Designing Cr complexes for a neutral Fe-Cr redox flow battery

Article abstract of DOI:10.1039/c9cc09704j

We lay out the design principles of Cr complexes to address issues of slow kinetics and parasitic reactions in the Fe-Cr redox flow battery (ICRFB). We identify theoretically and experimentally dipicolinic acid as a promising ligand, and synthesize its derivative to improve the solubility of the Cr complex to 0.7 M. We couple it with ferrocyanide for a neutral ICRFB delivering 120 stable cycles.

Full text of DOI:10.1039/c9cc09704j

ChemComm
View Article Online
View Journal
COMMUNICATION
Designing Cr complexes for a neutral Fe–Cr redox
flow battery†
Cite this:DOI: 10.1039/c9cc09704j
bc
Wenqing Ruan,abc Jiatao Mao,bcd Shida Yang,bcd Chuan Shi,b Guochen Jia
and
abcd
Received 17th December 2019,
Accepted 11th February 2020
Qing Chen
*
DOI: 10.1039/c9cc09704j
rsc.li/chemcomm
We lay out the design principles of Cr complexes to address issues of in the positive side, the reduction potential of Cr(EDTA)ꢀ/2ꢀ was
slow kinetics and parasitic reactions in the Fe–Cr redox flow battery too low to avoid hydrogen evolution even in a neutral electrolyte.
(ICRFB). We identify theoretically and experimentally dipicolinic acid Robb et al. recently hypothesized that the severe hydrogen
as a promising ligand, and synthesize its derivative to improve the evolution during the reduction of Cr(EDTA)ꢀ was due to its
solubility of the Cr complex to 0.7 M. We couple it with ferrocyanide coordination with water, and they used a similar but larger ligand,
for a neutral ICRFB delivering 120 stable cycles.
1,3-diaminopropanetetraacetate to prevent water coordination.9
This complex, however, displays an even lower reduction potential,
The iron–chromium redox flow battery (ICRFB) was invented in at a high risk of a homogeneous reaction with water. Sun Catalytic
the 1970s, much before the popular all-vanadium redox flow Corporation suggested Cr(CN)63ꢀ,2 which is however unstable upon
battery (VRB). The head start did not make the ICRFB the winner reduction.10 Recent studies have attempted to use Cr-bipyridine11
due to issues such as slow charge transfer kinetics, hydrogen and Cr-acetylacetonate complexes12 in non-aqueous RFBs for a
evolution, and crossover.1,2 Mixing the Fe- and Cr-electrolytes is high energy density, but these complexes either dissolve poorly or
a common engineering solution, but it inevitably increases the undergo decomposition upon reduction in an aqueous solution.
system complexity and decreases the efficiency.1 The progress of
Our work focuses on Cr complexes stable in a neutral
ICRFB research has been stagnant, while VRBs enjoy a steady electrolyte, because acids in the traditional ICRFB invite hydrogen
development. Yet as the high and volatile price of vanadium evolution and deprive chromium of many common ligands that
continues to threaten its competitiveness in grid-scale applica- bind instead to protons. We avoid alkaline electrolytes where
tions, we are drawn back to the ICRFB because of the global chromium cations easily precipitate as hydroxides. A neutral
abundance and low cost of iron and chromium.
electrolyte can also bring additional benefits as follows. First, it
In this work, we use the complexation chemistry of chromium extends the negative end of the electrochemical window of water to
ions to revamp the ICRFB. Complexation has been widely used in accommodate the low reduction potential of Cr complexes. Second,
iron chemistry for the positive sides of RFBs.3,4 The best example is a neutral electrolyte is compatible with many iron complexes like
ferrocyanide, whose stability and considerably high potential have ferrocyanide. At last, a neutral electrolyte makes the battery safe
made it a fixture in neutral and alkaline flow batteries.5–7 Among and environmentally friendly, and lessens the infrastructure
few reports on Cr complexes, Cr-ethylenediaminetetraacetic acid requirements.
