Journal of Power Sources 330 (2016) 261e272

Contents lists available at ScienceDirect

Journal of Power Sources

journal homepage: www.elsevier.com/locate/jpowsour

Cost-driven materials selection criteria for redox flow battery

electrolytes
Rylan Dmello a, b, 1, Jarrod D. Milshtein a, c, 1, Fikile R. Brushett a, d, Kyle C. Smith a, b, *
a
Joint Center for Energy Storage Research, USA
b
Department of Mechanical Science and Engineering and Computational Science and Engineering Program, University of Illinois at Urbana-Champaign,
Urbana, IL, 61801, USA
c
Department of Material Science and Engineering, Massachusetts Institute of Technology, Cambridge, MA, 02139, USA
d
Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, MA, 02139, USA

h i g h l i g h t s

We develop a techno-economic model for redox flow battery (RFB) electrolytes.

Electrolyte costs account for redox active material, salt, and solvent components.

Reactor, balance-of-plant, and additional costs are considered.

Design maps to build $100 kWh-1 RFBs guide electrolyte materials selection.

Multiple pathways to decreasing RFB electrolyte costs are proposed .

a r t i c l e i n f o a b s t r a c t

Article history: Redox flow batteries show promise for grid-scale energy storage applications but are presently too
Received 27 June 2016 expensive for widespread adoption. Electrolyte material costs constitute a sizeable fraction of the redox
Received in revised form flow battery price. As such, this work develops a techno-economic model for redox flow batteries that
10 August 2016
accounts for redox-active material, salt, and solvent contributions to the electrolyte cost. Benchmark
Accepted 30 August 2016
values for electrolyte constituent costs guide identification of design constraints. Nonaqueous battery
Available online 13 September 2016
design is sensitive to all electrolyte component costs, cell voltage, and area-specific resistance. Design
challenges for nonaqueous batteries include minimizing salt content and dropping redox-active species
Keywords:
Redox flow battery
concentration requirements. Aqueous battery design is sensitive to only redox-active material cost and
Electrolyte cell voltage, due to low area-specific resistance and supporting electrolyte costs. Increasing cell voltage
Techno-economic model and decreasing redox-active material cost present major materials selection challenges for aqueous
Energy storage batteries. This work minimizes cost-constraining variables by mapping the battery design space with the
techno-economic model, through which we highlight pathways towards low price and moderate con-
centration. Furthermore, the techno-economic model calculates quantitative iterations of battery designs
to achieve the Department of Energy battery price target of $100 per kWh and highlights cost cutting
strategies to drive battery prices down further.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction energy storage can alleviate the intermittency of renewable energy

technologies (i.e., wind, solar), facilitating broad implementation
Grid-scale energy storage technologies are becoming increas- [1]. Second, storage can also improve the cost-effectiveness of the
ingly critical to promoting sustainable electricity generation. First, existing fossil fuel infrastructure, decreasing electricity costs via
load-leveling and price arbitrage operations [2]. Third, grid-scale
storage can provide high value services such as back-up power,
frequency regulation, and voltage support [2]. Redox flow batteries
* Corresponding author. University of Illinois at Urbana-Champaign, Department
(RFBs) provide a promising technological pathway towards low-
of Mechanical Science and Engineering, 105 S. Matthews Ave., Urbana, IL, USA.
E-mail address: kcsmith@illinois.edu (K.C. Smith). cost grid-scale energy storage devices due to decoupled capacity
1
These authors contributed equally to this work. and power, long lifetimes, and facile thermal management [3e8].

http://dx.doi.org/10.1016/j.jpowsour.2016.08.129
0378-7753/© 2016 Elsevier B.V. All rights reserved.
262 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272

Current RFB prices, however, are too high for market penetration developed a comprehensive TE model (hereafter referred to as the
[9e11]. According to the Department of Energy (DOE) Office of DG model) to compare the price performance of aqueous and
Electricity Delivery and Energy Reliability, decreasing RFB system nonaqueous RFBs [9]. The DG model defined benchmark values for
price to $120 kWh-1 in the near term will enable widespread redox-active material concentration, molecular weight, cell voltage,
adoption for 4 h discharge grid-scale energy storage applications and area-specific resistance (ASR), for both families of RFBs, to
[3,9]. In comparison, the DOE's Advanced Research Projects Agency reduce battery price to $100 kWh-1. Although instrumental in
- Energy (ARPA-E) suggests a long-term energy storage battery elucidating future RFB prices, the DG model focused on a single set
price target of $100 kWh-1 for 1 h of discharge [12]. In 2014, RFB of benchmarks but did not explore alternative design iterations. A
prices exceeded $500 kWh-1 [9,10]. Despite the existing high prices, recent investigation into separator performance characteristics for
recent work has established that both aqueous and nonaqueous RFBs considered the tradeoffs among cell voltage, ASR, and reactor
RFBs can meet the desired $100 kWh-1 battery price target by cost [41], but no such sensitivity analysis has accounted for the
appropriately decreasing RFB reactor and materials costs contri- relative cost contributions from the electrolyte constituent mate-
butions [9]. rials: solvent, salt, and redox-active compounds.
To achieve the price target, the price-to-energy ratios of aqueous The present work addresses the lack of RFB design strategies by
and nonaqueous RFBs can drop by following different cost reduc- exploring the materials space mapped by an electrolyte-centric TE
tion pathways that optimize their fundamentally different reactor model, which identifies new RFB price reduction strategies. A
and materials characteristics [9]. Aqueous RFBs (AqRFBs) leverage detailed electrolyte cost model, explicitly accounting for redox-
inexpensive electrolytes, utilizing water as the solvent and typically active species, salt, and solvent cost contributions, combined with
a low-cost inorganic salt (e.g., H2SO4, KOH, and NaCl), while the existing DG model, enables a sensitivity study of aqueous and
exhibiting high power density due to low cell resistance. The typical nonaqueous RFB prices to various material and cost parameters. We
electrochemical stability window of water (less than 1.5 V), how- explore the available RFB design space and investigate the sensi-
ever, limits the maximum achievable AqRFB electrolyte energy tivity of both aqueous and nonaqueous RFBs to pertinent electro-
density. In contrast, nonaqueous RFBs (NAqRFBs) employ lyte constituent cost variables, cell voltage, and ASR. Further, maps
nonaqueous solvents with wide electrochemical stability windows of the available design space translate abstract price targets into
(3e4 V) and can thus enable electrolytes with greater energy quantitative performance targets, bridging the TE model to proto-
density as compared to aqueous systems. Despite attractive voltage type guidelines. As such, this paper demonstrates tradeoffs in RFB
capabilities, NAqRFBs suffer from relatively expensive nonaqueous constituent costs and performance to achieve a $100 kWh-1 battery
solvents (e.g., nitriles, glymes, and carbonates) and fluorinated salts price. While previous modeling efforts have highlighted cost per-
(e.g., tetrafluoroborates, hexafluorophosphates, and bis(tri- formance challenges with specific RFB chemistries (e.g. all-
fluoromethylsulfonyl)imides), as well as low power density due to vanadium, zinc-bromine) [9e11,40], our analysis culminates in a
low membrane conductivities. Considering the advantages and set of design maps to aid in selecting materials for new RFB elec-
drawbacks of each system, AqRFB cost cutting efforts should trolytes. We also suggest research pathways to most easily achieve
maximize cell voltage, while NAqRFB design should decrease the near-term target battery price ($100 kWh-1) and lower. This
electrolyte cost and improve power density. electrolyte-centric analysis can guide future research efforts in the
Redox-active materials for both families of RFBs require development and selection of new, promising materials for use in
continued research and development for widespread adoption. economically viable RFB prototypes.
Inorganic non-metallic (e.g., polysulfide-bromine) and transition
metal (e.g., all-vanadium) redox-active materials have traditionally
2. Methodology
been at the forefront of AqRFB development, although metal co-
ordination complexes have also been explored [7,13,14]. AqRFBs
2.1. Model definitions
utilizing certain inorganic, non-metallic redox-active materials,
such as bromine, have failed to penetrate the market due to their
Redox flow battery price is defined as the RFB's future-state
corrosive and toxic nature, making the practical design of flow
battery price P0 (excluding power conditioning systems) per unit
fields, pumps, storage tanks, and pipes difficult [15]. Additionally,
discharge energy Ed, delivered over a time td. The present TE model
transition metal based AqRFBs have struggled to achieve the bat-
(which builds on the DG model [9]) separates RFB price into four
tery price targets due to the high cost and limited abundance of the
major cost contributions from the reactor CReactor, electrolyte CEle-
redox-active material [6]. Early investigations into NAqRFBs
employed metal coordination complexes as redox-active materials ctrolyte, additional CAdditional, and balance-of-plant (BOP) CBOP:

that suffer from low solubility, poor stability, or expensive pre- P0

cursors [16e19]. A significant portion of recent RFB progress ¼ CReactor þ CElectrolyte þ CAdditional þ CBOP (1)
Ed
beyond vanadium RFBs, arguably the current state-of-the-art sys-
tems, has aimed at identifying low-cost redox-active materials such Table 1 provides variable definitions for all cost equations, as
as abundant inorganic species [20,21] and tailored organic mole- well as benchmark values and units. Here, a series of design maps
cules [22e34]. Organic redox-active molecules are particularly are presented in which certain model parameters vary. In addition,
attractive for use in both aqueous and nonaqueous RFBs; organic the supplementary information contains a MATLAB script that
molecules are comprised of earth abundant elements (e.g., generates the design maps presented here. The parameters that do
hydrogen, carbon, oxygen, sulfur) and offer a broad design space, not vary in the design maps assume benchmark values (Table 1),
allowing for rational control of molecular weight, solubility, and unless otherwise explicitly stated. In these design maps, thin dotted
redox potential by molecular functionalization [35]. black lines denote benchmark values from the original DG model.
RFB price relates to experimentally measurable chemical prop- This work builds on the reactor, additional, and BOP cost de-
erties, electrochemical performance, and cost parameters that scriptions from the DG model. In the DG model [9], the reactor cost
serve as critical inputs towards developing RFB cost projections via (in $ kWh1) depends on the reactor cost per unit area ca, which
a techno-economic (TE) model. TE models have quantified the price incorporates the costs associated with bi-polar plates, membranes,
performance of transportation [36,37] and grid-scale [9e11,38e40] and seals; the cost of each reactor hardware component is detailed
energy storage devices. In 2014, Darling, Gallagher, and co-workers in Ref. [9]. Additionally, the reactor cost varies with area-specific
 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272 263

