Escolar Documentos
Profissional Documentos
Cultura Documentos
Dirk Henkensmeier
Fuel Cell Research Center,
Korea Institute of Science and Technology,
Seoul 02792, South Korea;
ET-GT,
University of Science and Technology,
Seoul 02792, South Korea;
Green School,
Korea University,
Seoul 136-713, South Korea
Sangwon Kim
Transfercenter Sustainable Electrochemistry,
Saarland University,
Saarbr€ucken 66125, Germany;
Bio Sensor and Materials Group,
KIST Europe,
Campus E7 1,
Saarbr€ucken 66123, Germany
1
Ruiyong Chen
Transfercenter Sustainable Electrochemistry,
Saarland University,
Saarbr€ucken 66125, Germany;
Bio Sensor and Materials Group,
KIST Europe,
Campus E7 1,
Saarbr€ucken 66123, Germany
e-mail: r.chen@kist-europe.de
1
Corresponding author.
Manuscript received May 15, 2017; final manuscript received July 5, 2017;
published online September 19, 2017. Assoc. Editor: Kevin Huang.
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-1
C 2018 by ASME
Copyright V
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-3
attractive again to investigate AEMs with these functional poly(phenylene)-based AEM in VRFBs (Fig. 4). It is known that
groups. Zhang et al. [62] modified chloromethylated Radel with CEMs based on a similar structure degraded in the VRFBs,
pyridine. Zhang and coworkers [63] prepared crosslinked despite the fully aromatic structure. The similar materials, func-
membranes by adding bipyridine to a casting solution containing tionalized with quaternary ammonium groups, showed an
chloromethylated Udel. Zhao and coworkers [64] followed a very improved stability in VO2þ solution. AEMs with an intermediate
similar protocol, but crosslinked bromomethylated PPO (poly IEC value (0.8 mmol g1) showed the lowest vanadium perme-
(2,6-dimethyl-1,4-phenylene oxide) with 1,2-bis(4-pyridyl)ethane, ability (1.4 106 cm2 min1 versus 2.1 106 cm2 min1 for
which is a more flexible linker than bipyridine. All three groups Nafion 212) and a low area resistance of 0.21 X cm2 (compared to
reported excellent chemical stability, high CE, and good VE. Lu 0.32 X cm2 for Nafion 212). In general, the IEC value was shown
et al. [65] attached methylated imidazolium groups to a bromome- to be proportional to degradation rate and VE and negatively cor-
thylated poly(ether ether sulfone) backbone. Membranes with an related to the CE. A direct comparison between a CEM and an
IEC of 1.5 and 2.0 showed a very low permeability. A membrane AEM, both based on Diels–Alder polymers, showed unexpected
with a high IEC of 3.5 showed a permeability close to Nafion 117. findings [70]. Some sulfonated polymers showed a similar CE as
Another big challenge for development of hydroxide- the corresponding AEM, similar to that with Nafion 117. It is con-
conducting membranes is the alkaline degradation of the polymer sidered that the transport processes are strongly affected by water
backbone. It is considered that polymers with only C–C bonds uptake. The bulky, rigid structure of the Diels–Alder polymer
would be more stable than polymers with heteroatoms in the main chains may lead to a high free volume, limiting the effect of the
chain [66]. Similarly, for an Udel membrane functionalized and ionic groups. Finally, a Diels–Alder polymer, in which the ammo-
crosslinked by bipyridine, it was postulated that VO2þ attacks the nium groups are tethered by C6-spacers, was tested for application
ether groups, leading to chain scission [67]. For a sulfonated in a nonaqueous ionic liquid-based flow battery [71]. Membranes
Radel, Hickner and coworker proposed a mechanism which with an IEC of 1.5 were too brittle to be useful, while membranes
involves attack of a V-radical on the electron-rich phenyl rings of with an ICE of 2.5 showed excessive movement of solvent across
the biphenyl unit (which is activated by the ether groups in posi- the membrane. Therefore, an IEC of 2.0 was mostly chosen. No
tion 4 and 40 ) [48]. This suggests that for more stable polymer chemical degradation was found. An increase in membrane resist-
backbones, it is necessary to reduce the electron density of the ance was observed, probably due to mechanical and thermal
phenyl rings, to enhance phase separation between the ionic aging.
