Review Article: Flow battery systems with solid electroactive materials

Electrical energy storage devices have had dramatic increases in production due to the proliferation of personal digital devices, electric vehicles, and renewable energy technologies.^1–5 Dependent on the scale of the energy and power needs, the most common energy storage technologies include batteries, capacitors/supercapacitors, pumped hydropower, compressed air, and flywheels.^6–8 Among all of these energy storage technology options, batteries have attracted significant research attention recently in part because of the range of energy and power densities that can be achieved due to the variety of options with regard to battery chemistry and cell design.^8–11 Particular effort and success have been driven by high energy density rechargeable lithium-ion (Li-ion) batteries, which have seen their proliferation closely linked to the explosion in sales of the consumer devices where they dominate as the power source: smart phones, laptops, smart watches, fitness bands, and other portable electronics.¹² Li-ion batteries are also currently the dominant electric vehicle battery technology due to similar energy and power density advantages.^9,13

The battery technologies that are well-suited to portable electronics and transportation applications are not necessarily the best options for much larger scale stationary applications including emergency backup power and utility peak shaving or load leveling.^11,14 Even when hydrocarbon fuel sources are at low price points, renewable energy generation is still important due to political and environmental concerns.^11,15 Dealing with the intermittency of renewable sources such as solar and wind causes significant challenges for the electric grid due to the need for reliable power supply regardless of time of day and weather, and often supplemental fossil fuel power plants are operated to compensate for the dramatic change in power supply of renewables.^11,16 To operate renewable energies without nonrenewable fossil fuels, large scale electrical energy storage devices are needed to store the energy at peak hours and release it at off-peak hours. The redox flow battery (RFB) is a promising technology for this particular application due to its decoupling of power output and energy storage capacity and has been demonstrated in numerous large scale energy storage projects.^11,14,17 RFBs date back to at least the 1940s,^18–21 and recently interest in RFBs has increased in part because of the drive to enable larger scale energy storage applications to which RFBs are well suited, but also due to recent advances that have shown the potential to dramatically increase the energy density of these types of energy storage devices.^14,22–27 Transportation applications are also being explored for RFBs, and energy density advances would be particularly important in making RFBs more competitive for electric vehicles.²⁸ More details on RFB technology and its advantages and disadvantages will be discussed in Sec. I B after a brief review of static batteries in Sec. I A. Section I C will introduce the development of electrochemical systems that involve solid suspensions, relevant to the subset of RFBs that will be the focus of this review.

Batteries store electrical energy within chemical components of differing electrochemical potentials, the difference of which determines the battery voltage.²⁹ A wide variety of batteries have been developed and engineered depending on the application for different voltages, capacities, rate capabilities, geometries, energy densities, power densities, costs, etc.^12,29–33 Rechargeable Li-ion batteries have found widespread use due to their high energy density and retention of capacity with extended cycling.^34,35 We choose one of the most well-known commercial Li-ion battery material pairings, a lithium cobalt oxide (LiCoO₂, or LCO) cathode and a graphite anode, to explain the working principles of a rechargeable static battery. As shown in Fig. 1(a), a Li-ion battery consists of current collectors to transport electrons from the electrodes to the terminals of the battery, an anode (negative electrode), a cathode (positive electrode), a separator to prevent shorting of the electrodes, and electrolyte to provide ionic transport which is necessary to maintain charge neutrality during charge/discharge. LCO is an example of the material that would comprise the active material particles in the cathode and graphite would be the corresponding active material particles in the anode. Conductive carbon additives are mixed with the active materials to facilitate electron transfer within the composite electrode, which is held together by a polymeric binder that provides mechanical integrity, especially during cell manufacturing.^36–38 As indicated in Eqs. (1) and (2), Li ions are transferred from cathode to anode during the charging process and are transferred in the opposite direction during discharge. The total charge capacity is determined by the type and amount of active materials in the electrodes.

Another type of electrical energy storage device, the supercapacitor, functions differently. Supercapacitors store electrical energy within the electrical double layer near the surface of high surface area materials within an electrode, and redox reactions are not required.^29,39 Cations and anions migrate to the electrode surfaces on the appropriate sides of the cell during charging of the supercapacitor [shown in Fig. 1(b)]. Opposing surface charges on the electrodes balance the electric field created by the concentrated ions near the surface, generating a potential difference between the two electrodes. When the supercapacitor is discharged, the potential difference gradually diminishes while ions migrate away from the electrode surfaces, and eventually, the final result is a neutral electrolyte with minimal concentration gradient from anode to cathode.^29,39 The total active surface area within the electrodes and the concentration of ions in the electrolyte are the key factors that determine the total charge capacity and energy density. There is also significant research interest in hybrid supercapacitors, where one electrode functions as a supercapacitor and the other undergoes battery redox reactions.^40–45 More detailed discussion of supercapacitors can be found in a number of recent reviews.^{29,40,45–47}

The electrical energy stored in batteries is primarily stored as chemical energy that is released by redox reactions, as opposed to within a surface double layer. Thus, a straightforward approach to increase the cell capacity is to increase the amount of active material within the electrodes, either by increasing the relative fraction of active material within the electrodes or by increasing the electrode thickness. Unfortunately, this strategy only works within a limited range, because mass transfer resistance of the ions that must traverse through the electrodes will become limiting and the power output and current densities that the cell can operate at will become severely impaired.^48,49 We briefly mention here that some researchers have attempted to overcome the ion transport power limitations in the electrode by flowing the electrolyte through the electrode, which is different from a RFB since there is no active material being flowed through the system—only the electrolyte. One example was a convection battery proposed by Suppes et al., in which flowing electrolyte was used to facilitate ion transfer and reduce ionic resistance, resulting in a fivefold increase in achievable current densities; however, there were some performance limitations reported due to the separator and the operating efficiency after accounting for the pumping energy is unknown.^50–52 

One general challenge of using static batteries and static active material electrodes for stationary applications is that many smaller batteries must be connected in various combinations of parallel and series configurations to reach a final desired current and voltage output. The arrangement of many smaller batteries into a battery pack results in high costs due to the additional auxiliary parts, the need for a battery management system to safely and efficiently utilize all those cells, and typically the need for a heat removal system which also adds significant additional weight.^53–56 One promising approach to improve current density and also simplify battery system design and upgrades by decoupling the power and energy units in the cell is a RFB, which will be discussed below.

As shown in Fig. 2, a typical RFB consists of the following main components: power output cells—electrochemical cells with porous electrodes separated by ion-selective ionically conducting membranes; energy storage tanks—these tanks contain electrolyte comprised of the desired amount of dissolved redox species as well as soluble ions that transport across the membrane; and a pumping system to connect the power output cells to the energy storage tanks.^57,58 For the most common RFBs, the electroactive redox species are transition metal ions that have been dissolved in acidic aqueous electrolytes. During battery operation, the electrolytes are pumped through the electrochemical reaction cells to either oxidize or reduce the electroactive transition metal depending on which direction the current is flowing (charge or discharge) and which electrode the species is coming into contact with (anode or cathode). Note that in RFBs the electrolyte fed to the cathode is referred to as the catholyte, while the electrolyte fed to the anode is referred to as the anolyte. Reactions occur only on the surface of the porous electrodes in the reaction channel, which are typically comprised of porous carbon materials. The electrolytes are circled back to the original electrolyte energy storage tanks after exiting the electrochemical cell to mix with the remaining electrolyte solution while the new electrolyte material is fed from the tank back to the electrochemical cell. In this system, total battery capacity is limited by the volume of electrolyte within the electrolyte energy storage tanks, which is dependent on the tank size and electroactive material concentration. The power of the RFB, however, is limited by the performance and total number of electrochemical cells in the system.^57,58 As an example, all-vanadium RFB, the most successful RFB for large scale applications, involves four different oxidation states of vanadium ions forming two redox couples dissolved in highly acidic electrolyte. VO₂⁺ and VO²⁺ are dissolved in the catholyte and V³⁺ and V²⁺ in the anolyte.^14,57–59 The active species are the same chemical element, vanadium, on both sides of the cell. Issues limiting all-vanadium RFB include corrosion in the cell and limiting operating conditions to suppress gas evolution which results in low material utilization.⁵⁹ The crossover of active materials through the membrane, which is a major issue for RFBs with two different active material elements, is less of a concern for all-vanadium RFB because the cell can be rebalanced by redistributing anolyte/catholyte as opposed to requiring complete electrolyte replacement and reprocessing. This advantage is also shared by all-iron RFBs in which Fe³⁺/Fe²⁺ and Fe²⁺/Fe serve as redox couples in catholyte and anolyte, respectively.⁶⁰ More detailed discussion on active materials can be found in Sec. III.

The typical range of discharge current densities for RFBs based on dissolved transition metal species is between 20 and 80 mA cm⁻², though even higher current densities have been reported.^14,58,61,62 RFB systems are highly modular, providing the flexibility to independently design the power output by changing the design and number of electrochemical reaction cells; and to modify the total battery energy by optimizing the size and/or number of the storage tanks. This decoupling of power and energy is particularly advantageous in large scale energy storage applications because of the flexibility and potentially low cost.¹⁴ Many kW and MW scale RFB installations have been completed.¹⁷ For example, two 500 kW/6-h zinc-bromine RFBs were built in Massachusetts to lower peak energy demand and reduce the impact of power interruptions in 2016.^17,63 One of these RFBs is accompanied by a 605 kW photovoltaic array and the other with a 600 kW wind turbine to demonstrate the concept of integrating an intermittent power source with a scalable and large scale electrochemical energy storage platform. As another example, a 5 MW Vanadium RFB was installed in China supporting 10% of a 50 MW wind farm.^17,64 A few recent reviews provide good discussions on conventional RFBs.^{14,19,22,59,65–67}

Efforts to improve the performance of RFBs for next generation concepts have largely been focused on improving RFB energy density. RFB energy density is determined by the energy density of the electrolytes, which is relatively low compared to other battery technologies.^59,65,68 Commercial systems such as vanadium RFBs have reported energy densities on the order of 25 W h L⁻¹,^59,68 which is low in comparison to, for example, Li-ion technologies which range from 70 to 220 W h kg⁻¹ (100−450 W h L⁻¹), or even lead-acid (Pb-acid) batteries which have 25−50 W h kg⁻¹ (60−100 W h L⁻¹).⁶⁵ The lower the energy density, the greater the size of the storage tank needed to meet a specific energy requirement. The tank footprint is not necessarily an issue in stationary energy storage, but it makes it challenging for RFBs to penetrate other applications such as transportation, and even within stationary storage applications, the footprint is particularly important where space involves a cost and access premium such as in urban environments. The equipment size also influences the overall shipping and installation cost of the system. Therefore, energy density is a key metric for RFBs that research groups have aimed to improve.

