Operando Gas Monitoring of Solid Electrolyte Interphase Reactions on Lithium

Formation of stable solid electrolyte interphases (SEI) that protect Li against continuous electrolyte reduction is one of the remaining challenges to enable safe, secondary high-energy Li batterie...


Calculating gas evolution rate from experiments
Assuming that all gases are dilute in the Ar stream and thus behave as ideal gases, then molar fractions are equal to volume fractions. Hence, the rate of formation Ṅj of a given gas j is computed from where the first ratio (Ṅj/ṄA r ) is the stream molar concentration of gas j, which is measured using gas chromatography. The volumetric flow rate of argon V̇A r is known (setpoint at ~3.05 ml/min corresponding to a mass flow rate of 5 mg/min) and V STP = 22.4 L is used as the reference ideal gas volume at standard pressure and temperature (STP) conditions.

Baseline corrections
To calculate the baseline for each gas, each cell was rested under constant Ar flow for 2 hours before polarizing the electrodes while sampling the headspace every 15 minutes. Gas evolution rates measured at the three points (i.e., ~45 mins) before polarization were then averaged. A baseline was individually calculated for each gas and for each experiment . Baselines are illustrated in Figure 2 as dashed lines.
The mean gas evolution rate of individual gases as a function of current density was calculated by averaging the gas evolution rate throughout the polarization step before the cell shorted. From these averages, the respective baseline value was subtracted. The resulting value was then the baseline-corrected mean gas evolution rate, which is shown in Figure 4.

Water contamination
Undesired water contamination from the Ar stream over time was estimated by titration gas chromatography. 1 A clean online gas analysis cell was assembled with large symmetric 30 mm (7 cm²) Li metal electrodes, 32 mm Whatman filter separator and no electrolyte. The cell was purged inside the glovebox with clean argon (99.9997%, Airgas) for 5 minutes at high pressure (~27 psig), and then purged again at 100 mg/min outside the glovebox for 15 minutes. The argon mass flow rate was then lowered to 5 mg/min. Any water contamination uptake was assumed to immediately react with the Li electrodes to form LiOH and release hydrogen in a molar proportion of 2:1 (H2O:H2). Hydrogen evolution was measured in real time and, after a transient period of ~6 h, water uptake (due to that brought in from the Ar stream along with any low-level ambient leakage of the cell) was estimated to be 1.36 ± 0.02 nmol/min, or 11 ppm of the inflow argon stream ( Figure S3), which is similar to the water content of the electrolytes. In order to estimate total water uptake by the electrolyte during a measurement, an empty cell with a Whatman separator soaked in 250 µL of EC/DMC was assembled. The cell was then brought outside of the glovebox and Ar was allowed to flow through it at 5 mg/min, mimicking experimental conditions. After ~5 h, the cell was disassembled inside the glovebox and the separator was soaked in ~3 mL of dry EC/DMC (< 1 ppm). The resulting solution was then subject to Karl-Fischer titration, which was not able to measure any water. Based on the instrument sensitivity, the electrolyte water uptake was less than ~50 ppm.      Chemical evolution of CO2.
Following a 2-hour rest, a symmetric Li||Li cell with 1 M LiPF6 in EC/DEC was galvanostatically polarized at 0.5 mA/cm²Li to a capacity of 1.5 mAh/cm²Li, during which an SEI was electrochemically formed on the electrodes. After ~12 hours, the cell was disassembled inside the glovebox, the electrodes were rinsed with 1,2-dimethoxyethane and rapidly reassembled in a clean cell with a dry separator and its gas evolution was recorded. After another ~5 hours, over which the cell rested and the gas headspace was monitored, the cell was once again disassembled and the separator was rewetted with electrolyte and reassembled again with the same electrodes, and the its gas evolution recorded. In this experiment, to facilitate cell disassembly and avoid peeling the SEI, we used a Celgard 2325 separator instead of Whatman paper. Therefore, we chose to use EC/DEC as the solvent because EC/DMC does not effectively wet Celgard 2325. However, because DEC is also a carbonate solvent that, like EC and DMC, can decompose to form ROCO2Li species, our overall conclusion that ROCO2Li species release CO2 upon extended contact with the LiPF6-containing electrolyte is not affected by whether the co-solvent is DMC or DEC. In fact, as shown in Figure 2, we observed prolonged release of CO2 with 1 M LiPF6 in EC/DMC after an initial electrochemical SEI formation, even in the absence of electrochemical activity, supporting our observations with EC/DEC in Figure S9 below.

Nuclear magnetic resonance (NMR)
NMR spectra were acquired using a Bruker Avance Neo spectrometer operating at 500.34 MHz at 298 K. 1 H spectra were acquired in 256 scans, whereas 19 F spectra were obtained with 4096 scans. 1 H chemical shifts were referenced to ethylene carbonate at 4.5 ppm, and 19 F chemical shifts were referenced to LiPF6 at -72.4 ppm. 2 Water-contaminated samples were prepared directly in an NMR tube by adding 1 mL of electrolyte or lean solvent inside the glovebox, then taking the tube outside of the glovebox to quickly add 10 μL of DI water to the solution.

Mass spectrometry (MS) of cell headspace
For GC/MS, symmetric Li||Li gas-tight cells with a headspace of ~6 mL were assembled and purged with Ar for ~5 minutes inside the glovebox. After, following a 2-hour rest, they were galvanostatically polarized at 0.5 mA/cm²Li to a capacity of 2.5 mAh/cm²Li inside the glovebox ( Figure S12a). A gas sample (~700 µL) was then extracted from the headspace using a gas-tight syringe (SGE 2.5MDF-LL-GT, fitted with a Luer Lock valve and a 27-gauge needle). Within 30 minutes of extraction, the sample was manually injected into the inlet valve of a GC instrument (Agilent 7890B) fitted with a MS detector (Agilent 5977B) capable of observing species in the 10-1000 Da range. The results were integrated over all scans, from which we obtained the electron ionization mass spectrum ( Figure S12b). The only gas-phase product at a molar mass higher than 44 was a detection at m/z = 80, which could potentially be attributed to difluoropropane (C3H6F2), possibly originating from the reduction of two FEC molecules towards lithium oxalate (Li2C2O4) upon release of CO2 (i.e., 2 FEC + 2 Li -> Li2C2O4 + CO2 + C2H6F2). Li2C2O4 had been previously proposed as existing in SEIs formed from FEC-containing electrolytes, although its origin had been attributed to reduction of CO2 (i.e., 2 CO2 + 2 Li + + 2 e --> Li2C2O4). 4 Nonetheless, the strength of the ion signal at m/z = 80 indicates that C3H6F2 is not the major gas product, corresponding to approximately 1 % of the signal intensity. Comparatively, the m/z = 44 signal, which could be attributed to either CO2 or fluoroacetylene (C2HF), displayed ~10 % intensity. We nonetheless ruled out the possibility of having formed C2HF because its mass spectrum would require an equally strong signal at m/z = 43, 5 which we did not observe. We also observed a signal at m/z = 28 (~7 %), which could be attributed to CO, C2H4 and/or N2. However, we ruled out the possibility of N2 contribution because we did not observe significant fragmentation at m/z = 14.
We also observed weak signals (~2 %) at m/z = 32 and m/z = 36, the first which could be potentially attributed to O2 and the latter, of more elusive nature, to HOF.