A High Efﬁciency Iron-Chloride Redox Flow Battery for Large-Scale Energy Storage

We report advances on a novel membrane-based iron-chloride redox ﬂow rechargeable battery that is based on inexpensive, earth- abundant, and eco-friendly materials. The development and large-scale commercialization of such an iron-chloride ﬂow battery technology has been hindered hitherto by low charging efﬁciency resulting from parasitic hydrogen evolution at the negative electrode and high overpotential losses. We have demonstrated a high charging efﬁciency of 97% by maintaining the negative electrolyte at a pH value of 2 and by using indium chloride as an electrolyte additive. The high charging efﬁciency of the negative electrode was found to be stable over at least 50 cycles. Further, we have demonstrated that with a graphite felt electrode, the overpotential losses were substantially mitigated at the positive and negative electrodes allowing the electrodes to be operated at current densities as high as 100 mA/cm 2 . With these technical advancements, the iron-chloride redox ﬂow battery has an increased prospect of being a sustainable and efﬁcient solution for large-scale energy storage.

Redox flow batteries for large-scale energy storage.-Redox flow batteries are particularly well-suited for large-scale energy storage applications. 3,4,[12][13][14][15][16] Unlike conventional battery systems, in a redox flow battery, the positive and negative electroactive species are stored in tanks external to the cell stack. Therefore, the energy storage capability and power output of a flow battery can be varied independently to suit the desired application. For example, if an application requires a battery with high energy content, the amount of electroactive species in the tanks can be increased without significant modification to the cell stack. As the electroactive material is held in the external tanks, an increase in the amount of electroactive species does not require an increase in the volume of the cell stack. Thus, modifications to the battery design for increased energy output can be realized at minimal cost and without any impact on the power output of the system. In contrast, any attempt to increase the energy content of a conventional battery often results in a significant increase in the system cost if the specific power has to be maintained.
Among the various redox flow battery chemistries that have been studied, the all-vanadium redox flow battery has seen the widest commercial deployment. Systems as large as 250 kWh to 1 MWh have been * Electrochemical Society Student Member.
* * Electrochemical Society Fellow. z E-mail: srnaraya@dornsife.usc.edu demonstrated. 17 Since the all-vanadium redox flow battery uses just vanadium redox-active species in the positive and negative electrolyte, any electrolyte cross-over or capacity imbalance that occurs during operation can be addressed by appropriate electrolyte re-balancing external to the cells. However, the high cost, limited earth-abundance, and the relatively high toxicity of vanadium present challenges to the widespread adoption of vanadium-based batteries. Similar to the all-vanadium system, the iron-chromium redox flow battery also uses fully soluble redox species in both the positive and negative electrolytes. 18 However, preventing crossover of materials from one electrode compartment to the other requires special membranes. 19 While significant progress has occurred in the last five years in the deployment of the iron-chromium redox flow battery, the low natural-abundance of chromium, and the poor kinetics of the chromium electrode would be barriers for large-scale use. 15,20 Flowbattery systems based on zinc-bromine and iron-chloride chemistries are considered "hybrid" redox flow systems as metal deposition occurs from a solution of metal ions at the negative electrode, while the positive electrode operates on a redox couple that is soluble in both the oxidized and reduced forms. 16 Our interest focuses on the iron-chloride flow battery (also referred to as the "all-iron" battery) that relies on the use of iron-based materials at both electrodes. With the abundant earth resources of iron and chloride, this type of battery is particularly attractive for large-scale applications.
Operating principle of an iron-chloride redox flow battery.-A schematic of the principle of operation of this system is shown in Figure 1. The redox chemistry of the iron-chloride redox flow battery is based on the iron (II) chloride/iron (III) chloride couple at the positive electrode and the iron (II) chloride/metallic iron couple at the negative electrode. The reactions that take place during the charging and discharging of an iron-chloride redox flow battery are shown schematically in Figure 1 and in the following chemical equations.
