Open Archive TOULOUSE Archive Ouverte (OATAO)
OATAO is an open access repository that collects the work of Toulouse researchers and
makes it freely available over the web where possible.
This is an author-deposited version published in : http://oatao.univ-
toulouse.fr/
Eprints ID : 16801
To link to this article : DOI : 10.1038/NMAT4318
URL : http://dx.doi.org/10.1038/NMAT4318
To cite this version :
Griffin, John M. and Forse, Alexander C. and
Tsai, Wan-Yu and Taberna, Pierre-Louis and Simon, Patrice and
Grey, Clare P. In situ NMR and electrochemical quartz crystal
microbalance techniques reveal the structure of the electrical
double layer in supercapacitors. (2015) Nature Materials, vol. 14
(n° 8). pp. 812-819. ISSN 1476-1122
Any correspondence
concerning this service should be sent to the repository
administrator: staff-oatao@listes-diff.inp-toulouse.fr
In situ NMR and electrochemical quartz crystal
microbalance techniques reveal the structure of
the electrical double layer in supercapacitors
John M. Grin
1
, Alexander C. Forse
1
, Wan-Yu Tsai
2,3
, Pierre-Louis Taberna
2,3
, Patrice Simon
2,3
and Clare P. Grey
1,4
*
Super
capacitors store charge through the electrosorption of ions on microporous electrodes. Despite major eorts to
understand this phenomenon, a molecular-level picture of the electrical double layer in working devices is still lacking as few
techniques can selectively observe the ionic species at the electrode/electrolyte interface. Here, we use in situ NMR to directly
quantify the populations of anionic and cationic species within a working microporous carbon supercapacitor electrode. Our
results show that charge storage mechanisms are dierent for positively and negatively polarized electrodes for the electrolyte
tetraethylphosphonium tetrafluoroborate in acetonitrile; for positive polarization charging proceeds by exchange of the cations
for anions, whereas for negative polarization, cation adsorption dominates. In situ electrochemical quartz crystal microbalance
measurements support the NMR results and indicate that adsorbed ions are only partially solvated. These results provide new
molecular-level insight, with the methodology oering exciting possibilities for the study of pore/ion size, desolvation and
other eects on charge storage in supercapacitors.
T
he mechanism of charge storage in supercapacitors has
traditionally been attr ibuted to the electrosorption of ions on
the surface of a charged electrode to form an electrical double
layer. However, in recent years a number of empirical observations
have shown that the mechanism is more complex, with factors
such as relative pore/ion sizes
1–3
and desolvation effects
4,5
playing
important roles. Theoretical studies have led the way in under-
standing supercapacitor charging on the molecular level
6,7
, and have
demonstrated that charge screening
8–10
, ionic rearrangement
11
and
confinement
12
, and pore surface properties
13
can have significant
effects on the capacitance and charging dynamics. Nevertheless,
theoretical simulations necessarily depend on assumptions and sim-
plifications, and many questions concerning det ails of the charging
mechanisms in real devices remain unanswered. In particular, it
is not clear whether charging is a purely adsorptive process, or
whether exchange of one set of ions by those with opposite charge
(referred to in this paper as ‘ion exchange’) and co-ion expulsion
from the charged electrodes also contribute to t he formation of the
electrical double layer. Recently, in situ experimental methodolo-
gies based on electrochemical quartz crystal microbalance (EQCM)
techniques
14,15
and infrared spectroscopy
16,17
have started to address
these questions. These methods can observe ion adsorption and
expulsion in charged electrodes, and have been able to distinguish
purely adsorptive regimes from ion mixing during charging. How-
ever, neither of these techniques alone permits the direct quantifica-
tion of species within the electrical double layer in absolute terms,
with EQCM measuring total mass changes in the ele ctrode and
infrared spectroscopy measuring only the ions outside the pores,
and so an unambiguous picture of the charging mechanism has not
yet been obtained.
One approach that has recently shown promise for the study of
supercapacitorsis nuclear magnetic resonance(NMR) spectroscopy.
