Energy &

Environmental
Science
View Article Online
PAPER View Journal

Aqueous solutions of super reduced

This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

polyoxotungstates as electron storage systems†

Cite this: DOI: 10.1039/d3ee00569k
Open Access Article. Published on 21 April 2023. Downloaded on 5/12/2023 7:48:47 AM.

Tingting Zhao, Nicola L. Bell, Greig Chisholm, Balamurugan Kandasamy,

De-Liang Long and Leroy Cronin *

Due to the increasing energy density demands of battery technology, it is vital to develop electrolytes
with high electron storage capacity. Polyoxometalate (POM) clusters can act as electron sponges,
storing and releasing multiple electrons and have potential as electron storage electrolytes for flow
batteries. Despite this rational design of clusters for high storage ability can not yet be achieved as little
is known about the features influencing storage ability. Here we report that the large POM clusters,
{P5W30} and {P8W48}, can store up to 23 e and 28 e per cluster in acidic aqueous solution,
respectively. Our investigations reveal key structural and speciation factors influencing the improved
behaviour of these POMs over those previously reported (P2W18). We show, using NMR and MS, that for
these polyoxotungstates hydrolysis equilibria for the different tungstate salts is key to explaining
unexpected storage trends while the performance limit for {P5W30} and {P8W48}, can be attributed to
Received 22nd February 2023, unavoidable hydrogen generation, evidenced by GC. NMR spectroscopy, in combination with the MS
Accepted 20th April 2023 analysis, provided experimental evidence for a cation/proton exchange process during the reduction/
DOI: 10.1039/d3ee00569k reoxidation process of {P5W30} which likely occurs due to this hydrogen generation. Our study offers a
deeper understanding of the factors affecting the electron storage ability of POMs and provides insights
rsc.li/ees allowing for further development of these materials for energy storage.

Broader context
The storage of electrons by liquid electrolytes is the critical component of a flow battery. The material needs to be stable, safe, and to store as much charge as
possible without loss. The benefits of flow batteries include the ability to recharge them by removing the liquid electrolyte and the potential for coupling them
into chemical processes. Traditional flow battery electrolytes have been either simple salts or organic molecules, but these can only store 1–2 electrons per
molecule and either suffer from stability issues or are corrosive. Recently it was postulated that high nuclearity polyoxometalate clusters could be used to store a
large number of electrons per molecule in aqueous solution. In this study we show that large POM clusters, {P5W30} and {P8W48}, can store more than
20 electrons per cluster and we explore the mechanism of electron uptake as well as characterize the solutions as potential electrolytes suitable for use in
flow batteries.

Introduction optimising RFB performance: (1) potential difference between

the materials oxidised and reduced and (2) storage capacity.3,4
The transition to a green economy necessitates the develop- Significant work has been undertaken to address the former
ment of novel energy storage solutions and there is currently a using different redox pairs.5 Maximising capacity, or the number
great deal of interest in redox flow batteries (RFBs) for this of electrons able to be stored and released by a material is still an
application.1 Vanadium metal oxides have dominated the field important area of development.6 Polyoxometalates (POMs), are a
however their commercial development has been limited class of early transition metal oxide clusters with MOx (x = 5, 6) as
by their low energy density.2 Two key features are salient when basic construction units, which can both store and release
multiple electrons due to the reduction and oxidation of high-
valence transition metals, like W, Mo and V.7–10 Polyoxometa-
School of Chemistry, University of Glasgow, University Avenue, Glasgow, G12 8QQ,
lates are highly soluble yielding tungsten concentrations of up to
UK. E-mail: lee.cronin@glasgow.ac.uk
† Electronic supplementary information (ESI) available. CCDC 2173983. For ESI
8 M for PW12. Hence, investigating how POMs store and release
and crystallographic data in CIF or other electronic format see DOI: https://doi. electrons, and how this relates to their structure, will allow for
org/10.1039/d3ee00569k the further exploitation of their unique properties.

This journal is © The Royal Society of Chemistry 2023 Energy Environ. Sci.
 View Article Online

Paper Energy & Environmental Science

The exceptional electron storage ability of POMs was first of Li-ion battery charging and discharging processes.25,26 More
hinted at when Launay reported the reduction of [H2W12O40]6 , recently, NMR was used to monitor POMs cation–anion inter-
suggesting the reduction of the cluster by six electrons was also action in solution. Nyman and co-workers investigated ion-
accompanied by the transfer of six protons.11,12 Bond et al. also pairing between Cs+ cations and Nb/Ta POM anions using
reported the reduction of the Wells–Dawson [S2Mo18O62]4 inversion-recovery 133Cs-NMR spectroscopy demonstrating that
anion in a mixture of acetonitrile and water solution, indicating strong covalent orbital overlap occurs between the alkali metal
eight reversible one-electron reductions of this anion.13 A leap and the metal oxide ligands.28,29 Bloor and Kidd studied the 39K
to 24 electron reduction was achieved by Yoshikawa and Awaga, NMR chemical shifts in aqueous solutions with different POM
who investigated a Keggin-type POM, [PMo12O40]3 as a cathode anions and concluded that orbital overlap with metal oxides
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

active material.14 Through X-ray absorption near-edge structure can decrease paramagnetic shielding of the alkali metal ion
technique analysis they found all 12 Mo6+ centres in this Keggin relative to solvated ions.30 However, there is no NMR study to
Open Access Article. Published on 21 April 2023. Downloaded on 5/12/2023 7:48:47 AM.

