The Critical Role of Supporting Electrolyte Selection on Flow Battery Cost

Redox flow batteries (RFBs) are promising devices for grid energy storage, but additional cost reductions are needed to meet the U.S. Department of Energy recommended capital cost of $150 kWh−1 for an installed system. The development of new active species designed to lower cost or improve performance is a promising approach, but these new materials often require compatible electrolytes that optimize stability, solubility, and reaction kinetics. This work quantifies changes in RFB cost performance for different aqueous supporting electrolytes paired with different types of membranes. A techno-economic model is also used to estimate RFB-system costs for the different membrane and supporting salt options considered herein. Beyond the conventional RFB design incorporating small active species and an ion-exchange membrane (IEM), this work also considers size-selective separators as a cost-effective alternative to IEMs. The size selective separator (SSS) concept utilizes nanoporous separators with no functionalization for ion selectivity, and the active species are large enough that they cannot pass through the separator pores. Our analysis finds that SSS or H+-IEM are most promising to achieve cost targets for aqueous RFBs, and supporting electrolyte selection yields cost differences in the $100’s kWh−1. © The Author(s) 2017. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited. [DOI: 10.1149/2.1031714jes]

Energy storage has emerged as a key technology for improving the sustainability of electricity generation 1 by improving the efficiency of existing fossil-fuel infrastructure through load-leveling or price arbitrage, 2 alleviating the intermittency of renewables (i.e., solar, wind) to promote their broad implementation, 3 and providing high-value services such as frequency regulation, voltage support, or back-up power. 2 Redox flow batteries (RFBs) are promising devices for low-cost grid energy storage due to decoupled capacity and power scaling, long operational lifetime, easy thermal management, and good safety features. 2,[4][5][6][7][8][9] Unlike enclosed batteries (i.e., lithium-ion, nickel-metal hydride), RFBs implement soluble redox active species dissolved in liquid electrolytes, which are stored in large, inexpensive tanks. Specifically, the electrolyte is comprised of a supporting electrolyte, which contains solvent (e.g., water) and a supporting salt (e.g., sulfuric acid, sodium chloride), and the redox active species (e.g., bromine). The electrolyte is pumped through an electrochemical stack where the active species are oxidized or reduced to charge or discharge the battery. The size of the electrochemical stack determines the power rating, while the tank volume determines the energy capacity, enabling scalability unique to the RFB architecture. A variety of RFB chemistries have been researched in recent years 7,10-12 with many examples of successful commercial deployment, such as zinc-bromine (e.g., Redflow, 13 Primus Power 14 ), zinciron (e.g., ViZn Energy Systems 15 ), and organometallics (e.g., Lockheed Martin 16 ). The all-vanadium redox flow battery (VRFB) has been the most successful chemistry, 17,18 and several companies (e.g., Sumitomo Electric Industries, 19 Vionx Energy, 20 Gildemeister, 21 UniEnergy Technologies 22 ) are presently commercializing the technology at grid scale. The success of the VRFB has hinged on its facile engineering, given that all relevant redox couples lie just within the practical electrochemical window of the sulfuric acid (H 2 SO 4 ) based electrolyte, enabling high cell potential without dissociating water. Additionally, since the positive and negative electrolytes utilize the same parent species (vanadium sulfate), crossover does not irreversibly degrade cell performance. 18 Despite the technological success of the VRFB and other aqueous redox flow battery (AqRFB) chemistries, RFB costs in 2014 exceeded $500 kWh −1 23-25 , well above the U.S. Department of Energy (DOE) recommended capital cost target ($150 kWh −1 ) for an installed energy storage system with 4-h discharge at rated power. 2 The introduction of new redox chemistries is a strategy for substantially lowering the electrolyte (energy) cost contribution to the total battery cost via decreased chemical costs or increased electrolyte energy density. 23,26 Key active species characteristics in determining the RFB electrolyte cost are the solubility (M), molar mass (kg mol −1 ), number of electrons stored per molecule (-), material cost ($ kg −1 ), and cell potential (V). The latter two characteristics have an especially significant impact on total RFB cost. 23,26 Raising cell potential, by identifying active species with more extreme redox potentials, 26 is a particularly effective approach to reducing RFB costs because increasing cell potential decreases both the electrolyte and reactor (power) cost contributions. 23,26 Battery cost is also sensitive to active material cost, 26 and, as such, many recent studies have sought to identify active species that could serve as cheap replacements for the incumbent RFB chemistries. Abundant inorganic active species (e.g., metal polysulfides, [27][28][29] iodide 28,29 ) can be inherently inexpensive. Further, redox-active organic molecules (ROMs) are comprised of earth-abundant elements (i.e., hydrogen, carbon, oxygen, nitrogen, sulfur) and their cost is not determined by production rates of raw materials or material reserves. 11 New organometallic active species, [30][31][32][33][34][35][36] as well as ROMs, show promise through the addition of functionalizing ligands, enabling rational molecular design.
