Evolutionary Design of Low Molecular Weight Organic Anolyte Materials for Applications in Nonaqueous Redox Flow Batteries

Article abstract of DOI:10.1021/jacs.5b09572

The integration of renewable energy sources into the electric grid requires low-cost energy storage systems that mediate the variable and intermittent flux of energy associated with most renewables. Nonaqueous redox-flow batteries have emerged as a promising technology for grid-scale energy storage applications. Because the cost of the system scales with mass, the electroactive materials must have a low equivalent weight (ideally 150 g/(mol·e^-) or less), and must function with low molecular weight supporting electrolytes such as LiBF₄. However, soluble anolyte materials that undergo reversible redox processes in the presence of Li-ion supports are rare. We report the evolutionary design of a series of pyridine-based anolyte materials that exhibit up to two reversible redox couples at low potentials in the presence of Li-ion supporting electrolytes. A combination of cyclic voltammetry of anolyte candidates and independent synthesis of their corresponding charged-states was performed to rapidly screen for the most promising candidates. Results of this workflow provided evidence for possible decomposition pathways of first-generation materials and guided synthetic modifications to improve the stability of anolyte materials under the targeted conditions. This iterative process led to the identification of a promising anolyte material, N-methyl 4-acetylpyridinium tetrafluoroborate. This compound is soluble in nonaqueous solvents, is prepared in a single synthetic step, has a low equivalent weight of 111 g/(mol·e^-), and undergoes two reversible 1e^- reductions in the presence of LiBF₄ to form reduced products that are stable over days in solution.

Full text of DOI:10.1021/jacs.5b09572

Article
pubs.acs.org/JACS
Evolutionary Design of Low Molecular Weight Organic Anolyte
Materials for Applications in Nonaqueous Redox Flow Batteries
Christo S. Sevov,†,‡ Rachel E. M. Brooner,†,‡,# Etienne Chenard,†,§,# Rajeev S. Assary,†,⊥
́
Jeffrey S. Moore,†,§,∥ Joaquín Rodríguez-Lopez,†,§ and Melanie S. Sanford*,†,‡
́
†Joint Center for Energy Storage Research, Argonne, Illinois 60439, United States
‡Department of Chemistry, University of Michigan, 930 North University Avenue, Ann Arbor, Michigan 48109, United States
§Department of Chemistry, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801, United States
∥Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801,
United States
⊥Materials Science Division, Argonne National Laboratory, Argonne, Illinois 60439, United States
S
* Supporting Information
ABSTRACT: The integration of renewable energy sources into the
electric grid requires low-cost energy storage systems that mediate
the variable and intermittent flux of energy associated with most
renewables. Nonaqueous redox-flow batteries have emerged as a
promising technology for grid-scale energy storage applications.
Because the cost of the system scales with mass, the electroactive
materials must have a low equivalent weight (ideally 150 g/(mol·e−)
or less), and must function with low molecular weight supporting
electrolytes such as LiBF4. However, soluble anolyte materials that
undergo reversible redox processes in the presence of Li-ion
supports are rare. We report the evolutionary design of a series of pyridine-based anolyte materials that exhibit up to two
reversible redox couples at low potentials in the presence of Li-ion supporting electrolytes. A combination of cyclic voltammetry
of anolyte candidates and independent synthesis of their corresponding charged-states was performed to rapidly screen for the
most promising candidates. Results of this workflow provided evidence for possible decomposition pathways of first-generation
materials and guided synthetic modifications to improve the stability of anolyte materials under the targeted conditions. This
iterative process led to the identification of a promising anolyte material, N-methyl 4-acetylpyridinium tetrafluoroborate. This
compound is soluble in nonaqueous solvents, is prepared in a single synthetic step, has a low equivalent weight of 111 g/(mol·
e−), and undergoes two reversible 1e− reductions in the presence of LiBF4 to form reduced products that are stable over days in
solution.
Increasing the volume of the electroactive solution increases the
capacity of the system, while expanding the surface area of the
current collector increases the power. These features mitigate
mechanical fatigue, which is common in competing battery
technologies that involve deposition and dissolution of
electroactive materials at electrode surfaces. In addition, the
physical separation of the electroactive materials in RFBs
precludes exotherms that result from mixing.
These promising features of RFBs have motivated a techno-
economic model that provides quantitative guidelines for cost-
competitive grid-scale RFB storage systems.11 According to the
model, RFBs operating in nonaqueous media open up new
design space by taking advantage of high cell potentials without
decomposition of the solvent. To meet cost targets of $120/
kW·h, supporting electrolytes with low molecular weight, such
as LiBF4 (MW = 94 g/mol), are preferred to TBABF4 (TBA =
INTRODUCTION
■
Electrical energy is currently the single largest form of energy
consumed worldwide,1 and the consumption of electricity is
predicted to double by 2050. These increasing global demands
will require energy sources that do not contribute to the
accumulation of greenhouse gases or to the exhaustion of the
limited supply of fossil fuels.2 To address this challenge, major
effort is focused on the development of technologies that
convert renewables into electrical energy.3−6 However, the
integration of renewables into the electrical grid remains limited
due to the variable and intermittent nature of energy sources
like solar and wind. Energy storage systems (ESSs) that
respond rapidly to changes in flux from renewable sources will
enable the large-scale penetration of renewables into the grid.
Redox-flow batteries (RFBs) represent a promising technol-
ogy for grid-scale energy storage.7−10 In RFBs, anolyte and
catholyte fluids undergo electrochemical reactions as they are
passed over a current collector. As a result, the capacity and
power of the ESS scale independently, which reduces cost.
