Stable Electrolyte for High Voltage Electrochemical Double-Layer Capacitors

Rose E. Ruthera,z, Che-Nan Sunb, Adam Hollidaya, Shiwang Chenga, Frank M. Delnicka, Thomas
A. Zawodzinski Jr.a,c, and Jagjit Nandaa,c,z
a
Oak Ridge National Laboratory, Oak Ridge, TN 37831 USA
b
Electrosynthesis Company, Lancaster, New York 14086, USA
c
Department of Chemical and Biomolecular Engineering, University of Tennessee, Knoxville,
Tennessee 37996, USA
z
Corresponding Authors:
E-mail address: [email protected]. Telephone:(865)946-1578
E-mail address: [email protected]. Telephone:(865)241-8361

This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of
Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the
United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published
form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will
provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan
(http://energy.gov/downloads/doe-public-access-plan).

1
 Abstract

A simple electrolyte consisting of NaPF6 salt in 1,2-dimethoxyethane (DME) can extend the
voltage window of electric double-layer capacitors (EDLCs) to >3.5 V. DME does not passivate
carbon electrodes at very negative potentials (near Na/Na+), extending the practical voltage
window by about 1.0 V compared to standard, non-aqueous electrolytes based on acetonitrile.
The voltage window is demonstrated in two- and three-electrode cells using a combination of
electrochemical impedance spectroscopy (EIS), charge-discharge cycling, and measurements of
leakage current. DME-based electrolytes cannot match the high conductivity of acetonitrile
solutions, but they can satisfy applications that demand high energy density at moderate power.
The conductivity of NaPF6 in DME is comparable to commercial lithium-ion battery electrolytes
and superior to most ionic liquids. Factors that limit the voltage window and EDLC energy
density are discussed, and strategies to further boost energy density are proposed.

Keywords: supercapacitor, double layer capacitor, electrolyte, glyme, ether, electrochemical

window, voltage window

1. Introduction
Electric double-layer capacitors (EDLCs) store ionic charge electrostatically at the interface of
high surface area electrodes such as carbon in a liquid electrolyte.1, 2
Efforts to increase the
energy density of EDLCs focused mainly on developing higher surface area electrodes and
controlling pore size.3-6 Energy density of capacitors can also be increased through faradaic
mechanisms commonly known as pseudocapacitance, e.g. introduction of redox active groups
through carbon surface functionalization or incorporation of metal oxides.7-15 Despite significant
improvements in electrode and material design, virtually all state-of-the-art, non-aqueous
electrochemical capacitors use the same electrolyte: tetraethylammonium tetrafluoroborate
(TEABF4) in acetonitrile (ACN) or propylene carbonate (PC). Both electrolytes have very high
conductivities (>10 mS/cm for PC and >50 mS/cm for ACN), which minimize resistive losses
and enable capacitors to operate at very high power.16, 17 However, these electrolytes are limited
by a practical voltage window around only 2.5 – 3.0 V, beyond which the capacitor lifetime is
significantly shortened.2, 18-21 Since the energy stored in a capacitor increases quadratically with

2
voltage, extending this electrochemical window could significantly improve the energy density. 1,
2, 19
Other organic solvents (adiponitrile,22, 23
sulfones,24, 25
and carbonates25, 26
) have been
considered for high voltage electrolytes for EDLCs, but typically suffer from higher viscosity
and lower electrolyte conductivity. Ionic liquids have generated increased interest due to their
high stability, but remain limited by high cost, low purity, and low conductivity.19, 27-30

With prolonged cycling, the capacitance of EDLCs decreases and the resistance increases. 31 The
performance degrades more rapidly at elevated temperature or higher voltage.32-34 Degradation is
typically attributed to decomposition of the electrolyte, and is very sensitive to the electrolyte
composition, electrode polarity, carbon surface functionality, and trace moisture.31-35 The long-
term performance of EDLCs can also be limited by the stability of other components in the cell
including the polymer binders and current collectors (typically aluminum).32, 34, 35 Commercial
EDLCs with organic electrolytes operate over a voltage window of ~ 3 V (approximately 1.5 -
4.5 V vs. Na/Na+).35 Developing higher voltage electrolytes for EDLCs requires careful
consideration and control of all possible side reactions. For example, extending the positive
voltage limit beyond 4.5 V vs. Na/Na+ likely requires strategies to effectively suppress corrosion
of the aluminum current collector.36-39 Carbon oxidation may also occur at high voltage.34