groups and the hydrophobic backbones, and to prevent access of The strategy to embed metal oxide nanoparticles into mem-
the reactive species to the backbone. Recently, Zhang et al. [68] branes was also applied to AEMs. Recently, Ramani and
prepared poly(ether ether ketone)s with an adamantane moiety in coworkers [72] added a quaternized silesquioxane additive to a
the backbone. They hypothesized that the adamantane segments quaternized cardo-poly(ether ketone) membrane. The additive
restrict water uptake and membrane swelling, leading to a high was formed in the casting solution, by adding N-(trimethoxysilyl-
stability in the VRFB. As expected, all the AEMs showed a factor propyl)-N,N,N-trimethylammonium chloride to the polymer solu-
about 20 times lower vanadium crossover than Nafion 117. In an tion. Positively charged groups were added to increase the
ex situ stability test in a 0.15 M VO2þ solution, a higher IEC conductivity, while maintaining chemical, mechanical, and ther-
value leads to faster degradation. In terms of efficiencies, the mal stability. Most importantly, by introducing additives and nar-
AEMs with an IEC of 1.8 showed a slightly improved CE and EE. row channels (which form around the nanoparticles) and
Polymers without heteroatoms in the main chain have also been increasing the number of positively charged groups in the chan-
studied. Fujimoto and corkers [69] evaluated a Diels–Alder nels, it is expected that the vanadium crossover can be further
Fig. 4 A Diels–Alder poly(phenylene) with quaternary ammonium groups (left) and sulfonic
acid groups (right)
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-5
Fig. 5 (a) and (b) Cross-sectional SEM images of a CMPSF membrane immersed for 3 days in an imidazole solu-
tion, (c) dependence of the efficiencies on the current densities, and (d) cycling test at 120 mA cm22. (Reproduced
with permission from Zhao et al. [14]. Copyright 2016 by Wiley.)
conductivity increases exponentially with the phosphoric acid molecules. He and coworkers [82] preswelled membranes in phos-
uptake [78]. The area resistance of BIpPBI was about half of that phoric acid, followed by immersion in a sulfuric acid solution.
of PBI. An ex situ test showed a slightly increased VO2þ perme- Apparently, the ADL increased, leading to reduced resistance. In
ability (but far below that for Nafion 115), as expected for a mem- order to further minimize the membrane resistance, a porous
brane with higher acid uptake. After 200 cycles at 50 mA cm2, O-PBI membrane was fabricated recently with a 2–5 lm thin
both mPBI and BIpPBI showed a capacity retention of >80%, defect-free dense blocking layer [12]. When the porogen content
while Nafion 115 already lost >50% of the capacity after 100 (dibutylphthalate) reached 250%, the highly porous membranes
cycles. showed a similar resistance to Nafion 211, but no measurable
As discussed, the positively charged PBI backbones hinder VO2þ permeability. The EE reached 92% and 82% at 20 and
vanadium ions in the electrolyte solution from entering the poly- 100 mA cm2, respectively.
mer. Li and coworkers [13] hypothesized that establishing several
liquid/polymer interfaces, as in a membrane with a closed pores
structure, would block vanadium ions even more efficiently. A 3 Electrodes
cell with a 68-lm thick porous O-PBI membrane can withstand Electrodes used in an RFB are required to have a high surface
over 13,000 cycles between 80 and 120 mA cm2, with good effi- area, suitable porosity, good wettability for electrolytes, low elec-
ciencies (CE close to 100%, VE and EE around 80% after 13,500 tronic resistance, high activity toward the redox reactions (while
cycles, see Fig. 7). Furthermore, the material was used in a 1 kW suppressing water electrolysis), and good (electro)chemical stabil-
stack, which showed stable efficiencies over 100 cycles at 120 mA ity. Due to the commonly corrosive environment in a RFB, there
cm2. Such porous PBI membranes are promising for VRFBs in are limited choices of materials to fabricate electrodes. Carbon-
terms of performance and cost [13]. based materials have been used as electrodes in various RFBs
Moon and coworkers [16] reported the use of porous PBI copol- such as VRFBs, polysulfide/Br, Zn/Br, and V/Ce flow batteries
ymer membranes, which contain 50% para-phenyl and 50% C7 [83–101]. Because of inadequate electrochemical activity, cata-
alkyl chain links. The pores were rather large. Because of the lysts have been employed for facilitating the redox reactions
additional interfacial resistance and poor transport of vanadium [102,103]. So far, electrodes for VRFBs have been mostly investi-
ions through PBI membranes, the CE of a porous PBI membrane gated, as will be discussed in this section.
was higher than that of a dense PBI membrane. An improved VE Since the VRFB was first proposed by Skyllas-Kazacos et al.
was obtained. Wessling and coworkers [17] compared abPBI- [104], various approaches have been reported to modify the
based membranes. Again, a porous membrane filled with sulfuric carbon-based electrode materials, including thermal treatment,
acid showed a higher conductivity than the dense membrane. chemical treatment, electrochemical oxidation, and metal doping.
