Redox flow batteries—Concepts and chemistries for cost-effective energy storage

Electrochemical energy storage is one of the few options to store the energy from intermittent renewable energy sources like wind and solar. Redox flow batteries (RFBs) are such an energy storage system, which has favorable features over other battery technologies, e.g. solid state batteries, due to their inherent safety and the independent scaling of energy and power content. However, because of their low energy-density, low power-density, and the cost of components such as redox species and membranes, commercialised RFB systems like the all-vanadium chemistry cannot make full use of the inherent advantages over other systems. In principle, there are three pathways to improve RFBs and to make them viable for large scale application: First, to employ electrolytes with higher energy density. This goal can be achieved by increasing the concentration of redox species, employing redox species that store more than one electron or by increasing the cell voltage. Second, to enhance the power output of the battery cells by using high kinetic redox species, increasing the cell voltage, implementing novel cell designs or membranes with lower resistance. The first two means reduce the electrode surface area needed to supply a certain power output, thereby bringing down costs for expensive components such as membranes. Third, to reduce the costs of single or multiple components such as redox species or membranes. To achieve these objectives it is necessary to develop new battery chemistries and cell configurations. In this review, a comparison of promising cell chemistries is focused on, be they all-liquid, slurries or hybrids combining liquid, gas and solid phases. The aim is to elucidate which redox-system is most favorable in terms of energy-density, power-density and capital cost. Besides, the choice of solvent and the selection of an inorganic or organic redox couples with the entailing consequences are discussed.


Principles of operation for redox flow batteries
Electrochemical energy storage (EES) devices store energy for later use. To clearly showcase the unique features of redox flow batteries (RFBs), they and their principles will be contrasted with three other devices: electrochemical double layer supercapacitors (EDLSCs), solid-state batteries (SSBs), and fuel cells (FCs). These four archetypes capitalize on different physical and chemical principles, not even all of them employ redox reactions to store energy, and are therefore suitable for diverse EES applications. Examples for these applications are balancing and stabilizing the grid [1], electric vehicles [2,3] and mobile consumer electronics [4,5]. Schematics of these four devices are shown in Fig. 1. 1) EDLSCs ( Fig. 1(a)) work on the principle that a potential difference DU at an electrode-electrolyte interface, which establishes itself due to the different Fermi energies of (metal) electrode E F,Me and electrolyte E F,EL , is balanced by accumulating charges. Cations or anions in the electrolyte form an electrochemical double layer, conglomerating in order to keep the electrochemical potential e across the interface constant [6,7]. Charges in the electrode, counterbalanced by charges in the electrolyte form a capacitor with a differential capacitance C d equal to with accumulated charge Q, the distance between charges d, dielectric constant ε r , the permittivity of free space ε 0 and electrochemically active surface area A.
High specific capacitances of 10 Â 10 -6 F/cm 2 are reported for carbon electrodes in aqueous electrolytes [8], because the first layer of charges is very close to the electrode (some Ångströms). Carbon can have a high gravimetric surface area (~1000 m/g) [9][10][11]. The energy E SC of an EDLSC, which is stored entirely in the electrochemical double layer, is given by The maximum voltage ΔU is limited by the oxidative and reductive stability of the system comprising electrodes and electrolyte. Organic electrolytes and room temperature ionic liquids can sustain higher voltages than water (approx. 1.23 V) [12,13]. In commercial EDLSCs propylene carbonate or acetonitrile are used as solvent with quaternary ammonium salts as supporting electrolyte, and their maximum voltage is typically 2.8 V. Storage of energy in the double layer as only principle limits the energy density of EDLSCs to some watt-hours per kilogram, however, the build-up of the electrochemical double layers is a very fast process. This enables a high power density for EDLSCs, on the order of 10 kW/kg [12].
As shown in Fig. 1(b), SSBs also comprise two electrodes and an electrolyte (plus separator), but their energy is not stored in the ECDL, but in the electrodes. The energy content in SSBs is given by the product of the charge that can be transferred, the capacity Q, and the potential difference at which electrons flow through the external circuit E Bat ¼ QΔU : Atom density in condensed matter is roughly 10 23 atoms [14], which entails a high capacity Q for SSB because the active material is present at a very high concentration. This is a significant difference to RFBs, as will be discussed later. This can be illustrated for the example of a lithium metal electrode [15,16]. At a mass of roughly 7 u, each lithium atom can be oxidised to give a monovalent cation Li AE Li þ þ e -U 0 ¼ -3:04 V vs: SHE: The standard potential of this reaction is at -3.04 V vs. the standard hydrogen electrode (SHE). Hence, a lithium anode could store a theoretical specific capacity of q theo,g Li ¼ 3660 mAh/g or q theo,v Li ¼ 2061 mAh/cm 3 . Due to the problems associated with lithium metal anodes [17], graphite is commonly employed as the host material for the anode in a lithium ion battery (LIB). Because carbon is heavier than lithium, and six carbon atoms are required to store one lithium cation, the theoretical capacity of carbon is lower at q theo,g C ¼ 372 mAh/g.
Intercalation reactions are not conversion or alloying reactions, which limit the stress the electrodes experience during charge or discharge. As LIBs rely entirely on the intercalation of lithium ions into interstitial sites of host lattices, high cycle lives are possible.
The combination of high capacity, high voltage and high stability make LIB the premier choice for electric vehicles and consumer electronics. One of the major current challenges is safety, as the employed organic electrolyte has a large combustion enthalpy and can therefore explode violently.
FCs oxidize some fuel (hydrogen, methanol, ethanol, carbon, etc.) at the anode while reducing an oxidant (oxygen) at the cathode, see Fig. 1(c) [18]. The reaction product (H 2 O, CO 2 , etc.) is exhausted. The energy is stored in the chemical bonds of the fuel, which is pumped from external tanks into the power converter and the electrodes contain catalysts which facilitate the oxidation of fuel or the reduction of oxygen. Advantages of FCs are the high energy density of the fuel (ethanol stores 6.4 kWh/L) and their versatility in accepting various types of fuel. Also, FCs converting hydrogen are local-zero emission, a big advantage for FC electric vehicles when looking at the detrimental impact of internal combustion engine fumes on human health [3]. One of the main challenges is the sluggish redox reactions at anode and cathode. In acidic media on platinum, which is one of the best catalysts for the hydrogen related reaction, the hydrogen oxidation reaction (HOR) has an exchange current density of 1 mA/cm 2 [19]. Therefore, large over-potentials are required to obtain a substantial current from anode. For the oxygen reduction reaction (ORR), typically no exchange current density but a kinetic current is given [20,21]. It is agreed however, that the activity for the ORR is orders of magnitude lower than that of the HER on platinum. This is thought to be mostly caused by the high stability of reaction intermediates such as adsorbed oxygen and adsorbed hydroxyl [22]. To obtain a current density of 200 mA/cm 2 (at 1 bar O 2 and H 2 pressure) in a low temperature FC, the resting potential of 1.23 V is reduced to 0.78 V [7]. This kinetic limitation renders FCs low power devices.
RFBs share many characteristics with FCs, as can be seen in Figs. 1(c) and 1(d). Like FCs, RFBs store the active material in tanks which is then pumped to the anode or cathode for charge transfer. One of the differences is that the reaction products are not discarded in a RFB, but kept in their respective streams and pumped back into the anolyte and catholyte tanks. As a RFB is a secondary battery, the discharged species can be re-charged. For that the polarity of the RFB is reversed, the anode is now reducing the anolyte redox species, and the cathode is oxidizing the catholyte redox species (against electrochemical conventions). While a FC can use liquid (ethanol, methanol, liquid nitrogen) and gaseous fuels (compressed hydrogen), a RFB will usually operate with dissolved redox molecules, and it stores its energy entirely in solution. For the energy density of the RFB, Eq. (3) holds. The charge of the RFB is determined by how many redox molecules are in solution (governed by concentration c), and number of transferred electrons per molecule n: Because typical concentrations are on the order of 1 mole of redox molecules per liter (6 Â 10 23 L -1 % 6 Â 10 20 cm -3 ) and only a single electron is transferred for most RFB chemistries (n = 1), the charge contained in the liquid electrolyte is approximately two orders of magnitude lower than for SSBs (10 23 cm -3 ). However, storing energy in a liquid has many conceivable advantages, some of which are shared with FCs: The electrolyte can be easily exchanged by draining and refilling. This could be employed to replace faulty or aged electrolyte or to upgrade the battery with a more advanced electrolyte, or to instantaneously recharge a battery. In addition, it enables the scalability of RFBs: The size of the tanks determines the charge Q that is stored in the battery and therefore the battery's energy content. The power is determined by the size of the membrane electrode assembly (MEA). This means that RFBs can be easily adapted to specific needs, either by changing the size of the tanks, or by altering the size of electrodes and membranes (or the number of cells).
The power density of a RFB, which is normalized to the geometric surface area of the MEA, is limited by four resistances that introduce a potential drop in the cell: The total resistance R total is the sum of the resistances caused by the electronic resistance R elec between the various components (e.g. between current collector and electrode), the membrane resistance R mem , the charge transfer resistances on anode and cathode R CT , and the mass transport resistance R mass . Because R elec , R mem , and R mass are related to the engineering of the MEA [23,24], only R CT will be briefly covered here.
with the gas constant R, temperature T, Faraday constant F and electron transfer constant k 0 . Because k 0 can spread over several orders of magnitude, from facile couples such as ferrocene/ferrocenium with k 0 ≈ 1 cm/s [25] to sluggish ones such as the VO 2+ /VO 2 + redox reaction with k 0 &10 -6 cm/s [26,27], this fundamental parameter is of high importance for the power of a RFB. Maximizing it could lead to high-power RFBs, but current technology, such as the all-vanadium RFB (VRFB) provides only about 70 mW/cm 2 . With these principles established, the sequence in energy density and power density shown in the Ragone plot ( Fig. 2) can be understood: SCs store energy only in the electrochemical double layer, by placing charges in a potential field, which gives small energy densities. RFBs store energy in redox couples in liquid form, which entails a lower density of charge carriers than in SSBs. Lastly, FCs store energy in the chemical bonds of either the lightest element (H 2 ), or hydrocarbons (ethanol, methanol, gasoline) which makes them high-energy density devices. However, because the reactions in a FC are sluggish (ORR, HOR), they are low power devices. The intercalation reactions that take place in LIBs are faster than the redox reactions occurring in RFBs, making the latter slower than the former. SCs exhibit the highest power density, because the formation of the double layer is a purely physical process and no electron transfer is taking place.

