Development of organic redox‐active materials in aqueous flow batteries: Current strategies and future perspectives

Aqueous redox flow batteries, by using redox‐active molecules dissolved in nonflammable water solutions as electrolytes, are a promising technology for grid‐scale energy storage. Organic redox‐active materials offer a new opportunity for the construction of advanced flow batteries due to their advantages of potentially low cost, extensive structural diversity, tunable electrochemical properties, and high natural abundance. In this review, we present the emergence and development of organic redox‐active materials for aqueous organic redox flow batteries (AORFBs), in particular, molecular engineering concepts and strategies of organic redox‐active molecules. The typical design strategies based on organic redox species for high‐capacity, high‐stability, and high‐voltage AORFBs are outlined and discussed. Molecular engineering of organic redox‐active molecules for high aqueous solubility, high chemical/electrochemical stability, and multiple electron numbers as well as satisfactory redox potential gap between the redox pair is essential to realizing high‐performance AORFBs. Beyond molecular engineering, the redox‐targeting strategy is an effective way to obtain high‐capacity AORFBs. We further discuss and analyze the redox reaction mechanisms of organic redox species based on a series of electrochemical and spectroscopic approaches, and succinctly summarize the capacity degradation mechanisms of AORFBs. Furthermore, the current challenges, opportunities, and future directions of organic redox‐active materials for AORFBs are presented in detail.


| INTRODUCTION
Renewable energy sources, such as solar and wind energy, are taking a growing share of global energy production, which is predicted to be at least 32% in 2030 according to the target set by 2018 Renewable Energy Directive, to minimize the carbon footprint and to construct a green and sustainable society. [1][2][3] However, these renewable energy sources are intermittent and frequently unpredictable, requiring low-cost and durable energy storage systems to incorporate them into the electric grid. 4,5 Rechargeable battery systems represent simple and flexible technologies to store electricity for various applications, from small portable electronics to large-scale stationary storage. [6][7][8][9][10] Among them, lithium-ion batteries with high energy density and a relatively long cycling life have been commercialized in small portable devices and electric vehicles. 7,8 However, the static rechargeable batteries that rely on solid electrode materials to store electrical energy have relatively high costs and limited ability to decouple power and energy, thus hindering their widespread application for large-scale electricity storage.
Redox flow batteries (RFBs), which work via the reversible electrochemical reaction of redox-active materials in a circular flowing electrolyte, have been recognized as a promising technology for grid-scale electricity storage exceeding the level of MW/(MWh). [11][12][13] Specifically, RFBs store electrical energy in redox-active electrolytes that are circular flowing between the external reservoirs and the cell stacks, and herein, the redoxactive electrolytes are liquid electrolytes containing soluble redox species. The redox-active electrolytes flowing through the positive and negative electrodes are referred to as the catholyte and the anolyte, respectively. Thanks to the ingenious configuration design, RFBs are endowed with prominent features including decoupled energy and power, high current and power capability, scalability, and safety, showing excellent potential for grid-scale and long-duration energy storage. The first RFB, dating back to 1884, was the zinc-chlorine system. 14 Since then, RFBs with various inorganic redox species, such as zinc, iron, and vanadium ions, have been extensively developed. 15 However, the material scarcity, high cost, and environmental impacts of inorganic species are the inherent drawbacks of traditional inorganic RFBs. To this end, organic RFBs, particularly aqueous organic RFBs (AORFBs), with organic redoxactive molecules and nonflammable water as electrolytes, could be more sustainable and cost-effective options for green and grid energy storage. 16,17 In the past decade, AORFBs, with organic redoxactive molecules dissolved in flowing aqueous electrolytes (Figure 1), have received considerable interest primarily because of the following outstanding merits:  organic redox-active molecules have high element abundance, minimal environmental impacts, and potentially low cost. Their redox center(s) commonly is/are oxygen, nitrogen, carbon, or sulfur atom(s). The structural diversity and designability of organic redoxactive materials enable the high tunability of physical and chemical properties, for example, solubility, stability, and redox potential. To date, the reported organic redox species in the literature can be mainly classified into iron-based organic complexes, quinones, viologens, phenazines, phenothiazines, 2,2,6,6-tetramethyl-1piperidinyloxy (TEMPO) derivatives, azobenzenes, alloxazines, and so on. [66][67][68][69][70][71][72][73] Among them, the most studied organic redox species are ferrocene, TEMPO, quinone, viologen, and phenazine derivatives. Moreover, aqueous electrolytes composed of water and simple inorganic supporting electrolytes, in combination with welldeveloped selective ion-conductive membranes, have many advantages, such as high safety, low cost, and high conductivity.
To address the primary and long-standing challenges of AORFBs, including low energy density and cycling stability, molecular engineering of organic redox-active molecules is the key to achieving high-performance AORFBs since water-soluble organic redox-active materials serve as energy carriers to establish limits for overall battery performance including capacity, voltage, and cycling stability. Therefore, in this review, we focus on the redox chemistry and current concepts and strategies of organic redox-active materials, particularly molecular engineering, in terms of solubility, stability, redox potential, and electron number for AORFBs. We attempt to establish the relationship between molecular structures and battery performances and thereby provide guidance for the construction of advanced AORFBs.