The chemical structure of PBI can be varied. Not only can the Sun and Skyllas-Kazacos [97] adopted heat treatment of graphite
backbone structure be freely modified (Fig. 6) but it is also possi- felts (GFs) (at 400 C for 30 h) to improve its EE for the VRFBs.
ble to crosslink PBI or to sulfonate activated phenyl rings. Zhang The improved performance was attributed to the enhanced surface
and coworkers [79] prepared a series of crosslinked PBI mem- hydrophilicity and the formation of the oxygen functional groups
branes, in which the ratio of an aniline moiety and a sulfonated (C–O–H and C¼O groups) on the graphite felt surface, which act
phenylether was varied. These membranes showed stable CE as active sites to catalyze the electrochemical reactions, as illus-
close to 100% and EE above 80% for over 300 cycles at 60 mA trated in Fig. 8 [101]. Another method to modify the surface of
cm2. Note that the crosslinker is bisphenol A based, a structure graphite felts is chemical treatment with nitric acid or sulfuric
which is known to be acid labile and therefore will degrade over acid [96]. A significant enhancement in VRFB performance was
time [80,81]. achieved (EE of about 88% at 25 mA cm2) after the graphite felt
Since the membrane resistance decreases with the acid uptake, was treated with concentrated sulfuric acid solution for 5 h. A
efforts were made to increase the acid uptake. It is known that combination of thermal and chemical treatments has also been
PBI membranes absorb more phosphoric acid molecules per employed [105]. The graphite felt was first treated in 98 wt. % sul-
repeat unit (the acid doping level (ADL)) than sulfuric acid furic acid solution for 5 h and then kept at 450 C for 2 h. The
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-7
Fig. 7 (a) Cross-sectional SEM image of a 68 -lm thick porous O-PBI membrane, (b) 12 kW
stack made with that membrane, cycling performance for (c) a single cell, and (d) a 1 kW
stack. (Reproduced with permission from Yuan et al. [13]. Copyright 2016 by Royal Society of
Chemistry.)
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-9
(7)
Negative side
(8)
Although this flow battery yielded a peak power density of 0.6 W cm2 at 1.3 A cm2 and a high CE of around 95%, crossover of bro-
mine occurred. Bromine can react with the hydroquinone and oxidize it back to quinone, resulting in capacity loss. Yang et al. [151]
introduced an aqueous all-organic RFB, using 0.2 M BQDS in 1 M H2SO4 as catholyte and 0.2 M 9,10-anthraquinone-2-sulfonic acid
(AQS) as anolyte, equipped with carbon paper electrodes and a Nafion 117 membrane:
Positive side
(9)
Negative side
(10)
Argon gas was purged to the solutions to avoid reoxidation of the reduced form to quinone. A power density of only 0.025 W cm2 was
achieved. Later, the researchers demonstrated another aqueous organic RFB with BQDS (catholyte) and AQ26DS (anolyte), which
showed an EE of 70% at 100 mA cm2 and a high CE (about 100% over 100 cycles) [154]:
Positive side: the same as Eq. (9).
Negative side
(11)
AQDS exhibited fast reaction kinetics (a rapid two-electron redox reaction), high stability, low membrane crossover, and high aqueous
solubility (1 M at pH 0).
Li and coworkers [158] reported 3,4-dihydroxy-9,10-anthraquinone-2-sulfonic acid (ARS) with a redox potential of 0.082 V. For the
RFB test, 0.05 M ARS in 1 M H2SO4 and 0.05 M BQDS in 1 M H2SO4 were used as anolyte and catholyte, respectively:
Positive side: the same as Eq. (9), E0 ¼ 0.908 V (note that this is different from a previously reported value of 0.85 V [151,154])
Negative side
(12)
The electrolytes were sealed to prevent oxidation and pumped into a cell with carbon paper electrodes and a Nafion 212 membrane. A
CE of 99% was achieved. A maximal power density of 10.6 mW cm2 at 80% state-of-charge has been obtained.