Criteria for technology
A typical RFB cell consists of two electrode compartments divided by a separator. The separator is commonly an ion exchange membrane which prevents the crossover of active species between half-cells but allows the movement of ions between electrode compartments for the balancing of charge. The positive and negative electrolytes are stored in external tanks and are pumped into each half-cell. The electrodes of a RFB are normally inert, serving as the site for the redox reactions of the active species only which remain soluble in the electrolyte. Upon exiting the electrode compartments, the electrolytes are returned to the storage tanks to be recirculated through the cell. Individual RFB cells can be connected in series to produce cell stacks through the use of conductive bipolar plates which connect one cell to another. The array of cell stacks, stored electrolyte and the balance of plant constitute a complete RFB system. The balance of plant includes all other components necessary for operation of a RFB: pumps, plastic plumbing and tanks, a power conditioning system and systems for battery monitoring and control [28]. Figure 3 illustrates the creation of a RFB system through the building up of individual cells into modular stacks.
RFBs can be subdivided into 'true' RFBs and hybrid RFBs (see Section 3 'Various concepts' for further detail). In short, true RFBs utilize inert electrodes and redox species that remain in solution. Examples of classical true RFBs are the iron-chromium 1) , bromine-polysulphide [30], and the VRFB [31]. A schematic of the VRFB is shown in Fig. 4. The operation of a hybrid RFB involves a phase change during the cell reaction. An example is the zincbromine system in which the plating and dissolution of zinc at the anode occurs upon charge and discharge respectively [32].
The efficiency of a RFB can be characterized by several parameters. First, the Coulombic efficiency h CE which is the ratio of the current transferred upon discharge to the current transferred upon charge, with Q out the amount of charge during discharge and Q in Activation over-potentials deriving from the reaction kinetics, mass transport over-potentials and Ohmic losses are all sources of a discrepancy between the charging and discharging voltage (see Eq. (7)). The Coulombic efficiency h CE multiplied by the voltage efficiency h VE yields the energy efficiency h EE of a RFB which indicates how much of the energy that is supplied to the battery during charging can be extracted upon discharge.
RFBs have several benefits over conventional secondary batteries. For instance, external storage of the electrolyte prevents the self-discharge of the stored solutions. Further, as the electrodes are not subjected to continuous plating or intercalation reactions, electrode deterioration is minimised. When a conventional battery undergoes repeated charge/discharge cycles, the electrode materials expand and contract which results in their degradation over time [33]. Thus, the lifetime can be much improved when there is no phase change at the electrodes during cycling, as in a RFB. The predicted lifetime for a RFB tends to exceed 10 years and the VRFB is rated at 10000 cycles [34].
The life-limiting component of a RFB is typically the cell stack. Wearing of the membranes separating the electrode compartments is a source of performance degradation and for the VRFB in particular, the cell stack is expected to have a lifetime of 10 to 15 years for a RFB undergoing 1000 charge/discharge cycles per year [35]. The electrolyte storage tanks, plumbing, structural components, power electronics and controls of a RFB should have longer useful lifetimes [34], but pump replacement may also be required periodically. The crossover of active species between the half-cells due to an inefficiency of the separator is a source of capacity fade and as such, a degree of electrolyte maintenance may be necessary. However, with stack and pump replacement, VRFBs can operate for more than 20 years [35]. Hazards relating to the operation of RFBs include the use of flammable, toxic and/or corrosive electrolytes. While the flowing electrolytes aid heat dissipation and RFBs such as the VRFB classically used aqueous solutions, flammability should be a consideration if a RFB electrolyte is based on a non-aqueous solvent. Further, acidic supporting electrolytes are typically used which necessitates the use of acid-resistant components and toxicity associated with active species is of concern in the event of leakages.
It is often noted that cost is currently prohibitive for the widespread deployment of RFBs for energy storage. The cost of a RFB system includes numerous components: the cell stacks including separators, electrolyte solutions, storage tanks, power-electronics equipment, the control system, pumps, valves and plumbing as well as the cost of assembly and installation costs [33]. A target capital cost of $100 per kWh has been suggested for large-scale grid storage of energy from renewable sources [36], while a recent cost analysis that considered the VRFB set a base case capital cost of $380 per kWh for a 1 MW/12 MWh system [37]. Both the electrolyte and the cell stack and membrane are implicated as substantial contributors to the capital cost of VRFBs [37,38]. The cost of the electrolyte is dependent on the price of vanadium per kg, while the acidic and oxidising environment of the VRFB requires the use of hardwearing yet costly Nafion ion exchange membranes. Thus, at the electrolyte level, possible improvements to the capital cost of RFBs relative to the VRFB could come from the use of lower-cost active materials and RFB chemistries that allow less expensive porous separators to be used.
The different RFB chemistries have been summarized in several recent review papers [39][40][41][42][43][44][45] and there is a myriad of published reactions for the positive and negative electrode (see Table 1). Aqueous metal-based chemistries are complemented by both reports of non-aqueous RFBs and the emergence of active species based on organic molecules, as alternatives to the traditional and widely tested RFBs are continually devised. The themes of nonaqueous electrolytes versus aqueous electrolytes and nonmetallic RFBs versus metal-ion chemistries are discussed herein.
3 Various concepts, all-liquids, gas/liquid, semi-solid, slurries, redox mediators Since their inception in the 1970s at NASA, various combinations of electrolyte/electrode phases have been investigated for RFBs [39,41,42]. Classic examples, like the Fe-Cr system [46] or the VRFB [31,47,48] use two liquid electrolytes, anolyte and catholyte, to store energy. This all-liquid form exhibits all the advantages and disadvantages related to liquid electrolytes stated in the introduction: Lower concentration of charge carriers than in the solid state but a capacity that is only limited by the size of the tank (and the cost of the electrolyte).
The gaseous phase is lighter than liquid electrolytes, and oxygen, if it can be taken from the atmosphere, is almost free of costs. This motivated the design of a V-O 2 RFB or Vanadium/air RFB as presented by Hosseiny et al. [49]. This embodiment of a gas/liquid RFB hybrid employs the V 2+ /V 3+ redox couple as anolyte, and the oxygen evolution reaction (OER) and the ORR as cathode. The battery published suffered from low h EE (27% at 60°C), and the need for catalyst loading on the cathode. As this battery combines the more problematic half-cell of the VRFB [50,51] with the sluggish half-cell of a hydrogen-oxygen FC, the underwhelming performance can be understood.
Solid-liquid RFBs comprise one liquid half-cell and one metal electrode which typically undergoes a plating reaction [57]. One example is the zinc-bromine battery, that features a Zn or carbon plastic anode on which Zn is deposited during charge [58]. Metal electrodes allow for an increase in energy density as compared to regular flow batteries, however, they also introduce the problems of regular batteries into the RFB: Dendrite formation upon plating reactions that can lead to internal short-circuits and high-self-discharge rates make it questionable whether liquid-solid hybrids are useful.
Slurries of lithium-metal-oxides are well-studied system due to their importance for LIBs. Under the moniker "semisolid flow batteries," they were introduced for RFBs utilizing 20 vol% LiCoO 2 and 10 vol% Li 4 Ti 5 O 12 [59]. This concept allows for very high concentrations of redox species (estimated ranges from 20 to 80 mol/L) [60]. Conductive slurries are used as electrolyte, in which carbon forms a conductive network. The limited conductivity of these slurries (0.1 mS/cm) is one of the contemporary challenges of semisolid flow batteries [59,60].
A concept does not rely on dissolved redox species, but instead on a solid active material that is stored in tanks: mediated flow batteries. In these, the charge is transferred to the active material via redox-mediators such as ferrocene [61] or cobaltocene [62]. This type of flow battery utilizes similar active materials as semisolid flow batteries, but avoids the high viscosity and poor conductivity of slurries. To avoid losses, the potentials of the redox mediators must closely match the (de-)intercalation potentials of the active material.
Another concept is the addition of a solid energy booster to the tank of a RFB [63]. This approach proposes that the redox transitions of the polymer polyaniline are accessed by a redox mediator, storing charge in the emeraldinepernigraniline and leucoemeraldine-emeraldine redox transitions of the polymer. As redox-mediators, Fe 2+ /Fe 3+ and V 3+ /VO 2+ were employed at a concentration of 1 M. When polyaniline was combined with conductive carbon and abovementioned electrolytes the volumetric energy density was increased by a factor of three, compared to the base case with only iron-and vanadium-ions [63]. Current challenges are a "passivation" of polyaniline in certain redox state and sluggish kinetics of the V 3+ /VO 2+ couple.
The advantages and drawbacks of various RFB concepts are listed in Table 1.

