Interfacial Speciation Determines Interfacial Chemistry: X‐ray‐Induced Lithium Fluoride Formation from Water‐in‐salt Electrolytes on Solid Surfaces

Abstract Super‐concentrated “water‐in‐salt” electrolytes recently spurred resurgent interest for high energy density aqueous lithium‐ion batteries. Thermodynamic stabilization at high concentrations and kinetic barriers towards interfacial water electrolysis significantly expand the electrochemical stability window, facilitating high voltage aqueous cells. Herein we investigated LiTFSI/H2O electrolyte interfacial decomposition pathways in the “water‐in‐salt” and “salt‐in‐water” regimes using synchrotron X‐rays, which produce electrons at the solid/electrolyte interface to mimic reductive environments, and simultaneously probe the structure of surface films using X‐ray diffraction. We observed the surface‐reduction of TFSI− at super‐concentration, leading to lithium fluoride interphase formation, while precipitation of the lithium hydroxide was not observed. The mechanism behind this photoelectron‐induced reduction was revealed to be concentration‐dependent interfacial chemistry that only occurs among closely contact ion‐pairs, which constitutes the rationale behind the “water‐in‐salt” concept.


Methods
X-ray chemistry-X-ray probe experiments (XCXP) [1] experiments were performed at beamline 2-1 at the Stanford Synchrotron Radiation Lightsource (SSRL) using 17 keV X-rays. The X-ray beam size was defined by a set of slits to 300 μm vertically and 1000 μm horizontally. The X-ray flux on the sample was ≈ 10 11 photons/second, corresponding to a fluence of ≈ 3.3 e 13 photons/second/cm 2 . A Dectris Pilatus 100k area detector at 503 mm from the sample was used to collect scattered photons. The data was integrated into one-dimensional scattering patterns (intensity versus scattering vector q) using pyFAI [2] . The integrated intensity of the LiF (111) peak centered about qLiF111 = 2.702 Å -1 was extracted as follows: First, a linear background was calculated and subtracted via a line through the average intensity between q ≈ 2.650 and q ≈ 2.730 Å -1 . Subsequently, the peak area was obtained by numerical integrating from qmin = 2.66 to qmax = 2.73 Å -1 . Finally, the extracted values were normalized by the average intensity of a region on the detector that contained only background scattering.
X-ray photoelectron spectroscopy (XPS) experiments were performed using the PHI Versaprobe 1 Scanning XPS Microprobe at the Stanford Nano Shared Facility (SNSF). The X-ray source is an Al K-alpha at 1486.6 eV. The pass energy was set to 23.5 eV. Charge neutralization was used. Peak positions were shifted by -1.4 eV after normalization to the S 2p3/2 level of TFSIcentered on 169.4 eV [3] . XPS peak integration for compositional analysis was performed using a PHI MultiPak 9.8.0.19. Scanning electron microscopy energy-dispersive X-ray spectroscopy characterization was carried out using a FEI Magellan 400 XHR scanning electron microscope. Optical microscopy was conducted on a Leica DM4000M microscope. Nuclear Magnetic Resonance (NMR) spectra were recorded on a Bruker Avance 500 MHz NMR spectrometer and processed using Bruker TopSpin 4.0.1 Software. Ex-situ grazing in X-ray diffraction was performed at beamline 11-3 at the Stanford Synchrotron Radiation Lightsource (SSRL) using 12.7 keV X-rays. The X-ray beam size was defined by a set of slits to 50 μm vertically and 150 μm horizontally. A Rayonix MX225 MAR CCD area detector at 250 mm from the sample was used to collect scattered photons during 30 seconds of illumination.
Pulse radiolysis transient absorption kinetics of solvated electron decays with and without LiTFSI in a TFSI ionic liquid were measured at the BNL Laser-Electron Accelerator Facility [4] using previously described methods [5] . Samples of pure 1-butyl-1-methylpyrrolidinium TFSI (Pyrr1,4TFSI, IoLiTec) and 0.66 m LiTFSI (3M) in Pyrr1,4TFSI in septum-sealed 1 cm spectrophotometer cuvettes were purged with Ar to remove air for 30 minutes before the measurements.
For XCXP experiments, a solid-liquid interface scattering cell similar to the cell in Refs. [6] was used with a typical solution volume of 0.5 ml. The stainless steel (Grade 316, McMaster-Carr) substrates of size 4 x 30 mm were prepared by gentle polishing with ≈ 5 μm lapping film up to a mirror-like surface, and subsequently ultra-sonication in isopropanol prior to XCXP experiments.
Two aqueous electrolyte solutions were utilized. (1) 21 molal (m) LiTFSI in water; this is considered the water-in-salt (WiSE) case with 2.67 H2O per LiTFSI. (2) An ≈ 1 m aqueous LiTFSI solution prepared by mixing the 21 m solution 5:1 per volume with water to ensure no effects from possible preparation contaminations; this is considered the salt-in-water case with 55 H2O per LiTFSI.
Density functional theory (DFT) calculations were used to predict reduction potential for the TFSIanion decomposition in the highly concentrated electrolytes. The (Li2TFSI(H2O)6) + cluster was chosen a representative model with the water to Li + ratio similar to 2.67 as in 21 m LiTFSI. DFT calculations predict that this complex undergoes reduction around 2 V vs. Li/Li + either via the S-N bond breaking or defluorination and LiF formation as shown in Figure S9. After the TFSIanion defluorination, formation of the C-C bond between the TFSI(-F) -• radical anions is highly favorable as shown in Figure S10.
Gaussian 09 software was used for all QC calculations. [7] A computationally less expensive global hybrid functional M05-2X with a compact 6-31+G(d,p) basis set was used for QC studies of the larger clusters after its accuracy was examined for the for smaller complexes via comparison with G4MP2 calculations. In addition to SMD solvation model using water (ε = 78) parameters, SMD(ε = 20) implicit water model were used to estimate the influence of solvation environment on reduction potentials. [8] Figure S10 shows that a dramatic change of the dielectric constant from ε = 78 to ε = 20 lowered reduction potentials by 0. The difference between the Li + /Li and absolute reduction potential of 1.4 V was subtracted to convert results to Li + /Li scale as discussed extensively elsewhere. [9] 2. Absorbed dose Neglecting absorption by the electrolyte, the illumination corresponds to an absorbed dose by the stainless steel substrate of = ⋅ ⋅ = 3.5 kGy/second, where is the X-ray fluence of ≈ 3.3e 13 photons/second/cm 2 , the X-ray energy of 17 keV, and the mass energy absorption coefficient of ≈ 38.5 cm 2 /g for stainless steel (assumed composition is Mn2Cr17Ni12Mo3Fe66) taken from; [10] integrated over 180 minutes this corresponds to 37746 kGy. This constitutes a large cumulative dose. In comparison, solvents used in nuclear fuel reprocessing may see a dose of ~500 kGy during their useful lifetime [11] .

