Anodic Oxidation of Conductive Carbon and Ethylene Carbonate in High-Voltage Li-Ion Batteries Quantified by On-Line Electrochemical Mass Spectrometry

In order to investigate the effect of electrolyte additives and water on the cell chemistry of Li-ion and Li-air batteries, a novel cell design featuring separate compartments for anode and cathode was developed (see Figure 1). This design allows to add the investigated species selectively to one compartment and to study their effect on either anode or cathode. In half-cell studies, it shields the investigated species from exposure to the strongly reducing environment of the lithium counter-electrode.

Zoom In Zoom Out Reset image size

The sealed two-compartment cell was constructed on the basis of the Li-air cell design employed by our group.³³ It consists of a 316L stainless steel (SS) cell body (part M, in Figure 1), which exhibits a step in the bottom for aligning the lithium counter-electrode (L, diameter 17 mm, thickness 0.45 mm, 99.9%, Rockwood Lithium, USA) and the separator sheet (K, glass microfiber filter, 691, VWR, Germany) or any other suitable counter-electrode. Moreover, the SS body defines a cylindrical cell volume (8.5 ml) with a thread on its inner side that hosts a SS screw (F). The two compartments are separated by a Li⁺-ion conductive glass ceramic (LICGC, diameter 1 inch, thickness 150 μm, conductivity 10⁻⁴ S/cm, Ohara, Japan) that was also employed in some of our previous work and to which we will refer to as "Ohara glass" in the following text.^28,34 The Ohara glass (I) is only permeable for Li⁺-ions and serves as a diffusion barrier for liquids and gases between the anode and the cathode compartment. In order to enhance the stability of the fragile glass ceramic and to obtain an area that can be compressed for sealing, the glass was laminated with two concentric annuli of polypropylene foil (18 mm inner diameter and 30 mm outer diameter, thickness 40 μm, Profol, Germany) in a hydraulic heated-platen press (Dr. Collin GmbH, Germany) at 150°C and 10 bar for 5 min (this sealing method had been proposed previously³⁵). Thus, the area sealed exclusively by the foil, namely between the outer diameter of the Ohara glass and the outer diameter of the foil, amounts to 2.0 cm². To achieve an effective sealing between the compartments, a virgin-PTFE O-ring (H, cord diameter 1.07 mm, inner diameter 27 mm, Angst+Pfister, Switzerland) held by a SS guide ring (G) is pressed onto the lamination foil. The pressure on the O-ring is generated by screwing the SS screw (F) into the lower cell body (M). Once the lower cell compartment is sealed, separator sheets and the working-electrode (E) can be placed onto the Ohara glass, and even large volumes of electrolyte containing water or other additives can be pipetted into the upper compartment without exposing them to the counter-electrode. A gastight sealing of the two-compartment cell is ensured by another virgin-PTFE O-ring (C, cord diameter 2.62 mm, inner diameter 30 mm, Angst+Pfister, Switzerland) positioned in the upper cell body (A). Electrical contact between the upper cell body and the working-electrode is provided through a SS spring (not shown in Figure 1, 2.3 N/mm, 0.5 inch diameter and length, Lee Springs, UK) attached to the upper cell body (A) and pressing onto a SS mesh current collector (D, diameter 21 mm, wire diameter 0.22 mm, openings 1.0 mm, Spörl KG, Germany) placed on top of the working-electrode. For electrical insulation between the upper and the lower cell body, two Kel-F insulation rings (B) are placed into the sealing notch of the upper cell body (A). The cell assembly is completed by closing the cell with four screws at a torque of 6 Nm.

A schematic representation of the cell design is shown in Figure 2, illustrating the polypropylene edge-sealed Ohara glass (I) and the O-ring (H) which provide the seal between the working-electrode (E) and the counter-electrode (L).

Zoom In Zoom Out Reset image size

For the study at hand, non-labeled ¹²C carbon black (CB) electrodes are prepared from Super C65 carbon (BET surface area 68 m²/g, TIMCAL, Switzerland). The electrode slurries were prepared by adding 50 wt% of polyvinylidene fluoride binder (PVdF, Kynar HSV 900, Arkema, France) to the carbon and mixing both components with 1-methyl-2-pyrrolidinone (NMP, anhydrous, 99.5%, Sigma-Aldrich, Germany) for 5 min at 2000 rpm and 50 mbar in a planetary orbital mixer (Thinky, USA). The obtained slurry was coated onto a C480 separator (Celgard, USA) using a 250 μm gap bar coater (RK PrintCoat Instruments, UK). After drying on a hot plate at 60°C for at least 12 h, the coating was punched out to obtain electrodes with a diameter of 15 mm. All electrodes are further dried at 95°C for at least 12 h under dynamic vacuum in a glass oven (drying oven 858, Büchi, Switzerland) and then transferred for cell assembly into an argon filled glove box (O₂ and H₂O < 0.1 ppm, MBraun, Germany), avoiding any exposure to ambient air. The average carbon loading of the electrodes was determined to be ≈1 mg_C/cm², with an average porosity of 60%.