NMR has the advantage that it is element sele ctive, thereby allowing
individual ionic species to be obs erved independently
18,19
. Ex situ
NMR measurements on disassembled supercapacitor electrodes
have revealed changes in the populations and local environments of
ionic species that result from charging
20
. In situ NMR and magnetic
resonance imaging methods have also been de veloped, which allow
changes in the local environments of the ions in the electric double
layer to be observed for working devices
21–24
. These approaches
have provided qualitative insight into the charging mechanism for
a range of electrolyte systems. However, in principle, NMR is fully
quantitative and should enable the measurement of absolute ion
populations at the electrode/electrolyte interface. In particular, the
combination of NMR with EQCM, which tracks the displacement of
all electrolyte spe cies including s olvent molecules, should provide a
full description of the structure of the electric double layer.
Here, we use tetraethylphosphonium tetrafluoroborate(PEt
4
–BF
4
)
salt dissolved in acetonitrile (ACN) as the electrolyte, employing
31
P and
19
F NMR experiments to enable the selective observation of
the PEt
4
cations and BF
4
anions, respectively. Deuterated ACN was
also used to enable the obser vation of solvent species by
2
H NMR
(as described in Supplementar y Information). Experiments have
been performed on commercial YP-50F activated carbon, for which
gas sorption measurements show an average pore size of 0.9 nm
with 92% of pores being smaller than 2 nm. The specific surface
area is 1,730 m
2
g
−1
and the total pore volume is 0.75 cm
3
g
−1
(see
Supplementary Information).
Supercapacitor bag cells were constructed following a shifted
‘overlaid’ design, allowing a single electrode to be studied
1
Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, UK.
2
Université Paul Sabatier Toulouse III, CIRIMAT,
U
MR-CNRS 5085, F-31062 Toulouse, France.
3
Réseau sur le Stockage Electrochimique de l’Energie (RS2E), FR CNRS 3459, France.
4
Department of
Chemistry, Stony Brook University, Stony Brook, New York 11794-3400, USA.
*
e-mail: cpg27@cam.ac.uk
Ex-poreEx-pore
In-poreIn-pore
1.5 M
0.75 M
0.5 M
Cations observed Anions observed
a b
c
−150 −160−140 −170
−150 −160−140 −170
50 40
δ
31
P (ppm)
δ
18
F (ppm)
30 20
In-pore
δ
31
P (ppm)
δ
31
P (ppm) δ
19
F (ppm)
50 40 30 20
d
−150 −160−140
−170
δ
19
F (ppm)
In-pore
e f
In-pore
50 40 30 20
In-pore
Figure 1 | NMR spectra of individual supercapacitor electrodes showing in- and ex-pore cation and anion environments. a–f,
31
P (a,c,e) and
19
F (b,d,f)
NMR spectra were recorded for electrolyte concentrations of 1.5 M (a,b), 0.75 M (c,d) and 0.5 M (e,f). Supercapacitors were held at a cell voltage of 0 V.
The in-pore resonances in each spectrum are highlighted.
independently inside the NMR coil whilst maintaining good
capacitive properties
22
. Electrodes were fabricated using 95 wt%
YP-50F and 5 wt% PTFE binder (see Methods). Figure 1 shows
31
P and
19
F NMR spectra of individual supercapacitor electrodes
in cells held at 0 V with electrolyte concentrations of 1.5, 0.75 and
0.5 M. In each spectrum, intense ‘ex-pore’ resonances are observed
at 40 ppm (
31
P) and −150 ppm (
19
F), corresponding to ions located
in voids between carbon particles and in electrolyte that resides
in a small reservoir between the two electrodes. Small features
visible in the ex-pore resonances are due to susceptibility effects
associated with the bag cell components and geometric anisotropy
of the cell
21,22
. Weaker ‘in-pore’ resonances corresponding to ions
inside the micropores close to carb on surfaces are also observed,
shifted to lower frequency by 5–7 ppm. The shift of the in-pore
resonance from that of the ex-pore resonance is due to diamagnetic
nucleus-independent chemical shift (NICS, often referred to as
‘ring cur rent’) effec ts associated with the delocalized electrons in
the predominantly sp
2
-bonded carbon surface
22,25
. Calculations
indicate that the NICS effect should be significant for species within
a few ångstroms from the carbon surface. However, the effects
of dynamics must also be considered as fast exchange between
different positions within a pore will result in an averaging of the
observed chemical shift. Taking this into account, NICS values of
around 5 ppm have been predicted for species inside pores up to
2 nm in width, consistent with the shifts observed here
25
.