Negative electrode : Overall cell reaction : During discharge of the battery, iron (III) chloride is reduced to iron (II) chloride at the positive electrode. At the negative electrode, metallic iron is dissolved into the electrolyte as iron (II) chloride; these processes are reversed during battery charging (Eqs. 1 and 2). The open-circuit voltage of the iron-chloride redox flow battery is about 1.21 V.
Such an all-iron redox flow battery was first reported by Hruska and Savinell in 1981. 21 Several attributes make this type of battery suitable for large-scale energy storage applications. However, the successful commercialization of this iron-chloride redox flow battery has been hindered by technical challenges such as low charging efficiency of the negative electrode, self-discharge by electrolyte cross-over, and poor cycle-life. In their work, Hruska and Savinell reported the performance of an all-iron redox flow cell with titanium and carbon electrodes. 21 The cell used a microporous separator to hinder the mixing of the positive and negative electrolytes and was operated at 60 • C at current densities as high as 60 mA/cm 2 . The cell performance varied significantly during 50 cycles of testing and the highest energy efficiency observed was 50%. They also observed an increase in the resistance of the separator due to the precipitation of iron hydroxide. 21 Bartolozzi et al. also reported a similar all-iron based battery with a planar graphite electrode for the positive and negative electrodes and a semi-permeable membrane separating the two electrode compartments. 22 The authors observed that the concentration of ferric ions in the negative electrolyte increased from zero to 167 mg l −1 within eight hours of operation of the cell indicating leakage of ferric ions through the membrane. The authors have also re-affirmed that the iron-chloride redox flow battery has long life, is eco-friendly and uses raw materials that are inexpensive. 22 While the early work by Savinell and Bartolozzi et al. outlined the various factors that affect the performance of the iron-chloride redox flow battery system, subsequent reports aiming at addressing these technical challenges are scarce in the literature. Recently, the results from the operation of a 1 kW "all-iron" system operating at 50 A and 20 V has been reported by Trunov and Yminskii. It was a 20-cell system with 1 m 2 electrodes and was cycled for about 20 times yielding an energy density of 3.3 kWh/m 3 . 23 Low charging efficiency of the negative electrode.-During charging of the iron-chloride redox flow battery, the reaction at the negative electrode is the deposition of iron by the electro-reduction of ferrous ions (Eq. 2). The standard reduction potential of the Fe 2+ /Fe couple is −0.44 V and is about 440 mV more negative to that for the hydrogen evolution reaction (Eq. 4).
Thus, elemental iron is thermodynamically unstable under such conditions and hydrogen would evolve even at open circuit. Such corrosion of iron in acidic media accompanied by the hydrogen evolution reaction is well known. 24 Further, during charging of the iron-chloride redox flow battery, the electro-deposition of iron presents substantial overpotential thereby facilitating the hydrogen evolution reaction further. Thus, a significant amount of the input charge is directed towards hydrogen evolution resulting in a low charging efficiency. In addition, the evolution of hydrogen also results in an increase in the pH of the negative electrolyte. If the pH of the negative electrolyte increases beyond 3, precipitation of iron (III) and iron (II) hydroxides would occur. Maintaining the pH of the negative electrolyte below 3 is essential to the stable long-term performance of the iron-chloride redox flow battery. Thus, to achieve stable operation and high round-trip energy efficiency it is imperative that the charging efficiency at the negative electrode be close to 100%.
Approaches to suppress hydrogen evolution.-The rate of hydrogen evolution during charging is dependent upon the concentration of hydronium ions in the negative electrolyte. The natural pH of a 3 M solution of iron (II) chloride tetrahydrate is in the range of 0.7 -1.0. This low pH value facilitates the evolution of hydrogen during charging. We have focused on inhibiting the hydrogen evolution reaction by two approaches: (1) introducing limitations to the transport of hydronium ions to the electrode surface, and (2) inhibiting the kinetics of the charge-transfer process that leads to the evolution of hydrogen.