| CURRENT CONCEPTS AND STRATEGIES
A typical battery configuration of AORFBs, as shown in Figure 2, consists of two major units: the energyconverting unit (cell stack) and the energy storage unit (two separate reservoirs for storing the catholyte and the anolyte). The membrane in the cell stack, usually an ionexchange membrane, can separate the two half-cell electrolytes and allow the cross-transport of nonactive ions to balance the charge of the catholyte and the anolyte. Energy density and stability are the two most important elements for establishing advanced AORFBs. Energy density is principally dependent on the solubility F I G U R E 1 Typical organic redox species for aqueous organic redox flow batteries (AORFBs) (bracket: redox potential vs. standard hydrogen electrode (SHE) unless otherwise stated; red, light-blue, and navy-blue colors indicate pH acidic, neutral, and alkaline environments, respectively).  F I G U R E 2 A typical aqueous organic redox flow battery (AORFB) with organic redox-active materials dissolved in aqueous electrolytes. PAN ET AL. and the electron number of redox species, and the redox potential difference between the catholyte and the anolyte. Stability is principally determined by the chemical and electrochemical stability of redox-active materials and the crossover of the redox species. Hence, we summarize the typical design strategies and concepts based on organic or organometallic molecules in terms of solubility, stability, redox potential, and electron number for the construction of high-energy-density and stable AORFBs.

| Strategies for the construction of high-capacity AORFBs
The theoretical energy density of a full battery can be calculated using the equation: energy density (Wh/L) = ncFV/μ ν , where n is the electron transfer number; c is the lower concentration of the active material; F is the Faraday's constant (26.8 Ah/mol); V is the cell voltage; and μ ν is the volume factor. Therefore, when the cell voltage, the electrolyte concentration, and the electron transfer number increase, the battery energy density tends to increase. This indicates that it is very crucial to design innovative organic redox species with higher cell voltage, higher solubility, and larger electron transfer number for the application of highenergy-density AORFBs. The maximum electron concentration of redox-active materials, which is the multiplication of molecular solubility and electron transfer number in a molecule, determines the theoretical capacity of AORFBs. Solubility is the principal factor when designing redox-active materials for high-capacity AORFBs. Moreover, the number of electrons involved in redox reactions also has a positive and linear relationship with the capacity of AORFBs. The more the electrons in the molecule, the higher the capacity of AORFBs.