Through comparison of t he in-pore resonance intensities with
calibration samples, it is possible to quantify the number of in-
pore species in absolute terms and gain insight into the adsorption
properties. In-pore ion populations are plotted as a function
of electrolyte concentration in Fig. 2. For the three electrolyte
concentrations studied, in-pore anion and cation populations
at 0 V are balanced and vary approximately linearly with the
concentration. Using estimated s olvated ion diameters of 1.35 nm
(PEt
4
+
) and 1.16 nm (BF
4
−
; ref. 26; see Supplementary Information)
and the measured ion uptakes at 0 V, we find that the in-pore
ions should occupy a total volume of 1.12 cm
3
(1.5 M electrolyte),
0.56 cm
3
(0.75 M electrolyte) and 0.37 cm
3
(0.5 M electrolyte) per
gram of YP-50F. For the 1.5 M electrolyte, the total volume of the
0.25
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
0.50 0.75 1.00 1.25 1.50
Cations
Anions
1.0
0.0
Electrolyte concentration (mol L
−1
)
In-pore ion population (mmol g
−1
)
Figure 2 | In-pore ion populations per gram of YP-50F at 0 V plotted as a
function of concentration for the PEt
4
-BF
4
/ACN electrolyte. Error bars
represent the range of values obtained from measurements performed on
two separate electrodes within each calibration cell (see Supplementary
Information). For each concentration, equal populations of cations and
anions are adsorbed and the total in-pore ion population varies
approximately linearly with electrolyte concentration.
in-pore ions estimated on this basis is significantly larger than the
total pore volume of 0.75 cm
3
g
−1
for YP-50F. This indicates that for
this concentration, the assumption of each ion having a complete
ACN solvation shell is not valid; instead, the ions must be more
densely packed inside the micropores, with partial desolvation or
overlap of their solvation shells.
To investigate changes in the in-pore ion behaviour during
charging, in situ NMR experiments were performed as the
supercapacitor cells were charged sequentially from total cell
voltages of 0 to 1.5 V in steps of 0.25 V. They were then discharged
in a single step to 0 V, before being charged in steps of −0.25 to
−1.5 V.
19
F and
31
P in situ NMR spectra were acquired at each
voltage (Fig. 3) after cells had b een held for 60 min (1.5 and 0.75 M
electrolyte) or 90 min (0.5 M electrolyte), until a negligible constant
residual current was obtained (see Supplementary Information). In
both voltage ranges, the in-pore c ation and anion resonances move
to higher frequency in the NMR spectra as the cells are charged. This
is consistent with previous results
21–23
and is due to changes in the
NICSs that result from the electronic charge that is developed within
the carbon electrode: charging of π-bonded carb on systems gives
rise to so-called paratropic or anti-aromatic ring currents, which
result in positive NICSs for nearby species
22
.
Whereas the frequencies of the in-pore resonances depend on
the ele ctronic charge state of the electrode surface, importantly, the
intensities of the in-pore resonances correspond to the number of
in-pore ions w ithin the electrode. As the cell potentials are varied,
changes in the in-pore resonance intensities reflect the changing
ion populations within the electrical double layer that is formed
at the electrode/electrolyte interface within the micropores. In the
positive voltage range, the in-pore resonances in the
19
F NMR
spectra increase in intensity as the cell is charged, showing that
anions are absorbed into the micropores during charging. For the
negative voltage range, an increase in the in-pore cation population
is observed in the
31
P NMR spectra as cations are absorbed into
the micropores. These results provide a qualitative picture that is
consistent with the accepted view of supercapacitor charging: as
electronic charge accumulates within the electrode, ions of opposite
chargeare adsorbed ontothe surface to form an electric double layer.