Received: September 10, 2015
© XXXX American Chemical Society
A
DOI: 10.1021/jacs.5b09572
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
tetrabutylammonium, MW = 329 g/mol). Additionally, to meet
cost and solubility targets, the anolyte and catholyte materials
should have a molecular weight-to-charge ratio (also known as
equivalent weight) that is less than 150 g per mole of stored
charge g/(mol·e−)). These techno-economic constraints
motivate fundamental research to identify and understand the
electrochemistry of small molecules that are highly soluble, are
chemically compatible with Li-ion supports, and exhibit one or
more reversible redox reactions at especially low (anolytes) or
high (catholytes) potentials.
counter electrode, and a Ag/Ag+ reference electrode. The CV
of each anolyte candidate was initially conducted in acetonitrile
(MeCN) with TBABF4 as the supporting electrolyte. These
results were then compared to those obtained in MeCN/LiBF4
under otherwise identical conditions. The comparison of these
electrochemical results provided a feedback loop that guided
the iterative design of new generations of anolyte materials.
The chemical reversibility of each redox couple in the CV
was estimated by calculating the ratio of the diffusion-limited
peak-heights of the cathodic (ipc) and anodic (ipa) currents as a
function of scan rate (Figure 2). Redox processes that generate
Figure 1. Representative half-cell reactions of RFB anolyte and
catholyte materials in a lithium-ion supported media.
The use of organic molecules as soluble anolytes for RFBs is
an attractive strategy for meeting the challenging molecular
weight-to-charge goals. Persistent organic radicals have been
known for more than a century,12 and many soluble catholyte
materials that exhibit reversible electrochemistry in Li-ion
electrolytes are known (Figure 1).13 Most catholyte systems
employ molecules with extensive conjugation, such as
quinones,14−17 thioquinones,18 arylamines,19 alkoxyarenes,20−22
thiophenes,23−25 thiadiazoles,26,27 and N-oxides.28,29 The
compatibility of these catholyte materials with Li-ion electro-
lytes and nonaqueous solvents has facilitated their coupling
with other Li-ion half-cells (e.g., lithium intercalation supports
and lithium metal).7,30−32 In contrast, anolyte materials based
on organic compounds that operate with Li-ion electrolytes are
rare and currently remain limited to insoluble lithium salts of
arylcarboxylates.33−36
In this report, we seek to address the underdevelopment of
soluble organic anolyte materials that are compatible with Li-
ion electrolytes. In addition, we target anolyte candidates that
satisfy the cost constraints to equivalent weight. We address
these challenges through an iterative workflow that employs the
combination of cyclic voltammetry (to rapidly evaluate anolyte
candidates) with chemical synthesis of charged anolyte
materials (to monitor chemical stability and solubility). These
studies have ultimately led to the identification of N-methyl 4-
acetylpyridinium tetrafluoroborate: a molecule that exhibits two
reversible redox couples in MeCN/LiBF4. This compound is
soluble in nonaqueous solvents at all charge states, is prepared
from commercial materials in a single synthetic step, is stable
over days at all charge states, and, at 111 g/(mol·e−), falls
significantly below the target equivalent weight of 150 g/(mol·
e−).
Figure 2. Reductive couple of 4,4′-methoxybipyridine (0.01 M) in
MeCN with TBABF4 (0.1 M) at 100 and 10 mV/s scan rates. Second
cycles plotted: ipc/ipc at 100 mV = 1.02; ipc/ipc at 10 mV = 1.04.
reduced species with high persistence and stability generally
exhibit current peak-height ratios close to one (ipc/ipa ≈ 1).
Chemical irreversibility is preliminarily identified by high ipc/ipa
ratios that increase at slower scan rates. In addition to providing
an initial estimate of reversibility, these metrics serve as a tool
to rapidly screen and compare a series of prospective anolyte
materials.
Evaluation of Pyridine Derivatives as Anolyte Materi-
als. In the course of our studies of redox active metal-
coordination complexes,37,38 we discovered that a number of
bipyridine derivatives exhibit reversible CVs. For example, the
CV of 4,4′-dimethoxybipyridine displays a redox couple at low
potential (−2.6 V versus Ag/Ag+ in MeCN with TBABF4
supporting electrolyte; Figure 2) with ipc/ipc ≈ 1. On the basis
of this preliminary result, we assessed the CVs of a series of
lower molecular weight monopyridine analogues in MeCN/
TBABF4. No reduction was observed within the electro-
chemical window of MeCN for any pyridine derivative bearing
an electron-donating group (EDG) at the 2- or 4-position (e.g.,
4-methoxypyridine, 2-methoxypyridine, 4-methylpyridine, 2-
methylpyridine; see Supporting Information for a full list of
tested compounds). This led us to hypothesize that the
electron-withdrawing pyridyl ring ortho to nitrogen in 4,4′-
methoxybipyridine is critical for the observed redox couple. As
such, we next explored pyridine derivatives bearing electron-
withdrawing substituents. The redox potentials and current
peak-height ratios at scan rates of 100 and 10 mV/s were
determined for a series of picolinate (1), nicotinate (2), and
isonicotinate (3a−h) derivatives and are summarized in Chart
1.
RESULTS AND DISCUSSION
■
Methods for Electrochemical Evaluation of Com-
pounds. Our search for low molecular weight anolyte
materials required the exploration of a significant range of
chemical space. To do so in a rapid fashion, we initially
screened a series of heteroarenes by cyclic voltammetry (CV).