Recently, Savinell et al. have reported the effect of supporting electrolyte type and electrolyte pH on the rate of hydrogen evolution on iron electrodes. 25 As the electrolyte pH was increased from 1 to 3, the hydrogen evolution current was observed to decrease from 15 mA/cm 2 to 0.7 mA/cm 2 at −0.8 V (vs Ag/AgCl). Also, the hydrogen evolution current density was reported to be much lower in a chloridebased electrolyte compared to an electrolyte based on the sulfate anion. The highest faradaic efficiency of iron deposition on copper of 97% was observed in a sodium chloride electrolyte. 25 Advantages of the membrane-based system.-Our configuration of the iron-chloride redox flow battery uses an anion-exchange membrane separator between the two electrode compartments. This configuration has several advantages compared to that of the porousseparator-based cell configuration reported in the literature. 21 The membrane-based configuration prevents the crossover of reactants between the positive and negative electrode compartments. Loss in charging efficiency of the negative electrode due to any ferric ion that crosses over from the positive electrolyte is completely avoided. Also, the anion-exchange membrane separator allows different pH values to be maintained on each side of the cell. In the porous-separator-based cell design, the facile crossover of the electrolyte equalizes the pH in the two compartments leading to undesirable consequences. For instance, in the porous-separator-based design, the need to maintain the pH of the negative electrolyte around 2 (to suppress hydrogen evolution) requires the positive electrolyte to be kept around this pH value as well. At this pH, the ferric ions in the positive electrolyte will start forming hydroxide precipitates. To prevent the precipitation of these iron hydroxides, complexing agents are added to the positive electrolyte. While such complexing agents improve the solubility of the ferric ions, they can potentially hinder the kinetics of charge transfer for the positive electrode reaction. Further, the cross-over of these complexing agents to the negative electrolyte through the porous separator could have unintended consequences. A recent report from Prof. Savinell's group has documented the effect of various chelating ligands on preventing the precipitation of hydroxides in the positive electrolyte. 26 The effect of various ligands on the electrochemical and transport properties of the ferric/ferrous positive electrode couple has been studied. Among the various additives studied, glycine was found to inhibit the precipitation of iron (III) without significantly altering the diffusion coefficient and electrochemical characteristics of the ferrous/ferric couple. 26 However, the glycine additive was observed to decrease the plating efficiency of the negative electrode indicating the challenges of using a porous-separator-based flow cell design. 25,26 One of the disadvantages in the use of an anion-exchange membrane is the need for such a membrane to have adequate conductivity for the anion that is transferred -in this case, the chloride ion. Among other challenges are the stability of the membrane in acidic and oxidative environments, and the added cost of the membrane. Therefore, a thin, mechanically-robust, and inexpensive membrane with low ionic resistance and high chemical durability will be required.

Experimental
Measurements of the faradaic efficiency for the electro-deposition of iron in various electrolytes were conducted using a three-electrode "half-cell" with a rotating disk glassy carbon working electrode. The effect of additives and electrolyte pH on the charge/discharge performance of an iron-chloride redox flow battery was studied in a "full-cell" with electrolyte flowing across two electrodes separated by an anion-exchange membrane.
Faradaic efficiency of iron deposition.-The half-cell set up consisted of a glassy carbon rotating disk working electrode (RDE) (Pine Instruments), a silver/silver chloride (Ag/AgCl) reference electrode and a platinum wire counter electrode. The electrolyte was an aqueous solution of iron (II) chloride tetrahydrate (3 moles liter −1 ) with ammonium chloride (2 moles liter −1 ), the latter serving to increase the conductivity of the solution. Experiments were conducted with various electrolyte additives that included ascorbic acid, citric acid, indium chloride, or bismuth oxychloride. The concentration of each of these additives in the electrolyte has been indicated along with the performance results. During preparation and electrochemical testing, the electrolyte was de-aerated and a blanket of argon gas was maintained to prevent the oxidation of ferrous ions to ferric ions. The glassy carbon electrode was rotated at 2500 rpm in all the measurements to provide adequate mass transport of the ferrous ions. The experiments with the half-cell were performed using a PAR VMC-4 potentiostat. The faradaic efficiency of electro-deposition of iron was determined by plating iron on to the glassy carbon RDE at various values of current density in the range of 10-300 mA/cm 2 followed by anodic galvanostatic stripping of the iron deposit at 60 mA/cm 2 . The stripping step was terminated at a potential of 0 V vs the Ag/AgCl reference electrode. The total charge during deposition was a constant of 0.17 Coulombs at all the values of current density studied. Such a limit to the charge input curtailed the deposit thickness and avoided physical dislodgement of loosely-bound and dendritic deposits. The faradaic efficiency was calculated from the ratio of the charge obtained during stripping and the charge used in plating. The faradaic efficiency measurements were performed at different current densities of plating with different electrolyte compositions.