structure were reduced to Mo4+ in the discharging process, date of the interactions of Li+ counter cations with polyoxoa-
which means this Keggin can store 24 e per cluster. But this nions, nor investigations of their behaviour upon reduction of
performance was achieved in the solid state and not aqueous the metal centre.
solution. Since many energy storage devices use aqueous solu- Here, we show that both polyoxoanions {P5W30} and {P8W48}
tions as electrolytes, understanding the electron storage ability demonstrate notable proton-coupled electron redox processes,
of POMs in aqueous solution is important.1,15 Normally in which allow the two clusters to reversibly accept up to 23 and
aqueous solution, the electron storage ability is limited by side 28 e per cluster respectively in acidic aqueous solution. The
reactions, like hydrogen evolution, as the reduction of these electron storage ability is concentration based and dependent
polyoxoanions occurs at only slightly less cathodic potentials upon the type of counter cations. The excellent electrochemical
than hydrogen evolution as more electrons are put into system. performance is attributed to both aggregation of reduced
This means it becomes harder to further reduce the POMs and polyoxoanions and their protonation in solution, which was
side reactions become an even larger problem. studied by 7Li and 1H NMR. We also found there is perfor-
In previous work we demonstrated that the POM cluster, mance limit regarding the total W atoms in the structure, for
Li6[P2W18O62], has exceptional proton-coupled electron redox both {P5W30} and {P8W48}, and the possible reasons were
activity and can reversibly accept up to 18 protons and electrons analysed including cluster decomposition, competitive hydro-
in aqueous solution at high concentration and low pH.16,17 It is gen generation and the effect of different metal coordination
therefore of interest to examine whether this concentration and pH environments within the POMs. The ability for these clusters to
effect applies to other POMs offering the opportunity to increase readily be so highly reduced has importance for the develop-
the storage capacity by utilizing larger clusters. Also, inspired by ment of new energy storage materials as well as new strategies
this initial finding, it is important to investigate if electron storage to develop multi-electron storage systems.
potential can scale with cluster nuclearity and what properties
of a cluster promote electron storage. Therefore, we selected two
other phosphotungstates, [NaP5W30O110]14 (abbreviated {P5W30})
and [H7P8W48O184]33 (abbreviated {P8W48}). {P5W30}, known as the Results and discussion
Preyssler anion, has approximate D5h symmetry and consists of a
cyclic assembly of five {PW6O22} units. {P5W30} is stable within the Synthesis and cyclic voltammetry
pH window from 0 to 11 and is reducible to a blue-coloured K-{P5W30} (K14[NaP5W30O110]
22H2O) was synthesized according
species, which make it a promising candidate for multi-electron to a modified literature procedure.19,31 Li-{P5W30} (Li14[NaP5-
storage.18–20 {P8W48} is a wheel-shaped polyoxotungstate which is W30O110]
38H2O) was synthesized by ion exchange of K-{P5W30}
stable in pH range 1 to 8; it is also a potential candidate for multi- in excess LiNO3 and characterized by single crystal XRD structure
electron storage.21 Previous studies show that the electrochemistry determination. Refinement of the diffraction data for Li-{P5W30}
of POMs is dependent on the molecular charge, charge density,22 revealed the structure of polyoxoanion of [NaP5W30O110]14 , with
cation type, and cation/anion size.23,24 Therefore, as {P5W30} and comparable unit cell parameters to those of K-{P5W30} (Fig. 1a and
{P8W48} have bigger size and higher charge density than {P2W18}, Table S4, ESI†).18 The redox chemistry of {P5W30} in sulfuric acid
the effect of these factors on the electron storage ability suggest that solution was studied by cyclic voltammetry (CV) experiments
they should be great candidates to test our hypothesis to assess (Fig. 1b). At low concentration (2 mM), K-{P5W30} shows four
how many electrons can be stored in a single cluster molecule. reversible waves within the 0.6 V to 0.6 V range (versus standard
To explore the mechanism of reduction to form highly- hydrogen electrode, SHE). Upon moving to higher concentrations
reduced POMs, previous work suggests it is important to of the POM (10 mM), we found there is significant enhancement
consider the aggregation of polyoxoanions in solution. This in the current intensity of the redox waves for K-{P5W30} when
can be done by small-angle X-ray scattering (SAXS) e.g., for compared with the control study at a concentration of 2 mM
[PW12O40]3 and [P2W18O62]6 .14,17 Furthermore, we reasoned (Fig. S15, ESI†) in line with our previous studies.16 The CV curves
that NMR should also be a powerful technique to study the for K-{P5W30} and Li-{P5W30} show similar reduction and oxida-
structure in solution as a function of different redox state.25–27 tion waves at 10 mM concentration suggesting that cations may
For example, 7Li NMR has been used to analyse the mechanism have minimal interaction with the POM in acidic solution.

Energy Environ. Sci. This journal is © The Royal Society of Chemistry 2023
 View Article Online

Energy & Environmental Science Paper

electrochemically in a neighbour cell on the right-hand side

sharing the same POM solution. The electrons stored reversibly
in the POM solution could be deduced by comparing the total
charge (electrons) input to reduce the POM with the charge
(electrons) released out when it was re-oxidized.