The electrolyte was prepared with 2 M lithium perchlorate LiClO₄ (battery grade, 99.99%, trace metals basis, Sigma-Aldrich, Germany) vacuum-dried at 120°C for at least 3 days prior to mixing it with ethylene carbonate. Two types of ethylene carbonate were employed in this study: commonly used non-labeled ¹²C₃ ethylene carbonate (anhydrous, chemical purity 99%, Sigma-Aldrich, Germany), further on referred to as "¹²C₃ EC" and fully isotope labeled ¹³C₃ ethylene carbonate (isotopic purity 97 atom%, chemical purity 97%, ISOTEC by Aldrich, USA) referred to as "¹³C₃ EC". The latter was vacuum dried at room temperature for at least 3 days to remove any residual water. Upon mixing the two solid components LiClO₄ and EC, a liquid solution of 2 M LiClO₄ in EC is obtained due to freezing point depression.³¹ The as-prepared electrolyte contained < 20 ppm water as determined by Karl-Fischer titration (Titroline KF, Schott Instruments, Germany) and the amount of electrolyte added to the upper and lower compartment was 240 and 160 μl, respectively. Deionized water (18 MΩcm, Merck Millipore, Germany) was used for preparing electrolytes with 4000 ppm of H₂O (mass basis), a concentration which was chosen to mimic trace water in the battery cell, as this is an amount which could easily be introduced if, e.g., the electrodes or the separators had not been dried properly or if the cells had been assembled in ambient air, rather than in a glove box or a dry room (this may be compared to the 2000 ppm H₂O added in the experiments by Burns et al.^29,30). LiClO₄ was chosen as a salt in order to enable the addition of water without triggering any side reactions like HF formation known for fluorinated lithium salts (LiPF₆, LiBF₄, etc.)³⁶ and without leading to a partial consumption of the added water before the actual start of the experiment.^37,38 Note that in other studies examining the effect of water, LiTFSI²⁸ or LiClO₄^25,26 salt was used to avoid reaction between water and the salt; however, LiClO₄ was chosen for this study, since it does not contain any carbon atoms which would interfere with the interpretation of the CO/CO₂ signals.

After assembly and sealing in the glove box, the cell was placed into a controlled-temperature chamber held at temperatures of 10, 25, 40 or 60°C (KB 23, Binder, Germany). A crimped capillary leak (Vacuum Technology Inc., USA) connects the two-compartment cell to the mass spectrometer system (QMA 410, Pfeiffer Vacuum, Germany), permitting a constant flow of ≈1 μl/min from the cell head space (8.5 ml) to the cross-beam ionization source of the mass spectrometer. Our OEMS system is equipped with a secondary electron multiplier (SEM), allowing the detection of masses between 1 and 128 amu; for details see our previous work.³⁹ To avoid signal fluctuations due to minor pressure/temperature changes, all mass signal currents (I_Z) were normalized to the ion current of the ³⁶Ar isotope (referred to as I_Z/I₃₆). The cell was first held at OCV for 2 h; subsequently a linear sweep voltammetry (LSV) procedure was applied at a scan rate of 0.2 mV/s (Series G300 potentiostat, Gamry, USA) from OCV (≈2.8 - 3.2 V) to 5.5 V vs. Li/Li⁺. The gas evolution during this procedure was recorded by OEMS. Conversion of the mass spectrometer currents to concentrations was done for oxygen, the ¹²CO₂ and ¹³CO₂ carbon dioxide isotopes, the ¹²CO and ¹³CO carbon monoxide isotopes, and hydrogen using a calibration gas (Ar with 2000 ppm H₂, O₂, CO and CO₂, Westfalen, Germany) and considering the 1.1% natural abundance of ¹³C.

Correcting for isotopic impurity and system specific mass fractionation, the ion current signals were processed as described in the following. The isotopic purity of the ¹³C-labeled EC electrolyte was determined by measuring the ¹²CO₂/¹³CO₂ ratio of the CO₂ produced during the anodic oxidation of ¹³C₃ EC with 2 M LiClO₄ using a binder-free carbon black electrode based on ¹³C-labeled carbon (isotopic purity 99%, BET surface area 145 m²/g, Sigma-Aldrich, Germany). The resulting ¹²CO₂/¹³CO₂ ratio of 0.034 ± 0.005 corresponds to an isotopic purity of 96.6 ± 0.5%. The ¹²CO₂ signal (m/z = 44) which represents the corrosion of carbon and/or PVdF is obtained by subtracting 3% of the current at m/z = 45 (¹³CO₂ from oxidation of the labeled ¹³C₃ EC electrolyte) in order to account for the 3% isotopic ¹²C impurity in the labeled ¹³C₃ EC (isotopic purity 97 atom%, see discussion of Figure 11 for details): I_{44 (carbon+PVdF)} = I₄₄ - 0.03·I₄₅. An analogous correction was applied to obtain the ¹²CO signal (m/z = 28) derived from carbon and/or PVdF in addition to subtracting 14% of the measured m/z = 44 signal to account for the CO₂ fractionation in our mass spectrometer system: I_{28 (carbon+PVdF)} = I₂₈ - 0.03·I₂₉ - 0.14·I₄₄ (note that for the used electron energy of 70 eV, one would expect roughly 11% fractionation,⁴⁰ while the exact value for our system determined with a 10% CO₂ in Ar calibration gas was 14%). Similarly, the ¹³CO signal (m/z = 29) derived from electrolyte oxidation is corrected for the 14% mass fractionation of ¹³CO₂: I_{29 (electrolyte)} = I₂₉ - 0.14·I₄₅.