(EDTA) complex (Cr(EDTA)ꢀ/2ꢀ) was the focus of Bae et al.,8 given
We begin with Cr complexes from literature (Fig. 1a and Fig. S1,
the apparent reversibility on carbon cloth and the low redox ESI†) and employ two of their commonly known properties, the
potential. However, the as designed all-Cr RFB delivered a short association constant and the redox potential to screen for suitable
cycle life and low efficiencies. Apart from possible decomposition complexes. The redox potential, or more rigorously, the formal
0
potential of a pair of Cr3+ and Cr2+ complexes (E ) has to be low to
0
afford a high cell voltage, but not too low to evolve hydrogen during
reduction. We set the bounds as ꢀ1 to 0.6 V vs. the standard
hydrogen potential (SHE). The lower bound corresponds to the
typical potential of hydrogen evolution on carbon, and the upper
bound to allow a cell voltage of 1 V when paired with ferrocyanide.
Meanwhile, the stability here concerns the hydrolysis of both the
Cr3+ and Cr2+ complexes, referred to herein as CrIII(L)n and CrII(L)n
with L as the ligand. To simplify the discussion, we assume that a
a Department of Mechanical and Aerospace Engineering, HKUST, Hong Kong.
E-mail: chenqing@ust.hk
b Department of Chemistry, HKUST, Hong Kong
c The Energy Institute, HKUST, Hong Kong
d HKUST Shenzhen Research Institute, Shenzhen, China
† Electronic supplementary information (ESI) available: Derivation of the screening
criteria of Cr complexes, single crystal X-ray structure, EIS, UV-Vis spectrum, EPR
spectrum and experimental section. CCDC 1968718. For ESI and crystallographic
data in CIF or other electronic format see DOI: 10.1039/c9cc09704j
Chem. Commun.
This journal is © The Royal Society of Chemistry 2020

View Article Online
Communication
ChemComm
voltammetry (CV, Fig. 1c) was carried out in a three-electrode
cell in neutral electrolytes on a glassy carbon electrode. For
CrCl3 without any complexant, the poor electrochemical rever-
sibility is evident from the large overvoltage in the reduction
and the lack of oxidation peak even up to 0 V vs. SHE, similar to
previous reports.17 [Cr(IDA)2]ꢀ has too low a reduction potential
to show any current before hydrogen evolution, in a sharp
contrast to an earlier report of its electrochemical reversibility
on a mercury electrode,13 likely due to the differences between
carbon and mercury in catalyzing the reductions of water and
[Cr(IDA)2]ꢀ. [Cr(EDTA)2]ꢀ, instead, shows a pair of quasi-
reversible peaks, but the low reduction potential and the
relative slow kinetics push its cathodic peak well into hydrogen
evolution.
The CV of [Cr(DPA)2]ꢀ displays the most rapid kinetics.
Using the classic Nicholson’s method to fit the voltammetry,
we can estimate a lower limit for the charge transfer rate (k0) to
be 0.05 cm sꢀ1. It is four orders of magnitude higher than the
upper limit of k0 of Cr3+/2+, and comparable to that of
Fig. 1 Theoretical and experimental screening of Cr complexes. (a)
Molecular structures of CrIII investigated in this work. (b) A diagram of
formal potentials vs. association constants of CrIII complexes. The blue
area sets the lower limits for the potential and association constants for Cr
complexes suitable for a neutral ICRFB. Cyclic voltammetry of the synthe-
sized complexes and the solubilities of [Cr(DPA)2]ꢀ at the presence of
various supporting electrolytes. (c) CV of the complexes. The dashed lines
indicate the reported formal potentials of the species with the same color.
(d) The solubility at the addition of 0.3 M supporting cations. The inset is
the crystal structure of Na[Cr(DPA)2](H2O)2, highlighting the sodium cation
(light blue) tightly coordinated by the carboxylates.