Electrochemical analysis was conducted in a three-electrode cell
composed of a glassy carbon working electrode, a platinum
We first examined the effect of varying the position of an
electron-withdrawing ester substituent in molecules 1a, 2a, and
3a (Chart 1, entries 1−3). In MeCN/TBABF4, all three
compounds are reduced at standard potentials lower than −2 V
vs Ag/Ag+. No reoxidation current is observed for methyl
nicotinate 2a. The reduction of methyl picolinate 1a shows
partial chemical reversibility at 100 mV/s, but is completely
B
DOI: 10.1021/jacs.5b09572
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
to isopropyl (3c) to tert-butyl (3d) derivative (Chart 1, entries
3−6 and Figure 3). In contrast, reductions of esters bearing the
electron-withdrawing cyanoethyl, trifluoroethyl, and hexafluor-
oisopropyl substituents appear at approximately −2.0 V and are
all irreversible (entries 8−10). Replacement of the ester with a
less electron-withdrawing diethylamide substituent (3e) results
in a significant lowering of the reduction potential (to −2.5 V);
however, this redox couple is irreversible at lower scan rates
(entry 7).
Chart 1. Ratios of Current Peak Heights from CV for Alkyl
Pyridinecarboxylates
The electrochemical data for 3a−h provides insight into
possible mechanisms for decomposition of the radical anion
that is formed following reduction. Decomposition could occur
by either homolytic or heterolytic cleavage of bonds in the ester
moiety (Figure 4).39,40 Homolytic cleavage of the O−R bond
Figure 4. Two possible decomposition pathways for the reduced
anolytes (isonicotinate radical anions). The experimental data is more
consistent with pathway b.
(Figure 4, top) would generate a carboxylate and an alkyl
radical. According to this proposition, accelerated decom-
position is expected for compounds with EDGs (which stabilize
the alkyl radical) and retarded by EWGs. Conversely,
heterolytic cleavage of the (O)C−O bond releases an alkoxide
anion (Figure 4, bottom). Accelerated decomposition by this
pathway is expected for compounds with EWGs that stabilize
the alkoxide. Our experimental results show that isonicotinates
bearing electron withdrawing ester substituents (3f−h) exhibit
irreversible electrochemistry (Chart 1, entries 8−10). Fur-
thermore, the electrochemical behavior of the tert-butyl
analogue 3d is superior to that of the methyl analogue 3a
(Chart 1, entries 6 vs 3). These data are most consistent with
decomposition pathway b in Figure 4.
Evaluation of N-Protected Isonicotinates as Anolyte
Materials with LiBF4 Electrolyte. Compound 3d displays the
most promising electrochemistry of the tested compounds in
MeCN/TBABF4. Thus, we next performed CV of this
compound in MeCN/LiBF4 and propylene carbonate (PC)/
LiBF4 (Figure 3). Although the cathodic wave in Figure 3
suggests that reduction generates a similar radical anion to that
in TBABF4, no corresponding anodic wave is observed for CVs
in solutions with LiBF4.
a
b
c
0
0.01 M substrate. 0.10 M TBABF4 or LiBF4 support. E1 is
d
calculated as (E1c + E1a)/2 and referenced to Ag/Ag+. Ratios were
calculated for the second cycle following deconvolution of the
voltammogram. No anodic current was observed.
e
irreversible at the lower 10 mV/s scan rate. Methyl
isonicotinate 3a, exhibits a redox couple with significantly
improved reversibility over 1a and 2a, even at scan rates of 10
mV/s (Chart 1, entry 3).
We next explored the impact of the ester substituent on the
electrochemical behavior of isonicotinates 3a−h. The methyl,
ethyl, isopropyl, and tert-butyl esters all show reductions at
comparable potentials (approximately −2.2 V; Figure 3). A
sequential improvement in current peak-height ratio is
observed upon moving from the methyl (3a) to ethyl (3b)
A recent report has shown that the complexation of BF3 to
the nitrogen of quinoxaline dramatically increases the
reversibility of its reduction in Li-ion electrolytes.41 On the
basis of this report, we hypothesized that analogous behavior
might be observed for the isonicotinate series. To test this
possibility, the isonicotiate−BF3 adducts 4a−d were synthe-
sized and analyzed by CV, and the results are summarized in
entries 1−6 of Chart 2. The methyl and tert-butyl analogues 4a
and 4d were initially examined in MeCN/TBABF4. In this
media, the first reductive couple appears at approximately −1.8
V, with current peak-height ratios approaching unity for both
materials. In MeCN/LiBF4, the couple shifts to approximately
−1.4 V, but the current ratios remain close to 1, even at the
slow 10 mV/s scan rate (Chart 2, entries 3−6).
Figure 3. Voltammograms of selected isonicotinates. Conditions:
second cycles plotted for CVs of 0.01 M 3a−d in specified media at 10
mV/s scan rate.
C
DOI: 10.1021/jacs.5b09572
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
that the N-alkyl bond is much stronger than that of N-oxides or
BF3-adducts. In addition, literature reports have shown that a
variety of pyridinium salts undergo one-electron chemical
reduction to form stable radical intermediates.44,45 Further-
more, a CV study of related pyridinium salts in MeCN with
tetrabutylammonium salts demonstrated the accessibility of a
second, reversible reduction.46
Chart 2. Current Peak Height Ratios from CV for BF3-
Adducts and N-Oxides of Isonicotinates
N-Ethylpyridinium iodide salts 6a,b were synthesized by
reactions of 3a−d with iodoethane. CV of these materials in
MeCN/LiBF4 revealed a reversible redox couple at −1.1 V with
peak height ratios close to one (Chart 3, entries 1 and 2). Upon
Chart 3. Current Peak Height Ratios for Isonicotinates As
Determined by Cyclic Voltammetry
a
b
c
0
0.01 M substrate. 0.10 M TBABF4 or LiBF4 support. E1 is
d
calculated as (E1c + E1a)/2 and referenced to Ag/Ag+. Ratios were
calculated for the second cycle following deconvolution of the
e
voltammogram. No anodic current was measured.
