3. Results and Discussion
The voltage window of the electrolyte, 1 m NaPF
6
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-
7
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
8
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/cm
2
(< 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 NaPF
6
/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
9
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 TEABF
4
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.
10
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 NaPF
6
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
11
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.
Chia sẻ với bạn bè của bạn: |