Typically when a new active material is discovered, the supporting electrolyte must simultaneously be tuned to facilitate its implementation. Key technical considerations when designing a supporting electrolyte are the active material solubility, chemical stability, reaction kinetics, and safety. Numerous experimental examples of such electrolyte development campaigns can be found in literature. As a solubility example, 2,6-dihydroxyanthraquinone is only sufficiently soluble (>0.5 M) in alkaline electrolytes, 37 and anthraquinone disulfonic acid (AQDS) is more soluble in the protonated state, as opposed to the sodium form. 38,39 Regarding charge transfer kinetics, modifying the supporting electrolyte composition, without adding catalysts or pretreating electrodes, has been shown to dramatically affect the electrochemical reaction rates for both metallic and organic A3884 Journal of The Electrochemical Society, 164 (14) A3883-A3895 (2017) active species. [40][41][42] Considering chemical stability, charged methyl viologen 31,43,44 is known to react with molecular hydrogen, making it unsuitable for use in acidic electrolytes where hydrogen evolution is a likely side reaction. 45 Finally, as for safety, metal polysulfides 27 or ferrocyanide 27,37 can undergo chemical decompositions in acid to produce toxic hydrogen sulfide or cyanide gases, respectively. While designing a new supporting electrolyte may be beneficial for enhancing the electrochemical behavior of a new active material, other celllevel performance metrics, such as area specific resistance (ASR) or cell potential, can suffer as a consequence; these cell-level performance metrics are often overlooked when developing a cost-effective alternative to the state-of-the-art VRFB. For example, neutral and alkaline RFBs tend to exhibit significantly lower power density 27,30,37 than VRFBs 46-48 due to more resistive membranes. Moreover, many new proposed chemistry combinations exhibit theoretical open-circuit potentials (OCP) well below that of the VRFB (1.4 V). 18,30,31,38,39,43,49 This study explores how variations in RFB performance due to supporting electrolyte and membrane selection impacts system cost. We focus on aqueous electrolytes, as opposed to nonaqueous electrolytes, 12,50,51 due to the higher technology-readiness level of AqRFBs and, consequently, the larger amount of associated device information for grid-relevant operation. First, we identify governing physical parameters, namely the membrane ASR, electrolyte conductivity, electrolyte viscosity, and cell potential, which significantly impact RFB cost. Membrane ASR can vary drastically with membrane type, working ion, or electrolyte pH. [52][53][54] The latter two factors can also impact electrolyte conductivity and viscosity. A typically overlooked characteristic, electrolyte viscosity is of critical importance as it directly impacts mass transfer rates in the porous electrodes of a RFB, as well as required pumping power through the entire battery. [55][56][57] As mentioned before, cell potential is a key parameter that affects the power and energy density of the RFB, both of which define RFB cost. 23,26 Second, to link these materials properties to RFB cost, fullcell ASR is calculated by implementing a one-dimensional porous electrode model that solves for electrode polarization as a function of electrolyte resistivity, charge-transfer kinetics, and convective mass transfer rate. The electrode ASR contribution is combined with ASR values of ion-exchange membranes (IEMs) 27,31,58 and newly proposed size selective separators (SSS) [59][60][61][62][63] in various supporting electrolytes. Third, we develop a techno-economic model to estimate RFB capital cost as a function of electrolyte composition and cell potential, among other detailed device parameters. The reactor cost contribution is calculated by leveraging prior literature, 23,25,26 as well as developing new and more nuanced descriptions of the electrolyte cost, which enable consideration of cost differences from various supporting salts and specific inclusion of tank costs. We also implement a new estimate of balance of plant (BOP) costs, which accounts for variations in pumping costs with changes in electrolyte flow rate and viscosity, as well as reactor geometry. Combining the reactor, electrolyte, and BOP costs, as well as literature estimates of unit cost less materials, 23,25 permits evaluation of variations in RFB capital cost for different AqRFB supporting electrolytes. Through this dual analysis of electrochemical performance and techno-economics, we have found that changes in supporting electrolyte or membrane selection can yield battery cost differences in the $100's kWh −1 .

Estimating Materials Parameters
Ion-exchange membrane and electrolyte conductivities.-Conductivities of IEMs implemented with new RFB electrolytes are rarely reported and can vary drastically depending on the working ion 52 and the properties of the surrounding supporting electrolyte. 64 As such, we derive membrane (κ mem ) and electrolyte (κ electrolyte ) conductivities from a limited amount of available literature data (Table I). For H + -IEMs, we assume the conductivity of a membrane employed in a state-of-the-art VRFB, 58 and the associated electrolyte conductivity is taken as an average of VRFB electrolyte conductivities across all states-of-charge (SOCs). 18 For Na + -and Cl − -IEMs, we estimate membrane conductivity from the high-frequency intercept of full-cell impedance measurements from published flow cells employing these membranes. 27,31 As thick membranes (≥ 120 μm) were employed in the relevant studies, the high-frequency intercept on the experimental Nyquist plot is dominated by the membrane resistance. The conductivity of a NaCl-based RFB electrolyte is taken as an average value from a recent literature study. 31 Although a K + -IEM has been utilized in two experimental RFB studies, 37,65 the associated conductivity has not been measured, and no impedance data is available to derive an experimental estimate. In the present work, we estimate the conductivity of K + -IEM by scaling the conductivity of the H + -IEMs by the ratio of ionic conductivities of the two membranes without active species present. 52 In a similar fashion, we estimate the conductivity of a KOH-based electrolyte by scaling the conductivity of an average VRFB electrolyte by the ratio of the peak ionic conductivities 66 of H 2 SO 4 and KOH. From prior systematic studies of Nafion conductivity with various working ions in supporting electrolyte (no active species), we anticipate that cation-exchange membrane conductivity will decrease in the following order: H + >Na + >K + . Similarly, from prior data reporting on the conductivity of various acid and salt solutions, we anticipate that the electrolyte conductivity should decrease as: H 2 SO 4 >KOH >NaCl. 66 The membrane conductivities reported in Table I are converted to a membrane ASR (R mem , cm 2 ), where Figure 1a denotes the relationship among R mem , the working ion, and the membrane thickness. Table I and Figure 1a also specify example performances for two hypothetical size-selective separators (SSS). Recent reports have recognized that the membrane of a RFB does not need to be selective to a single ion in the electrolyte, but rather must only reject the active species, permitting any supporting ions to maintain electroneutrality across the cell. From this realization emerged a different approach, first proposed in 2011, 62 of pairing SSS with larger active species where the separator allows smaller supporting ions to exchange between the two electrolyte streams while blocking crossover of the larger active species. [59][60][61][62][63][67][68][69] Since SSS are comprised of non-functionalized polymers, they promise higher conductivities and lower costs as compared to their functionalized IEM counterparts. However, to date, SSS have only been reported in a small number of literature studies, 60,62,63,[68][69][70][71] limiting the amount of available experimental data. As such, we estimate their effective conductivities (κ eff , mS cm −1 ) using the Bruggeman relation (Equation 1), where κ is the electrolyte conductivity (mS cm −1 ) in the pore phase of the SSS, ε is the SSS porosity (-), and b is the Bruggeman coefficient (b = 1.5). 72 The SSS porosity is assumed to be that of Celgard 2500 (ε = 0.55), 73 a typical nanoporous separator employed in lithium-ion batteries, 59 and the electrolyte conductivities of the pore phase are taken from literature reports (Table I). 18,31 In the present estimates of SSS conductivities, perfect separator wetting is assumed. For comparison, we estimate the κ mem for a Na + -SSS to be 41 mS cm −1 , and a recent experimental study on SSS, without active species, measured κ mem to be 79 mS cm −1 . 68 This discrepancy is attributed to the higher conductivity of a NaCl electrolyte, in the separator pore phase, without any active species present. Transport properties.-To begin estimating transport rates of active materials in RFBs, diffusion coefficients are calculated using the Stokes-Einstein relationship (Equation 2), 74 where k B is the Boltzmann constant (1.38 × 10 −23 J K −1 ), T is the nominal electrolyte temperature (298 K), μ is the electrolyte viscosity (Pa · s), and r is the solvated radius of the active species in solution (m). This work will consider electrolyte viscosities in the range of 1-10 mPa · s, and given that SSS will likely require larger active species for successful RFB operation, two representative solvated radii are selected. For small active species (e.g., inorganic ions, ROMs) paired with IEMs, r = 1 nm, corresponding to a typical diffusion coefficient of V 2+/3+ in aqueous media (μ = 1 mPa · s, T = 298 K) of ≈ 2 × 10 −6 cm 2 s −1 . 75 For larger active species (e.g., redox-active oligomers) paired with SSS, we assume a radius 3 × that of the small species (r = 3 nm), which is a similar increase as two recent experimental reports on SSS for blocking oligomeric or polymeric active species in RFBs. 60,68 The relationship between diffusion coefficient and electrolyte viscosity is plotted (Figure 1b) for both solvated radii, indicating that larger active species will exhibit slower transport rates. While this analysis does not explicitly consider the case of transport for large redox-active polymers 68 (RAPs) in AqRFBs, the case of a high viscosity electrolyte, with large active species, likely mimics their transport rates, assuming Newtonian flow behavior.