Zoom In Zoom Out Reset image size

The performance of our two-compartment cell is assessed in the following by electrochemical charge/discharge tests and by its ability to truly separate anolyte and catholyte compartment. Figure 3a compares the galvanostatic charge/discharge curves at a C-rate of 0.1 h⁻¹ of a 80/10/10 LiFePO₄/CB/PVdF working-electrode (3.3 ± 0.2 mg_LFP/cm²) vs. a lithium metal anode in the two-compartment cell and in a conventional Swagelok T-cell with LP30 electrolyte (1 M LiPF₆ in EC:DMC 1:1 wt/wt, Merck, < 20 ppm H₂O). The good agreement in specific capacity of ≈160 mAh/g_LFP, which is close to the theoretical value for LiFePO₄, demonstrates that the two-compartment cell functions properly.⁴¹ The only difference is that the voltage hysteresis between the charge and discharge plateaus is substantially larger in the two-compartment cell (≈380 mV, i.e., ≈190 mV in each direction) compared to a conventional one-compartment Swagelok cell (≈160 mV, i.e., ≈80 mV in each direction), indicating a higher internal resistance in the former. This potential difference is substantially larger than one would expect based on the difference in ohmic resistance: the areal resistance of the Ohara glass is ≈190 Ω·cm² (0.8·10⁻⁴ S/cm conductivity at a thickness of 150 μm)⁴² vs. ≈5 Ω·cm² in the conventional cell, yielding an iR-based voltage drop of only ≈20 mV for the applied current density of 0.12 mA/cm² (referenced to the smaller LFP electrode area), i.e., roughly 5 times lower than the observed ≈110 mV difference. Therefore, the observed voltage hysteresis cannot be explained by a simple iR-based voltage drop and it is most likely due to an additional interfacial resistance between the Ohara glass and the liquid electrolyte, analogous, as was proposed previously for the interface between solid Li⁺-ion conductors and liquid electrolytes.³⁵ This, however, should not significantly affect our slow potential scan experiments.

Zoom In Zoom Out Reset image size

Another test was conducted to demonstrate the good sealing between anolyte and catholyte, i.e., that liquid diffusion between the electrolyte in the upper compartment and the electrolyte in the lower compartment can be suppressed effectively by the laminated Ohara glass and the O-ring seal. This was done by deliberately adding 4000 ppm H₂O to the 2 M LiClO₄ EC electrolyte in the upper compartment, while using "dry" electrolyte (< 20 ppm H₂O) in the lower compartment. Any diffusion of water-containing electrolyte to the lithium metal electrode in the lower compartment would lead to the instantaneous formation of hydrogen according to 2Li + 2H₂O → 2LiOH + H₂↑. The comparison between the I₂/I₃₆ signals for the two-compartment cell and a one-compartment cell (identical cell design, except that it has no Ohara glass between anode and cathode) reveals that over an OCV phase of 7 h almost no hydrogen is evolved in the former (red line in Figure 3b), while substantial hydrogen evolution is observed for the latter (blue line in Figure 3b). This clearly confirms that the two-compartment cell indeed provides an excellent barrier to electrolyte diffusion between anode and cathode compartments.

For the detection of small signals in mass spectrometry, it is critical to have a low level of background signals on the m/z channels of interest. For this study, we are particularly interested in detecting the CO₂ isotopes ¹²CO₂ and ¹³CO₂ as well as the CO isotopes ¹²CO and ¹³CO, which have their main ion current signal on m/z = 44 and m/z = 45 as well as on m/z = 28 and m/z = 29, respectively. Unfortunately, most alkyl carbonate-based electrolyte formulations have a relatively high vapor pressure due to the ethyl methyl carbonate (EMC) or dimethyl carbonate (DMC) component (see Table I).

Consequently, electrolytes containing these volatile components exhibit large OEMS background signals as shown in Figure 4a and 4b for the commonly used formulations LP30 (1 M LiPF₆ in EC:DMC 1:1) and LP57 (1 M LiPF₆ in EC:EMC 3:7). For assessing the background signals of these electrolytes, we assembled two cells with 240 μl of the respective electrolyte contained in a glass fiber separator. Apart from that, the cell was empty and its head space was filled with Ar. Upon scanning for 3 h (2 h rest for signal equilibration not shown), a stable background from the gas phase components can be obtained. Both, LP30 and LP57 exhibit the largest background signals on m/z = 15, 29, 31 and 45. This is in agreement with the mass spectroscopic fractionation pattern of the DMC molecule (m/z [%signal]: 15 [100%], 45 [95%], 29 [75%], 31 [70%]) and the EMC molecule (m/z [%signal]: 45 [100%], 29 [80%], 31 [60%], 15 [20%]).⁴⁰ That we do not get an exact agreement in the relative signal strengths is probably related to the fact that the fractionation strongly depends on the electron energy (here 70 eV) and other setup specific aspects (note that the same fractionation pattern is observed for pure EMC and pure DMC solvents). In particular the background signals at m/z = 45 and m/z = 29 would be problematic for the quantification of ¹³CO₂ and ¹³CO, since small signal changes would be difficult to quantify on top of these high background signals. On the other hand, Figure 4c shows the OEMS background signals for the purely EC-based electrolyte with 2 M LiClO₄. By avoiding the volatile components DMC and EMC, the background signals from the electrolyte could be lowered by a factor of ≈100 (note the 100-fold magnified y-axis scale in Figure 4c), thus rendering this type of electrolyte particularly suitable for mass spectrometry. There is effectively no background on the channels for ¹²CO₂, ¹³CO₂, and ¹³CO and only a small background on the channel for ¹²CO, which was the reason to employ this electrolyte formulation for the subsequent experiments.

Zoom In Zoom Out Reset image size

Since the corrosion stability of conductive carbon and ethylene carbonate shall be studied in the presence of water, it is essential to first determine any background signals from a possible chemical reaction between EC and H₂O. According to Yang et al., ethylene carbonate undergoes a potential-independent H₂O-driven hydrolysis reaction to form ethylene glycol and CO₂:⁴⁵

Figure 5 shows the potential-independent CO₂ evolution from the reaction of EC with H₂O as a function of temperature, which was measured by pipetting 240 μl of 1.5 M LiClO₄ in EC containing 4000 ppm H₂O into an empty OEMS cell and holding the temperature at 10, 25, 40 and 60°C for 4 h each. The plot shows that essentially no H₂O-driven hydrolysis is observed at room temperature or below. At 40°C, a slow but steady increase of the CO₂ signal is observed, which increases substantially at 60°C, leading to a CO₂ concentration of ≈700 ppm in the cell. Consistent with the mechanism proposed by Yang et al.,⁴⁵ no gases other than CO₂ are evolved and, in the absence of H₂O, no CO₂ evolution is observed.