Characterization of iron-chloride redox flow cells.-A schematic of the iron-chloride redox flow cell used in this study is shown in Figure 2. Polypropylene was used for the cell housing. The electrodes were made from densified graphite (Graphtek LLC) with a facial cross-sectional area of 25 cm 2 .
A platinum-plated titanium rod was threaded into the graphite electrode and used as the current collector. An anion-exchange membrane (Tokuyama A901, 11 μm thickness) was used to separate the positive and negative electrode compartments. Silver/silver chloride (Ag/AgCl) reference electrodes were introduced into the positive and negative electrode compartments to facilitate the measurement of the individual electrode potentials during cell charging and discharging. A photograph of the flow-cell set up used for the experimental studies is shown in Figure 3.
The electrolyte reservoirs contained about 250 ml of solution. The positive and the negative electrolyte solutions were circulated during the tests at about 1 to 1.2 liter/min. The charging efficiency of the iron-chloride redox flow cell was determined by charging the cell at 20 mA/cm 2 for 150-1200 seconds followed by discharge at 20 mA/cm 2 . A cut-off potential of 0 V (vs Ag/AgCl) on the negative electrode was used to terminate the discharge step. The ratio of the output capacity during discharge to the input charge yielded the charging efficiency of the negative electrode. A PAR Versastat 3 potentiostat coupled to a booster (KEPCO) was used to perform polarization and cycling studies with the flow cell. During the charging and discharging experiments on the flow cell, the potentials of the positive and negative electrodes were simultaneously monitored against the reference electrodes in the respective compartments using the potentiostat and a Keithley multimeter (Model 3706).
Correction of electrode potentials for ohmic drop.-The electrode potentials measured during the charging and discharging of the ironchloride redox flow cell were corrected for the voltage drop across the series equivalent ohmic resistance. The impedance of the positive electrode and negative electrode was measured in a three-electrode configuration as a function of the frequency of the alternating current excitation signal. For this measurement, the working electrode lead of the potentiostat was connected to the electrode of interest, the reference electrode lead was connected to the silver/silver chloride reference electrode in the same electrode compartment, and the counter electrode lead was connected to the opposite electrode (positive or negative). The high-frequency intercept of the impedance on the complex-plane plot was noted as the series equivalent ohmic resistance.

Results and Discussion
Faradaic efficiency for the electro-deposition of iron.-An example of the potential-charge curves measured during the constant current deposition and stripping of iron on the glassy carbon RDE is shown in Figure 4.
The faradaic efficiency of iron deposition is calculated from the ratio of the charge delivered during the stripping step to the charge  input during the deposition of iron as %Faradaic Efficiency = Q stri pping /Q deposition * 100 [5] where Q stripping and Q deposition are the quantities of charge transferred during the stripping and plating process, respectively. In the baseline electrolyte (3 M iron (II) chloride tetrahydrate with 2 M ammonium chloride) at low current densities of deposition, the faradaic efficiency was about 60% ( Figure 5). As the plating current density was increased above 50 mA/cm 2 , the faradaic efficiency decreased further; the faradaic efficiency was about 40% at 150 mA/cm 2 . Thus, in the absence of any mass transport limitations, the rate of charge transfer for the evolution of hydrogen increases with the current density and increasing overpotential.