One disadvantage of the BF3 adducts is their high molecular
weight. For instance, the equivalent weight of 4d as a single
electron anolyte is 247 g/(mol·e−), which significantly exceeds
the target of 150 g/(mol·e−). In an attempt to address this
issue, we scanned to more negative potentials to explore the
possibility of charging 4d with a second electron. While a
second reduction is observed at −2.3 V, reoxidation occurs with
a much lower ipa (red trace in Figure 5). Nonetheless, these
data demonstrate that N-functionalized isonicotinates exhibit
multiple redox processes at low potentials in MeCN/LiBF4.
a
b
c
0.01 M substrate. 0.10 M LiBF4 support. E10 is calculated as (E1c
+
d
0
E1a)/2. E2 is calculated as (E2c + E2a)/2 and referenced to Ag/Ag+.
e
Ratios were calculated for the second cycle following deconvolution
of the voltammogram.
scanning to more negative potentials, a second couple was
observed at −1.9 V. In contrast to the BF3−adducts,
compounds 6a−d exhibit well-shaped anodic peaks after this
second reduction with ipc/ipa ≈ 0.95 at scan rates of 10 mV/s.
The voltammograms of the iodide salts also show an
additional current response at lower potentials. Undesirable
redox processes of the iodide counterion were suspected. To
test this hypothesis, we next prepared the hexafluorophosphate
(PF6) analogues 7a−d by salt metathesis of the iodide salts with
NH4PF6 to exchange the iodide counterion. Voltammograms of
these compounds exhibit more ideal behavior compared to
those of the iodide analogues, and the identity of the
counterion does not appear to influence the peak height ratios
for the N-ethylpyridinium (Chart 3, entries 3−6). In addition,
unlike with compounds 3a−d, substitution of the carboxylate or
at nitrogen has minimal influence on the current peak height
ratios for the two redox couples (Chart 3, entries 3−6, 10 mV/
s). As a result, N-methyl isonicotinate 9a with a BF4 counterion
was prepared. This molecule has the lowest equivalent weight
(127 g/(mol·e−)) of the isonicotinate salts studied while
maintaining analogous electrochemical properties to those of
6a, 7a, and 8a (Chart 3, entry 8).47
Figure 5. (a) CVs of 4d through one (black trace) and two (red trace)
redox couples with LiBF4 supporting electrolyte. Conditions: 100 mV/
s scan rate.
We next examined the substitution of BF3 with lower
molecular weight substituents on nitrogen, starting with N-
oxide derivatives. In MeCN/TBABF4, the alkyl isonicotinate-N-
oxides 5a−d undergo a single reduction at approximately −2 V,
with current peak-height ratios close to one (Chart 2, entries
7−8).42 However, this reduction is irreversible in MeCN/LiBF4
(Chart 2, entry 9).
We noted that the weak N−O bond43 of the N-oxide is
susceptible to many side reactions. As a result, we next targeted
N-alkylpyridinium salts. These molecules offer the advantage
Synthesis, Solubility, and Stability of Anolyte 9a and
Its Reduced Analogue 9a(−). The isonicotinate salt 9a was
prepared in a single step from the reaction of methyl
D
DOI: 10.1021/jacs.5b09572
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
isonicotinate with trimethyloxonium tetrafluoroborate in
dichloromethane. This air- and moisture-stable solid precip-
itates over the course of the reaction, and is isolated on a gram
scale via simple filtration.
The CV of 9a was measured in both MeCN/LiBF4 and PC/
LiBF4 (Figure 6, i and ii, respectively). Although two reductive
the CV of this brown solution shows no redox couples
characteristic of 9a. Instead, a single, irreversible oxidation at
+0.5 V is observed (Figure 6, iv). These data demonstrate that
the charged anolyte Na-9a(−) is unstable in MeCN/LiBF4.
We next conducted studies to gain insight into the
decomposition pathways of Na-9a(−) in MeCN/LiBF4.
Preliminary analysis of the material responsible for the anodic
1
current at +0.5 V was performed using H NMR spectroscopy
and CV. While multiple decomposition products are observed
by H NMR spectroscopy, H−1H NMR COSY experiments
show resonances that are consistent with products of
protonation alpha to the carbonyl. The decomposition of Na-
9a(−) via protonation was further explored by conducting CV
with water intentionally added to the electrolyte solution.
These studies revealed that CVs of Na-9a(−) in MeCN/LiBF4
with 22 mM of added water are identical to those shown in
Figure 6, iv. Notably, this decomposition process persists even
upon our most rigorous attempts to dry the electrolyte
solution.48 These data suggest that the basicity of Na-9a(−)
(which is an ester enolate) is sufficiently high to react with trace
amounts of water and/or the MeCN solvent (pKa49 ∼ 31).50
Synthesis, Solubility, and Stability of Anolyte 10a and
Its Reduced Analogues 10a(•) and 10a(−). To minimize this
proposed decomposition pathway, we next targeted anolyte
materials that are less basic in their fully reduced form. It is well
know that ketone enolates are approximately 3 orders of
magnitude less basic than the analogous ester enolates. For
example, the pKa’s of acetone and ethyl acetate are 26.5 and
29.5, respectively.51 As such, we hypothesized that the reduced
species derived from ketone analogues of 9a would exhibit
increased stability under the electrochemical conditions.
