Impact of Fluorine‐Based Lithium Salts on SEI for All‐Solid‐State PEO‐Based Lithium Metal Batteries

LiF‐rich solid‐electrolyte‐interphase (SEI) can suppress the formation of lithium dendrites and promote the reversible operation of lithium metal batteries. Regulating the composition of naturally formed SEI is an effective strategy, while understanding the impact and role of fluorine (F)‐based Li‐salts on the SEI characteristics is unavailable. Herein, LiFSI, LiTFSI, and LiPFSI are selected to prepare solid polymer electrolytes (SPEs) with poly(ethylene oxide) and polyimide, investigating the effects of molecular size, F contents and chemical structures (F‐connecting bonds) of Li‐salts and revealing the formation of LiF in the SEI. It is shown that the F‐connecting bond is more significant than the molecular size and F element contents, and thus the performances of cells using LiPFSI are slightly better than LiTFSI and much better than LiFSI. The SPE containing LiPFSI can generate a high amount of LiF, and SPEs containing LiPFSI and LiTFSI can generate Li3N, while there is no Li3N production in the SEI for the SPE containing LiFSI. The preferential breakage bonds in LiPFSI are related to its position to Li anode, where Li‐metal as the anode is important in forming LiF, and consequently the LiPFSI reduction mechanism is proposed. This study will boost other energy storage systems beyond Li‐ion chemistries.


Introduction
The emergence of lithium-ion batteries (LIBs) has accelerated technological advancements, and LIBs have become an inseparable part of our daily lives.To extend their usage in portable electronics and electric vehicles (EVs), high power/energy density combined with long cycle stability is needed, which cannot be met by using the current LIBs.To address this challenge, the Li metal anode (theoretical specific capacity: 3860 mA h g −1 , density: 0.534 g cm −3 ), with a capacity of an order of magnitude greater than the state-of-the-art graphite anode, has attracted great interest, and the corresponding batteries are named lithium metal batteries (LMBs).However, there are challenges in using the Li metal anodes due to the high activity of Li metal and its most negative electrochemical potential (−3.04 V versus a standard hydrogen electrode), the Li metal reacts spontaneously with all polar organic solvents, leading to thermal runaway, short-circuit, and even explosion. [1]o mitigate the above-mentioned issues in LMBs, SPEs have attracted significant attention from academic and industrial sectors because of their low cost, lightweight, chemical stability, and compatibility with current large-scale manufacturing.Armand proposed poly(ethylene oxide) (PEO)-based SPEs for rechargeable batteries in 1978. [2]Since then, significant progress has been achieved, especially after Bolloré Bluecar commercialized the PEO-based SPEs in car-sharing services in 2010.4a] Among these reasons, the issue linked to SEI is a rate-control step on the Li + transport, and thus impacts the formation of lithium dendrites; furthermore, the microstructure and SEI composition will significantly influence the cycle life and Coulombic efficiency. [5]Forming a stable SEI with uniform Li + flux, wide ESW, and enough mechanical strength is essential for the PEO-based LMBs.
Two strategies have been proposed to construct a stable SEI.One is the fabrication of artificial SEI by ex-situ process, covering organic, inorganic, or organic-inorganic composite materials to the surface of Li metal. [6]The other pathway is the formation of in-situ SEI based on the decomposition of Li-salts in the electrolyte, which is viewed as one more promising strategy due to the convenient craft.Within the area of forming in-situ SEI, exploration has been conducted on Li-salt chemistry.For example, Zhang et al. [7] studied the asymmetric feature of Li-salts in Li-S batteries.They found that the electrolyte containing an asymmetric salt showed a higher content of LiF than symmetric ones, and the Li-S batteries further delivered high specific/areal capacity and good Coulombic efficiency.Lian et al. [8] developed a new "polymer-in-salt" electrolyte based on the composite Li-salts, verifying that the formation of a LiF-rich SEI can prevent the growth of lithium dendrites.Zhang et al. [9] developed a new Lisalt to regulate the LiF amount and SEI properties, endowing the NMC111//Li cell with superior cycling stability.These results indicate that the salt chemistry strongly affects the SEI characteristics, and it is crucial to forming a LiF-rich SEI to achieve efficient Li + transport and reasonable cell performance due to the low Li + diffusing energy barrier and high surface energy of LiF, [10] which strongly links to the salts from their configuration to type.However, the effect of the element F in Li-salts, including the molecular size, the amount of F elements, and chemical structures (F connecting bonds), as well as the formation mechanisms of LiF and its role have not yet been concerned, calling for future study.
Herein, three Li-salts (lithium bis(fluorosulfonyl)imide (LiFSI), lithium bis (trifluoromethylsulfonimide) (LiTFSI) lithium bis(pentafluoroethanesulfonyl)imide (LiPFSI)) were selected purposefully to investigate the effects of molecular size and configuration (the F-amount and connecting bonds) of F-based salts on the SEI characteristics and the LiF formation mechanism.To prepare the SPEs, PEO was selected as the polymer host due to its high solvation power of ethylene oxide (−CH 2 CH 2 O-, -EO-) unit and the industrial maturity, while polyimide (PI) was used as the substrate to enhance the overall mechanical strength. [11]When exploring the effects of salts on the battery performances, the Li//Li symmetric cells and coin cells with different cathodes (low-voltage LiFePO 4 (LFP) and high-voltage cathodes (LiCoO 2 (LCO), LiNi 0.5 Co 0.2 Mn 0.3 O 2 (NCM 523)) were assembled and studied.The Li + flux related to the dendrite suppression ability was tracked by COMSOL simulations, and the composition and LiF content in the formed SEI were detected by X-ray photoelectron spectra (XPS) and time-of-flight secondary ion mass spectrometry (ToF-SIMS).The LiF generation process in the SEI for the SPE with LiPFSI was proposed based on the ab-initio molecular dynamics (AIMD) simulation results.