| Improving the solubility of redox-active materials
High solubility of a solute can be achieved by strengthening the solute-solvent interaction or weakening the solutesolute interaction because dissolution is the solvation process of the solute, where the solute-solvent attraction should be strong enough to overcome the solute-solute forces. Recently, Zhang et al. 68 proposed the solubility standard for the evaluation of the solubility level of redoxactive materials. For example, high solubility of a redoxactive material means that the maximum molecular concentration should reach or exceed 1 mol/L in the solution under different conditions, such as with the exsitence of different supporting salts and at different redox states. Note that most reports only presented the solubility data of redox-active materials in the discharged state and overlooked their solubility in the charged state. 70 The solubility of redox-active materials in the charged state is always difficult to measure but should also be considered to be practically meaningful because the solubilities of a redox species in the charged state and the discharged state are always different, particularly when the solubility in the charged state has lower solubility. According to the rule of "like dissolves like," hydrophilic functional groups such as ammonium/amino, pyrrolidinium, hydroxy, sulfonate, and carboxylate groups are frequently used for improving molecular aqueous solubility. In most cases, the hydrophilic functional group is not redox-active. Under certain circumstances, some redox-active centers such as pyridinium can also act as hydrophilic functional groups. When the hydrophilic functional group has an impact on the redox center, the physiochemical properties, such as redox potential and cycling stability, of the molecule can be accordingly modified. Moreover, hydrogen bonding may also have a nonnegligible impact on the electrochemical properties of redox-active materials. Other strategies such as breaking the symmetry of molecular structures, replacing counter ions, and adding hydrotropic additives are also effective to increase the water solubility.
Introducing hydrophilic functional groups: The incorporation of high-polar and hydrophilic functional groups including ammonium/amino, pyrrolidinium, hydroxy, sulfonate, and carboxylate groups into organic redoxactive materials is the most common way to increase molecular aqueous solubility. In some cases, the redoxactive center such as pyridinium itself can also act as the hydrophilic functional group. Herein, we present some representative examples of organic redox-active materials with high solubility in water or different pH media, as shown in Scheme 1. For example, grafted ammonium and pyrrolidinium groups were incorporated into the terminal of TEMPO units (namely, TMAP-TEMPO and Pyr-TEMPO), resulting in extremely high water solubilities of 4.62 and 3.35 mol/L, respectively, which correspond to theoretical catholyte capacities of 123.8 and 89.8 Ah/L, respectively. 39,64 The redox-active complex, (ferrocenylmethyl)trimethylammonium chloride, has high water solubility of 4.0 mol/L, but undergoes severe light-induced decomposition. 26 The potassium salt of hydroxy-substituted benzoquinone, such as 2,5-dihydroxy-1,4-benzoquinone, showed ultrahigh solubility of 4.31 mol/L in 1.0 mol/L KOH, which is equivalent to an 8.62 mol/L electron concentration and corresponds to a theoretical anolyte capacity of 231 Ah/L. 35 Methyl viologen, with two methyl pyridinium units coupling with a single bond, has high solubility of 2.5 mol/L in water. 24 Moreover, the two ammonium terminals were introduced into the viologen molecule, resulting in water solubility of 2.0 mol/L. 27 Note that the suitable positive or negative charges of ionic group (s)-substituted organic molecules can effectively avoid the molecular aggregation due to the inherent strong Coulombic repulsion, and thus increase the molecular aqueous solubility. 27,77 However, it may result in a decrease in water solubility on introducing excess charges into the molecule because of the strengthened cation-anion interactions between the solute and the solute. 24,27,39,61 In addition, liquid molecular design, for example, using polyethylene glycol groups in the quinone molecules such as 1,8-bis(2-(2-(2-hydroxyethoxy)-ethoxy)ethoxy)anthracene-9,10-dione (PEGAQ), can lower the melting point, which seems to be a promising approach to overcome the solubility limit. 38 A high volumetric capacity of 80.4 Ah/L and energy density of 25.2 Wh/L for the redox pair of PEGAQ and ferrocyanide were demonstrated. However, high viscosity in the highconcentration cell would result in lower capacity utilization due to mass-transport limitations. Moreover, the use of a phenazine derivative, 7,8-dihydroxyphenazine-2-sulfonic acid (DHPS), yielded high solubility of 1.8 mol/L in 1 mol/ L KOH. 30 A reversible anolyte capacity of 67 Ah/L at 1.4 mol/L DHPS in 1 mol/L KOH with a capacity fade of 0.0195% per cycle (0.68% per day) was achieved ( Figure 3).
Breaking the symmetry of molecular structures: The solubility of a substance can be modified by changing the molecular symmetry. According to Carnelley's rule, molecules with high symmetry have lower solubility than the analogues with low symmetry. 78 Thus, it is feasible to design organic redox-active molecules with low symmetry for high solubility. In this regard, Zhu and colleagues 53 designed symmetry-breaking iron complexes (M 4 [Fe II (Dcbpy) 2 (CN) 2 ], M = Na, K), which showed 4.2 times higher solubility (up to 1.22 mol/L) than that of the symmetrical M 4 [Fe II (Dcbpy) 3 ] ( Figure 4). The flow cell constructed with 1.02 mol/L Na 4 [Fe II (Dcbpy) 2 (CN) 2 ] and 1.2mol/L SPr-Bpy demonstrated a high capacity of 19.4 Ah/L and a practical energy density of 12.5 Wh/L. Moreover, Yu and colleagues 48 selected a high-soluble 4-amino-1,1′-azobenzene-3,4′-disulfonic acid monosodium salt (AADA) with two sulfonic groups and an amino group to form an asymmetrical configuration to achieve high solvation energy because the asymmetric charge distribution of AADA offers strong preferential solvation sites. Some other groups also observed that the asymmetrically substituted redox-active molecules showed higher solubility than the symmetrically substituted analogues. 79 Replacing counter ions: By replacing the counter ion with another ion, a change in the solubility of ionic redox-active species may occur. Anthraquinone derivatives have been widely used in acidic and alkaline electrolytes; however, their chemical degradation S C H E M E 1 Molecular structures, electron transfer numbers, and solubilities of some typical organic redox-active materials. 24,26,27,29,30,32,35,38,39,41,54,64,77 In the bracket, the electron transfer number (n) and molecular solubility in water or aqueous solution are summarized. BTMAP-Vi, bis(3-trimethylammonio)propyl viologen tetrachloride; 2,6-DBEAQ, 4,4′-((9,10-anthraquinone-2,6-diyl)dioxy) dibutyrate; DHBQ, 2,5-dihydroxy-1,4-benzoquinone; DHPS, 7,8-dihydroxyphenazine-2-sulfonic acid; FcNCl, (ferrocenylmethyl) trimethylammonium chloride; FQH 2 , 2,3,5,6-tetrakis((dimethylamino)methyl)hydroquinone; MV, methyl viologen; PEGAQ, 1,8-bis(2-(2-(2-hydroxyethoxy)-ethoxy)ethoxy)anthracene-9,10-dione; Pyr-TEMPO, 4-[3-(methylpyrrolidinio)propoxyl]-2,2,6,6-tetramethylpiperidine-1-oxyl; Spr-Bpy, 1,1′-bis(3-sulfonatopropyl)-4,4′-bipyridinium; TMAP-TEMPO, 4-[3-(trimethylammonio)propoxy]-2,2,6,6-tetramethylpiperidine-1-oxyl. remains challenging. 69 To address this challenge, Liu and colleagues 36 reported the 9,10-anthraquinone-2,7disulfonic diammonium salt, AQDS(NH 4 ) 2 , as an anolyte material for pH-neutral AORFBs by exchanging the Na + cation in AQDSNa 2 with an NH 4 + cation, resulting in a significant increase in solubility up to 1.9 mol/L in water, which is more than three times that of AQDSNa 2 ( Figure 5). This is because the NH 4 + cation is more hydrophilic than the Na + cation and can form a hydrogen bond with the AQDS anion, thus enhancing the solvation of the AQDS anion in water and improving its solubility. The AQDS(NH 4 ) 2 /NH 4 I AORFB achieved a high practical capacity of 30.7 Ah/L, an energy density of 12.5 Wh/L, and outstanding cycling stability over 300 cycles. It may be a simple and straightforward way to replace the counter ion with a more hydrophilic ion to improve the solubility of ionic redox-active materials. It is worth mentioning that the compatibility issue of the cation-exchange membrane such as ion selectivity and conductivity, when changing the counter ion from one to another, needs to be considered. Adding hydrotropes: A hydrotrope or hydrotropic agent is an additional solute that can induce an increase in the aqueous solubility of the investigated solute. Adding a hydrotropic agent may be an effective way to increase the solubilities of organic redox-active materials in the aqueous electrolyte. It is conceivable that a large amount of hydrotropic agent can alter the nature of the solvent and interact with the solute via weak interactions such as hydrogen bonds and host-guest interactions. 80 For example, the addition of 4 mol/L urea can significantly increase the aqueous solubility of benzene-1,4-diol (H 2 BQ) from 0.5 to 1.5 mol/L due to the hydrotropic effect. 81 The practical energy density of AORFB at 1 mol/L H 2 BQ with urea addition can reach 25.3 Wh/L. Interestingly, the precipitation of BQ did not occur in the battery test due to the hydrotropic effect of urea. Moreover, urea as the hydrotropic agent can promote the stability of concentrated electrolytes of AADA through hydrogen-bonding interaction, thus resulting in a high reversible capacity of up to 41.5 Ah/L. 48 Recently, Chen and colleagues 82 used ionic liquids, that is, imidazolium chlorides, as novel hydrotropes to increase the water solubility of 4-OH-TEMPO from 2.1 to 4.3 mol/L using an "interaction-mediating" strategy. It is notable that the supporting electrolytes composed of 3 mol/L imidazolium chlorides in water have a high ionic conductivity of up to 83 mS/cm and water-like flowability with a low viscosity of up to 1.8 mPa s. Furthermore, Yang and colleagues 83 demonstrated that the water solubility of ferrocene derivatives can be increased from nearly zero to 0.28 mol/L by simply mixing water-insoluble ferrocenes with a hydroxypropyl-β-cyclodextrin additive via host-guest addition. Moreover, nicotinamide (vitamin B 3 ) was used as a hydrotropic agent to improve the aqueous solubility of flavin mononucleotide from 0.1 to 1.5 mol/L in 1.0 mol/L KOH. 84 However, the solubility data of flavin mononucleotide seem to be controversial because the solubility was measured to be as high as 1.3 mol/L in 1.0 mol/L KOH without adding the nicotinamide hydrotrope, as reported by another group. 85 Briefly, using hydrotropic agents may increase the solubility of redox-active materials, but may also introduce some new factors, such as the chemical/ electrochemical stability variation of hydrotropes, the possible change of electrochemical properties, and the compatibility change with the membrane, which should be taken into consideration.