The root causes of the relatively low energy density for conventional RFBs are as follows: (1) limited cell voltages due to the narrow electrolyte stability window and (2) low volumetric capacity due to solubility limits of the redox compounds.^57,58 The electrolyte stability window is limited because the solvent for conventional RFBs is water, and the thermodynamic stability range for water is ∼1.23 V (with the location of the upper and lower potentials highly dependent on the composition and concentration of the various solutes and additives).⁶⁹ To increase RFB operating voltages, much work has been and continues to be done to replace water with organic solvents with larger stability windows, and then to explore new redox compounds that are soluble and stable within organic electrolytes. Organic electrolytes provide a wider stability window to increase net cell voltage, and many new active materials and electrolytes have been reported.^67,70–72 The cost and flammability of organic RFB systems are significant considerations; however, the possibility to increase the voltage and subsequently the performance of these RFBs has driven significant interest. Most of the organic RFB reports have achieved at or above 2 V cells, though at this stage ionic conductivity limits the current densities to well below aqueous systems.^22,73 The voltage has even been reported as high as ∼3.5 V when paired with a Li metal anode.^65,74 There have also been efforts to construct hybrid organic-aqueous RFBs to take advantage of the low potential achievable at the anode due to the organic electrolyte and suitable potential and high ionic conductivity offered by the aqueous electrolyte on the cathode side.⁷⁵ More comprehensive discussions on organic and/or Li-based RFBs can be found in these recent reviews.^{19,22,65,67,76}

The second limitation, electroactive material solubility, is a challenge for both aqueous and organic RFBs. The higher the concentration of redox compounds in the electrolyte, the greater the number of electrons that can be exchanged for a given volume or mass of electrolyte and hence the higher the capacity and energy density of the RFB. Unfortunately, above the solubility limit for the electrolyte, the redox compounds precipitate out as inactive solid particles, and thus, the solubility limit provides one limitation on the cell energy density. Extensive research has been done to develop electrolytes with high solubility of active materials. For example, a number of reports demonstrated active material concentrations greater than 1 M for both aqueous and nonaqueous RFBs,^77–82 and even greater concentrations of over 5 M for the active material concentrations have been reported.^83–85 In addition, as the concentration of the redox compounds increases, and, in particular, if solid particles start to form, the viscosity of the electrolyte increases and the parasitic energy lost to pumping increases, in some cases dramatically.^25,86 Approaches have been developed for innovative solution preparations and solution chemistry, including the use of ionic liquids; however, significant increases in energy density and/or keeping viscosities reasonable is still a challenge.^{22,25,87–89} One increasingly popular approach is to start with insoluble solid particles as the electroactive materials that participate in the redox reactions. By starting with electroactive solid particles solubility limitations are no longer a relevant barrier, though the trade-off between particle loading and electrolyte viscosity is still a consideration. This strategy of starting with solid electroactive particles in a RFB can be applied with both aqueous and organic electrolytes, and is the focus of this review paper.

Solid suspensions are composed of nano- or micrometer sized solid particles dispersed within a continuous liquid phase. Solid suspensions have found a broad range of applications, and common design elements include the chemistry of the particles and/or liquid, solid particle size and morphology, particle surface chemistry, particle concentration and aggregate formation, and suspension rheological properties.^42,90–97 For example, thermal conductivity can be dramatically modified with changes in particle concentration within a suspension, and thus, solid dispersions have been popular for thermal storage and transfer applications.⁹³ Similarly, electrical conductivity of suspensions can also be manipulated by appropriate suspension design, which has particular relevance to electrochemical applications.⁹⁸ 

One application of solid suspensions with active material dispersed within an electrolyte is flowing supercapacitors. As mentioned in Sec. I A, supercapacitors store electric charge in the form of electric double layers on the surfaces of electrodes. An electrochemical flow capacitor (EFC) combines aspects of both RFBs and supercapacitors (shown in Fig. 3). Carbon suspensions, with the electrode surface being the surface area of the particulates within the suspension, are dispersed in storage tanks.^42,90–92 Similar to conventional supercapacitors, cations and anions migrate to carbon surfaces at two opposite electrodes under applied potentials during the charging step. Instead of using static and limited carbon materials as in a conventional supercapacitor, in an EFC the charged carbon is pumped to storage tanks while fresh uncharged carbon materials are pumped to the electrochemical cell to accept more ions. The discharge process is similar to the charge step with continuous carbon suspension supplied from storage tanks until the ions migrate to depletion from the carbon surfaces. The tank size is a parameter that can be changed to tune the total device capacity.⁹⁰ While EFCs in principle expand potential applications for supercapacitors, there are still challenges including low energy density and fast self-discharge. Research efforts have pursued new materials, flow cell designs, surface modification, and mathematical modeling.^{42,44,92,99–102} More detailed discussion on the current status and research perspectives for supercapacitors and EFCs can be found in these recent reviews.^{29,40,45,46,103,104}

Capacitive water deionization is another application where solid suspensions have been coupled into a system with electrodes and electrical potential driving forces.^104–107 In a conventional capacitive deionization system, a potential is applied between the two electrodes which forms electric double layers and drives ions toward the two electrode surfaces [shown in Fig. 4(a)]. Cations and anions are aggregated around the anode and cathode surfaces, respectively, which are both typically high surface area carbon materials.¹⁰⁸ As a result, relatively deionized water exits the cell. In a conventional capacitive water deionization system, the device needs to be periodically stopped and regenerated when it reaches its maximum capacitance. This intermittent operation limits the productivity and causes extra energy consumption during the regeneration procedure.¹⁰⁶ Solid carbon dispersions have also been applied to capacitive deionization applications [see Fig. 4(b)]. Once flowing carbon electrodes are integrated into the system, the device can run continuously with high capacity because the active carbon can be regenerated without interfering with the deionization process by simply mixing the two carbon suspensions, filtering the carbon suspensions to separate the liquid with concentrated ions, and reinjecting the carbon into the fresh water flow.^106,109,110 The system can be adjusted to fit particular working loads by changing channel size and/or flow rate. More details and discussions on capacitive water deionization and electrochemical flow electrodes can be found in recent reviews.^{104,106,108,111–113}

Research on electrochemical flow capacitors and water capacitive deionization has demonstrated the applicability of solid suspensions for capacitive and electrochemical devices. Recently, progress has also been made in using solid suspensions in RFBs—where the move to solid particles provides a route to increase the battery energy density by overcoming the solubility limitation of active species. Many other considerations will be important for these new types of RFBs beyond electrolyte energy density, including but not limited to operating efficiency, cost, safety, and reliability.^22,114 The operating efficiency is hindered by the pressure drop of the flowing liquid. Depending on the cell design, the pressure drop could be due to pumping liquid through a porous and tortuous electrode and/or due to high viscosity of the carrier fluid. In either case, significant energy is needed to keep the fluid moving and overcome the pressure drop. Efforts to address this challenge will be discussed in more detail in Secs. II, IV, and V. The cost of RFBs includes material cost, equipment cost, transport cost, maintenance cost, etc., and is an important factor to commercialization. A number of researchers are analyzing the cost metrics related to RFBs.^{18,26,115–117} For example, ion-selective membranes are one of the primary contributors to material costs,¹¹⁶ and hence, research has been reported toward membrane-free RFBs.^118,119 Addressing safety concerns is very important and necessary for wide scale adoption. Conventional RFBs use transition metals as active species, which in some cases can be expensive and/or have environmental concerns. Thus, research has pursued RFBs with abundant and nontoxic active materials.^{18,67,77,80,120} For example, Lin et al. proposed an alkaline quinone RFB without transitional metals.⁷⁷ In addition, reliability and equipment life are also important for a long-term stationary battery application, and these topics have been addressed by many researchers as well.^18,70,121

This paper contains a comprehensive review of the growing research area of using solid electroactive materials in RFBs as a subset of the RFB field, as well as some perspectives for future research. This review contains five additional sections. Section II introduces different reported design architectures for RFBs with solid active materials. Section III discusses active material chemistry. Section IV introduces the flowing methods used in the literature. Section V will summarize characterization methods that have been applied to these batteries, and finally, Sec. VI discusses perspectives on the development of these technologies.

Due to the complexity of dealing with the two-phase system of a suspension, RFBs with solid electroactive materials have had a number of different innovative engineering designs to enable their characterization and operation. We segment these designs into four main types dependent on the flowing conditions used and the role of carbon in the electrolyte. The mode of flowing the material through the system is a very important consideration for these systems, in particular, because the viscosities in some cases become very high at increased particle loading. Carbon, with its relatively high electronic conductivity and low density, is a key component in many electrochemical devices and plays various roles in the RFBs described below.^91,122–124 We categorize the designs for RFBs with solid electroactive dispersions as follows: type I—flowing carbon as the electrochemical reaction electrodes, type II—flowing solid active materials within a carbon conducting network, type III—flowing active material particles colliding on current collectors without carbon, and type IV—soluble redox mediators for power with solid active materials within tanks for energy. Each one of these systems is suited to different materials and has different advantages and disadvantages. The system designs will first be described in greater detail before discussion of the relevant chemistry.

Carbon is a common electrode material in conventional RFBs, for example, in the form of activated carbon foams, glassy carbon sheets, and carbon fiber cloth.¹²² Porous carbon is preferred in order to increase the total surface area, which increases the net rate of electrochemical reactions achievable per projected area of the electrochemical cell with the same cell volume. Thus, higher surface area porous electrodes result in higher area specific current density and increased total current and power from the RFB power module. Design of the porous electrodes must also take into consideration the pressure-drop of the electrolyte being forced through the electrode, which is relatively high and requires significant pumping power. For a solid electroactive material RFB type I design, one way to conceptualize this type of device is that the porous carbon electrode of a conventional RFB has been broken apart into micro/nanosized carbon particles flowing while dispersed in the energy-containing electrolyte. There would result in significantly less pressure-drop across the electrochemical reaction cells (the current collectors would be planar instead of porous carbon). As illustrated in Fig. 5(a), these flowing carbon particles form percolated aggregates and electrochemical reactions occur on the surface of carbon materials while in contact with the current collector—either directly or through particle–particle connections to the current collector surface. This type of design has been demonstrated for multiple battery systems including lithium–polysulfide (Li-PS),^125,126 lithium–air (O₂),¹²⁷ and metal ions in aqueous solvents.^60,68

One example of a type I design was a Li-PS half-cell system reported by Fan et al., where nanocarbon material was dispersed within a Li-PS electrolyte before pumping through an electrochemical reaction cell.¹²⁵ The counter electrode was Li metal, and a microporous separator film was adopted to separate the catholyte and the Li metal, with LiNO₃ added to the electrolyte to passivate the Li surface.^{118,119,125,128} In a conventional RFB design with Li-PS and a porous carbon electrode, the Li-PS electrochemical reaction window was limited to the range including only the soluble species between Li₂S₈ and Li₂S₄ and did not proceed to insoluble species such as elemental sulfur or Li₂S, restricting the total achievable capacity and energy in the cell.^119,128 However, with the flowing nanocarbon dispersed in the electrolyte, the electrochemical reaction window was expanded to include the solid species, which increased the energy density of the electrolyte significantly. The precipitation of the solid sulfur species onto the nanocarbon surfaces within the dispersion did not impact the flow of the catholyte and kept the solid species confined within the already solid and electronically conductive particles dispersed in the electrolyte. As shown in Fig. 5(b) (adapted from Fan et al.¹²⁵), four- to fivefold higher reversible capacity was achieved for the dispersed nanocarbon system over the conventional static porous carbon electrode.¹²⁵ In this design, the overpotential was also lower than conventional electrodes because the charge transfer resistance was lower even at a lower loading of carbon (1.5 vol. % nanocarbon suspension over 6 vol. % stationary carbon fiber within the electrochemical cell channel). This reduced resistance was attributed to the unique size and surface chemistry of the carbon nanoparticles compared to the microsized porous carbon fiber. The overall pumping cost was estimated to be lower for the dispersed nanocarbon relative to the conventional porous electrode because of the much lower pressure drop for the fluid flowing through the channel in the absence of the tortuous carbon fiber, though the viscosity of the catholyte with the carbon dispersion was higher than the carbon-free catholyte.

In this particular case, the addition of solid electrically conducting particles facilitated more complete oxidation and reduction using Li-PS chemistry due to the ability to go all the way to the solid products efficiently. This enabled higher total energy density in the electrolyte. For the type I system, the added particles to the electrolyte are not themselves the electroactive material undergoing redox chemistry, but rather facilitate and improve the rate or utilization of the redox chemistry that occurs within the electrolyte. The next type of RFB with particles dispersed in the electrolyte will involve the addition of electrochemically active particles that undergo redox reactions in the electrolyte. Some further discussion of optimization efforts relevant to type I systems can be found in Sec. III.