Table 1 theoretical capacity accessed):

Parameters utilized in the present techno-economic model. Benchmark values were
obtained from the DG model [9]. *These material-specific targets were set in Ref. [9]

1 sþ Mþ s M
as guidelines to achieve $100 kWh-1 battery price, based on the values of the other CElectrolyte ¼ cm;þ þ c
parameters listed here. εsys;d εq;rt Fεv;d U cþ neþ c ne m;

2
Modeling parameter Benchmark values þ 2ravg Msalt csalt þ c (3)
bavg solvent
Nonaqueous Aqueous

Reactor parameters The redox-active materials used in either the positive or nega-
Cost per unit area, ca $107.5 m2 $122.5 m2 tive electrolytes (denoted with ‘þ’ or ‘’ subscripts, respectively)
Area-specific resistance, R 5.0 U-cm2 0.5 U-cm2 store ne electrons per s formula units of the particular redox-active
Open-Circuit Cell Voltage, U 3V 1.5 V
Discharge time, td 5h 5h
species (also called the stoichiometric coefficient) that has molec-
System discharge efficiency, εsys,d 0.94 0.94 ular weight M with a cost per unit mass cm. Equation (3) also ac-
Voltage discharge efficiency, εv,d 0.916 0.916 counts for the benefits of employing a cheaper salt or solvent in the
Electrolyte parameters RFB. Several variables specify the type and amount of salt
Round-trip coulombic efficiency, εq,rt 0.97 0.97
employed, including the molar ratio of salt to redox-active species r
Stoichiometric coefficient, s± 1 1
Allowable state-of-charge range, c± 0.80 0.80 (in moles of salt per mole of electrons stored), salt molecular
Actives molecular weight,* M± 100 g mol1 100 g mol1 weight Msalt, and the salt cost per unit mass csalt. Solvent costs
Actives cost per unit mass, cm,± $5 kg1 $5 kg1 depend on the redox-active species concentration b (molality in
Electrolyte cost per unit mass, cme,± $5 kg1 $0.1 kg1 units of moles of electrons transferred per kilogram of solvent) and
Actives solubility,* S± 1.0 kg kg1 0.2 kg kg1
the cost per unit mass of the solvent csolvent. Note here that the
Additional parameters
Addition to price, cadd $112.5 kW1 $87.5 kW1 molar ratio of salt to redox-active species and the redox-active
Balance-of-plant cost, cbop $102.5 kW1 $102.5 kW1 species concentration appear as average values of the positive
DG model parameters and negative electrolytes, assuming the symbols ravg and bavg,
Salt cost per unit mass, csalt $20 kg1 e
respectively [ravg is an arithmetic mean with ravg ¼ ðrþ þ r Þ=2,
Solvent cost per unit mass, csolvent $2 kg1 $0.1 kg1
Salt solubility,* Ssalt,± 0.16 kg kg1 e while bavg is a harmonic mean with bavg ¼ 2bþ b =ðbþ þ b Þ]. The
Mean molar salt ratio,* ravg 0.20 mol mol1 e Appendix relates these variables (ravg and bavg) to parameters from
Salt molecular weight, Msalt 100 g mol1 e the DG model. This approach also accounts for the ion-transfer
Mean actives molality,* bavg 9.6 mol kg1 1.6 mol kg1 configuration of the RFB by modeling salt concentration varia-
*
These material-specific targets. tions. This capability enables comparison between rocking-chair
and salt-splitting ion transfer configurations that are discussed in
detail later (Section 3.2).
resistance R (including resistance contributions from the mem-
brane, porous electrode, reaction kinetics, mass transfer, and elec- 2.2. Material cost assumptions
trical contacts), open-circuit cell voltage U, discharge voltage
efficiency εv,d, system efficiency during discharge εsys,d (including The costs-per-unit-mass ($ kg1) of redox-active materials, salts,
losses due to auxiliary equipment and pumping), and discharge and solvents in Table 1 assume future commodity-scale production
time td: and can vary depending on the choice of material. While these costs
may appear optimistic, certain materials could achieve these values
ca R today; for example, the 2006 bulk price of acetonitrile, a typical
CReactor ¼
(2) NAqRFB solvent, was ~$1.50 kg1 [43]. Additionally, the two-fold
εsys;d U 2 εv;d 1  εv;d td
decrease in lithium-ion battery electrolyte (1 mol L1
The balance-of-plant cost accounts for the ancillary equipment salt þ solvent) costs from $40 kg1 to $18 kg1 between 1999 and
(such as pumps, controls, sensors, and pipes) required to build a 2011 [37,44], suggests that other solvent and salt costs could real-
working system, while the additional cost accounts for economic istically fall to the values listed in Table 1 over the next decade.
factors like depreciation, overhead, labor, and profit margin. A full Redox-active materials have an estimated future-state cost of
discussion of these costs is provided in Ref. [9]. Importantly, this $5 kg1, however, tailored molecules can cost more if complicated
work considers a battery price, which excludes power conditioning synthetic procedures are required for manufacture [9]. Anthraqui-
systems (i.e., inverters) and installation costs [9]. The battery price none, a precursor to several other redox-active materials
is not to be confused with the system price, which does include [22,23,45,46], has an estimated price of ~$4.40 kg1 [23,47].
power conditioning systems costs, as the system price may be the Alternatively, inherently low-cost materials, such as those con-
focus of other techno-economic modeling literature. The $120 taining sulfur [48] or bromine [23] (the 2006 prices of S8 and Br2
kWh-1 system price target used in the DG model is thus converted were $0.20 kg1 and $1.41 kg1, respectively [43]), can decrease
to a $100 kWh-1 battery price target by excluding a power condi- future-state costs compared to tailored redox-active molecules.
tioning system that costs $100 kW1 for 5 h of discharge [42]. Furthermore, this study considers RFBs implementing a fluorinated
A new model is used for the electrolyte cost CElectrolyte, in $ ion-exchange membrane (i.e., Nafion) with an estimated future-
kWh1, that incorporates cost contributions from individual ma- state price of $50 m2 [9]. Our study does not consider how vari-
terials in the electrolyte. Specifically, the costs from redox-active ations in membrane cost affects the available RFB design space
materials (used in the positive and negative electrolytes), sup- because an excellent prior study (Ref. [40]) offers a comprehensive
porting salt, and solvent are included explicitly. This electrolyte cost analysis of tradeoffs in membrane cost and performance.
model captures the state of RFB materials as purchased from a
chemical supplier. Further, the model normalizes the electrolyte 3. Results and discussion
materials costs by the total system energy, accounting for discharge
efficiencies (as included in the reactor cost), round-trip coulombic 3.1. Mapping the RFB materials design space
efficiency εq,rt (accounting for crossover and shunt current effects),
open-circuit cell voltage, and depth-of-dischargec (the fraction of We map the available materials design space for aqueous and
264 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272