Extending the negative voltage limit below 1.5 V vs. Na/Na+ also presents certain challenges.
The solvents most commonly used in lithium-ion batteries and EDLCs (carbonates and ACN)
passivate electrodes at potentials below about 1.2 V vs. Na/Na+.40-42 Effective passivation of the
negative electrode is critically important for the operation of lithium-ion batteries, but
detrimental for double-layer capacitors. Even very thin insulating surface films can reduce the
double layer capacitance and block small pores.31-34, 43 Binders based on polytetrafluoroethylene
(PTFE), which are commonly used in commercial EDLC electrodes, are also reduced below
about 1.0 V vs. Na/Na+.32 Finally, the stability of the carbon itself with respect to reduction
and/or intercalation of cations must be considered.34, 35, 44, 45

In this contribution, the possibility of using ether-based electrolytes to increase the voltage of
EDLCs is explored. Ethers are highly stable with respect to reduction and, therefore, promising
solvents for extending the negative voltage limit.40, 46-49 In particular, this report focuses on an

3
electrolyte consisting of NaPF6 in 1,2-dimethoxyethane (DME or monoglyme). While many salts
could potentially be used, NaPF6 salt was chosen because it inhibits corrosion of the aluminum
current collectors, forms electrolytes with high ionic conductivity, and has a wide
electrochemical window.50-54 DME was selected because it has a low viscosity (0.4 cP at 25
°C),55 which also promotes high electrolyte conductivity.56, 57
However, glymes with higher
molecular weight should all have a similar voltage window. Sodium carboxymethyl cellulose is
used as an electrode binder primarily for its excellent stability over a wide voltage window. 58-61
This binder also has the advantage of being water-soluble and environmentally benign. The
DME-based electrolyte shows an electrochemical window up to 3.5 V in full cells with high-
surface-area carbon electrodes. The high voltage performance could significantly increase the
overall energy density of EDLCs.

2. Materials and Methods

2.1 Electrolyte preparation:

All procedures were carried out in a glove box filled with high-purity argon (O2 and H2O < 1
ppm). Sodium hexafluorophosphate (NaPF6, 98% Sigma Aldrich) was dissolved in 1,2-
dimethoxyethane (DME, battery grade, Mitsubishi Chemical Company) at a concentration of 1
molal (m). Some control experiments were performed with a standard acetonitrile-based
electrolyte. Tetraethylammonium tetrafluoroborate (TEABF4, electrochemical grade, Sigma
Aldrich) was dissolved in acetonitrile (ACN, 99.8%, anhydrous, Sigma Aldrich) at 1 m
concentration. Both electrolytes were dried over 4A molecular sieve for several days. After
drying, the DME-based electrolyte was stored over sodium metal to remove other impurities.
Electrolyte conductivity was measured using a conductivity cell (YSI 3401) calibrated with
standard KCl solutions. Electrolyte viscosity was measured at 20 °C using a rheometer with 60
mm parallel plates and a gap of 1000 μm (AR2000ex, TA instruments). The shear rate used to
measure viscosity was 10 rad/s.

2.2 Electrode preparation:

High-surface-area carbon electrodes were prepared using 85 wt.% Black Pearls 2000 (BP2000,
Cabot Corporation, BET surface area 1500 m2/g) and 15 wt.% sodium carboxymethyl cellulose

4
(CMC, Sigma Aldrich, molecular weight 700,000). BP2000 and CMC were ultrasonicated in
water to produce a homogenous suspension. Electrodes were deposited by spray-coating onto
aluminum foil. Electrode thickness ranged from 25 to 130 µm with loadings from 0.4 to 2.2
mg/cm2.

2.3 Electrochemical testing:

Two-electrode button cells (316L stainless steel, size CR2032, Hohsen Corp. Japan) were
prepared using high-surface-area carbon electrodes and a polymer separator (Celgard 2325).
Two-electrode cells with DME-based electrolyte were fabricated using electrodes with different
loadings. The positive electrode in the DME cells had a loading of 2.2 mg/cm2 and the negative
electrode had a loading of 0.4 mg/cm2. Two-electrode cells with ACN-based electrolyte were
fabricated using electrodes with the same loading of 2.2 mg/cm2 each. The capacity of full cells
(in mAh/g) is normalized to the total mass of carbon in the capacitor including both electrodes.
In contrast, the capacitance is normalized to F/g of carbon for each individual electrode. The
normalization was calculated by treating the capacitance of the full cell (Ccell) as the series
combination of the positive and negative electrodes (Cpos and Cneg):
1 1 1
=𝐶 +𝐶 (Equation 1)
𝐶𝑐𝑒𝑙𝑙 𝑝𝑜𝑠 𝑛𝑒𝑔

We make the approximation that the capacitance is proportional to the mass loading. For
symmetric cells where the positive and negative electrodes have the same loading, C pos = Cneg
and Ccell = ½ Cpos = ½ Cneg. For asymmetric cells where the positive and negative electrodes have
different loadings Cpos ≠ Cneg. For the cells made with DME-based electrolyte Cpos = 5.5∙Cneg and
Ccell = 0.15∙Cpos = 0.85∙Cneg.