In the type I design, the carbon or other conductive additive needs to be small (nanosize) to provide a large surface area for electrochemical reactions with a relatively low loading of particles into the electrolyte. In contrast, the type II design does not have electrochemical reactions occur on the carbon particles, and in this case, the carbon functions as the conducting network for electrons between the electroactive solid particles and the current collector. In a type II architecture shown in Fig. 6(a), both the active material particles and carbon additives are dispersed in the liquid electrolyte before flowing into the electrochemical reaction cells. This has previously been referred to as a semisolid flow cell (SSFC), as proposed by Duduta et al.²⁷ This design has been exploited for Li-ion active materials as well as sodium-ion (Na-ion) materials in both organic and aqueous fluids,^76,129–133 and some material choices will be discussed in more detail in Sec. III. The type I design is suitable for soluble or liquid electrochemically active materials, including those that form solid deposits such as Li-PS, while the type II design is desirable for systems where the electroactive material is always in the solid phase as particles dispersed in the electrolyte.

In the demonstration of the SSFC by Duduta et al., half-cells were demonstrated with Li-ion active materials in the slurry dispersion versus a Li metal anode, and also, a full cell with both a slurry anode and a slurry cathode was reported.²⁷ The electrochemical charge and discharge was completed using both continuous and intermittent flow modes.²⁷ The cathode active material used was LCO which was dispersed into a suspension of 22.4 vol. % LCO with 0.7 vol. % Ketjen Black (KB) carbon conductive additive, with the balance of the dispersion being organic Li-ion electrolyte. During the continuous flow mode experiments the slurry had 127 mA h g⁻¹_LCO reversible capacity; however, there was significant energy lost to pumping under continuous flow mode due to the high viscosity of the dispersion. The pumping power dissipation was calculated to be 44.6% of the total discharge power at a flow rate of 15 ml min⁻¹. To optimize the operating efficiency, an intermittent flow mode was also demonstrated which reduced the reported average pumping requirements down to 0.6% of the discharge power. During intermittent flow mode, the process includes (1) pumping the suspension into the electrochemical cell, (2) fully charging and/or discharging the materials in the cell, and (3) pumping the suspensions out of the cell and replacing with fresh electrode materials for the next charge/discharge. The pumps run intermittently to save pumping energy and hence increase the operating efficiency. As shown in Fig. 6(b) (adapted from Duduta et al.²⁷), a full cell under intermittent mode was demonstrated with 20 vol. % LCO and 1.5 vol. % KB suspension as the catholyte and 10 vol. % Li₄Ti₅O₁₂ (LTO) with 2 vol. % KB suspension as the anolyte, and both suspensions contained the same organic Li-ion electrolyte for the balance of the volume.²⁷ Although the coulombic efficiencies of the first two iterations for the intermittent flow (73% and 80%) were lower than the first two charge/discharge cycles of static suspensions (98% and 88%), there are many possible optimization opportunities, for example, the channel geometry and flow rate. The company 24 M was originally founded based on this technology, though the company subsequently pivoted toward using the slurries for static semisolid batteries due to more favorable results from economic projections.¹³⁴ 

The main advantage of this type II design is unlocking the opportunity to use solid active materials with high energy density in RFBs, in this case Li-ion battery active materials. Moving to solid active materials overcomes the solubility limitation on capacity per volume of conventional RFBs, and the use of Li-ion materials and organic electrolytes expands the possible voltage range to that of Li-ion batteries, which reaches >4 V for existing commercial systems and is even higher for next-generation materials.^135–137 For the system reported by Duduta et al., an optimized system with Li intercalation active materials is expected to achieve a theoretical energy density of 300–500 W h L⁻¹, compared with a theoretical value of 40 W h L⁻¹ for a 2 M aqueous vanadium RFB.²⁷ The type II design can be further adapted to many other battery materials with high energy densities. Another attractive feature of this system is that the solid active materials and conductive carbon within the two slurries can be kept within separate channels by size exclusion. In principle, this enables the use of porous separators, which are generally less expensive and when paired with liquid electrolytes have higher ionic conductivities than ion-conducting polymeric or solid-state membranes. There are also a few challenges for researchers with this system. The first major drawback is the pumping energy loss due to the high viscosity. For example, the viscosities reported for the slurries used by Duduta et al. were greater than 2 Pa s at a shear rate of 10 s⁻¹. The viscosities need to be reduced to lower the pumping energy, which results in net improvements in the total energy efficiency. Pumping high viscosity slurries also requires larger duty and more complex or expensive pumps. Using intermittent flow mode was one approach described above aimed at minimizing pumping losses, but there are challenges with this mode of operation, which will be discussed in more detail in Sec. IV A. Another challenge is that not all of the active material particles are in sufficient contact with the conductive carbon network in this mechanically mixed suspension, and as a consequence, there are losses in capacity utilization and coulombic efficiency.¹²⁶ Different carbon architectures have been reported to impact the electrochemical performance; thus, carbon morphology and loading must be carefully controlled as well as the aggregation of the carbon particles.^125,138,139 Coatings on the carbon particles and suspension additives may provide opportunities to reduce suspension viscosity and/or reduce the total amount of carbon that must be dispersed in the slurry.^140–143 In addition, there is a tradeoff between the power density and the coulombic efficiency depending on the slurry conductivity. A high conductivity is needed to provide high power density with reasonable overpotential; however, shunt currents will be much more significant with a highly conductive carbon network, resulting in a loss of coulombic efficiency for a system with cells in series.^{121,144–146} This conductivity tradeoff will be a more significant challenge for any flow battery with a conductive material network (e.g., types I and II). Exploration of flow profile control within the channel and the balance between active material and carbon material morphology and loading to improve the performance and efficiency of type II systems could lead to major improvements for these relatively unusual battery cells.

The addition of carbon to the electrolyte results in a percolated network which has a significant impact on the suspension viscosity. For example, in the type II SSFC system described earlier with 22.4 vol. % LCO suspension, the addition of only 0.6 vol. % carbon additive resulted in a >10-fold increase in the viscosity.²⁷ Therefore, eliminating carbon from the electrolyte, in principle, provides a route to significantly reduce pumping requirements while maintaining the high energy density that results from using high energy density materials and solid electroactive particles. The removal of carbon from the system results in the type III system [schematic in Fig. 7(a)], which has two key features that differentiate it from the type II system. First, due to the lack of carbon or other conductive additives, the type III system has reduced viscosity while still relying on solid electroactive particles for redox reactions. The lack of carbon also reduces the mass and volume of components in the electrolyte that do not contribute to the cell energy density. Second, without the percolating network, the electrochemical reaction no longer occurs throughout the cell channel facilitated by the electronic conductivity of the carbon, and thus, only particles in contact with the current collector (directly or indirectly through other particles in, for example, a particle aggregate) participate in electrochemical reactions at any given time. In operation, type III systems rely on the collisions of active material particles with the current collector and have very low electrochemical activity in the absence of flow.^147,148 This type of carbon-free active material suspension as an energy containing fluid has been explored for both Li-ion materials and polymer suspension based cells,^147–153 and the chemistry of the active materials will be discussed in more detail in Sec. III.

Both half- and full cell demonstration of LCO and LTO suspensions with type III design have been reported by our group.^147,148 As shown in Fig. 7(b) (adapted from Qi et al.¹⁴⁷), 10 vol. % carbon-free LTO suspension was reversibly galvanostatically reduced and oxidized, demonstrating the use of this design with Li-ion chemistries. However, the cell was only discharged for 10 min instead of fully discharging and the coulombic efficiency was only 38%. Both of these limitations were due to low power output. The electrochemical reactions occurred during the collision of active material particles with the current collector, and thus, the ambipolar diffusion of electrons and Li ions limits the reaction rate.^147,154 Active materials with high electronic and ionic conductivities are particularly desired for this conductive additive-free design. Montoto et al. reported a solid suspension of functionalized cross-linked poly(vinylbenzyl chloride) (xPVBC) and multiple full cycles of a 0.74 wt. % suspension were demonstrated with high coulombic efficiencies, which will be discussed in more detail in Sec. III B.¹⁵³ This design has a wide variety of potential chemistry options to choose from and, similar to type II, it also holds the advantage of using cheaper porous separators due to the advantage of preventing particle crossover via size-exclusion. An additional feature of this design is its simplicity, where the influences of other additives can be excluded. This simplicity provides the opportunity to understand the properties of the active materials in the flow cell without the complication of accounting for the electrochemical properties and interactions with other materials in the electrolyte. An analytical method based on this concept was reported to correlate to the rate capabilities of active materials, with the advantage being that the flow method does not require electrode fabrication or full cell cycling to characterize the material.¹⁵⁵ However, more effort is needed to understand the material properties and flow profiles best suited to maximize the energy efficiency, energy density, and power output for the type III design. One possible direction would include surface modifications on the active material particles. Such modification would be expected to be beneficial to (1) facilitate transport of electrons and ions; and (2) further reduce the suspension viscosity to increase the operating efficiency.¹⁴⁰ Another potential route to improve the system may include designing an innovative flow channel which maximizes the active surface area while minimizing pressure drop. Such a modification to the current collector is important because type III has to be operated such that the flow facilitates contact between active material particles and the current collector.

In addition to research on battery geometry design and analysis, studies of battery material particles in isolation can also provide insights to understand and improve the performance of type III batteries.^156,157 Over the last two decades, a number of researchers have more generally explored particle collisions on electrode interfaces for a number of different systems and applications.¹⁵⁸ These studies can largely be divided into three groups of activities: (1) inert particles colliding with an electrode that block electrochemical reactions, (2) electrocatalytic particles that facilitate electrochemical reactions only during collisions with an electrode, and (3) direct oxidation/reduction of electrochemically active particles that only occurs upon collision with the electrode. In the first case, inert and electrically insulating particles partially block electrochemical reactions on the surface of the electrochemically active electrode upon collision, creating transient current changes by reducing the active area of the electrode (and in some cases, the adsorption was permanent).^159–161 In the second scenario, electrocatalytic particles are suspended in the electrolyte and only facilitate electrochemical reactions when a sufficient potential driving force is available, which is only provided when Brownian motion causes the particles to collide with the electrode surface resulting in a collision-dependent current response.^162,163 For this electrocatalytic system, research has been performed to specifically understand the impact of surface chemistry,^164,165 electrode size,¹⁶⁵ electrode material, and mass transfer effects.¹⁶⁶ One study of similar phenomena reported measurements of single particle collisions for a photovoltaic system that took advantage of semiconducting nanoparticle collisions to interpret spikes of photoelectrochemical current as individual particle events.¹⁶⁷ For the third case of direct oxidation/reduction of nanoparticles, the particles are electrochemically active for oxidation or reduction upon collision with the otherwise inert electrodes.¹⁵⁹ The measured current was quantitatively correlated to particle size, providing the capability to extract particle size information from electrochemical measurements.^168–171 More detailed discussions on the background and developments related to particle collisions on electrodes can be found in these recent reviews.^158,168

RFBs are known for decoupling of the energy storage and power output components, providing the flexibility for customized designs for different applications.^57,58 In a typical RFB design and all designs introduced above, the energy storage is provided by redox-active materials which are flowing through the electrochemical reaction cell for the storage or delivery of power. The function of energy storage and power delivery is provided by the same materials, and thus, even though the energy and power components are decoupled, the performance of the electrochemical cell is still coupled with the properties and state of the energy storage fluid. Early in 2006, Wang et al. proposed the concept of using targeted redox shuttles to improve the electrochemical performance of poorly conducting Li-ion battery materials to potentially eliminate the use of carbon additives.¹⁷² Soluble redox mediators were introduced into the electrolyte within the cell that underwent electrochemical reactions, and then, the lithiation/delithiation of Li-ion active materials proceeded via chemical reduction/oxidation by the mediators.¹⁷² Therefore, the energy storage and power output for the cell were provided by two different sets of materials; the energy storage was provided by the solid active material while the dissolved mediators actually underwent the electrochemical reactions during power delivery or charging. Integrating this concept with RFBs enabled further increases in capacity and flexibility. Wang's group proposed a design with flowing dissolved redox mediators to provide power output and static high energy density solid materials in separate tanks for energy storage.^24,173–175 As shown in Fig. 8(a), solid active materials are stored in the energy tanks and only the dissolved mediators undergo electrochemical reactions in the flow channel. For example, after being discharged in the electrochemical reaction cell, the mediators get “recharged” through chemical redox reactions with the solid energy storage materials in the tanks until the solid materials have been fully chemically discharged. The charging process is then run in reverse. This design has been applied mainly to Li-ion active materials.^{24,172–174,176,177}