nonaqueous RFBs, within physical reason, to achieve a $100 kWh-1

battery price. The analysis presented here remains within a design
space commonly accessible by laboratory and industrial scale RFBs,
even though extreme RFB electrolyte systems may be possible. To
begin, this work explores the tradeoffs among cell voltage, redox-
active material molecular weight, and redox-active material con-
centration for both aqueous and nonaqueous RFBs. Contours of
constant concentration in Fig. 1a represent possible RFB designs
with a $100 kWh-1 price. Note that design maps throughout this
work present molecular weight in units of “grams per mole of
electrons” to accommodate active species that undergo multiple
electron transfer events.
The thermodynamic dissociation potential of water is 1.23 V, but
due to the sluggish kinetics of the hydrogen and oxygen evolution
reactions on porous carbon electrodes, AqRFB cell voltages can
typically reach 1.5 V (Fig. 1a, horizontal dashed line) [49]. In some
exceptional battery systems, such as lead-acid and zinc-bromine,
the water stability limit has exceeded 1.7 V [7,50]. RFBs with cell
voltages greater than 2 V, however, will require the use of an aprotic
nonaqueous solvent [9], which can easily exhibit electrochemical
windows from 3 to 4 V [51]. Each concentration contour in Fig. 1a,
for both aqueous and nonaqueous RFBs, demonstrates that as the
molecular weight of the redox-active species increases, the
required cell voltage to achieve $100 kWh-1 also increases.
Increasing molecular weight subsequently increases RFB price (in
$). Therefore, to offset higher redox-active material costs, the cell
voltage must also rise, increasing the available system energy while
simultaneously decreasing reactor and electrolyte costs (in $
kWh1).
Notably for NAqRFBs, as the redox-active species molality de-
creases, either the cell voltage must increase or the molecular
weight must decrease significantly to attain the target battery
price, and this sensitivity results from higher solvent costs ($2 kg1)
than water. Redox-active species molality is directly proportional to
electrolyte energy density, which subsequently defines the total
available energy of the RFB. As redox-active material concentration
decreases, the volume of electrolyte required to achieve a fixed
system energy increases, and subsequently the amount and total
cost of solvent also increases. Therefore, to achieve the target price,
the battery energy must increase via a voltage increase, or the
electrolyte cost must reduce by utilizing redox-active compounds
with lower molecular weight. As a quantitative nonaqueous
example, a 100 g mol1 redox-active material at 20 mol kg1
concentration requires a 2.8 V cell, but the same redox-active ma-
terial requires a 4 V cell if the operating concentration drops to
2 mol kg1. Further, the NAqRFB design space is insensitive to
molality changes at high redox-active material concentrations
(greater than 20 mol kg1) because, in this regime, the solvent cost
contribution approaches zero. Ultra-high NAqRFB concentrations
(greater than 200 mol kg1) correspond to redox-active materials
in near-neat form with minimal solvent content, and only liquid
redox-active species can achieve such high concentrations. Liquid Fig. 1. Contours of constant redox-active species concentration for aqueous and
redox-active species are a new concept for NAqRFBs, demonstrated nonaqueous RFBs as a function of cell voltage and (a) redox-active species molecular
weight or (b) reactor ASR. Each contour achieves a battery price of $100 kWh-1. Con-
in Ref. [32]. Due to the decreasing solvent cost contribution at ultra- tours above 2 V represent the NAqRFB feasible design space, while the shaded region
high redox-active material molality, the concentration contours for below 1.5 V represents the AqRFB design space. Horizontal dotted lines at 1.5 V and 2 V
such redox-active liquids will closely match the 200 mol kg1 denote the typical electrochemical stability window and upper stability limit of
contour, and thus, the NAqRFB feasible region in Fig. 1a, highlighted aqueous electrolytes, respectively. (c) Contours of constant cell voltage as a function of
in green, exists above the 200 mol kg1 contour. Additionally,
electrolyte cost (CElectrolyte) and reactor ASR, where each contour achieves a battery
price of $100 kWh-1. For all sub-figures, NAqRFB and AqRFB contours are represented
Fig. 1a shows that a NAqRFB with a cell voltage less than 2 V will be as solid and dashed lines, respectively.
financially infeasible.
In contrast to nonaqueous systems, AqRFB designs demonstrate
negligible sensitivity across order of magnitude changes in redox- or cell voltage can substantially affect the design space. At
active species concentration (0.5e200 mol kg1). For AqRFBs, the extremely low concentrations (less than 0.5 mol kg1), however,
supporting electrolyte cost contribution ($0.1 kg1) is extremely AqRFBs require cell voltages greater than the electrochemical sta-
low, and thereby, only the redox-active material molecular weight bility window of water to meet the cost targets. This voltage
 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272 265

requirement suggests a minimum concentration requirement of aqueous and nonaqueous RFBs. The difference in design sensitivity
0.5 mol kg1 to maintain electrolyte stability and eliminate redox- between the two systems leads to fundamentally different chal-
active material molality as a cost constraint. The maximum stable lenges in materials selection at fixed battery price. Recent reports
cell voltage in aqueous solution and the minimum cell voltage have already demonstrated NAqRFBs with cell voltages greater than
required to recover electrolyte costs bound the AqRFB feasible 2 V [52,53], and low molecular weights less than 200 g mol1
design space (highlighted in yellow in Fig. 1a). In the regime of [28,33,54]. These early advances suggest that the cell voltage and
ultra-low redox-active species concentration (less than molecular weight benchmarks of 3 V and 100 g mol1 may be
0.5 mol kg1), AqRFBs may become sensitive to variations in other possible in the future. The corresponding redox-active material
cost parameters such as pumping losses, cycle efficiencies, or tank concentration target of 9.6 mol kg1 (~4e5 mol L1, assuming
sizes, which the design maps in Fig. 1a do not capture. specific electrolyte volume of 1 L kg1 and solvent mass fraction of
RFB design is also sensitive to reactor ASR. A recent study has ~0.4e0.5) for NAqRFBs, however, would be difficult to achieve
shown that reactor ASR can drastically impact the required cell experimentally. State-of-the-art tailored organic redox-active ma-
voltage for economically feasible RFBs [41], but changes in ASR can terials developed by Sevov et al. and Huang et al. had solubility
also affect the required redox-active material concentrations for limits less than 2 mol L1 [33,54]. Decreasing the required redox-
NAqRFBs. Fig. 1b plots contours of constant concentration as a active material concentration becomes a critical design optimiza-
function of cell voltage and reactor ASR for both aqueous and tion pathway for economically viable NAqRFBs. Since AqRFB design
nonaqueous RFBs. First, this analysis establishes that an upper is relatively insensitive to solvent costs, AqRFBs can operate in a
bound on a maximum plausible ASR is approximately 20 U-cm2; at cost effective manner even at low redox-active material concen-
this ASR value all NAqRFB designs would require cell voltages above trations, but viable AqRFBs will require cell voltages in the range of
4.5 V, which would be difficult due to electrolyte breakdown [51]. 1.2e1.5 V and ASR values below 1.5 U-cm2. While many AqRFBs,
Similarly, a maximum plausible ASR for AqRFBs is approximately including vanadium systems, demonstrate cell voltages exceeding
1.5 U-cm2, beyond which an AqRFB would require a cell voltage 1 V, low-cost redox-active species that maximize use of aqueous
exceeding 1.5 V, leading to imminent water dissociation. Consid- electrochemical stability windows are essential to achieve a low
ering the nonaqueous contours, Fig. 1b demonstrates a rapid battery price.
decrease in required cell voltage or redox-active species concen-
tration as ASR decreases in the range of 20 to 1 U-cm2. The DG 3.2. Nonaqueous RFB design optimization
model recommended employing 3 V NAqRFB reactors with 5 U-cm2
ASR and redox-active species concentration of 9.6 mol kg1 [9], but To enable NAqRFBs with sufficiently high cell voltages, positive
a later study recommended decreasing the ASR of NAqRFBs down electrolyte materials must have relatively high redox potentials,
to 2.3 U-cm2 [41], which could reduce the required redox-active while negative electrolyte materials must have relatively low redox
species concentration to 4 mol kg1. For ASR values below 1 U- potentials. The difference between the redox potentials of the
cm2, NAqRFB cell voltage targets become relatively insensitive to positive and negative redox-active materials will define the total
further decreases in ASR. Again, due to low solvent costs for con- NAqRFB cell voltage, and thus, the positive and negative electro-
centrations greater than 0.5 mol kg1 AqRFB cell voltage and ASR lytes each require unique materials selection criteria. Fig. 2 quan-
requirements are less sensitive to redox-active species concentra- tifies required changes in individual electrolyte material redox
tion, even over order of magnitude changes. AqRFBs with concen- potential as a function of molecular weight for various redox-active
trations less than 0.5 mol kg1 are infeasible at $100 kWh-1 due to material concentrations.
high cell voltage requirements that extend beyond the stability To allow such an analysis, Fig. 2 pairs positive electrolyte ma-
window of aqueous electrolytes. Fig. 1b, however, also demon- terials with a benchmark negative electrolyte material that has the
strates that decreasing ASR for aqueous systems below 1 U-cm2 same molecular weight per electron transferred as the positive
could decrease cell voltage requirements down to under 1.2 V, active material of interest, but with a redox potential of 1 V vs. Li/
broadening the number of viable redox-active materials for use in Liþ. Conversely, the analysis pairs negative electrolyte materials
AqRFBs. Due to the inherent constraint of the narrow AqRFB elec-
trochemical window, small improvements in ASR could lead to a
critical decrease in required AqRFB cell voltage.
We directly illustrate the relationship among reactor ASR and
electrolyte cost (CElectrolyte) with a series of concave-down cell
voltage contours in Fig. 1c. Generally, as ASR decreases, electrolytes
that are more expensive enter the available design space, but as ASR
increases, electrolyte cost must decrease to compensate for the
associated rise in reactor cost. In addition, higher cell voltage per-
mits both higher electrolyte cost and ASR, as the cell voltage is a
critical parameter in determining both the reactor and electrolyte
costs. For example, the higher cell voltages possible in NAqRFBs can
offset the higher costs and ASR values associated with nonaqueous
electrolytes. Given the electrochemical stability windows of typical
aqueous and nonaqueous electrolytes, Fig. 1c reaffirms our sug-
gested bounds on reactor ASR for both RFB families. While Fig. 1c
demonstrates interesting tradeoffs in reactor and electrolyte cost
contributions, the total electrolyte cost is difficult to decompose
into price optimization pathways for individual electrolyte
components. Fig. 2. Contours of constant of redox-active species concentration, as a function of
redox-active species molecular weight and redox potential, for NAqRFBs, that achieve a
Considering only cell voltage, ASR, redox-active material mo- battery price of $100 kWh-1. Solid contours correspond to positive redox-active ma-
lecular weight, and redox-active material concentration as design terials, and dashed contours correspond to negative redox-active materials. The right
parameters, the variability of battery price is evident for both hand y-axis displays the benchmark counter-electrode potentials.
266 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272