Some measurements were performed using a three-electrode cell (EL-Cell GmbH) with a high-
surface-area carbon electrode for the working electrode and sodium metal for the reference and
counter electrodes. One polymer separator and one glass fiber separator were used in these three-
electrode cells to provide sufficient electrode separation to accommodate the reference electrode.
Some cyclic voltammetry measurements were performed with glassy carbon as the working
electrode in three-electrode T-cells. T-cells were constructed using polypropylene tees. The
glassy carbon disc was freshly polished with alumina (Buehler MicroPolish) prior to use. Sodium
metal was used as counter and reference electrodes in T-cells. Potentials measured with three-

5
electrode cells are referenced to the Na/Na+ potential (ENa/Na+ ≈ +0.13 V vs. ELi/Li+). Unless stated
otherwise, 1 m NaPF6 in DME was used as the electrolyte. Cyclic voltammetry (CV),
electrochemical impedance spectroscopy (EIS), galvanostatic charge-discharge cycling, float
tests, and leakage current measurements were acquired using Bio-Logic instruments (VSP and
MPG2). All tests were conducted at room temperature.

2.4 Raman spectroscopy:

Raman spectra were collected on high-surface-area carbon electrodes before and after charge-
discharge cycling. Cycled electrodes were rinsed with pure DME and dried under vacuum prior
to analysis. All sample preparation was carried out in an argon glove box, and electrodes were
sealed under glass in a special cell to prevent air exposure during Raman analysis. Raman spectra
were acquired with an Alpha 300 confocal Raman microscope (WITec, GmbH) using a solid-
state 532 nm excitation laser, a 20x objective, and a 600 grooves per millimeter grating. The
laser spot size was approximately 1 µm, and the laser power was attenuated to 1 mW.

3. Results and Discussion

The voltage window of the electrolyte, 1 m NaPF6 in DME, was initially evaluated in three-
electrode cells with a glassy carbon working electrode (Figure 1). A CV acquired between -0.1
and 4.8 V (dashed red line in Figure 1) clearly shows the stability limits. Electrolyte oxidation
leads to a large increase in the current at potentials above 4.5 V, and sodium reduction occurs at
potentials below 0 V. A CV acquired between 0.1 and 4.5 V is also shown (solid blue line in
Figure 1) to confirm that the DME-based electrolyte is stable with respect to reduction and
oxidation over a 4.4 V window.

However, voltage window measurements on relatively inert glassy carbon electrodes are an
idealized case. Additional side-reactions are possible on electrodes that more closely resemble
commercial EDLC electrodes (high-surface-area carbons with polymer binders and aluminum
current collectors). Also, small amounts of electrolyte decomposition may passivate the
electrodes and impede the onset of bulk electrolysis in voltammetry measurements.40, 42

Therefore, the voltage window of high-surface-area carbon electrodes was measured in three-

6
electrode cells using EIS (Figure 2). The low-frequency EIS response of an ideal capacitor is a
vertical line parallel to the imaginary axis in the Nyquist plot.20 Any deviation from vertical
indicates the presence of a leakage current or undesirable side-reactions.43 EIS spectra acquired
near the positive voltage limit are shown in Figure 2a. Up to 3.9 V vs. Na/Na+ the Nyquist plots
show vertical lines at 90° with respect to the real axis. At 4.05 V the line is slightly skewed to
89°, and at even higher voltages the EIS spectra transition to a semicircular arc. The electrolyte
and electrode are stable up to 3.9 V vs. Na/Na+, but some side reaction begins at higher voltage.
The most likely side-reactions at high voltage are either corrosion of the aluminum collector or
oxidation of DME. Our EIS results cannot distinguish these two mechanisms. Results from
voltammetry with glassy carbon (Figure 1) would suggest that corrosion is responsible for the
narrower voltage window when aluminum current collectors are used. However, it is also
possible that the oxidation of DME is catalyzed by the BP2000 carbon black compared to glassy
carbon. Voltammetry of glyme-based electrolytes on platinum electrodes show the onset of
oxidation around 4.0 V vs. Li/Li+, suggesting that oxidation of the DME solvent could limit the
upper voltage on more active electrodes.62, 63

The negative voltage limit extends almost to the sodium potential as evident by EIS spectra
collected between 3.9 and 0.05 V vs. Na/Na+ (Figure 2b). At low-frequencies there is no
deviation from vertical lines over this entire voltage range, indicating no side-reactions occur
within this window. Importantly, the electrodes are not passivated even within 50 mV vs.
Na/Na+.

