Anion Coordination Improves High ‐ Temperature Performance and Stability of NaPF6 ‐ Based Electrolytes for Supercapacitors

: Electrolyte stability can be improved by incorporating complexing agents that bind key decomposition intermediates and slow down decomposition. We show that hexamethyl ‐ phospho ‐ ramide (HMPA) extends both the thermal stability threshold of sodium hexafluorophosphate (NaPF 6 ) in dimethoxyethane (DME) electrolyte and the cycle life of double ‐ layer capacitors. HMPA forms a stable complex with PF 5 , an intermediate in PF 6 anion thermal degradation. Unbound, this intermediate leads to autocatalytic degradation of the electrolyte solution. The results of electro ‐ chemical impedance spectroscopy (EIS) and galvanostatic cycling measurements show large changes in the cell without the presence of HMPA at higher temperatures ( ≥ 60 °C). Fourier trans ‐ form infrared spectroscopy (FTIR) on the liquid and gas phase of the electrolyte shows without HMPA the formation of measurable amounts of PF 5 and HF. The complimentary results of these measurements proved the usefulness of using Lewis bases such as HMPA to inhibit the degradation of the electrolyte solution at elevated temperatures and potentially lead to improve cycle life of a nonaqueous capacitor. The results showed a large increase in capacitance retention during cycling (72% retention after 750,000 cycles). The results also provide evidence of major decomposition pro ‐ cesses (0% capacitance retention after 100,000 cycles) that take place at higher temperatures without the additive of a thermal stability additive such as HMPA. a at To discern the formation of PF


Introduction
Energy storage devices have taken on a monotony of various forms (fuel cells, capacitors, batteries, etc.) to satisfy numerous operating conditions: energy density, voltage windows, cycle rate, power density, high temperature ranges, et cetra [1][2][3][4][5]. Commercial supercapacitors or electric double-layer capacitors (EDLC) generally have lower energy densities but higher power densities compared to Li-ion batteries, 10 vs. 200 Wh kg −1 , and 10 4 vs 10 2 W kg −1 [1,[6][7][8][9][10][11], respectively. Unlike batteries, supercapacitors do not utilize the Faradaic redox process to store charge in the bulk of the materials; rather, they store charge along the electric double layer (surface) of pores within high surface area electrode materials. Double-layer charging is fast, which enables supercapacitors to have excellent power (rate) performance and long cycle life (>10 6 cycles) at the cost of poor energy density. As such, EDLCs are used for cold start assists for generators and memory back-up, as well as to deliver high power for smart grids, transport, and aerospace applications [10][11][12]. Aqueous-based EDLC are limited to 1.2 V due to the electrochemical window of water. The maximum energy stored (W) within capacitors is directly related to the specific capacitance (C) and the voltage window of the system (V) as follows: Equation (1) suggests that increasing the working voltage of the electrolyte will dramatically improve the amount of stored energy. Nonaqueous EDLC electrolytes have been developed based on acetonitrile and propylene carbonate as solvent systems. These electrolytes were shown to increase energy stored, at the expense, however, of the power [6,9,13,14]. Clearly, successful development of new electrolyte compositions needs optimization of a number of materials, as well as electrochemical properties such as working potential window, cycling rate, compatible carbon and binder systems, and other factors. Operating temperature is also an important factor as some electrolytes are unstable at temperatures above 60 °C. This instability leads to the creation of byproducts that inevitably increase the effective series resistance (ESR), reduce the operating voltage and power, and may cause cell pressure build-up. Another major issue is the corrosion of the current collector (aluminum) at higher oxidation potential and instability of the binder material, Teflon, at voltages <1 V. Simply put, these issues not only result in lower performance, device failure, and dangerous conditions, but can cause high pressure build-up, open flames, leakage of toxic chemicals, and fumes [9,15,16].
In this study, we build upon previous reports [15,28] and demonstrate that the thermal management additive, HMPA, also improves the stability of NaPF6 salt in glymebased solvents without compromising device performance. Specifically, we performed electrochemical impedance spectroscopy (EIS) to monitor the thermal decomposition (degradation) of the electrolyte by observing increases in the overall resistance of the cell and corresponding changes in capacitive behavior. We further performed Fourier transform infrared spectroscopy (FTIR) on both the solution and gas phase of the electrolyte to better understand the inhibition of electrolyte decomposition. The gas-phase IR spectra provides evidence of the formation of PF5 gas without the HMPA, but these peaks do not exist in the presence of HMPA. Our results provide evidence that HMPA can be used as an additive to increase the thermal stability of NaPF6 in DME at elevated temperatures (>60 °C). Based on these observations, we propose a simplistic pathway on how the HMPA-PF5 complex inhibits the autocatalytic decomposition reaction as reported in other systems [15,28] and improves cycle life and performance.