1
1
Figure 6. CVs of 9a in MeCN (i) or PC (ii). CV of 9a(−) immediately
after dissolution in MeCN (iii) and 2 min later (iv). Conditions: 0.01
M 9a or Na-9a(−) with 0.1 M LiBF4 in the specified solvent at 100
mV/s scan rate. First cycles are plotted. Gray dashed lines indicate the
axis of origin. Arrows indicate start and direction of CV scan.
waves are observed in both solvents, the chemical reversibility
of the second reduction is lower in PC than in MeCN.
Furthermore, the CV in PC shows the presence of byproducts
that undergo oxidation at positive potentials (Figure 6, ii).
These initial results warranted further studies of the solubility
and stability of 9a in MeCN.
To test this hypothesis, the acetylpyridinium salt 10a was
prepared in 93% yield via methylation of 4-acetylpyridine with
trimethyloxonium tetrafluoroborate. The CV of 10a in MeCN/
LiBF4 shows two reductive couples with peak height ratios
close to 1 (Figure 7, i). A direct comparison of 9a and 10a
As discussed above, the CV experiments were used as an
initial screen for the accessibility and reversibility of reductive
couples for anolyte candidates. However, for a redox flow
battery application, it is crucial to have more detailed
information about the long-term stability of the reduced
species. As such, our next efforts focused on chemically
generating 9a(−), the material formed upon the 2e− reduction of
9a. The addition of sodium metal to a white suspension of 9a in
THF resulted in the formation of a red solution within 30 min
at room temperature. After filtration, the red solution was
concentrated to yield a dark orange solid. Analysis by 1H NMR
spectroscopy and elemental analysis confirmed the formation of
the diamagnetic sodium salt Na-9a(−) (Scheme 1).
The CV of a freshly prepared solution of Na-9a(−) in
MeCN/LiBF4 shows peaks at −1.1 and −1.8 V, analogous to
those observed for 9a (Figure 6, iii). However, when the red
solution of Na-9a(−) in MeCN/LiBF4 is allowed to stand for 2
min, it changes color from red to brown, suggesting that
decomposition is occurring. Consistent with this hypothesis,
Figure 7. CVs of 10a in MeCN (i) and PC (ii). CVs of 10a(•) (iii) and
10a(−) (iv) in MeCN. Conditions: 0.01 M 10a, 10a(•), or 10a(−) with
0.1 M LiBF4 in the specified solvent at 100 mV/s scan rate. First cycles
are plotted. Gray dashed lines indicate the axis of origin. Arrows
indicate start and direction of CV scan.
Scheme 1. Chemical Reduction of 9a, and Decomposition
Pathways of 9a(−)
reveals that the first reductions occur at similar potentials (0.1
V difference) but that the second reduction of 10a is 0.4 V
more positive than that of 9a. Notably, the CV of 10a in PC/
LiBF4 shows reversible couples at comparable potentials to
those in MeCN/LiBF4 (Figure 7, ii). In contrast, as discussed
above, the CV of 9a in PC/LiBF4 is indicative of significant
decomposition (Figure 6, ii).
E
DOI: 10.1021/jacs.5b09572
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
We next independently synthesized the materials generated
during the reduction of 10a. Compound Na-10a(−) was
prepared via the 2e− reduction of 10a with sodium metal in
THF. The product was isolated in 88% yield as an orange solid,
and was characterized by elemental analysis, NMR spectrosco-
py, UV−vis spectroscopy, and CV. The 1H NMR resonances of
Na-10a(−) are shifted upfield relative to those of 10a, consistent
with a disruption of aromaticity (Figure 8, panel c vs b).
Figure 9. (a) Synthesis of 10a(•) by comproportionation of 10a and
Na-10a(−). (b) EPR spectrum of a 1.0 mM solution of 10a(•) in MeCN
at 130 K.
1
resonances are observed in the H NMR spectrum of 10a(•)
.
The EPR spectrum of a frozen solution of 10a(•) in MeCN
shows a strong isotropic resonance (Figure 9b). This signal is
centered at 334 mT, and corresponds to a g-factor of 1.996.
This value is consistent with a carbon centered radical, which
generally has a g-factor near the value of a free electron
(2.0036).55
Figure 8. (a) Synthesis of Na-10a(−) by chemical reduction of 10a. (b)
1H NMR spectrum of 10a and labeled resonances. (c) 1H NMR
spectrum of Na-10a(−) and labeled resonances.
A comparison of the CV of 10a(•) to that of 10a and Na-
10a(−) shows two couples at identical potentials (Figure 7,
iii).56 Similar to Na-10a(−), a non-zero current is observed at
the initial scanning potential of +1 V. This current (23 μA) is
half of the initial current (45 μA) measured for Na-10a(−) at the
same potential (compare baselines of CVs in Figure 7 iii vs iv).
At the initial potential of +1 V, the reduced species Na-10a(−)
undergoes two, 1e− oxidations, while 10a(•) undergoes only
one, 1e− oxidation. As such, these data are consistent with the
different redox states of the molecule.
Furthermore, all four protons on the ring of Na-10a(−) are
inequivalent. This is consistent with an enolate structure, in
which the carbonyl is not freely rotating. The CV of Na-10a(−)
in MeCN/LiBF4 shows two couples at identical potentials to
10a (Figure 7, iv). In addition, CVs measured at starting
potentials of +1 V show baseline currents of 40−50 μA. This
non-zero initial current is consistent with the presence of a
prereduced material (Na-10a(−)) that undergoes oxidation at
high potentials. Importantly, no analogous current is observed
at +1 V when the material is in its discharged state 10a
(compare baseline currents relative to 0 μA lines in Figure 7, i
vs iv). Unlike with Na-9a(−), the CV of Na-10a(−) does not
change after standing in MeCN/LiBF4 for 20 min at room
temperature. This indicates that Na-10a(−) is stable under these
conditions.