| Designing multiredox organic materials
Due to the flexible designability and tunability of organic species, it may be possible to create multiredox organic materials that can undergo multiple electron transfer reactions (electron number, n ≥ 2). However, in the reported literature, most of the organic redox-active molecules showed one or two reversible electron storage. When more than two electron transfer reactions occurred in a redoxactive molecule, the induced solubility and redox stability/ reversibility issues would become challenging. To the best of our knowledge, only a few organic redox-active materials capable of three or more electron transfers are reported and, to date, they either have low stability or low aqueous solubility, making them impractical for real-world applications. 52,86 Currently, small organic redox-active molecules with reversible two-electron storage and high aqueous solubility are the most promising candidates for highcapacity AORFBs.

| Redox-targeting strategies
Redox-targeting strategies that use diluted solutions of soluble redox mediators to catalyze the charge/discharge of solid energy storage materials in tanks can significantly boost the capacity of RFBs. 87 As a result, the capacity no longer depends on the concentration of the redox electrolyte, but on the solid materials stored in the tank. The position of the redox potential between the soluble redox mediator and the solid energy-storage material should be as close as possible to ensure charge extraction from solid materials. To date, the redox-targeting strategy has achieved great success in inorganic flow batteries, but is rarely applied in AORFBs. [88][89][90] Recently, Wang and colleagues 90 reported that poly(anthraquinonyl sulfide)/carbon black (PAQS/CB) was used as a capacity booster for a 1,5dihydroxyanthraquinone (1,5-DHAQ)-based anolyte system via single-molecule redox-targeting reactions, enabling a high volumetric capacity of 47.3 Ah/L with a low capacity fading rate of 0.02% per day for over 1000 h ( Figure 6). However, compared with soluble redox-active materials that directly undergo redox reactions, the redox-targeting strategy needs an additional catalytic process for charge transfer, thus resulting in slower reaction kinetics and limiting its applicability for AORFBs under high current density.