{P5W30} electron storage capacity studies

The electron storage ability of K-{P5W30} was studied versus
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

concentration (Fig. 2a and Fig. S16 and Table S5, ESI†). When
increasing the concentration from 2 mM to 5 mM in water (the
Open Access Article. Published on 21 April 2023. Downloaded on 5/12/2023 7:48:47 AM.

pH decreased from 4.2 to 3.7 accordingly), the electron storage

ability of K-{P5W30} increased from 15.7 e to 20.7 e . The trend
at lower concentration is similar to that observed for the
previously reported Li-{P2W18},16 however the trend breaks
Fig. 1 (a) Structure of Li-{P5W30} by single crystal XRD (grey polyhedra: W; down above 5 mM as the 10 mM solution (20.7 e ) did not
red ball: O; orange polyhedra: P; cyan ball: Na)18 (b) cyclic voltammograms show an increase in performance over the 5 mM one (20.7 e ).
of K-{P5W30} (10 mM) and Li-{P5W30} (10 mM) in 1 M H2SO4,18 (c) structure Considering the effect of pH,33 we dissolved the cluster in 0.2 M
of KLi-{P8W48}21 by single crystal XRD (grey polyhedra: W, red ball: O; H2SO4 which led to an increase in the solutions storage capacity
orange polyhedra: P), (d) cyclic voltammograms of KLi-{P8W48} (10 mM)
and LiNH4-{P8W48} (10 mM) in 1 M H2SO4.32 Scan rate: 10 mV s 1. (See ESI†
Section 4.1 for further details)

Similarly, mixed cation salts K28Li5H7[P8W48O184]
92H2O (KLi-

{P8W48}) and Li17(NH4)21H2[P8W48O184]
85H2O (LiNH4-{P8W48})
were synthesized according to modified literature procedures
and the structure of KLi-{P8W48} is shown in Fig. 1c.21 CV curves
of KLi-{P8W48} and LiNH4-{P8W48} (Fig. 1d) show different wave
shapes, indicating potential different redox mechanisms, with the
LiNH4-{P8W48} salt displaying similar reduction behaviour down
to 0.4 V but slightly different oxidation behaviour cf. the KLi salt,
suggesting different cation incorporation, modulating behaviour,
upon reduction including hydrogen evolution which will be
further detailed below.32 However, many of the differences, inclu-
ding the large oxidation peak at ca 0.15 V likely arise from the
presence of [NH4+] ions.

Flow-cell testing
Previous studies showed the storage of multiple electrons in
Li-{P2W18} was proton-coupled, and as the concentration of
Li-{P2W18} increased, an increasing number of electrons could be
stored. To explore if this trend is followed here we explored the
cluster redox performance in a three-electrode electrochemical
flow cell with an Hg/HgSO4 as a reference electrode.16 This device Fig. 2 Reversible multi-electron redox chemistry of [NaP5W30O110]14 . (a)
can quantify the number of electrons that the polyoxometalate Relationship between K-{P5W30} concentration, solution pH (measured
solutions could store. The oxygen evolution reaction (see before electro-reduction) and the number of electrons that can be
Fig. S14, ESI†) was used to be the counterpart reaction of the extracted from a reduced solution. Solutions were charged 30 electrons.
(b) Representative 23 e reduction/reoxidation curves of a 10 mM solution
POM reduction reaction. Water is oxidized and simultaneously
of K-{P5W30}, 0.2 M H2SO4 was used as supporting electrolyte. (c) Rela-
produces O2, protons and electrons (2H2O - O2 + 4H+ + 4e ). tionship between Li-{P5W30} concentration, and the number of electrons
The protons generated can cross over Nafion membrane and that can be extracted from a reduced solution. (d) Representative 22 e
combine with electrons drawn from the external circuit to reduction/reoxidation curves of a 10 mM solution of Li-{P5W30}, 1 M H2SO4
protonate and reduce an aqueous solution of [NaP5W30O110]14 was used as supporting electrolyte. In (a) and (c), the number of electrons
used to reduce the K-{P5W30}and Li-{P5W30} was 30 per cluster. In (b) and
on the right-hand side of this cell, forming reduced polyoxoa-
(d), a current density of 10 mA cm 2 was applied. A comparison of
nion [NaP5W30O110](14+n) (n = equivalents of electrons charged). K-{P5W30} and Li-{P5W30}: (e) electrons extracted vs. concentration; (f)
Once the desired number of electrons per cluster had been coulombic efficiency (the total charge extracted out of the total charge put
stored, the reduced POM solution was then re-oxidized into the POM solutions) vs. concentration.

This journal is © The Royal Society of Chemistry 2023 Energy Environ. Sci.
 View Article Online

Paper Energy & Environmental Science

to 22.1 e (a comparison of 10 mM K-{P5W30} in H2O, 0.2 M

H2SO4 and 0.2 M Li2SO4 can be seen in Fig. S18, ESI†), however
upon increasing the concentration further to 25 mM and
50 mM the performance decreased to 17.2 and 8.5 e respec-
tively. A representative 23 e reduction/reoxidation curves of a
10 mM solution of K-{P5W30} in 0.2 M H2SO4 with the best
performance in this group is shown in Fig. 2b.
Lithium salts often have better aqueous solubility than the
corresponding potassium salts and some previous studies
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

showed small cations such as Li+ can stabilise the reduction of

POMs.23 Therefore we synthesized Li-{P5W30} with the expectation
Open Access Article. Published on 21 April 2023. Downloaded on 5/12/2023 7:48:47 AM.

this system might show better performance than K-{P5W30}.