The diffusion coefficient of an active material describes transport in the absence of forced convection and must be mathematically connected to a description of mass transport within a flow-through porous electrode. The total effective mass transfer rate (k m , m s −1 ), in a porous carbon electrode, can be described in two steps (Equation 3), including the macro-scale mass transfer rate from the flow-field channel to the electrode pores (k r , m s −1 ) and the pore-scale mass transfer rate (k p , m s −1 ): 76 The macro-scale transport rate is provided in Equation 4, where α (-) and β ((m s −1 ) 1−α ) are constants specific to the porous electrode, and v e is the intra-electrode electrolyte velocity (m s −1 ). 76,77 α and β values for transport to a carbon fiber electrode are available in Ref. 78.
We employ an empirical relationship (Equation 5) for the porescale mass transfer rate (Figure 1c), measured for the case of vanadium transport in a porous carbon-felt electrode, 55 where d f is the electrode fiber diameter (m) and Re is the Reynolds number (-) for the electrolyte within the porous electrode. For this work, the electrode fiber diameter is assumed to be that of SGL 25AA (d f ≈ 7 μm). 79 Note that the diffusion coefficient is not constant in the present analysis and has a dependence on the electrolyte viscosity (Equation 2).
The Re associated with the porous electrode, sometimes referred to as the Blake number, is similarly defined as in the case of a packed particle bed reactor (Equation 6). In Equation 6, ρ is the electrolyte density (kg m −3 ) and ε is the electrode porosity (-). For this work, the electrolyte density is assumed to be that of water (ρ = 1000 kg m −3 ), and the electrode porosity is that of SGL 25AA carbon paper under ≈ 20% compression (ε = 0.75). 79 The intra-electrode velocity (v e ) is calculated as the mean velocity over the rib of an interdigitated flow field (IDFF) from a typical high performance VRFB experiment (v e = 0.025 m s −1 ). 80 All relevant electrode and transport properties are listed in Table II.

Computing Cell Area Specific Resistance
The full-cell ASR (R DC , cm 2 ) can broadly be described as the summation of membrane (R mem ), contact (R contact ), and electrode resistances (R electrode ) in the cell (Equation 7). Contact resistances assume a fixed experimental value of 35 m cm 2 for SGL 25AA carbon paper under 20% compression on a typical carbon composite bipolar plate with an IDFF. 80 The membrane resistances are derived from Figure 1a, assuming an optimistically thin membrane of 25 μm, the dry thickness of Nafion N211. 80 Electrode resistance model.-The electrode resistance contribution to R DC is computed using a one-dimensional, steady-state porous electrode polarization model from a recent publication, and this model was validated with experimental flow cell polarization data employing various flow fields and electrolyte flow rates. 82 The model calculates individual electrode polarization, accounting for overpotential losses due to the convective mass transfer, Butler-Volmer reaction kinetics, and the electrolyte resistivity in the pore-phase of the porous electrode. Importantly, this steady-state model assumes that variations in active species concentration in an electrode are negligible in the directions parallel to the separator, which holds true when the electrolyte flow rate is very high relative the rate of change in battery SOC (i.e., low reactant conversion per pass). 82 This assumption aligns with high power RFB cell design, where typically the mass transfer increases afforded by pumping the electrolyte faster lead to improvements in power density that vastly outweigh the pumping losses associated with the faster flowing electrolyte. 46 To briefly summarize the full derivation and analysis available in Ref. 82, Equations 8 through 16 describe how the overpotential distribution and, subsequently, the current distribution in the porous electrode are computed, using all dimensionless parameters.
The second derivative of the overpotential distribution (d 2η /dx 2 ) with respect to the position in thickness of the electrode is proportional to the overall electrochemical reaction rate (Equation 8), which is a function of several dimensionless parameters. Equation 8 is specifically derived for an electrode at 50% SOC, with transfer coefficients α a = α c = 0.5, and the solid-phase electrode conductivity (2200 -2700 mS cm −1 ) 80 is much higher than the electrolyte conductivity (see Table I).
η is the dimensionless overpotential, which is defined as the overpotential (η, V) normalized by the thermal potential (RT/F, V), where R is the gas constant (8.314 J mol −1 K −1 ) and F is the Faraday constant (96485 C mol −1 ):η = Fη RT [9] x is the dimensionless position in the porous electrode, where x is position (m) and L e is the electrode thickness (m): x = x L e [10] c is the dimensionless concentration, which is defined as the active species concentration (c, mol m −3 ) normalized by the reference concentration (c θ , typically 1000 mol m −3 ): ν 2 is the dimensionless exchange-current density (Equation 12), where a is the electrode area per volume (m 2 m −3 ), i 0 is the exchange-current density (A m −2 ) and κ eff is the effective electrolyte conductivity in the porous electrode (S m −1 ). Physically, ν 2 represents the ratio of chargetransfer and ion-conduction rates in the porous electrode. [12] θ is the dimensionless limiting current (Equation 13), defined as the ratio between the exchange and limiting current densities (i l , A m −2 ). Physically, θ represents the ratio of charge-transfer and mass transfer rates in the porous electrode.