3ꢀ/4ꢀ 3
Fe(CN)6
. The different CV behaviors among the tested
Cr complexes raise an intriguing question on the structural
dependence of the charge transfer rate. Previously the differ-
ences in the rates on a mercury electrode were attributed to the
structural effect of the ligands on the Jahn–Teller distortion
experienced by the CrII complex.13 This proposition would
suggest fast kinetics for [Cr(IDA)2]ꢀ and slow kinetics for
[Cr(DPA)2]ꢀ independent of the electrode, which is clearly
different from our observation. Therefore, other effects, such
as the interactions between the complexes and the carbon, play
complex will lose its electrochemical activity and precipitate only bigger roles here. The conjugated structure in the DPA
when fully hydrolyzed in the reactions CrIII(L)n + 3OHꢀ - Cr(OH)3 + ligand could lead to its dispersive interaction with the glassy
nL and CrII(L)n + 2OHꢀ - Cr(OH)2 + nL. Given association carbon electrode to promote charge transfer. We synthesized a
constants KCr(OH) = 1017 and KCr(OH) = 1030, to stabilize the Cr-bipyridine complex, which also shows decent kinetics, in
3
2
complex up to a concentration of 1 M at pH = 7, we can derive line with this hypothesis.
the lower bounds of KCrIIIL and KCrIIL to be 109 and 103, respectively
The above results support our earlier assessment of
(see ESI† for details). The three criteria above can be reduced to [Cr(DPA)2]ꢀ as a suitable candidate to be applied in a neutral
two given the relationships between KCrIIIL, KCrIIL, and E0 through RFB. In addition, its relatively large size would address the
0
the Nernst equation. Given that KCrIIL is rarely assessed experi- issue of crossover. In a home-made cell, we separated two
mentally due to the instability of CrII complexes against air, we chambers of electrolytes with a cationic-selective membrane
re-formulate its lower bound in terms of KCrIIIL and E0 (ESI†). The (Nafion 212). As monitored with UV-Vis spectroscopy,
0
result is a diagram of KCrIII and E0 (Fig. 1b), where a blue area [Cr(DPA)2]ꢀ in the donating side permeates through the
0
L
corresponds to the bounds of properties. The two horizontal membrane to the receiving side at a mere rate of 6.4 ꢁ 10ꢀ9
dashed lines correspond to the upper and lower limits for a cm2 minꢀ1. The permeability of Cr3+ measured in the same cell
desirable potential. The vertical dashed line separates regions of was estimated 1.3 ꢁ 10ꢀ5 cm2 minꢀ1, four orders of magnitude
sufficient (right) and insufficient (left) CrIII(L)n stabilities, and the higher.
diagonal line for the stability of CrII(L)n. Many of the complexes
have too low formal potentials. Iminodiacetate (IDA2ꢀ 13
picolinic acid (Picꢀ)14,15 do not bind to Cr2+ strongly enough. sized salt NEt4[Cr(DPA)2] can dissolve up to 0.4 M in water,
The dipicolinic acid (DPA2ꢀ 16
complex falls into the blue area, precipitate appears if we add another salt to the saturated
However, a shortfall of [Cr(DPA)2]ꢀ is its limited solubility in
)
and the presence of supporting electrolytes. Although the synthe-
)
and as we will show later, the complex does provide the desirable solution for a higher conductivity. As summarized in Fig. 1d,
properties. The diagram, though skimming over complications the solubility drops to below 0.1 M even with small cations like
such as the impact of partial hydrolysis, offers a convenient Na+. The difference in the solubilities can be understood based
assessment of Cr complexes for RFB applications.
on the cationic hydration energy, which produces the trend
With the insights above, we synthesized three Cr complexes, among the alkaline ion salts,18 and the lattice energy, which
[Cr(IDA)2]ꢀ, [Cr(EDTA)2]ꢀ and [Cr(DPA)2]ꢀ to be compared with limits the solubility of salts of divalent cations. The low
CrCl3 in terms of their electrochemical properties. Cyclic solubility of the ammonia salt is a result of hydrogen bonds
Chem. Commun.
This journal is © The Royal Society of Chemistry 2020

View Article Online
ChemComm
Communication
Table 1 Key properties of active species in neutral and acidic ICRFB.
