Mechanism of “shuttle-relay” Li metal battery. Fig. 1a illustrates the working mechanism of the quasi-solid-state SRLMB, which is realized by the well-designed HGPE. Currently, Li-ion batteries extensively apply organic electrolytes containing cyclic carbonate solvents (e.g. ethylene carbonate (EC)) with high dielectric constant to dissolve lithium hexafluorophosphate (LiPF6) salt, and linear carbonate solvents (e.g. ethylmethyl carbonate (EMC)) to reduce the electrolyte viscosity11. However, such carbonate-based electrolytes generally show poor compatibility with both Li metal anode and 5 V-class cathodes (e. g. LRO for “rocking-chair” chemistry and graphitic carbon for “dual-ion” chemistry). On the anode side, the electrolyte solvents cannot form stable SEI layer on the Li metal surface, leading to Li dendrite growth and low Columbic efficiency18. On the cathode side, the insufficient oxidation resistance of carbonate solvents triggers severe electrolyte decomposition and constructs a thick CEI with high-resistance, which dramatically degrades the battery performance (Fig. 1a, upper panels). The interfacial issues are even more severe for graphite cathodes, since the carbonate molecules tend to co-intercalate into graphite interlayers, giving rise to low reversible anions insertion/extraction capacity accompanying structural deterioration16.
To address these issues, we developed an all-fluorinated electrolyte for high-voltage Li metal batteries, which contains 1 M LiPF6 dissolved in a mixture of fluoroethylene carbonate (FEC): 2,2,2-trifluoroethylmethyl carbonate (FEMC): 1,1,2,3,3,3-hexafluoropropyl-2,2,2-trifluoroethylether (HTE) with a volume ratio of 1: 6: 3. In this liquid electrolyte, FEC is beneficial for improving the compatibility with Li anodes, meanwhile FEMC ensures a reversible insertion/extraction of PF6- into graphite (Supplementary Fig. 1). Moreover, as a novel electrolyte component, the HTE not only functionsas a diluent to reduce the electrolyte viscosity, but also optimizes the localized solvation structure of cation/anion aggregates, thus further stabilizing the Li|electrolyte interfaces. On this basis, 3 wt% DAP monomer and 1.5 wt% PETEA crosslinker were in-situ polymerized in this fluorinated electrolyte to form a HGPE, in which the three-dimensional polymer matrix effectively improves the electrolyte safety by preventing liquid leakage (Supplementary Fig. 2). To verify the effect of each electrolyte component, we performed density functional theory (DFT) calculation on the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO) energies of the solvent molecules. Based on molecular orbital theory, the HOMO energy correlates to the oxidative decomposition potential, while the LUMO energy is associated with the reductive decomposition potential19. As shown in Fig. 1b, the HOMO energy of fluorinated solvents (i. e. FEC: -7.3278 eV, FEMC: -7.001 eV and HTE: -8.1353 eV) is much lower than those of EC (-6.8853 eV) and EMC (-6.4791 eV), demonstrating the superior oxidation resistance of fluorine solvents owing to the strong electron-withdrawing effect of fluorine atoms on the core of solvent molecules20. Meanwhile, DAP monomer presents the highest HOMO value (-6.2292 eV). As a result, the residual DAP monomer after polymerization acts as a CEI-forming additive in the HGPE to further inhibit the electrolyte oxidation and the co-intercalation of solvent molecules into graphite. Moreover, FEC and DAP exhibit the lowest LUMO energies of -0.8835 eV and -1.0021 eV, respectively. Consequently, FEC and the residual DAP monomer will be preferentially reduced on the Li anode to form a protective LiF-rich and phosphorus-containing SEI inhibiting dendrites growth. To evaluate the effect of HTE, we calculated the binding energy of electrolyte components with Li+ cation and PF6- anion (Fig. 1c). It is seen that the HTE shows the lowest absolute values of binding energy with both Li+ and PF6-, indicating a weak interaction between HTE and ions. This is consistent with the lower solubility of LiPF6 salt in HTE compared with other solvents (Supplementary Fig. 3). Therefore, the introduction of HTE enables formation of a highly concentrated electrolyte in local regions by increasing the ratio of ion: fluorinated carbonate in the solvation structures. Such a unique solvation structure can minimize the excessive side reactions between electrolyte solvents and electrodes, thus improving the battery performance21.