A minimum in the capacitance measured by EIS occurs near 3.0 V vs. Na/Na+ (Figure 3a). This
minimum, which also corresponds to the open circuit potential, reflects the potential of zero
charge of the carbon electrode.64 Moving away from the potential of zero charge, the capacitance
steadily increases, confirming the electrodes do not passivate over this voltage range. The same
trend is also seen in the cyclic voltammogram (Figure 3b). The capacitance values measured by
EIS range from 50 to 250 F/g, which is comparable to what is obtained with high surface area
electrodes in ACN-based electrolyte with three-electrode cells.32, 35

7
To verify that no side-reactions occur between 0.05 and 3.9 V vs. Na/Na+ leakage currents were
also measured in the three-electrode cells (Figure 4). The carbon electrode was held at different
potentials for 12 h while the decay in the current was monitored. The leakage currents are ≤ 1
µA/cm2 (< 20 µA/F) at both the low and high limits of the voltage window. Impedance spectra
were collected before and after the leakage current measurements to verify that the low currents
were not due to electrode passivation (data not shown). Together, the data obtained from the
three-electrode cells indicate that the NaPF6/DME electrolyte has a voltage window >3.8 V.
The voltage window was also measured in full cells with two high-surface-area carbon
electrodes and without a sodium metal reference electrode. The electrochemical window for a
full cell EDLC is very sensitive to the electrode balance (the relative mass loading of each
electrode).65-67 Since the open circuit potential of each electrode is near 3.0 V vs. Na/Na+, the
positive electrode can only be charged +0.9 V before reaching the positive voltage limit of this
electrolyte. The negative electrode must charge almost -3.0 V to reach the negative limit of the
electrolyte, which is essentially the sodium electrode potential. Therefore, the full voltage
window of the electrolyte cannot be demonstrated in a symmetric cell, where each carbon
electrode has the same loading. Since capacitance adds reciprocally for capacitors in series
(Equation 1), the specific capacitance is maximized for symmetric cells where each electrode has
the same capacitance. In asymmetric cells, the total capacitance is dominated by the electrode
with the lower capacitance. From a practical perspective, therefore, the specific capacitance and
specific energy would be highest if the open circuit potential of the carbon could be shifted to lie
near the middle of the voltage window of the electrolyte (approximately 2 V vs. Na/Na+). Efforts
are currently underway in our laboratory to accomplish this. Here, in order to demonstrate the
voltage window in full cells, we show results for cells built with different mass loadings for the
positive and negative electrodes. The mass loading of the positive electrode is approximately a
factor of five greater than the negative electrode.

The voltage window for the full cells was first checked by EIS (Figure 5). Up to 3.8 V the
Nyquist plots exhibit vertical tails, indicating the cell behaves as an ideal capacitor with no
measurable side-reactions (Figure 5a). At 4.0 V the Nyquist plot is no longer vertical,
demonstrating that the voltage window has been exceeded. The same behavior is also seen in the
Bode plots of the phase angle (Figure 5b) where the phase approaches -90° at low frequency up

8
to 3.8 V, but begins to deviate from ideal behavior at 4.0 V. The capacitance values extracted
from the EIS data are shown in Figure 5c. The specific capacitance for the full cell with DME-
based electrolyte is comparable to what can typically be achieved with ACN-based electrolytes
in similar two-electrode cells.32, 33

The long-term performance of the full cells was tested using a combination of charge-discharge
cycling and voltage holds (float tests).68 One duty cycle of the testing protocol is illustrated in
Figure 6a. For comparison, cells using a standard ACN-based electrolyte (1 m TEABF4 in ACN)
were subjected to the same tests. The cells built with the ACN electrolyte used two carbon
electrodes with the same loading for both positive and negative electrodes. The voltage window
of the ACN electrolyte is more symmetric with respect to the point of zero charge of carbon
electrodes (approximately 3.0 V vs. Na/Na+),35 and EDLCs made with ACN-based electrolyte
typically use electrodes with identical loading. As expected, the ACN-based electrolyte was very
stable when charged to 3.0 V. Figure 6b shows EIS spectra collected after the cell with ACN-
based electrolyte was held at 3.0 V for different periods. The EIS spectra remain unchanged after
up to 50 hours of floating at 3.0 V. An identical cell with ACN electrolyte charged to 3.5 V
degrades rapidly (Figure 6c). The resistance increases significantly and the capacitance
decreases. By contrast, a cell with the DME electrolyte charged to 3.5 V is much more stable
(Figure 6d).