In the report of Jia et al. of a full cell demonstration, LiFePO₄ (LFP) and TiO₂ were used as the cathode and anode solid energy storage materials, respectively.^24,173,174 Two pairs of redox mediators were used, dibromoferrocene (FcBr₂) and ferrocene (Fc) for the catholyte, and cobaltocene [Co(Cp)₂] and bis(pentamethylcyclopentadienyl)cobalt [Co(Cp*)₂] for the anolyte. A fully charged cathode includes static FePO₄ in the cathodic tank and dissolved FcBr₂⁺ and Fc⁺ flowing through the reaction channel and energy storage tank. As shown in Fig. 8(b) (adapted from Jia et al.²⁴), the FcBr₂ redox potential is above LFP and Fc is below LFP. When the cell is being discharged, FcBr₂⁺ will be first reduced to FcBr₂. Then, the only species participating electrochemically in the flowing liquid in the electrochemical reaction cell is Fc⁺, which will be reduced to Fc thereafter. Fc will then be oxidized back to Fc⁺ by FePO₄ through chemical oxidation when it comes into contact with the surface of the solid FePO₄ particles in the tank. This process continues until all FePO₄ is reduced to LFP. The residual Fc⁺ continues to be reduced to Fc until depletion, and the cell was then considered fully discharged. Corresponding reactions occur on the anode side, and collectively, the full cell is discharged in the order of reactions (1)–(6) [reactions in Fig. 8(b)]. The charging process proceeds in the reverse sequence from reactions (6) to (1). Galvanostatic cycling at different current densities were demonstrated and are shown in Fig. 8(c) (adapted from Jia et al.²⁴). The discharge curves showed three plateau regions, with the highest potential plateau corresponding to reaction (1) and progressively the lowest attributed to reaction (5). This promising demonstration provides the opportunity to further increase the capacity and energy density by storing more static LFP/TiO₂ materials in the tanks. Increases in energy density by adding more solid particles to the tank does not change the electrolyte viscosity as it would in types II and III systems. Therefore, this design has a high theoretical energy density, ∼500 W h L⁻¹ for TiO₂/LFP assuming 50% porosity is practically achievable.²⁴ The amount of redox shuttles in this system is, in principle, much less than that required for other system designs and would result in significant volume reductions relative to an organic RFB that did not contain solid active material particles. This reduction in the amount of redox shuttles and electrolyte required provides a possibility for cost reductions. There are also safety advantages provided by the reduction in volume of flammable electrolyte required. In addition, the pumping energy loss is expected to be low because only low concentration mediator solutions with relatively low viscosities need to be circulated.

However, the type IV system still has some opportunities for innovation. At high loadings of solids in the energy storage tanks which are desired to increase system energy density, the pressure drop arising from the particle bed is likely to be high. The pressure drop associated with the active material packed in the tank may even exceed the pressure drop from passing through porous carbon in the flow channel, which is also present in this design incorporating mediator redox couples. To overcome this challenge, the particle bed would need to be designed to minimize the pressure drop, possibly at the expense of other metrics such as active material loading. A second challenge comes from the separator. The size-exclusion benefit of solid active materials in types II and III is not applicable in this design because the redox active mediators are dissolved in the electrolyte and must be kept segregated via a more resistive membrane or thin film. Therefore, Li-ion conductive membranes compatible with organic solvents are needed. Although this is a greater issue relative to other RFBs with solid active materials that can rely on size exclusion at the separator, this challenge is shared by all other RFBs that rely on Li ions for ionic conductivity and have soluble redox shuttles. Jia et al. reported a custom Nafion/polyvinylidene difluoride (PVDF) composite membrane for this application after finding ceramic and commercial Nafion membranes to have performance limitations in their electrolyte.²⁴ Membrane cost will be a major factor to consider for Li-ion conductive membranes. A low-cost, stable, low resistance and low mediator crossover Li-ion conductive membrane will be needed to drive this design forward. The third challenge is the relatively low voltage efficiency due to the voltage difference needed between the mediators and the solid active materials to drive the additional chemical redox process. There was a 0.81 V difference between the expected primary charge reaction [2.11 V, reaction (1) in Fig. 8(b)] and discharge reaction [1.30 V, reaction (5) in Fig. 8(b)]. This significant voltage difference reduces the cycling energy efficiency. Future research will be needed to find more active material–mediator pairs that are effective with a lower potential difference, although voltaic efficiency loss due to the additional chemical oxidation/reduction processes between the redox shuttles and solid particles in this system is unique for RFBs and cannot be eliminated.

In batteries, the active materials store the chemical energy and participate in redox reactions. In RFBs, the active materials are typically dissolved or dispersed in electrolytes, stored in energy storage tanks, and pumped through electrochemical reaction cells during charge/discharge (illustrated in Fig. 2). There have been many types of active materials developed since the RFB concept was introduced.^21,22 The first group of active materials was dissolved single-elements in aqueous electrolytes, most commonly transition metal ions (e.g., Fe, V) and halogens (e.g., Br, Cl).¹⁷⁸ Some of the redox couple systems that have reached the greatest maturity are iron-chromium, soluble metal-bromine, iron-vanadium, and all-vanadium.¹⁷⁹ A more detailed discussion can be found in a recent review.²² The second group of compounds are ligand-modified ions, which may be dissolved in aqueous or organic solvents.^{22,65,178,180–182} For example, additional capacity has been reported from the ligands of a vanadium complex.¹⁸⁰ More complex organic redox compounds can be modified by chemical functionalization to optimize both the redox potential and solubility.^19,178 Beyond soluble compounds in RFBs, more recently research activity has increased toward solid active material tailored for RFBs.

Pb-acid batteries are one of the most widely known types of cells in part due to reliable performance and low cost.^183,184 A conventional Pb-acid battery uses two electrochemical redox couples, Pb/Pb²⁺ as the anode and PbO₂/Pb²⁺ as the cathode. The electrochemical reactions are shown in Eqs. (3) and (4) (note: precipitation of Pb²⁺, as is typical in Pb-acid cells, has been omitted). In the discharged state, both anode and cathode active materials have Pb in the Pb²⁺ state, typically as a PbSO₄ precipitate. However, if the Pb is under conditions where it remains as a soluble Pb²⁺ species, the possibility opens up of adopting Pb to a RFB system. For example, a RFB using Pb/Pb²⁺ as the anode and Fe²⁺/Fe³⁺ cathode was reported.¹⁸⁵ More interestingly, Pletcher et al. reported a Pb-acid based RFB using a single flow channel without separators.¹⁸⁶ Electrolyte with dissolved Pb²⁺ pumped through the flow channel, and Pb and PbO₂ are electrochemically deposited on the anode and cathode, respectively, while being charged. Current densities in the range of 10–60 mA cm⁻² were demonstrated and >85% coulombic efficiency and ∼65% energy efficiency were reported.¹⁸⁶ The energy efficiency was improved to 79% with a cycle life of >2000 cycles after further optimization of the electrolyte, deposition conditions, and current collectors.^187–190 However, there are a few challenges to apply this system widely. First, the energy density is highly limited by the solubility of Pb²⁺, which may undergo precipitation at both the anode and cathode. Therefore, the operating efficiency is a challenge due to low energy density, low coulombic efficiency, low energy efficiency, and pumping energy dissipation. Second, the concentration overpotential is expected to be high due to a wide gap between the electrodes at low state of charge (SOC) after dissolution of solid material during discharge, although it is relatively low compared with conventional Pb-acid battery where often a thick layer of liquid sits in between two electrodes at all SOC. This challenge could potentially be overcome by promoting more compact surface deposition and/or engineering adjustable current collectors which minimize electrode separation during charge/discharge. Third, the benefit over conventional Pb-acid is not obvious because the energy density of the chemistry has not changed, but there is additional cost and complexity of RFB components including the pumps and tanks. The flexibility of decoupling power and energy is reduced due to high concentration overpotential at high overall capacity, in contrast to a conventional RFB where the electrodes have similar efficiency across a wide range of total capacity for the external tanks. In addition, the toxicity of Pb needs to be considered, particularly in the target large scale applications of RFBs. Dong et al. have noted, though, that recycled Pb could be used as the active materials for Pb-acid RFBs [standard hydrogen electrode (SHE)]¹⁹¹ 

A similar design to the Pb-acid RFB has been reported using Zn/Ni chemistry [redox reactions shown in Eqs. (5) and (6)], and for Zn/Ni, the reported coulombic efficiency was 98% and energy efficiency was 88%, higher than those of Pb-acid.^192,193 Optimization of electrolytes, Zn morphology, and current collectors were done to improve the overall performance.^193–196 Zn is also used as an electrode in other RFBs and has been paired with bromine,¹⁹⁷ polyhalides,¹⁹⁸ cerium,^199,200 and polymer suspensions.¹⁴⁹ All-copper RFB based on reactions shown in Eqs. (7) and (8) was reported using aqueous solution of Cu⁺.²⁰¹ The achieved energy density (20 W h L⁻¹), cell voltage (0.6 V), and current density were low relative to other RFBs.^201,202 Cu has also been paired with PbO₂/PbSO₄ electrodes, which increases the cell voltage to 1.29 V with an energy efficiency of 83%.²⁰³ A similar design was applied to Cd/Cd²⁺ couple with a H₂SO₄–(NH₄)₂SO₄-CdSO₄ intermixture electrolyte as a RFB, achieving ∼1.0 V discharge voltage and 82% energy efficiency.²⁰⁴ MnO₂/Mn²⁺ is another commonly used solid cathode²⁰⁵ and has been paired with polymer suspension electrodes as well, to be discussed in more detail later.¹⁵¹ MnO₂ has also been reported as a suspension electrode for flow capacitors in the form of percolating network; thus, this material has shown the versatility to have been successfully demonstrated as a RFB active material in multiple different systems and cell designs⁴² 

All the materials just discussed share the same challenges as the Pb-acid RFB. One possible approach to improve performance is to apply type I design discussed in Sec. II A. This design would be beneficial by keeping the separation between electrodes consistent during charge/discharge, which would reduce IR drop across the electrolyte. In type I design, the solid material is deposited on or dissolved from the surface of the percolated conductive additive material and flows within the electrochemical reaction cell and the storage tank, instead of building up as thick layers on the current collectors. Recently, Mubeen et al. demonstrated such a RFB.⁶⁸ For example, a Zn-Cu active solid suspension battery has the electrochemical redox reactions in Eq. (9). In a fully charged state, Zn is deposited on carbon particle surfaces, and Cu²⁺ is dissolved in the electrolyte, and both electrolytes contain flowing carbon suspensions. During discharge, Zn dissolves, and Cu is electrochemically deposited onto the surface of carbon. The reaction is reversed during charging. A discharge voltage of 0.97 V was achieved with an overall energy efficiency of 70% at a constant current density of 5 mA cm⁻² (Ref. 68)

The all-iron RFB is another chemistry that relies on solid transition metal deposition. The redox reactions are shown in Eqs. (10) and (11). A conventional RFB with all-iron chemistry uses a Fe²⁺/Fe³⁺ solution as the cathode flowing through a porous electrode and Fe/Fe²⁺ (Fe plating/stripping) as the anode.²⁰⁶ Petek et al. demonstrated an all-iron RFB with type I design and carbon particles added to the electrolytes.⁶⁰ Current density as high as 75 mA cm⁻² was demonstrated, although the charge/discharge voltage efficiency was just over 50%.⁶⁰ Applying type I design to this transition metal chemistry provided the possibility to expand capacity significantly. All these cells share the common challenges of conventional RFBs including relatively low voltage, energy density, and round trip efficiency. On the other hand, the cost of many of these metal materials are relatively low due to their abundance.^60,68 These technologies are also competitive in applications where space and footprint limitations are less of a concern.