with a benchmark positive electrolyte material, again, with the

same molecular weight per electron transferred as the negative
active material of interest, but exhibiting a redox potential of 4 V vs.
Li/Liþ. In Fig. 2 contours with Eo > 3 V vs. Li/Liþ (solid) correspond to
positive electrolyte materials, while contours with Eo < 2 V vs. Li/Liþ
(dashed) correspond to negative electrolyte materials. The positive
and negative electrolyte contours exhibit complementary trends to
the cell voltage contours in Fig. 1a; specifically, increasing redox-
active material molecular weight requires a more extreme redox
potential to achieve $100 kWh-1. In addition, as redox-active ma-
terial concentration decreases for a fixed molecular weight, the
target system design requires more extreme redox potentials. The
region between the positive and negative electrolyte contours
represents an infeasible region for redox-active material use in a
NAqRFB. Redox potentials in this region are too moderate to enable
high enough cell voltages to offset the associated reactor and
electrolyte costs. Fig. 2 ultimately demonstrates that by identifying
redox-active materials with more extreme redox potentials, or by
decreasing redox-active material molecular weight, nonaqueous
electrolytes with lower concentrations of redox-active materials
become cost effective.
Until now, the TE analysis has only considered the redox-active
material and solvent cost contributions to NAqRFB electrolytes, but
salt costs ($20 kg1) will be higher than either redox-active ma-
terials ($5 kg1) or solvent costs ($2 kg1) due to the high cost
associated with fluorinated anions. Fig. 3a demonstrates the rela-
tionship among salt cost, salt molecular weight, and redox-active
species concentration. For each contour of constant concentra-
tion, as salt cost increases, the salt molecular weight must decrease,
and vice versa. This simple trend arises to maintain the same
overall cost of salt (in $) for a fixed redox-active material concen-
tration. The DG model estimated a salt cost of $20 kg1, but the
variation in cost among lithium salts [55], suggests that cheaper
materials, below $20 kg1, could be used for NAqRFBs. Salt candi-
date searches should consider new lithium-ion battery salts, such
as chelated phosphates, borates, imides, and heterocyclic amines Fig. 3. (a) Contours of constant redox-active species concentration as a function of salt
cost and salt molecular weight for NAqRFBs. Each contour achieves the target $100
[56], as a possible pathway to decrease materials cost. Identifying kWh-1 battery price. (b) Contours of constant redox-active species concentration for
cheaper or lower molecular weight salts can minimize redox-active NAqRFBs, as a function of average molar salt ratio and redox-active species molecular
species concentration requirements for NAqRFBs. weight, which satisfy the $100 kWh-1 battery price target. The bottom half of the
Aside from identifying overall cheaper salts, carefully selecting design space (ravg < 0.5) is only accessible by rocking-chair (RC) configuration RFBs,
while the upper half (ravg  0.5) is available to either rocking-chair or salt-splitting (SS)
NAqRFB redox reactions can minimize the total salt cost contribu-
cells. The shaded upper half region represents the viable design space for salt-splitting
tion to the electrolyte cost. Consider that the salt plays three roles in cells. Solid contours correspond to a salt cost of $20 kg1, and dashed contours
a NAqRFB electrolyte. First, dissolved salt imparts conductivity on correspond to a salt cost of $5 kg1. Dotted black lines denote benchmark values.
an otherwise insulating nonaqueous solvent, allowing for ionic
conduction through the pore phase of a porous electrode. Second,
the salt ions act as ionic charge carriers through the membrane of The first reaction (Eq. (4)) represents a salt-splitting configura-
the RFB, which is a key attribute of any electrochemical cell. Third, tion, where both redox-active materials begin as neutral species at
the salt will serve to maintain electroneutrality in the bulk elec- 0% state of charge (SOC), but then A oxidizes to a cation and B re-
trolyte throughout the entire RFB while redox-active materials duces to an anion at 100% SOC. The use of dissimilar charged species
undergo reduction or oxidation. Importantly, the requirement of at 100% SOC will drive salt cations to the negative electrolyte, while
bulk electroneutrality can lead to unnecessarily high concentra- salt anions migrate to the positive electrolyte. Thus, the salt-
tions of salt and subsequently unnecessarily high salt costs if the splitting configuration requires a minimum of one salt molecule
charges on the redox-active materials are dissimilar [57]. for every two redox-active molecules (i.e., including redox-active
Equations (4)e(6) show three distinct reaction schemes for RFBs molecules in both the positive and negative electrolytes) to main-
(assuming one electron stored for each redox-active species), tain electroneutrality across all SOCs. This condition restricts the
where A is the positive electrolyte redox-active material and B is molar salt ratio ravg to values in excess of 50% for salt-splitting
the negative electrolyte redox-active material: configurations. Many NAqRFBs presented in literature exhibit a
salt-splitting configuration due to the wider availability of stable
A þ B4Aþ þ B (4) redox-active materials in neutral state [24e26,34,53]. In contrast,
Eqs. (5) and (6) (where n is an integer greater than or equal to zero)
represent a special case of rocking-chair configuration RFBs, also
Aðnþ1Þ þ Bn 4An þ Bðnþ1Þ (5) sometimes referred to as common-ion exchange RFBs [57]. In these
systems, either a single cation (Eq. (5)) or a single anion (Eq. (6))
Anþ þ Bðnþ1Þþ 4Aðnþ1Þþ þ Bnþ (6) transfers across the membrane to maintain electroneutrality [57],
resembling ion transfer in a traditional lithium-ion battery [58]. By
 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272 267

utilizing a single ion to facilitate charge transfer across the mem- electrolytes. Figure 4 demonstrates the difficulty in identifying
brane and redox-active materials that maintain the same sign of viable redox-active species for AqRFBs, by plotting contours of
charge (cation or anion) across all SOCs, rocking-chair RFBs do not constant redox-active material concentration in the space of redox
require any salt to charge balance (i.e., ravg  0); the salt in a potential (vs. the Reversible Hydrogen Electrode (RHE)) and mo-
rocking-chair cell merely imparts ionic conductivity to the elec- lecular weight. Here, the respective benchmark counter-electrodes
trolyte and membrane. Rocking-chair RFBs require that at least one at 0.1 V and 1.4 V vs. RHE are paired with positive and negative
of the redox-active materials be ionic at 0% SOC, and this ionic electrolyte materials of interest for AqRFBs in a manner similar to
redox-active material must pair with an associating counter ion NAqRFBs in Fig. 2. As a quantitative example, a positive electrolyte
[57]. While uncommon in recent literature, some reports demon- material with a molecular weight of 100 g mol1 must have a redox
strate rocking-chair NAqRFBs [16,17,59e62]. potential in the range 1.1 V < Eo < 1.4 V vs. RHE; this is a narrow
Switching NAqRFB configuration towards rocking-chair systems 300 mV design space to investigate, eliminating many possible
will allow for decreasing salt concentrations, which can dramati- redox-active material candidates. If the redox-active species mo-
cally widen the NAqRFB design space by eliminating costly salt lecular weight increases beyond 100 g mol1, the available design
from the system. Fig. 3b relates the required redox-active species space shrinks even further. Selecting redox-active materials that
concentration to molar salt ratio and redox-active species molec- enable AqRFBs with cell voltages >1.1 V is of paramount
ular weight for various NAqRFB designs, assuming two different salt importance.
costs of $5 kg1 (dashed lines) and $20 kg1 (solid lines). Each iso- Beyond individual redox-active species selection, tradeoffs be-
concentration contour exhibits a near linear decrease of molar salt tween reactor and electrolyte cost contributions can broaden the
ratio with increasing redox-active species molecular weight; to available design space. Figure 5a shows the relationship among
offset higher total redox-active material costs, the total salt cost, required cell voltage, ASR, and redox-active material molecular
and thereby salt concentration, must decrease. When constructing weight to achieve a battery price of $100 kWh-1. As previously
a RFB in rocking-chair configuration, redox-active species in at least shown in Fig. 1b, the ASR must be under 1.5 U-cm2 if the redox-
one of the electrolytes must be an ion. Such a material would be active material exhibits a reasonable molecular weight
purchased with an associated counter ion that increases the mo- (~100 g mol1), and any molecular weight greater than 150 g mol1
lecular weight of the redox-active material (relative to its molecular would require unrealistically high cell voltages for aqueous sys-
weight as a neutral species). Thus, Fig. 3b can assist in balancing tems. If extremely low molecular weight redox-active materials
molecular weight with the amount of dissolved salt in the NAqRFB. (<50 g mol1) were available, the maximum allowable ASR could
Further, as the redox-active species concentration decreases, either increase to 4 U-cm2, but such low weights would require molecular
the molar salt ratio or redox-active species molecular weight must simplicity similar to that of an ethanol molecule (46 g mol1). Such
decrease to offset higher solvent costs. In Fig. 3b, values of ravg < 0.5 simple species are likely to undergo only chemically irreversible
represent a design space that is only accessible by employing a redox events. Additionally, drastically decreasing cell ASR to an
rocking-chair NAqRFB design. The DG model assumed 1 mol L1 ultra-low value of 0.1 U-cm2 affords only a 20% decrease in cell
salt concentration for NAqRFB electrolytes [9], which, through our voltage target. The practical difficulties in decreasing molecular
analysis, corresponds to ravg ¼ 0.20. As we show here, this salt ratio weight and ASR suggest that these are unviable design pathways
is only compatible with a rocking-chair configuration NAqRFB, towards decreasing AqRFB cell voltage requirements. Thus, we
where fewer moles of salt are present in the electrolyte than moles conclude that the most viable pathways towards achieving the
of redox-active material. For a salt cost of $20 kg1, salt-splitting desired battery price for AqRFBs are either by minimizing redox-
cells are financially infeasible, unless the redox-active materials active material cost ($ kg1) or by maximizing cell voltage.
exhibit unrealistically low molecular weights (<50 g mol1) and As an illustration, Fig. 5b reveals how decreasing redox-active
high concentrations (>8 mol kg1). The salt-splitting design space, material cost balances variations in cell voltage to achieve $100
however, does become accessible for a salt cost of $5 kg1. As such, kWh-1. The benchmark value for redox-active material cost of
NAqRFB design is sensitive to salt cost and salt concentration due to $5 kg1 corresponds to a required cell voltage of 1.2 V, but
the anticipated high costs of NAqRFB salts relative to redox-active employing a redox-active material cost of $1 kg1 reduces the cell
material and solvent costs, and, by carefully minimizing salt cost
and concentration, lower redox-active species concentrations, un-
der 4 mol kg1, become feasible.