Results and Discussion
To systematically investigate the effect of F-based Li-salts on the properties of SPEs and their performance in LMBs, firstly, the properties of SPEs were characterized.Following the identified ratio of PEO to Li-salt (18:1) (Figure S1 and Table S1, Supporting Information), the SPEs containing PEO and Li-salt (LiFSI, LiTFSI, or LiPFSI) on the PI substrate were denoted as PPF, PPT, and PPP, respectively.The SPE containing PEO on the PI substrate without Li-salt was marked as PPI for comparison.Then, the Li + plating-stripping in Li//Li cells and the electrochemi-cal properties in practical cells, including LFP//Li, LCO//Li, and NCM523//Li, were tested.After that, the SEI composition and morphology were characterized by XPS, ToF-SIMS, and scanning electron microscope (SEM).Furthermore, the formation process of LiF in the SEI for PPP was tracked by the AIMD simulation, and the corresponding LiPFSI decomposition mechanism was proposed.

Characterization
The chemical structures of LiFSI, LiTFSI, and LiPFSI are shown in Figure 1a, and their anion radii follow the sequence of FSI − <TFSI − <PFSI − .Obviously, all the anions are composed of nitrogen (N) with strong electronegativity; while for LiTFSI and LiPFSI, the two sulfur (S) atoms are connected with the strong electron-attracting groups (CF 3 ) (i.e., S-C-F), being different from the F in LiFSI (i.e., S-F).The structure with the CF 3 connection can disperse negative charges, allowing the salt to be dissociated easily.Furthermore, the contents of the F element in each molecule are various, which are 2F for LiFSI, 6F for LiTFSI, and 10F for LiPFSI. 7Li NMR spectra provide information about the local chemical environment of the Li + , and the corresponding spectra for LiFSI, LiTFSI, and LiPFSI are shown in Figure S2 (Supporting Information).In the case of LiFSI, the spectrum shows three distinct peaks from −70 to 60 ppm.However, for LiTFSI and LiPFSI, two additional peaks are observed at ≈−129 and 110 ppm, which are ascribed to the presence of the -CFgroup. [12]This suggests that the chemical environments of Li + in LiTFSI and LiPFSI are similar but different from that in LiFSI, which indicates that the connecting bond plays an essential role in determining the properties of the prepared SPEs.
The SPEs containing the chosen salts (LiFSI, LiTFSI, LiPFSI) were characterized.Firstly, the microstructure of free-standing SPE was characterized by SEM (Figure S3, Supporting Information).It is shown that the SPE surface is dense and smooth, with a thickness of 59 μm.The 7 Li NMR spectra of these SPEs follow the characteristic peaks of Li-salts, while their positions were broadened and shifted slightly, as illustrated in Figure 1b.More specifically, the prominent EO peak of PPI is located at −3.97 ppm, while for PPF, PPT, and PPP, the main peaks are shifted to −3.87, −3.85, and −3.81 ppm, respectively.The more positive positions of PPP and PPT, compared to PPF, indicate that LiPFSI and LiTFSI possess a stronger interaction with PEO than LiFSI. [13]In addition, there are no peaks for Li-salt because of the relatively low content.The enlarged figure of the dotted box is depicted in Figure S2b (Supporting Information), showing similar positions for the tiny peaks of PPP and PPT, which again demonstrates their comparable chemical environment.
The SPEs were further characterized.Their Raman spectra (Figure 1c) were divided into three peaks, corresponding to the free anion and ion pair (TFSI − coordinated with Li + (Li-TFSI) and TFSI − coordinated with PEO), respectively (Figure S4, Supporting Information), and relatively quantitative results on salt decomposition were thus obtained.It is shown that, compared to PPF, the integral intensity of the free anion peak for PPP and PPT is weaker, suggesting that more anions in PPP and PPT are immobilized within the polymer frameworks, leading to reduced anion transport and improved Li + transference number. [14]he XRD patterns of SPEs with Li-salts in Figure 1d only show two characteristic peaks at 2 = 19.3 and 23.7, both belong to the crystalline phase of PEO, proving the complete dissolution of Li-salt in PEO via the complexion of Li + with the strong electron-donating EO units and a favorable entropy factor. [7]The relative crystallinities were calculated, which are 31.9%,30.6%, 23.8%, and 18.8%, respectively, for PPI, PPF, PPT, and PPP, verifying that adding Li-salts can decrease the crystallinity of PEO and the effects of LiTFSI and LiPFSI are more significant compared to LiFSI.
The TG curves in Figure 1e show that the decomposition temperature of PEO is ≈341.0°C, indicating a good thermal tolerance.The decomposition temperatures of PPT and PPP are comparable with pristine PEO, but that of PPF drops to 236.0 °C, owing to the low decomposition temperatures of LiFSI originating from the low bond energy of S-F.The results in thermal stability are consistent with those from the Raman spectra and XRD curves, suggesting that LiFSI is not as effective as LiTFSI and LiPFSI.
A free-standing SPE with high mechanical strength is highly desirable to achieve exceptional compatibility with the electrodes.Figure 1f and Table S2 (Supporting Information) summarize the mechanical strength of each SPE together with PPI.For PPI, its tensile strength (s) is 8.51 MPa, and the elongation at break ( b ) reaches 803.60%.As usual, the addition of Lisalt weakens the mechanical strength of PPI, but the prepared SPEs still maintain acceptable mechanical strengths.In addition, the toughness was calculated to study the strength and ductility of SPEs quantitatively.The corresponding values for PPF, PPT, and PPP are 0.11×10 4 , 0.13×10 4 , and 0.35×10 4 kJ m −4 , respectively.Since high strength favors the suppression of Li dendrite growth and high strain is required for large-scale fabrication, PPP is considered as a desirable candidate from this aspect.
According to the characterization in this section, the SPEs were successfully prepared, and, in general, adding Li-salts is beneficial to develop effective SPEs.According to the results for different salts, PPP and PPT are considered better candidates than PPF, indicating that the anion configuration, especially the bonds (i.e., S-C-F in this work), can greatly affect the physical properties of SPE, further resulting in differences in electrochemical performances.