All the previously mentioned AQs were tested under acidic conditions. An alkaline quinone RFB was first introduced by Aziz and
coworkers [159]. A commercially available 2,6-dihydroxy-9,10-anthraquinone (2,6-DHAQ), which has a solubility of >0.6 M in 1 M
KOH, was used as negative material. Ferrocyanide was used as positive material:
Positive side
(13)
Negative side
(14)
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-11
(18)
Negative electrode reaction
(15)
(19)
The first reduction step is highly reversible. The second reduction This flow battery with a high concentration and a high voltage
is less reversible, because the fully reduced state is an insoluble delivered an energy density of 126 Wh L1. Another flow cell
uncharged species. Methyl viologen (1,10 -dimethyl-4,40 -bipyridi- with a lithium–graphite hybrid anode, using a diluted catholyte of
nium, MV) is chemically stable, but will dealkylate in alkaline 0.1 M TEMPO in 1.0 M LiPF6, was able to operate at 10 mA
solutions [162]: cm2 with an EE above 75% and a CE above 99%. An enhanced
EE (86%) was achieved when operating at 5 mA cm2. Over 100
cycles, the average capacity retention was 99.8% per cycle.
A 4-methoxy-2,2,6,6-tetramethylpiperidine-1-oxyl, a methoxy-
(16) substituted TEMPO (MeO-TEMPO) was investigated by Takechi
The potential of the first reduction for MV dichloride (MVCl2) is et al. [169]. MeO-TEMPO was mixed with lithium bis(trifluoro-
0.446 V and the second one is 0.88 V [163]. The solubility of methanesulfonyl)imide (LiTFSI) at a molar ratio of 1:1 to form a
MV in water can reach 2.6 M [164]. Aqueous RFBs using violo- self-melting liquid, which could be recognized as a super-cooled
gens were demonstrated by Liu et al. [165] and Schubert and liquid. This liquid, with a concentration of over 2 M consisting of
coworkers [21,166]. Viologens were combined with TEMPO- 17 wt. % of water, was used as catholyte. Li-metal and 1 M
based species in these RFBs, which will be discussed later. LiTFSI in propylene carbonate were used as the anode and elec-
trolyte, respectively. A piece of Liþ-conducting glass ceramics
was employed as separator. MeO-TEMPO exhibited a higher
4.4 TEMPO. TEMPO (2,2,6,6-tetramethylpiperidine-1-oxyl)
redox potential than TEMPO. The charge and discharge capacities
is a heterocyclic nitroxide radical compound. TEMPO and its
at 0.1 mA cm2 were 93% and 92% of the theoretical capacity,
derivatives were investigated and used as organic electrode mate-
respectively. The CE was 99% and the capacity retention was
rials [167]. The nitroxide radical has two redox couples:
84% after 20 cycles. The energy density and specific energy were
200 Wh L1 and 155 Wh kg1, respectively.
By introducing the hydrophilic –OH group at the para-position
for the nitroxide radical compound, Liu et al. [165] obtained a
water solubility up to ca. 2.1 M and demonstrated an all-organic
aqueous RFB using 4-hydroxy-TEMPO as catholyte and MV as
(17) anolyte (0.5 M for both redox-active materials in 1.5 M NaCl):
Positive side
(20)
Negative side
(21)
Fig. 11 (a) A schematic representation of a polymer-based RFB. The anolyte and catho-
lyte are separated by a size exclusion membrane and (b) electrode reactions. (Reproduced
with permission from Janoschka et al. [21]. Copyright 2015 by Nature Publishing Group.)
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-13
5 Suspension Electrolytes
Different from the conventional solid-state lithium-ion bat-
teries, RFBs store energy in redox-active electrolytes, contained
in external reservoirs. The energy density of RFBs is limited by
the concentration of the active species (typically below 2 M) and
the narrow electrochemical potential window of aqueous electro-
lytes (1.23 V). The development of semi-solid flow batteries,
with a high concentration of active species (in solid form) up to
10–40 M, can lead to an enhanced energy storage capability [9].