Organic molecules have also been used as the active materials in both aqueous and organic RFBs. Some of these materials may offer cost and toxicity advantages over dissolved transition metals, and organic molecules provide a wide range of design flexibility and desirable attributes with regard to redox potentials and other physicochemical properties.^22,65,73,79 For example, an acidic solution of 9,10-anthraquinone-2,7-disulphonic acid was used to undergo fast two-electron two proton reduction.⁸⁰ The quinone/hydroquinone couple was paired with a Br₂/Br⁻ redox couple, yielding >99% capacity retention per cycle.⁸⁰ The quinone couple was demonstrated in alkaline solutions to improve safety and inhibit corrosion.⁷⁷ More detailed discussion on organic active materials and redox polymers can be found in recent reviews.^22,65,73,207 One challenge facing dissolved organic molecules as active species is membrane crossover, which reduces cell capacity and operating lifetime. One option to mitigate this issue is to design new membrane materials; however, Montoto et al. proposed instead using redox active colloids (RAC) as the active materials to address this issue.¹⁵³ Solid xPVBC particles were synthesized through emulsion polymerization as backbones for redox couples, in this particular case ethyl viologen and (dimethylaminomethyl) ferrocene. The redox couples were then grafted onto xPVBC through ion exchange by replacing –Cl on xPVBC with the functional redox species to form the final RAC particles. The functionalization efficiency was nearly 100%, indicating a high loading of redox species on the RAC polymer particles. These RACs had good size and shape stability during charge/discharge, chemical stability of the redox groups, and minimal membrane crossover.¹⁵³ A prototype RFB using low concentration RAC and a commercially available porous separator with a type III flow cell design was demonstrated. The overall performance was comparable to typical RFBs based on dissolved organic active materials, showing a coulombic efficiency of ∼94% at 43 μA cm⁻², and both energy and voltage efficiencies were above 90%.¹⁵³ 

Although RACs are a very promising technique, there are still a few challenges to be addressed. First, the capacity utilization was low in the initial demonstration, achieving 21% of the theoretical capacity. Notable causes including low loading and sedimentation at the counter electrode were mentioned by the authors. More research is needed to engineer the RAC design (e.g., particle size, morphology, flow pattern, and loading) and/or explore new backbone materials. Second, full charge/discharge cycling of a RAC fluid with high loading to increase energy and power density will be needed, and parameters will need to be explored at the higher loadings including particle contact with the current collector, electrolyte tortuosity, and rheological properties. Successfully increasing the loading may also improve capacity utilization, as noted in the report.¹⁵³ A third challenge is the low energy density relative to conventional RFBs, which will likely be an issue even if higher particle loading in the electrolyte are achieved. Assuming a high concentration of 40 wt. % and an average discharge voltage of 0.85 V, the energy density would be approximately 12.5 W h L⁻¹, which is significantly lower than a typical conventional all vanadium RFB (25 W h L⁻¹).^19,22 We note this energy density is limited not just by the mass loading, but also the cell voltage, which could potentially be improved by grafting other redox couples with different redox potentials. Increasing the cell voltage should be achievable because many organic redox couples are available and also because the system was designed for organic electrolytes which have stability windows exceeding 0.85 V. With regard to volumetric and gravimetric capacity limitations, future research could be directed toward other backbone materials with lower molecular weights and higher densities while maintaining or improving electrochemical properties. The key advantage of the RAC design is that it provided an opportunity to bypass the solubility limitation of the active species by grafting them onto backbone materials that were already solid particles. This feature is exciting because it opens a variety of materials that could potentially increase the energy density of RFBs with organic redox couples. The fourth challenge is the high viscosity at high RAC loadings. A 40 wt. % RAC showed >10 Pa s viscosity at 100 s⁻¹ shear rate.¹⁵³ This is significantly higher than many other suspension-based systems.^{27,147,148,208} The high viscosity may result in low operating efficiencies as discussed earlier in Sec. I B. Further research efforts will likely explore lowering the viscosity by optimizing the RAC surface and electrolyte. Overall, the RAC system is promising due to the use of relatively low cost materials and their high initial capacity retention and lifetime estimates, though more research needs to address energy density, power density, and viscosity challenges.

Another polymer active material demonstrated in RFBs is polyaniline (PANI), where the redox couple involves Cl⁻ addition/subtraction with the PANI backbone [reactions in Eq. (12)]. PANI has attracted attention for rechargeable batteries due to its high electronic conductivity, low cost, environmental stability, and good redox reversibility.^149,209,210 Zn–PANI batteries offer high capacity and discharge voltages close to 1.2 V [redox reactions in Eqs. (12) and (13)].^{149,152,209,211} Different methods have been reported to synthesize well controlled PANI films/particles.^210,212,213 Zhao et al. demonstrated a RFB based on PANI microparticle suspensions in aqueous electrolytes using a half-cell type III design with a Zn counter electrode.¹⁴⁹ The results were promising, particularly relative to Zn-PANI thin film batteries, with a discharge capacity of 115.2 mA h g⁻¹_PANI, coulombic efficiency of 97% and minimal capacity decay at a current density of 20 mA cm⁻².¹⁴⁹ The high current density enables the possibilities of high power applications and fast cell charging, getting closer to the current densities achievable with RFBs based on dissolved transition metals.¹⁴⁹ In addition, the suspension viscosities were lower than Li-ion solid suspensions. This cell was further optimized by doping Ag on PANI particles, achieving high discharge capacity from oxidation of PANI particles.¹⁵² PANI suspension redox has also been paired with PbO₂ electrodes, increasing the cell voltage.¹⁵⁰ Another microsized polymer particle suspension RFB cathode was reported by the same group.¹⁵¹ This system included a polypyrrole (PPy) microparticle suspension paired with manganese dioxide for an average discharge voltage of ∼0.95 V [anode redox reaction shown in Eq. (14)].¹⁵¹ This cell also showed a high capacity retention (97.2% after 90 cycles) and a stable coulombic efficiency of 92.1%.¹⁵¹ These two micropolymer suspensions share similar promising advantages such as high capacity retentions and high current densities; however, they share similar challenges as well. First, these are aqueous systems, and thus, the cell voltage is low relative to organic electrolyte chemistry. Incorporating organic solvents may be possible, but particle stability and solubility in the organic electrolyte would need careful consideration. Second, the voltage efficiency is relatively low; researchers are currently addressing this challenge with new conducting polymers, modified polymer synthesis methods, and optimizing the supporting electrolyte.^{209,210,213–215} Potential optimization on the counter electrode by adopting a type I design may also be helpful to decrease the overpotential and to take advantage of the power-energy decoupling feature of RFBs. Oh et al. reported an all-organic RFB with polythiophene microparticle suspensions as both anode and cathode showing redox activities through two opposite electrochemical process, n- and p-doping.²¹⁶ Although a relatively high discharge voltage of ∼2.4 V was achieved, the current density and efficiency were limited even with a high loading of KB to facilitate electron transfer.²¹⁶ Winsberg et al. demonstrated a RFB using 2,2,6,6-tetramethylpiperidinyl-N-oxyl containing methacrylate/styrene block copolymers micelles as cathode paired with Zn/Zn²⁺ anode achieving ∼1.1 V discharge voltage with high coulombic efficiencies (99.8%) but low energy density (1.6 W h L⁻¹) with limited current densities (fast capacity decline with increasing current densities with maximum at only 0.2 mA cm⁻²).²¹⁷ Overall, these redox active polymer particles and grafted redox couples on inert backbone particles are promising approaches to overcome the solubility limitation of conventional RFBs, and more innovations on material selection and engineering are expected to increase the performance significantly

RFBs using Li metal or Li-ion battery redox chemistry stand out with regard to voltages and energy densities. One option is to use soluble cathode active redox couples dissolved in aqueous or organic solvents to form the flowing catholyte, and metallic Li or soluble anode active material as the anolyte separated by suitable ion conducting ceramics or polymers that serve the dual role of separator and electrolyte.⁶⁵ While Li metal anodes provide high cell voltages due to the anode potential, safety issues arise dependent on the ability of the electrolyte and separator to control the high reactivity of Li and the formation of Li dendrites that can cross to the cathode and short the cell.^{61,116,218,219} Another challenge of using soluble catholyte/anolyte species is that energy density is still limited by the solubility of the active species, just as is the case for conventional RFBs. More detailed discussion on this topic can be found in a recent review.⁶⁵ Given this limitation, it is not surprising that efforts have been made to incorporate solid Li-ion battery intercalation redox couples into RFBs in the past few years. As static battery cells, Li-ion batteries stand out among rechargeable battery materials with regards to energy density and cycling capacity retention.^34,35 In this section, both cathode and anode solid active materials that have been adopted in RFBs will be discussed.

Li-PS is a low voltage but high capacity battery chemistry that typically uses sulfur as the cathode active material and Li as the anode. The overall cathode redox reaction, assuming complete conversion to Li₂S, is shown in Eq. (15). The theoretical capacity of a Li-PS battery is 1672 mA h g⁻¹, significantly higher than transition metal-based commercial Li-ion cathode materials. However, this lithiation/delithiation is not a single step process; instead, Li₂S_x (2 < x ≤ 8) forms throughout the process. While some of the product/intermediate species are insoluble in the electrolyte, the Li-PS Li₂S_x (6 < x ≤ 8) are highly soluble. These dissolved PS have a shuttling effect, migrating to the counter electrode and reacting with Li resulting in capacity fade and low coulombic efficiency.²²⁰ Numerous research efforts have attempted to address the issue of PS solubility in static battery configurations;^221–227 however, soluble PS can be conveniently incorporated into a RFB system. Li-PS RFBs have been reported where the operating voltage maintains the sulfur species always as the soluble Li₂S₈ and Li₂S₄ PS species.^119,128,228 Functioning like a conventional RFB, this Li-PS RFB only operates within the voltage window of the soluble PSs, and thus the theoretical capacity drops to 418 mA h g⁻¹. The experimentally achieved energy density was still high, at 108 W h L⁻¹.¹¹⁹ Conversion all the way to solid species is required to fully take advantage of the high capacity of sulfur. Therefore, a type I RFB design using percolating carbon network as flowing electrodes to provide both electron conduction and a surface for solid deposition was reported,¹²⁵ as was discussed in detail earlier in Sec. II A. A four- to fivefold increase in capacity was achieved by expanding the redox window to the insoluble species.¹²⁵ Li-PS chemistry has attracted increased attention for RFBs in recent years, for example, with new separator design,²²⁹ applying type IV mediator design,¹⁷⁷ and innovations on flow design,²³⁰ which will be discussed in more details in Sec. IV B.