The rate of hydrogen evolution during iron deposition can be decreased by increasing the pH of the electrolyte and limiting the mass transport of hydronium ions. However, when we attempted to increase the pH of the baseline electrolyte by adding a weak base such as ammonium hydroxide, rapid precipitation of iron hydroxides was observed. To increase the electrolyte pH without causing precipitation of the hydroxides, we tested the effect of complexing agents such as ascorbic acid and citric acid. These additives prevented the precipitation of the iron hydroxides in the electrolyte during operation. Ascorbic acid, citric acid, tartaric acid and malonic acid are frequently used in iron plating baths to prevent the precipitation of hydroxides. 27 Further, complexation by ascorbic acid is also used in the nutrition industry to enhance the bio-availability of iron. 28,29 In the presence of ascorbic acid or citric acid, the pH of the electrolyte could be increased to 2 by adding ammonium hydroxide without any precipitation of Fe 2+ or Fe 3+ hydroxides. The faradaic efficiency of iron deposition using this modified electrolyte (3 M iron(II) chloride tetrahydrate + 2 M ammonium chloride + 0.3 M ascorbic acid) at pH 2 was greater than 95% ( Figure 5). The amount of ascorbic acid was chosen to be 0.3 M based on the minimum amount required for ensuring complete solubility of iron (II) chloride in the negative electrolyte. Similar high faradaic efficiencies were also observed when citric acid was used as the stabilizing agent ( Figure 5). Ascorbic acid and citric acid with their pK a values in the range of 3.1 to 4.2 provide some buffering action, preventing rapid changes in the electrolyte pH during the electrodeposition and dissolution of iron. 27 When depositing iron from the electrolyte containing ascorbic acid, the faradaic efficiency was about 90% at 10 mA/cm 2 . The faradaic efficiency increased further with increasing current density reaching a value of 96% at 50 mA/cm 2 . With further increase in the plating current density, the faradaic efficiency stayed constant; no further increase in efficiency was observed ( Figure 5). Due to the relatively high pH of the electrolyte, the hydrogen evolution reaction was limited by the mass-transport of hydronium ions even at low values of plating current. The low concentration of protons in the bulk establishes a steep-gradient for proton transport to the electrode surface. With increasing total current density, the fraction of the current attributed to hydrogen evolution decreases because of the foregoing mass transport limitation. Thus, the fraction of the plating current diverted to the iron deposition reaction increased with increasing current density. However, as the deposition of iron from the ferrous ions in the electrolyte also became limited by mass transport, the partial current of iron deposition did not increase. Thus, the charging efficiency reached a constant value in the mass-transport-limited regime for both iron(II) and hydronium ions.
The maximum faradaic efficiency of iron deposition that can be obtained at a given concentration of iron (II) species and electrolyte pH can be estimated from the limiting currents for the iron deposition and hydrogen evolution reactions. The steady-state limiting current for iron deposition and hydrogen evolution can be related to the concentration from the Nernst diffusion layer model as, i lim = n F DC * /δ [6] Where i lim is the limiting current density, n is the number of electrons involved in the reaction, F is the Faraday's constant, D is the diffusivity of the electroactive species, C * is the concentration of the species in the bulk of the electrolyte and δ is the diffusion-layer thickness at the rotating disk electrode. The thickness of the diffusion layer at the rotating disk electrode during iron deposition can be calculated using the following expression.