When applied in SRLMBs with LRO as active materials and KS6 graphite as conductive agents, the well-designed HGPE endows a novel “shuttle-relay” battery chemistry. As shown in the lower panels of Fig. 1a, Li ions are stripped from the LRO cathode in the voltage range of 2.0-4.8 V, followed by an insertion of the PF6 anions into the conductive graphite at 4.8-5.0 V. In the HGPE, the synergistic effect of liquid-state electrolyte components constructs robust SEI/CEI to improve the electrode|electrolyte compatibility; meanwhile, the polymer matrix efficiently retards the migration of Li+/PF6- ions to SEI/CEI surface defects through strong interaction (see the high binding energy absolute values between DAP and Li+/PF6-, Fig. 1c), thereby favoring a uniform Li+/PF6- ion flux to promote uniform Li plating and anions intercalation into graphite7.
Characterization of the HGPE. As seen from Fig. 2a, the as-synthesized HGPE appeared as non-flowing white gel. The conversion degree of the monomer and crosslinker to form HGPE was calculated to be around 71 % (Supplementary Note 1). Fig. 2b exhibits the Fourier transform infrared spectra (FTIR) of DAP monomer, PETEA crosslinker and HGPE polymer matrix. The characteristic peaks at around 1013 cm-1 (P-O stretching), 1099 cm-1 (P=O stretching), 1260 cm-1 (C-O antisymmetric stretching), 1470 cm-1 and 1406 cm-1 (CH2 bending), and 1720 cm-1 (C=O stretching) presented in the FTIR spectra of DAP and PETEA22,23. Upon polymerization, the absorption peak at about 1630 cm-1 correlated to the stretching vibration of C=C bond nearly disappeared in the polymer matrix, indicating a high-degree polymerization of monomer and crosslinker in the HGPE. Fig. 2c demonstrates the combustion tests of 1 M LiPF6-EC: EMC electrolyte and the HGPE. It is seen that the traditional liquid electrolyte easily caught fire on ignition, and kept on burning even after removing the torch with a self-extinguishing time (SET) of ≈92 s g-1 (Supplementary Video 1). This significantly differs from the HGPE which exhibited zero SET after removal of the torch, indicating its excellent non-flammability (Supplementary Video 2). This is attributed to facts that the substitution of hydrogen atoms by fluorine atoms in fluorinated solvents significantly reduces the generation of hydrogen radicals, which hence diminishes combustion hazard (Supplementary Fig. 4 and Supplementary Video 3 and Video 4), and the thermal decomposition products of DAP at high temperature can further capture the radicals and prevent the unwanted combustion chain reactions24. Moreover, it is worth noticing that the polymer matrix in the HGPE effectively immobilizes the solvents and decreases their volatility, thus preventing the risk of liquid leakage (Supplementary Fig. 5)25. Such high safety of the HGPE is a critical asset for the practical application of high-energy Li metal batteries.
Raman spectra was measured to characterize the coordination environment in the electrolyte. It is seen that in the mixture of FEC: FEMC (1: 6 by volume), peaks at around 730cm-1 were recorded (assigned to free FEC) and one at about 840 cm-1 corresponding to the free FEMC molecule (Fig. 2d and Supplementary Fig. 6). After dissolving 1 M LiPF6 into the mixture, the peak intensity of free solvent molecules diminished accompanying the appearance of new bands at about 849 cm-1 (Li+-coordinated FEMC), 921 and 745 cm-1 (Li+-coordinated FEC)26. With the addition of HTE, an extra peak of free HTE molecules was observed at approximately 706 cm-1 in the spectrum of LiPF6-FEC: FEMC: HTE electrolyte, meanwhile the peak intensity of Li+-coordinated carbonates increased, which verifies that more fluorinated carbonate molecules are coordinated with Li ions in the solvation sheaths27. This is well-consistent with the binding energy calculation results in Fig. 1c, which efficiently alleviates the excessive side reactions between free solvent molecules and Li metal.