As illustrated in Fig. 14, in the semi-solid flow batteries, the solid
active compounds and the conductive additives are well mixed in
the electrolytes [174–178]. With sufficient conductive additives,
an electron transfer percolation network can form, which ensures
that the active materials can effectively accept or dispatch
charges. In this section, we discuss several reported semi-solid
suspensions, including inorganic intercalation-based materials,
redox-active organic compounds, and polysulfides.
Fig. 14 (a) Schematic illustration of a semi-solid RFB and (b) a percolation network
between the suspension
form at the graphite anode due to the decomposition of alkyl car- in the liquid phase and fast redox reaction kinetics of ferrocene/
bonate solvents [9], which hindered the electron transfer between ferrocenium, the full-cell showed small polarization and high
the fluid electrode and the current collector. Using anode materials capacity retention (94% of the theoretical capacity of ferrocene at
with a suitable lithiation potential (1.55 V versus Liþ/Li for 60 C). The power density and energy density based on the vol-
Li4Ti5O12, for instance) can minimize the formation of SEI [179]. ume of catholyte reached 1400 W L1 and 40 Wh L1, respec-
Biendico et al. [180] studied the role of carbon black on tively. Excellent cycling stability (a capacity retention of 80%)
the electrochemical and rheological properties based on was observed over 500 cycles.
LiNi1/3Co1/3Mn1/3O2 suspensions. With an increase in the volume Tomai et al. [191] reported a promising approach to impregnate
percentage of Ketjenblack (KB), the electrolyte resistance and nanoporous carbon beads with redox-active tetrachlorohydroqui-
charge transfer resistance decreased. Meanwhile, the viscosity of none and 1, 5-dichloroanthraquinone to form a flowable, low vis-
the suspensions increased. During static and dynamic measure- cosity slurry electrolyte. Through shuttling protons between the
ments, large capacities can be delivered as the volume percentage slurry electrolytes, the energy density of this flow cell was
of KB increased. Moreover, Lewis and coworkers [181] intro- 19.4 Wh kg1 at 0.26 A g1, based on the total weight of solid
duced a nonionic dispersant of polyvinylpyrrolidone (PVP) into compounds.
the LiFePO4/KB suspensions. Such electrode suspensions allowed
for high content of active materials, without deteriorating the flow
behavior and electronic conductivity. Recently, carbon-free sus- 5.3 Li–S Flow Battery. A concept of semi-solid Li/polysul-
pensions with a low viscosity have been reported, functioning fide batteries was introduced by Cui and coworkers [192]. Carbon
through the collision of active particles with the current collector fiber paper was used as current collector for the polysulfide catho-
[182,183]. However, only a low concentration and low operating lyte and assembled against a passivated Li film. The catholyte was
currents (from 0.001 to 0.02 mA) were applied. The utilization of cycled only between sulfur and Li2S4, to avoid the formation of
the active particles was too low from their preliminary work. insoluble Li2S2 and Li2S. When the concentration of Li2S8
An aqueous RFB based on LiFePO4–LiTi2(PO4)3 was first reached up to 5 M, an energy density of 108 Wh L1 was
reported by Chiang and coworkers [184]. Aqueous suspensions achieved, which is about three times that of the VRFBs. The addi-
with high ionic conductivity are of interest for safe and cost- tion of LiNO3 in the ether solvent contributed to the formation of
effective applications [7]. In static cell configuration, the suspen- an SEI film, which prevented parasitic reactions between the Li
sion containing KB in 1 M LiNO3 aqueous electrolyte showed a anode and the polysulfide.