The EIS data from Figure 6 were used to determine the effective series resistance (ESR) and
specific capacitance as a function of the total time held at the maximum voltage (float time)
(Figure 7). The ESR is measured by extrapolating the low frequency impedance (vertical line) to
the real axis, and the capacitance is also calculated from the low frequency impedance. The ESR
of cells with the DME electrolyte is initially higher than with ACN electrolyte, which is expected
based on the lower conductivity of the DME electrolyte (Figure 7a). However, no resistance
increase is measured after 50 hours of floating at 3.5 V with the DME electrolyte. In contrast, the
cell with ACN shows a rapid increase in resistance after 20 hours at 3.5 V. At first the specific
capacitance of cells with the DME electrolyte is lower than with ACN, but the specific
capacitance of the DME cell increases during the float test at 3.5 V (Figure 7b). This is in sharp
contrast to the ACN cell, which shows a steep drop in specific capacitance after 20 h at 3.5 V.

9
The origins of the capacitance rise for the DME cell are not known, but could derive simply from
better wetting of the electrolyte with time. These results clearly show that the DME electrolyte
enables cells to operate at higher voltage with similar specific capacitance compared to cells built
with ACN-based electrolytes.

Full cells with the DME electrolyte were also tested by charge-discharge cycling at constant
current between 0 and 3.5 V without voltage holds. EIS data were also collected after every 100
cycles after the cell was fully discharged (0 V cell voltage). Representative voltage profiles are
shown in Figure 8 up to 3000 cycles. The cycling stability was also evaluated from EIS and
coulombic efficiency (Figure 9). The capacitance measured by EIS (at 0 V cell voltage) increases
with cycling (Figure 9a), consistent with the results of the float tests (Figure 7b). However, the
capacity measured by charge-discharge cycling steadily decreases with cycle number (Figure 8).
The capacity fade (Figure 8) may reflect an increase in the leakage current due to electrolyte
impurities or other side-reactions that occur at high voltage, but do not occur when the cell is
fully discharged to 0 V.69 The results highlight the need to use multiple measurement techniques
to fully characterize the performance of EDLCs.

One concern is that sodium ions may intercalate into the negative electrode.35, 44, 45
If small
amounts of sodium intercalate, the process is highly reversible. The coulombic efficiency
exceeds 99% after 20 cycles and 99.9% after 500 cycles (Figure 9b). Raman spectroscopy was
also performed on electrodes after 3000 charge-discharge cycles to investigate if the carbon
structure changed due to intercalation or other side reactions (Figure 10).70, 71 No changes were
observed in the D and G bands at either the positive or negative electrode, confirming the
electrode stability.

While increasing the voltage window of the electrolyte is one approach to improve EDLC energy
density, high electrolyte conductivity is required for high power performance. Acetonitrile-based
electrolytes have conductivities > 50 mS/cm, which are unmatched by any other non-aqueous
electrolyte.16, 17 Jow and Shacklette measured the conductivity of NaPF6 in DME and found the
maximum conductivity is 13 mS/cm for a 1.5 molar solution.72 We verified that the conductivity
of the 1 molal solution used in this study is 12 mS/cm at 24 °C. The viscosity of the same

10
solution is 1.4±1 cP at 20 °C. The conductivity is similar to what can be achieved with
carbonate-based electrolytes used in lithium batteries and some commercial capacitors.17, 27, 73

The conductivity of the DME-based electrolyte is superior to many ionic liquids under
consideration for EDLC applications.30 While DME-based electrolytes are unlikely to compete
with ACN electrolytes for applications that demand very high power, they could provide a
significant improvement in energy density over today’s commercial EDLCs. DME electrolytes
could enable EDLCs with high energy density and moderate power, potentially expanding the
EDLC market.

4. Conclusions

A DME-based electrolyte shows great promise for increasing the energy density of
supercapacitors. This electrolyte clearly shows an electrochemical window of 3.8 V in three-
electrode half cells and 3.5 V in two-electrode full cells. This increased voltage window could
double the energy density compared to conventional EDLC electrolytes based on acetonitrile.
Further improvements in cycle life and energy density are anticipated by improving electrode
fabrication, optimizing the carbon pore structure, and tailoring the carbon surface chemistry.

Acknowledgement

This work was sponsored by the Oak Ridge National Laboratory Technology Innovation
Program using technology transfer license royalties. Oak Ridge National Laboratory is managed
by UT-Battelle, LLC, for the U. S. Department of Energy under Contract No. DE-AC05-
00OR22725. SC acknowledges financial support by the U.S. Department of Energy, Office of
Science, Basic Energy Sciences, Materials Sciences and Engineering Division.

References

1. B. E. Conway, Electrochemical Supercapacitors: Scientific Fundamentals and

Technological Applications, Plenum Publishers, New York (1999).
2. R. Kotz and M. Carlen, Electrochim. Acta, 45, 2483 (2000).
3. A. G. Pandolfo and A. F. Hollenkamp, J. Power Sources, 157, 11 (2006).
4. P. Simon and Y. Gogotsi, Nat. Mater., 7, 845 (2008).