3.3. Aqueous RFB design optimization

While NAqRFB price is sensitive to all constituent electrolyte

material costs, cell voltage, and ASR, the cost contributions of salt
and solvent in AqRFBs are small. Consequently, the battery price of
AqRFBs is sensitive neither to redox-active species concentration
nor cell configuration (i.e., either rocking-chair or salt-splitting).
These insensitivities stem from the extremely low-cost support-
ing electrolytes afforded in aqueous systems. Therefore, AqRFB
design optimization focuses on varying cell voltage, ASR, redox-
active material molecular weight, and redox-active material cost
per unit mass to achieve $100 kWh-1.
Aqueous redox-active materials require sufficiently extreme
redox potentials (high potential for positive redox-active materials,
low potential for negative redox-active materials) to construct cells Fig. 4. Contours of constant redox-active species concentration for AqRFBs, as a
function of redox-active species molecular weight and redox potential, which achieve
with voltages that are high enough to achieve the target battery a battery price of $100 kWh-1. Solid contours correspond to positive electrode mate-
price, but the redox potentials of the redox-active species are also rials, and dashed contours correspond to negative electrode materials. The right hand
constrained by the electrochemical stability window of aqueous y-axis displays the benchmark counter-electrode potentials.
268 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272

Fig. 5. (a) Contours of constant cell voltage for AqRFBs, as a function of ASR and redox- Fig. 6. Suggested design pathways that minimize cost and expand the design space for
active species molecular weight, which achieve a battery price of $100 kWh-1. (b) critical design constraints for (a) NAqRFBs and (b) AqRFBs. The horizontal dashed line
Contours of constant cell voltage for AqRFBs, as a function of redox-active species cost represents the $100 kWh-1 battery price target. “Active Material” is abbreviated as AM
and molecular weight, which achieve a $100 kWh-1 battery price. Dotted lines illus- in the legend.
trate the benchmark values for an AqRFB.

for various design improvements over the DG model baseline. As

voltage requirement as low as 0.7 V. We can thus identify redox- observed in Fig. 6a, the DG model suggestions achieve the target
active material cost and cell voltage as the two most critical pa- $100 kWh-1 battery price at challenging redox-active material
rameters in building economically viable AqRFBs. concentrations (greater than 6 mol kg1) [9]. As a first possible
design improvement, decreasing cell ASR down to 2.5 U-cm2, a
4. Recommended RFB design pathways value recommended by a study of RFB area-specific resistance [41],
affords a small decrease in battery price. Experimental studies of
The RFB materials selection maps presented in this work are membrane performance demonstrate significant variability in
powerful tools for quantifying the tradeoffs among various elec- preventing crossover and facilitating ion transfer [53,59,63,64].
trolyte material costs, cell voltage, and ASR, but extracting design Finding a membrane (or separator) for NAqRFBs with high selec-
rules from them can be overwhelming due to the large number of tivity and that performs at high currents is one major materials
variables and wide range of values each variable may assume. To challenge to overcome [41]. Beyond the ohmic contribution to
this end, we propose generalized RFB design guidelines aimed at NAqRFB resistance, the transport capabilities of nonaqueous sol-
assisting in electrolyte materials selection. The DG model began vents in porous electrodes present a challenge. Due to the high
such a design process by suggesting single iterations of aqueous viscosity of some NAqRFB electrolytes [65,66], resistance due to
and nonaqueous RFB designs [9], but the plethora of available it- pore-scale mass transfer of redox-active species and ionic con-
erations outlined in this work hints that even more realistic and duction through the electrode thickness could possibly be similar in
cost effective pathways may exist. magnitude to the separator/membrane resistance.
Employing extremely cheap or lightweight redox-active mate-
4.1. Nonaqueous RFB design pathways rials (~50 g mol1) could afford similar cost savings as the decrease
in ASR described above (Fig. 6a), but, as previously mentioned, such
As NAqRFB price is sensitive to electrolyte components, cell light species are unlikely to be electrochemically reversible com-
voltage, and ASR, many possible NAqRFB design iterations become pounds. Identifying redox-active materials with molecular weights
available, and Fig. 6a outlines new design pathways, showing between 100 and 200 g mol1, which participate in multiple elec-
NAqRFB price as a function of redox-active species concentration tron transfer events, such as those developed by Sevov et al. [54],
 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272 269

Table 2
Alternative NAqRFB and AqRFB design iterations that decrease redox-active material molality targets and cell voltage targets by changing other parameters (bolded). All cells
shown achieve the $100 kWh-1 price target.

Nonaqueous RFB

Cell type Cell voltage (V) Actives Mol. Weight (g mol1) ASR (U-cm2) Salt ratio Actives Molality (mol kg1)

1. Benchmark 3 100 5 0.20 6.3

2. High Cell Voltage 4 100 5 0.20 1.8
3. Low Actives Molecular Weight 3 50 5 0.20 3.1
4. Low ASR 3 100 2.5 0.20 3.3
5. No Salt 3 100 5 0 2.7
6. High Cell Voltage, Low ASR, and No Salt 4 100 2.5 0 1.1

Aqueous RFB

Cell type Actives cost ($ kg1) Actives Mol. Weight (g mol1) ASR (U-cm2) Cell voltage (V)

1. Benchmark 5 100 0.5 1.21

2. Low Actives Cost 2 100 0.5 0.79
3. Low Actives Molecular Weight 5 50 0.5 0.85
4. Low ASR 5 100 0.1 1.04
5. Low Actives Cost and Low Actives Molecular Weight 2 50 0.5 0.67

may present a viable cost-cutting pathway. Recent reports potentials less than 1.5 V vs. Li/Liþ are not available. Identifying
demonstrate metal coordination complexes exhibiting up to six redox-active species with low redox potentials for use in the
redox events [60], but these compounds have very high molecular negative electrolyte of a NAqRFB remains a major materials design
weights and offer only moderate redox potentials. Two electron (or challenge [35,54].
more) transfer materials could decrease the equivalent weight To complement the price minimization trends offered in Fig. 6a,
(molecular weight normalized by number of electrons transferred) Table 2 presents quantitative design iterations for NAqRFBs to ach-
into the range of 50e100 g mol1. ieve a $100 kWh-1 battery price, providing tangible performance
Salt cost contributions have a large effect on NAqRFB system guidelines for materials selection. By pushing NAqRFB design to
design options. In particular, Fig. 6a shows that the battery price of a incorporate any one of the proposed cost cutting pathways (high
NAqRFB with no salt is lower than the benchmark value by voltage, low ASR, low salt concentration, or low redox-active ma-
approximately $10 kWh-1 for moderately high redox-active species terial weight), redox-active material concentration requirements
concentration. Removing salt from NAqRFBs may actually be a shrink by more than 50% of the DG model benchmark (9.6 mol kg1).
practical option in decreasing battery price by employing ionic Driving redox-active material concentration below 1 mol kg1 is
liquid solvents, deep eutectic melts [67,68], or all ionic redox-active unlikely, as demonstrated by the divergence to infinity of every price
materials. First, ionic liquid solvents can cost 5e20 times more than curve in Fig. 6a at low values of redox-active species molality. If a
molecular solvents [69], but the cost decrease afforded by removing NAqRFB can leverage all of the cost cutting pathways presented
salt from the electrolyte could make certain ionic liquid solvents here, the battery price could easily drop below $100 kWh-1 for
viable. Additionally, we recommend investigation into RFBs utiliz- redox-active species concentrations greater than 2 mol kg1.
ing protic ionic liquids (e.g., methanesulfonic acid, triethylammo-
nium hydrogen sulfate), which can exhibit costs [70] and 4.2. Aqueous RFB design pathways
electrochemical windows [71] in between those of water and mo-
lecular nonaqueous solvents. Second, deep eutectic melts While this work presents many pathways towards low-cost
employing a redox-active material (e.g., FeCl3 [67]) and a miscible NAqRFBs, design optimization pathways for AqRFBs are substan-
host (e.g., choline chloride [67,68]) may offer an attractive pathway tially more limited since AqRFBs are only sensitive to variations in
to no-salt NAqRFBs with moderate redox-active species concen- cell voltage, ASR, and redox-active species cost. Fig. 6b displays
trations (approximately 3.6 mol kg1) [67]. Third, redox-active AqRFB prices as a function of cell voltage for a limited number of
materials that maintain ionic nature throughout all SOCs (e.g., possible cell improvement pathways. In addition, Table 2 also
metal centered bypiridines [16e18,62]) could serve as redox-active presents quantitative iterations of AqRFB designs that achieve $100
charge carriers, also eliminating the need for a supporting salt. kWh-1. Much like the NAqRFB, employing redox-active materials
Identifying any such multi-function materials that assume two or with low molecular weights only affords a small decrease in battery
more roles in the electrolyte (i.e., redox-active, charge carrier, sol- price, and, as previously described, synthesizing redox-active spe-
vation) could enable large cost savings for NAqRFBs. cies with molecular weight much below 100 g mol1 is unlikely.
One final recommendation towards decreasing NAqRFB price is Decreasing redox-active species cost, however, can provide the
simply to increase cell voltage to approximately 4 V, which yields most drastic savings, alleviating cell voltage requirements or
the most drastic decrease in NAqRFB price considered. Nonaqueous driving battery prices down below $100 kWh-1; low-cost tailored
electrolytes offering a 4 V electrochemical window could easily organic [22,23,29,30,45,74] or abundant inorganic [20] materials
enable low concentration batteries with a low price. Cell voltage could offer redox-active species costs under $5 kg1. If an AqRFB
affects the total battery price so dramatically because increased exploits both low redox-active material cost and high cell voltage
voltage decreases both the electrolyte (Eq. (3)) and reactor (Eq. (2)) (approaching the stability limit of 1.5 V), AqRFB price could drop
cost contributions simultaneously. Molecular nonaqueous solvents, below $100 kWh-1.
such as propylene carbonate, exhibit large electrochemical win-
dows allowing for 4 V electrochemical cells [51], and some work 5. Conclusions
has demonstrated fluorinated organic solvents designed for use in
5 V lithium-ion batteries [72]. Additionally, soluble redox-active In this work, we present a detailed electrolyte cost model, which
compounds have proven redox potentials as high as 4.6 V vs. Li/ explicitly accounts for redox-active material, salt, and solvent
Liþ [73], but stable, soluble redox-active compounds with contributions to RFB price, as an adaptation to and an extension of
270 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272