Ionic conductivity is considered as an important property of electrolytes. Fig. 2e and Supplementary Fig. 7 show the temperature dependences of ionic conductivities for 1 M LiPF6-EC: EMC (1:6 by volume), 1 M LiPF6-FEC: FEMC (1:6 by volume), 1 M LiPF6-FEC: FEMC: HTE (1:6:3 by volume) and HGPE electrolytes within the temperature range 0 °C to 90 °C. The plot of log σ versus T-1 for electrolytes presents a nonlinear relationship that can be well fitted by the following empirical Vogel-Tamman-Fulcher (VTF) formula28:
where σo is the pre-exponential coefficient, Ea is the pseudo-activation energy, To is the parameter related to the ideal glass transition temperature, and R is the gas constant. The fitted parameters and ionic conductivity values are presented in Supplementary Table 1. It is seen that the 1 M LiPF6 in FEC: FEMC and 1 M LiPF6 in FEC: FEMC: HTE electrolytes showed a slight decrease in ionic conductivity at 25 oC due to the high viscosity of the fluorinated solvents. After the in-situ gelation, the ionic conductivity of HGPE maintained 1.99 ×10-3 S cm-1 at 25 oC. Meanwhile, the Ea value (1.00×10-2 eV) was very close to that of the 1 M LiPF6 in FEC: FEMC: HTE liquid electrolyte (9.30×10-3 eV). This indicates that the hindrance of the gel matrix for wanted ion transport is negligible. Such a high ionic conductivity value of the HGPE enables excellent operation of batteries at high current densities.
The electrochemical stability window of the electrolytes was investigated by cyclic voltammetry (CV). As shown in Fig. 2f, the strong current appeared from around -0.43 V vs. Li/Li+ in the negative scan was related to the plating of Li on the stainless steel; while a peak of Li stripping occurred at about 0.32 V in the subsequent reverse scan29. Apart from in the plating-striping range of -0.5 to 0.5 V, no noticeable peaks or oxidation currents peaks were observed in the CV curve of the cell using the HGPE. This demonstrates the HGPE remained stable in the testing range (up to 5.5 V vs. Li/Li+). In sharp contrast, the irreversible oxidation voltages of 1 M LiPF6-EC: EMC, 1 M LiPF6-FEC: FEMC and 1 M LiPF6-FEC: FEMC: HTE were around 3.5 V, 4.2 V and 4.9 V vs. Li/Li+ (inset of Fig. 2f and Supplementary Fig. 8). respectively. The high electrochemical stability of HGPE mainly originates from the fluorination of the electrolyte solvent and the robust CEI formed by the oxidation of DAP at about 3.3 V (Supplementary Fig. 8), which enables its application in 5 V-class Li metal batteries. Furthermore, the PETEA-DAP polymer framework can restrict the movement of anions, resulting in an increased Li ion transfer number (tLi+) for the HGPE (i. e. 0.43) compared with the 1 M LiPF6-FEC: FEMC: HTE electrolyte (0.37, Supplementary Fig. 9)9. Such increased tLi+ is close to 0.5, which facilitates the balance of the active ions in the DIBs30.