Materials and Methods
Electrolyte preparation: Electrolytes were prepared in an argon-filled glovebox. Before mixing, DME and HMPA were dried over a 4A molecular sieve. NaPF6 (98%, Sigma Aldrich) was dissolved into DME (battery grade, Mitsubishi Chemical Company) to give a 1 molal solution (1 mol NaPF6 /1 kg of DME). The electrolyte was stored over sodium to remove impurities for at least one week. HMPA (Sigma Aldrich) was added to the elec-trolyte solution prior to testing in specified amounts-typically 20% by weight of the solution. It is important to note that HMPA cannot be in the presence of sodium metal as it will violently react.
Electrode preparation: The carbon electrodes were prepared from ultrasonicated aqueous slurries containing Black Pearls 2000 (BP2000, Cabot Corporation, Boston, MA, USA, BET surface area: around 1460 m 2 /g) with carboxymethyl cellulose (CMC, Sigma Aldrich, St. Louis, MO, USA, molecular weight 700,000) as a binder. The electrode slurry was made by tape casting 85% BP2000 and 15% CMC onto an aluminum current collector. The electrodes were dried at 120 °C under vacuum. The weight balance between the two electrodes is 1 to 1.
Electrochemical measurements: Two electrode button cells (316 L stainless steel, CR2032, Hohsen Corp., Osaka, Japan) containing two carbon electrodes and a Celgard separator (2325) were used for all measurements. All impedance measurements were made in the frequency range of 20 kHz to 10 mHz and acquired using Bio-Logic instruments (VSP3). The tests were performed at elevated temperatures and were controlled in a Cincinnati Sub Zero temperature chamber. Galvanic cycling measurements were performed between 0 and 1 V or 3 V at 5 mA using Bio-Logic instruments (VSP3). Cyclic voltammetry was performed at 10 mV/s. Characterization: The infrared spectroscopy spectra of the solution was measured using a diamond attenuated total reflection accessory (ATR) attached to a Bruker ALPHA FTIR spectrometer within an argon-filled glovebox. Gas-phase FTIR (Cary 680 FTIR, Agilent Technologies, Santa Clara, CA, USA) spectra were measured every 3 min over 15 h inside a gas cell specifically made for studying gaseous products without interference from the ambient atmosphere as seen in Appendix A1. The cell was purged with helium and then sealed from the environment using Viton gaskets and KBr windows. The cell was sealed with KBr windows and wrapped in a heating coil and heated to 85 °C.

Results and Discussion
Cyclic voltammetry was performed to establish the electrochemical stability window of the electrolyte. Figure 1a,b shows that the voltage window is much smaller in the presence of HMPA of around 2.5 V (green curve, Figure 1b) versus the 3 V of the HMPA-free electrolyte 2.8 V (purple curve, Figure 1a). Though the voltage window is lower, cycling at 80 °C at both 1V and 3V show a remarkable difference in cycling performance. Without HMPA, the capacitance of the cell quickly drops from around 50 to ~0 F/g, which can be corroborated by the EIS measurements taken after 5000 cycles, which show significant increase in cell resistance (7 to 150 Ω at the end of the semicircle). We attribute these changes to electrolyte breakdown and the passivation of the carbon surface. A similar effect on high-voltage, high-temperature cycling is seen in the HMPA-bearing electrolytespecifically, the capacitance decays from around 50 to 35 F/g after 25,000 cycles. This shows that in addition to impeding electrolyte decomposition at 80 °C, HMPA limits the destructive effect of high-voltage cycling. The impedance measurements in Figure 1e,f suggest that HMPA prevents surface passivation as the cell impedance (the semicircle in the Nyquist plots) growth is much slower (50 Ω) than without HMPA (150 Ω). Given that cycling at 3 V induces electrochemical degradation, all measurements beyond this point were performed at 1 V and focus solely on the thermal chemical degradation of the electrolyte. Figure 1. The cyclic voltammetry of 1m NaPF6 without HMPA (a) and with HMPA (b). The galvanostatic cycling measurements of 1m NaPF6 in DME without HMPA (c) and with HMPA (d) at 5 mA at 80 °C. The spike in capacitance is an artifact from performing EIS at OCV before every 5000 cycles and normalizes after 500 cycles. The Nyquist plots of cells were taken every 5000 cycles without HMPA (e) and with HMPA (f). An increase of around 40 F/g of stable cycling can be observed with the addition of HMPA at 80 °C and cycling at 3 V.