Solubilities were measured in pure water. Values of k0 were estimated from
CV at 50 mV sꢀ1. The permeabilities through Nafion 212 for CrCl3 and
FeCl2 were estimated in one experiment by separating an electrolyte of
0.1 M CrCl3 from another of 0.1 M FeCl2. Although the measurement could
not separate the permeability values, it represents the scale of crossover in
the acidic ICRFB
that favor crystal formation. To understand the impact from the
added salt, we grew a single crystal of Na[Cr(DPA)2]ꢂ(H2O)2 and
analyzed it with X-ray diffraction (XRD). The inset of Fig. 1d shows
that each sodium cation bridges three [Cr(DPA)2]ꢀ via the coordina-
tion of oxygen atoms of a carboxylate group from each [Cr(DPA)2]ꢀ
unit. Two of the carboxylates work as a bidentate while the other
one as a monodentate ligand. The bond lengths are in the range of
2.395 to 2.460 pm,19–21 indicative of strong interaction underlying
the low solubility.
E0
Solubility (M) (V vs. SHE) (cm sꢀ1
k0
Permeability
0
)
(cm2 minꢀ1
)
CrCl3ꢂ6H2O
FeCl2ꢂ6H2O
1.66
3.85
ꢀ0.41
0.77
o10ꢀ6
10ꢀ5
10ꢀ3
0.1
10ꢀ5
To improve the solubility, we synthesized a derivative of DPA2ꢀ
,
10ꢀ5
[Cr( f-DPA)2]Br 0.71
ꢀ0.71
o10ꢀ9
10ꢀ9
the dipotasium salt of 2,6-pyridinedicarboxylic acid functionalized
with an ammonium group, 3-((2,6-bis(ethoxycarbonyl)pyridin-4-
yl)oxy)-N,N,N-trimethylpropan-1-aminium bromide, referred herein
K4Fe(CN)6
B0.60
0.36
as [f-DPA]Br. The compound was synthesized by the nucleophilic given a cation-selective membrane (Nafion 212). All other battery
attack of 3-bromo-N,N,N-trimethylpropan-1-aminium bromide with components are commercial products. Two other independent
deprotonated diethyl 4-hydroxypyridine-2,6-dicarboxylate (nuclear cells were tested, whose results are summarized in Fig. S6 and
magnetic resonance shown in Fig. S2–S4, ESI†). The complex, S7 (ESI†).
[Cr(f-DPA)2]+, can be easily synthesized by the reactions of CrCl3
with the ligands as shown in Fig. 2a.
Fig. 3a shows representative charging and discharging
curves. We ran the cell at 10 mA cmꢀ2 to a cutoff voltage of 1.4 V
The complex shows desirable properties for RFB applications. in charging to access most of the capacity without triggering
It displays a pair of reversible peaks near ꢀ0.7 V, much like hydrogen evolution. The first cycle has a low Coulombic efficiency
[Cr(DPA)2]ꢀ (Fig. 2b). In pure water, the solubility is about likely due to the residual oxygen in the system. The cell has an
0.71 M. The reason for this improved solubility, as confirmed open-circuit voltage 41.15 V, close to the difference in the formal
by the single crystal XRD (Fig. S5, ESI†), is that charge repulsion potentials of the two redox couples. At higher currents, the cell
and steric effects between the pendent ammonium groups delivers higher power, reaching 42 mW cmꢀ2 at 60 mA cmꢀ2
weaken the interaction between molecules and frustrate the (Fig. 3b), which is about the peak power if we linearly extrapolate
formation of a solid. As the complex carries a positive charge, the polarization curve. The linear polarization suggests resistances
its solubility is less sensitive to the addition of cations. In the as current-limiting factors. From the slope of the curve, we can
presence of 0.3 NaCl, the solubility stays 40.5 M.