11
5. Z. Li, L. Zhang, B. S. Amirkhiz, X. H. Tan, Z. W. Xu, H. L. Wang, B. C. Olsen, C.
M. B. Holt and D. Mitlin, Adv. Energy Mater., 2, 431 (2012).
6. H. Wang, Z. W. Xu, A. Kohandehghan, Z. Li, K. Cui, X. H. Tan, T. J. Stephenson, C.
K. King'ondu, C. M. B. Holt, B. C. Olsen, J. K. Tak, D. Harfield, A. O. Anyia and D.
Mitlin, ACS Nano, 7, 5131 (2013).
7. V. Subramanian, S. C. Hall, P. H. Smith and B. Rambabu, Solid State Ion., 175, 511
(2004).
8. J. W. Long, M. B. Sassin, A. E. Fischer, D. R. Rolison, A. N. Mansour, V. S.
Johnson, P. E. Stallworth and S. G. Greenbaum, J. Phys. Chem. C, 113, 17595
(2009).
9. D. Hulicova-Jurcakova, M. Seredych, G. Q. Lu, N. Kodiweera, P. E. Stallworth, S.
Greenbaum and T. J. Bandosz, Carbon, 47, 1576 (2009).
10. E. Frackowiak and F. Beguin, Carbon, 39, 937 (2001).
11. B. E. Conway and W. G. Pell, J. Solid State Electrochem., 7, 637 (2003).
12. M. Toupin, T. Brousse and D. Belanger, Chem. Mater., 16, 3184 (2004).
13. G. Pognon, T. Brousse, L. Demarconnay and D. Belanger, J. Power Sources, 196,
4117 (2011).
14. V. Augustyn, P. Simon and B. Dunn, Energy Environ. Sci., 7, 1597 (2014).
15. Z. Li, Z. W. Xu, H. L. Wang, J. Ding, B. Zahiri, C. M. B. Holt, X. H. Tan and D.
Mitlin, Energy Environ. Sci., 7, 1708 (2014).
16. H. O. House, E. Feng and N. P. Peet, J. Org. Chem., 36, 2371 (1971).
17. M. Ue, K. Ida and S. Mori, J. Electrochem. Soc., 141, 2989 (1994).
18. M. Winter and R. J. Brodd, Chem. Rev., 104, 4245 (2004).
19. F. Beguin, V. Presser, A. Balducci and E. Frackowiak, Adv. Mater., 26, 2219 (2014).
20. R. Kotz, M. Hahn and R. Gallay, J. Power Sources, 154, 550 (2006).
21. O. Bohlen, J. Kowal and D. U. Sauer, J. Power Sources, 172, 468 (2007).
22. A. Brandt and A. Balducci, J. Electrochem. Soc., 159, A2053 (2012).
23. A. Brandt, P. Isken, A. Lex-Balducci and A. Balducci, J. Power Sources, 204, 213
(2012).
24. K. Chiba, T. Ueda, Y. Yamaguchi, Y. Oki, F. Shimodate and K. Naoi, J.
Electrochem. Soc., 158, A872 (2011).
25. O. Borodin, W. Behl and T. R. Jow, J. Phys. Chem. C, 117, 8661 (2013).
26. K. Chiba, T. Ueda, Y. Yamaguchi, Y. Oki, F. Saiki and K. Naoi, J. Electrochem.
Soc., 158, A1320 (2011).
27. A. Burke, Electrochim. Acta, 53, 1083 (2007).
28. A. B. McEwen, H. L. Ngo, K. LeCompte and J. L. Goldman, J. Electrochem. Soc.,
146, 1687 (1999).
29. A. Lewandowski and M. Galinski, J. Phys. Chem. Solids, 65, 281 (2004).
30. M. Galinski, A. Lewandowski and I. Stepniak, Electrochim. Acta, 51, 5567 (2006).
31. P. Azais, L. Duclaux, P. Florian, D. Massiot, M. A. Lillo-Rodenas, A. Linares-
Solano, J. P. Peres, C. Jehoulet and F. Beguin, J. Power Sources, 171, 1046 (2007).
32. P. W. Ruch, D. Cericola, A. Foelske-Schmitz, R. Kotz and A. Wokaun, Electrochim.
Acta, 55, 4412 (2010).
33. P. W. Ruch, D. Cericola, A. Foelske, R. Kotz and A. Wokaun, Electrochim. Acta, 55,
2352 (2010).