This promising chemistry, however, still faces a few technical challenges in RFBs. First, the PS shuttling effect remains a problem, causing self-discharge and capacity fade. The larger PSs migrate toward the anode, react with Li forming smaller PSs, and migrate back to the cathode to reform the larger PSs again.²²⁷ This is a shared problem with conventional Li-PS batteries, and extensive research has been focused on this topic. There are currently three main reported approaches to reduce this effect. (1) Immobilizing sulfur in sulfur-porous carbon composites or other composite materials.^221–226 Sulfur immobilization strategies have already been applied to Li-PS RFBs.^125,126 Although some Li-PS RFBs achieved very high capacity and good cycling performance, the S to C ratio in the cathode was very low, resulting in low energy density based on the total cathode mass or volume as pointed out by Zhang et al.^220,231 (2) Nonporous ion-selective membranes are needed to provide a physical barrier for PS to cross between the cathode and anode in the electrochemical cell. Porous polypropylene separators, which are commonly used in static Li-PS batteries, have limited capability in blocking PS migration. There are a few reports demonstrating good blocking of PS by using selective ion-conducting membranes, for example, lithiated Nafion membranes²³² and perfluorinated polymer membranes with lithium sulfonyl dicyanomethide functional groups.²³³ The membranes used in other Li RFBs are also good candidates for Li-PS RFBs, for example, Nafion/PVDF membranes used in mediator RFBs by Jia et al.^24,65 (3) Electrolytes have been designed to suppress PS dissolution, including the use of high Li salt concentration^234,235 and ionic liquids,^236,237 although these methods are expected to be less effective in RFB systems as these strategies significantly increase the viscosity. The second challenge for Li-PS RFB, for systems that choose Li metal anode, is the stability of Li metal. Metallic Li has been used as the anode in both conventional Li-PS batteries and RFBs to achieve high energy density. Li anodes are typically stabilized by a passive layer on the surface; however, the passive layer can be unstable and particularly in flowing systems the Li will be exposed to a relatively large amount of electrolyte and will face shear forces, which may impact the stability of the Li interface. Innovations on cell configuration and electrolyte design may help in this area.^238,239 More detailed discussions on Li-S chemistry, especially for static Li-S cells, can be found in these recent reviews^{32,65,220,240,241}

The most well-known Li-ion battery active cathode materials are based on transition metal oxides with high redox potentials, from 3.0 to approaching 5.0 V versus Li/Li⁺.²⁴² In addition to the specific examples in Eqs. (1) and (2), a more general redox reaction is shown in Eq. (16), where M refers to one or multiple transition metals, and O refers to oxygen, though other anion species have been reported such as phosphates.^242–245 During the discharge process, the metal oxides are lithiated by Li⁺ insertion coupled with transition metal reduction by electrons originated from the anode. The charge process is the reaction reversed. The cycling profiles of a few representative cathode materials are shown in Fig. 9 (adapted from Patoux et al.²⁴⁶). Some of these materials have already been demonstrated in RFBs and will be discussed first, followed by a discussion of other candidate materials.

LCO is one of the most widely used Li-ion cathode materials, especially in portable devices, and has a layered structure with its electrochemical properties widely reported.^12,247–252 It has been demonstrated in both type II and type III designs from Sec. II.^27,148 A suspension of 22.4 vol. % LCO with 0.7 vol. % KB showed a reversible capacity of 127 mA h g⁻¹_LCO,²⁷ close to the capacity of 137 mA h g⁻¹_LCO corresponding to 50% delithiation/lithiation, a common approximation for the practical capacity of LCO materials. LCO has a discharge voltage of ∼4.0 V versus Li/Li⁺ as shown in Fig. 9 and higher energy density can be achieved if charged to higher potentials that further delithiate LCO, although charging to such potentials generally accelerates material degradation or failure.^253–256 Compared with other common Li-ion cathode materials, LCO has relatively high electronic and ionic conductivities (∼10⁻⁴ S cm⁻¹ electronic conductivity and 10⁻¹¹–10⁻⁷ cm² s⁻¹ Li⁺ diffusion coefficient reported),^257–259 both of which minimize the overpotential while operating in flow cell geometries.¹⁴⁸ There are a few additional considerations for the use of LCO in RFBs. First, the environmental impacts of Co needs to be addressed.²⁶⁰ Although the recycling of Co from a flow cell system should be easier than static laminated Li-ion batteries, the toxicity of Co means that recovery and recycling is important. Second, the cost of Co is relatively high because of its relatively low earth abundance; and high energy inputs are needed during processing and manufacturing LCO, further demonstrating the need for implementation of effective material recycling.^261,262 Third, the particle volume is expected to change by 2.32% during cycling.²⁶³ The concern with volume change is that it may result in particle cracking, isolation of active materials, and hence capacity fade. However, particle fracture is less of an issue in a flow cell system because the particles are not statically connected in a solid composite and therefore there is no isolation problem in the dynamic solid suspension.¹⁴⁸ The volume change in large scale RFB system may cause pressure fluctuations during cycling, which is a potential technical challenge. Additionally, particle fracture during cycling in a flow system would change the rheological properties of the electrolyte, which requires appropriate design to account for these changes in electrolyte properties.

LiMn_1.5Ni_0.5O₄ (LMNO) is a spinel-phase material with a high voltage of 4.7 V versus Li/Li⁺ (shown in Fig. 9) and high theoretical capacity of 146 mA h g⁻¹, suggesting a theoretical energy density of 686 W h kg⁻¹.¹³⁷ It also uses Ni and Mn, which are both less expensive than Co. These promising features have previously been noted, and LMNO has been demonstrated in a type II design flow cell.²⁷ A suspension of 20 vol. % LMNO and 2.5 vol. % carbon was cycled as a half cell achieving close to theoretical voltage and high capacity.²⁷ Research effort has improved the electrochemical performance of LMNO in conventional Li-ion batteries using metrics such as rate capability, stability, and cycle life.^{137,264–268} Although the high energy density of LMNO is appealing, there are still challenges, in particular, long-term cycle life due to the high potential of LMNO which is outside of the stability window of many Li-ion battery electrolytes.^137,269 More detailed discussion about LMNO materials can be found in recent reviews.^137,270

Other transition metal oxide cathode materials may be potential candidates for RFBs. LiMn₂O₄ (LMO) has a spinel structure providing an average voltage of ∼4.0 V versus Li/Li⁺ (shown in Fig. 9) and a theoretical capacity of 148 mA h g⁻¹.^271–273 It has a lower cost and is more environmental friendly than other transition metal oxides such as LCO or LMNO, but capacity loss is an issue due to manganese dissolution in the electrolyte and/or particle crystallinity change.^274,275 The ionic and electronic conductivities are also relatively low (∼10⁻⁶ S cm⁻¹ electronic conductivity and 10⁻¹¹–10⁻⁹ cm² s⁻¹ Li⁺ diffusion coefficient reported),^{148,257,276,277} suggesting potential challenges of electrochemical performance in RFBs. Other materials like Li(Ni_xMn_yCo_z)O₂ (with varying compositions of x, y, and z) and vanadium oxides are also under active research.^278–281 To improve the performance of these cathode materials, doping with various metals has been explored. Dopants (e.g., aluminum, zinc, titanium, and magnesium) have been reported in specific material systems to increase crystal structure stability, battery capacity and capacity retention, material rate capability, and improve conductivity and dissolution rate.^262,282 More detailed discussions on these materials are available in recent reviews.^{135,242,262,281,283–285}

LFP is another appealing cathode material with a discharge voltage of ∼3.5 V versus Li/Li⁺ (shown in Fig. 9) and a theoretical capacity of 169 mA h g⁻¹.²⁸⁶ LFP features cost, environmental, and safety advantages and has been adopted in electric vehicles and hybrid electric vehicle batteries.^287,288 It also has a flat discharge voltage profile, which can be beneficial for type III design based on stochastic particle collision because a stable voltage output is expected even for particles at different SOC.^147,289 Extensive research has been done on synthesis methods, morphology control, dopants, and carbon coatings to improve the electrochemical performance of LFP.^{154,287,289–297} LFP was demonstrated as a cathode material for RFBs in both type II and type IV designs described in Sec. II.^{24,131,132,173} Other phosphates that have been researched in conventional Li-ion batteries can also potentially be adopted in RFBs such as LiMnPO₄,²⁹⁸ LiCoPO₄,^244,299,300 Li₃V₂(PO₄)₃,³⁰¹ and mixed phosphates.³⁰² More discussions on this group of phosphate cathode materials, especially LFP, can be found in recent reviews.^142,286,303

Li-ion solid active cathode materials generally have high voltage and capacity, providing the opportunity for RFBs with high energy densities. Development of these materials has been a very active field, but there are a few shared challenges and/or potential improvements to be addressed specifically for the application of RFBs. First, the suspension loaded with solid Li-ion battery particles generally results in high viscosity, as will be further addressed in Sec. V, and causes operating complexity and cost to pump the high viscosity electrolyte. Second, the effects of particle size, morphology, and microstructure are not well studied for RFBs, although there is a rich history with static Li-ion batteries to draw from. There are potential opportunities to improve the electrochemical performance of active material suspensions by designing the particle physical and electrochemical properties to be matched to a flowing suspension, rather than a static composite, system. In addition, surface modifications on active material particles could improve both the cycling performance and fluid properties (further discussion in Sec. V).

Besides metallic Li, which was discussed earlier with Li-PS chemistry, Li₄Ti₅O₁₂ (LTO) is a commonly used anode material in RFBs and has been demonstrated in type II and type III design geometries.^{27,131,147,148,155,304} LTO has material properties that make it desirable for RFB applications. It has reasonably high theoretical capacity of 175 mA h g⁻¹_LTO and a Li insertion/extraction voltage of ∼1.55 V versus Li/Li⁺, which lies within the stability window of common organic electrolytes. The electrochemical potential being within the stability window of common Li-ion battery electrolytes mitigates solid electrolyte interphase (SEI) formation and electrolyte decomposition.³⁰⁵ The discharge curve is very flat, providing a stable voltage output.¹⁴⁷ LTO is also reported as a zero-strain material, suggesting good particle morphology and crystal stabilities due to the lack of strain during Li insertion/extraction because there is no volume change during these processes.³⁰⁶ In addition, LTO has excellent ionic and electronic conductivities (10⁻⁶–10⁻⁸ cm² s⁻¹ Li⁺ diffusion coefficient and 10⁻⁶–10⁻¹³ S cm⁻¹ electronic conductivity), particularly after slight lithiation of the material which results in 10⁰S cm⁻¹ electronic conductivity.^{257,307–310} As an example, a suspension of 25 vol. % LTO and 0.8 vol. % KB showed close to theoretical capacity with reasonably low overpotential at a rate of C/3 (the current that is able to theoretically charge or discharge the electrode in 3 h).²⁷ Our group also took advantage of the flat discharge profile of LTO and designed a method using a type II flow cell design to measure a resistance parameter for the material which showed good correlation with the relative rate capability of the LTO materials.¹⁵⁵ LTO shares common challenges related to using solid dispersions in flowing electrolytes. A recent review provides a more detailed discussion on the materials development of LTO.³¹¹ 

Anatase TiO₂ is another titanium-based Li-ion anode material with an average potential of ∼1.8 V versus Li/Li⁺ and a theoretical capacity of 330 mA h/g, and it has been reported in RFBs.^{140,174,312,313} A type IV cell configuration showed ∼1.25 V versus Li/Li⁺ discharge voltage for CoCp*₂ and CoCp₂ mediators and using TiO₂ particles as the energy storage material that facilitated chemical redox with the mediators.¹⁷⁴ TiO₂ has cost, environmental, and safety advantages and a high ionic conductivity, but the electronic conductivity and rate capability are low relative to other anode materials.³¹³ More detailed discussions about the synthesis and development of TiO₂ as a Li-ion anode material can be found in these recent reviews.^313–315 Another titanium-based material that has been applied to RFBs is LiTi₂(PO₄)₃, which was paired with LFP and cycled in aqueous electrolyte, resulting in a ∼0.9 V full cell.¹³² The high redox potential (∼1.8 V versus Li/Li⁺) is beneficial to minimize hydrogen evolution and enables using this material in an aqueous electrolyte; however, in an organic system, the sacrifice in cell voltage would make this material unlikely to be chosen over other Ti-containing materials.³¹⁶ 

Silicon is a Li-ion anode material under active research and development with a high theoretical capacity (3590 mA h g⁻¹ by forming Li₁₅Si₄, ∼10 times higher than graphite, the most commonly used anode material).^290,317 A mixture of silicon and carbon in organic electrolyte was cycled showing a large capacity and low polarization.¹³⁰ However, silicon performance can suffer from large SEI formation which results in low coulombic and energy efficiency. The material also undergoes large volume change during cycling which can result in particle fracture during lithiation and delithiation cycles. More detailed discussions on silicon Li-ion anode material can be found in these recent reviews.^318–322 There are many new anode materials besides graphite and those discussed above, and these reviews provide a good perspective on anode development of Li-ion batteries.^{135,315,323–325} Anode materials require consideration of the tradeoff between energy and stability—lower potential materials are desirable because they increase the net cell voltage, but below the stability window of the electrolyte there will be electrolyte decomposition and formation of an interfacial layer, the stability of which is critical to long-term charge/discharge cycling of the material.

