polarization and mass transfer resistance. On the other hand, the electron transfer rate constant is in direct relationship with exchange current. The slower the kinetics of redox couples, the lower the current density provided by the RFB will be. As can be seen in Table S14, Na 2 AQDS at pH = 8, has higher diffusion coefficient and electron transfer rate constant than both Na 2 AQDS at pH = 0 and vanadium redox couple. These high values affect the supplied current density of RFB and power density which will be discussed in the following paragraph.
The peak galvanic power density of three different battery configurations was measured under the same conditions and using the same flow cell. The results were compared and are shown in Fig. 4. The acidic quinone-bromide battery using the electrolyte components in Table I, yields an area resistance of 0.46 .cm 2 and peak galvanic power density of 0.35 W/cm 2 at room temperature ( Fig. S11a and S11b show polarization curves and power density calculations, respectively). The area resistance and power density values are comparable with previously reported values of 0.4 .cm 2 and 0.6 W/cm 2 at 40 • C 20 and 0.56 .cm 2 at 30 • C. 57 However, our measured values are higher than 0.25 .cm 2 and lower than 1 W/cm 2 at 40 • C which have been obtained for higher bromine concentration (2 M) and flow rate (400 ml/min). 21 Additionally, comparison between the acidic and differential pH battery at room temperature shows that the acidic quinone-bromide battery, despite a lower cell resistance, has lower power density than the differential pH battery. In spite of lower conductivity of electrolytes in differential pH compared to the acidic battery ( Supplementary Information, section S12), the cell potential is 33% higher. Here the power density generally scales with U 2 Cell / R tot and it underlines the importance of a high cell potential. Also, electron transfer rate constant and diffusion coefficient as discussed above are higher, which leads to higher power density value .The vanadium redox flow battery as shown in the Fig. 4 has a peak galvanic power density of 0.6 W/cm 2 at room temperature which is within the range of the values (0.1-1.3 W/cm 2 ) reported in the literature. 39,58,59 From the polarization curve (Fig. S13) the power density is clearly limited by mass transport and the resistance in the linear region is 0.4 .cm 2 . The power density value is comparable with literature values while the area resistance in our work is lower, 37,60 indicating that a power density above 0.6 W/cm 2 can be obtained with higher circulation rates. Compared to the acidic and differential pH quinone-bromide battery, vanadium redox flow battery has lower ohmic resistance because of high conductivity of electrolytes (2 M sulfuric acid) and lower diffusion coefficient and electron transfer rate constant 61,62 while higher peak galvanic power density than both quinone-bromide batteries.

Conclusions
High cell potential and power density are key parameters for redox flow batteries. Previously, it has been shown that differential pH quinone-bromide battery delivers high cell potential of 1.3 V and poor power density. In this work, we have studied effect of several factors to minimize the cell resistance and consequently improve the output power density. The cell resistance was reduced by using optimal conditions for the thickness of electrodes, flow rate, membrane thickness and concentration of redox species. The differential pH quinonebromide battery under optimal conditions has an area resistance of 1 .cm 2 and a maximum power density of 0.45 W/cm 2 . The power density value is 12% higher than acidic quinone-bromide battery and somewhat more lower than the highest value for vanadium redox flow battery under the identical conditions. Given the results shown in this work it is realistic that with further cell optimization, e.g. flow field geometry, electrode compression, and membrane morphology, the differential pH quinone-bromide battery can reach performance similar to the acidic quinone-bromide and vanadium RFB.