Absolute ion populations determined from the deconvoluted in-
pore resonance intensities, plotted in Fig . 4, offer more quantitative
insight and provide a detailed compositional picture of the electrical
double layer during charging. Interestingly, for the positive cell
voltage range, we find that the charging mechanism is not a purely
adsorptive process. For al l electrolyte concentrations, t he in-pore
anion population increases together with a simultaneous decrease
in the in-pore cation population. This shows that charge storage is
actually driven by ion exchange, whereby anions are absorbed into
the micropores while cations are simultaneously ejected. This leads
to an overall negative ionic charge, which forms an electrical double
layer with the positive electronic charge that accumulates on the
electrode surface. Over the voltage range studied, the number of
cations ejected from the micropores is approximately equal to the
number of anions adsorbed, meaning that the total number of in-
pore ions does not change significantly. For the 0.5 M electrolyte,
an apparent increase in in-p ore cation population is observed
between 1.0→ 1.5 V. However, for these voltages we note that there
are larger uncertainties in deconvoluting the very low-intensity
in-pore resonances in the experimental
31
P NMR spect ra. In the
negative voltage range, a different charging mechanism is observed:
between 0→ −1.5 V, the in-pore cation population increases for all
electrolyte concentrations, but there are no significant changes in
the in-pore anion populations. For each ele ctrolyte concentration,
the relative change in the cation population is approximately double
that observed in the positive charging regime, thus preserving
electroneutrality. Therefore, charge storage in the negative voltage
range is dominated by counter-ion adsorption, with an overall
increase in the number of in-pore ions. Note that whereas the
electrochemical stability window of ACN-based electrolytes can be
as large as 3 V, the potential range studied here was limited to
±1.5 V owing to the significant overlap of the in-pore and ex-pore
resonances at >1.5 V.
It is straightforward to determine the total ionic charge within the
carbon electrode at each voltage step by summing the positive and
negative charges associated with the in-pore ion populations. For
the three concentrations studied, we find good overall agreement
between the ionic charge inside t he micropores and the stored
electronic charge (Fig. 5), the l atter being determined from
integration of the current versus time plots at each voltage step.
At high voltages between 1 and 1.5 V for the 0.5 M electrolyte, a
noticeable deviation is observed; this is, at least in part, related
to the difficulty in accurately deconvoluting the weak
31
P in-
pore resonance intensity at these voltages, with the weak signal
overlapping with the stronger signal from the ex-pore ions.
The overall agreement between the ionic and electronic charge
shows that the ions in t he in-pore environment (within a few
nanometres fromthe electrode surface) are primarily responsible for
charge storage.
To probe the local environments of the solvent molecules,
in situ
2
H NMR spectra were recorded for a supercapacitor cell
containing 1.5 M electrolyte (see Supplementary Information). The
resonances corresponding to in-pore and ex-pore/ACN solvent
molecules are much broader, probably as a result of faster exchange
processes affecting the highly mobile solvent molecules, precluding
an accurate deconvolution of in-pore resonance intensities. Instead,
to gain furt her information about the charging mechanisms and
behaviour of the solvent molecules, t he same system was studied
in situ using an EQCM. Particles of YP-50F carbon were deposited
on a piezoelectric quartz crystal resonator whose resonance
frequency can be related to the mass of the crystal electrode
through the Sauerbrey equation
27
, thus providing information
about the ion and solvent molecule fluxes during the charging of
the porous carbon electrode. The carbon-coated resonator was used
as a working electrode in a three-electrode EQCM cell containing
PEt
4
–BF
4
/ACN electrolyte at the intermediate concentration of
0.75 M. In contrast to previous EQCM studies, which probed
dynamic charging using cyclic voltammetry exper iments
14,15
,
measurements were performed under steady-state conditions (that
is, at fixed voltages) to mirror the NMR experimental methodology.
