Probing a battery electrolyte drop with ambient pressure photoelectron spectroscopy

Operando ambient pressure photoelectron spectroscopy in realistic battery environments is a key development towards probing the functionality of the electrode/electrolyte interface in lithium-ion batteries that is not possible with conventional photoelectron spectroscopy. Here, we present the ambient pressure photoelectron spectroscopy characterization of a model electrolyte based on 1M bis(trifluoromethane)sulfonimide lithium salt in propylene carbonate. For the first time, we show ambient pressure photoelectron spectroscopy data of propylene carbonate in the liquid phase by using solvent vapor as the stabilizing environment. This enables us to separate effects from salt and solvent, and to characterize changes in electrolyte composition as a function of probing depth. While the bulk electrolyte meets the expected composition, clear accumulation of ionic species is found at the electrolyte surface. Our results show that it is possible to measure directly complex liquids such as battery electrolytes, which is an important accomplishment towards true operando studies.


Ambient conditions
In a previous publication, we showed that solvent evaporation occurs from an electrolyte drop in ambient conditions of N 2 1 . Therefore, we also investigated the stability of the Elect-Drop sample in ambient PC vapor environment. In Supplementary Figure 1, the spectra from the Elect-Drop samples after 10 and 70 minutes under ambient conditions of 0.2 mbar PC (835 eV excitation energy) and 2 mbar Ar (1487 eV excitation energy) are compared. Clearly, the ambient conditions of PC stabilizes the stoichiometry of the electrolyte drop well beyond the timeframe of the measurement, since the only difference between the initial and the final measurement is a slight increase in hydrocarbon species. Specifically, the salt to solvent ratio is fully preserved. However, for the ambient conditions of 2 mbar Ar, clear changes in the stoichiometry of the electrolyte are visible and the relative ratio of salt (CF 3 ) to solvent (C-O) has changed from 1:11 to 1:2 after 70 minutes. Importantly, stabilizing the solvent was only possible in 0.2 mbar PC ambient conditions, i.e. not under Ar or N 2 ambient conditions, preventing a similar comparison for the Solv-Drop sample.

Electrolyte composition -Stoichiometric calculations
To determine the ratio between the solvent to salt ratio in the probed volume of the Elect-Drop sample, the integrated intensities of the C-O component for PC (C2) and the intensity of the CF 3 component for TFSI were compared for 5 different measurement positions on the same drop. This estimation does not consider the small changes in the PC related intensities in the electrolyte (Elect-Drop sample) as compared to the solvent (Solv-Drop sample) as discussed in the main manuscript. Since these calculations are performed for peaks of the same element in the same spectra intensity variations due to the spectrometer, analyser, sample setup and cross-sections are eliminated. Thus, the only parameter necessary for the evaluation is the peak area. This value could vary slightly depending on the peak fitting if there are several overlapping peaks. Therefore, only the C2 component of PC at 287 eV was considered and not the C3 component at 290.6 eV corresponding to the carbonate moiety in PC, since the latter might be overlapping with CF x components. The C-O and CF 3 peaks are well separated from other contributions and therefore we believe the results to be reliable.

Radiation sensitivity
LiTFSI salt decomposition was observed during APPES measurements of the Elect-Drop sample. In the battery environment, LiTFSI is considered as a relatively stable salt. To follow its electrochemical stability in operando APPES experiments, it is therefore important to control radiation exposure and monitor changes if radiation damage cannot be avoided. We compared the radiation sensitivity of the Solv-Drop sample (pure PC) to the Elect-Drop sample by repeatedly measuring on the same spot. Since the carbon spectrum contains information from the solvent and salt, this region was chosen for evaluation, and the results for the Solv-Drop and the Elect-Drop samples are shown in Supplementary Figure 3 (top and middle). From these measurements we determine, that within the time of the experiment PC is more stable versus radiation, while the electrolyte shows instability during X-ray exposure. The stability of PC is seen from the constant relative intensities of the C 1s peaks. Comparing the successive C 1s measurements of the Elect-Drop sample shows instead a relative increase of the -C-O peak with increasing measurement time. In both cases, the radiation exposure lasted roughly 40 min at a photon energy of 835 eV, i.e. at the synchrotron light source. Methodologically, it is thus vital to minimize the radiation exposure time of the electrolyte and constantly monitor the spectral changes due to radiation damage. The radiation sensitivity of the LiTFSI salt most likely stems from the fluorine atoms, which generally seem sensitive 2 . For the evaluation of the salt composition as presented in Figure 2, the measurement spot was therefore repeatedly changed to minimize radiation exposure and a C 1s line was recorded first to ensure that the electrolyte had not started to degrade. It is notable, that this radiation sensitivity of the electrolyte is most pronounced at the synchrotron facility and thus seems to scale with the incoming photon intensity, since the repeated measurement of the C1s emission at 2 mbar Ar with Al Kα after 60 min exposure showed no additional peaks but merely the already described solvent evaporation (see Supplementary Figure 3, bottom).

Salt Stoichiometry in Electrolyte Drop
For the evaluation of the electrolyte composition and stoichiometry of the salt, the relative atomic ratios were calculated according to Supplementary Equation 1.
Tabulated cross sections σ i for the specific elements for an excitation energy of 800 eV were used 3 . The value for S2p was used without further polarization correction. Further correction factors such as transmission functions were not considered. As all spectra were recorded with the same photon energy, the different element specific photoelectrons will have different kinetic energies and thus different inelastic mean free paths (IMFP) λ i . These were therefore calculated for photoelectrons traveling through PC based on the TTP-2M equation implemented in the NIST IMPF database. A density of 1.204 g cm -3 and a band gap of 9 eV for PC 4 were used as input parameters. When evaluating the salt stoichiometry only peaks that could be identified as stemming from the intact salt were included. This means that some elements (N, Li) could be overestimated since possible degradation products would still appear at similar binding energies. For S, C and F distinguishable chemical environments can be seen, allowing us to omit these contributions for the calculations. For oxygen further difficulties occur to exactly determine the salt content since the peak is overlapping with the solvent peak. Keeping these difficulties in mind, we acknowledge that we can have rather large variations in the calculated ratios, also due to some uncertainties in the assumptions made to estimate σ i and λ i 3 . Therefore, we refrain from drawing conclusions for deviations on the scale up to ~50 %, but note that the lithium content exceeds all other variations by far.

LiF evaluation
In order to evaluate if Fcan act as a counter-anion for the clear intensity increase in ionized lithium at the droplet surface, the relative intensity contribution from F -(seen at a binding energy of 684.7 eV) to the Li 1s spectrum is evaluated. The intensities of the F 1s and Li 1s spectra of the Elect-Drop were firstly normalized to the corresponding CF 3 intensity, since each elemental core line was recorded individually at a fresh sample spot and the overall spectral intensity varied significantly between the different measurement spots. In the F 1s spectrum, the contributions from the TFSIand Fare clearly separated. Therefore, these normalized intensities are used together with the known atomic ratios of F to Li (6 to 1 in LiTFSI and 1 to 1 in LiF, respectively) to reversely calculate the expected Li 1s intensity contributions using Supplementary Equation 2.
From these calculations, we estimate that Fcan act as a counter-anion to roughly 70% of the total Li 1s intensity and LiTFSI to roughly 10%.