estimate a polarization resistance of 9.45 O cm2, which is very close
Therefore, we pair [Cr(f-DPA)2]+ with Fe(CN)64ꢀ for a neutral to the high frequency resistance estimated via the electrochemical
ICRFB. In Table 1, we compare key properties of the active impedance spectroscopy (Fig. S8, ESI†). Since the electronic and
species of this neutral ICRFB to those of the conventional acid contact resistances of the electrodes are at the order of 0.1 O cm2,
ICRFB. With the complexation, we sacrifice the solubilities to the results suggest the membrane resistance as the major con-
address the more critical issues of the ICRFB. The kinetics of tributor, which can be an area of future improvement in both the
both sides in the neutral ICRFB, particularly the negative side, selections of the membrane and the supporting electrolytes. The
is drastically improved without even a catalyst. Both complexes charge transfer kinetics does not seem to limit the power output, in
show much lower rates of crossover. The complexation shifts an agreement with the rapid kinetics shown in CV.
the potentials of both sides negatively, but the shifts are
Fig. 3c shows relatively stable cycling performance of the
positive against a reversible hydrogen electrode, which will same battery at 10 mA cmꢀ2. The coulombic efficiency starts
alleviate the other critical issue of the ICRFB, the hydrogen below 85% likely due to the residual oxygen in the cell, and it
evolution. In the negative side, we have 0.13 M of [Cr(f-DPA)2]Br stays above 99.5% afterwards. The energy efficiency starts just
dissolved in 1 M KCl. In the positive side, 0.15 M K4Fe(CN)6 is below 90%, but slowly drops to 85%, likely due to the increase
dissolved in 0.4 M KCl to afford the same concentration of K+ in the membrane resistance as shown by post-cycling EIS
(Fig. S8, ESI†). The capacity also decays at a steady rate of
0.17% per cycle, retaining B73% of the initial capacity after
120 cycles. The loss can come from crossover, decomposition,
and leakage. We rule out crossover via UV-Vis spectroscopy,
which shows no peak from the complex in the positive electro-
lyte. The decomposition is suggested also by UV-Vis, which
shows a 28% decrease in the absorption peak of the Cr complex
(Fig. S9, ESI†), consistent with the capacity loss. No other peak
can be identified as the decomposition product. We further
analyzed the cycled electrolyte with electron paramagnetic
resonance, which shows an additional peak at low magnetic
Fig. 2 Cr complexes with DPA derivatives. (a) Synthetic routes to [Cr(f-
field absent in the spectra of either a [Cr(f-DPA)2]Br, f-DPA, or CrCl3
solution (Fig. S10, ESI†). We suspect that the decomposition is
DPA)2]Br. (b) CV of the complexes. CV of the free ligands is shown as the
dashed lines.
Chem. Commun.
This journal is © The Royal Society of Chemistry 2020

View Article Online
Communication
ChemComm
Fig. 3 A neutral ICRFB. (a) Charging and discharging curves at 10 mA cmꢀ2. (b) A polarization curve constructed by plotting the average discharging
voltage vs. the current density (black), which is used for calculating the power density (red). (c) Capacity and efficiencies vs. cycle number from
galvanostatic charging–discharging cycling at 10 mA cmꢀ2
.
initiated or at least aggravated by oxygen in the system, as the cell Notes and references
tested outside the glovebox shows a more rapid capacity loss
1 N. H. Hagedorn, NASA Redox Storage System Development Project,
(Fig. S9, ESI†). In control test, we retrieved the charged negative
electrolyte and exposed it to air. The solution quickly changed color
and showed no peak at 567 nm in its UV-Vis spectrum, suggesting a
complete loss of the complex (Fig. S4, ESI†). Therefore, instead of
just oxidizing the reduced complex back to the oxidized state,
oxygen likely catalyzes a decomposition reaction. Although we have
not identified the exact decomposition mechanism, further work
along this path will need to address the vulnerability of the complex
against oxygen or improve the containment of the cell to better
prevent the attack by oxygen.
NASA, 1984.
2 G. L. Soloveichik, Chem. Rev., 2015, 115, 11533–11558.
3 K. Gong, F. Xu, J. B. Grunewald, X. Ma, Y. Zhao, S. Gu and Y. Yan,
ACS Energy Lett., 2016, 1, 89–93.