However, after conducting flow cell tests under low concentration
of 2 mM, it was shown to only store 9.1 e per cluster, while the
5 mM solution can only store 9.3 e per cluster (Fig. 2c and
Fig. S19 and Table S6, ESI†). We were surprised that the perfor-
mance is even lower than the K salt given our previous work with
Li-P2W18, but upon investigating the 31P NMR of this salt in D2O
we found that this cluster degraded readily in solution. In contrast
a solution in D2SO4 showed no degradation (Fig. S20, ESI†). Thus,
H2SO4 was used as supporting electrolyte in experiments and
the effect of acid was studied. A 5 mM solution of Li-{P5W30} in
0.2 M H2SO4 was able to store 14.9 e , while a 10 mM solution
was shown to store 18.0 e . Increasing the acid concentration to Fig. 3 Reversible multi-electron redox chemistry of [H7P8W48O184]33 . (a)
1 M while maintaining Li-{P5W30} at 10 mM allowed storage Relationship between KLi-{P8W48} concentration and the number of
of up to 22.0 e . At 25 mM substrate in 1 M acid, we found the electrons that can be extracted from a reduced solution. The number of
performance dropped slightly to 21.0 e and performance was not electrons used to reduce the KLi-{P8W48} was 30/32 per cluster. 2 mM,
increased by further increasing acid concentration to 2 M. Thus, 5 mM and 10 mM are under room temperature; 25 mM and 35 mM are
under 70 1C. (b) Representative 27 e reduction/reoxidation curves of a 25
it seems these conditions represent the performance limit of mM solution of KLi-{P8W48}, 1 M H2SO4 was used as supporting electrolyte
Li-{P5W30}, and the representative reduction/reoxidation curves and a current density of 25 mA cm 2 was applied. (c) Relationship
are shown in Fig. 2d. In total, the electron storage ability of Li- between LiNH4-{P8W48} concentration and the number of electrons that
{P5W30} is affected by both concentrations of POM and supporting can be extracted from a reduced solution. (d) Representative 28 e
acid, increasing with greater concentration of POM from 2 mM to reduction/reoxidation curves of a 25 mM solution of LiNH4-{P8W48}. In
(b) and (d), a current density of 25 mA cm 2 was applied. A comparison of
10 mM and with decreasing pH. However, increasing the concen- KLi-{P8W48} and Li(NH4)-{P8W48}: (e) electrons extracted vs. concen-
tration further to 25 mM or lowering pH, does not improve the tration; (f) coulombic efficiency (the total charge extracted out of the total
performance more. A comparison of K-{P5W30} and Li-{P5W30} charge put into the POM solutions) vs. concentration.
with respect to extracted electrons and coulombic efficiency is
shown in Fig. 2e and 2f. Both K-{P5W30} and Li-{P5W30} show a
similar trend of first increasing and then decreasing with a peak carefully, there is an extra step appearing around 28–32 e , so
performance under 10 mM. In total, K-{P5W30} is better than for the following experiment we charged 30 e per cluster,
Li-{P5W30} in terms of electron storage ability and coulombic whereby a normal reoxidation curve was observed. KLi-{P8W48}
efficiency under lower concentration however Li-{P5W30} begins showed an increased electron storage ability with an increase
to show an advantage under higher concentration. in concentration from 2 mM, 5 mM, 10 mM to 25 mM (heating
was applied from 25 mM due to the solubility limit under
{P8W48} electron storage capacity studies room temperature), when charging 30 e per cluster, to a
Next the electron storage ability of the larger tungstate wheel, maximum of 26.9 e . We then tried to charge more (32 e
KLi-{P8W48}, was studied as a function of concentration (Fig. 3a per cluster) into this system, the performance did not increase
and Fig. S21 and Table S7, ESI†). Here 1 M H2SO4 was used as significantly. Increasing the concentration to 35 mM, the
the supporting electrolyte for all flow cell tests to overcome the performance dropped to 24.7 e . The best performance was
lower solubility of KLi-{P8W48} in H2O. Since {P2W18} was achieved at 25 mM with 26.9 e retrieved after charging by
effectively charged with 18 e and {P5W30} was able to be 30 e , shown in Fig. 3b. Similarly, for LiNH4-{P8W48} among
charged with 30 e we elected to first charging 48 e per cluster 10 mM, 25 mM and 35 mM, the concentration of 25 mM still
for this salt (i.e. assuming each WVI can be reduced to WV). displays the best performance (27.9 e out of 30 e , Fig. 3c, d
At this level of charge the POM shows decomposition with a and Table S8 and Fig. S22, ESI†). Overall the LiNH4 salt slightly
rapid decrease in electrochemical performance (Fig. S21, ESI,† outperformed the KLi salt at all concentrations measured
25 mM charge 48 e one). After observing the charge curve (Fig. 3e).