Electrochemical Properties and Performances
To clarify the effect of Li-salt anion configuration on the electrolyte properties and cell performances, the electrochemical properties of SPEs, not only in the Li//Li symmetric cells but also in full LMBs with both low and high-voltage cathodes, were tested.Meanwhile, a theoretical study was conducted to explain the experimental observations and to clarify the mechanism of LiF formation.

Electrochemical Properties
The lowest unoccupied molecular orbital (LUMO) and highest occupied molecular orbital (HOMO) levels of PEO, PI, LiFSI, LiTFSI, and LiPFSI were calculated via density functional theory (DFT) to assess the components stability of SPEs, i.e., antioxidation/reduction ability.As shown in Figure 2, all Li-salts exhibit lower HOMO energy levels than PEO, and among the studied three salts, the HOMO energy levels of LiPFSI and LiTFSI are lower than that of LiFSI, indicating their higher antioxidation ability derived from the strong electron withdrawing effect of CF 3 .This means that adding these Li-salts can improve the compatibility of PEO with high-voltage cathodes, and PPP and PPT are superior than PPF. [10]Also, LiTFSI and LiPFSI have a lower LUMO energy level than LiFSI and can thus be The linear sweeping voltammetry (LSV) profiles were determined to test the oxidative stability of SPEs experimentally as presented in Figure 3a.PPF shows a low oxidation potential, as there is a rapid increase in current at about 3.8 V, while for PPT and PPP, their oxidation occurs at a potential higher than 3.8 V. Therefore, the oxidative potentials follow the order of PPP>PPT>PPF as V PPP > V PPT > V PPF .Based on the above theoretical and experimental results, in principle, all these SPEs are electrochemically stable and can be coupled with high-voltage cathode materials (such as NCM and LCO). [15]Furthermore, the oxidative potentials of PPP and PPT are higher than PPF, and these resultsagree with the HOMO values shown in Figure 2.
The temperature-dependent ionic conductivities for SPEs were recorded using the alternating current impedance method.As shown in Figure S5 (Supporting Information), the addition of Lisalt has a positive effect on the ionic conductivity over the whole temperature range from 25 to 65 °C.The ionic conductivities for PPF, PPT, and PPP are 0.08, 0.22, and 0.27 mS cm −1 at 60 °C, respectively (Figure 3b).Combining the results here and the features of the studied salts, the connecting bond of S-C-F in the studied Li-salts possesses higher ionic conductivity than the bond of S-F.
Lithium transference number (t Li + ) is an essential factor to evaluate the mobility of Li + , since the moving of cations exerts a great effect on the cell performance.As shown in Figure 3c and Figure S6 (Supporting Information), the calculated results of t Li + for PPP, PPT, and PPF are 0.53, 0.50, and 0.49, respectively, being consistent with the observations from the Raman spectra, i.e., more anions in PPP and PPT are immobilized in the polymer frameworks, and thus PPP and PPT present lower anion migration ability and higher Li + transference number than PPF.