Zoom In Zoom Out Reset image size

The reaction rates for the different temperatures are calculated using the slope of the CO₂ mass trace and the cell volume of the standard one-compartment OEMS cell (9.5 ml). This results in H₂O-driven hydrolysis rates of 0.5·10⁻¹² mol/s and 10.9·10⁻¹² mol/s at 40 and 60°C, respectively. While the rates at 10 and 25°C were unfortunately too low to be quantified, one can still estimate an activation energy of ≈133 kJ/mol from the data at 40 and 60°C, which seems reasonable for the ring-opening reaction. Referenced to the mass of EC, the H₂O-driven hydrolysis rates at 40 and 60°C in the presence of 4000 ppm H₂O would be roughly 2.1·10⁻¹² mol/(s·g_EC) and 45·10⁻¹² mol/(s·g_EC). Assuming second order reaction kinetics and an Arrhenius-type behavior, the H₂O-driven EC hydrolysis rate could be roughly estimated for any given water content (x ppm_H2O) by:

where T is the temperature in Kelvin. While measuring the kinetics of this reaction has not been the focus of the present work, Equation 1 might be useful to understand electrolyte degradation at high temperature and potential, as water was shown to be produced at potentials above 4.6 to 4.9 V.^34,46,47 For the following anodic oxidation experiments, any CO₂ evolution from the potential-independent hydrolysis of EC will be subtracted as constant background.

Having determined the rate of H₂O-driven potential-independent EC hydrolysis, which is only noticeable at ≥ 40°C, we will now turn to examine the anodic oxidation of conductive carbon, PVdF, and ethylene carbonate, using the sealed two-compartment cell in a configuration with dry electrolyte (<  20 ppm H₂O) in the lower compartment (counter-electrode compartment) and either dry electrolyte or electrolyte with 4000 ppm H₂O in the upper compartment (working-electrode compartment). As for all of the following experiments, a Super C65 carbon PVdF-bonded working-electrode and a lithium metal counter-electrode were employed.

Starting out with a configuration where no isotopically labeled ethylene carbonate is used (i.e., ¹²C₃ EC only), Figure 6a depicts the evolution of the ¹²CO₂ (m/z = 44) and the ¹³CO₂ (m/z = 45) isotope upon an initial 2 h OCV phase and a subsequent LSV scan from OCV to 5.5 V vs. Li/Li⁺ for cells with and without water. As expected, no ¹³CO₂ signal is observed in either case (light and dark purple lines in Figure 6a). In the absence of water, no ¹²CO₂ is evolved until about 4.7 h (i.e., until ≈5 V), at which point ¹²CO₂ evolution becomes discernible and its concentration is continuously increasing (light magenta line in Figure 6a). When a 4000 ppm water-containing electrolyte is used in the upper compartment (dark magenta line in Figure 6a), there is an unexpected strong evolution of the ¹²CO₂ isotope already during the OCV period. This constant ¹²CO₂ evolution rate (indicated by its constant slope) is at first potential-independent and then increases at about the same time when the ¹²CO₂ evolution was observed in the absence of water. Having seen that the H₂O-driven hydrolysis of EC only yields negligible amounts of CO₂ at 25°C (see Figure 5), this potential-independent ¹²CO₂ evolution has to be of different origin. It should be noted, that the only gas evolved in the initial ≈4.7 h is ¹²CO₂ and H₂ (not shown, but discussed in the next section).

Zoom In Zoom Out Reset image size

In order to elucidate the origin of the unexpected constant CO₂ evolution rate during OCV and during the initial part of the potential scan, we have repeated the experiment, this time using labeled ¹³C₃ EC-based electrolyte in the upper compartment and non-labeled ¹²C₃ EC-based electrolyte in the lower compartment. In this case, we expect the evolution of ¹²CO₂ from the anodic oxidation of carbon and PVdF and ¹³CO₂ from the anodic oxidation of the electrolyte. In the absence of water, Figure 6b shows the onset of ¹²CO₂ and ¹³CO₂ evolution at again ≈4.7 h (light magenta and purple line, respectively), while no gases are evolved during OCV and the initial part of the voltage scan, as one would expect. The same holds true for the ¹³CO₂ signal in the presence of 4000 ppm water in the working-electrode compartment (dark purple line in Figure 6b), except that the onset of ¹³CO₂ evolution is somewhat earlier than in dry electrolyte and that a much stronger ¹³CO₂ evolution is observed, clearly indicating that the anodic oxidation of electrolyte is enhanced by the presence of water. What is surprising again is the potential-independent increase of the ¹²CO₂ signal when water is present in the working-electrode compartment, starting already at OCV and continuing throughout the initial part of the potential scan (dark magenta line in Figure 6b). Only later on in the scan, at ≈4.3 h (≈4.7 V), the ¹²CO₂ evolution rate accelerates, which could be consistent with the onset of the anodic oxidation of carbon and/or PVdF. However, since it is without question that carbon and PVdF (only ¹²C components in the upper working-electrode compartment) are stable at the OCV potential (≈2.8 - 3.2 V), the initial potential-independent ¹²CO₂ evolution rate must have a different cause.