EIS measurements were performed to verify ideal capacitive behavior in the cells for this study. The low frequency region shown in Nyquist plots should be parallel with the imaginary axis for an ideal capacitor [13,29]. Deviation from this ideality can be attributed to surface roughness, adsorption processes, and side reactions with the electrolyte [30]. With that said, near-ideal capacitive behavior in the EIS measurements was found when the cell voltage was kept between 0 and 1 V with and without HMPA- Figure 2a,b. The figures also show that the addition of HMPA did not affect the room temperature impedance except in small deviation of capacitance, though ideal capacitive behavior was observed. Room temperature galvanostatic cycling was performed to measure the cycling performance of cells containing 0% and 20% HMPA by weight. During cycling, the EIS measurements performed every 1000 cycles show no change in capacitive behavior at well over 200,000 cycles-see Figure 2c,d. The specific capacitance is retained with a 99% re-tention over 200,000 cycles as shown in Figure 2e,f. This provided evidence that the addition of HMPA does not cause a detrimental effect on the cycling performance of the electrolyte. The impedance of the cells was measured periodically during cycling (every 1000 cycles) to track changes in the capacitive behavior of the cell. The cells underwent testing at 80 °C after being equilibrated for at least 1 h. The cells were then cycled between 0 and 1 V, while an impedance measurement was taken every 5000 cycles. Without HMPA, a noticeable growth in a semicircle in the high-frequency domain indicates a possible passivation of the electrode surface. This would lead to a decrease in capacitance while cycling, which can be observed in Figure 3b. The cell is essentially dead after 60,000 cycles after a quick decrease in capacitance. The cell continued cycling for 1,000,000 cycles without any change in capacitance, but a larger growth in passivation on the surface. The addition of HMPA lead to stable cycling over 739,000 cycles at 80 °C-see Figure 3c,d. The EIS measurements also showed no significant change in impedance during cycling. There was slight fading with each cycle, but this electrode and electrolyte composition was not optimized, as it falls outside the focus of this study. These results provided evidence that the addition of HMPA led to longer cycle life at elevated temperatures. Representative impedance data with 20% by wt. HMPA at 80 °C; (d) galvanic cycling data (The spikes in the data are an artifact of how the data is taken. They occur after every EIS measurement). Note also that the initial capacitance of cells without HMPA at 80 °C vary due to unaccounted for degradation while equilibrating at 80 °C.
Visual evidence to the improved thermal stability brought on by HMPA is shown in Appendix A2. Here, the addition of 5 and 10% of HMPA keeps the electrolyte from turning dark brown after 40 h of heating at 85 °C. The mechanism for increasing the thermal stability of PF6-based electrolytes includes scavenging PF5 or acidic byproducts. An equilibrium is suspected to develop between PF6and PF5 in the solution phase [15] 15 .