Li metal morphology and interfacial chemistry. To investigate the stability of Li metal anodes in different electrolytes, the voltage variation of symmetric Li||Li cells was assessed during galvanostatic cycling at a constant current of 0.5 mA cm-2. As shown in the inset of Fig. 3a, the Li|Li symmetric cell using 1 M LiPF6-EC: EMC electrolyte exhibited a dramatically increased overpotential with cycling time (≈5 V at 420 h), mainly due to the thickening of SEI layer and continuous Li dendrite growth18. The overpotential of Li|1 M LiPF6-FEC: FEMC|Li cells and Li|1 M LiPFF6-FEC: FEMC: HTE|Li cells were ≈120 mV and ≈150 mV, and the cells failed at 600 h and 800 h, respectively (Supplementary Fig. 10). Meanwhile, the Li|HGPE|Li cell delivered a stable voltage hysteresis of ≈100 mV with no obvious oscillation throughout a 1400 h cycling (the general decrease in overpotential in the initial cycles is related to the activation of Li anode with a pristine oxide layer on the surface31), indicating a dendrite-free Li deposition when use HGPE. The temperature-dependent electrochemical impedance spectroscopies (EISs) of the Li||Li cells after 20 cycles were performed to further evaluate the activation energies during Li deposition/stripping. The activation energies derived from the SEI (Rsei) and ion transfer resistance (Rct) are denoted as Ea1 and Ea2, corresponding to the energy barriers for Li ions transport across the SEI layer and their desolvation from Li+ solvation shells, respectively (Supplementary Note 3)32. The Ea1 for HGPE (46.07 kJ mol-1) is significantly lower than those for the 1 M LiPF6-EC: EMC (76.65 kJ mol-1), 1 M LiPF6-FEC: FEMC (50.63 kJ mol-1) and 1 M LiPF6-FEC: FEMC: HTE (59.86 kJ mol-1) electrolytes (Fig 3b and Supplementary Fig. 12). This implies that the structure and composition of the SEI layer formed in the presence of HGPE endows a fast Li ion transport kinetics. Moreover, although the Ea2 value for HGPE (58.28 kJ mol-1) is slightly higher than that for 1 M LiPF6-EC: EMC electrolyte (51.13 kJ mol-1) due to the stronger interaction between PF6--EC than that between PF6--fluorinated carbonate that facilitates Li+ desolvation from the ion pairs and aggregates33,34, the Ea2 for HGPE is still lower than that for 1 M LiPF6-FEC: FEMC electrolyte (62.52 kJ mol-1) and 1 M LiPF6-FEC: FEMC: HTE electrolyte (69.67 kJ mol-1, Supplementary Fig. 11). This could be attributed to the fact that DAP-based polymer matrix can promote the dissociation of Li ions from the solvation sheath35. The low Ea1 and Ea2 of the HGPE-based cell promote a low-resistance Li|HGPE interface with fast ion diffusion and conversion.
The average Li plating/striping Coulombic efficiency (CEavg) measurement was further conducted in Li||Cu cells in various electrolytes (Supplementary Note 4)36. The cells using the HGPE exhibited a CEavg of 99.7 %, which is one of the best values for state-of-the-art electrolytes6,37,38 and much higher than those using 1 M LiPF6-EC: EMC (69.9 %), 1 M LiPF6-FEC: FEMC (98.4 %) and 1 M LiPF6-FEC: FEMC: HTE electrolyte (98.6 %, Fig. 3c and Supplementary Fig. 13). The plating morphologies of Li on Cu substrates were examined by field emission scanning electron microscope (FE-SEM). As shown in Fig. 3d, the Li|1 M LiPF6-EC: EMC|Cu cell present a highly loose and mossy deposition structure with a thickness of ≈29.0 μm, far exceeding the theoretical value (≈9.7 μm). After fluorinating the carbonate solvents and introducing the HTE diluent, the surfaces of deposited Li gradually became smoother and denser, and the thicknesses of Li deposition decreased to around 21.4 and 16.1 μm, respectively (Supplementary Fig. 14). In the Li|Cu cell employing the HGPE, for comparison, the plating Li showed a compact morphology as aggregated large particles, and the plated thickness (≈10.7 μm) was very close to the theoretical value (Fig. 3e). Such a dense Li deposition with a smaller surface/volume ratio effectively minimizes the parasitic reaction between metallic Li and electrolyte, and thus enables the high CEavg of Li|HGPE|Cu cells.