At this point, the only feasible explanation is that this ¹²CO₂ signal comes from the counter-electrode compartment. While we have shown in Figure 3a that the Ohara glass seal clearly prevents the diffusion of liquid electrolyte and water between the working- and the counter-electrode compartment, the polypropylene edge-seal area nevertheless has a non-negligible permeability for water vapor, H₂, and CO₂,⁴⁸ which must be considered. It was shown recently, that water can get electrochemically reduced at potentials below 2 V,²⁸ so that water vapor diffusing to the lithium counter-electrode compartment will get reduced to H₂ and OH⁻:

Hydroxide formed at the counter-electrode is a much stronger nucleophile than water, so that the nucleophilic decomposition of ethylene carbonate by OH⁻ should clearly be a much faster reaction than the reaction with H₂O:

Further evidence for this unintended side reaction is the fact that we observe a H₂ concentration at ≈4.3 V (at the end of the potential-independent ¹²CO₂ evolution rate) which is exactly half of that of the CO₂ concentration, as would be predicted by reactions 3 and 4. The reason why the H₂ formation from the water vapor permeating through the polypropylene edge-seal was not evident in Figure 3b (red line), is the difference in scale (note that the y-axis scale in Figure 6b is ≈25-fold enlarged compared to Figure 3b). The OH⁻-driven decomposition of ethylene carbonate (reaction 4) has also been observed previously.^28,49

If the above hypothesis were correct, the ¹²CO₂ evolution rate during the OCV period and during the initial part of the voltage scan would have to be the same (or smaller) than the water permeation rate through the polypropylene edge-seal. The latter can be calculated from the polypropylene water vapor permeability of 2.5·10⁻¹² cm³_STP·cm/(s·cm²·Pa) at 25°C,^28,50 considering the edge-seal area (2.0 cm²) and thickness (8·10⁻³ cm). What is also required is the water vapor pressure, p_H2O, above the electrolyte containing 4000 ppm water (corresponding to a water mole fraction of x_H2O ≈ 0.02), which for an ideal solution could be estimated by Raoult's Law:

With these values, the water permeation rate would amount to 1.6·10⁻¹² mol/s, which is about one order of magnitude lower than the CO₂ evolution rate of 14·10⁻¹² mol/s obtained from Figure 6b. This apparent discrepancy, however, is most likely due to the inaccuracy of Raoult's Law for low x_H2O in EC, assuming that the H₂O/EC binary system behaves similar to the H₂O/PC (propylene carbonate) binary system. For the latter, p_H2O at x_H2O < 0.04 is roughly one order of magnitude larger than what would be predicted based on Raoult's Law,⁵¹ which suggests that the actual water permeation rate is indeed consistent with the CO₂ evolution rate.

Since the permeability increases with temperature with an apparent activation energy of ≈40 kJ/mol,⁵⁰ and since the water vapor pressure increases with temperature, the water permeation rate through the edge-seal will also increase with temperature. Therefore, based on the above mechanism, the ¹²CO₂ evolution rate during OCV and in the initial part of the voltage-scan should also increase with temperature. As shown in Figure 7, this is the case, and the initial ¹²CO₂ evolution rate increases with an apparent activation energy of ≈44 kJ/mol. This is a rather low activation energy for a chemical reaction and indeed confirms that the reaction is likely limited by water vapor diffusion.

Zoom In Zoom Out Reset image size

In summary, the evolution of ¹²CO₂ is an experimental artifact derived from the gas permeability of the polypropylene edge-seal. We are currently investigating gas impermeable edge-sealing concepts, but have not yet found a suitable solution. Nevertheless, owing to the perfectly constant ¹²CO₂ evolution rate (i.e., the ¹²CO₂ signal from this process following a straight line as indicated by the gray lines in Figure 7), it can be subtracted with high precision from the overall signal, yielding the potential-dependent ¹²CO₂ evolution due to carbon and/or PVdF oxidation.

In order to quantify the potential-dependent anodic oxidation of carbon, PVdF, and ethylene carbonate as a function of temperature and water content, the above discussed potential-independent signals from the H₂O-driven (see Figure 5 and reaction 1) and OH⁻-driven (see Figure 7 and reactions 3 and 4) hydrolysis of ethylene carbonate must be subtracted. This is illustrated in Figure 8 for non-labeled electrolyte (i.e., ¹²C₃ EC in both compartments), where the CO₂ signal derived from the potential-independent EC hydrolysis reaction (see gray baselines in Figure 6 and 7) was subtracted from the overall CO₂ signal (note that 14% of this signal were also subtracted from the CO signal, as explained in the Experimental Section). The resulting potential-dependent concentrations of CO, CO₂, and O₂ as well as the corresponding oxidation current in ¹²C₃ EC with 2 M LiClO₄ are shown vs. the applied potential in the absence of added water (Figure 8a) and with 4000 ppm H₂O (Figure 8b).

Zoom In Zoom Out Reset image size

For the dry electrolyte, ¹²CO (m/z = 28) and ¹²CO₂ (m/z = 44) evolution is observed for potentials > 4.8 V during the potential scan from OCV to 5.5 V vs. Li/Li⁺ at 0.2 mV/s (see Figure 8a). Since none of the working-electrode constituents are isotopically labeled in this experiment, the mass traces of the ¹³CO (m/z = 29) and ¹³CO₂ (m/z = 45) isotopes stay flat for the entire potential scan and, furthermore, one cannot differentiate between carbon, PVdF, and electrolyte corrosion. At the end of the potential scan, the concentrations of ¹²CO₂ and ¹²CO in the battery head space amount to ≈95 ppm and ≈40 ppm, respectively. In the presence of 4000 ppm H₂O, both ¹²CO₂ and ¹²CO evolution are enhanced by a factor of ≈10, yielding ≈1010 ppm and ≈345 ppm, respectively (see Figure 8b). This finding suggests that the anodic oxidation of carbon, PVdF, and/or electrolyte are substantially enhanced in the presence of water. This is also indicated by the fact that the onset of the corrosion current and of the ¹²CO/¹²CO₂ evolution is shifted negatively by ≈100 mV to ≈4.7 V. In the presence of 4000 ppm H₂O, the onset of oxygen evolution is observed at ≈5.3 V vs. Li/Li⁺, which is produced by the anodic oxidation of water on the carbon surface (2H₂O → O₂↑ + 4H⁺ + 4e⁻).