Where ν and ω are the kinematic viscosity of the electrolyte and the angular velocity of rotation, respectively. The diffusion coefficient for the H + ion is 9.31×10 −5 cm 2 /s and that of Fe 2+ is about 0.72×10 −5 cm 2 /s; 30 the kinematic viscosity of water is 10 −2 cm 2 /s. Using these parameters and the corresponding bulk concentration of protons and iron (II) ions in the electrolyte, the limiting current density for hydrogen evolution and iron deposition are 0.043 A/cm 2 and 4.67 A/cm 2 , respectively. From the ratio of the limiting currents, the maximum efficiency of iron deposition for the electrolyte at a pH of 2 is calculated to be about 99%. This value of faradaic efficiency is consistent with the 96% efficiency observed in our experiments ( Figure 5). Thus, we understand as to how an iron electrode operated at high current densities in electrolyte of pH∼2 exhibits a high faradaic efficiency of iron deposition. However, operating the cell at such high current densities may not be feasible in the full-cell configuration both from a voltage efficiency standpoint and also due to the variable charging demands of a particular energy storage application. To overcome this challenge, we must inhibit the kinetics of charge-transfer for the hydrogen evolution reaction. To this end, we have studied the effect of indium chloride and bismuth oxychloride as additives to the negative electrolyte.
Flow cell experiments -Effect of additives and electrolyte pH on charging efficiency.-The potential-time curves measured during the charging efficiency measurement of the iron-chloride redox flow cell using a negative electrolyte containing ascorbic acid are shown in Figure 6. The measured charging efficiency on the negative electrode in the flow cell was consistent with the faradaic efficiency of iron electro-deposition measured in the half-cell testing ( Figure 5). When the baseline negative electrolyte (3 M iron (II) chloride tetrahydrate + 2 M ammonium chloride) was used at its natural pH (0.9-1.0), the charging efficiency of the negative electrode was measured to be about 63% (Figure 7). In the presence of ascorbic acid and raising the electrolyte pH to 2 (with ammonium hydroxide), the charging efficiency increased to 83% (Figure 7). Based on the results from the half-cell experiments with a similar electrolyte, ( Figure 5) the charging efficiency values are expected to be in the range of 80-90% at 20 mA/cm 2 . Thus, the measured charging efficiency at the negative electrode in the flow cell was consistent with the results of half-cell measurements ( Figures 5, 7).
Another approach to suppress hydrogen evolution on the negative electrode during charging is to increase the overpotential for hydrogen evolution. Indium and bismuth are known for their sluggish kinetics of hydrogen evolution. The exchange current densities for hydrogen evolution on indium and bismuth in acid electrolytes are about four orders of magnitude lower than that on iron. 31 Further, indium and bismuth exhibit extremely limited miscibility with iron and thus tend to segregate to the surface. 32,33 Therefore, when indium and bismuth are employed as electrolyte additives, the charge-transfer kinetics of hydrogen evolution can be expected to be inhibited in the same manner as with elemental indium or bismuth.
When indium chloride was present in the electrolyte (3 M iron (II) chloride tetrahydrate + 2 M ammonium chloride + 0.3 M ascorbic Acid, pH∼0) at even a fairly small concentration of 0.2 mM, the charging efficiency increased to 83% (Figure 7). We restricted the amount of indium chloride to be small enough so that the limiting current density for the deposition of indium is <0.1% of that for the deposition of iron. When the pH of this indium chloride containing electrolyte was increased to 2, the charging efficiency of the negative electrode in the flow cell reached a value of 97%. From the half-cell studies (Figure 7), we observe that at the same current density, an indium-free electrolyte with pH∼2 had a faradaic efficiency of only 86%. The increase in efficiency in the presence of indium chloride in the electrolyte results from the inhibition of the charge transfer process for hydrogen evolution in the presence of in situ electrodeposited indium metal. Therefore, the addition of indium chloride has enabled high charging efficiencies to be obtained even at low charging current densities -before mass-transport limitations set in.
When the negative electrode is charged in the presence of an electrolyte containing indium chloride, in addition to the reduction of iron (II) to metallic iron, the reduction of indium (III) ions to metallic indium also occurs. The standard reduction potential for the indium (III)/indium couple is -0.34 V. 34 Thus, the deposition of indium can occur as the potential of the negative electrode during charging is sufficiently negative to the reversible potential of the In(III)/In couple ( Figure 6). As discussed earlier, the deposited indium tends to be segregated and stays on the surface because of its immiscibility with iron. Even a monolayer of indium that stays on the surface of the iron is sufficient to hinder the kinetics of hydrogen evolution. This type of immiscibility is important to the success of this approach, as alloying will reduce the surface concentration of indium and the effective inhibition of hydrogen evolution.