prior work by Darling, Gallagher, and co-workers [9]. This techno- Storage Research, an Energy Innovation Hub funded by the U.S.
economic model explores the available design space for both Department of Energy, Office of Science, Basic Energy Sciences.
aqueous and nonaqueous RFBs by considering variations in elec- Jarrod Milshtein also acknowledges financial support from the
trolyte cost parameters, cell voltage, and reactor ASR, as well as National Science Foundation Graduate Research Fellowship Pro-
identifying critical cost constraining variables for RFBs. In a broad gram (DGE 1256260). The authors would like to thank the JCESR
sense, this analysis first defines lower bounds on cell voltage re- Flow Chemistry Sprint team for their continued input and support,
quirements of 1.1 V and 2.0 V for aqueous and nonaqueous RFBs, including Dr. Xiaoliang Wei, Dr. Rama Vemuri, Dr. Wentao Duan, Dr.
respectively, to obtain a $100 kWh-1 battery price. Additionally, Levi Thompson, Dr. Krista Hawthorne, Dr. Tanya Breault, and Syd-
upper bounds on reactor ASR for aqueous and nonaqueous RFBs are ney Laramie. We especially thank Dr. Robert Darling and Dr. Kevin
1.5 U-cm2 and 20 U-cm2, respectively. Gallagher for their valuable input regarding this project. Further-
NAqRFBs are sensitive to every cost parameter considered in this more, we thank members of the Brushett Group at MIT and the
analysis due to comparable cost contributions from the electrolyte Smith Group at UIUC for constructive feedback throughout the
components and reactor, but the largest potential cost savings for project.
NAqRFBs come from either increasing cell voltage above 3 V or
minimizing the amount of supporting salt. We propose identifying Appendix
materials that provide two or more functions in the electrolyte (i.e.,
charge balancing, electroactivity, solubilization, and ionic conduc- The present work uses a detailed electrolyte model that builds
tivity), which remove the need for a true salt or solvent and could on the DG model by quantifying the effect of salt concentration and
enable much more cost effective nonaqueous electrolytes. In salt molecular weight on RFB price. The DG model accounts for salt
comparison, NAqRFB cost cutting by decreasing cell ASR below 5 U- and solvent cost using an electrolyte cost per unit mass (cm,e ± in
cm2 or reducing redox-active material molecular weight below units of $ kg1). By lumping salt and solvent costs together in this
100 g mol1 only affords small gains. These same techniques can manner, the DG model did not capture the sensitivity of battery
also reduce the required NAqRFB redox-active species concentra- price to salt concentration and molecular weight. Thus, the present
tion to reasonably low values of 2e4 mol kg1, which are near detailed electrolyte model expands the electrolyte cost per unit
experimental realization. Unlike nonaqueous systems, AqRFBs only mass in terms of the mass ratio of salt to total mass of salt and
exhibited large cost sensitivities to cell voltage and redox-active solvent Ssalt, as well as the costs per unit mass of the salt and solvent
material cost. Identifying low-cost (less than $5 kg1) redox- (csalt and csolvent, respectively):
active materials that enable a cell voltage in the range of
1.1e1.5 V is the most promising pathway towards economically cm;e ¼ Ssalt csalt þ ð1  Ssalt Þcsolvent (A1)
viable AqRFBs.
To capture salt and solvent costs explicitly, the electrolyte cost
Another application of the present cost-driven materials selec-
per unit mass for each electrolyte (cm,eþ and cm,e-) was substituted
tion approach is to curate databases of redox-active molecules
into the battery price expression from Ref. [9]. The resulting
compiled either experimentally [75e78] or computationally [79].
expression for battery price expressed in terms of the average
By comparing the performance of a real, new RFB electrolyte ma-
molar salt ratio ravg and the average redox-active species concen-
terial with these design maps, experimentalists can quickly deter-
tration bavg. In terms of parameters from the DG model, ravg and bavg
mine if the new electrolyte chemistry could achieve the $100 kWh-
1 are expressed as:
price target. Further, computational data sieving is already un-
derway, in conjunction with the Materials Project, which stores  
1 sþ Mþ Ssaltþ s M Ssalt
electrochemical data for thousands of ab initio predicted redox- ravg ¼ þ ; (A2)
2 cþ neþ Msaltþ Sþ c ne Msalt S
active molecules as candidates for RFB electrolytes [79]. Accord-
ingly, the present model has been implemented as an interactive  
online tool within the Materials Project, called the RFB Dashboard 1 1 sþ Mþ ð1  Ssaltþ Þ s M ð1  Ssalt Þ
¼ þ ; (A3)
[80]. The RFB Dashboard filters redox-active molecules with suit- bavg 2 cþ neþ Sþ c ne S
able redox potentials and molecular weights to build a $100 kWh-1
aqueous or nonaqueous RFB by using the design maps presented where the redox-active species concentration S is expressed in
here (see Figs. 2 and 3). This online tool also allows users to adjust units of kilograms per kilogram of solvent.
model input parameters to assess impact on materials selection, as
well as to accommodate future changes to material cost factors and List of symbols
RFB performance parameters.
Beyond the immediate application to RFB materials selection, b± individual electrolyte active species molality, mol kg1
this analysis presents a framework for cost-conscious research ef- bavg mean active species molality, mol kg1
forts. The design maps translate system-level price and perfor- ca reactor cost per unit area, $ m2
mance metrics to quantitative guidelines for materials properties cadd addition to price, $ kW1
and performance. Bridging the gap between abstract cost models cbop balance-of-plant cost, $ kW1
and focused experimental research will enable rapid transition of cm,± active material cost per unit mass, $ kg1
new materials into economically viable prototypes. Design maps cme,± electrolyte cost per unit mass, $ kg1
also highlight promising regions of design space that may be csalt salt cost per unit mass, $ kg1
underexplored in the contemporary literature. We hope that this csolvent solvent cost per unit mass, $ kg1
methodology will apply to other systems where cost is a major CAdditional additional cost, $ kWh1
inhibitor to success by creating tangible experimental targets from CBOP balance-of-plant cost, $ kWh1
detailed techno-economic modeling. CElectrolyte electrolyte cost, $ kWh1
CReactor reactor cost, $ kWh1
Acknowledgements Ed battery discharge energy, kWh
F Faraday constant, kAh mol1
This work was supported as part of the Joint Center for Energy M± active species molecular weight, g mol1
 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272 271