In order to distinguish between the anodic oxidation of carbon/PVdF and the electrolyte, the following experiments are now conducted with dry ¹³C-labeled EC electrolyte (¹³C₃ EC) in the working-electrode compartment. Figure 9 shows the evolution of CO and CO₂ as a function of the applied linear potential sweep from OCV to 5.5 V vs. Li/Li⁺ at different temperatures, ranging from 10 to 60°C. At 10°C, the anodic oxidation rates are so slow that hardly any gases are released when the potential of the carbon electrode is ramped up to 5.5 V. At 25°C, however, both CO₂ isotopes are clearly detected, with the formation of ¹²CO₂ marking the anodic oxidation of carbon/PVdF and with the formation of ¹³CO₂ marking the anodic oxidation of ethylene carbonate. Consequently, both carbon/PVdF and ethylene carbonate are anodically oxidized to CO₂ at 25°C once the potential reaches roughly 4.8 V vs. Li/Li⁺, even in the absence of added water. With increasing temperature, the onset potential for CO₂ evolution moves negatively (≈4.6 V at 60°C), and the CO₂ concentration increases dramatically, indicating strongly temperature activated processes. In addition to CO₂, we also observe the formation of CO from the anodic oxidation of both the carbon/PVdF (¹²CO) and the electrolyte (¹³CO), whereby the CO concentration is always minor compared to that of CO₂. The gas concentrations produced once the positive potential limit of 5.5 V is reached are shown in Table II.

Zoom In Zoom Out Reset image size

Comparing the evolution of ¹²CO₂/¹²CO from carbon/PVdF with that of ¹³CO₂/¹³CO from the electrolyte (see Figure 9 and Table II), it is apparent that the initially higher anodic oxidation rate of carbon/PVdF increases less strongly with temperature compared to that of the electrolyte, so that at 40°C and above, electrolyte oxidation is the major contributor to the overall CO₂/CO formation. This is qualitatively consistent with the study by Onuki et al.²¹ who showed that roughly two thirds of the CO₂ and CO produced during 85°C storage of a charged cell (LiNi_xCo_yAl_l-x-yO₂/graphite charged to 4.2 V using ¹³C-labeled EC/DEC electrolyte) derived from the anodic oxidation of electrolyte vs. roughly one third from non-electrolyte components.

In Figure 10, the same set of temperatures was tested with 4000 ppm H₂O in the upper compartment, containing ¹³C-labeled EC with 2 M LiClO₄. In contrast to the 10°C data in dry EC (Figure 9a), significant gas evolution is observed at this temperature in the presence of 4000 ppm H₂O (see Figure 10a). At all temperatures, the ¹²CO₂/¹²CO formation from the anodic oxidation of carbon/PVdF and the ¹³CO₂/¹³CO formation from the anodic oxidation of electrolyte are substantially enhanced in the presence of 4000 ppm H₂O (see Table II). As in the case of the dry electrolyte (Figure 9), the increase of CO₂/CO formation with temperature is less pronounced for the carbon/PVdF (¹²CO₂, ¹²CO) compared to the ethylene carbonate (¹³CO₂, ¹³CO), with the latter exceeding that of the former at 40°C and above. This will also be apparent when comparing the anodic oxidation rates at 5.0 V in the Discussion Section, after we have determined the contribution of PVdF binder to the ¹²CO₂/¹²CO gas evolution.

Zoom In Zoom Out Reset image size

As a consistency check, the gas formation at 25°C with 4000 ppm H₂O in the ¹³C-labeled EC (Figure 10b) may also be compared with those under identical conditions in the ¹²C-labeled EC (Figure 8b), whereby the amount of ¹²CO₂ + ¹³CO₂ at the end of the scan of the former (1040 ppm, see Figure 10b and Table II) should be the same as that of ¹²CO₂ at the end of the scan of the latter (1010 ppm, see Figure 8b). Reasonably good consistency is observed for the CO concentration at the end of the scan (i.e., c(¹²CO + ¹³CO) = 295 ppm in Figure 10b vs. c(¹²CO) = 345 ppm in Figure 8b).

Up to this point, we had only been able to distinguish between the anodic oxidation of carbon/PVdF and ethylene carbonate based on the evolution of ¹²CO₂/¹²CO vs. ¹³CO₂/¹³CO, respectively. In order to investigate the contribution of PVdF corrosion to the ¹²CO₂ and ¹²CO signals, the Super C65 carbon working-electrode was replaced by an electrode composed of fully ¹³C-labeled carbon (isotopic purity 99 atom%, BET surface area 145 m²/g) bonded with PVdF (33 wt% PVdF) at a comparable loading of ≈1 mg_C/cm². By using ¹³C₃ EC with 2 M LiClO₄ as electrolyte, PVdF is now the only source of ¹²C in the working-electrode compartment, apart from the ¹²C impurities in the ¹³C₃ EC (isotopic purity of 96.6 ± 0.5%; see Experimental).