The limiting current density relates to the mass transfer coefficient (Equation 14), where n is the number of electrons transferred per active molecule (-).
The dimensionless overpotential distribution in the electrode can be determined by setting a constant overpotential boundary condition at the membrane-electrode interface (x = 0) and a zero ionic current boundary condition at the electrode-current collector interface (x = 1). Once the overpotential distribution has been solved numerically, the dimensionless current density (δ) passing through the membrane can be derived (Equation 15).
The geometric cell current density (J, A m −2 ) is related to δ by Equation 16. [16] Now that current density can be related to the overpotential at the membrane-electrode interface, we can generate polarization curves that describe the current-voltage characteristic of RFB electrodes at 50% SOC. Relevant values of the θ parameter can be computed by relating θ to k m (Figure 1c). The effective electrolyte conductivity in the porous electrode (κ eff ) is found by using the Bruggeman relation (Equation 1) and the relevant electrolyte conductivity (Table I). The electrode area per unit volume (a, m 2 m −3 ) is estimated using Equation 18 and listed in Table II. 83 Given these relations, values of θ can be generated as a function of experimentally measured mass transfer coefficients and the dimensionless exchange current density (ν 2 ).
Figure 2a illustrates electrode polarization for various Re with optimistically fast reaction kinetics (ν 2 = 2) and a large active species (r = 3 nm), with a H 2 SO 4 -based electrolyte. Polarization calculations are performed assuming 1.5 M total active species concentration, a target value for AqRFBs. 23 Varying active species concentration is a commonly explored electrolyte design parameter, so, to focus on the underreported role of supporting electrolyte composition and viscosity, the present study does not vary active species concentration. Further, while the 50% SOC model assumption does not capture ASR variations due to changes in active species concentration that may affect charge or mass transfer resistances, the ASR at 50% SOC represents the average cell performance throughout the battery's discharge; at high SOC (>50%), cell ASR for discharge will be smaller, whereas at low SOC (< 50%), ASR for discharge will be larger. Incorporating a SOC-dependent ASR model would require knowledge of a characteristic discharge profile, which would also narrow the applicability of our modeling efforts. Such a model and necessary experimentation would not exhibit the transparency and computational efficiency of the polarization model used here. Fixing the SOC at 50% in the present analysis permits mathematical simplification, offers a reasonable estimate of cell performance throughout discharge, and provides multiple parameters to analyze.
For all Re in Figure 2a, as the current density through the cell increases, the overpotential drop across the electrode increases, but, as Re increases, a smaller overpotential is required to drive the same current through the electrode. This trend occurs because increasing Re correlates with a higher rate of convective mass transfer (i.e., smaller viscosity, increased flow rate), yielding lower mass transfer resistances. Re = ∞ represents the case of infinitely fast mass transport, where the electrode experiences losses only due to charge-transfer and electrolyte resistivity. The low overpotential regime (≤ 60 mV) of the polarization curves is linearized to derive a value R electrode for each Re. An overpotential of 60 mV aligns with voltaic efficiency targets (91.6%) 23 during both charge and discharge for AqRFBs, assuming a large cell potential of 1.5 V; 23,59 hence, during typical operation, a RFB electrode at 50% SOC should be polarized by a maximum of ≈ 60 mV. Figure 2b plots R electrode as a function of Re −1 for three values of the dimensionless exchange current density (small active species, r = 1 nm, H 2 SO 4 -based electrolyte), highlighting the critical role of reaction kinetics in defining R electrode . Increasing ν 2 can deliver significant reductions in R electrode (>0.25 cm 2 ), while varying Re changes R electrode by as much as 0.03 cm 2 across the range of Re under consideration for the small active species. As ν 2 increases, R electrode becomes more sensitive to Re because the relative contribution of charge-transfer losses shrinks. Since one motivation for implementing a new active material or supporting electrolyte in a RFB is improved reaction rates, 30 all subsequent calculations will incorporate optimistically large exchange current densities. While the kinetic rate constant of an electrochemical reaction is an important factor in determining overall cell performance and cost, reaction kinetics within RFBs has been studied extensively, with many reports identifying new catalysts or electrode treatments to reduce activation polarization losses. 5,7,8,18 Moreover, in a previous report, Darling et al. contemplate the cost implications of active species with varying kinetic rates. 23 Thus, the present report focuses on the case of optimistically fast kinetics (i.e., ν 2 = 2.0) to highlight the cost implications of underreported factors: charge carrier identity, salt cost, and electrolyte viscosity. Our value of i 0 is estimated by first assuming that cells with even the highest electrolyte conductivities (i.e., H 2 SO 4 -based supporting electrolyte) will be ohmic-limited. As such, the H 2 SO 4 -based electrolyte is assigned ν 2 = 2.0, corresponding to i 0 = 7.26 mA cm −2 , and this exchange current is fixed in all subsequent analyses. For reference, our estimated value of i 0 is similar to that of AQDS on carbon paper (12.3 mA cm −2 ), 84 which is known to exhibit rapid kinetics in H 2 SO 4 -based electrolytes. 39,84 Figure 2c shows electrode resistances as a function of Re −1 for H 2 SO 4 -, NaCl-, or KOH-based electrolytes with small active species (r = 1 nm), as well as H 2 SO 4 -or NaCl-based electrolytes and large active species (r = 3 nm). All resistance curves in Figure 2c employ the high value of exchange current density previously mentioned. As the electrolyte conductivity decreases (i.e., H 2 SO 4 to KOH), the electrode resistance increases due to a larger ohmic loss through the porous phase of the electrode. This phenomena is especially apparent in the limit of infinitely fast mass transfer (Figure 2c, y-intercept), where the variation in R electrode among the three supporting electrolytes is strictly imposed by differences in the electrolyte conductivity. Between the H 2 SO 4 and NaCl electrolytes, changes in the electrolyte conductivity can increase the total electrode resistance by a small value of 0.04 cm 2 ; this difference will be exacerbated if thick (i.e., >>300 μm) porous electrodes are selected. Further, the sensitivity of electrode resistance to changes in Re increases as the active species size increases; the large active species (r = 3 nm) induces a change in R electrode of 0.10 cm 2 across the range of Re considered, while the small active (r = 1 nm) species causes an increase of only 0.03 cm 2 . The larger active species exhibits lower diffusion coefficients (Figure 1b), invoking slower mass transport rates (Figure 1c) and subsequently causing greater electrode sensitivity to Re (Figure 2c).