With 10a, 10a(•), and Na-10a(−) in hand, we next
quantitatively assessed solubility and stability at each oxidation
state. The solubilities of 10a, 10a(•), and Na-10a(−) were
determined by UV−vis spectroscopy in MeCN under an inert
atmosphere. A solubility of 1.6
0.1 M in MeCN was
measured for the parent compound 10a by analyzing the UV
absorbance at 281 nm.57 This corresponds to 0.94 kg of a one-
electron anolyte material per 1 kg of solvent, which exceeds the
target of 0.8 kg anolyte per 1 kg solvent required to meet an
RFB system price target of $120/kW·h.11 In addition to these
solubility parameters, the low equivalent weight of 111 g/(mol·
e−) for 10a is well below the target limit of 150 g/(mol·e−).
A saturated solution of 10a(•) was prepared and analyzed at
322 nm by UV−vis spectroscopy. The saturation concentration
of 10a(•) was 5.4 0.5 M. An analogous study with Na-10a(−)
(monitoring the absorbance at 426 nm) revealed a solubility of
62 7 mM. We believe that the low solubility of Na-10a(−) is
partially a reflection of the Na+ counterion, which is not the
charge-balancing counterion during electrochemical reduction
in solutions with a LiBF4 support. Consistent with this
proposal, a moderate increase in the solubility of Na-10a(−)
(to 80 6 mM) was observed using a 0.1 M solution of LiBF4
in MeCN as the solvent. Nonetheless, the solubility of the
We also sought to prepare 10a(•), which is the intermediate
redox-state between the fully charged and discharged states of
10a. In the literature, related N-alkylpyridine radicals have been
generated electrochemically52 or via chemical reduction with
magnesium or zinc.53,54 However, these methods generally
afford the radical products in low yields. As such, we pursued an
alternative strategy to prepare 10a(•), involving comproportio-
nation of 10a and Na-10a(−) (Figure 9a). The dissolution of
equimolar quantities of 10a (a white solid) and Na-10a(−) (an
orange solid) in MeCN resulted in the instantaneous formation
of a dark green solution. Dilution of this solution with diethyl
ether led to the precipitation of NaBF4. The solution was then
collected, and the solvent was removed to afford 10a(•) as a
dark green solid in 91% yield.
The neutral radical 10a(•) was characterized by elemental
1
analysis, H NMR spectroscopy, EPR spectroscopy, UV−vis
spectroscopy, and CV. As expected for a free radical, no
F
DOI: 10.1021/jacs.5b09572
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
doubly reduced anolyte will clearly need to be enhanced prior
to battery cycling studies at high concentration. Both literature
precedent21,37,58,59 and our preliminary results suggest that this
will be feasible by appending solubilizing substituents on the
scaffold of 10a.60
salts. The latter materials are stable in their singly and doubly
reduced states for days without measurable decomposition.
Advanced electrochemical analysis and implementation of these
materials in battery devices is currently ongoing and will be
reported in due course.
The stabilities of 10a, 10a(•), and Na-10a(−) as solutions in
MeCN were investigated by 1H NMR spectroscopy or UV−vis
spectroscopy. Separate solutions of 10a, 10a(•), and Na-10a(−)
in MeCN were prepared at 480, 540, and 39 mM
concentrations, respectively. The parent compound 10a was
monitored by 1H NMR spectroscopy against an internal
standard of 1,3,5-trimethoxybenzene. No measurable decom-
position was detected over 4 days even in the presence of air
and moisture. Solutions of the reduced species 10a(•) and Na-
10a(−) were stored at room temperature under nitrogen, and
the concentrations of the solutions were measured periodically
over 36 or 48 h by UV−vis spectroscopy (Figure 10). During
ASSOCIATED CONTENT
* Supporting Information
The Supporting Information is available free of charge on the
ACS Publications website at DOI: 10.1021/jacs.5b09572.
■
S
Experimental procedures and characterization of all new
compounds including spectroscopic data and potentio-
metric data (PDF)
AUTHOR INFORMATION
Corresponding Author
*mssanfor@umich.edu
■
Author Contributions
#R.E.M.B. and E.C. contributed equally.
Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS
■
Research described in this paper was supported by the Joint
Center for Energy Storage Research (JCESR), a Department of
Energy, Energy Innovation Hub.
REFERENCES
Figure 10. Concentrations of 10a(•) (540 mM initial concentration,
blue) and Na-10a(−) (39 mM initial concentration, red) in MeCN
measured by UV−vis over 36−48 h.
■
(1) Yang, Z.; Zhang, J.; Kintner-Meyer, M. C. W.; Lu, X.; Choi, D.;
Lemmon, J. P.; Liu, J. Chem. Rev. 2011, 111, 3577.
(2) Dunn, B.; Kamath, H.; Tarascon, J.-M. Science 2011, 334, 928.
(3) Lin, Y.; Li, Y.; Zhan, X. Chem. Soc. Rev. 2012, 41, 4245.
(4) Huang, X.; Qi, X.; Boey, F.; Zhang, H. Chem. Soc. Rev. 2012, 41,
666.
(5) Li, G.; Zhu, R.; Yang, Y. Nat. Photonics 2012, 6, 153.
(6) Cappillino, P. J.; Pratt, H. D.; Hudak, N. S.; Tomson, N. C.;
Anderson, T. M.; Anstey, M. R. Adv. Energy Mater. 2014, 4, 1300566.