To analyze the microstructure of the SEI, 1 mAh cm-2 Li was repeatedly plated on and stripped off a Cu grid for 10 cycles at 0.2 mA cm-2 to obtain a vacant SEI shell for transmission electron microscopy (TEM) characterization. It is seen that a large amount of “dead Li” residues appeared on the Cu grid cycled in 1 M LiPF6-EC: EMC electrolyte, indicating an irreversibility of Li plating/striping (Fig. 4a, inset). The SEI was mainly composed of Li2O particles distributed in an amorphous matrix (Fig. 4a), and the composition was further identified as organic compounds (e. g. ROCO2Li, where “R” represents functional groups) originating from the decompositions of carbonate solvents32, and LixPOyFz/Li2O as the decomposition products of LiPF6 salt, respectively (see the in-depth X-ray photoelectron spectroscopy (XPS) results in Supplementary Fig. 15-18)39. The Young's modulus of this SEI layer was as low as 398 MPa (Fig. 4c). For comparison, the amount of residual inactive Li obviously decreased on the Cu substrates in fluorinated electrolytes (Supplementary Fig. 18, insets). Meanwhile, the proportion of LiF, mainly originating from the reduction of fluorinated solvents, greatly increased in the SEI (Supplementary Fig. 15-18). It is well-known that LiF with high mechanical strength (i. e. a shear modulus of 55.1 GPa, almost 11 times higher than that of Li metal (4.9 GPa)) can significantly enhance the robustness and interfacial energy of SEI layers, thus blocking Li dendrite growth24. As expected, the mechanical strength of SEIs in 1 M LiPF6-FEC: FEMC and 1 M LiPF6-FEC: FEMC: HTE electrolytes increased to 911 MPa and 1426 MPa, respectively (Supplementary Fig. 20-21). In the Cu grid retrieved from the cell employing the HGPE,a high Young's modulus up to 2768 MPa (Fig. 4d) has been achieved, owing to the co-existence of LiF and phosphorous-containing compounds (i. e. P-O-C and P=O) derived from the residual DAP monomer in the SEI (Fig. 4b and Supplementary Fig. 15-18). The robust SEI can efficiently suppresses the formation of Li dendrite and dead Li (Fig. 4b, inset) and leads to a smooth surface morphology of the cycled Cu grid (Fig. 4d, inset). The above results are well-consistent with the electrochemical behavior in traditional Li metal batteries.
The stabilization effect of HGPE on the Li|electrolyte interface can be elucidated as follows. It is known that the ion transfer kinetics and Li deposition behavior are mainly determined by the composition and morphology of the SEI32. As shown in Fig. 4e, in the traditional 1 M LiPF6-EC: EMC electrolyte, the Li+ solvation shell consists of large amounts of carbonate solvent molecules but with negligible PF6- anions solvation40. Upon electrochemical reaction, EC and EMC molecules in the solvation shell will be reduced and constitute the main component of SEI. Such an organic component (e. g. ROCO2Li)-rich SEI is of insufficient strength to inhibit the dendrite growth, and repeatedly breaks down/reconstructs during the cycling, which causes raised thickness and resistance. This gives rise to a large energy barrier (Ea1) for Li ions to transport through the SEI, triggers inhomogeneous charge distribution and aggravated polarization, further promoting dendrite formation4. In contrast, in HGPE, although the activation energy for dissociating the Li ions from the solvation sheath (Ea2) is similar to that for the traditional liquid electrolyte, the Ea1 is significantly reduced to facilitate Li ion diffusion through the SEI. This is because the addition of HTE as diluent leads to a formation of localized highly concentrated regions in the electrolyte, in which fluorinated carbonates and Li+-anion ion pairs participate in the solvation shell32. This solvation shell structure and residual DAP monomer in the HGPE endow a formation of an inorganic component (e. g. LiF-rich) SEI on the Li metal surface, which is highly robust to suppress dendrite formation and maintains low resistance throughout cycling (Fig. 4f). Additionally, the crosslinked DAP-PETEA matrix not only generates a relatively homogeneous Li+ flux, but also effectively eases the volume changes upon Li deposition, thus inhibiting any incipient dendrite growth7. Such a stable Li|HGPE interface with low Li+ diffusion energy barrier contributes to the superior performance of HGPE in Li metal batteries.