Energy Environ. Sci. This journal is © The Royal Society of Chemistry 2023
 View Article Online

Energy & Environmental Science Paper

Comparison of Li-{P2W18}, Li-{P5W30} and LiNH4-{P8W48} The concentration also affects the impact of different coun-
We compared the electrochemical performance of Li-{P2W18}, terions on performance. At low concentration protonation
Li-{P5W30} and LiNH4-{P8W48} in terms of extracted number of predominates over metalation, particularly for larger, more
electrons and coulombic efficiency versus concentration for a highly charged cations. Mass Spectrometry data of the two
more comprehensive understanding, shown in Fig. 4a and b. W30 salts in both acid and water shows that the potassium salt
From these two figures, we can see the electron storage ability becomes more protonated than the lithium salt in solution at
of POMs is affected by concentration of POMs and their the same concentration (Fig. S4, S9, S23, S24 and Tables S3, S4,
structure among {P2W18}, {P5W30} and {P8W48}. The electron S9, S10, ESI†). This difference in protonation can potentially
alter the ability of the POM to stabilise a high charge upon
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

storage ability for these three POMs shows both similarities

and differences. In terms of similarities, the electron storage reduction through aggregation. As concentration increases the
equilibrium shifts towards metalation, as described above. As
Open Access Article. Published on 21 April 2023. Downloaded on 5/12/2023 7:48:47 AM.

ability of them is all dependent on: (1) the concentration of

POMs, (2) the nature of the counter cations and (3) the such, at high concentration lithium salts show increased
supporting electrolyte pH or proton concentration. performance over Na+ and K+ salts for all of our POTs (W18,
W30 and W48). For {P8W48}, as both of the samples here have
mixed counter cations (K+ and Li+, Li+ and NH4+), it is hard to
Factors influencing performance analyse the specific effect of a single cation. Mass spectrometry
The concentration effect for the large clusters P5W30 and P8W48 showed poor solubility upon the water dilution required for
is more complex than for P2W18, whereby increasing concen- injection, however we can observe several broad peaks in the
tration leads to a continuous increase in performance. Instead envolopes expected for each of the salts. Exemplar spectral
optimal performance is observed at 10 mM for P5W30 and windows and simulated patterns for proposed salts can be
25 mM for P8W48, although the region between these two found in Fig. S41–S43 (ESI†). Importantly, the performance
concentrations is relatively flat for both (Fig. 4b). For the larger for the Li–(NH4)– salt is higher at all concentrations for this
POMs cation concentration increases more rapidly with cluster.
increasing POM concentration leading to differences in the For coulombic efficiency (Fig. 4b), both {P5W30} and {P8W48}
protonation/metalation equilibrium which occurs through showed an advantage over {P2W18} at low POM concentration
POM hydrolysis in aqueous media (Fig. 4d and Fig. S26, ESI†).33 (o20 mM), but for {P2W18}, the coulombic efficiency continues
These factors lead to an electron storage performance limit, to increase, approaching 100% when concentration reaches
and even lower performance due to the reduced precipitates 100 mM. It is interesting to note that normalising for W
which can form during this reduction/reoxidation process, concentration allows us to observe that all three clusters follow
affecting mass transfer in the flow system. a similar efficiency trajectory (Fig. 4c). Regarding the utilization
ratio of W atoms (the ratio of W metal centres which are redox-
active for storing electrons among all W atoms in the structure),
{P5W30} still showed an advantage at low concentration, but at
concentrations higher than 25 mM the advantage of {P2W18}
begins to emerge; {P8W48} has the lowest utilization rate of W,
only part of W in the structure are redox active and participate
in the reduction/oxidation process. (Fig. S25, ESI†)

NMR and mass spectrometry studies

NMR and mass spectrometry next allowed us to probe the effect
of concentration and reduction/oxidation on speciation, speci-
fically the cation–anion interaction in solution.29 Normally 7Li
NMR shows a sharper signal than 39K NMR, so here we use Li-
{P5W30} as an example to study changes in the samples with
varying concentration and reduction state through 7Li and
1
H NMR spectroscopy. Firstly, 10 mM of Li-{P5W30} was studied
in 1 M D2SO4, and both the 7Li and 1H NMR of original,
reduced, and re-oxidized POM solutions were monitored and
Fig. 4 A comparison of electrochemical performance between Li-
illustrated in Fig. 5a. Before treatment Li-{P5W30} resonates at
{P2W18}, Li-{P5W30} and LiNH4-{P8W48}. (a) Electrons extracted vs. concen- around 0.32 ppm in the 7Li spectrum reflecting the mass
tration (b) coulombic efficiency vs. concentration of polyoxotungstate spectra which show a mixture of lithium and proton cations for
(POT) (For W18, charge 18 e ; for W30, charge 30 e ; for W48, charge 30 the POM in solution (See Fig. S28, ESI†). After reduction, it
e , per cluster) (the data for W18 are from our prior work16). (c) Coulombic
shifted downfield by +0.06 ppm indicating a decrease in para-
efficiency vs. concentration of W. (d) Variation of total non-protic counter
ions with increasing cluster concentration for the three POMs compared
magnetic shielding of the lithium ion upon coordination to the
herein. Inset: POM hydrolysis equilibrium between metalation and POM over solvation29 (mass spec suggests all of the fourteen
protonation. available lithium ions are coordinated). (Tables S12 and S14, ESI†).

This journal is © The Royal Society of Chemistry 2023 Energy Environ. Sci.
 View Article Online

Paper Energy & Environmental Science

no extra peaks in the ESI-MS meaning this degradation may

occur due to the different rate of reaction of deuterons and
protons in electrochemical kinetics. To further confirm the
NMR chemical shift trends on 7Li and 1H, we ran an electro-
chemical test, charging 20 e of Li-{P5W30} in 1 M H2SO4, and
the results are shown in Fig. 5b. Similar trends on both 7Li and
1
H were observed here albeit with a smaller change in chemical
shift in both 7Li and 1H spectra due to the lower charge on the
POM. An illustration of this process is shown in Fig. 5c.
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

We also investigated both the 7Li and 1H NMR for the original
solution with varied concentration from 2 mM, 5 mM, 10 mM,
Open Access Article. Published on 21 April 2023. Downloaded on 5/12/2023 7:48:47 AM.