Full cell area-specific resistance.-One can now estimate the fullcell ASR (R DC , Figure 3) of an AqRFB employing various working ions, membranes (IEMs or SSS), or electrolyte viscosities by incorporating the values of R electrode (Figure 2c) into Equation 7. Note that the present analysis considers a fixed IEM or SSS thickness of 25 μm, and lower (or higher) ASR values could be achieved if the thickness were further reduced (or increased). Estimates of R DC for IEM-based RFBs assume transport rates for small active species (r = 1 nm) with H 2 SO 4 -, NaCl-, or KOH-based supporting electrolytes, whereas estimates for SSS-based RFBs assume transport rates for large active species (r = 3 nm) with H 2 SO 4 -or NaCl-based supporting electrolytes. Changing the type of ion passing through a membrane leads to a linear increase  Table II. or decrease in R DC due to variations in membrane conductivity, while increasing Re (lower Re −1 ) decreases R DC due to a higher rate of convective mass transfer. When considering the SSS case, both the NaCl and H 2 SO 4 systems offer the lowest R DC at high Re, however, at sufficiently low Re, the H + -IEM outperforms both SSS options. While we do not explicitly consider the effects of temperature on ASR, the resistance contribution plots (e.g., Figure 1a, Figure 2b, Figure 2c, Figure  3) can illustrate the role of increasing temperature on decreased cell ASR; raising temperature will increase charge and mass transfer rates, as well as electrolyte and membrane conductivities, all of which will decrease cell ASR. Figure 3 ultimately illustrates a critical balance in membrane and supporting electrolyte selection, where the membrane conductivity and electrolyte viscosity must be optimized to deliver the lowest R DC possible. The Na + -, Cl − -, and K + -IEMs afford much higher ASR (up to 3.1×) than the H + -IEM or SSS. Multiple techno-economic analyses have recommended a target R DC ≈ 0.5 cm 2 for costeffective AqRFBs, 23,26,59 indicating that only a few of the membrane and working ion combinations presented in Figure 3 could achieve DOE cost targets, in the regime of high Re. The next section uses a techno-economic model to estimate complete RFB capital costs for the different membrane and supporting salt options under consideration here, so that the reader may better appreciate the cumulative cost impact of all the RFB components.

Pressure Drop and Pump Power Requirement
Pressure drop.-We assume that AqRFBs will employ state-ofthe-art IDFFs, which have been shown to balance excellent electrochemical performance with an acceptable pressure drop. 80, 85 Darling and Perry described a series of analytical equations to calculate the pressure drop through an IDFF for flow batteries 80 using a formulation originally targeted for polymer-electrolyte fuel cells. 86 The pressure drop through an IDFF ( P, Pa) can be calculated from Equation 19, where P ch is the pressure drop through a channel in the flow field (Pa) and ζ is a dimensionless geometric factor (-).
The pressure drop through a channel within the IDFF is computed according to Equation 20, where v ch is the electrolyte velocity in the channel (m s −1 ), L ch is the channel length (m), and d h is the channel hydraulic diameter (m).
Since a rectangular channel is assumed, the d h can be calculated in Equation 21, where w ch is the channel width (m), and h is the channel height (m). In this work, w ch and h have fixed dimensions of 0.00117 m and 0.00076 m, respectively, based on a recent experimental RFB study employing IDFFs. 80 The channel velocity is defined in Equation 22, where N is the number of inlet channels in the flow field (-), and Q is the electrolyte flow rate through one side of a single cell (m 3 s −1 ). v ch = Q N ch w ch h [22] The channel and intra-electrode velocity for the IDFF are related, as shown in Equation 23. [23] The geometric factor, ζ, in Equation 19 is defined in Equation 24, where k is the electrode permeability (m 2 , Table II) and S is the mean path length of the electrolyte through the electrode and over the rib (m).
While S can be numerically computed for the IDFF, 87 such a calculation is beyond the scope of this work. We estimate S as given in Equation 25, where w rib is the width (m) of the rib in the IDFF. w rib has a fixed dimension of 0.00089 m, based on a recent experimental RFB study employing IDFFs. 80 S ≈ w ch + w rib + L e [25] Later estimates of pump cost will consider cells with a total width (W) of 26 cm and L ch = 31.2 cm (L ch /W = 1.2), which are similar dimensions to kW-scale flow cells, [88][89][90] resulting in a cell active area of 811 cm 2 . For fixed channel and rib widths, W defines the number of inlet channels and ribs in the IDFF (for W = 26 cm, N ch = 80). To briefly illustrate how pressure drop varies with cell aspect ratio (L ch /W), consider Figure 4a, which plots the pressure drop through a single side of one cell against inverse mass transfer coefficient within the porous electrode, both of which are functions of electrolyte flow rate. Figure 4a illustrates the tradeoff between pressure drop and mass transfer rate for a small (r = 1 nm) active species. As expected, larger convective mass transfer rates (smaller k m −1 ) will require higher pressure drops. Further, as the aspect ratio increases (higher L ch /W), a larger pressure drop is required to sustain the same mass transfer rate, due to a larger flow rate requirement to sustain the same intra-electrode velocity.
Pump power requirement.-The required pump power (P, W) to pass electrolyte through a single side (positive or negative electrolyte) of a RFB can be computed as shown in Equation 26, where ε pump is the pump efficiency (-), and N cells is the number cells in the electrochemical stack (-). 24 We assume that the pump efficiency is ε pump = 0.85 91 and that the RFB stack contains 60 cells, in accordance with a prior RFB cost model. 24 Figure 4b shows how the required pump power varies with mass transfer rate and cell aspect ratio. The trends are similar to the pressure drop calculation in Figure 4a, but the scaling from P to pump power is nonlinear due to a nonlinear dependence of Q on P (Equations 19 and 22). P = P Q N cells ε pump [26] The primary goal in evaluating pump power requirement here is to account for the capital cost of appropriately sized pumps for an aqueous RFB stack, however, the importance of designing a RFB with  sufficiently low pump power requirement, relative to the stack output power, should not be overlooked. The pump power requirements plotted in Figure 4b align with state-of-the-art system efficiencies at lower transport rates for our assumed stack geometry. A stack with 1.4 V OCP, 0.4 cm 2 ASR, 91.6% voltage efficiency, aspect ratio of 1.2, and 60 cells delivers a total power of ≈ 18.3 kW, indicating that ≈ 0.92 kW of pump power, for two electrolyte streams, can be afforded to maintain ≤ 5% system efficiency loss. In subsequent cost calculations, we will assume an optimistic system efficiency of 94%, which accounts for energy losses due to pumping, power electronics, and thermal management equipment. Critically, when designing a specific, new electrolyte, the pump power requirement must be compared with the nominal battery power output to ensure high system efficiency. For example, a battery with relatively low cell potential and low mass transfer rates (e.g., high viscosity, poor flow field design) could engage a RFB design where the pump power actually exceeds the power output from the electrochemical stack. Identifying the viable regions of the efficiency design space is outside the scope of the present analysis, but must be considered when evaluating new, specific RFB chemistries.