This equilibrium is disrupted, and the reaction shifts towards the formation of PF5 at higher temperatures and begins the process of continual electrolyte degradation. In a typical lithium-ion battery electrolyte, PF5 reacts with impurities found within the cell, including trace water in the electrolyte solution, surface impurities of the electrode, and the solvent of the electrolyte [15]. The plausible decomposition pathway for NaPF6 in DME is shown in Figure 4. The reaction products are primarily a combination of hydrofluoric acid, POF3, fluorine containing organic compounds, and carbon dioxide. The reaction is considered autocatalytic since PF5 will also react with products from its own decomposition pathway as shown in Figure 4. The decomposition pathway shows a large combination of possible products that can be consumed to further the decomposition of the electrolyte. The decomposition products could also act as reactants for surface passivation of the carbon electrodes leading to the loss of capacitance. This phenomenon was measured in Figure 3a, which shows a large increase in resistance with continuous exposure to elevated temperatures. The reaction pathway also highlights that all components of the electrolyte solution are affected by the presence of PF5. The more important factor is that as PF5 reacts, it disrupts the equilibrium and results in more PF5 forming and further degradation of the electrolyte solution in the cell. On the contrary, HMPA binds with PF5, as shown in Figure 5, and forms a highly stable complex. The kinetics of this complexation are thought to be rapid, as PF5 is unable to induce the autocatalytic decomposition of PF6and other electrolyte species. As with many sacrificial additives, there is a limit to the amount or quantity of HMPA in the electrolyte that can suppress the formation of PF5. When all the HMPA in the system is utilized, NaPF6 thermal degradation resumes [28]. The scope of the paper is not to optimize the necessary amount of HMPA for sodium supercapacitors, but to prove the concept of using a Lewis base such as HMPA as an additive for inhibiting the thermal degradation of NaPF6. FTIR was used to track the decomposition reactions within the electrolytes with and without HMPA added-see Figure 6. The peak at 807 cm −1 corresponds to the asymmetric P-F stretching band of the PF6 -anion, and the peaks centered at about 1087 and 1110 cm −1 correspond with the C-O stretching of DME [31]. Signals from HMPA are at 740, 985, 1200, and 1295 cm −1 . The relative intensity between 807 and 1087 cm −1 drastically decreases after 24 h at 85 °C without HMPA-see Figure 6a. This decrease indicates a decomposition of PF6. The same bands show little change when 20% HMPA is present-see Figure 6b. This could be because PF5 exists as a gas at room temperature. To discern the formation of PF5, FTIR was performed on the gases evolving from the electrolyte solution at 85 °C-see Figure 7. After 7 h of exposure, new peaks at 480 and 735 cm −1 appear, which are attributed to the P-F stretching bonds of PF5 [32]. Another notable feature is the generation and consumption of HF corresponding to the peaks between 3600-4000 cm −1 -see Figure 7b. After 7 h, these peaks begin to diminish which is representative of consumption due to the previously discussed decomposition pathway. Both HF and PF5 are typical decomposition products of LiPF6 in organic solvents [33,34]. Figure 6. (a) FTIR spectra 1 M NaPF6 in DME before and after decomposition at 85 °C; (b) FTIR spectra of 1m NaPF6 in DME after exposure to 85 °C (compared to pristine 1m NaPF6 in DME). Liquid-phase FTIR shows changes in peak intensity of the PF6 P-F stretch (807 cm −1 ). Figure 7. FTIR spectra of gas phase of 1M NaPF6 in DME at 85 °C. Top spectra are without HMPA, with (a) highlighting the HF region, (b) highlighting the PF5 region during degradation, and (c) comparing the effect of HMPA addition after 7 h. NaPF6 will degrade into PF5 and HMPA suppresses PF5 formation in the gas phase.
Neither of these compounds are observed in the FTIR spectra for electrolyte solutions containing HMPA-see Figure 7c. The gas-phase FTIR provides evidence that without HMPA, the electrolyte solution degraded and formed PF5 gas. The correlation of the lack of degradation of NaPF6 in the liquid phase and the absence of the formation of PF5 in the gas phase means that HMPA complexed with PF5. This complexation led to inhibition of the degradation reaction and stability of the electrolyte at higher temperatures.

Conclusions
Electrochemical and infrared spectroscopy provided evidence that without HMPA, fluorophosphate-based electrolytes undergo rapid thermal degradation. FTIR results indicate that PF6undergoes thermal breakdown into PF5 and HF. With the addition of HMPA, EDLC cells retained their capacitive behavior and showed stable cycling performance at elevated temperatures. Here, no major changes in the FTIR spectra were observed. Our results support the notion that HMPA complexes with PF5 in a similar manner as that previously hypothesized in Li-ion systems. Similar research could be performed pairing Lewis bases with other PF6-based or Lewis acid anion salts for a myriad of applications and working temperatures. This provides one possible avenue for the development of PF6-based supercapacitors for high-temperature applications. Figure A1 shows the typical setup used for the in situ gas-phase FTIR that was used for this experiment. The sample is placed inside the cell while the different values across the top were used to outgas and replace air within the cell with inert gases such as helium and argon. Figure A2. Visual degradation of NaPF6 in DME with varying concentrations of HMPA at 85 °C. Figure A2 shows that the electrolyte leads to certain degradation unless an adequate amount of HMPA is added to the electrolyte. One noticeable feature is that a small amount of HMPA is enough to prolong the decomposition of the electrolyte.