| Strategies for building high-stability AORFBs
Stability is one of the most important factors that should be considered when evaluating the performance of AORFBs. There are many factors that can affect the stability of AORFBs during the operation of AORFBs. Among them, water-splitting reactions (i.e., hydrogen evolution reaction [HER]/oxygen evolution reaction [OER]), decomposition or other side reactions of redox-active materials, and membrane crossover issues are the major factors that affect the capacity degradation of AORFBs. It is anticipated that pH-neutral or near-neutral systems can alleviate the competitive HER or OER at an extreme pH level. 67 By molecular engineering, there are lots of strategies, including creation of a πconjugated extension, design of systems with stable radicals or nonradicals, and the use of hydrogen bond effects, that can be used to enhance the intrinsic chemical and electrochemical stability of redox-active materials. The anion-exchange membrane crossover issues can be mitigated by the design of high-selectivity and low-resistance anionexchange membranes and non-or ultra-low crossover redoxactive materials. 68 Moreover, a detailed discussion on capacity decay mechanisms has been well presented in an excellent review, providing guidance for the rational design of redox-active materials. 69 Herein, we mainly present the strategies in molecular engineering of redox-active materials for high-stability AORFBs rather than in membrane or electrode engineering.

| Creation of π-conjugated extension
The strategy of "π-conjugated extension" is used to enhance the molecular stability in both charged and discharged states, due to the broader delocalization effect. For example, a conjugation-extended viologen was designed for reversible two-electron storage by using thiazolo [5,4-d]thiazole as a central π-conjugated framework linked in two pyridinium moieties (Figure 7). The [(NPr) 2 TTz]Cl 4 /N Me -TEMPO AORFB with a cell voltage of 1.44 V delivered 99.97% capacity retention per cycle and 70% energy efficiency. 33 However, the planar extended π-conjugation results in a relatively high redox potential (−0.44 V vs. the standard hydrogen electrode, SHE), which is only comparable to methyl viologen. To address this issue, we and Liang's group introduced a phenylene group into two pyridinium moieties, which are linked but not coplanar with the two pyridinium redox centers. 51,54 The bridging phenylene group plays a dual role in the extended bipyridinium system: (i) it prevents communication between the two pyridinium redox centers and (ii) it promotes a certain degree of π-conjugation and mitigates the intramolecular Coulombic repulsion between the two positively charged pyridinium centers. As a result, an extremely low redox potential (−0.77 V vs, SHE, 2e − ) and enhanced electrochemical stability were obtained. 54 By such molecular

| Design of systems with stable radicals or nonradicals
To guarantee sufficient electrochemical stability, the redox centers, such as C, N, and O, should be properly distributed in the molecular structure to either avoid the formation of possible radicals or stabilize the radical. Almost all the organic redox-active materials will undergo redox reactions accompanied by valence state and radical variations. Stable organic radicals, most frequently, TEMPO analogues, play an important role in building stable pH-neutral AORFBs. 39,60,61,64 Besides TEMPO derivatives, molecular engineering of fluorenone through the introduction of carboxylate and sulfonate groups on the aromatic ring enables reversible ketone hydrogenation and dehydrogenation via the formation of a stable radical intermediate for two-electron and stable energy storage in AORFBs ( Figure 8A). 50 Interestingly, the AORFB based on 4-carboxylic-7-sulfonate fluorenone showed stable cycling for more than 700 cycles at 50°C without inert gas protection, and slight capacity decay was observed with capacity retention of 87.8% over 780 cycles ( Figure 8B). Moreover, for metal-organic ligand complexes, the redox reactions generally occurred via the valence state change of metal ions and the formation of stable radicals. 49,53,63 Thus, the design of systems with stable radicals or nonradicals is very important for the establishment of oxygen-insensitive or less sensitive AORFBs.

| Use of hydrogen bond effects
In some cases, hydrogen bonding can have a significant impact on the stability of redox-active materials. 36,48,90 For example, a hydrogen bond was proposed to mediate the degradation and protection of the reduced states of 2,6-dihydroxyanthraquinone (DHAQ) species such as 2,6-DHAQ 2− and 1,5-DHAQ 2− ( Figure 8C). 90 Specifically, the intermolecular hydrogen bonds between two 2,6-DHAQ 2− may bridge the electron transfer and induce a disproportionation reaction, thereby forming dimers and resulting in a capacity decay. The reduced state, 1,5-DHAQ 2− , would preferentially bond with H 2 O on the two neighboring oxygen sites to form 1,5-DHAQ 2− ·H 2 O, preventing its coupling with another DHAQ molecule.