Electrochemical performances evaluation. Fig. 5a and b show the charge-discharge profiles and rate performances of the Li|HGPE|KS6 graphite (sheet size of 4 μm, Supplementary Fig. 22) cell, respectively. During the charging process of Li|HGPE|KS6 graphite cells, three slopes at 4.21-4.50 V (stage III), 4.50-4.85 V (stage II) and 4.85-4.95 V (stage I) corresponded to the staged phase transition of graphite during anion insertion9,41. Subsequently, three plateaus associated with anions deintercalation from the graphite appeared in the discharge curves with potentials downshifted to 4.95-4.79 V, 4.79-4.40 V and 4.40-4.0 V, respectively. This is in agreement with the CV curves in Supplementary Fig. 23 and the dQ/dV plot in Supplementary Fig. 24. The Li|HGPE|KS6 graphite cells delivered specific discharge capacities of 101.5, 99.8, 98.2, 95.8, 92.0, 88.3 and 81.0 mAh g-1 at 0.1, 0.2, 0.5, 1, 2, 3 and 5 C (1 C=100 mAh g-1 based on the mass of graphite), respectively (Fig. 5b). These are much higher than those of cells using 1 M LiPF6-FEC: FEMC and 1 M LiPF6-FEC: FEMC: HTE electrolytes (Supplementary Fig. 25). When the current density was switched back to 0.1 C, the capacity retention of the HGPE-based cell was 98.2 % of the initial value, demonstrating that this battery system is highly robust and stable. In contrast, the capacity of Li||KS6 graphite cell with 1 M LiPF6-EC: EMC rapidly decreases to ≈0 at a current density of 3 C, indicating a sluggish Li ions diffusion kinetics at the graphite|electrolyte interface.
Fig. 5c shows the long-term cycling performance of Li|KS6 graphite DIBs employing various electrolytes at 1 C. The Li|1 M LiPF6-EC: EMC|KS6 graphite cell exhibited an initial discharge capacity of 29.4 mAh g-1, and the capacity suddenly droped to 21.7 mAh g-1 at the 169th cycle. This is probably caused by serious structure exfoliation and destruction of graphite originating from the co-intercalation of solvent molecules into the graphite interlayers, as well as the thickening of the CEI induced by the severe electrolyte oxidation42. The lifespan and reversible capacity of DIBs significantly increased with the adoption of fluorinated solvents (Supplementary Fig. 26). The Li||KS6 graphite cell using the HGPE demonstrateda high initial discharge capacity of 89.8 mAh g-1 with a capacity retention of ≈93 % after 1000 cycles, and the Coulombic efficiency was maintained at ≈98.9 % except for the activation process in the first 10 cycles (Fig. 5c). The above results were further corroborated by the small interfacial resistances (Rsei and Rct) of the HGPE-based cells, and the interfacial resistance changes were much smaller than in the cells using other electrolytes during cycling (inset of Fig. 5c, Supplementary Fig. 27 and Supplementary Note 6). This superior cycling stability is mainly because the HGPE effectively suppresses solvent co-intercalation and protects the structural integrity of graphite, thus allowing a highly reversible and durable insertion/extraction of anions into/from the KS6 graphite.