Although the energy density, power density, and cycle life advantages of Li-ion battery chemistry has been established, Na-ion battery materials have attracted attention, in part, because Na is more earth abundant and thus cheaper than Li, and in some cases allows for alternative material choices in the battery cell (e.g., Na does not alloy with aluminum, whereas Li does).^129,326 Ventosa et al. reported a nonaqueous RFB using P2-type Na_xNi_0.22Co_0.11Mn_0.66O₂ (NaNCM) and NaTi₂(PO₄)₃ (NaTP) as the cathode and anode active materials, respectively.¹²⁹ NaTP operates at a flat potential of ∼2.1 V versus Na/Na⁺, which is within the electrolyte stability window, and had a reported capacity of 125 mA h g⁻¹_NaTP.^129,327 P2-type NaNCM was demonstrated to have a capacity of ∼130 mA h g⁻¹_NaNCM with a range of potential plateaus between 2.1–4.3 V versus Na/Na⁺.^129,328 In their RFB system, a reversible 9 W h L⁻¹ energy density was demonstrated, although it was suggested 150 W h L⁻¹ would be achievable with a better selection of active materials and performance optimization.¹²⁹ The coulombic efficiencies were considerably low, 53% in the first cycle and ∼88% in subsequent cycles. The main challenges in this system were (1) high overpotential, causing incomplete utilization of the active material capacities; (2) low gravimetric capacity relative to other demonstrated Li-ion materials, which is even further reduced when pumping power is included in the total power output for the viscous suspension. Na-ion materials provide an alternate route to using solid intercalating particles in RFB systems. Recent reviews provide more detailed discussions on Na-ion battery materials.^{326,329–332}

Power needs to be applied in a RFB to drive the fluid for mass exchange between the electrochemical reaction cells and the energy storage tanks. In most scenarios, this energy is provided by pumps and the fluids run continuously as shown in Figs. 2, 5(a), 6(a), 7(a), and 8(a). There are other methods for providing this force to move the electrolyte fluid, as well as different operating modes, and both factors can have significant influence on the electrochemical performance of the cell. In this section, three operating modes will first be discussed, and then, two types of driving forces will be introduced that have previously been demonstrated for RFBs.

In a continuous pumping mode, both the catholyte and the anolyte are circulated between the reaction cells and the energy storage tanks continuously during charge/discharge. Such a mode is necessary for conventional soluble transition metal RFBs because (1) the energy densities of the fluids are limited and continuous refreshing of the active material solution is required to provide a stable power output; (2) the fluids that have been charged/discharged in the electrochemical reaction cells will be quickly mixed with the rest of the liquid in the energy tank, allowing a gradual change in electrolyte concentration of the oxidized/reduced electroactive species; (3) the fluid viscosities are typically low, making continuous pumping relatively easy to implement. This situation also applies to a type IV RFB where dissolved mediators are pumped to circulate and the mediators have relatively low energy density.^24,173,174 Therefore, continuous pumping of electrolyte is also the best method for type IV batteries. In a type III design where the reactions are based on collisions of active material particles on current collectors, suspension agitation is needed to charge/discharge the battery; therefore, a continuous pumping mode may be applicable, provided the fluid is pumped with sufficient force to keep the particles well-dispersed.¹⁴⁷ 

In other designs; however, the active materials have sufficient energy density to halt the flow before the active materials are fully discharged to save pumping energy. An intermittent flow mode, in which the active material suspensions are pumped to the electrochemical reaction cell in a batch manor and replaced after they are fully charged/discharged, was proposed and estimated to reduce the energy loss due to pumping. In a type II system, the estimated loss of energy due to pumping dropped from 44.6% in the continuous mode to 0.6% when an intermittent pumping was implemented.²⁷ However, there are additional items to consider for this operation mode. First, it is important to note that the estimates from the report referenced were based on a flow rate assumption that did not account for the energy to initiate flow and the initial fluid viscosity. At very low shear rates, the viscosity of the suspension is multiple orders of magnitude higher due to the shear thinning behavior of the solid suspensions.^27,148,208 Second, four suspension storage tanks are needed for intermittent flow operation because one tank is required for charged suspensions and one for discharged suspensions for each electrode.¹⁰⁴ Such a system must then account for the additional cost, space, and complexity requirements of four tanks as opposed to two per system. In addition, a stable method to switch active material batches is needed to avoid power interruptions between batches of electrolyte suspension being pumped into the electrochemical reaction cell intermittently.¹³² 

Another design is to add electroactive suspensions in a custom electrochemical cell that uses stirring to agitate the suspension and create a simulated flow environment.^147,148 This design is a modified version of the intermittent flow mode as fresh active material is pumped in after full charge/discharge. This method provides extra agitation via stirring while pumping is only needed during material exchanges. It is potentially useful in a type III flow cell design since the reaction is based on collision between active materials and the current collectors and reliable agitation of the suspension is needed to run the cell. It also provides an approach to boost the power output in other designs because the quick mixing of active materials may facilitate ion diffusion and hence lower the cell overpotential. An extra energy input is needed to drive the system agitation; therefore, the advantage of the additional agitation relative to its additional energy cost needs to be evaluated in detail for a system considering this design.

Other innovative approaches in addition to pumping have been reported for flow cells. A gravity-induced flow cell design (GIFcell) was proposed by Chen et al.²³⁰ As shown in Fig. 10, the energy storage tanks and electrochemical reaction cells are directly connected, and the mass transfer between these two units is induced by the force of gravity. The energy input to mechanically flip the cell is expected to be significantly smaller than pumping the viscous suspensions. One or more passes, or movements of the suspension from one end of the cell to the other, may be required to fully charge/discharge the cell. The number of passes is dependent on the flow rate, which is further controlled by channel thickness and width, channel surface properties, tilt angle of the channels, and cycling rates.²³⁰ In the first reported demonstration of this concept, a Li-PS chemistry was cycled within the potential window of Li₂S₈ and Li₂S₆ for the catholyte, and the counter electrode was Li metal. Chronoamperometry (CA) tests were conducted to charge the cell at 2.6 V and discharge at 2.05 V, with resulting current densities in the range of 0.5–1 mA cm⁻².²³⁰ A complete charge/discharge required 25 passes (12.5 h in total) for a tilt angle optimized cell. The mechanical energy for flipping the cell 25 times was calculated to be only ∼0.01% of the electrochemical energy stored in the cell.²³⁰ Although the low energy input is appealing for a flow inducing method, there are a few challenges to be addressed in addition to the quick capacity fade, which may have been due to PS shuttling. First, optimizing channel design for various active material suspensions for system scale up will be challenging because the flow can have a viscous fingering effect which impacts the uniformity of the flow profile, though the authors note this can be mitigated by reducing the channel width to height ratio; and channel surface properties must be designed to form high wall slip interactions between the surface and the fluid to result in the desired unyielding plugs. Thus, channel surface modifications will be needed depending on the composition of the active material suspensions. Second, although the flow rate can be controlled by tilt angle, the range and control over flow rate is limited, and a large number of flips may still require high energy inputs. Third, other electrochemical characterization, including galvanostatic charge/discharge, needs to be investigated for these systems. The drop in current density during CA testing as the number of passes increases, including 54% during discharge, suggests high overpotential which needs to be addressed. The difficulties of controlling flow in tilted channels suggests adjusting elevation of only the energy storage tanks in a more conventional flow cell design may be worth exploring, and would simplify design by separating the energy storage and power components.

Magnetic fields are another potential force to drive fluid movement in RFBs. γ-Fe₂O₃ nanoparticle suspensions are well-known as ferrofluids, and have also been reported as additives to energy storage fluids to promote using magnetic fields to provide transport advantages.³³³ Li et al. reported a Li-PS battery with added γ-Fe₂O₃ particles in a conventional static cell geometry and used applied magnetic fields to improve electrochemical properties.¹³⁹ An optimized suspension of PS, γ-Fe₂O₃ particles and carbon nanotubes achieved a high capacity of ∼350 mA h g⁻¹_sulfur (corresponding to ∼66 W h L⁻¹ of catholyte) while cycling between sulfur and Li₂S₄ under applied magnetic field, where the magnetic field improved cycling performance by concentrating the active material near the current collector.¹³⁹ In this suspension, γ-Fe₂O₃ nanoparticles absorbed PS on the surface and were concentrated near the current collector under an applied magnetic field, providing benefits with regard to mitigating PS shuttling and increasing the achievable discharge current densities. The magnetic particles also provide the possibility of driving the fluid with magnetic fields rather than pumps. The benefits of transporting the fluid with magnetic fields may include more precise control of the flow rate and direction, a simpler system, and possibly savings on pumping energy.¹³⁹ However, there a few challenges to implement these flow cell systems, especially at large scale. First, the composition of the electrolyte needs further research and optimization to be able to absorb PS and carry the flow with a minimum amount of γ-Fe₂O₃ loading to save both space and energy. Second, the electrochemical reaction cell has a relatively complex geometry, especially for large scale designs; therefore, the design of the magnetic field and the auxiliary components needed to generate the field and provide the flow pattern desired for numerous pipes and channels simultaneously will be needed. Third, while γ-Fe₂O₃ nanoparticles worked well with PS, it is not clear how general the use of this material will be, and surface modifications and/or alternative magnetic nanoparticles may need to be designed to match a given battery chemistry.

As discussed above, desirable properties such as high electrolyte energy density make RFBs with solid electroactive materials attractive battery systems to pursue. However, there are a number of challenges that span many of the cell designs that need to be addressed, and researchers have developed various methods to probe the fundamental properties of these systems or optimize the various battery components. We introduce some of these advances below.