Msalt salt molecular weight, g mol1 [14] J. Goeltz, D. Amadeo, A.J. Esswein, T.D. Jarvi, E.R. King, S.Y. Reece, et al.,
Aqueous Redox Flow Batteries Comprising Metal Ligand Coordaination
ne ± number of electrons stored per active molecule,
Compounds, 2014. US 2014/0028260 A1.
mole- mol1 [15] Z. Yang, J. Zhang, M.C.W. Kintner-Meyer, X. Lu, D. Choi, J.P. Lemmon, et al.,
P0 battery future-state price, $ Electrochemical energy storage for green grid, Chem. Rev. 111 (2011)
r± individual electrolyte molar salt ratio, mol mol1 3577e3613.
[16] Y. Matsuda, K. Tanaka, M. Okada, U. Takasu, M. Morita, A rechargeable redox
ravg mean molar salt ratio, mol mol1 battery utilizing ruthenium complexes with non-aqueous organic electrolyte,
R area-specific resistance, U-cm2 J. Appl. Electrochem. 18 (1988) 909e914.
s± stoichiometric coefficient, mol mol1 [17] M.H. Chakrabarti, R.A.W. Dryfe, E.P.L. Roberts, Evaluation of electrolytes for
redox flow battery applications, Electrochim. Acta 52 (2007) 2189e2195.
S± active species solubility, kg kg1 [18] M.H. Chakrabarti, E.P. Lindfield Roberts, M. Saleem, Chargeedischarge per-
Ssalt,± salt solubility, kg kg1 formance of a novel undivided redox flow battery for renewable energy
td battery discharge time, h storage, Int. J. Green Energy 7 (2010) 445e460.
[19] A.A. Shinkle, T.J. Pomaville, A.E.S. Sleightholme, L.T. Thompson, C.W. Monroe,
U open-circuit cell voltage, V Solvents and supporting electrolytes for vanadium acetylacetonate flow bat-
εq,rt round-trip coulombic efficiency teries, J. Power Sources 248 (2014) 1299e1305.
εsys,d system discharge efficiency [20] X. Wei, G.G. Xia, B. Kirby, E. Thomsen, B. Li, Z. Nie, et al., An aqueous redox
flow battery based on neutral alkali metal ferri/ferrocyanide and polysulfide
εv,d voltage discharge efficiency electrolytes, J. Electrochem. Soc. 163 (2016) A5150eA5153.
c± depth-of-discharge [21] H. Pan, X. Wei, W.A. Henderson, Y. Shao, J. Chen, P. Bhattacharya, et al., On the
way toward understanding solution chemistry of lithium polysulfides for high
energy Li-S redox flow batteries, Adv. Energy Mater. 5 (2015) 150113.
List of abbreviations [22] K. Lin, Q. Chen, M.R. Gerhardt, L. Tong, S.B. Kim, L. Eisenach, et al., Alkaline
þ positive electrolyte quinone flow battery, Science 80 (2015) 1529e1532, 349.
e negative electrolyte [23] B. Huskinson, M.P. Marshak, C. Suh, S. Er, M.R. Gerhardt, C.J. Galvin, et al.,
A metal-free organic-inorganic aqueous flow battery, Nature 505 (2014)
AM active material 195e198.
Aq aqueous [24] F.R. Brushett, J.T. Vaughey, A.N. Jansen, An all-organic non-aqueous lithium-
ARPA-E Advanced Research Projects Agency e Energy ion redox flow battery, Adv. Energy Mater. 2 (2012) 1390e1396.
[25] S.H. Oh, C.W. Lee, D.H. Chun, J.D. Jeon, J. Shim, K.H. Shin, et al., A metal-free
ASR area-specific resistance and all-organic redox flow battery with polythiophene as the electroactive
BOP balance-of-plant species, J. Mater. Chem. A 2 (2014) 19994e19998.
Conc concentration [26] S.K. Park, J. Shim, J.H. Yang, K.H. Shin, C.S. Jin, B.S. Lee, et al., Electrochemical
properties of a non-aqueous redox battery with all-organic redox couples,
DG Darling and Gallagher et al.
Electrochem. Commun. 59 (2015) 68e71.
DOE Department of Energy [27] X. Wei, W. Xu, J. Huang, L. Zhang, E. Walter, C. Lawrence, et al., Radical
e electron compatibility with nonaqueous electrolytes and its impact on an all-organic
NAq nonaqueous redox flow battery, Angew. Chem. Int. Ed. 54 (2015) 8684e8687.
[28] X. Wei, W. Xu, M. Vijayakumar, L. Cosimbescu, T. Liu, V. Sprenkle, et al.,
RFB redox flow battery TEMPO-based catholyte for high-energy density nonaqueous redox flow
SOC state of charge batteries, Adv. Mater. 26 (2014) 7649e7653.
TE techno-economic [29] J.D. Milshtein, L. Su, C. Liou, A.F. Badel, F.R. Brushett, Voltammetry study of
quinoxaline in aqueous electrolytes, Electrochim. Acta 180 (2015) 695e704.
[30] Y. Xu, Y.H. Wen, J. Cheng, G.P. Cao, Y.S. Yang, A study of tiron in aqueous
Appendix A. Supplementary data solutions for redox flow battery application, Electrochim. Acta 55 (2010)
715e720.
[31] A.P. Kaur, N.E. Holubowitch, S. Ergun, C.F. Elliott, S.A. Odom, A highly soluble
Supplementary data related to this article can be found at http:// organic catholyte for non-aqueous redox flow batteries, Energy Technol. 3
dx.doi.org/10.1016/j.jpowsour.2016.08.129. (2015) 476e480.
[32] J. Huang, L. Cheng, R.S. Assary, P. Wang, Z. Xue, A.K. Burrell, et al., Liquid
catholyte molecules for nonaqueous redox flow batteries, Adv. Energy Mater.
References 5 (2015) 1401782.
[33] J. Huang, L. Su, J.A. Kowalski, J.L. Barton, M. Ferrandon, A.K. Burrell, et al.,
[1] P. Denholm, E. Ela, B. Kirby, M. Milligan, The Role of Energy Storage with A subtractive approach to molecular engineering of dimethoxybenzene-based
Renewable Electricity Generation: NREL/TP-6A2e47187, 2010. redox materials for non-aqueous flow batteries, J. Mater. Chem. A 3 (2015)
[2] I. Gyuk, M. Johnson, J. Vetrano, K. Lynn, W. Parks, R. Handa, et al., Grid Energy 14971e14976.
Storage, US Department of Energy, Washington DC, 2013. [34] T. Liu, X. Wei, Z. Nie, V. Sprenkle, W. Wang, A total organic aqueous redox flow
[3] U.S. Department of Energy, Office of Electricity Delivery and Energy Reliability, battery employing a low cost and sustainable methyl viologen anolyte and 4-
Energy Storage: Program Planning Document, 2011. HO-TEMPO catholyte, Adv. Energy Mater. 6 (2016) 1501449.
[4] C. Ponce de Leo
n, A. Frias-Ferrer, J. Gonzalez-Garcia, D.A. Szanto, F.C. Walsh, [35] J.A. Kowalski, L. Su, J.D. Milshtein, F.R. Brushett, Recent advances in molecular
Redox flow cells for energy conversion, J. Power Sources 160 (2006) 716e732. engineering of redox active organic molecules for nonaqueous flow batteries,

n, L. Berlouis, C.T.J. Low, F.C. Walsh, Progress in
[5] P.K. Leung, X. Li, C. Ponce de Leo Curr. Opin. Chem. Eng 13 (2016) 45e52.
redox flow batteries, remaining challenges and their applications in energy [36] K.G. Gallagher, S. Goebel, T. Greszler, M. Mathias, W. Oelerich, D. Eroglu, et al.,
storage, RSC Adv. 2 (2012) 10125e10156. Quantifying the promise of lithiumeair batteries for electric vehicles, Energy
[6] M. Skyllas-Kazacos, M.H. Chakrabarti, S.A. Hajimolana, F.S. Mjalli, M. Saleem, Environ. Sci. 7 (2014) 1555.
Progress in flow battery research and development, J. Electrochem. Soc. 158 [37] P.A. Nelson, K.G. Gallagher, I. Bloom, D.W. Dees, Modeling the Performance
(2011) R55. and Cost of Lithium-ion Batteries for Electric-Drive Vehicles, 2012.
[7] A.Z. Weber, M.M. Mench, J.P. Meyers, P.N. Ross, J.T. Gostick, Q. Liu, Redox flow [38] C. Zhou, K. Qian, M. Allan, W. Zhou, Modeling of the cost of EV battery wear
batteries: a review, J. Appl. Electrochem 41 (2011) 1137e1164. due to V2G application in power systems, IEEE Trans. Energy Convers. 26
[8] W. Wang, Q. Luo, B. Li, X. Wei, L. Li, Z. Yang, Recent progress in redox flow (2011) 1041e1050.
battery research and development, Adv. Funct. Mater. 23 (2013) 970e986. [39] G. Kear, A.A. Shah, F.C. Walsh, Development of the all-vanadium redox flow
[9] R.M. Darling, K.G. Gallagher, J.A. Kowalski, S. Ha, F.R. Brushett, Pathways to battery for energy storage: a review of technological, financial and policy
low-cost electrochemical energy storage: a comparison of aqueous and aspects, Int. J. Energy Res. 36 (2012) 1105e1120.
nonaqueous flow batteries, Energy Environ. Sci. 7 (2014) 3459e3477. [40] M. Zhang, M. Moore, J.S. Watson, T.A. Zawodzinski, R.M. Counce, Capital cost
[10] V. Viswanathan, A. Crawford, D. Stephenson, S. Kim, W. Wang, B. Li, et al., Cost sensitivity analysis of an all-vanadium redox-flow battery, J. Electrochem. Soc.
and performance model for redox flow batteries, J. Power Sources 247 (2014) 159 (2012) A1183eA1188.
1040e1051. [41] R. Darling, K. Gallagher, W. Xie, L. Su, F. Brushett, Transport property re-
[11] S. Ha, K.G. Gallagher, Estimating the system price of redox flow batteries for quirements for flow battery separators, J. Electrochem. Soc. 163 (2016)
grid storage, J. Power Sources 296 (2015) 122e132. A5029eA5040.
[12] U. S. Department of Energy, Headquarters Advanced Research Projects Agency [42] D. Rastler, Electric Power Research Institute, Market Driven Distributed En-
e Energy ( ARPA-E ), Grid-Scale Rampable Intermittent Dispatchable Storage ergy Storage System Requirements for Load Management Applications, 2007.
Funding Opportunity Announcement, 2010. [43] ICIS Indicative Chemical Prices A-Z, Indicative Chemical Prices A-Z, 2006.
[13] P. Alotto, M. Guarnieri, F. Moro, Redox flow batteries for the storage of http://www.icis.com/chemicals/channel-info-chemicals-a-z/.
renewable energy: a review, Renew. Sustain. Energy Rev. 29 (2014) 325e335. [44] L. Gaines, R. Cuenca, Costs of Lithium-Ion Batteries for Vehicles, 2000.
272 R. Dmello et al. / Journal of Power Sources 330 (2016) 261e272