Cell Electrochemical Performances
To clarify the effect of anion configuration in the Li-salts on the electrochemical performance, Li/SPEs/Li cells were assembled and their plating/stripping performance was tested.As presented in Figures 3d,e, at a low current density of 0.1 mAh cm −2 , the cells using PPP and PPT demonstrate an overpotential of 32 mV within 2500 h, while for the cell using PPF, although its overpotential is similar to PPP and PPT, it occurs short circuit at ≈2200 h  (Figure 3d).Furthermore, the plating/stripping was carried out from 0.1 to 0.3 mAh cm −2 to explore the actual performance.As shown in Figure 3e, the overpotential of the cell using PPF drops sharply at 1900 h, evidencing a short circuit due to the dendrite growth.In contrast, the cells using PPP and PPT exhibit stable cycling until 2500 h.All these results indicate that PPP and PPT can favor the formation of a robust SEI that can suppress dendrite growth. [16]o further study the effects of anion configuration in the Lisalts on the cell performances, all solid-state LMBs with three representative cathodes from low to high voltage were assembled and tested at 60 °C.As shown in Figure 4a, for the cell with a low-voltage cathode (LFP), the one using PPF took 20 cycles to achieve a uniform Li + dispersion, enhancing the contact with LFP, [17] while once reaching equilibrium, its capacity decays at an extremely fast rate, and only 30 mAh g −1 remains after 200 cycles.In contrast, the cell with PPP exhibits excellent cycling performance with a capacity retention of 73% after 400 cycles and obtains an average Coulombic efficiency of ≈99%.However, the cell using PPF only exhibits 61.3 mAh g −1 after 400 cycles, and such a capacity is almost half of the cell with PPP.The cell performance using PPT is worse than that using PPP but much better than the one with PPF.The corresponding GCD (Galvanostatic Charge-Discharge) curves for these three cells at 200 cycles are shown in Figure 4b.Obviously, the discharge platforms for the cells using PPP and PPT are analogous, ≈3.38 V, which is higher than that using PPF (3.33 V).In addition, all these cells can work at a high rate of 1 C (Figure 4c), and the cells using PPP and PPT can deliver a high-capacity retention of ≈85%, which is much higher than that with PPF (60%).
These three SPEs were assembled with two representative high-voltage cathodes (LCO and NCM 523) to further investigate their compatible capacity.The test potential windows were set to be 3.0-4.3V and 2.8-4.3V for using LCO and NCM 523 cathodes, respectively, as shown in Figures 4d,e.The performance of the cells follows the order of PPP > PPT > PPF.More specifically, for the LCO//Li cells, the corresponding capacities are 148, 140, and 130 mAh g −1 and their retentions are 63%, 57%, and 55% after 45 cycles for the cells using PPP, PPT, and PPF, respectively (Figure 4d); while for the NCM523//Li cells, their capacities are 125, 112, and 104 mAh g −1 and retentions maintain 82%, 90%, and 32% after 50 cycles, respectively (Figure 4e).The corresponding GCD curves (Figures 4f,g) at 20 cycles also indicate that the cells using PPP and PPT have a higher discharge platform than that using PPF.
In summary, the electrochemical properties and performance of the SPE and cell using PPP are slightly better than or similar to PPT, which is much better than PPF.Combining the observation in this section with the features of three studied salts, we can state that the anion configuration of the Li-salts, especially the connecting bond (S-F or S-C-F), plays a significant role in electrochemical properties and performances.