(7) Wei, X. L.; Cosimbescu, L.; Xu, W.; Hu, J. Z.; Vijayakumar, M.;
Feng, J.; Hu, M. Y.; Deng, X. C.; Xiao, J.; Liu, J.; Sprenkle, V.; Wang,
W. Adv. Energy Mater. 2015, 5, 1400678.
this period, the concentrations of the reduced anolytes did not
change, indicating that these materials are also stable and
amenable to storage for days in solution. Furthermore, CVs of
solutions of 10a(•) and Na-10a(−) that were stored for 2 months
in MeCN exhibited identical traces to those of the freshly
prepared solutions. Importantly, these materials are also stable
in mixed redox states. For example, CVs of mixtures of 10a and
10a(•) as well as of 10a(•) and Na-10a(−) showed no detectable
impurities after aging at room temperature for 16 h in MeCN/
LiBF4.
(8) Ponce de Leon
D. A.; Walsh, F. C. J. Power Sources 2006, 160, 716.
́ ́ ́
, C.; Frías-Ferrer, A.; Gonzalez-García, J.; Szanto,
(9) Shin, S.-H.; Yun, S.-H.; Moon, S.-H. RSC Adv. 2013, 3, 9095.
(10) Wang, W.; Luo, Q.; Li, B.; Wei, X.; Li, L.; Yang, Z. Adv. Funct.
Mater. 2013, 23, 970.
(11) Darling, R. M.; Gallagher, K. G.; Kowalski, J. A.; Ha, S.;
Brushett, F. R. Energy Environ. Sci. 2014, 7, 3459.
CONCLUSIONS
■
In conclusion, this report describes an iterative workflow that
enabled the design of anolyte materials with tailored electro-
chemical properties. Specifically, we report an all-organic
anolyte that undergoes two reversible reductions in solutions
with Li-ion electrolytes. An initial systematic evaluation of
anolyte candidates by CV led us from materials based on free
pyridines to BF3−pyridine adducts, pyridine N-oxides, N-ethyl
isonicotinate salts of iodide, PF6, and BF4, and ultimately N-
methyl acetylpyridinium BF4 salts. These latter materials
address the challenge of identifying soluble anolytes that
operate in Li-ion supports and that have equivalent weights
below 150 g/(mol·e−).11
(12) Gomberg, M. Chem. Rev. 1924, 1, 91.
(13) Song, Z.; Zhou, H. Energy Environ. Sci. 2013, 6, 2280.
(14) Huskinson, B.; Marshak, M. P.; Suh, C.; Er, S.; Gerhardt, M. R.;
Galvin, C. J.; Chen, X.; Aspuru-Guzik, A.; Gordon, R. G.; Aziz, M. J.
Nature 2014, 505, 195.
(15) Wain, A. J.; Wildgoose, G. G.; Heald, C. G. R.; Jiang, L.; Jones,
T. G. J.; Compton, R. G. J. Phys. Chem. B 2005, 109, 3971.
(16) Li, Q.; Batchelor-McAuley, C.; Lawrence, N. S.; Hartshorne, R.
S.; Compton, R. G. Chem. Commun. 2011, 47, 11426.
(17) Fujinaga, T.; Izutsu, K.; Nomura, T. J. Electroanal. Chem.
Interfacial Electrochem. 1971, 29, 203.
(18) Iordache, A.; Maurel, V.; Mouesca, J.-M.; Pec
Gutel, T. J. Power Sources 2014, 267, 553.
(19) Feng, J. K.; Cao, Y. L.; Ai, X. P.; Yang, H. X. J. Power Sources
2008, 177, 199.
(20) Nesvadba, P.; Folger, L. B.; Maire, P.; Novak
́
aut, J.; Dubois, L.;
Over the course of these studies, we developed synthetic
procedures to access the materials in their singly and doubly
reduced states. Studies on these materials allowed us to identify
decomposition pathways of isonicotinate salts. These decom-
position pathways were resolved by preparing acetylpyridinium
́
, P. Synth. Met.
2011, 161, 259.
G
DOI: 10.1021/jacs.5b09572
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX

Journal of the American Chemical Society
Article
(21) Huang, J.; Cheng, L.; Assary, R. S.; Wang, P.; Xue, Z.; Burrell, A.
K.; Curtiss, L. A.; Zhang, L. Adv. Energy Mater. 2015, 5, 1401782.
(22) Huang, J. H.; Shkrob, I. A.; Wang, P. Q.; Cheng, L.; Pan, B. F.;
He, M. N.; Liao, C.; Zhang, Z. C.; Curtiss, L. A.; Zhang, L. J. Mater.
Chem. A 2015, 3, 7332.
(55) Bunce, N. J. J. Chem. Educ. 1987, 64, 907.
(56) Rate constants of 6.0 × 10−3 and 4.7 × 10−3 cm/s were
measured for the reductions of 10a to 10a(•) and 10a(•) to 10a(−) at
the electrode, respectively.
(57) Rheometric measurements on 0.5 and 1.0 M solutions of 10a in
MeCN reveal viscosities of 0.42 and 0.72 cP, respectively.
(58) Cabrera, P. J.; Yang, X.; Suttil, J. A.; Hawthorne, K. L.; Brooner,
R. E. M.; Sanford, M. S.; Thompson, L. T. J. Phys. Chem. C 2015, 119,
15882.
(59) Huang, J.; Su, L.; Kowalski, J. A.; Barton, J. L.; Ferrandon, M.;
Burrell, A. K.; Brushett, F. R.; Zhang, L. J. Mater. Chem. A 2015, 3,
14971.
(23) Henderson, J. C.; Kiya, Y.; Hutchison, G. R.; Abruna, H. D. J.
̃
Phys. Chem. C 2008, 112, 3989.
(24) Burkhardt, S. E.; Conte, S.; Rodriguez-Calero, G. G.; Lowe, M.