12
34. A. M. Bittner, M. Zhu, Y. Yang, H. F. Waibel, M. Konuma, U. Starke and C. J.
Weber, J. Power Sources, 203, 262 (2012).
35. M. Hahn, O. Barbieri, R. Gallay and R. Kotz, Carbon, 44, 2523 (2006).
36. X. Zhang, R. S. Kuhnel, S. Passerini and A. Balducci, J. Solut. Chem., 44, 528
(2015).
37. R. S. Kuhnel, J. Reiter, S. Jeong, S. Passerini and A. Balducci, Electrochem.
Commun., 38, 117 (2014).
38. R. S. Kuhnel and A. Balducci, J. Power Sources, 249, 163 (2014).
39. R. S. Kuhnel, M. Lubke, M. Winter, S. Passerini and A. Balducci, J. Power Sources,
214, 178 (2012).
40. D. Aurbach, M. Daroux, P. Faguy and E. Yeager, J. Electroanal. Chem., 297, 225
(1991).
41. D. Aurbach, M. Moshkovich, Y. Cohen and A. Schechter, Langmuir, 15, 2947
(1999).
42. K. Xu, S. P. Ding and T. R. Jow, J. Electrochem. Soc., 146, 4172 (1999).
43. S. Ishimoto, Y. Asakawa, M. Shinya and K. Naoi, J. Electrochem. Soc., 156, A563
(2009).
44. D. A. Stevens and J. R. Dahn, J. Electrochem. Soc., 148, A803 (2001).
45. A. Laheaar, A. Janes and E. Lust, ECS Trans., 50, 153 (2013).
46. D. Aurbach, M. L. Daroux, P. W. Faguy and E. Yeager, J. Electrochem. Soc., 135,
1863 (1988).
47. D. Aurbach and E. Granot, Electrochim. Acta, 42, 697 (1997).
48. Z. Lu, A. Schechter, M. Moshkovich and D. Aurbach, J. Electroanal. Chem., 466,
203 (1999).
49. S. A. Campbell, C. Bowes and R. S. McMillan, J. Electroanal. Chem., 284, 195
(1990).
50. H. Yang, K. Kwon, T. M. Devine and J. W. Evans, J. Electrochem. Soc., 147, 4399
(2000).
51. S. S. Zhang and T. R. Jow, J. Power Sources, 109, 458 (2002).
52. R. Vali, A. Janes and E. Lust, J. Electrochem. Soc., 163, A851 (2016).
53. R. Vali, A. Laheaar, A. Janes and E. Lust, Electrochim. Acta, 121, 294 (2014).
54. A. Laheaar, A. Janes and E. Lust, Electrochim. Acta, 82, 309 (2012).
55. J. Barthel, R. Neueder and H. Roch, J. Chem. Eng. Data, 45, 1007 (2000).
56. K. Hayamizu, Y. Aihara, S. Arai and C. G. Martinez, J. Phys. Chem. B, 103, 519
(1999).
57. A. Jänes, J. Eskusson, T. Thomberg, T. Romann and E. Lust, J. Energy Chem., 25,
609 (2016).
58. J. Li, R. B. Lewis and J. R. Dahn, Electrochem. Solid State Lett., 10, A17 (2007).
59. J. Drofenik, M. Gaberscek, R. Dominko, F. W. Poulsen, M. Mogensen, S. Pejovnik
and J. Jamnik, Electrochim. Acta, 48, 883 (2003).
60. H. Buqa, M. Holzapfel, F. Krumeich, C. Veit and P. Novak, J. Power Sources, 161,
617 (2006).
61. J. T. Xu, S. L. Chou, Q. F. Gu, H. K. Liu and S. X. Dou, J. Power Sources, 225, 172
(2013).
62. D. Aurbach, Nonaqueous Electrochemistry, Taylor & Francis (1999).

13
63. K. Yoshida, M. Nakamura, Y. Kazue, N. Tachikawa, S. Tsuzuki, S. Seki, K. Dokko
and M. Watanabe, J. Am. Chem. Soc., 133, 13121 (2011).
64. V. Lockett, M. Horne, R. Sedev, T. Rodopoulos and J. Ralston, Phys. Chem. Chem.
Phys., 12, 12499 (2010).
65. J. H. Chae and G. Z. Chen, Electrochim. Acta, 86, 248 (2012).
66. C. Peng, S. W. Zhang, X. H. Zhou and G. Z. Chen, Energy Environ. Sci., 3, 1499
(2010).
67. S. Vaquero, J. Palma, M. Anderson and R. Marcilla, Int. J. Electrochem. Sci., 8,
10293 (2013).
68. D. Weingarth, A. Foelske-Schmitz and R. Kotz, J. Power Sources, 225, 84 (2013).
69. B. E. Conway, W. G. Pell and T. C. Liu, J. Power Sources, 65, 53 (1997).
70. J. Syzdek, M. Marcinek and R. Kostecki, J. Power Sources, 245, 739 (2014).
71. L. J. Hardwick, M. Hahn, P. Ruch, M. Holzapfel, W. Scheifele, H. Buqa, F.
Krumeich, P. Novak and R. Kotz, Electrochim. Acta, 52, 675 (2006).
72. T. R. Jow and L. W. Shacklette, J. Electrochem. Soc., 136, 1 (1989).
73. J. T. Dudley, D. P. Wilkinson, G. Thomas, R. Levae, S. Woo, H. Blom, C. Horvath,
M. W. Juzkow, B. Denis, P. Juric, P. Aghakian and J. R. Dahn, J. Power Sources, 35,
59 (1991).