Mechanism Study
To explore the mechanism underlying the electrochemical and cell performances, the ex-situ SEM of Li metal after cycling was obtained, as shown in Figures 5a-c.It exhibits that the Li metal surface in Li/PPP/Li and Li/PPT/Li cells is uniform without Li dendrites.Relatively, the surface of Li metal using PPF as the SPE is random and rough, demonstrating that the cells using PPP and PPT can suppress Li dendrites, leading to highly reversible Li plating and stripping.
In parallel, the COMSOL simulation was executed to clarify the Li + flux through the anode/SPE interface.Figures 5d-f shows the electric field variations at different times.At the initial stage of Li + plating, the most active reaction area in these SPEs is the tip protrusion, which possesses the highest current density.However, when the time lasts for 10 s, the cells using PPP and PPT show homogeneous Li + transport, implying that they can homogenize Li + flux and inhibit the high reaction activity at the protrusion.However, the cell using PPF shows inhomogeneous Li + transport and exhibits the highest current density, which is consistent with the obtained SEM images, i.e., the surface of Li metal is observed to be random and rough.We assume that the inhomogeneous Li + transport for the cell using PPF may be due to the lack of some critical components in the constructed SEI layer, leading to the SEI exhibiting a high exchange current density and surface energy, increasing the lithium nucleation density and promoting the continuous growth of small Li nuclei. [18]To prove the above hypothesis and identify the specific critical components, ex-situ XPS was conducted in the following section.
As the cell performance is related closely to the interfacial characteristics, XPS analysis was conducted to compare the SEI compositions in Li//Li cells after cycling and analyze the effect of Lisalts.In the experiments, the XPS spectra of C 1s, O 1s, F 1s, S 2p, and N 1s of PPF, PPT, and PPP before and after sputtering 60 s were obtained (Figure S7, Supporting Information) to identify the compositions and mainly quantify the LiF content.The detailed explanation of peaks was described in supporting materials together with Figure S8 (Supporting Information).According to the detailed analysis, two observations can be obtained, as described in the following text.
1) there are some inorganic components, such as LiF, Li 2 O, Li 2 CO 3 , Li 2 S, and Li 3 N in the inner SEI when using PPP and PPT; in contrast, no Li 3 N is formed in the inner SEI when using PPF.This observation can be explained by the salt decomposition mechanism linked to the apparent electron transfer number.Based on the study by Zhang's group, PPF is inclined to undergo mostly four electron reactions that generate LiF and Li 2 O, whereas the other reaction pathways with 8-16 electron transfer that generate other fragments like Li 3 N or Li 2 S take place less frequently. [19]Additionally, in the work by Budi, [20] the XPS results also confirmed the non-existence of Li 3 N in the SEI for PPF.Based on this observation, the decomposition mechanism of these three salts was proposed, as shown in Figure 6a.Actually, besides LiF, other inorganic and Li 2 CO 3 , [21] and Zhang et al. stated higher ionic conductivities (exceeding 10 −4 S cm −1 ) and higher Li + mobilities of Li 2 S and Li 3 N than LiF. [22]Therefore, all these inorganic species are advantageous and contribute to the electrochemical stability of the formed SEI, enabling the long-term cycling of Li symmetric cells and LMBs.For example, the Li//Li cells using PPP and PPT can achieve a stable cycle for 2500 h, and the corresponding LMBs can reserve a capacity greater than 70% after 400 cycles.This observation proves the hypothesis, as mentioned earlier, that the lack of Li 3 N in PPF is one of the reasons for the worse performance of the assembled cells compared to those using PPP and PPT. 2) the weight percentages of the F atom in the SEI after sputtering are increased, while the increasing degree depends on the used salts.Specifically, the F atom contents increase from 12.2% to 16.2%, 14.7% to 22.1%, and 13.0% to 24.0% when using PPF, PPT, and PPP, respectively (Figures 6b-d); i.e., for PPT and PPP, their weight percentages of the F atom in the SEI after the sputtering are similar and much higher than PPF.This result indicates the importance of forming a LiF-rich SEI and its role in the Li + transport and cell performances.
To further investigate the SEI composition in the Li/PPP/Li cell, the distribution of ion fragments was characterized by ToF-SIMS.The 2D surface mappings under negative mode in Figure S9 (Supporting Information) prove the uniform distributions of organic fragments of CH − , CF 3 SO 2 N − , and inorganic fragments of LiF 2 − , Li 2 F 3 − , LiPFO 2 − , C 2 H − , LiCO 3 − , CH − , CF 3 SO 2 N − , and LiO − in the SEI after cycling.Impressively, the intensity of the LiF 2 − and Li 2 F 3 − fragments at the early stage of ToF-SIMS depth profiles is much higher than those of organic fragments (Figure 6e), indicating a relatively high ratio of LiF in the inner SEI. [23]Besides, the corresponding 3D mappings in Figure 6f further indicate the higher intensity of LiF contents throughout the sputtering depth than those of organic/inorganic species.This reveals that LiF is the principal component of SEI, and it contributes to homogenized Li + transport and protects the Li anode during cycling, leading to solid-state Li batteries with high reversibility, rate capability, and long cycle time.For example, the LMB using PPP can retain 83% capability from 0.1 C to 1 C. [10,24]