capacity retention of about 85% after 100 cycles at 10 mA cm2. Lu and coworkers [193] investigated a concentrated sulfur-
Side chemical reactions, such as (1) oxidation reactions due to the impregnated carbon suspension in order to increase the volumetric
presence of dissolved oxygen in the aqueous electrolytes, (2) capacity of a Li/polysulfide RFB. The catholyte with 20 vol % sul-
hydrolysis reaction, and (3) precipitation reaction, caused capacity fur and 26 vol % KB showed a higher volumetric capacity of up
fade over time. to 155 Ah L1 at 4 mA cm2. Furthermore, Chen and Lu [194]
Naþ-based batteries have attracted increasingly attention due to demonstrated a catholyte suspension with mixed liquid lithium
abundance in resources. Recently, Ventosa et al. [185] applied iodide and solid sulfur/carbon active components. As observed
Naþ intercalation materials as suspensions for nonaqueous RFBs, from the cyclic voltammetry analysis for the individual S/S2
using a P2-type NaxNi0.22Co0.11Mn0.66O4 suspension cathode and redox couple (Fig. 15(a)), I3/I redox couple (Fig. 15(b)), and
a NASICON-type NaTi2(PO4)3 suspension anode. NaTi2(PO4)3 their mixture (Fig. 15(c)), there were no electrochemical interfer-
has a flat potential plateau at about 2.1 V versus Naþ/Na and is ences, confirming that LiI is compatible with sulfur. A high volu-
suitable as SEI-free anode [186]. The full-cell delivered a dis- metric capacity of 550 Ah L1 catholyte has been achieved using such
charge capacity of 80 mAh g1 (based on cathode material) at multiple redox reactions of LiI and sulfur. For further application,
0.17 mA cm2 with an average discharge voltage of about 0.8 V. it is necessary to replace the lithium metal anode with semi-solid
Large overpotential was observed from the positive electrode. negative flow electrode, such as silicon anolyte [195]. Table 1
summarizes the test conditions and performance for several
5.2 Organic Materials. Metallocenes have demonstrated selected compounds.
good electrochemical performance in nonaqueous electrolytes Considering the high-energy consumption for pumping suspen-
with high-energy density and CE [187,188]. Ferrocene (Fc) den- sions with a typically high viscosity, Carter et al. [196] modeled
drites were first introduced as a redox mediator in RFBs by Wang the hydrodynamic (residence time of a fluid volume within a cell)
and coworkers [189]. Recently, a novel membrane-free semi-solid and electrochemical efficiency properties (time required to
RFB with a high cell voltage (3.0 V) using a ferrocene-based charge/discharge a single volume of suspensions within a cell) for
catholyte and a passivated lithium anode has been investigated the semi-solid RFBs. High-flow rates increase the pumping-
by Yu and coworkers [190]. Owing to the fast mass diffusion energy dissipation, whereas low-flow rates will give rise to shunt
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-15
Capacity/ Average
Testing Current energy voltage CE
Cathode Anode Electrolytes model density density (V) (%) Ref.
11.87 vol % LiNi1/3Co1/3 Li LiPF6 þ ethylene carbonate Static 2.5 mA cm2 48.8 Ah L1 About 3.6 90 180
Mn1/3O2 þ 13.97 vol % (EC)/dimethyl carbonate
KB (DMC) (1:1 w/w)
20 vol % LiFePO4 Li Propylene Intermittent 1.67 mA cm2 93 Wh L1 — 72.3 181
þ 1.25 vol % KB carbonate þ LiTFSI
þ 0.3 wt % PVP
NaxNi0.22Co0.11 NaTi2(PO4)3 NaPF6 in EC/DMC Static 0.17 mA cm2 9 Wh L1 About 0.9 88 185
Mn0.66O4 þ 1.3 wt % KB þ 1.3 wt % KB
Tetrachlorohydroquinone 1,5- dichloroanthraquinone A mixture of 0.5 M H2SO4 Static 0.26 A g1 19.4 Wh kg1 0.55 — 191
þ 9.8 wt % nanoporous þ 9.8 wt % and 0.05 M HCl aqueous
carbon nanoporous carbon solutions
20 vol % sulfur Li 0.2 M LiClO4/0.1 M Static 4 mA cm2 221 Ah L1 2.1 80 193
þ 26 vol % KB LiNO3 þ 1,3-dioxolane
(DOL)/1,2-dimethoxy-
ethane (DME) (1:1 v/v)
5 M LiI-20 vol % S Li LiClO4/LiNO3 þ DOL/ Static 2.5 mA cm2 550 Ah L1 About 1.9 95 194
þ 26 vol % KB DME (1:1 v/v) (based on
catholyte)
currents. In addition, at low-flow rates, the presence of gradients Electric Industries (SEI) have developed VRFB systems for a wide
in the state-of-charge, current distribution, and overpotential along range of applications [206]. They demonstrated that overall EE of
the length of the flow channels needs to be considered. Rational the previously mentioned systems could be maintained about 80%
operational mode of semi-solid flow cells is thus important. In over 270 thousand cycles [207]. To store wind energy, in 2005 SEI
addition, a suitable battery module design may improve the rheo- installed 4 MW/6 MWh VRFB plant at Subaru Wind Farm in Japan
logical properties of the suspension and reduce the energy con- [208]. The power of each stack is 45 kW, which consisted of 108
sumption from the pump. cells. Recently, MW-class energy storage system has emerged to
meet the market needs. Although SEI demonstrated already several
6 Systems MWh scale VRFB systems, they seem to be impractical due to the
expensive cost. For large-scale systems, the modularization and
6.1 Historical Development. The earlier Fe/Cr systems optimization are important. It is necessary to understand the sys-
(1 kW prototype) were developed in 1980 by NASA Glenn tem configuration from the engineering perspective.