Techno-Economic Modeling
The following sections describe the integration of the flow battery ASR model and pump power requirement calculation with a techno-economic model that evaluates RFB cost as a function of many design, chemical, and cost input parameters. Benchmarking the quality of a techno-economic model is challenging considering the lack of available detailed engineering designs, costs, and profit margins for commercially available RFBs. In an attempt to validate the following techno-economic model, we input literature parameters Reactor cost.-To begin bridging membrane and supporting electrolyte selection to RFB cost, the full-cell ASR can be incorporated into a description of RFB reactor cost. In the present analysis, RFB reactor cost (C reactor , $ kW −1 ) is defined similarly to prior literature (Equation 27), 23,26 where U is the cell potential (V), ε sys is the system efficiency (-), and ε v is the voltaic efficiency (-). Additionally, c stack is the areal cost of the electrochemical stack materials ($ m −2 ), including bipolar plates, gaskets, electrodes, and current collectors, and c mem is the areal membrane (or separator) cost ($ m −2 ). [27] Note that in the present analysis, we separate the stack hardware cost from the membrane cost to offer a more flexible framework for incorporating various membranes. This treatment contrasts prior studies that lump these two costs into one term. 23,26 We also consider present and future-state areal reactor and membrane costs, since these materials costs are anticipated to decrease with bulk purchasing required for future large-scale manufacturing, which will be described later (see Unit cost less materials section). Estimates of the electrochemical stack costs and IEMs are adapted from prior literature for present and future-state (Table IV). 23,24 Note that emerging IEM technologies, such as hydrocarbon-based options, 93 could reduce membrane costs relative to Nafion. Estimates for present and future-state costs of SSS are assumed to be that of reverse-osmosis membranes (≈ $30 m −2 ) 94 and lithium-ion battery separators (≈ $1 m −2 ), 23 respectively. The voltaic efficiency is assumed to be 91.6%. 23 R DC values are extracted from Figure 3. Hence, the reactor cost is described as a function of the full-cell ASR. As R DC increases, a larger area electrochemical stack is required to drive the same total current at the same voltaic efficiency, yielding a more expensive reactor overall. Figure 5a shows reactor cost as a function of cell potential for the various working ion/membrane combinations for present-day costs of the electrochemical stack and membrane. The extreme values of Re can be interpreted as electrolytes at fixed flow rate with viscosities of 1.0 and 10 mPa · s, respectively, which are the assumed upper and lower bounds on AqRFB electrolyte viscosity. As indicated by Equation 27, reactor cost decreases as U −2 and thus the reactor cost sensitivity to variations in cell performance, caused by the working ion type or electrolyte viscosity, decreases as cell potential increases. For the IEM cases, changing working ion type or electrolyte viscosity can vary reactor cost by over $900 kW −1 at lower cell potentials (< 1.0 V). Despite the need for large molecules, the SSS offer the lowest reactor costs, across all cell potentials, for present-day production , as the combined stack and membrane cost (c stack + c mem ) per unit area decreases, the total reactor cost (C reactor ) becomes less sensitive to differences in R DC . Figure 5b explicitly illustrates how the total reactor cost grows as a function of c stack + c mem for the different membrane/working ion configurations, where larger values of R DC necessitate that the reactor cost grows more quickly with larger c stack + c mem . This sensitivity to the stack and membrane costs indicates that RFBs with more expensive cell components Figure 5. Present-day reactor cost as a function of (a) cell potential or (b) areal reactor cost for various working ion/membrane types and two extreme values of Re. For $ kWh −1 cost estimates, a 5 h discharge time is assumed. must offer lower cell ASR to keep reactor costs down. Considering that present-day stack and membrane costs are much higher than their anticipated future-state costs at mass-scale production, engineering low ASR cells today via rational selection of membranes and supporting salts is of the upmost importance in facilitating early adoption of RFBs. Given that SSS are anticipated to be more than an order-ofmagnitude less expensive than IEMs, assuming present cost estimates, the SSS is an attractive option to pursue when reactor costs (c a , $ m −2 ) are high. 28 describes the total electrolyte cost normalized by the energy delivered upon discharge (C electrolyte , $ kWh −1 ) with the following parameters: ε q is the coulombic efficiency (-), F is the Faraday constant (0.0268 kAh), M is the molar mass of the active material (kg mol −1 ), s is the stoichiometric coefficient of the discharge reaction (-), χ is the depth-of-discharge (-), n e is the number of electrons stored per mole of active material (-), c m is the active species cost per unit mass ($ kg −1 ), r salt is the arithmetic mean ratio of moles of salt per mole of active species across both electrolytes (-), 26 M salt is the molar mass of the supporting salt (kg mol −1 ), c salt is the salt cost per unit mass ($ kg −1 ), c t is the tank cost per unit volume ($ L −1 ), andc is the mean active species molarity across both sides of the cell (M). Note that the present model explicitly accounts for the salt and tank cost contributions to the electrolyte cost, in contrast with prior reports. 23,26 The + / − subscripts denote the positive/negative sides of the battery, respectively. The cost of water as a solvent is neglected in this estimate as the cost contribution of deionized water (≈ $0.001 kg −1 ) 95 is at least one order-of-magnitude lower than all other electrolyte components. Tank cost is fixed at $0.15 L −123 for all electrolytes under consideration since inexpensive polypropylene or polyethylene tanks should be chemically compatible across all pH. Additionally, the coulombic efficiency and allowable SOC range are set to 97% and 80%, respectively. 23

Electrolyte cost.-Equation
Since one motivation for designing a new AqRFB electrolyte is to leverage inexpensive active materials, we assume an optimistically low cost for the active material of $5 kg −1 . 23 Precursors to some proposed ROMs come close to this value today. 26 For example, anthraquinone is a precursor to multiple proposed ROMs 37,39 and costs ≈ $4.40 kg −1 . 39 For comparison, vanadium sulfate costs ≈ $22 kg −123 and sulfur (S 8 ) costs ≈ $0.20 kg −1 . 96 The active species molecular weight is another key parameter that defines electrolyte cost because the molecular weight normalized by the number of electrons transferred (equivalent weight (g mol e-−1 )) specifies the mass of active material that must be purchased to store a certain amount of charge. In the present analysis, we assume a moderate equivalent weight of 100 g mol e-−1 for small active species. 26 For comparison, the equivalent weight of vanadium is 51 g mol e-−1 , whereas AQDS has an equivalent weight of 184 g mol e-−1 , assuming that AQDS has 2e − transfer capability. 39 In the case of RFBs with SSS, we estimate that the active species equivalent weight must be ≈ 1.25 × that of a small active species, paired with an IEM, to successfully implement size exclusion; this equivalent weight increase is similar to that of the added molecular weight imposed by linkers between redox-active centers on oligomerized ROMs. 60 Finally, the active-species concentration is set to 1.5 M, for consistency with previous cell polarization calculations, which defines the total tank size required to store a particular amount of energy. Other electrolyte cost parameters are listed in Table III.