Molecular Dynamic Simulation of Bond Cleavage
The decomposition mechanisms of LiFSI and LiTFSI to generate LiF and other inorganic species have been well established, [25] while, it is unclear for LiPFSI, to the best of our knowledge.Thus, in this work, the decomposition steps for LiPFSI at different positions in the SPE system to form LiF were traced through the AIMD simulations.The initial structure of the AIMD simulation is shown in Figure 7a, which contains five layers of Li atoms, four LiPFSI molecules (I, II, III, IV) with different distances from the Li metal anode, one PI molecule, and ten PEO molecules.It should be noted that the lower three layers of Li atoms were fixed during the simulation, and the MD trajectories were monitored.The results are depicted in Figure 7 and Figure S10 (Supporting Information).
For the studied four LiPFSI molecules, the bond cleavage only occurs for Li (I) and Li (III), which are closer to the Li metal anode; while for Li (II) and Li (IV), no bond cleavage occurs at all.Further, the bond cleavages for Li (I) and Li (III) are different.For Li (I), two bonds, S2-C2 and N-S2, were cleaved during the whole process, and the S2-C2 bond was cleaved first (Figure 7b).For Li (III), three bonds, S2-C2, N-S2, and S1-C1, were cleaved within 5 ps, and N-S2 was the first one to be cleaved among these three bonds, but there is no cleavage for the N-S1 bond (Figure 7c).The bond breakage reflects the salt decomposition and the formation of inorganic species, including LiF.Based on the track of the bond cleavage, the possible decomposition routes for PPP to generate inorganic species can be proposed as those described in reactions 1-7, where reactions 1-3 are for Li (I), and those of 4-7 are for Li (III).

LiN
( SO 2 C 2 F 5 ) To further reveal the processes of bond cleavage and LiF formation, we traced the LiPFSI decomposition steps for Li (I) owing to its nearest location to the Li metal anode (Figure 7h).It is shown that there is no reaction before 40 s; when the process reaches 320 ps, the Li atom starts to connect with the F atom of LiPFSI due to the |e| charge being transferred from the system to the LiPFSI molecule, further affecting the C-F bond cleavage at 500 ps to form LiF. However, from 500 up to 1200 ps, due to the finiteness of the F element, some Li atoms still combine with the O atom, and thus the N-S bond cleavage occurs.The snapshots of AIMD simulations of LiPFSI decomposition steps further confirmed the proposed decomposition mechanism in reaction (1).
The radial distribution functions for these four LiPFSI molecules were obtained, as shown in Figures 7d-g To further confirm the important role of the Li-metal anode in the LiF formation, one more simulation was conducted for the cell with graphite as the anode (i.e., Li-ion batteries, LIBs).The corresponding radial distribution function is shown in Figure S11 (Supporting Information).The average length of Li-F is 6.2 Å, which is much larger than that of Li-O (2.2 Å), indicating that the first coordination layer of Li is the O atom.Comparing the radial distribution functions of Li-O and Li-F in LMBs and LIBs, it is demonstrated that LiPFSI can decompose and generate LiF more easily in LMBs, suggesting that the presence of Li metal as the anode is crucial for the LiF formation.
To provide a further comparison and discussion of decomposition mechanisms for different Li-salts, the decomposition mechanisms of LiFSI and LiTFSI proposed by others [10b,25a] were summarized (Figure S12, Supporting Information).Specifically, the decomposition of the FSI − anion begins with the cleavage of the S−F bond, and then the negatively charged fluorine radicals F• − and F(SO 2 ) 2 N − were formed.The formed F• − radical is rapidly bound with the Li atom of the outermost surface layer, resulting in the generation of LiF.After that, one of the ejected surface Li atoms is cleaved and combined with the remaining F − from one of the two F(SO 2 ) 2 N• radicals, which triggers a complete dissociation of the radical, starting with the S−N bond and leaving an SO 2 molecule and a SO 2 N• radical.One of the O atoms in the SO 2 N• radical reacts with the surface Li atom, forming LiO.For TFSI − , the sequence of bond breaking is N-S, S-C, C-F, S-O, C-F, and S-O, as evidenced by the AIMD simulation in Figure S12e (Supporting Information).The reduction of LiTFSI begins via the cleavage of the S-N bond, followed by breaking the C-S bond and forming three fragments.Three of the six C-F bonds of TFSI − are broken within a short period, thereafter yielding fragments such as NSO 2 CF 3 , SO 2 , carbon, and three F ions on the anode surface.The detached F ions are bonded with the Li ions and form LiF. Following this step, the S-O bond in the SO 2 fragment breaks and forms a carbonyl group (C=O) with the cleaved C atom on the Li metal surface.Therefore, the decomposition processes of FSI − , TFSI − , and PFSI − are different, resulting in the difference in SEIs from components and compositions to properties and performance, as evidenced by the experimental results obtained in this work and others [7,26]