In addition to indium, bismuth has been used as an additive for suppressing hydrogen evolution in alkaline zinc-, and iron-based batteries because of the high hydrogen overpotential on bismuth. [35][36][37] Therefore, we investigated the effect of adding bismuth oxychloride to the electrolyte on the charging efficiency of the negative electrode. However, we observed that bismuth oxychloride did not have a noticeable impact on the charging efficiency of the negative electrode. We also noted that bismuth oxychloride was relatively insoluble in the negative electrolyte, unlike indium chloride. Thus, it is quite likely that we were not able to introduce sufficient concentration of bismuth ions into the electrolyte solution to have an impact on the hydrogen evolution reaction.
Given the high charging efficiency of 97% with indium chloride additive, the negative electrolyte with ascorbic acid and indium chloride at a pH of 2 was chosen for further studies. The specific composition of the negative electrolyte used in these studies was 3 M iron (II) chloride tetrahydrate, 2 M ammonium chloride, 0.3 M Ascorbic Acid, and 0.2 mM indium chloride at pH∼2.
Flow cell experiments-Effect of charging and discharging rate on efficiency.-The charging efficiency of the negative electrode at different current densities of charge and discharge is shown in Figure  8. Consistent with the results of the half-cell experiments (Figure 5), the charging efficiency of the negative electrode showed an increase with increasing current density. The charging efficiency of the negative electrode also showed a lower value when the pH was ∼0 instead of ∼2.2. The charging efficiency reached a value as high as 97% at current densities as high as 40 mA/cm 2 . The high charging efficiency observed even at 40 mA/cm 2 suggests that mass transport limitations in the flow cell may be reached at lower current densities than in the RDE experiment.  charging, the oxidation of ferrous chloride to ferric chloride takes place; the reverse process occurs during discharge (Eq. 1). The kinetics of electron-transfer for the Fe 2+ /Fe 3+ couple are facile and high rates can be sustained without significant activation overpotential. The standard exchange current density for the Fe 2+ /Fe 3+ reaction has been measured to be about 1-5 mA/cm 2 . 24 Due to its facile and reversible kinetics, the Fe 2+ /Fe 3+ reaction is a very attractive candidate as a positive electrode for redox flow batteries. 18,38 The composition of the positive electrolyte used in this study was a mixture of 2 M iron (II) chloride tetrahydrate with 1 M iron (III) chloride and 2 M ammonium chloride in 1 M hydrochloric acid solution. When a graphite block was used as the current collector, the limiting current density on the positive electrode was 50 mA/cm 2 (diffusion layer thickness is about 0.3 mm) during charging and 30 mA/cm 2 during discharging. At 20 mA/cm 2 for example, the charging and discharging overpotentials were about 50 mV. In order to increase the current at the positive electrode, graphite felt (Sigracell) was used to pack the space between the graphite current collector and the membrane. The enhancement of the electrode performance was clearly seen in the polarization curves where the overpotential at high currents was found to be low (Figure 9). For example, both the cathodic and anodic overpotentials at 30 mA/cm 2 decreased by more than 50 mV after the addition of graphite felt to the positive electrode compartment. The polarization curves also did not exhibit any mass-transport limitation even at high current densities of charging and discharging. Current densities as high as 100 mA/cm 2 were achieved with <20 mV of overpotential. The area enhancement resulting from the graphite felt was calculated from double layer capacitance measurements to be about 80. Thus, a substantial decrease in effective current density is expected to result from the use of the graphite felt. Modifications to the graphite felt (similar to that reported by Nguyen et al: growth of carbon nanotubes) will result in even greater area enhancements. 39 Further, the fluid flow field used in the full cell has not been optimized for the complete utilization of the active area of the graphite felt. Increasing the operating temperature will decrease the viscosity and also increase the diffusion coefficient facilitating mass transport in the diffusion layer. These modifications are expected to increase the limiting current density on both the positive and negative electrodes. The cell voltage (E cell ) during charge and discharge was determined from the potentials of the positive and negative electrodes measured against the silver/silver chloride reference electrode in the respective compartments. The measured electrode potentials were corrected for all the voltage losses arising from series equivalent ohmic resistances at the electrodes. This arrangement also allowed us to ignore the significant voltage loss contribution across the membrane and electrolyte layers that were not part of the electrode. Thus, the charge-transfer and mass transfer polarization losses at the electrodes could be studied.