Figure 11 shows the evolution of the ¹²CO₂/¹²CO and ¹³CO₂/¹³CO isotopes for a potential scan of such a PVdF-bonded ¹³C carbon electrode (against a metallic lithium counter-electrode) from OCV to 5.5 V vs. Li/Li⁺ under the most extreme conditions of 60°C and 4000 ppm H₂O in the electrolyte. As expected at these conditions (compare to Figure 10d), a strong evolution of the ¹³CO₂ signal is observed in Figure 11, originating from the anodic oxidation of the electrolyte and of the ¹³C carbon in the electrode. On the other hand, the ¹²CO₂ signal amounts to only ≈3.2% of the ¹³CO₂ signal (i.e., c(¹²CO₂)/c(¹³CO₂) = 945 ppm/29280 ppm ≈ 3.2%) and could either stem from the oxidation of PVdF to CO₂ and/or the anodic oxidation of the ¹²C impurities in the ¹³C₃ EC-based electrolyte to CO₂. The fact that this number reflects the ¹²CO₂/¹³CO₂ ratio of 3.4 ± 0.5% obtained with a binder-free ¹³C carbon pellet (corresponding to an isotopic purity of the ¹³C₃ EC of 96.6 ± 0.5%, see Experimental Section) suggests that the ¹²CO₂ signal in Figure 11 is essentially completely accounted for by the ¹²C impurities in the ¹³C₃ EC. The evolution of ¹³CO and ¹²CO is negligible compared to the strong ¹³CO₂ evolution at these conditions (compare to Figure 10d). Therefore, we can conclude that the ¹²CO₂/¹²CO signals observed in Figures 9 and 10 do not derive from the anodic oxidation of PVdF and can be exclusively attributed to the anodic oxidation of the Super C65 carbon. This is consistent with the high anodic stability of PVdF reported previously.¹⁴

Interestingly, the concentration of ¹³CO₂ in the cell head space after the LSV scan to 5.5 V is ≈2.3 times higher than for the same experiment with Super C65 carbon (viz., 29280 ppm in Figure 11 vs. 12980 ppm in Figure 10d). This ratio corresponds to the BET surface area ratio of the ¹³C-labeld carbon and that of Super C65 (145 m²/g for ¹³C carbon vs. 68 m²/g for Super C65), suggesting that the anodic carbon oxidation current is directly proportional to its BET surface area, as one would expect if both carbons have a similar degree of graphitization. The latter was indicated by Raman spectroscopy measurements, which showed similar D/G band ratios for both carbons (data not shown).

In order to put the obtained values for the gas evolution from the anodic oxidation of carbon and electrolyte into perspective with the anodic decomposition of conductive carbon and electrolyte in actual Li-ion batteries, the local slopes of the ¹²CO/¹²CO₂ and ¹³CO/¹³CO₂ concentrations vs. potential were determined at E = 5.0 V vs. Li/Li⁺. The potential of 5.0 V was chosen, since the anodic stability of battery components at this potential is relevant for the new class of 5 V cathode materials.^3–6 The slope of the CO₂ and CO concentration vs. potential for each isotope at 5.0 V was estimated by a linear regression fit between 4.975 V and 5.025 V from Figures 9 and 10, yielding the anodic oxidation rates (in units of [ppm/s]) of carbon and electrolyte into CO and CO₂. These were then converted into molar evolution rates (in [mol/s]) using the volume of the employed cell (8.5 ml). In order to quantify the overall molar anodic oxidation rates of carbon and ethylene carbonate, the molar evolution rates of CO₂ and CO were added for each isotope, yielding r_C (in units of [mol_C/s]) from ¹²CO + ¹²CO₂, and r_EC (in units of [mol_EC/s]) from ¹³CO + ¹³CO₂. Finally, since the carbon oxidation rate is proportional to the carbon mass (for the same type of carbon) and since the decomposition of electrolyte on the carbon surface is a heterogeneously catalyzed process that scales with the available electrochemically active surface area, all rates were normalized by the carbon loading of the electrodes (≈1 mg_C/cm²). The resulting anodic oxidation rates of carbon (in units of [mol_C/(s·g_C)]) and of ethylene carbonate (in units of [mol_EC/(s·g_C)]) are summarized vs. temperature and water content in Figure 12 as an Arrhenius-type plot. As will be shown later, these normalized rates allow for an estimate of the stability of Super C65 conductive carbon and ethylene carbonate in high-voltage cathodes of Li-ion batteries.

Zoom In Zoom Out Reset image size

Figure 12 shows that the corrosion rates with 4000 ppm H₂O (solid symbols) are higher than without water (open symbols) for both, carbon (magenta symbols) and electrolyte (purple symbols). Accordingly, water does not favor the corrosion of one particular component, but enhances both carbon and electrolyte corrosion. The enhanced corrosion of carbon in the presence of water can be understood from the well-known anodic oxidation mechanism of carbon in proton exchange membrane fuel cells,⁴³ which would be consistent with the enhanced formation of CO and CO₂ in the presence of water:

Based on our measurements (see Figure 10 and Table II), the ratio of CO₂ to CO from the anodic oxidation of carbon is ≈2:1, independent of temperature. We hypothesize that a similar reaction takes place in the water-enhanced anodic decomposition of EC (¹³C₃ EC + H₂O → ¹³CO₂ + ...), whereby CO₂ is found to be the main product in the absence of added water (CO₂/CO ratio > 6:1, see Table II), consistent with the observations by Onuki et al.²¹ In the presence of water, the CO₂/CO ratio increases substantially (≫ 10:1, see Table II), suggesting that water favors the reaction pathway of EC to CO₂. A mechanism for this reaction is proposed in our forthcoming article.⁴⁶