| Crossover mitigation of redox species
The crossover of organic redox-active species, particularly small organic redox species, is an inevitable challenge in AORFBs. Crossover mitigation strategies are highly desired to further improve the stability of AORFBs. Note that N Me -TEMPO and SPr-Bpy have high water solubility of 3.2 and 2.0 mol/L, respectively, but both have the disadvantage of severe crossover issues. 53,54 One general method that is used to alleviate the crossover is to increase the ionic charges of redox-active materials by molecular engineering. For example, THAQ presents a high solubility of 1.88 mol/L at pH 14, and the four protons on THAQ were deprotonated by the neutralization reaction with OH − . 77 This highly anionic anthraquinone molecule can avoid electrochemical dimerization and membrane crossover, resulting in an excellent cycling stability with ultrahigh capacity retention of 95.2% over 1100 cycles. As one of the most popular redox molecules, N Me -TEMPO showed very good chemical/ electrochemical stability; however, its remaining existing crossover issue limits the long-term cycling performance. 53 Recently, a stable and low permeable TEMPO catholyte was designed by the incorporation of a dual-ammonium dicationic group. 61 The dual-ammonium functionalized TEMPO molecule with more positive charges and a larger molecular size showed an excellent capacity retention of 90% for 400 cycles at 60 mA/cm 2 at 1.0 mol/L electrolyte concentrations.

| O 2 -balance strategy
For AORFBs that operate in an air atmosphere, continuous capacity fading generally occurs due to the effect of O 2 . For example, Wang and colleagues 91 used a previously reported organic redox-active material, that is, DPivOHAQ, as an anolyte for constructing air-operable AORFBs. As a result, the trace dissolved O 2 in the electrolyte led to the oxygen reduction reaction (ORR), thus breaking the capacity balance between the anolyte and the catholyte. To address this issue, an O 2 -balance strategy was proposed by using the OER in the catholyte to balance the ORR in the anolyte. 91 The OER in the catholyte could be induced using Ni(OH) 2 -modified carbon felt as the electrode for the catholyte. As a result, the AORFBs showed improved stability in air atmosphere, with a fade rate of 0.0012% per cycle for 1200 cycles.

| Strategies for building high-voltage AORFBs
High cell voltage is definitely beneficial to improving the power and energy densities of AORFBs. However, compared with an organic solvent, an aqueous electrolyte provides lower cell voltage due to the relatively narrow thermodynamic stability window of water (1.23 V under standard conditions). The HER and the OER of water can be significantly suppressed by using suitable electrode materials and electrolyte composition. 92,93 Because of the slow kinetics of HER and OER, the electrochemical stability window of aqueous electrolytes can be substantially larger than their thermodynamic limits. Similarly, the electron transfer kinetics of redox molecules for the redox reaction is always much faster than those of water-splitting reactions, which is why the redox molecules remain electrochemically stable. Notably, under high voltage conditions, trace amounts of HER and OER may induce a change in pH, which requires all redox states of redox molecules to be stable in this range of pH. Moreover, polarization on the electrode and a decrease in cell voltage may occur during cycling. For example, the voltage is closely correlated with internal resistance; this in turn considerably increases the overpotential between charging and discharging and thus decreases the cycling capacity or cell voltage. 68 Liu and colleagues 26 demonstrated that the overpotential of the OER (1.5 V vs. SHE) and the HER (−1.0 V vs. SHE) on the carbon electrode is as large as 2.5 V. Moreover, aqueous electrolyte composition can also considerably affect the electrochemical stability window. For example, highly concentrated aqueous solutions can markedly broaden the electrochemical stability window to 3 V or above. 94 Based on these findings, it seems promising to build high-voltage AORFBs with a cell voltage higher than 1.5 V by molecular engineering, although the cell voltage of most aqueous AORFBs is currently below this value. However, to the best of our knowledge, high-voltage AORFBs with a cell voltage of >1.5 V always show inferior stability than conventional AORFBs with a cell voltage of <1.3 V primarily because of the unavoidable water splitting and more severe crossover and degradation issues of the membrane. Therefore, it is highly desired but remains challenging to develop high-voltage and long-duration AORFBs with acceptable capacity fade.
Recent works have shown significant progress in the field of high-voltage AORFBs. 41,54,64,95 For example, by introducing a phenylene group into two pyridinium moieties, the resulting [(NPr) 2 PV]·4Cl anolyte presented two overlapped single-electron reductions at an exceptionally low redox potential (−0.77 V vs. SHE). 54 When paired with N Me -TEMPO, the AORFB yielded a cell voltage of 1.71 V and delivered reversible cycling with a capacity retention of 99.94% per cycle, an ultrahigh energy efficiency of~89%, and a Coulombic efficiency of~100% (Figure 7). Furthermore, pyrrolidinium cation-functionalized TEMPO and extended viologen redox pair were designed for long cycling life of over 1000 cycles and high cell voltage of 1.57 V (Figure 9). 64 F I G U R E 9 (See caption on next page) PAN ET AL.