Solvent: aqueousnon-aqueous
In most of the invented RFB-systems, acidic, aqueous electrolytes are used. One disadvantage of electrochemical application in aqueous solutions is the electrolysis of water. The corresponding half-reactions can be described for acidic and basic solutions: Cathodic reaction: Anodic reaction: The potential at which those reactions occur depends on the electrode material used. The thermodynamic potential window in which no reactions take place is 1.23 V. Out of this range, water electrolysis is promoted which leads to parasitic side reactions and therefore to efficiency loss and capacity imbalance in a full RFB. This limits the selection of electrochemical active species for an aqueous RFB. Furthermore, the application of aqueous electrolytes is restricted by the temperature range of 0°C to 100°C. Mostly, sulphuric acid is added as supporting electrolyte to enhance the conductivity of the electrolyte without the addition of additional salts as supporting electrolyte.
Often, to ensure higher conductivity, hydrochloric acid is added to the aqueous electrolyte [64]. However, the dissociated chloride ions can promote corrosion processes, thus having a negative effect on the lifetime of the battery attachments. Chloride ions are not oxidizing agents themselves, but, they can influence the corrosive conditions and are able to catalyze the process depending on the used materials, which results in higher maintenance costs [65].
The use of non-aqueous electrolytes is a possibility to enlarge the potential window and the temperature range. An example for an organic solvent used in a RFB is acetonitrile [66][67][68][69]. To obtain a reasonable conductivity, a supporting electrolyte like tetrabutylammonium hexafluorophosphate (TBA PF 6 ) or sodium perchlorate (NaHClO 4 ) has to be added to the electrolyte, which can limit the potential window because of decomposition reactions [70]. Furthermore, these supporting electrolytes increase capital costs. The potential windows of acetonitrile with 0.1 M additive on a platinum electrode are from +3.0 V to -2.6 V vs. saturated calomel electrode (SCE) in the case of TBA PF 6 and from +1.8 V to -1.6 V vs. SCE in the case of sodium perchlorate NaHClO 4 [71]. This suggests that the potential window can be enlarged in non-aqueous media but is limited by the decomposition of the additive. The useable temperature range of acetonitrile (anhydrous) is between -43.8°C and 81.7°C which is beneficial for low temperature applications. In general, solvents with nitrile groups have a good stability toward oxidizing and reducing conditions in electrochemical cells [71].
Another example for non-aqueous electrolytes are organic carbonates like propylene carbonate (PC), ethylene carbonate (EC) and methyl ethyl carbonate (EMC), which are well studied due to their application in LIBs [70,72]. Those solvents have the capability to dissolve salts such as tetrabutylammonium trifluoromethanesulfonate (TBAOTf) and lithium hexafluorophosphate (LiPF 6 ) to function as supporting charge carrier in the electrolyte.
The use of organic solvents could potentially require complex maintenance operations, especially if the organic solvent is hygroscopic and prone to water intake from the atmosphere which would lead to a decreased potential window. Acetonitrile, for example, is strongly hygroscopic, and will readily absorb water from air if not stored under inert conditions. For technical application, this would mean that the battery would have to run under nitrogen or argon gassing which would result in higher Self-discharge; Oxygen gas permeation through membrane need for catalyst loading on electrode; Pt leaching into the cell [49,52] maintenance costs compared to the aqueous system. Also, non-aqueous solvents like acetonitrile corrode rubber and polymer materials which could damage the tubes the electrolyte are pumped through and the innards of the cell which would reduce the lifetime of the battery [65]. While intensifying maintenance requirements, electrolytes such as PC, EC, EMC, and others also carry a significant safety risk. This is aggravated in LIBs because they contain, besides the flammable electrolyte, also oxidizing material in the form of the lithium-metal oxides that typically make up the cathode [5]. Therefore, two of the three elements of the combustion triangle (fuel and oxidant) are present in LIBs, and the addition of a heat source can lead to catastrophic failure. In a large scale RFB application of organic electrolytes, the fuel would be present in a much higher quantity than in a LIB, and surrounding oxygen in the atmosphere could function as oxidizer. Organic solvents are, in most cases, ecologically questionable, and their application should be reduced as much as possible because, in spite of safety measures, solvents are released into the environment and have a polluting impact. For the synthesis of PCs, dichloromethane is used for which carcinogenic effect is suspected and it harms aquatic organisms (H-351, H-373) [73]. Acetonitrile as a solvent should be considered critically as well. Acetonitrile is absorbed by the digestive system, the skin, and the lungs and releases cyanide in the human body. Hence, poisoning symptoms of cyanide exposure occur [74]. Due to these properties, the toxicity of the used solvent must be taken into account when developing a battery [75].
The selection of a suitable solvent for conducting experiments requires careful consideration, as the solvent can have a significant influence on the electrochemical properties of the redox couples. The standard potential of a reaction like depends on the solvent or on the solvation energy of the metal ion. For instance, the standard potential is shifted to more negative values in solvents in which the metal ion is solvated more strongly [65]. In some cases, solvents have the capability to change the whole reaction mechanism, which affects the electrochemistry of a redox system. For example, a one-step reaction in water could be a two-step reaction in an organic solvent. Copper in the oxidation state I þ is unstable in water but is stabilized in acetonitrile, which means the reduction of Cu(II) to Cu(0) in water occurs at another potential than in acetonitrile. Besides, the permittivity and Lewis acidity of a solvent affects the standard potential of metal ions significantly. The electron transfer constant k 0 , an important factor in electrochemistry, is affected by the solvent, too, due to its influence on the reorganization energy. Another type of electrolyte that receives more and more attention to enhance the energy density of RFBs are room temperature ionic liquids (RTILs) [76]. They are solvents that consist entirely of ions, are non-volatile, nonflammable, chemically stable, and highly conductive [77]. Furthermore, if free of water, they make a huge potential window and temperature range accessible [78,79]. RTILs offer environmentally friendly properties and can be used as green solvents in electrochemical application, especially RFBs. There are two main types of RTIL: one is a mixture of AlCl 3 and quaternary ammonium chloride (R + Cl -) and the other type are salts of cations like 1-butyl-3-methylimidazolium (BMI + ), 1-ethyl-3-methylimidazolium, and 1-butyl-pyridinium and anions like BF 4 The potential window depends on the acidity and basicity of the solution, respectively. In the case of 1-butyl-3-methylimidazolium-PF 6 -, the potential window is 7.1 V [65]. Some properties can be adjusted by the anion and in which way the ligand coordinates to the metal center. The group of Anderson et al. investigated aforementioned properties in copper ionic liquids [80]. Cu{NH 2 CH 2 CH 2 OH} 6 [BF 4 ] 2 showed the best results regarding viscosity, conductivity and electrochemical reversibility, and was liquid at 25°C, whereas the copper complex with one anion exchanged with triflate was solid at RT. In addition, the conductivity changes with the anion. With one 2-ethylhexanoate anion instead of tetrafluoroborate, the conductivity decreases from 6.8 to 0.586 mS/cm [81].
Membranes in RFBs are used to separate the liquid electrolyte compartments and are permeable ions to enable charge balancing [82,83]. The most important properties of those membranes are high ionic conductivity and low permeability toward the electrochemical active species to inhibit crossover of the catholyte into the anolyte and vice versa. In aqueous systems, a cation exchange membrane like Nafion is usually used [84]. However, anion-exchange membranes [85][86][87] or porous membranes [88] have also been used. In non-aqueous media polyethylene-based porous separators [72], anion-exchange membranes [67] and cellulose-based dialysis membranes [89] are reported.
The advantages and challenges of various types of electrolytes used in RFBs are tabulated in Table 2.