SRLMBs have been further developed by applying KS6 graphite as conductive agent in the cathode of the LRO|HGPE|Li cells. As shown in Fig. 5d, during the charging of LRO|HGPE|Li and Li|HGPE|LRO/graphite cells, a sloping potential below 4.5 V corresponded to Li ion extraction from LRO cathode44. For the hybrid LRO/KS6 graphite cathode an extra plateau at 4.9 V appears, which is ascribed to a “relay” intercalation step of PF6- into the graphite. Supplementary Fig. 28 further validated that KS6 contributed ≈ 6.2 % of the areal capacity and ≈ 7.1 % of the areal energy density (Supplementary Note 7). Such a “shuttle relay” process was highly reversible in the subsequent discharging. The SRLMB delivered a discharge capacity of 205.3 mAh g-1, based on the total mass of the cathode active material and conductive agent, which was higher than that of the Li|HGPE|LRO cell with Super P as the cathode conductive agent (191.1 mAh g-1, Fig.5e). The cell can maintain a capacity of 188.0 mAh g-1 after 100 cycles at 0.2 C (1 C=250 mAh g-1 based on the mass of LRO) with a high capacity retention of 91.6 %. This indicates that the PF6- intercalation/deintercalation after Li extraction/insertion can increase the battery capacity without sacrificing its cycling stability. In addition, the Li||LRO/graphite cell applying traditional 1 M LiPF6-EC: EMC electrolyte suffered from a quick capacity fading during cycling, demonstrating a poor electrode|electrolyte compatibility (Supplementary Fig. 29).
Single-layer SRLMB pouch cells with 50 μm-thick Li foil as anodes were assembled to further evaluate the battery performance under abuse conditions (Supplementary Fig. 30a). The Li|HGPE|LRO/graphite pouch cell not only showed excellent cycling performance (Supplementary Fig. 31), but also exhibited superior flexibility (i. e. consistently powering a red light-emitting diode (LED) under flatted, bent, or even clustered states (Fig. 5f, lower panels and Supplementary Video 5). Whereas the cell using traditional liquid electrolyte losing power supply ability in bent or clustered states (Fig. 5f, upper panels and Supplementary Video 6). This verifies that the electrode|HGPE interfaces can maintain tight adhesion under significant shape deformations. Moreover, when aging the fully charged cells at 130 oC, an Li||LRO/graphite pouch cell with 1 M LiPF6-EC: EMC liquid electrolyte suffered from severe swelling and bulging due to the sever volatilization and thermal decomposition of the liquid electrolyte (Fig. 5g, inset), and the open circuit potential suddenly dropped to ≈0 V at 1964 s, illustrating a contact failure inside the cell (Fig. 5g). In sharp contrast, owing to the high thermal stability of fluorinated solvents and leakage-free property of the gel, the shape and open circuit voltage of Li||LRO/graphite pouch cell with HGPE did undergo not change at the 130 oC test (Fig. 5g). Meanwhile, the temperature excursion of a fully charged Li|HGPE|LRO/graphite pouch cell was lower than that of a Li|1 M LiPF6-EC: EMC|LRO/graphite pouch cell during the nail penetration safety tests (Supplementary Fig. 30b). All these enable a highly safe operation of SRLMBs in practical applications.
Electrochemical mechanism of PF6- intercalation/deintercalation in the HGPE. In-situ X-ray diffraction (XRD) tests were conducted to further investigate the operation mechanism of PF6- intercalation/deintercalation in the presence of different electrolytes at 0.05 C. The in-situ XRD patterns and charge/discharge curves during the initial cycle are shown in Supplementary Fig. 32, and the corresponding intensity contour maps are presented in Fig. 6a and b, respectively. In the Li||KS6 graphite cells using traditional 1 M LiPF6-EC: EMC liquid electrolyte, the graphite (002) diffraction peak gradually shifted from 26.6° to 24.1° during the charging process. The corresponding interlayer spacing (d) values of graphite can be calculated from the XRD pattern according to the Bragg equation43:
where θ is the diffraction angle between the incident X-rays and the corresponding crystal plane, and λ is the X-ray wavelength (i.e., 0.15406 nm). The increase of graphite d(002) interplanar spacing from 0.335 nm at 3.0 V to 0.370 nm at 5.0 V is consistent with the intercalation of the PF6- anions into the graphite interlayer (Supplementary Fig. 33). In the subsequent discharge process, however, no distinct position change of the graphite (002) peak was observed, indicating a constant graphite d(002) spacing caused by the blocked PF6- anion stripping from the graphite host (Fig. 6a). This has been further confirmed by the TEM image of the graphite cathode after cycling, in which the lattice spacing from XRD (0.370 nm) is well consistent with the calculated value from the TEM (Fig. 6c, inset). The thickness of the CEI derived from the oxidative decomposition of carbonate solvents is as high as 7.8 nm in 1 M LiPF6-EC: EMC liquid electrolyte, which strongly hinders the PF6- stripping and causes the irreversibility during cycling (Fig. 6c). In sharp contrast, in the Li|HGPE|graphite cell, the graphite (002) diffraction peak gradually shifted to 24.1°(i. e. interlayer spacing of 0.370 nm) during the charge process, and reversibly reverted to 26.50° (i. e. interlayer spacing of 0.336 nm) when discharged back to 3.0 V (Fig. 6b). The TEM image of the cycled graphite cathode exhibited lattice stripes with a spacing of 0.336 nm (Fig. 6d, inset), which is in accordance with the in-situ XRD results and almost the same as that of the pristine graphite powder (0.335 nm, Supplementary Fig. 34). This validates a highly reversible PF6- intercalation into/deintercalation from the graphite without structural deterioration. Moreover, the thickness of the CEI formed in the HGPE is only 1.4 nm, indicating a suppressed electrolyte oxidation with reduced interfacial resistance.