Experiments to test X-ray-induced LiF formation in bulk WiSE
While the bulk WiSE electrolyte significantly absorbs X-rays (1/e absorption length of ≈ 2 mm at 17 keV), we do not observe LiF formation in the bulk liquid. This was tested by exposing the liquid just above the surface to the X-ray beam for several hours, and subsequently measuring XRD of the surface below which would show LiF diffraction intensity for precipitated LiF. Attempts to investigate possible colloid formation in the irradiated WiSE solution by dynamic light scattering (after filtering with 200 nm PTFE syringe filter) suggested no presence of LiF colloids smaller than 200 nm. Experiments using unfiltered solutions were not reproducible and were hence inconclusive; we attribute this to particulate impurities in the sample, in particular after irradiation, potentially due to impurities in the X-ray cell or the formation of soluble reaction products. The unfiltered samples were filtered through a 200 nm Watman AlOx filter which was subsequently rinsed with ethanol (which solubilizes LiTFSI but not LiF). The filter was then investigated using grazing incidence X-ray diffraction (Figures S4) and XPS and no LiF/Li signal was found in the case of the irradiated WiSE solution (Figures S5 and S6). Figure 1, the intensity remains unchanged in the 1m salt-in-water case, whereas it increases steadily with the exposure time (denoted in minutes in the legend) in the 21m water-in-salt case (as indicated by the magenta arrow). Figure S2: X-ray photoelectron spectroscopy (XPS) in the Li 1s, F 1s, and N 1S spectral range of stainless-steel after x-ray exposure in 1 m LiTFSI solution. The dashed blue lines correspond to the expected peak positions of LiF in the Li 1s spectral range (55.6 eV [12] ) and F 1s spectral range (684.8 [13] ). No evidence for LiF or left-over salt after washing is observed. The peak at around 53 eV in the Li 1s spectral range corresponds to the Fe 3p peak originating in the stainless steel electrode.    [12] ) and F 1s spectral range (684.8 [13] ). The pristine and irradiated solution filtered filter show no detectable Li signal (see also Figure S6), which suggests that no radiation chemistry occurs to form LiF via reaction of an photoexcited electron with Li + and TFSIin the bulk solution. Some left over LiTFSI after washing is present on the irradiated and un-irradiated solution as evident from the peak in the N 1s spectral range. This is consistent with a decrease upon ion sputtering. The un-irradiated solution shows some evidence for Li, even after sputtering, which may be consistent with some LiF (peak after sputtering in the F 1s at around 685 eV) or interaction of LiTFSI with AlOx [14] ; subtle changes in the Al spectral ranges (not shown) are also observed but a more detailed investigation is outside the scope of this manuscript. Figure S6: Same as Figure S5 but with higher data point density and longer averaging times to yield better statistics. Li 1s and N 1s was not measured.