[45] B. Yang, L. Hoober-Burkhardt, F. Wang, G.K. Surya Prakash, S.R. Narayanan, An Electrochem. Solid-State Lett. 15 (2012) A80eA82.
inexpensive aqueous flow battery for large-scale electrical energy storage [63] L. Su, R.M. Darling, K.G. Gallagher, W. Xie, J.L. Thelen, A.F. Badel, et al., An
based on water-soluble organic redox couples, J. Electrochem. Soc. 161 (2014) investigation of the ionic conductivity and species crossover of lithiated
A1371eA1380. nafion N117 in nonaqueous electrolytes, J. Electrochem. Soc. 163 (2016)
[46] R.A. Potash, J.R. McKone, S. Conte, H.D. Abrun ~ a, On the benefits of a symmetric A5253eA5262.
redox flow battery, J. Electrochem. Soc. 163 (2016) A338eA344. [64] S.H. Shin, Y. Kim, S.H. Yun, S. Maurya, S.H. Moon, Influence of membrane
[47] A.J. Cofrancesco, Anthraquinone, in: Kirk-Othmer Encycl. Chem. Technol. structure on the operating current densities of non-aqueous redox flow bat-
Anthraquinone, John Wiley & Sons, Inc., New York, 2001, pp. 1e13. teries: organiceinorganic composite membranes based on a semi-
[48] F.Y. Fan, W.H. Woodford, Z. Li, N. Baram, K.C. Smith, A. Helal, et al., Polysulfide interpenetrating polymer network, J. Power Sources 296 (2015) 245e254.
flow batteries enabled by percolating nanoscale conductor networks, Nano [65] L. Su, M. Ferrandon, J.A. Kowalski, J.T. Vaughey, F.R. Brushett, Electrolyte
Lett. 14 (2014) 2210e2218. development for non-aqueous redox flow batteries using a high-throughput
[49] S.H. Shin, S.H. Yun, S.H. Moon, A review of current developments in non- screening platform, J. Electrochem. Soc. 161 (2014) A1905eA1914.
aqueous redox flow batteries: characterization of their membranes for [66] J.D. Milshtein, J.L. Barton, R.M. Darling, F.R. Brushett, 4-acetamido-2,2,6,6-
design perspective, RSC Adv. 3 (2013) 9095. tetramethylpiperidine-1-oxyl as a model organic redox active compound for
[50] D. Pletcher, R. Wills, A novel flow battery: a lead acid battery based on an nonaqueous redox flow batteries, J. Power Sources 327 (2016) 151e159.
electrolyte with soluble lead(II) Part II. Flow cell studies, Phys. Chem. Chem. [67] M.A. Miller, J.S. Wainright, R.F. Savinell, CommunicationdIron ionic liquid
Phys. 6 (2004) 1779e1785. electrolytes for redox flow battery applications, J. Electrochem. Soc. 163
[51] D. Aurbach, Y. Talyosef, B. Markovsky, E. Markevich, E. Zinigrad, L. Asraf, et al., (2016) A578eA579.
Design of electrolyte solutions for Li and Li-ion batteries: a review, Electro- [68] L. Bahadori, M.A. Hashim, N.S.A. Manan, F.S. Mjalli, I.M. AlNashef,
chim. Acta 50 (2004) 247e254. N.P. Brandon, et al., Investigation of ammonium- and phosphonium-based
[52] J. Mun, M.-J. Lee, J.-W. Park, D.-J. Oh, D.-Y. Lee, S.-G. Doo, Non-aqueous redox deep eutectic solvents as electrolytes for a non-aqueous all-vanadium redox
flow batteries with nickel and iron tris(2,20 -bipyridine) complex electrolyte, cell, J. Electrochem. Soc. 163 (2016) A632eA638.
Electrochem. Solid-State Lett. 15 (2012) A80eA82. [69] N.V. Plechkova, K.R. Seddon, Applications of ionic liquids in the chemical in-
[53] I.L. Escalante-García, J.S. Wainright, L.T. Thompson, R.F. Savinell, Performance dustry, Chem. Soc. Rev. 37 (2008) 123e150.
of a non-aqueous vanadium acetylacetonate prototype redox flow battery: [70] L. Chen, M. Sharifzadeh, N. Mac Dowell, T. Welton, N. Shah, J.P. Hallett,
examination of separators and capacity decay, J. Electrochem. Soc. 162 (2015) Inexpensive ionic liquids: [HSO4]-based solvent production at bulk scale,
A363eA372. Green Chem. 16 (2014) 3098.
[54] C.S. Sevov, R.E.M. Brooner, E. Che
nard, R.S. Assary, J.S. Moore, J. Rodríguez- [71] J. Luo, J. Hu, W. Saak, R. Beckhaus, G. Wittstock, I.F.J. Vankelecom, et al., Protic
Lo
pez, et al., Evolutionary design of low molecular weight organic anolyte ionic liquid and ionic melts prepared from methanesulfonic acid and 1H-
materials for applications in nonaqueous redox flow batteries, J. Am. Chem. 1,2,4-triazole as high temperature PEMFC electrolytes, J. Mater. Chem. 21
Soc. 137 (2015) 14465e14472. (2011) 10426e10436.
[55] R. Younesi, G.M. Veith, P. Johansson, K. Edstrom, T. Vegge, Lithium salts for [72] L. Hu, Z. Xue, K. Amine, Z. Zhang, Fluorinated electrolytes for 5-V Li-Ion
advanced lithium batteries: Li-metal, Li-O2, and Li-S, Energy Environ. Sci. 8 chemistry: synthesis and evaluation of an additive for high-voltage LiNi0.
(2015) 1905e1922. 5Mn1. 5O4/Graphite cell, J. Electrochem. Soc. 161 (2014) A1777eA1781.
[56] K. Xu, Electrolytes and interphases in Li-Ion batteries and beyond, Chem. Rev. [73] Z. Chen, J. Liu, A.N. Jansen, G. GirishKumar, B. Casteel, K. Amine, Lithium borate
114 (2014) 11503e11618. cluster salts as redox shuttles for overcharge protection of lithium-ion cells,
[57] S.M. Laramie, J.D. Milshtein, T.M. Breault, F.R. Brushett, L.T. Thompson, Per- Electrochem. Solid-State Lett. 13 (2010) A39.
formance and cost characteristics of multi-electron transfer, common ion [74] T. Janoschka, N. Martin, U. Martin, C. Friebe, S. Morgenstern, H. Hiller, et al., An
exchange non-aqueous redox flow batteries, J. Power Sources 327 (2016) aqueous, polymer-based redox-flow battery using non-corrosive, safe, and
681e692. low-cost materials, Nature 527 (2015) 78e81.
[58] B. Scrosati, Lithium rocking chair batteries: an old concept? J. Electrochem. [75] N.G. Connelly, W.E. Geiger, Chemical redox agents for organometallic chem-
Soc. 139 (1992) 2776. istry, Chem. Rev. 96 (1996) 877e910.
[59] N.S. Hudak, L.J. Small, H.D. Pratt, T.M. Anderson, Through-plane conductivities [76] Y. Liang, Z. Tao, J. Chen, Organic electrode materials for rechargeable lithium
of membranes for nonaqueous redox flow batteries, J. Electrochem. Soc. 162 batteries, Adv. Energy Mater. 2 (2012) 742e769.
(2015) A2188eA2194. [77] Z. Song, H. Zhou, Towards sustainable and versatile energy storage devices: an
[60] P.J. Cabrera, X. Yang, J.A. Suttil, K.L. Hawthorne, R.E.M. Brooner, M.S. Sanford, overview of organic electrode materials, Energy Environ. Sci. 6 (2013)
et al., Complexes containing redox noninnocent ligands for symmetric, mul- 2280e2301.
tielectron transfer nonaqueous redox flow batteries, J. Phys. Chem. C 119 [78] R.T. Boere
, T.L. Roemmele, Electrochemistry of redox-active Group 15/16
(2015) 15882e15889. heterocyles, Coord. Chem. Rev. 210 (2000) 369e445.
[61] J.H. Kim, K.J. Kim, M.S. Park, N.J. Lee, U. Hwang, H. Kim, et al., Development of [79] A. Jain, S.P. Ong, G. Hautier, W. Chen, W.D. Richards, S. Dacek, et al., The
metal-based electrodes for non-aqueous redox flow batteries, Electrochem. Materials Project: a materials genome approach to accelerating materials
Commun. 13 (2011) 997e1000. innovation, Apl. Mater 1 (2013) 11002.
[62] J. Mun, M.J. Lee, J.W. Park, D.J. Oh, D.Y. Lee, S.G. Doo, Non-aqueous redox flow [80] RFB Dashboard, The Materials Project, 2016. https://materialsproject.org/
batteries with nickel and iron tris (2, 20 -bipyridine) complex electrolyte, #apps/RFBDashboard.