Viscosity of the electrolyte suspension has been noted as a key parameter of operating efficiency for RFBs and understanding the underlying causes of changes in viscosity and methods to decrease this parameter are important to improve overall battery performance. Active material suspensions were found to be non-Newtonian fluids showing shear-thinning behavior,^{27,131,138,147,148,208} which is common for solid particle suspensions, especially metal oxides.^{94,138,334–336} This shear-thinning behavior was described by a power law or Ostwald–de Waele relationship as shown in Eq. (17),^147,337 where η is viscosity, γ is shear rate, and K and n are fitting parameters. K is known as the flow consistency index, giving the viscosity at a shear rate of 1 s⁻¹, and n is a number smaller than one for a shear-thinning fluid. This relationship indicates that (1) the fluid viscosity can be decreased significantly with increasing shear which corresponds to faster flow, predicting a minimum energy input for the highest practical fluid flow rates; (2) the energy to initiate flow can be high since the electrolyte initially has a shear rate close to 0, which results in challenges for the intermittent flow mode as discussed in Sec. IV A. The fluid viscosity is also highly dependent on solid material loading as described by the relationship between fluid viscosity and volume fraction of solid particles shown in Eq. (18),^148,336,338 in which $η$ is viscosity, $ϕ$ is volumetric solid particle loading, and a and b are fitting parameters. This power law relationship predicts a rapid increase in viscosity as the particle loading increases, indicating a tradeoff between fluid energy density and disproportional pumping energy loss.³³³ The relationship between particle loading and viscosity reinforces the need to find methods to reduce the fluid viscosity, in particular, for large scale applications where particle loadings are expected to be high. Narayanan et al. reported a rheoimpedance study on KB suspensions and indicated that a high flow rate would minimize the electronic resistance for RFBs in continuous flow mode and that a high preshear through fast pumping or stirring would help minimize the electronic resistance and maximize the yield stress in intermittent flow mode³³⁹ 

Carbon additives have significant impact on both the conductivities and viscosities of active material suspensions. Carbon material was added in type I and type II systems to improve the electrochemical performance, and at the same time, the viscosities of the suspensions were also increased dramatically. For example, the addition of only 0.6 vol. % carbon resulted a >10-fold increase in viscosity for a 22.4 vol. % LCO suspension.²⁷ There is a tradeoff between conductivity and viscosity with varied carbon additive loading, with multiple groups previously reporting on this phenomena in the literature.^104,339 Particle size, morphology, and salt concentration have previously been reported to have significant influence on the rheological properties of solid suspensions,^{138,208,338,340–342} and hence provide potential directions to decrease the fluid viscosity. There are a few reports in the literature aimed at decreasing the fluid viscosities of battery material solid suspensions in particular. A nonionic dispersant, polyvinylpyrrolidone, was applied in LFP/KB suspensions to selectively stabilize only LFP particles, but not KB particles, resulting in a decrease in viscosity as well as an increase in the conductivity of the suspension, both of which resulted in improved electrochemical performance.¹³¹ Similarly, the addition of nonionic surfactant (isooctylphenylether of polyoxyethylene) to a LTO/KB suspension decreased the viscosity and improved the mixture homogeneity.³⁴³ Surface modification of TiO₂ nanoparticles with a monolayer of propyl sulfonate groups was demonstrated to achieve high particle loading, low viscosity (below 10 cP with 50 wt. % particle loading), and high colloidal stability while maintaining good electrochemical activity.¹⁴⁰ Surface grafting of small organic molecules onto γ-Fe₂O₃ was used to decrease suspension viscosity significantly, although the electrochemical activity was partially suppressed by the surface grafting.³³³ Polyethyleneimine, a cationic surfactant, was found to effectively help disperse LFP and carbon suspensions and improve the homogeneity of LFP-C slurry electrodes, hence improving the electrochemical performance via processing of the slurry electrodes for conventional Li-ion cells.¹⁴¹ These results may be applicable to LFP/C suspensions for type II semisolid designs for improving suspension homogeneity and hence preventing particle agglomeration and decreasing viscosities.^126,141 While these examples are specific to battery particle suspensions, there have been many innovative methods reported to modify particle surfaces for improved rheological and/or electrochemical properties and are highlighted in recent related reviews.^344–346 

Mathematical modeling and simulations provide perspectives to understand the operating fundamentals and to optimize system performance. A 3D model of a RFB cell integrating ionic flux, electronic current, and hydrodynamic flow was developed to describe a type II design flow cell.³⁴⁷ Based on these calculations, active materials with flat charge/discharge potential profiles, i.e., electrochemical potential showing little variation with SOC such as LTO and LFP, are preferred to achieve more complete charge/discharge and higher energy efficiency than those materials with less flat potential profiles such as LCO.³⁴⁷ Smith et al. simulated the energy loss mechanisms using LCO, LFP, and the VO²⁺/VO₂⁺ couple, and provided four strategies to maximize the operating efficiency: (1) controlling the flow volume per pumping stroke; (2) promoting slip at interfaces; (3) reducing suspension rheology; and (4) selecting appropriate active material thermodynamics.³⁴⁸ A simulation based on a LTO-carbon suspension in a battery stack model system was conducted and compared with experimental results, indicating early-cycle coulombic inefficiencies, and three recommendations were made to improve the energy inefficiency: (1) extend the length of the flow channel/current collector; (2) operate at high charge/discharge rates at the stack level; and (3) control the cutoff voltage window.³⁴⁹ In addition, a half-cell model of a type IV cell was reported to capture the distinct charge and discharge profiles, providing the possibility of accurate full cell simulations.¹⁷⁶ 

Electronic and ionic limitations to electrochemical performance have been studied with LTO-carbon suspensions in static cells, noting the effects of the electrode thickness and LTO loading per area on charge/discharge profiles.³⁰⁴ For example, anolytes with 15 and 20 wt. % LTO accompanied by 3 wt. % KB can form a perfect percolated network connecting all LTO particles with no isolated KB clusters; however, when the anolyte is loaded at 25 wt. % LTO, the KB network (3 wt. %) is fragmented with isolated KB clusters, and only a fraction of the LTO mass can contribute to the reversible cell capacity.³⁰⁴ Different carbon black materials or carbon nanotubes also showed notable impacts on rheological and electrochemical properties.^{60,126,138,139,350,351} Ventosa et al. noted the importance of considering the SEI formation with flowing electrodes since within flow systems the SEI formation occurs on the full surface of the current collector, as well as the anode particles.³⁵² As discussed earlier, SEI is a metastable layer typically formed on the surface of the low potential anode when the potential is below the stability window of the electrolyte. The SEI is critical to stabilizing the electrode in conventional Li-ion batteries. SEI production, however, also causes capacity loss and results in additional cell overpotential, hence lower energy efficiency. For an electrode contacting a flowing electrolyte, new SEI can form continuously as new active material particles collide with the current collector, resulting in a much larger quantity of SEI formed and therefore lower energy efficiency. This effect was noted in a previous report as was the importance of selecting anode materials within the electrolyte stability window.³⁵² Petek et al. proposed the concept of an additional overpotential due to the distributed nature of the current between the electronic and ionic phases of the slurry to describe the high operating overpotential of suspension systems.³⁵⁰ Electrochemical impedance spectroscopy results showed an extra component in addition to traditional overpotential descriptions (Ohmic, activation and mass transfer overpotential), which is a function of slurry electrode charge transfer resistance and the ratio of the electronic and ionic phase conductivities.³⁵⁰ Increasing the suspension electronic conductivity was found to be the key factor to decrease overpotential and hence improve the electrochemical performance.³⁵⁰ These results provided insights to improve RFB performance, and additional studies are needed to further understand the different types of solid active material RFBs and to further develop these technologies.

RFBs with solid electroactive materials have attracted increased attention, and recent research progress was discussed in this review. Much of the increased interest is driven by the energy density improvements that are possible for RFBs with solid active materials, as this route overcomes the intrinsic energy density thresholds set by solubility limitations of more established RFB systems that rely on soluble redox species. Significant achievements have been reported in research on flow cell design, material development, operating methods, and fundamental understanding. These projects all aim to provide the knowledge and tools necessary to enable large scale energy storage devices using these RFBs by improving energy density, energy efficiency, and cost. The trend of using high energy density Li-ion solid active materials was noted, and this battery chemistry has been reported to provide high density energy storage in both conventional batteries and RFBs. In RFBs, the use of Li-ion battery materials was made possible due to the innovation of flow cell designs, introducing carbon percolating networks, taking advantage of collision reactions, and mediating reactions with external chemical oxidants/reductants. Li-S chemistry, in particular, has attracted substantial attention due to its high capacity, earth abundancy, and flexibility to operate completely soluble in the electrolyte or only soluble during different stages of charge/discharge. The challenges and possible research directions for further development of RFBs with solid electroactive materials are briefly discussed below.

1. Increasing energy density is desirable in every battery system. Using multiple electron redox couples in the suspension is an approach to further increase the energy density in solid active material RFBs, and such materials have been reported for static battery cells.³⁵³ Another option to increase energy density is to have multifunctional materials in the RFB. Chen et al. reported a catholyte containing solid sulfur/carbon composites and liquid lithium iodide (LiI) electrolyte, where the LiI serves as both a Li⁺ conductor and energy storage active material.³⁵⁴ This design achieved a high energy density (580 W h L⁻¹) and columbic efficiency (>95%), providing the possibility to increase the energy density by storing energy in both the solid and liquid phases. Further investigations applying similar designs with other materials, studying and optimizing the suspension viscosity, and finding more stable soluble species will improve the energy density and electrochemical performance. Electrolytes with high energy density, high ionic conductivity, and low viscosity are preferred.

2. Another potential way to increase the energy density is to combine RFBs and EFCs, resulting in a high energy density and fast response energy storage system. Hybrid systems with one electrode providing high surface area for high capacitance and the other electrode with LTO for Li⁺ intercalation have been reported.⁴¹ Research on a hybrid system using nanosized active material particles providing both electrochemical storage and capacitive energy storage could be possible and would help achieve relatively high energy and power densities.

3. Increased operating efficiencies are needed. Operating improvements could be accomplished by reducing the fluid viscosity or investigating surface coatings on the tubing and channels to reduce flow resistance, especially for high loading systems where the resistance due to the fluid being in contact with the tubing surface may be significant. Solid particle surface modifications were shown effective for reducing the fluid viscosity.³³³ Research on lowering viscosity of Li-ion electrolytes explicitly for RFBs has not yet been pursued to the authors' knowledge.

4. Increasing the electronic conductivity of solid electroactive particles is expected to improve the cell performance.³⁵⁰ Methods to increase particle electronic conductivity includes changing to intrinsically higher conductivity materials and carbon coating particles,^{142,143,154,289} which has previously been demonstrated to improve electrochemical performance for active materials with low conductivities such as LFP and LMO.²⁵⁷ Electronically conductive polymers may also be beneficial as particle surface coatings, particularly those providing high electronic conductivity as well as energy storage capability.^149,151 These coatings might also reduce the fluid viscosity, given previous reports of such effects from experiments on the rheological properties of polymer–particle suspensions.^149,150,217 The polymer coatings will also need to be conductive to Li ions to be effective.

5. Safety is always a concern for batteries, but is particularly important for large scale applications such as those suited to RFBs because at larger scales the failures could be catastrophic. A unique issue for RFBs relative to static cells is that clogging and unexpected pressure changes can be potential operating issues, which may cause unexpected failures. The safety concerns of fires and gas generation of Li-ion batteries are of course also relevant to the RFB systems with Li-ion materials and electrolytes.³⁵⁵ Future solid electroactive material RFB reports will likely present safety information on these systems, including abuse test and analysis.^356,357 Fault detectors and/or mitigation strategies will be needed within the battery.³⁵⁸ 

6. Although almost all RFBs are suited toward large scale applications, no large scale demonstrations have been done with solid electroactive material RFBs to date. Such demonstrations and experiments require significant capital investments; however, intermediate scale battery prototypes will need to be demonstrated in the near term to identify more realistic estimates of operating parameters.

7. Aqueous–nonaqueous hybrid designs, which have an aqueous suspension as the catholyte and a nonaqueous suspension as the anolyte, are promising because they combine the aqueous suspension high ionic conductivity and low cost with the nonaqueous suspension low operating potential which enables high cell voltages.^{74,75,359,360} The key challenge in such a system is designing a membrane compatible with both electrolytes with long lifetime, reliable separation, high and selective ionic conductivity, and low cost.

8. Battery temperature control is an important component of battery research, particularly for large scale applications to prevent potential safety hazards and ensure long battery lifetimes.^{13,55,361–364} RFBs have temperature regulation advantages because heat exchangers may be constructed as an additional unit operation seamlessly integrated with the flowing electrolyte through tubing, as opposed to complicated heat removal systems designed for static battery stacks. The efficiency of such a design for temperature regulation is expected to be tested in the future with flowing solid suspensions.

Overall, RFBs with solid electroactive material suspensions show promise to provide reliable energy storage with high capacity, efficiency, safety, and low cost. Future research may accelerate the development of these technologies and demonstrate these systems at increasing scales of applications.

The authors acknowledge financial support from the National Science Foundation through Award No. ECCS-1405134.