| 13 of 20
Moreover, the assembled AORFBs showed a high energy density of 16.8 Wh/L and a peak power density of 317 mW/ cm 2 . Furthermore, zinc-organic hybrid flow batteries could be engineered with high voltage. For example, Park et al. 41 designed a hybrid RFB based on a functionalized 1,4hydroquinone bearing four (dimethylamino)methyl groups and a Zn/[Zn(OH) 4 ] 2− redox couple to achieve a high operating voltage of 2.0 V by utilizing a three-electrolyte, two-membrane configuration. The complicated twomembrane, three-electrolyte system as well as the short cycling life (~50 cycles) limit further application of RFB. Instead, Schubert's group 95 just used Zn 2+ salt as an anolyte and polymeric TEMPO as a catholyte, separated by a sizeexclusion membrane derived from regenerated cellulose, to construct hybrid RFBs with a simple two-electrolyte, onemembrane configuration (Figure 9). Note that the zinc anode has a very high overpotential for hydrogen evolution, expanding the applicable potential window up to 2 V. As a result, the static hybrid RFBs had a high cell voltage of 1.7 V and showed excellent long-term cycling over 1000 cycles. However, the formation of Zn dendrites would limit the long-term stability.

| REDOX REACTION MECHANISM STUDIES
Redox reaction is any chemical reaction in which the oxidation state of substance changes by gaining or losing an electron. A clear understanding of the redox reaction mechanism can provide an important foundation for the construction of advanced AORFBs. However, the redox reaction mechanism is always very complicated and can include not only the major reaction pathway(s) of redoxactive materials but also minor or side reactions. For simplification, herein, we mainly discuss the electron transfer mechanisms of major reaction pathways involved in one-/two-step, one-electron or one-step, two-electron reactions, or with/without proton coupling. For example, Grey's group 96 first used in situ nuclear magnetic resonance (NMR) technology to study the reaction mechanisms of organic molecules in AORFBs ( Figure 10A,B). The two-step, single-electron process for anthraquinone species such as DHAQ via radical formation and consumption was directly verified. In situ pseudo-two-dimensional 1 H NMR spectra were acquired as a function of electrochemical cycling. Upon charging, the proton signals closest to the carbonyl redox center (A and C) disappeared almost immediately, which can be ascribed to the electron delocalization over the semiquinone radical anion, whereas the proton signal farthest from the redox center (B) broadened and shifted downfield. As charging proceeded, the chemical shift of signal B reached a maximum, then reverted to lower values and became narrower as a consequence of further reduction of semiquinones. When the cut-off voltage (1.7 V) was reached and the charge potential was held at 1.7 V, the narrow proton signals A″, B″, and ″ clearly emerged, indicating the formation of the final diamagnetic product DHAQ 4− .
Considering the technical complication of in situ NMR analysis, Pan et al. 54 developed a series of ex situ spectroscopic approaches for a simplified analysis of the redox reaction mechanism of organic redox-active materials such as [(NPr) 2 PV]·4Cl. The two single-electron reductions of [(NPr) 2 PV]·4Cl, which underwent transformation from the initial cationic form to the monoradical form and then to the quinoid form during charging, were clarified by no-deuterium NMR (No-D NMR), electron paramagnetic resonance (EPR), and ultraviolet-visible (UV-vis) absorption spectra ( Figure 10C-F). In the No-D NMR method, the [(NPr) 2 PV]·4Cl anolytes at different charge states were directly used and the deuterated solvent such as D 2 O was sealed in a capillary, thus preventing contamination due to the deuterated solvent and reflecting the intrinsic signals of the sample. During the charging process, the alkyl ammonium groups such as α and β, which are away from the pyridinium redox centers, showed similar variations of the 1 H NMR chemical shift to the water peak that initially shifted downfield and then upfield. The NMR results demonstrated that the concentration of the formed radicals first increased and then decreased, revealing the two single-electron reductions and the generation of the quinoid form. Moreover, EPR spectroscopy was used to directly monitor the formation and concentration variations of free radicals. During charging, the initial increase and then decrease in the EPR response confirmed the formation of the monoradical form and then the quinoid form, indicating the two singleelectron reductions of [(NPr) 2 PV]·4Cl. In addition, UV-vis  2 PV]·4Cl, could also be confirmed using a simple electrochemical approach (cyclic voltammogram [CV] and differential pulse voltammetry curves). 64 CVs and the corresponding Pourbaix diagrams obtained by measuring the redox potentials at different pH values were effective electrochemical approaches to determine whether the redox reaction is pH-dependent or not. For example, in a study by Liu and colleagues, 36 the redox potential of AQDS(NH 4 ) 2 remained almost unchanged in the pH region of 4-7, indicating the twoelectron process of AQDS(NH 4 ) 2 without proton coupling, whereas the redox potential negatively shifted with a linear slope of −30 mV/pH from pH 7 to 10, revealing a two-electron process coupled with a single proton. Moreover, Aziz and colleagues 38 reported that the redox reaction of PEGAQ (AQ-1,8-3E-OH) involved a transfer of 2H + /2e − from pH 5 to 10.8, 1H + /2e − from 10.8 to 12.6, and 0H + /2e − from 12.6 to 14, revealing a pH-dependent electrochemical behavior of AQ-1,8-3E-OH.