Redox center: metallicnon-metallic
There are two types of metal-based batteries: It can be distinguished between Me/Me z+ and Me z+ /Me (z+ 1)+ systems (with integer z). An example for Me/Me z+ system is the Zn/Br 2 battery in which metallic Zn is deposited on the cathode during charging and Brions are oxidised to elementary bromine simultaneously [32,58]. In this system the metal is an electrochemically active species and forms part of the electrode as well [90]. A problem that often occurs with this kind of battery is the formation of dendrites, which can lead to a short circuit when these protrusions pierce the separator [91,92].
An example for a Me z+ /Me (z+1)+ system is the VRFB [93]. In this battery, the redox couples V 3+ /V 2+ and VO 2+ / VO 2 + are used. During charging, V 3+ is reduced to V 2+ and VO 2+ is oxidised to VO 2 + simultaneously. During the entire process, the ions remain dissolved and are not part of the electrode. The disadvantage of this type of system is the crossover of the ions across the membrane into the other half-cell which decreases the efficiency. In the case of different elements, the capacity is significantly reduced by the crossover and the associated irreversible contamination. Another point which limits the application is the limited solubility of the metal salts used. In the case of the VRFB, the maximum concentration of the VRFB electrolyte is 1.7 to 2 M vanadium if only sulphuric acid is used as supporting electrolyte [94]. It should also be mentioned that at higher concentrations, the V 3+ and V 2+ ions precipitate in the negative electrolyte if the temperature is lower than 10°C. While the solubility of V 2+ , V 3+ and VO 2+ increases with increasing temperature, thermal precipitation of the VO 2 + species into V 2 O 5 occurs when the positive electrolyte is over 40°C [95]. The electrolyte of non-aqueous system contains a concentration of electrochemically active species from 0.01 M [67] to 2 M [72].
The locus of the redox reaction is the electrode surface and this is why the selection of the right electrode plays an important role. The electrode requirements for all systems are, in general, high conductivity, good resistance toward the chemical environment, mechanical stability toward compression, and inexpensive costs. In Me z+ /Me (z+1)+ systems, commonly carbon-felt materials are used because they do not undergo any redox reaction within the stability window of water [26]. The surface of the electrode should be active toward the desired redox reactions and simultaneously inhibit unwanted side reaction such as the HER.

Review of battery chemistries
The development of novel chemistries for RFBs is a research area that attracts a tremendous amount of attention at the moment. Therefore, it is impossible to present a holistic compendium of all the (half-cell) chemistries investigated. The present paper should be seen as a complimentary to recently published reviews on the topic of chemistries for RFBs [39][40][41][42][43][44][45]. In the selection, chemistries that target the capital costs of RFBs are focused on, which is currently the main inhibitor for their wide-spread application. Avenues to achieve these costsavings are high capacity chemistries that allow for a reduction in footprint of the battery, high power chemistries that allow for a reduction in size of the cells and, therefore, reduce the footprint and costly materials such as membranes and low-cost molecules. 6.1 High energy density all-vanadium RFBs 6.1.1 VRFBs with increased concentration As described, for VRFBs, V 2+ and V 3+ species in the negative half-cell precipitate below 10°C while VO 2 + precipitates as V 2 O 5 above 40°C in the positive half-cell which allows a maximum vanadium concentration of 2 M in the sulphuric acid electrolyte and limits the energy density to 25 Wh/kg [95]. Reformulation of the electrolyte is a possible way to access higher vanadium concentrations and increase energy density. Indeed, Li et al. demonstrated use of a mixed sulfate and chloride electrolyte that allowed a 2.5 M concentration of vanadium to be achieved [96]. This represented about a 70% increase in energy density relative to current sulfate-only systems. It was found that vanadium in all four oxidation states was stable in a solution of 2.5 M SO 4 2and 6 M Clfrom -5°C to 40°C,  [96]. The formation of Cl 2 gas during cycling is a concern, but it has been reported that no significant gas evolution is observed. Our group was able to show that k 0 of the VO 2+ /VO 2 + reaction in 1 M H 3 PO 4 is up to 67 higher than in 1 M H 2 SO 4 , which is attributed to a different chemical coordination in the electrolytes [97].
Electrolyte additives are also a possible measure that can be implemented to increase vanadium concentration and achieve higher energy densities for the VRFB. For instance, Roe et al. recently investigated various stabilizing additives for the purpose of preventing the thermal precipitation of VO 2 + species [98] (Fig. 5). Several inorganic additives were studied as it was noted that finding effective organic additives that had long-term stability was a challenge due to the high oxidising power of VO 2 + . Sodium pentapolyphosphate, K 3 PO 4 , H 3 PO 4 and (NH 4 ) 2 SO 4 were thus screened as stabilizers for 3 M supersaturated VO 2 + solutions. H 3 PO 4 (1 wt%) was found to be most effective at maintaining VO 2 + concentration at 30°C while a 1 wt% H 3 PO 4 + 2 wt% ammonium sulfate formulation performed best at 50°C and was investigated further in cell cycling tests. Such stabilizing behavior could be attributed to the formation of V(V)-phosphate complexes and the increase in H + concentration due to phosphoric acid addition causing thermal precipitation of V(V) as V 2 O 5 to become disfavored. A 3 M VO 2 + solution with 5 M total sulfate concentration and containing 1 wt% H 3 PO 4 + 2 wt% ammonium sulfate demonstrated stable efficiencies over 90 charge/discharge cycles with a slight decrease in cell capacity observed [98]. No precipitation was evident during the experiments. A viable 3 M vanadium electrolyte could allow a 60% -90% increase in energy density relative to the practical concentrations of 1.6 M -1.8 M that are currently used in VRFBs.
A number of organic additives were proposed [99] and researchers found that amino acids [100], coulter dispersant [101], polyacrylic acid and its mixture with CH 3 SO 3 H [102] as well as fructose, mannitol, glucose, and D-sorbitol [103] and other [104] organic compounds could improve the temperature stability of the catholyte in the VRFB. However, Nguyen et al. studied the effect of additives glucose and ascorbic acid, and found that they only led to an apparently enhanced stability [104]. They claimed that the catholyte with organic molecules was more stable because these molecules were oxidised by the VO 2+ ions, thereby reducing the state of charge (SOC) of the VRFB. As precipitation occurred at high SOCs, when the catholyte was mostly present at VO 2 + and these additives effectively diminished the concentration of VO 2 + , the time until precipitation occurred was prolonged. 6 (Fig. 6), degradation mechanisms [106] and the influence of different solvents and supporting electrolytes [107,108].
A1 and C1 : VðacacÞ 3 þ e -AE ½VðacacÞ 3 -, A2 and C2 : VðacacÞ 3 AE ½VðacacÞ 3þ e -: RFBs that discharge to the same active species in both half-cells mitigate the issue of cross-contamination of halfcells resulting from active species crossover. In a similar way, a non-aqueous chromium acetylacetonate RFB was also investigated [109]. A cell potential of 3.4 V was indicated for a one-electron disproportionation reaction of Cr(acac) 3 in which Cr III (acac) 3 was oxidised to [Cr IV (acac) 3 ] + and reduced to [Cr II (acac) 3 ] -. The charge-discharge performance in an H-type cell of 0.05 M Cr(acac) 3 in acetonitrile with a 0.5 M TEABF 4 supporting electrolyte was evaluated. The cycling between 0% to 50% of the theoretical SOC yielded h CE of 53% -58% and h EE of 21% -22%. A non-aqueous RFB utilizing manganese acetylacetonate was also later investigated [67]. A 1.1 V cell potential was indicated for a RFB based on Mn III (acac) 3 /[Mn II (acac) 3 ]and Mn III (acac) 3 /[Mn IV (acac) 3 ] + redox couples. A cell containing 0.05 M Mn(acac) 3 in 0.5 M TEABF 4 /acetonitrile supporting electrolyte showed a discharge plateau at around 0.3 V over 10 cycles. An increase in h CE from~74% to~97% was attributed to an unknown side reaction, while the h EE remained at around 21% during cycling.