A.; Qian, H.; Zhou, W.; Gao, J.; Hennig, R. G.; Abruna, H. D. J. Mater.
Chem. 2011, 21, 9553.
(25) Hernan
́
dez-Burgos, K.; Rodríguez-Calero, G. G.; Zhou, W.;
Burkhardt, S. E.; Abruna, H. D. J. Am. Chem. Soc. 2013, 135, 14532.
(26) Gao, J.; Lowe, M. A.; Conte, S.; Burkhardt, S. E.; Abruna, H. D.
Chem. - Eur. J. 2012, 18, 8521.
̃
(60) For example, our preliminary studies show that changing the
methyl ketone to a phenyl ketone leads to a >25-fold increase in the
solubility of the doubly reduced anolyte.
̃
(27) Kiya, Y.; Henderson, J. C.; Abruna, H. D. J. Electrochem. Soc.
̃
2007, 154, A844.
(28) Nakahara, K.; Oyaizu, K.; Nishide, H. Chem. Lett. 2011, 40, 222.
(29) Janoschka, T.; Hager, M. D.; Schubert, U. S. Adv. Mater. 2012,
24, 6397.
(30) Wei, X.; Cosimbescu, L.; Xu, W.; Hu, J. Z.; Vijayakumar, M.;
Feng, J.; Hu, M. Y.; Deng, X.; Xiao, J.; Liu, J.; Sprenkle, V.; Wang, W.
Adv. Energy Mater. 2015, 5, 1400678.
(31) Brushett, F. R.; Vaughey, J. T.; Jansen, A. N. Adv. Energy Mater.
2012, 2, 1390.
(32) Zhao, Y.; Ding, Y.; Song, J.; Li, G.; Dong, G.; Goodenough, J.
B.; Yu, G. Angew. Chem., Int. Ed. 2014, 53, 11036.
(33) Burkhardt, S. E.; Bois, J.; Tarascon, J.-M.; Hennig, R. G.;
Abruna, H. D. Chem. Mater. 2013, 25, 132.
̃
(34) Hernan
́
dez-Burgos, K.; Burkhardt, S. E.; Rodríguez-Calero, G.
G.; Hennig, R. G.; Abruna, H. D. J. Phys. Chem. C 2014, 118, 6046.
̃
(35) Walker, W.; Grugeon, S.; Vezin, H.; Laruelle, S.; Armand, M.;
Wudl, F.; Tarascon, J.-M. J. Mater. Chem. 2011, 21, 1615.
(36) Armand, M.; Grugeon, S.; Vezin, H.; Laruelle, S.; Ribiere, P.;
Poizot, P.; Tarascon, J. M. Nat. Mater. 2009, 8, 120.
(37) Suttil, J. A.; Kucharyson, J. F.; Escalante-Garcia, I. L.; Cabrera, P.
J.; James, B. R.; Savinell, R. F.; Sanford, M. S.; Thompson, L. T. J.
Mater. Chem. A 2015, 3, 7929.
(38) Cabrera, P. J.; Yang, X.; Suttil, J. A.; Hawthorne, K. L.; Brooner,
R. E. M.; Sanford, M. S.; Thompson, L. T. J. Phys. Chem. C 2015, 119,
15882.
(39) Webster, R. D.; Bond, A. M.; Schmidt, T. J. Chem. Soc., Perkin
Trans. 2 1995, 1365.
(40) Webster, R. D.; Bond, A. M. J. Org. Chem. 1997, 62, 1779.
(41) Carino, E. V.; Diesendruck, C. E.; Moore, J. S.; Curtiss, L. A.;
Assary, R. S.; Brushett, F. R. RSC Adv. 2015, 5, 18822.
(42) Miyazaki, H.; Matsuhisa, Y.; Kubota, T. Bull. Chem. Soc. Jpn.
1981, 54, 3850.
(43) Acree, W. E.; Pilcher, G.; Ribeiro da Silva, M. D. M. C. J. Phys.
Chem. Ref. Data 2005, 34, 553.
(44) Kosower, E. M.; Cotter, J. L. J. Am. Chem. Soc. 1964, 86, 5524.
(45) Kosower, E. M.; Poziomek, E. J. J. Am. Chem. Soc. 1964, 86,
5515.
(46) Leventis, N.; Rawaswdeh, A.-M. M.; Zhang, G.; Elder, I. A.;
Sotiriou-Leventis, C. J. Org. Chem. 2002, 67, 7501.
(47) Solubility measurements by UV−vis indicate that saturated
solutions of 9a in MeCN have concentrations of 2.5 0.1 M, which
corresponds to 1.5 kg of a one-electron anolyte material per 1 kg of
solvent.
(48) Williams, D. B. G.; Lawton, M. J. Org. Chem. 2010, 75, 8351.
(49) Zhang, X. M.; Bordwell, F. G.; Van Der Puy, M.; Fried, H. E. J.
Org. Chem. 1993, 58, 3060.
(50) This high reactivity of Na-9a(−) is also consistent with the
irreversible CVs in PC.
(51) Bordwell, F. G. Acc. Chem. Res. 1988, 21, 456.
(52) Schwarz, W. M.; Kosower, E. M.; Shain, I. J. Am. Chem. Soc.
1961, 83, 3164.
(53) Kosower, E. M.; Schwager, I. J. Am. Chem. Soc. 1964, 86, 5528.
(54) Kosower, E. M.; Poziomek, E. J. J. Am. Chem. Soc. 1963, 85,
2035.
H
DOI: 10.1021/jacs.5b09572
J. Am. Chem. Soc. XXXX, XXX, XXX−XXX