25 mM to 50 mM for LiCl, Li6{P2W18}, Li14{P5W30} and Li17-

(NH4)21{P8W48} (Fig. S31–35, ESI†). While no change in the shift
was observed for LiCl over the range (Fig. S31, ESI†) a shift of 7Li
NMR to downfield was observed when increasing the concen-
tration of all the POMs. This observation supports our assertions
above about the effect of increasing alkali metal concentration
upon the protonation/metalation equilibrium (effectively decreas-
ing POM hydrolysis). The upfield peak shift in the 1H NMR
spectrum with an increase in POM concentration is consistent
with a similar observation for Li salts in aqueous solution.34
Fig. 5 7Li and 1H NMR study (a) 10 mM of Li-{P5W30} in 1 M D2SO4 charge
30 e ; (b) 10 mM of Li-{P5W30} in 1 M H2SO4 charge 20 e ; (c) schematic Performance limit analysis
diagram of protonation/aggregation (reduction and reoxidation process)
From above electrochemical results, we found that both
{P5W30} and {P8W48} have an electron extraction limit which
Li orbitals are overlapping with the POM oxo ligand orbitals which is lower than the number of tungsten atoms in the cluster,
is also consistent with our previous work.17 Meanwhile, we mon- unlike {P2W18} which is able to release 17.2 electrons per
itored 1H NMR and an opposite trend compared to 7Li NMR was cluster, with a coulombic efficiency of 96%. Neither {P5W30}
observed. The peak around 4.8 ppm represents the protons from nor {P8W48} can store 30 e /48 e per cluster respectively with
H2O (or HOD) which in D2SO4 solution is in equilibrium with {P5W30} storing a maximum of 23 e per cluster, while {P8W48}
DH2O+ meaning an intermediate resonance representing the rela- can store 28 e per cluster. The first possible reason for this
tive position of this equilibrium is observed. After reduction, the performance limit could be POM decomposition during this
acidic protons (deuterons) are bound to the POM oxo ligands reduction/reoxidation process. During the electrochemical test,
shifting this equilibrium and resulting in a more upfield resonance we found that when charging 48 e per cluster into {P8W48} the
for the H2O/HOD protons. The reoxidation process of both 7Li and discharge curve rapidly decreased, likely due to degradation
1
H showed an opposite trend as the reduction process, but crucially of the POM. Secondly, when comparing the structures of the
the resonances did not return to the original peak position. (Fig. three POMs, we found that the {WO6} polyhedra could be either
S26, ESI†) This is due to the fact that production of H2 from the edge-sharing or corner-sharing with their neighbours. In
system reduces the concentration of H+ which shifts the equili- {P2W18} all of the W atoms are in edge sharing polyhedra
brium towards the formation of H2O/HOD (see performance limit whereas in {P5W30} this number is only 20 and {P8W48} has
section). This altered speciation upon reduction can provide some 32 edge-sharing polyhedra. Edge sharing polyhedra are known
explanation of the differences in the CV curves shown in Fig. 1d, if to undergo reduction more easily than their corner sharing
we assume the Li- and K- salts have different metalation levels. counterparts.35 Hence a higher barrier to reduction of the
Besides 1H and 7Li NMR, we also monitored ESI-MS and GC corner sharing could limit our ability to charge the cluster.
for this study and found there is an extra peak on 31P NMR at Another possible factor, which may work in concert with the
11.03 ppm after reoxidation (Fig. S27, ESI†), showing the prior issue, could be competitive hydrogen evolution within the
possible partial decomposition of the cluster. From the ESI- electrochemical potential range utilised here.
MS spectrum, the re-oxidized {P5W30} has extra peaks at
1091.69 and 1454.92 m/z, compared with the original and POM decomposition
reduced one. These two peaks can be assigned as species of We utilised 31P NMR and ESI-MS to study the performance limit
[H2P2W18O62]4 and [H3P2W18O62]3 , which is consistent with of Li-{P5W30} by charging 20, 24 and 30 electrons per cluster
the b-Dawson as shown in NMR spectrum (Fig. S29, ESI†). At respectively into Li-{P5W30} and monitoring the structure
the same time, the flow cell test result shows 11.8 e are change for the original, reduced and re-oxidized POMs of each.
extracted after the POM has been charged 30 e (See Fig. S30, At each of the three charging levels only [NaP5W30O110]14 is
ESI†), which is much less than using H2SO4 as solvent (22.0 e ). observed, indicating there is no decomposition occurring in the
Samples of the reduced and reoxidised POM in H2SO4 showed experiment (Fig. 6 and Fig. S24, S29, S30, ESI†).

Energy Environ. Sci. This journal is © The Royal Society of Chemistry 2023
 View Article Online

Energy & Environmental Science Paper

Consistent results were obtained; no decomposition was

observed. (See Fig. S38–S40 and Table S16–S19, ESI†). The
reason for this unavoidable hydrogen generation is possibly
related to the degree of protonation, the structure of the POM/
coordination environment of the W, the charge density of the
reduced species and the pH of the battery solution.16

Conclusions
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

This work shows that polyoxotungstate clusters like M14[NaP5-

W30O110] (M = K+, Li+) and M33[H7P8W48O184] (M = K+, Li+ and
Open Access Article. Published on 21 April 2023. Downloaded on 5/12/2023 7:48:47 AM.