VRFBs with novel reactor design
A co-laminar flow cell (CLFC) is a microfluidic device with a 'membrane-less' design: achieving laminar flow in adjacent liquid streams allows two solutions to flow side by side with mixing occurring only by diffusion of active species between streams [110,111]. The prevention of turbulent mixing thus allows the two electrolytes to flow side by side in a CLFC, with ionic conduction still permitted but no requirement for a membrane, thereby omitting a significant cost contribution [37] and source of overvoltage through membrane resistance. A high power density RFB that utilizes vanadium chemistry was presented in the form of a CLFC by Goulet et al. [111]. The high power density was mainly attributed to a novel in operando deposition method of carbon nanotubes (CNTs) that was applied to the flow-through porous carbon paper electrodes of the CLFC. This involved suspending the CNTs directly in the electrolyte that was to be flowed through the cell. The deposition of CNTs enhanced the electrochemical active surface area by adhering to the carbon paper electrodes and forming a conducting nanoporous layer. Further, an improvement in mass transport was indicated which was attributed to both the deposition of material directly in the reactant flow path and a reduced average pore size of the electrodes. The power density of the flow cell subjected to CNT deposition was 2 W/cm 2 (a typical power density for a RFB cell is 0.1 W/cm 2 [33,111]). CLFCs are classed as microfluidic electrochemical cells and as such, it is noted that the cell is limited in scale. However, the addition of CNTs directly to the redox electrolyte presents as a simple and inexpensive method that lead to an enhancement of both the surface area and mass transport properties of the electrodes. A realtime image of the CLFC discharging with color changes indicating the different vanadium ion oxidation states and a schematic of the construction of the cell are shown in Fig. 7. A tubular vanadium redox flow cell was presented by Ressel et al. [112]. This approach was chosen to enable reduced manufacturing costs and less shunt currents in a cell stack. The tubular cell was constructed from extruded current collectors and a welded tubular membrane, it was able to produce 70 mA/cm 2 at an h EE of 55%.

Vanadium chloride/polyhalide RFB
To avoid the precipitation of V 2 O 5 at high temperatures, a RFB which used a polyhalide solution in the catholyte and a vanadium(II)/vanadium(III) chloride redox couple in the anolyte was developed [113]. The occurring reactions are During the charging process, the bromide ions in the positive half-cell are oxidised to the polyhalide ion Br 2 Cl -; the reaction between Br 2 and Clleads to the Br 2 Clion, while Cl 2 dissolved in a Brsolution generates the Cl 2 Brion that has high oxidation potential. The application of those compounds in a vanadium chloride/polyhalide redox flow cell would lead to a cell potential of around 1.3 V. A CV was conducted on a graphite electrode and the authors claimed that the redox reaction of VCl 3 /VCl 2 was reversible [113]. A correction for the high surface area and porosity of the electrode, which could elucidate k 0 [26], was not performed. A higher chloride ion concentration apparently shifts the peak potentials. In 8.48 M Clsupporting electrolyte, the anodic peak appears at a potential of -0.27 V vs SHE (recalculated from SCE) and the corresponding cathodic peak at -0.33 V vs. SHE. Another advantage is that no hydrogen evolution is observed at potentials below the V(III) reduction peak which is favorable for the charging process in RFBs . A short-term cell test was conducted and the composition of the negative half-cell electrolyte was 1 M VCl 3 in 1.5 M HCl, while that of the positive half-cell electrolyte was 1 M NaBr in 1.5 M HCl. The h CE and h VE values were calculated as 83% and 80%, respectively. The long-term experiments to investigate the crossover across the membrane of electrolyte and the stability tests of the bromine-polyhalide mixture have to be conducted.

Bromine-polysulphide RFB
The bromine-polysulphide RFB was patented in 1984 [30]. Upon charge, bromide ions are oxidised to bromine at the positive electrode, which is complexed as the tribromide ion Br 3 in solution, and the S 4 2anion is reduced to S 2 2at the negative electrode. Upon discharge the reverse occurs, with S 2 2anions oxidised to the polysulphide anion S 4 2at the anode and bromine reduced to bromide ions at the cathode. The catholyte and anolyte originally reported were an aqueous 1 M NaBr solution saturated with bromine and a 2 M Na 2 S solution, the optimal pH for the electrolytes being neutral or slightly basic pH, while a cation exchange membrane separated the half-cells to allow the transfer of Na + ions between the electrodes to complete the circuit [30]. The open circuit voltage of the system is around 1.5 V [42].
Anolyte:  kW power output and a 1 MW pilot scale facility that was built at Aberthaw Power Station near Cardiff, UK [114]. Interest in the system culminated in the construction of a 15 MW/120 MWh demonstration plant at Little Barford power station, also in the UK. However, the plant was never fully commissioned and the funding for the technology was later withdrawn [115]. Problems encountered with the bromine-polysulphide system include crosscontamination of electrolyte solutions due to membrane inefficiency, possible deposition of sulfur species in the membrane, as well as concern over the formation of H 2 S and Br 2 gases [42].

Zinc/polyiodide hybrid RFB
On the subject of higher energy density RFBs, Li et al. presented a zinc/polyiodide hybrid flow battery [116]. The aqueous electrolyte, a ZnI 2 solution in both compartments, was ambipolar; both the cationic and anionic ions were redox active which eliminated the need for counterions. Further, the Zn 2+ species also functioned as a charge carrier so a supporting electrolyte was not needed. Upon charge, zinc is deposited at the anode and polyiodide ions are formed in solution at the cathode while the reverse reactions occur upon discharge. Anolyte: Zn AE Zn 2þ þ 2e -U 0 ¼ -0:76 V vs: SHE: Catholyte: A high h CE of 99% was demonstrated in charge/ discharge cycling using concentrations of ZnI 2 from 0.5 M to 3.5 M. Over 40 cycles, a 3.5 M ZnI 2 electrolyte showed no obvious efficiency or capacity decay. While h EE was shown to decrease with higher ZnI 2 concentrations due to the increasing electrolyte resistance, a 5 M ZnI 2 solution reached an energy density of 167 Wh/L based upon the discharge energy density which approached the energy density of some LIBs. An electrolyte stability window of -20°C to 50°C was reported. However, as discussed in a previous section for the zinc-bromine RFB, an issue with the operation of the Zn/Zn 2+ redox couple was the formation of dendrites when zinc was deposited at the anode during charging which was observed in the zincpolyiodide flow battery.

Semi-solid lithium slurry RFBs
Optimisation of the VRFB remains a compelling topic, but alternative RFB chemistries also garner attention in the pursuit of improved energy densities. For instance, concepts of LIBs are naturally carrying over into flow battery research. As well as the increased cell voltages allowed by the use of a non-aqueous electrolyte, LIB operation involves ionic transport within solid electrodes which allows greater storage of active species and higher energy densities. As such, a semi-solid flow cell utilizing Li-ion chemistry and slurries of suspended electrode material was presented by Duduta et al. [59]. The electrolytes were semi-solid suspensions containing a nanoscale conducting carbon network that was formed by Ketjen black in alkyl carbonate/LiPF 6 solution, with micrometre-scale particles of electrode material (e.g. LiCoO 2 ) distributed throughout the network. The flowable suspensions that were demonstrated had up to 12 M active material concentration. A full-cell test using intermittent flow mode in which a single volume of the semi-solid suspensions was pumped into the cell, discharged and later displaced by a new volume gave a first demonstration of a fully operational semi-solid flow cell. Suspensions containing LiCoO 2 were utilized at the cathode (20 vol% (10.2 M), 1.5% Ketjen black) and suspensions containing Li 4 Ti 5 O 12 were active at the anode (10 vol% (2.3 M), 2% Ketjen black), both in 1 M LiPF 6 in dimethyl carbonate. A h CE of 73% and 80% was achieved in the first and second cycles. A calculated theoretical energy density for a semi-solid system utilizing LiCoO 2 and Li 4 Ti 5 O 12 with 40 vol% solid in each suspension at an average discharge voltage of 2.35 V was 397 Wh/L. More generally, it was estimated that optimised semi-solid flow cells utilizing lithium intercalation compounds could have energy densities of 300 to 500 Wh/L. However, higher viscosity electrolytes are associated with increased parasitic energy losses, owing to the additional energy that is required to pump the electrolyte.