In-depth XPS measurements were performed on graphite cathodes cycled in various electrolytes to further analyze the components of CEI layers. For the Li|1 M LiPF6-EC: EMC|graphite cell, peaks at approximately 532 eV (C=O), 530 eV (ROCO2-) and 528 eV (Li2O) in O 1s spectrum and 687.5 eV (PF6-), 686 eV (LixPOyFz) and 685 eV (LiF) in F 1s spectrum appeared on the cycled cathode surface (a sputtering time of 0 s, Fig. 6d)9,20,32, which is in agreement with Li 1s spectrum in Supplementary Fig. 35. In addition, the peaks at about 136 eV, 134 eV and 131 eV in the P 2p spectrum are related to PF6-, LixPOyFz and P-C/P-O-C44, respectively (Supplementary Fig. 36). These results suggest that in 1 M LiPF6-EC: EMC liquid electrolyte, the CEI on the graphite cathode is mainly composed of abundant alkyl carbonate (e. g. ROLi) and polycarbonate as oxidation products of carbonate solvents (as further verified in the C 1s spectrum in Supplementary Fig. 37, and LixPOyFz and LiF originated from the decomposition of LiPF6. More importantly, a large amount of intercalated PF6- anions remained in the graphite interlayers, and the intensity of ROCO2- from carbonates abruptly increased at a depth of 20 nm (i. e. a sputtering time of 240 s). These demonstrate the inhibited stripping of PF6- anions from the graphite host and the co-intercalation of solvent molecules, which causes the irreversibility of batteries at 0.05 C.
In contrast, for the surface of the graphite cathode from the cycled HGPE-based cell, two new O 1s peaks at about 533 eV and 531 eV are assigned to P-O-C and O-P=O as oxidative decomposition products of DAP45,46, which is consistent with the P 2p spectra in Supplementary Fig. 36. Additionally, the peak intensities of PF6-, LixPOyFz and LiF in F1s spectrum and the ROCO2- from the C 1s spectrum significantly decreased. The above results suggest that phosphorus-containing substances in the CEI (i.e., P-O-C and O-P=O) can suppress the decomposition of electrolytes and form a thin CEI to ensure a reversible PF6- de-intercalation (Fig. 5d and 6d). Considering allyl groups can easily undergo polymerization45, such a CEI film may originate from polyphosphoesters generated by the electropolymerization of residual DAP monomer on the carbon-oxygen rich graphite surface (Supplementary Fig. 39). Moreover, in the XPS depth profiles of the Li|HGPE|graphite cell, no noticeable peaks were observed from the carbonate solvents or their decomposition products. This confirms that solvent molecule co-intercalation can be effectively inhibited by such protective CEI, which preserves the cathode against structure destruction and facilitates the superior electrochemical performance of the HGPE-based DIBs and SRLMBs.