Research Center (Cleveland, OH) [197]. Since then, several
commercial attempts were followed [198–200]. Although many
pilot-scale systems were demonstrated, they were not sufficiently 6.2 System Configuration. A VRFB system is mainly cate-
satisfactory for commercial systems [201]. Commercial attempts gorized by the cell stack, electrolyte storage, and the balance of
using the polysulfide-Br flow batteries were conducted by Regen- plant (BOP). Electrolytes are circulated between the tanks and the
esys [202]. Moreover, Exxon, Johnson Controls (Milwaukee, WI), stack during operation. VRFB system needs a power conditioning
and Meidensha developed large-scale Zn/Br flow batteries. NASA system and a system controller. Through the proper design of
Glenn Research Center and the U.S. Electric Power Research direct current (DC)/alternating current (AC) converter, the system
Institute (Palo Alto, CA) installed a large-scale electric storage can obtain high round-trip efficiency. In the case of system con-
system using a Fe/Cr RFB [203]. Japanese industries featured troller, there is a battery management system (BMS) to operate
prominently for the early development of RFBs. The Japanese the system safely and to maximize the stack performance and sys-
Ministry of International Trade and Industry initiated a tem lifetime by reducing undesired side reactions and local deple-
“Moonlight Project.” Kansai Electric Power Co., Inc. (KEPCO, tion of the redox reactants. A sufficient quantity of electrolyte
Amagasaki, Japan) tested 60 kW Fe/Cr redox battery prototypes should be circulated between storage tank and stack for preventing
[203]. Meidensha Electric Company developed a Zn/Br battery, side reactions. The upper and lower limits of the state-of-charge
and Furukawa Electric Company developed a Zn/Cl battery [203]. are approximately 80–85% and 15–20%, respectively, to avoid
Although previous studies can demonstrate the technical possibil- undesired side reactions. Furthermore, a VRFB system typically
ities, commercialization of large-scale system was not successful, uses centrifugal pumps to prevent leaking or corrosion [209].
due to inherent cross-contamination. Low energy density and high VRFB systems have several advantages from the engineering
primary capital cost also restricted the commercialization [103]. aspects. First, such systems have dominant feature for safety and
In contrast, VRFB systems were successfully commercialized thermal management. Flowing electrolytes can remove heat gen-
since 1990s. The first system appeared in 1993. Thai Gypsum erated in the cell stack. Second, the unique system configuration
Products, Ltd., installed a vanadium battery system for Solar can simplify the manufacturing process. Through the modulariza-
House in Thailand in cooperation with UNSW Center for Photo- tion of stacks, electrolyte containers, plumbing, and electrical sys-
voltaic Devices and Systems [204]. This integrated system tems, VRFB can reduce expensive and complicated production
included solar cells of 2.2 kW and the VRFB system with 12 kWh. steps. Finally, the rebalancing process in the VRFB system is sim-
They developed a 47 V, 36 cell stack for long-term field trials. In ple. For VRFB systems, it is necessary to manage the active spe-
1997, Mitsubishi Chemicals (Kashima-Kita Electric Power, cies crossover and water transport through the membrane.
Tokyo, Japan) installed a 200 kW/800 kWh grid-connected vana- Each single cell of a VRFB stack includes several components,
dium battery for load-leveling. This system could run over 150 such as the ion-exchange membrane, electrodes, and bipolar
charge–discharge cycles and showed a high EE of 80% at plates. In addition, nonconductive plastic frames and sealing
80–100 mA cm2 [205]. Since the installation of a VRFB system materials are also required. A complex manifold design is often
in 1996 at Tatsumi Sub-Station for peak shaving, Sumitomo necessary to reduce undesirable shunt currents in the ionically
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-17
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-19
Journal of Electrochemical Energy Conversion and Storage FEBRUARY 2018, Vol. 15 / 010801-21