While the apparent activation energies determined from the linear regression lines in Figure 12 differ substantially for the anodic oxidation of carbon and electrolyte, they are very similar without (thin lines in Figure 12) and with added water (thick lines in Figure 12): ≈65 kJ/mol vs. ≈54 kJ/mol for carbon in dry and 4000 ppm H₂O-containing electrolyte (this may be compared to the apparent activation energy of ≈67 kJ/mol for carbon oxidation in fuel cells⁵³) and ≈104 kJ/mol vs. ≈92 kJ/mol for ethylene carbonate in dry and 4000 ppm H₂O-containing electrolyte. This finding indicates that the underlying mechanism for the anodic oxidation of carbon and ethylene carbonate is identical, no matter if water is added or not. To understand this, it is important to note that water is generally believed to be produced in the anodic oxidation of alkyl carbonate electrolytes at potentials above 4.6 to 4.9 V.^34,46,47 The observed similar apparent activation energies with and without water are thus an indication that in-situ produced water might largely be responsible for the anodic oxidation process of carbon and electrolyte. This would be consistent with the somewhat larger oxidation rate enhancement by water at the lowest temperature of 10°C (see Figure 12), since the in-situ formation through electrolyte oxidation at 5.0 V and 10°C is much smaller than at higher temperatures, which in turn leads to a stronger effect of the externally added water at low temperatures.

Using the carbon weight-normalized anodic oxidation rates shown in Figure 12 (in units of [mol_C/(s·g_C)] and multiplying them by the molecular weight of carbon (12 g/mol), the relative carbon weight loss rate (in units of [g_C/(s·g_C)] vs. time at 5.0 V can be estimated. Figure 13a shows the relative carbon weight loss over 100 h (in units of [wt%/100 h]) for 25, 40 and 60°C in electrolyte with < 20 ppm H₂O and with 4000 ppm H₂O (note that the anodic oxidation rates at 10°C were too low to be of any significance). This mimics, for example, a scenario in which a battery is cycled at a rate of 1C for 50 charge/discharge cycles with a 5 V cathode material, i.e., at a charge/discharge voltage plateau of ≈5.0 V.

Zoom In Zoom Out Reset image size

Figure 13a reflects the observed trends in Figure 12, illustrating that the carbon weight loss increases with temperature and that it is strongly amplified by water. In nominally dry electrolyte at 25°C, only ≈1% of the carbon would be oxidized at 5.0 V over 100 h (e.g., over 50 cycles at a rate of 1C), so that the loss of conductive carbon is likely to not influence the cathode performance over even 1000 h of charge/discharge cycling. On the other hand, at 60°C ≈15 wt% carbon would be oxidized in only 100 h, which is likely to lead to a noticeable loss of electrical conductivity of the cathode electrode. While there are no detailed studies in the literature, we would assume that a loss of conductive carbon on the order of 10% could lead to a severe decrease in electrode performance, particularly when contact points in the carbon network and between the carbon and the cathode active-material are affected. In the presence of 4000 ppm water, the projected life-time of the conductive carbon is reduced by another factor of 2 to 6. In consequence, elevated temperature cycling of high-voltage cathode electrodes is expected to lead to a significant loss of conductive carbon in the cathode electrode, which would be drastically increased in the presence of the intrusion of small amounts of water. The pronounced anodic oxidation of carbon at 5.0 V, particularly at high temperatures, also puts into question the viability of conductive carbon coatings on high-voltage cathode materials, as will be shown in our forthcoming publication.⁴⁶

In the following, the weight loss of electrolyte shall be calculated from the anodic oxidation rates of the ethylene carbonate based electrolyte shown in Figure 12. For this calculation, we assume: (i) that the active-material (AM) in the cathode does not have an enhanced catalytic activity for electrolyte oxidation, i.e., that its rate is simply proportional to the overall surface area in the electrode; (ii) that the total surface area of active-material is small compared to that of the conductive carbon (e.g., in a typical electrode composed of 90 wt% AM with 1 m²/g and 5 wt% Super C65 with 68 m²/g, ≈80% of the total surface area are that of the carbon); and, (iii) that the carbon weight-normalized ethylene carbonate oxidation rates given in Figure 12 therefore provide a reasonable estimate for the oxidation rate in a typical cathode electrode. According to Wagner et al.,⁵² the weight ratio of the electrolyte and the cathode active-material in a typical battery is ≈0.35/1 g_electrolyte/g_AM, which for an assumed conductive carbon content of 5 wt% results in an electrolyte to carbon ration of ≈7/1 g_electrolyte/g_C. By dividing the electrolyte oxidation rates in Figure 12 (in units of [mol_EC/(s·g_C)]) by the g_electrolyte/g_C ratio and by multiplying the latter by the molecular weight of EC (88 g/mol), one obtains the fractional electrolyte loss rate.

Figure 13b shows a projection of the obtained values to 100 h for the temperatures 25, 40 and 60°C. In the absence of water, the projected EC weight loss would be minor at 25°C (≈20 wt% over 1000 h), but quite drastic at 60°C, where more than 50 wt% of the electrolyte are projected to be oxidized after only 100 h. Thus, the operation of high-voltage cathodes at 5.0 V at elevated temperature over even 100 h is not feasible with EC-based electrolytes. In addition to electrolyte consumption, the accumulation of electrolyte degradation products on the electrode surface is likely to further reduce battery life-time at these conditions. As in the case of carbon, the weight loss of EC is strongly enhanced in the presence of water.

As a sense-check, we have also determined the anodic oxidation of ethylene carbonate at 60°C and 4.4 V, which is the typical charging potential of LiNi_xMn_yCo_zO₂ (NMC).⁵⁴ In this case, the anodic oxidation rate in < 20 ppm H₂O electrolyte obtained from Figure 9d is ≈1.1·10⁻⁸ mol_EC/(s·g_C), which under the above listed assumptions would correspond to an EC weight loss of ≈5 wt%/100 h. This is roughly one order of magnitude lower than the value at 5.0 V (see Figure 13b) and would suggest a life-time of hundreds of hours, consistent with what one would expect based on life-time data in EC-based electrolytes at 60°C and 4.4 V.^54–56