Redox active polymers for RFBs
The use of organic molecules or polymers as the active materials in RFBs is also gaining prominence in RFB research as such species can be synthesized from inexpensive organic raw materials. An all-polymer RFB utilizing a polymer bearing the TEMPO moiety at the cathode and a viologen-based polymer at the anode was presented by Janoschka et al. [89]. The TEMPO unit (2,2,6,6-tetramethylpiperidine-1-oxyl) is a heterocyclic, stable radical nitroxide molecule that can be oxidised to an oxammonium cation, while viologens are dicationic 4,4'bipyridine derivatives that can be reduced to monocationic radicals. Both polymers also contain quaternary ammonium units to aid solubility. A study of the redox properties of the polymers by CV indicated the TEMPO-containing polymer underwent oxidation to TEMPO + at 0.7 V vs. Ag/ AgCl while the viologen polymer underwent a reduction at around -0.4 V vs. Ag/AgCl which corresponded to the formation of a monovalent radical Viol + species. The electrolyte for the all-polymer battery was an aqueous sodium chloride solution and a low-cost cellulose-based dialysis membrane was used to prevent crossover of the polymeric species but allow the movement of ions between half-cells. Figure 8 shows a schematic representation of the polymer-based RFB and the fundamental electrode reactions of the TEMPO and viologen radicals.
The open-circuit voltage in cell testing was 1.1 V and energy densities of around 10 Wh/L are reported for the polymer solutions. After 10000 cycles, 80% of the initial capacity was retained in a static, unpumped cell. A faster capacity fade was observed in a pumped cell, thought to be attributable to oxidation of the viologen radical species by oxygen entering the electrolyte. The cytotoxicity of the redox-active polymers was also tested and compared with the cytotoxicity of VCl 3 , VOSO 4 and two other cationic polymers, poly(L-lysine) and branched poly(ethylene imine), materials chosen because of their wide use. The TEMPO polymer showed less toxicity than the viologen polymer, while both polymers were less cytotoxic than poly(L-lysine), poly(ethylene imine) and the vanadium salts.
The use of size-exclusion membranes that prevent polymeric active species from crossing into the adjacent electrode compartment instead of more costly Nafion ion exchange membranes illustrates a promising aspect of RFBs that utilize polymers. However, as the concentration or the molecular weight of a dissolved polymer increases, so too does the viscosity of a polymer solution. While for any RFB, a higher concentration of active species leads to a greater energy density, increasing the concentration of a polymer solution may be problematic as viscous solutions increase the energy cost of pumping the electrolytes. Thus the advantage of the use of a size-exclusion membrane in a polymer RFB could be counterbalanced by the viscosity that is intrinsic to polymer solutions.
The use of TEMPO and viologen as small organic molecules in a RFB is an alternative to TEMPO and viologen-based polymers. For instance, a RFB that used aqueous solutions of TEMPTMA (N,N,N-2,2,6,6-heptamethylpiperidinyl oxy-4-ammonium chloride) (a TEMPO derivative) and methyl-viologen (N,N'-dimethyl-4,4bipyridinium dichloride) had an energy density of 38 Wh/L for a theoretical cell voltage of 1.4 V and excellent capacity retention over 100 cycles [117].
Winsberg et al. presented a polymer-based RFB that utilized two polymers bearing the boron-dipyrromethene (BODIPY) unit [118] in propylene carbonate as solvent. CVof the monomer containing the BODIPY group showed two redox reactions at -1.51 V vs. AgNO 3 /Ag and 0.69 V vs. AgNO 3 /Ag which corresponded to reduction of the monomer to BODIPYand oxidation to BODIPY + respectively. The BODIPY monomer was copolymerised with either TASt ((vinylbenzyl)trimethylammonium perchlorate) or TEGSt ((vinyl-benzyl)-triethylene glycol monomethyl ether) to generate two different polymers that could be used as the anolyte and catholyte in cell testing. Two static cell tests were performed (see Fig. 9). One employed poly(BODIPY-co-TEGSt) as anolyte with a further polymer that bore the TEMPO group used as the catholyte (poly(TEMPO-co-PEGMA). The second cell utilized poly(BODIPY-co-TASt) and poly(BODIPY-co-TEGSt) as the catholyte and anolyte respectively. A sizeexclusion membrane separated the half-cells and a solution of 0.5 M tetrabutylammonium perchlorate (Bu 4 NClO 4 ) in propylene carbonate was used as the supporting electrolyte. The TEMPO/BODIPY battery outperformed the all-BODIPY battery, with an average discharging voltage of 1.82 V and a h VE and h EE of 89% and 88% respectively. A h CE of 99% was achieved over 100 cycles, with 70% of the initial discharge capacity preserved. The all-BODIPY system had a charging plateau at 2.06 V while the mean discharge voltage was 1.28 V leading to a h VE and h EE of 62% and 55% respectively, with a h CE reaching 89%. The energy density of the electrolyte was 0.5 Wh/L for both polymer RFBs. 6.6 Polyoxometalate RFB Pratt et al. presented a flow battery containing vanadiumand tungsten-polyoxometalates (POMs), in which the POMs underwent multi-electron reactions [119,120]. POMs are suitable for electrochemical energy storage applications because they are often stable over a wide range of temperatures, exhibit multi-electron transfer with high kinetics [121]. Also, oxidation or reduction of POMs is usually accompanied by proton or cation transfer, a mechanism that avoids the generation of highly charged radicals [122] The h CE in aqueous solution was greater than 95% with a low capacity loss observed during more than 100 chargedischarge cycles and no decomposition of the molecule was reported [119]. The POM also dissolved in a nonaqueous electrolyte (0.5 M TBAOTf in propylene carbonate) and the non-aqueous system had a higher operating voltage (1.1 V, 0.3 V higher than the aqueous system) but a drop of h CE (initially 87%, after 10 cycles dropped by half) occurred. With a concentration of 20 mM POM and 0.5 M H 2 SO 4 , the observed current densities were one order of magnitude lower than in conventional RFBs. A reasonable approach to increase the current would be to enhance the POM concentration. The stability and costs were not reported. In the paper, the dimerization and eventual deposition of POMs containing W-ions at negative potentials was not discussed [119]. Keita and Nadjo reported that at negative potentials (approx. 1 V vs. SHE) in acidic solutions, POMs will modify the electrode surface [123]. The deposited material is usually a catalyst for the HER, thereby reducing the stability window of the electrolyte [124]. For [SiW 12 O 40 ] 4-, the precursor for the redox active material used in Ref. [119], electrode coverages of 33 to 120 monolayers are reported [123].
A potential candidate for a POM catholyte was reported: [Mn II 3 SiW 9 O 34 ] 7- [121,125]. While this molecule was able to transfer six electrons at high potentials, by oxidising three Mn(II) to Mn(IV), the POM adsorbed on the electrode and was too difficult to synthesize to make upscaling reasonable.
Another symmetric POM RFB was presented by Liu Reprinted with permission from Ref. [118]) et al. [126]. They employed H 6 [CoW 12 O 40 ] in both anolyte and catholyte of an aqueous battery. The reactions for the anolyte were two two-electron waves at -0.04 V vs. SHE and -0.16 V vs. SHE (in 1 M H 2 SO 4 , recalculated form the SCE) [127]. As catholyte, the single one-electron redox reaction of the Co(II)-heteroatom at 1.1 V vs. SHE was used. The POM was exceptionally soluble, up to 0.8 M, and 14 AhL -1 were reached as capacity. However, for that result four times more catholyte volume than anolyte volumes were used (and above volumetric capacity is given for the anolyte) which was necessitated by the imbalance in charge transfer reactions at low and high potentials.