14
Figures

Figure 1. Cyclic voltammograms measured with a glassy carbon electrode in 1 m NaPF6/DME

electrolyte. The dashed red line shows the scan between -0.1 and 4.8 V vs. Na/Na+. The solid
blue line shows the scan between 0.1 and 4.5 V vs. Na/Na+. The scan rate is 10 mV/s.

15
Figure 2. EIS (20 kHz – 10 mHz) of high-surface-area carbon electrodes in 1 m NaPF6/DME
electrolyte at different voltages. EIS data were measured in three-electrode cells with sodium
counter and reference electrodes. The carbon electrode was held at each voltage step for 2 h prior
to measurement. (a) Nyquist plots of the impedance response between 3.75 and 4.35 V vs.
Na/Na+. (b) Nyquist plots of the impedance response between 0.05 and 3.9 V vs. Na/Na+.

16
Figure 3. (a) Capacitance (normalized to F/g of carbon) extracted from the EIS response of a
high-surface-area carbon electrode in 1 m NaPF6/DME electrolyte at different voltages. A subset
of the EIS spectra are shown in Figure 2b. (b) Cyclic voltammogram of a high-surface-area
carbon electrode in 1 m NaPF6/DME electrolyte taken at a scan rate of 2 mV/s between 0.05 and
3.9 V vs. Na/Na+. Both EIS and CV data were measured in three-electrode cells with sodium
counter and reference electrodes.

17
Figure 4. Leakage current of high-surface-area carbon electrode in 1 m NaPF6/DME electrolyte
at different voltages. Leakage currents were measured in three-electrode cells with sodium
counter and reference electrodes.

18
Figure 5. EIS (20 kHz – 100 mHz) of a full cell (two-electrode button cell) with 1 m
NaPF6/DME electrolyte at cell voltages between 0 and 4.0 V. The cell was held at each voltage
step for 2 h prior to measurement. (a) Nyquist plot. (b) Bode plot of the phase angle. (c)
Capacitance extracted from the EIS results shown in panel a. The capacitance is normalized to
F/g of carbon for each individual electrode.

19
Figure 6. (a) Schematic of one duty cycle of the protocol used to test full cells (two-electrode
button cells). Cells were cycled four times at 1 mA/cm2 followed by a two hour hold at the
maximum voltage. Cells were then fully discharged prior to EIS analysis (20 kHz – 100 mHz).
EIS results for up to 25 consecutive duty cycles are shown corresponding to a total float time of
50 h at the maximum voltage. (b) Nyquist plots for a full cell with 1 m TEABF4/ACN electrolyte
after 10-50 hours of floating at 3.0 V. (c) Nyquist plots for a full cell with 1 m TEABF4/ACN
electrolyte after 10-40 hours of floating at 3.5 V. (d) Nyquist plots for a full cell with 1 m
NaPF6/DME electrolyte after 10-50 hours of floating at 3.5 V.

20
Figure 7. (a) Effective series resistance (ESR) of full cells subjected to the cycling protocol
shown in Figure 6a as a function of float time. (b) Capacitance of full cells subjected to the
cycling protocol shown in Figure 6a as a function of float time. Capacitance is normalized to F/g
of carbon for each individual electrode. Both specific capacitance and ESR are derived from EIS
of fully discharged cells (0 V cell voltage).

21
Figure 8. Voltage profiles for a full cell (two-electrode button cell) with 1 m NaPF6/DME
electrolyte cycled at 1 mA/cm2 between 0 and 3.5 V. The capacity is normalized to the total mass
of carbon in the capacitor including both positive and negative electrodes.

22
Figure 9. Cycling stability of a full cell (two-electrode button cell) with 1 m NaPF6/DME
electrolyte cycled at 1 mA/cm2 between 0 and 3.5 V. (a) Specific capacitance and ESR as a
function of cycle number calculated from EIS collected after every 100 cycles. EIS was collected
on fully discharged cells (0 V cell voltage). Capacitance is normalized to F/g of carbon for each
individual electrode. (b) Coulombic efficiency as a function of cycle number.

23
Figure 10. Raman spectra (D and G bands) of electrodes before and after electrochemical
cycling. Raman spectra for both the positive and negative electrode are shown after 3000 cycles.

24