| CAPACITY DEGRADATION MECHANISM STUDIES
To obtain ideal stability, the proposed redox reactions should reversibly and entirely occur without any competing side redox reactions. The capacity degradation mechanisms in the reported literature mainly include the instability of the redox species and the supporting electrolyte, the crossover issues due to the membrane and redox species, and degradation of the membrane and electrode components. Postelectrolyte electrochemical and spectroscopic characterizations including CV, NMR, and mass spectrometry spectra are common approaches that are used to study the decomposition of organic redox species. A number of groups, particularly Aziz's group, have carried out excellent capacity degradation mechanism studies of organic redox-active materials in AORFBs. 69,70 The possible chemical decomposition mechanisms of organic redox-active materials include dimerization, tautomerization, hydrolysis, disproportionation, and nucleophilic addition/substitution. 27,37,68,97,98 In addition, biological electrolyte regeneration systems, such as the use of Escherichia coli to regenerate the degraded/decomposed redox-active phenazines in the anolyte, may provide a possibility to further enhance the stability of RFBs. 99 So far, several recent reviews have well summarized the possible capacity degradation mechanisms of AORFBs. [68][69][70] Interested readers can access these studies and reviews for more detailed information.

| CONCLUSION AND PERSPECTIVES
Despite the significant developments that have been made in terms of AORFBs, there is huge room for improvement of battery performances including capacity, stability, cell voltage, and efficiencies. The design of the organic redox molecule, membrane, and electrode should be effective to realize AORFBs for practical applications. In this review, we mainly focus on the molecular engineering of organic redox-active materials to improve the battery performances of AORFBs. Considerable progress has been made in AORFBs based on molecular engineering strategies. However, the long-standing problems, in particular, low energy density and cycling stability, must be addressed. In other words, novel and effective strategies for high-capacity, high-stability, and high-voltage AORFBs are highly desired. For highcapacity AORFBs, there are three major solutions, i.e., (1) improving solubility of redox-active materials, (2) designing multiredox organic materials, and (3) introducing redox-targeting strategy. Among them, solubility is the primary consideration when designing organic redox species for high-capacity AORFBs. There are four main strategies for improving the solubility of organic redox-active materials, such as (1) employing hydrophilic functional groups, (2) breaking the symmetry of molecular structures, (3) replacing counter ions, and (4) adding hydrotropes. For high-stability AORFBs, molecular engineering strategies mainly comprise the creation of π-conjugated extension, design of systems with stable radicals or nonradicals, use of hydrogen bond effects, crossover mitigation of redox species, and the O 2 -balance strategy. For high-voltage AORFBs, the redox pair needs to have a large redox potential gap. Appropriate selection or design of stable redox species with low or high redox potentials is particularly important. However, the unwanted water-splitting reaction remains challenging, which requires suitable design of electrode materials and electrolyte composition. In most cases, high-voltage AORFBs (>1.5 V) always have inferior stability than low-voltage AORFBs (<1.2 V), probably caused by the unavoidable water splitting and more apparent membrane crossover and degeneration. Although the cycling stability of AORFBs can be improved to the practical level, the energy density of AORFBs remains less competitive to the traditional inorganic RFBs. Moreover, most organic radicals such as viologen radicals are sensitive to oxygen, preventing them from operating under ambient air conditions.
In view of the development history (2009-current), AORFBs are now at a rapidly growing stage full of opportunities and challenges. Compared with traditional inorganic RFBs, AORFBs are undoubtedly more environmentally friendly and not limited by resource scarcity. Due to the designability, diversity, and tunability of organic species, we anticipate that molecular engineering of organic redox-active materials will play a critical role in addressing the long-standing low energy density and cycling stability issues of AORFBs. To advance AORFBs for practical applications, in terms of molecular design, promising organic redox molecules should primarily have high water solubility, high chemical/electrochemical stability, and low cost. 100 Ideally, all redox states of organic materials are expected to be air-stable. 50,101 Moreover, the redox species should have low or high redox potential to ensure a high redox potential gap between the redox pair and should be able to function at high voltage. For high-voltage AORFBs, suitable electrode materials and electrolytes need to be designed to avoid the underlying water splitting. Meanwhile, advanced characterization methods including electrochemical and spectroscopic analyses and theoretical calculations are important to examine the redox reaction mechanisms and capacity fade mechanisms of AORFBs during the cycling process. Highthroughput DFT computations for the assessment of properties and high-throughput synthesis could be efficient methods in artificial molecular engineering for the selection and design of ideal organic redox species for advanced RFBs. The ideal organic redox molecules should have high aqueous solubility, high electrochemical/chemical stability, high positive or negative redox potential, fast kinetics, and low cost. Additionally, the synthesis should be energy-and time-saving, scalable, environmentally friendly, and involve simple steps without complicated purification treatments. Moreover, pH-neutral or weak alkaline electrolyte (pH 7−10) should be most promising for future AORFBs, especially if the solubility of organic redox species is further improved in this pH region, because the corrosion and undesired side reactions including water splitting and decomposition of redox-active materials can be efficiently alleviated. In this regard, pH-neutral or near-neutral AORFBs show great potential for the applications towards safe, clean, and sustainable grid energy storage.