Metal-free organic-inorganic aqueous RFBs based on anthraquinones
A RFB with quinones as aromatic redox-active organic molecules instead of redox-active metals was developed with the main objective to reduce capital costs [128]. The molecule used in the anolyte was 9,10-anthraquinone-2,7disulphonic acid (AQDS) dissolved in 1 M H 2 SO 4 (Fig. 10). According to the authors, AQDS can be synthesized from cheap, conventional educts.
In the positive half-cell, the redox couple was Br 2 /Brcomplementing the quinone/hydroquinone redox couple in the anolyte. For this type of battery, a tremendous power density of 1 W/cm 2 was reported at a relatively low cell voltage of 0.86 V [129,130]. The storage capacity retention reached was more than 99% per cycle (99.90% per cycle over 40 cycles). These losses are thought to originate from the leakage of the anolyte, crossover of bromide, and destruction of redox species [131].
The theoretical capacity of 35 Wh/L cannot be reached in the cell; only 23 Wh/L could be realized. A possible explanation was provided by Carney et al., who found intermolecular dimerization of AQDS at concentrations greater than 10 mM [132]. This behavior is pH dependent and might reduce the electrons that can be carried per AQDS molecule from 2 to 1.5 in acidic media.
A further advantage is that the chemical and electrochemical properties of the AQDS molecules can be engineered by adding functional groups. Addition of a hydroxyl group reduces the standard potential of the redox reaction, thereby increasing the open circuit potential U OCV . A CV of 1 mM solution of AQDS in 1 M sulphuric acid on a glassy carbon disk working electrode was conducted and a fast reversible two-electron two-proton redox reaction occurred. The peak separation was 34 mV and correspond to a two-electron process. The kinetic reaction rate constant k 0 was calculated to 7.25 Â 10 -3 cm/s which was greater than that found for other components used for RFBs such as V 3+ /V 2+ (compare Table 1). The main advantages of the AQDS/Br system are low electrolyte costs and fast kinetics. The crossover of bromine into the negative half-cell could affect the lifetime and the stability of the organic compound should be investigated. The array of p-aromatic redox-active organic compounds constitutes a new opportunity for low-cost large-scale energy storage.
Another cell chemistry employing quinones is the 2,6dihydroxyanthraquinone (DHAQ)/Fe(CN) 6 system in alkaline media [133]. At pH 14, the Fe(CN) 6 3-/Fe(CN) 6 4-avoids the use of the problematic Br 2 /HBr employed in the acidic quinone chemistry [128], and the cell reaches a higher cell voltage of 1.2 V. The modest solubility of both DHAQ and Fe(CN) 6 , however, limits the theoretical energy density to 9.2 Wh/L, of which 74% have been shown experimentally [133].

A RFB with an alloxazine-based organic electrolyte
In this work an aqueous RFB utilizing an organic redox compound was reported. An electrochemically active molecule alloxazine, which was a tautomer of the isoalloxazine backbone of vitamin B 2 was used [134]. It could be obtained by a single-step coupling of ophenylenediamine derivatives and alloxan with a yield of 95% at room temperature. The synthesis route was inspired by nature and the educts were reported to be inexpensive.
The alloxazine was functionalized with a carboxylic acid group (alloxazine 7/8-carboxylic acid(ACA), see Fig. 11) to increase its solubility in aqueous solution. ACA was investigated by CVand the standard potential was -0.62 V vs. SHE. The rate constant was calculated from rotatingdisc electrode measurement and determined to be k 0 = 1.2 AE 0.2 Â 10 -5 cm/s, which was an order of magnitude higher than that of the reactions of the VRFB [26]. The full-cell test was conducted using 0.5 M ACA (1.5 mmol) as the negative electrolyte and 0.4 M ferrocyanide (4.5 mmol) + 40 mM ferricyanide (0.46 mmol) as the positive electrolyte. Both solutions were adjusted to pH 14 by KOH. An excess concentration of the positive electrolyte was used to determine the electrochemical stability of the negative electrolyte. The resulting alkaline aqueous RFB exhibited an OCV of 1.2 V and h CE and capacity retention reaching 99.7% and 99.98% per cycle, respectively. The alloxazine redox center demonstrated a high electrochemical and chemical stability and a high solubility. It was shown in theoretical studies that structural modification such as adding electron-donating groups to the alloxazine could increase the battery voltage. This aza-aromatic molecule undergoes reversible redox reaction in aqueous electrolyte and constitutes a class of radical-free redoxactive compounds for application in RFBs with high power density and low costs. 6.9 TEMPO-based catholyte for non-aqueous RFBs Although only a catholyte was presented, due to its potential it was included in this review. A lithium hybrid flow battery containing TEMPO, a stable, heterocyclic nitroxide radical, as the organic active material was developed [72]. Figure 12 shows the redox reactions taking place in anolyte and catholyte. CVs were conducted in an electrolyte of 0.005 M TEMPO and 1.0 M LiPF 6 dissolved in EC/PC/EMC (4:1:5 by weight) and clear redox peaks in the potential range of 2.5 -4.0 Vof TEMPO were obtained. The peak separation of 61 mV at 10 mV/s corresponds to a one-electron-transfer between the nitroxide radical and the oxoammonium cation.
The electrochemical performance was tested in a nonaqueous hybrid RFB with a polyethylene-based porous separator. The anode was a hybrid anode consisting of lithium foil and graphite felt, and as cathode, a graphite felt was used. Both electrodes were compressed by 15%. An additive, fluoroethylene carbonate (FEC), was added to protect the Li anode. TEMPO was dissolved in a solution containing EC/PC/EMC (weight ratio of 4:1:5) and LiPF 6 as supporting electrolyte. The TEMPO based electrolyte was static within the anode compartment and did not flow, whereas in the cathode compartment the electrolyte was circulated through the graphite felt at a flow rate of 50 mL/ min. Galvanostatic charge/discharge cycling with various concentrations of TEMPO (0.8/1.2 M, 1.5/1.8 M, and 2.0/ 2.3 M) were conducted. Charge and discharge experiments were conducted and 100 cycles were obtained with a h CE of 99%, a h VE of 87% and a h EE of 86%, and an average capacity retention of 99.8% per cycle were achieved. Due to high viscosities of the electrolytes, the flow cells were cycled at current densities of 2.5 mA/cm 2 at 1.5 M and 1.0 mA/cm 2 at 2.0 M.
The theoretical energy density of the system with 2.0 M TEMPO is 126 Wh/L and is much higher than that for a conventional VRFB. Due to the high energy density and overall voltage of 3.5 V, TEMPO is a promising candidate for flow batteries.