4 N. Arroyo-Curras, J. W. Hall, J. E. Dick, R. A. Jones and A. J. Bard,
J. Electrochem. Soc., 2015, 162, A378–A383.
5 K. Lin, Q. Chen, M. R. Gerhardt, L. Tong, S. B. Kim, L. Eisenach,
A. W. Valle, D. Hardee, R. G. Gordon, M. J. Aziz and M. P. Marshak,
Science, 2015, 349, 1529–1532.
´
´
6 K. Lin, R. Gomez-Bombarelli, E. S. Beh, L. Tong, Q. Chen, A. Valle,
A. Aspuru-Guzik, M. J. Aziz and R. G. Gordon, Nat. Energy, 2016, 1, 16102.
7 J. Luo, B. Hu, C. Debruler, Y. Bi, Y. Zhao, B. Yuan, M. Hu, W. Wu and
T. L. Liu, Joule, 2019, 3, 149–163.
8 C. Bae, E. P. L. Roberts, M. H. Chakrabarti and M. Saleem, Int.
J. Green Energy, 2011, 8, 248–264.
9 B. H. Robb, J. M. Farrell and M. P. Marshak, Joule, 2019, 16,
2503–2512.
10 S. W. Feldberg and L. Jeftic, J. Phys. Chem., 1971, 75, 2381–2387.
11 P. J. Cabrera, X. Yang, J. A. Suttil, R. E. M. Brooner, L. T. Thompson
and M. S. Sanford, Inorg. Chem., 2015, 54, 10214–10223.
12 J. D. Saraidaridis, B. M. Bartlett and C. W. Monroe, J. Electrochem.
Soc., 2016, 163, A1239–A1246.
13 M. Hecht, F. A. Schultz and B. Speiser, Inorg. Chem., 1996, 35,
5555–5563.
In conclusion, we demonstrate the feasibility of designing
metal complexes for solving the critical issues of ICRFB. The
use of chemical equilibria in the ligand selection builds a
quantitative connection between physical chemical properties
and requirements by RFBs, which should be widely applicable
to other systems. The dependence of the charge transfer
kinetics, the stability, and the solubility on the structure of a
complex suggests molecular design as a promising path to
addressing challenges in RFBs running on metal ions. The
performance of the neutral ICRFB demonstrates the effective-
ness of complexation in stabilizing metal ions with a strong
tendency of hydrolysis at a moderate or high pH.
14 M. L. Lugo and V. R. Lubes, J. Chem. Eng. Data, 2007, 52, 1217–1222.
´
15 B. S. Parajon-Costa, C. C. Wagner and E. J. Baran, Z. Anorg. Allg.
Chem., 2003, 629, 1085–1090.
16 J. Drze˙zd˙zon, A. Piotrowska-Kirschling, J. Malinowski, A. Kloska,
´
B. Gawdzik, L. Chmurzynski and D. Jacewicz, J. Mol. Liq., 2019, 282,
441–447.
The work was supported by the Research Grants Council of
the Hong Kong Special Administrative Region, China (T23-601/
17-R) and the Science and Technology Innovation Commission,
Shenzhen under the grant number JCYJ20170307173900343.
17 C. D. Wu, D. A. Scherson, E. J. Calvo and E. B. Yeager, J. Electrochem.
Soc., 1986, 133, 2109.
18 Y. Marcus, J. Solution Chem., 1986, 15, 291–306.
19 V. Kumar, T. Pilati, G. Terraneo, F. Meyer, P. Metrangolo and
G. Resnati, Chem. Sci., 2017, 8, 1801–1810.
20 S. Wustoni, C. Combe, D. Ohayon, M. H. Akhtar, I. McCulloch and
S. Inal, Adv. Funct. Mater., 2019, 1904403.
21 G. Cavallo, P. Metrangolo, T. Pilati, G. Resnati, M. Ursini and
G. Terraneo, Acta Crystallogr., Sect. E: Struct. Rep. Online, 2013, 69,
m387–m388.
Conflicts of interest
There are no conflicts to declare.
Chem. Commun.
This journal is © The Royal Society of Chemistry 2020