During charging, During discharging, Here, E P and E N are the measured potentials of the positive and negative electrodes, respectively. R ,P and R ,N are the ohmic resistances of the positive and negative electrode determined from the impedance data at high frequency. I charge and I discharge are the charging and discharging currents, respectively. We observed that with increasing current density, the discharge voltage of the cell decreased ( Figure 10). During discharge, we observed a well-defined plateau that corresponded to the dissolution of iron from the negative electrode. Once all the iron was stripped from  the electrode surface, the cell voltage dropped rapidly. The second plateau at low voltages during discharge at 10 mA/cm 2 could be due to the oxidation of iron (II) to iron (III) at the negative electrode. The cell voltage difference (E cell,charge -E cell,discharge ) between charge and discharge is shown as a function of the operating current density in Figure 11.
Comparing the cell voltage losses with the changes in electrode potential at the negative electrode (Figures 12a, 12b), it is clear that 95% of the cell voltage losses could be attributed to processes at the negative electrode. At 10 mA/cm 2 , the overpotential at the negative electrode is about 140 mV during charging and 100 mV during dis-  To verify the effect of mass transport and increased electrode area on the performance of the negative electrode, we changed the electrode configuration by including a graphite felt layer. With the graphite felt, the overpotential at the negative electrode decreased by almost 50% at all the current density values ( Figure 13). While we had considered a planar electrode initially for the negative electrode, it appears that a porous electrode structure has significant benefits in reducing the overpotentials for the deposition and stripping of iron.
Flow cell studies -Stable operation of the cell during prolonged cycling.-Repeated cycling of the iron-chloride flow cell with the high-efficiency electrolyte at 40 mA/cm 2 demonstrated that a high charging efficiency (> 95%) could be maintained over 50 cycles (Figure 14). Visual examination of the membrane did not suggest any signs of degradation or swelling after this cycling test. This observation is consistent with the chemical stability of the Tokuyama membrane under acidic conditions. These results of repeated cycling confirm the viability of the new electrolyte composition and the use of the anion-exchange membrane separator for designing a high-efficiency iron-chloride flow battery.

Conclusions
We have demonstrated a high-efficiency iron-chloride redox flow battery with promising characteristics for large-scale energy storage applications. The advances demonstrated in this study show a path for the deployment of large-scale systems based on the iron-chloride flow battery concept originally discussed by Savinell and others. With a new electrolyte formulation consisting of complexing agents such as ascorbic acid, we were able to avoid precipitation of the iron hydroxides and stabilize the electrolyte at a pH value of 2. Further, with small amounts of indium chloride in the negative electrolyte, we were able to achieve charging efficiency values up to 97%. The high charging efficiency of the negative electrode was found to be stable over at least 50 cycles. High-surface area graphite felt electrodes reduced the overpotentials not only at the positive electrode but also the negative electrode. As a result, current densities >100 mA/cm 2 could be achieved even at room temperature. We have demonstrated that an iron-chloride flow battery design using an anion-exchange membrane separator with a unique formulation of the negative electrolyte provides significant advantages for long periods of operation, avoiding the difficulties of electrolyte crossover or precipitation. With the potential to achieve the objectives of high efficiency and low cost, the membrane-based iron chloride flow battery could be a very attractive candidate for large-scale energy storage.