Overview of some redox reactions of importance for RFBs
The overview of some redox reactions of importance for RFBs is shown in Table 3. If approximate realized energy density is given, the energy content is calculated from the given publication. Chemistries not demonstrated in flow batteries were usually tested in a stationary cell configuration.
8 Conclusions: What would be the ideal RFB?
The primary properties of the ideal RFB chemistry are high energy-and power-density, long-time stability and low capital costs. Secondary features are high efficiencies and safety, and low toxicity of the chemistries. However, the detail in these secondary features will not be discussed, because efficiencies are very much related to the issues power density and stability, and the ideal toxicity level is easily assessed: The ideal RFB chemistry is not harmful to the health of humans or the environment, such as claimed by the nanoFlowcell Holdings Ltd. for their undisclosed nanoFlowcell technology 1) , which, due to its unverified claims, is regarded very sceptical by the scientific community.  Energy density: The (volumetric) energy density E for a RFB is given by the combination of Eqs. (3) and (6). There are three parameters that can be adjusted to increase E: the number of electrons transferred per molecule n, the concentration c, and the cell voltage ΔU . For most molecules discussed in this paper, n = 1, with some exceptions such as the AQDS and Br 2 /Brsystems which transfer n = 2 electrons [42,128,139] and POM systems with n ³3 [119,126]. In the latter case, this property is achieved only by employing heavy molecules (molar mass larger than 2000 g/mol). Due to the high solubility of POMs, this can increase the volumetric energy density, but hardly the gravimetric energy density. Highly charged species are often not stable, and therefore, multi-electron transfer is usually accompanied by proton transfer [121,122,153]. While this prevents the formation of radicals, balancing of protons adds an additional challenge to the system, if it does not operate at very low pH. As shown in Section 4, the question of maximum cell voltage ΔU is intricately connected to the question of aqueous or non-aqueous chemistry. While the limit of ΔU in water is roughly 1.6 V, this value can be significantly higher in organic electrolytes. However, as the use of non-aqueous electrolytes comes with a serious penalty in terms or costs (as we will show later) and ease of operation, the cell voltage has to be increased at least to 3 V in order to warrant drastic step [154].
Concentration c is probably the most freely adjustable parameter that determines the energy density. Employing slurries of lithium ion battery cathode materials concentrations as high as 12 M for the active material are reported [59]. Classical RFBs such as the VRFB reach concentrations of 1-3 M, often with the use of stabilizers or in concentrated acids.
Looking at the discussion of the three parameters above, something like the expected value can be defined in terms of volumetric energy density for an aqueous RFB: This cell transfers one electron n = 1 per molecule, anolyte and catholyte are present as c = 2 M solutions, and the cell voltage takes the maximum advantage of the stability window of carbon electrodes in water, ΔU ¼ 1:6V.The result is a battery with energy density E % 43 Wh=L: Any RFB chemistry that features a higher volumetric energy content, while simultaneously satisfying the other constraints (power, cost, safety, efficiency), would be desirable.
Power density: The power that can be drawn from a battery is limited by the overvoltage R total I that has to be applied to reach a certain current I. This total resistance R total , often given as area-specific resistance (ASR) of the power converter, comprises the single resistances R elec , R mem , R CT , and R mass as given in Eq. (7). R elec and R mass are from the realm of power converter design and electrochemical engineering, and are, therefore, not discussed here [23,24,42,114]. The membrane resistance R mem is determined by the membrane used or separator, but that is pre-determined by the chemistry used [82,155]. Ideally, size-exclusion membranes [89] or fluorinated exchange membranes with high proton concentrations are used to keep R mem low. In non-aqueous solvents or in neutral aqueous electrolytes, the membrane can have a significant contribution to the ASR. One major factor is the charge transfer resistance R CT that depends on the employed redox couple. The electron transfer constant k 0 which determines R CT (see Eq. (8)) spreads over more than three orders of magnitude for different redox couples used for RFBs (see Table 1). With c = 2 M/L, n = 1 and k 0 = 0.001 cm/s, the R CT = 0.1 Ω$cm 2 would contribute only minimally to the ASR for which an upper bound of 1.5 Ω$cm 2 was given [154]. Therefore, the ideal RFB chemistry has an electron transfer constant of k 0 ³ 0:001 cms -1 : Long-time capacity stability is an obvious criterion for an RFB. In terms of the chemistry, this can be subdivided into cycling stability and chemical stability. The former is diminished by side-reactions during charge and discharge, and crossover through the membrane. Only the fraction of electrolyte volume in the power converter is subjected to it. The latter, chemical stability, concerns either individual oxidation states or redox molecules themselves. For most of these phenomena, mitigation strategies can be found. For example, in a VRFB, hydrogen evolution takes place [56], ion-specific crossover through the membrane occurs [156], and V 2+ is oxidized by O 2 . All these effects lead to an imbalance in the electrolyte, thus anolyte and catholyte are not at the same SOC during battery operation. Remixing of the electrolytes or electrochemical rebalancing can be employed to mitigate this capacity loss [157]. More problematic is chemical instability, if the employed redox shuttles react to form other species. This is because even if the temporal capacity fade, such as for the bis((3trimethylammonio)propyl)-ferrocene dichloride/bis(3-trimethylammonio)propyl viologen tetrachloride seems extremely low, at the reported 0.10%/d [158], the capacity has dropped to 70% of its initial value after one year.
Degradation effects that do not stem from the electrolytes can be due to membrane fouling [159], bipolar plate corrosion, and electrode aging [50,160]. These are serious issues, but they could be remedied by replacing the cell stack which might be cheaper than replacing the electrolyte.
The ideal RFB chemistry features high Coulombic efficiency, however, ion crossover and side reactions can be tolerated if the battery is symmetric as in the case of the VRFB, and viable rebalancing strategies exist. Chemical instability cannot be tolerated, as even tiniest decay rates add up very quickly.
The discussion of costs for a RFB is dominated by the goal set by the Advanced Research Projects Agency-Energy (ARPA-E) of the Department of Energy (DOE) of the United States to limit the capital cost to US$100 per kWh for widespread adoption [161,162]. Since then, a number of cost calculations for RFBs have appeared [37,154,161], comparing the costs for the various components. For the VRFB, Zhang et al. found that the electrolyte and the membrane were the major cost contributors [37]. However, it is also reasonable that when the main application of RFBs shifts from frequency regulation to grid storage, the ratio of power (kW) to energy content (kWh) will decrease, favoring longer storage times over power output. In that case, the contribution of the power converter will decrease in importance, and the main cost factor will be the electrolyte. The cost for battery grade V 2 O 5 , a convenient benchmark, fluctuated a lot over the last decade, with maximum costs of US$28 kg -1 and minimum costs of US$8 kg -1 [163]. Assuming a cost of US$21.13 kg -1 for vanadium, the capital cost price tag for a 1 MW/12 MWh VRFB battery was put at US$400 kWh -1 [37]. Acknowledging that even at a power to energy ratio of 1:12, this value greatly exceeds the DOE's target. There are two conceivable pathways for capital cost reduction: First, it is considered that the vanadium employed does not degrade over the course of the battery's lifetime. Therefore, it can be fully recuperated at end of operation, potentially with minimal purification necessary. The capital cost spent is, therefore, not lost but is merely an investment. With respect to recycling, the VRFB, and RFBs in general, have a clear advantage over LIBs in which the single components are more intimately entangled and cannot be separated easily.
The second possibility is to replace vanadium as active element by cheaper molecules. Substitution of vanadium by organic molecules that only contain cheap elements [89,118,128] is envisioned, or by replacing it with cheaper metals such as iron and chromium [164] or manganese [142]. Whatever the strategy, Darling et al. concluded that the active species should not cost more than US$5 kg -1 and the electrolyte not more than US$ 0.1 kg -1 (for aqueous systems) assuming a weight of not more than 150 g per mole of electrons stored [161]. They further concluded that while the design space for non-aqueous systems seemed wider than for aqueous chemistries, the additional costs for the electrolyte (must not cost more than US$5 kg -1 ) imposed additional hurdles that needed to be taken by the chemistry such as ΔU > 3 V and c > 4 M. Therefore, the conclusion of this paper is that the ideal system, in terms of cost, is either a combination of two aqueous redox couples that can be easily produced on large-scale and is cheaper than US$5 kg -1 , or that there is rethinking regarding the VRFB system, with the electrolyte not seen as something that is spent after the battery's operation, but as an investment that keeps its value, or might even increase in value during the operation of the battery. Matthäa Verena Holland-Cunz, née Schwob, studied Bachelor and Master of Chemistry at University of Cologne, Germany, and completed her master thesis with focus on the investigation of new electrode materials for redox flow batteries at Fraun-hoferInstitut in Freiburg im Breisgau, Germany. Since 2016 she is enrolled as Ph.D. student in Physical Chemistry at Newcastle University. Her research is about the electron transfer kinetics of the vanadium redox reaction and the investigation of new advanced materials for redox flow batteries.
Faye Cording is a postgraduate student in Chemistry of the School of Natural and Environmental Sciences at Newcastle University. She completed undergraduate studies in Chemistry at Northumbria University before joining Newcastle University to study new materials for redox flow batteries.
Dr. Jochen Friedl is a research associate in Chemistry of the School of Natural and Environmental Sciences at Newcastle University. For the past 7 years he has been researching redox flow batteries, with a focus on the electron transfer kinetics of the vanadium redox reactions and novel cell chemistries. His investigations of the electron transfer of vanadium revealed the influence of oxygen functional groups on the redox reactions that are at the heart of the anolyte and the catholyte of the all-vanadium redox flow battery. He helped to develop and scale up a novel redoxchemistry based on nano-sized electron shuttles from batteries that produced some mA to a single cell that can deliver 100 A. Author of 20 publications, he has received his doctorate from the Technical University of Munich, Germany, with highest distinction in 2015.
Ulrich Stimming was educated at the Free University of Berlin, Germany, where he received his Diploma degree in Chemistry and a Ph.D. in Physical Chemistry.
Prof. Stimming is currently a Professor of Physical Chemistry at Newcastle University, UK. Previously, he was Head of the School of Chemistry at Newcastle University. He was CEO and Scientific Advisor and Principal Investigator of TUM Create in Singapore. He had a Chair of Technical Physics and was Professor of Chemistry at Technical University Munich (TUM). Prior to that he was a Director at the Research Center Jülich and before a member of the faculty of Columbia University in New York, USA. He has visiting appointments at various universities including Shanghai Jiao Tong University and University of Science and Technology of China.
Currently, he is the Director of the North-East Centre of Energy Materials (NECEM) funded by EPSRC. He also directs a large battery project at Newcastle University on degradation of Li-Ion batteries in cooperation with Cambridge, UCL and Glasgow universities. Prof. Stimming is the founder and Editor-in-chief of the scientific journal Fuel Cells-From Fundamentals to Systems, VCH-Wiley. He coordinated for the Association of Leading Technical Universities in Germany (TU9) a research network of a total of 8 universities in electro-mobility between Germany and China, and was the co-director of the Joint "Institute for Advanced Power Sources" of TU Munich and Tsinghua University, Beijing, from 2010 to 2014. He has 300 + publications and numerous patents.