We choose to acknowledge the cost contribution of the supporting salt, which is typically overlooked, to evaluate its relative magnitude in determining RFB cost. 26 The molar masses and costs for H 2 SO 4 , NaCl, and KOH are listed in Table V. Figure 6 shows how electrolyte cost varies as a function of r salt for the salt types listed in Table V for small active species with H 2 SO 4 , NaCl, or KOH as the supporting salt, or large active species with H 2 SO 4 or NaCl. r salt represents the number of moles of salt per mole of active species on one side of the RFB, and typical r salt values for VRFB electrolytes can vary between 0.7-2. 80,97 For comparison, recent studies on new AqRFB active species experimentally illustrated that by employing multifunctional, ionic active species, the need for a separate supporting salt can be eliminated. 30,98 The slopes of the electrolyte cost curves in Figure 6 are defined by the salt costs ($ kg −1 ), and the y-intercepts ( Figure 6) represent the cost contributions from the tanks and active species. As such, the higher equivalent weight active species yield a larger electrolyte cost in the limit of zero salt. Overall, variations in electrolyte cost with supporting electrolyte type and amount are relatively small in comparison to other cost contributions for AqRFBs.

Balance of plant.-
The balance-of-plant (BOP) costs (C BOP , $ kW −1 ) for RFBs typically comprise costs associated with ancillary equipment, normalized by the power output of the RFB stack. In the  present work, the BOP cost accounts for thermal-management systems (e.g., heat exchangers), controls, and pumps. Note that this work does not consider power-conditioning equipment (e.g., inverters) or installation costs. 26 Since we do not anticipate the cost of thermal management or controls to vary significantly with supporting electrolyte composition, we assume that their cost (C control ) is $60 kW −1 , as estimated in a recent RFB techno-economic analysis. 23 Note that in the particular case of thermal management equipment, an analysis of the thermal balance of a RFB is presently unavailable, and additional research is required to define the size, design constraints, and costs associated with this equipment. Electrolytes that are tolerant to large temperature variations, unlike the VRFB, 18 could reduce the controls cost contribution to the BOP. The total C BOP is calculated as shown in Equation 29,23 where N pump is the number of pumps required to operate the RFB and C pump is the pump cost per power ($ kW −1 ). Since the BOP cost is normalized by the RFB stack power, N pump = 2 because each electrolyte stream will require one pump. [29] Pump cost (c $ pump , $) is estimated using the cost correlation in Equation 30,91 where P is the required pump power (kW). The pump cost ($) as a function of required pump power is depicted in Figure  7a.
c $ pump = −1100 + 2100 · P 0.6 [30] We normalize the pump cost by the power capability of the electrochemical stack (Equation 31). The variations in battery energy efficiency with changes in pumping energy requirements are typically small 56,57 and would also require details of the battery's thermal management system. Thus, for simplicity, we choose to hold ε sys constant at 94%. 23 Figure 7b shows how pump costs ($ kW −1 , left axis) vary as a function of cell potential for the upper and lower Re values considered in this work. As the reactor performance improves via increased cell potential, the effective pump cost decreases because less total pumping power is required to deliver a certain power upon discharging the battery. Additionally, the high Re case (e.g., low viscosity) has a much weaker dependence of pump cost on cell potential because of a small pressure drop and improved mass transfer rates. Note that when normalizing the pump cost by discharge time ($ kWh −1 , Figure  7b right axis), the pump cost is < $2 kWh −1 for nearly all design conditions, indicating that pump costs are small as compared to reactor ( Figure 5) and materials costs ( Figure 6) in the target discharge duration time. Thus, although varying electrolyte viscosity will impact pumping costs, the overall contribution to battery cost will be relatively small; viscosity has a significantly larger impact on cell ASR, which affects the reactor cost contribution.
Unit cost less materials.-The final cost contributions to consider for estimating RFB capital cost are termed unit cost less materials (C UCLM ), 25 which is also sometimes referred to as additional costs. 23,26,101 The unit cost less materials accounts for all RFB costs excluding purchased goods. Purchased goods refers to all RFB parts manufactured by specialty firms and implemented in the electrochemical stack, electrolyte, or BOP. 25 For instance, carbon-paper electrodes are an example of a purchased good that will be used in the electrochemical stack, but the manufacturing cost of implementing the carbon paper in the stack falls under the unit cost less materials. The unit cost less materials in this work specifically accounts for depreciation, research and development, sales, administration, variable overhead, direct labor, and warranty costs. 25 Note that the present analysis excludes an estimate of profit margin, which has been incorporated in prior analyses. 23,25,26,101 As such, this work considers battery cost, as opposed to battery price.