Cell voltage (V) Cell voltage (V)
Postive polarization Negative polarization
1.25
1.50
1.00
0.75
0.50
0.25
0.00
−1.25
−1.50
−1.00
−0.75
−0.50
−0.25
0.00
1.25
1.50
1.00
0.75
0.50
0.25
0.00
−1.25
−1.50
−1.00
−0.75
−0.50
−0.25
0.00
−145 −150 −155 −16045 40 3035 25
−145 −150 −155 −160
45 40 3035 25
−145 −150 −155 −160
45 40 3035 25
−145 −150 −155 −160
45 40 3035 25
−145 −150 −155 −16045 40 3035 25
−145 −150 −155 −16045 40 3035 25
a b
c d
e f
δ
31
P (ppm) δ
19
F (ppm) δ
19
F (ppm)δ
31
P (ppm)
Cations observed Cations observedAnions observed Anions observed
1.5 M
0
.75 M
0.5 M
Figure 3 | In situ
31
P and
19
F NMR spectra of individual supercapacitor electrodes at dierent states of charge. a–f, Spectra recorded in the range 0 to
1.5 V (a,c,e) and in the range 0 to −1.5 V (b,d,f). Electrolyte concentrations are 1.5 M (a,b), 0.75 M (c,d) and 0.5 M (e,f). In-pore anion intensities increase
for positive charging, whereas in-pore cation intensities increase for negative charging.
These steady-state experiments should be less influenced by the
kinetics of adsorption; rather, a complete reorganization of the ions
and solvent molecules in the pores to approach the lowest-energy
arrangements is possible.
The electrode was polarized from the open-circuit voltage
(OCV) of 0.43 V versus the Ag reference electrode up to +0.7 V
versus Ag in steps of 0.1 V. The electrode was then discharged
to the OCV before being charged to 0 V versus Ag in steps of
−0.1 V. The electrode was held at each voltage for 120 s until
a constant residual current was obtained (Supplementary Fig. 9
in Supplementary Information). The initial OCV was close to
the point of zero charge (PZC) of the electrode (0.49 V versus
Ag), with the PZC being measured in a separate experiment
(see Supplementary Information). The sequence was repeated
several times to ensure reproducibility of the results. The low
carbon loading on t he resonator (tens of micrograms) restricts the
electrochemical window such that purely capacitive behaviour is
observed only in the potential range 0↔ +0.7 V versus Ag. Outside
this window, there are significant redox contributions to the total
current (see Supplementary Fig. 10 in Supplementary Information).
Measured mass changes (normalized by the electrochemically active
area of the quartz resonator, 1.27 cm
2
) are plotted as a function
of the ele ctrode potential in Fig. 6a. The mass of the electrode
is observed to increase at negative potentials relative to the OCV.
This behaviour is qualitativelyconsistent with the adsorption-driven
charging mechanism inferred from the NMR data, the absorption of
cations into the micropores with no significant change in the anion
population resulting in an increase in the total number of ions and
hence mass of the electrode. For positive potentials relative to OCV,
the electrode mass is found to decrease slightly over the potential
range studied. This behaviour is consistent with the ion exchange-
driven charging mechanism determined from the NMR data. The
adsorption of anions and simultaneous expulsion of cations from
the electrode can result in an overall decrease in the total in-pore
ionic mass because the BF
4
anions have a significantly smaller mass
(86.8 g mol
−1
) than the PEt
4
cations (147 g mol
−1
).
It is possible to gain a more quantitative interpretation of
the EQCM results by comparing the exper imental mass changes
with theoretical values calculated assuming different models for
the charging mechanism. Figure 6b shows the experimental mass
changes plotted as a function of the capacitive charge stored
on the electrode, c alculated by integrating current versus time
plots using Faraday’s law
15
. Theoretical mass changes based on
two different charging mechanisms are also shown. For the first
model, a purely adsorptive mechanism is considered, based on the
traditional assumption that one negative (positive) charge stored
in the electrode is balanced by the adsorption of a single cation
(anion) on the electrode surface. For the second model, an ion
exchange mechanism was assumed whereby for positive (negative)
electrode polarization, two charges stored on the electrode sur face
are compensated by the adsorption of one anion (cation) and
desorption of one cation (anion). For negative polarization, the
purely adsorptive model (green triangles) predicts the largest mass
change, although the measured mass changes are much higher
than those predicted by either model across the entire potential
range. In contrast to the NMR data shown in Fig. 3, which
selectively observe in-pore cation and anion populations, mass
changes measured by EQCM originate from all electrolyte species