NH4+) can achieve very high proton-electron storage capacities

in aqueous solution. {P5W30} can store up to 23 e and {P8W48}
can store up to 28 e per cluster, which is the best electron
storage ability reported to date for a single molecule.
However, unlike previously reported {P2W18} increasing con-
Fig. 6 A typical mass spectrum of sample of the polyoxometalate centration improves storage ability only up to around 10 mM
solution taken from a 10 mM solution of Li-{P5W30} that was (a) original
(P5W30)–25 mM (P8W48) whereupon the rapidly increasing
(b) reduced by 30 e (c) re-oxidized and then injected into the mass
spectrometer (after dilution with degassed methanol). The reduced sample concentration of cations may impede the ability of the clusters
was taken out immediately after charging finished and protected with Ar to to aggregate through proton bridges. Using acidic electrolyte
avoid oxidation by air, before injection it was also diluted with degassed media can mitigate this effect to a degree and stabilise the
methanol. POM. At low concentration heavy alkali metal salts (e.g. K) show
increased storage likely due to cluster hydrolysis however these
are surpassed by more electropositive metal salts (e.g. Li) as
Hydrogen evolution vs. W edge-sharing effect
concentration and thus metalation increases. The performance
To analyse for hydrogen evolution and the edge sharing effect limit in coulombic efficiency for these clusters is caused by
we designed a series of three experiments, observing the unavoidable hydrogen generation. In this regard this work
coulombic efficiency and detection of H2. If the edge sharing helps understand how polyoxometalates might be exploited
effect is the principal reason for the performance limit, then for flow batteries and how clusters with greater stability and
when we charge to 20 e we expect to see 100% coulombic even higher capacities may be designed.
efficiency while if we charge to more than 20 e we expect to
observe hydrogen formation. However, if, no matter how many
electrons are added, the coulombic efficiency is always the Author contributions
same (or similar), then hydrogen is formed under all circum-
L. C. conceived the idea, and L. C and T. Z designed the project.
stances and the performance limit is caused by unavoidable
T. Z synthesized {P5W30} (including all physical characteriza-
hydrogen generation.
tion), conducted all electrochemistry tests and analysis of
The electrochemical result and gas analysis for this series of
{P5W30} and {P8W48}, conducted all NMR, MS and GC study
experiment are shown in Table 1. When charged to 30 e , 22 e
of {P5W30} for mechanisms study and relevant analysis, ana-
was extracted and with H2 detected; when charged to 24 e , less
lysed all data together, and wrote the paper; N. L. B contributed
electrons (16.4 e ) were extracted with H2 detected (Fig. S36,
on NMR and MS discussion and analysis and helped write the
ESI†); when charged to 20 e , a significantly lower number of
paper; G. C helped with building up the flow cell and con-
electrons (14.8 e ) were released out (Fig. S37, ESI†), with H2
tributed on GC measurement; B. K synthesized {P8W48} and
still detected. This demonstrates that the performance limit of
finished physical characterization of {P8W48}; D-L. L contribu-
Li-{P5W30} is due to unavoidable hydrogen generation rather
ted on single crystal XRD characterization and analysis.
than the edge-sharing effect. At the same time, we also mon-
itored the structure change (through ESI-MS and 31P NMR) for
reduced and oxidized ones for charging 20 and 24 e . Conflicts of interest
There are no conflicts to declare.
Table 1 A comparison of battery performance and hydrogen detection of
Li-{P5W30} charge to different electrons (30, 24 and 20 e per molecule)
Acknowledgements
Electrons Electrons Coulombic effi- H2
input output ciency (%) detected We gratefully acknowledge financial support from the EPSRC
(no. EP/J015156/1; EP/L023652/1; EP/I033459/1; EP/J0151; EP/
Li-{P5W30} 10 mM in 30.0 22.0 73 Yes
1 M H2SO4 24.0 16.4 68 Yes K023004/1; and EP/L023652/1); the European Research Council
20.0 14.8 74 Yes (project 670467 SMART-POM); and the University of Glasgow.

This journal is © The Royal Society of Chemistry 2023 Energy Environ. Sci.
 View Article Online

Paper Energy & Environmental Science

We thank Dr Chang-gen Lin providing helpful suggestion on 17 J.-J. Chen, L. Vilà-Nadal, A. Solé-Daura, G. Chisholm,
the synthesis of Li-{P5W30}; we thank Dr Maria D. Castro, Dr T. Minato, C. Busche, T. Zhao, B. Kandasamy, A. Y. Ganin,
Jennifer S. Mathieson and Mr Jim McIver’s help on performing R. M. Smith, I. Colliard, J. J. Carbó, M. P. Josep, M. Nyman
mass spectra, TGA and relevant data analysis. and L. Cronin, J. Am. Chem. Soc., 2022, 144, 8951–8960.
18 M. H. Alizadeh, S. P. Harmalker, Y. Jeannin, J. Martin-Frere
and M. T. Pope, J. Am. Chem. Soc., 1985, 107, 2662–2669.
References 19 A. Hayashi, T. Haioka, K. Takahashi, B. S. Bassil, U. Kortz,
1 M. R. Horn, A. Singh, S. Alomari, S. Goberna-Ferrón, T. Sano and M. Sadakane, Z. Anorg. Allg. Chem., 2015, 641,
R. Benages-Vilau, N. Chodankar, N. Motta, K. Ostrikov, 2670–2676.
This article is licensed under a Creative Commons Attribution 3.0 Unported Licence.