Conclusions
In this work, the effect of the anion configuration in the Li-salts on the formation and composition of the naturally formed SEI in LMBs was studied, where the experimental determinations and theoretical analyses were organically combined.It is shown that the F-connecting bond in the Li-salts is more important than the salt molecular size and the F element contents, and thus, the performances of SPEs and cells using LiPFSI are slightly better than LiTFSI and much better than LiFSI.The XPS and AIMD simulation demonstrate that the bond cleavage and sequence for the salt (LiPFSI) are related to its position; the closer to the anode, the easier the bond cleavage.Also, the bond cleavage from type to sequence resulted in the formation and characteristics of SEI, and LiTFSI and LiPFSI induced the formation of LiF-rich SEI together with Li 3 N, while no Li 3 N was found when using PPF.Anyhow, all the studied Li-salts show a positive effect on the cell performances, regardless of the low-voltage or high-voltage cathodes.The present work provides a deep understanding of the SEI formation and characteristics when developing SPEs with Lisalts, which can inspire the rational design of SPEs with robust electrochemical performance.
Preparation of Electrospinning PI Film: The PI powder was synthesized with the method described in a previous article. [27]Specifically, 2 g PI powder was dispersed in 10 g DMF, stirring the mixture for 12 h at 60 °C to form a transparent solution.The Al foil was used in the negative roll of the spinning machine to collect the spinning PI film.The negative and positive pressures were set to −1 and 12 V, respectively, the humidity was ≈40%, the distance between the needle and the foil was ≈20 cm, and the spinning speed was 0.1 cm min −1 .Finally, the film was hot-pressed three times by a roller press and dried under vacuum for 12 h at 80 °C.
Preparation of SPEs: The SPEs were prepared by a typical solvent casting method.Briefly, a certain amount of PEO was dissolved into ACN, and the Li-salt (LiFSI, LiTFSI, LiPFSI) was added.After casting, the PI film was pasted on it and vacuum dried at 50 °C for 24 h.Then, the prepared SPEs were cut into 16 mm and stored in the glove box (Mikrouna, Universal 3660, H 2 O and O 2 < 0.01 ppm).The electrolyte properties and performances were influenced by the ratio between PEO and Li-salts.Thus, an orthogonal assay was designed to obtain an optimal ratio, where LiTFSI was chosen as the representative salt, and the evaluation was based on the ionic conductivity and ESW.Five mole-ratios of PEO to LiTFSI (X(= 17, 18, 19, 20, 22):1), denoted as PEO-X, were selected.The one with the ratio of 17:1 could not form a self-supporting film.A higher ionic conductivity was achieved for PEO-18 and PEO-20 at 60 °C (Figure S1a, Supporting Information), and a wider ESW of 4.0 V was obtained for PEO-18 and PEO-19 (Figure S1b, Supporting Information).Therefore, the ratio of 18:1 was considered as the optimal one, which was used for all other Li-salts studied in this work.Furthermore, Differential scanning calorimetry (DSC) for the corresponding electrolytes was detected (Figure S1c, Supporting Information), showing that their glass transition temperature (T g ) substantially decreases with increasing LiTFSI proportion.The detailed chemical additive amount is shown in Table S1.
Preparation of Cathodes and Assembling of Cells: The cathodes were fabricated by casting the homogeneous slurry of commercial active material (LFP, LCO, or NCM523), PVDF, and super P (w/w/w = 80/10/10) on the Al foil and dried at 100 °C under vacuum overnight.Then, the foil was cut into 14 mm disks and stored in the glove box.The coin cells were fabricated in a glove box filled with argon using the Li metal anode, the fabricated cathode, and the as-prepared SPEs as a separator and electrolyte.
Materials Characterization: The X-ray diffractometer (XRD) patterns were obtained from a ray diffractometer (Rigaku, Smartlab) with a Cu K radiation (40 kV, 40 mA).The relative crystallinity can be calculated with Equation (8).
where I c is the diffraction integral intensity of the crystal part, and I a is the diffraction integral intensity of the amorphous part.The microstructure was examined by the scanning electron microscope (SEM, JSM-7001F).Thermogravimetric analysis (TGA, STA7200RV, Hitachi High-Tech) was conducted to analyze the thermal stability.Differential scanning calorimetry (DSC) analysis was performed on Perkin Elmer Diamond to determine the glass transfer temperature (T g ) from −80 to 150 °C with a heating rate of 10 °C min −1 .The mechanical property was investigated by an electronic tensile machine (TA, Q800).The toughness can be calculated by integrating the area under tensile curves, as expressed by Equation ( 9): where T is the toughness,  is the strain,  b is the elongation at break, and S is the stress.The 7 Li NMR spectra were acquired on a Bruker AVANCE II 400 spectrometer with NMP as solvent.The X-ray photoelectron spectra (XPS) were performed using an ESCALAB 250Xi spectrometer.Time of Flight Secondary Ion Mass spectrometry (ToF-SIMS) was performed using PHI nanoTOFII Time-of-Flight SIMS under a vacuum.
Electrochemical Measurements: The electrochemical performances of the CR2025-type coin cells were detected by the CHI660E electrochemical workstation (Shanghai Chenhua Instrument Co., Ltd).The ionic conductivity of the developed SPEs was measured based on the stainless steel (SS)/SPE/SS cells by using electrochemical impedance spectroscopy (EIS) tests in the frequency ranging from 0.1 to 10 6 Hz from 25 to 65 °C and calculated with Equation (10), where d is the thickness of the SPE, R presents the intrinsic impedance value obtained from the EIS data, and S denotes the surface area of stainless steel (1.95 cm −2 in this work).