The unit cost less materials depends on the production volume of RFB components. As production volume increases, the unit cost less materials will decrease due to increased utilization of capital-intensive manufacturing infrastructure. As before, we first consider a presentday estimate of unit cost less materials at the low RFB production volumes at the time of publication. We also consider a future-state estimate of unit cost less materials, assuming that RFB annual production achieves the volume to store ≈ 1% of the world's energy consumption in 2013 for 5 h (2 GW, 10 GWh). 25,102 The presentday estimate of unit cost less materials is taken to be $1550 kW −1 , which was originally computed by engaging a gap analysis between present-day costs of RFB materials and costs of energy storage systems in the field. 23 For the future-state estimate, we assume values calculated by Ha and Gallagher, which are listed in Table VI for convenience. 25 Battery cost.-Now that the reactor, electrolyte, BOP costs, as well as the unit cost less materials, have been computed for all membranes and supporting electrolytes under investigation, these cost contributions can be combined in Equation 32 to describe the battery cost. The total RFB cost (C battery , $ kWh −1 ) is defined as the total cost of the battery normalized by the energy delivered upon discharge. Since the reactor, BOP, and additional costs are all defined in units of $ kW −1 , their costs are normalized by a total discharge time (t d ) of 5 h. The DOE target capital cost of $150 kWh −1 includes the costs of power conditioning equipment and installation, which has been estimated to be ≈ $250 kW −1 , 23 so the $150 kWh −1 DOE target translates to a battery cost of ≈ $100 kWh −1 in the present analysis. 26 Further, note that the DOE target cost is an aggressively low value that would enable ubiquitous adoption of an energy storage technology for nearly all grid-level services. Some grid-level energy storage services are more valuable than others (e.g., frequency regulation), and estimated affordable costs of energy storage for specific grid services are available in market reports, such as Lazard's Levelized Cost of Energy Storage. 92 Figure 8 shows battery costs as a function of cell potential for the membrane and supporting salt combinations under consideration, along with a low and high value of Re number, corresponding to a 10 × viscosity change. Present-day ( Figure 8a) and future-state estimates (Figure 8b) illustrate how changes in stack and membrane cost with production volume impact battery cost. Along with Figure 8, Table  VII highlights select RFB design iterations, varying cell potential, salt identity, membrane type, and electrolyte viscosity to quantitatively illustrate the range of impact on ASR, C electrolyte , C reactor , and C battery . For present-day, differences in the RFB performance and electrolyte costs can manifest as cost differences >$300 kWh −1 , in the limit of low cell potentials (< 1 V). At high cell potentials (i.e., 1.4 V), the differences are smaller, but still reaches ≈ $100 kWh −1 between the SSS and K + -IEM designs (Table VII). Viscosity changes can induce a difference in battery cost as high as $30 kWh −1 for designs with higher R DC and low cell potential. Surprisingly, however, a 10-fold variation in electrolyte viscosity has a relatively small impact on RFB cost, as quantitatively indicated in Table VII. For present-day costs, RFBs implementing SSS have an advantage over the IEM options. The separator cost is much lower than that of an IEM; we calculate that the SSS options are least expensive across the design space. The use of the higher conductivity electrolyte (H 2 SO 4 ) with the SSS offers marginally better cost performance due to reduced ASR. When the stack and IEM costs decrease, as assumed for future state, SSS options with high electrolyte viscosity perform worse than IEMs with H + or Na + charge carriers due to the slower transport and higher materials cost associated with large active species.
The future-state cost estimates (Figure 8b) only show the high Re case, and all cost curves are much closer together as compared to present-day estimates. The smaller differences ($10 -30 kWh −1 ) in future-state battery cost stem from the lower anticipated stack and membrane costs, which makes the battery cost less sensitive to variations in reactor performance (Figure 5b). In future state estimates, chemical costs comprise a larger fraction of the total battery cost, resulting in the following notable trend. While the H + -SSS option yields a lower ASR than the H + -IEM and is anticipated to be less expensive, the SSS requires utilization of larger active materials, with higher molecular weights, than the IEM-based designs. Hence, in the future state cost projections, the higher chemical cost associated with the larger active material results in the H + -SSS being marginally more expensive than the H + -IEM according to future state costs. Finally, despite substantial cost future state cost reduction, battery costs only surpass the $100 kWh −1 guideline for cell potentials >1.2 V, and the H + -IEM offers the best opportunity to exceed the benchmark cost, whereas the K + -IEM only reaches the benchmark cost with precisely a 1.5 V cell.

Conclusions
RFBs are promising electrochemical devices for grid-scale energy storage, but their capital costs must be reduced for ubiquitous adoption. As such, recent reports have investigated new active species for RFBs geared toward lower cost or improved performance. The identification of a promising active material typically involves tailoring the supporting electrolyte to optimize stability, solubility, and reaction kinetics. The choice in electrolyte composition, however, can have a significant impact on the low cost potential of a new redox chemistry, which, to date, has yet to be systematically evaluated.
This work quantifies changes in RFB cost performance for different supporting electrolytes paired with ion-selective membranes and size-selective separators. Membrane ASR is derived from RFB reports employing cation or anion exchange membranes with various chargecarriers (H + , Na + , Cl − , and K + ) in concentrated electrolytes. The ASR contribution of the RFB electrodes is quantified by implementing a steady-state, one-dimensional porous-electrode model that describes cell overpotential as a function of current density, incorporating losses from the electrolyte resistivity, Butler-Volmer reaction kinetics, and convective mass transfer. The consideration of mass transfer losses al-lows us to link the description of electrode resistance to the electrolyte viscosity through a mass transfer coefficient power law correlation. The physical description of cell ASR is then integrated into a technoeconomic model that estimates RFB cost for the different membrane and supporting electrolyte options under consideration, accounting for the reactor, electrolyte, BOP, and additional costs. Variations in cell performance due to the working ion selection and electrolyte viscosity can yield battery cost differences in the $100's kWh −1 , and this analysis allows for quantification of cost performance changes by selecting certain electrolyte characteristics. The battery cost curves in Figure 8 can also be used to quantify how much extra cell potential is required to overcome a performance loss due to selecting a less mobile ion or more viscous electrolyte.
Beyond the conventional RFB design incorporating small active species and an IEM, this work also considers the combination of nanoporous separators, with no functionalization for ion selectivity, and larger active species that cannot pass through the separator pores. Supporting electrolyte ions, however, can pass freely through the SSS, imparting higher ionic conductivities that their IEM counterparts. Drawbacks of the SSS concept are that the larger active species will exhibit slower transport rates and will have higher active material costs, but these performance setbacks are offset by the lower cost associated with the SSS. With the benchmark performance and cost values assumed in this analysis, the H + -IEM option offers the best opportunity to achieve the ultimate DOE target capital cost at cell potentials >1.2 V, however, for present costs, SSS combined with low cost active materials and electrolytes are an attractive design option.

Symbol
Description Units