J. MacLeod, P. Sonar, P. Gomez-Romero and D. Dubal, 20 A. Hayashi, M. N. K. Wihadi, H. Ota, X. Lopez, K. Ichihashi,
Energy Environ. Sci., 2021, 14, 1652–1700. S. Nishihara, K. Inoue, N. Tsunoji, T. Sano and
Open Access Article. Published on 21 April 2023. Downloaded on 5/12/2023 7:48:47 AM.

2 Z. Li and Y.-C. Lu, Adv. Mater., 2020, 32, 2002132. M. Sadakane, ACS Omega, 2018, 3, 2363–2373.
3 W. Wang, Q. Luo, B. Li, X. Wei, L. Li and Z. Yang, Adv. Funct. 21 R. Contant and A. Teze, Inorg. Chem., 1985, 24, 4610–4614.
Mater., 2013, 23, 970–986. 22 M. Segado, M. Nyman and C. Bo, J. Phys. Chem. B, 2019, 123,
4 J. Zhang, G. Jiang, P. Xu, A. Ghorbani Kashkooli, 10505–10513.
M. Mousavi, A. Yu and Z. Chen, Energy Environ. Sci., 2018, 23 A. Misra, K. Kozma, C. Streb and M. Nyman, Angew. Chem.,
11, 2010–2015. Int. Ed., 2020, 59, 596–612.
5 M. L. Perry, K. E. Rodby and F. R. Brushett, ACS Energy Lett., 24 V. A. Grigoriev, D. Cheng, C. L. Hill and I. A. Weinstock,
2022, 7, 659–667. J. Am. Chem. Soc., 2001, 123, 5292–5307.
6 X.-P. Gao and H.-X. Yang, Energy Environ. Sci., 2010, 3, 25 A. B. Gunnarsdottir, C. V. Amanchukwu, S. Menkin and
174–189. C. P. Grey, J. Am. Chem. Soc., 2020, 142, 20814–20827.
7 D. L. Long, E. Burkholder and L. Cronin, Chem. Soc. Rev., 26 K. A. See, M. Leskes, J. M. Griffin, S. Britto, P. D. Matthews,
2007, 36, 105–121. A. Emly, A. Van der Ven, D. S. Wright, A. J. Morris, C. P. Grey
8 D. L. Long, R. Tsunashima and L. Cronin, Angew. Chem., Int. and R. Seshadri, J. Am. Chem. Soc., 2014, 136, 16368–16377.
Ed., 2010, 49, 1736–1758. 27 E. W. Zhao, T. Liu, E. Jonsson, J. Lee, I. Temprano,
9 H. D. Pratt, N. S. Hudak, X. Fang and T. M. Anderson, R. B. Jethwa, A. Wang, H. Smith, J. Carretero-Gonzalez,
J. Power Sources, 2013, 236, 259–264. Q. Song and C. P. Grey, Nature, 2020, 579, 224–228.
10 J. Friedl, F. L. Pfanschilling, M. V. Holland-Cunz, R. Fleck, 28 D. Sures, M. Segado, C. Bo and M. Nyman, J. Am. Chem. Soc.,
B. Schricker, H. Wolfschmidt and U. Stimming, Clean 2018, 140, 10803–10813.
Energy, 2019, 3, 278–287. 29 D. J. Sures, S. A. Serapian, K. Kozma, P. I. Molina, C. Bo and
11 J. P. Launay, Phys. Chem. Chem. Phys., 1976, 38, 807–816. M. Nyman, Phys. Chem. Chem. Phys., 2017, 19, 8715–8725.
12 C. Boskovic, M. Sadek, R. T. C. Brownlee, A. M. Bond and 30 E. G. Bloor and R. G. Kidd, Can. J. Chem., 1972, 50,
A. G. Wedd, J. Chem. Soc., Dalton Trans., 2001, 187–196, DOI: 3926–3930.
10.1039/b006264m. 31 M. J. Turo, L. Chen, C. E. Moore and A. M. Schimpf, J. Am.
13 D. M. Way, A. M. Bond and A. G. Wedd, Inorg. Chem., 1997, Chem. Soc., 2019, 141, 4553–4557.
36, 2826–2833. 32 B. Keita, Y. W. Lu, L. Nadjo and R. Contant, Electrochem.
14 H. Wang, S. Hamanaka, Y. Nishimoto, S. Irle, T. Yokoyama, Commun., 2000, 2, 720–726.
H. Yoshikawa and K. Awaga, J. Am. Chem. Soc., 2012, 134, 33 N. I. Gumerova and A. Rompel, Chem. Soc. Rev., 2020, 49,
4918–4924. 7568–7601.
15 E. Sánchez-Dı́ez, E. Ventosa, M. Guarnieri, A. Trovò, C. Flox, 34 S. Cai, T. Bai, H. Chen, W. Fang, Z. Xu, H. Lai, T. Huang,
R. Marcilla, F. Soavi, P. Mazur, E. Aranzabe and R. Ferret, H. Xu, X. Chu, J. Ling and C. Gao, J. Phys. Chem. Lett., 2020,
J. Power Sources, 2021, 481, 228804. 11, 303–310.
16 J. J. Chen, M. D. Symes and L. Cronin, Nat. Chem., 2018, 10, 35 R. I. Buckley and R. J. H. Clark, Coord. Chem. Rev., 1985, 65,
1042–1047. 167–218.

Energy Environ. Sci. This journal is © The Royal Society of Chemistry 2023