𝜎 = d∕RS
(10) Li + transference number (t Li + ) was determined by combining the measurements of alternating current impedance and direct current polarization using the Li/SPE/Li cell.Specifically, the polarization currents of the cell, including those of initial (I 0 ) and steady-state (Is), were recorded under a direct current polarization voltage of 50 mV (ΔV), and the interfacial resistances before (R 0 ) and after (Rs) polarization were tested by alternating the current impedance.Subsequently, t Li + was calculated from the Bruce-Vincent-Evans Equation (Equation 11).
Theoretical Simulations: In this work, the geometry of all molecules was optimized through density functional theory (DFT).All the DFT computations were performed at the B3LYP/TZVP theoretical level with the GD3BJ dispersion correction.The SMD implicit solvation model was used to account for the solvation effect of MeCN.All these calculations were performed with the Gaussian 16 software package.
Due to the uneven Li + flux, the dendrite formation is related to electrode dynamics and ion transport caused by diffusion.In this work, COM-SOL Multiphysics 6.0 was used to track the moving boundary of lithium sheet surface by coupling the "tertiary current distribution, Nernst Plank" interface and "deformation geometry" to simulate the process of Li + deposition at the interface.Besides, the Nernst-Einstein equation was used to calculate the ion mobility.The diffusion coefficient and ionic conductivity were determined by EIS, and the electrode reaction and deposition rate were described by the Butler-Walmer equation and the Faraday's law, respectively.
The ab-initio molecular dynamics (AIMD) simulation was carried out by using the projector augmented wave method in the framework of the density functional theory (DFT), [28] as implemented in the Vienna abinitio Simulation Package (VASP).The generalized gradient approximation (GGA) and Perdew-Burke-Ernzerhof (PBE) exchange functional [28] were used.The Monkhorst-Pack method [29] was employed for the Brillouin zone sampling.The convergence criteria of energy and force calculations were set to be 10 −5 eV atom −1 and 0.01 eV Å −1 , respectively.The AIMD simulations were performed by the supercells with a Gammacentered 1 × 1 × 1 k-point grid.A small plane-wave energy cutoff of 300 eV was chosen.The time step was set to 2 fs, and all supercell systems were simulated for 10 000 steps with a total time of 20 ps in a statistical ensemble with a fixed particle number, volume, and temperature (NVT).

Figure 2 .
Figure 2. The HOMO and LUMO levels.Red, yellow, light blue, dark blue, and gray balls stand for O, S, F, N, and C atoms, respectively.

Figure 4 .
Figure 4. Electrochemical performances of full cells.a) Cycling performance of LFP//Li cells at a current density of 0.5 C (1 C = 170 mA g −1 ) and b) the corresponding GCD curves.c) Rate performance of LFP//Li cells.Cycling performance of cells using d) LCO cathode and e) NCM 523 cathode with the Li anode at a current density of 0.1 C (1 C = 140 mA g −1 for LCO, and 1 C = 160 mA g −1 for NCM523) and f,g) the corresponding GCD curves.

Figure 5 .
Figure 5. Microstructure and Li + flux.Ex-situ SEM of Li metal after cycling when using a) PPP, b) PPT, and c) PPF as the SPEs.Evolution snapshots at different time steps from the phase-field simulation of Li deposition within d-d2) PPP, e-e2) PPT, and f-f2) PPF.

Figure 6 .
Figure 6.Analysis of the SEI composition.a) The schematic illustration of the SEI using three salts (from left to right: LiFSI, LiTFSI, LiPFSI).The XPS characterization of the SEI elemental compositions that use b) PPF, c) PPT, and d) PPP as the SPEs at different etching times.e,f) ToF-SIMS depth profiles with accompanying 3D renders of LiF 2 − , Li 2 F 3 − , LiPFO 2 − , C 2 H − , LiCO 3 − , CH − , CF 3 SO 2 N − , and LiO − after cycling in Li/PPP/Li cell.
. A peak arises at 2 Å for Li (I)−F and Li (III)-F, while, Li (II) and Li (IV) show weak signals with F but strong signals with O instead.This indicates that the LiPFSI molecule near the Li metal will generate LiF.Oppositely, the LiPFSI molecule far away from the Li metal is preferred to connect to O rather than F to generate Li 2 O.It indicates that the generation of LiF is related to the position of LiPFSI with respect to the Li metal, and only the LiPFSI molecules close to the Li metal anode will decompose and form LiF, i.e., not all the LiPFSI molecules in SPEs are involved in the LiF formation.

Figure 7 .
Figure 7. Probability distribution functions of Li + with F and O elements.a) The simulation structure of PPP.Bond cleavage of b) Li (I) and c) Li (III) in LiPFSI.Radial distribution functions of d) Li (I)-F/O, e) Li (II)-F/O, f) Li (III)-F/O, g) Li (IV)-F/O.h) Snapshots of AIMD simulations with instantaneous Bader charges (the unit of charge is |e|) illustrating the degradation dynamics.