All-solid-state Li–S batteries with fast solid–solid sulfur reaction

a, XRD patterns for the LBPSI electrolytes with x of 0, 0.11, 0.17, 0.29, 0.50, 0.67, 0.75 and 0.80. b, Raman spectra of LBPSI electrolytes with x from 0 to 0.80. c, Enlarged view focused on the [P_mS_n] clusters. d, Schematic drawing of the clusters with their characteristic Raman shift. e, ³¹P MAS NMR spectra of LBPSI electrolytes with x of 0.17 and 0.29. f–h, DC polarization curves with an ion-blocking cell configuration of Ti|electrolyte|Ti for the three LBPSI electrolytes with x of 0 (f), 0.17 (g) and 0.29 (h). The thicknesses of the pellets during measurement for the three electrolytes are 0.028 cm, 0.022 cm and 0.023 cm, respectively. The measurements were conducted at 25 °C. i, The trend of electron conductivities obtained from DC polarization curves and the corresponding E_a values. The electronic conductivities for the three LBPSI electrolytes with x of 0, 0.17 and 0.29 are 6.4 × 10⁻¹⁰ S cm⁻¹, 2.8 × 10⁻¹¹ S cm⁻¹ and 2.8 × 10⁻¹⁰ S cm⁻¹, respectively. It is clear that, by fine-tuning the P/B ratio in the sulfide glass, the electronic conductivity of the electrolytes can be reduced. The values are much lower than the typical electron conductivity of argyrodite Li₆PS₅Cl and Li_5.5PS_4.5Cl_1.5 (typically measured to be 10⁻⁸–10⁻⁹ S cm⁻¹)^51,52, which may lead to low self-discharge in a cell. Discussions on the evolution of local clustering structures: on the basis of the XRD patterns (a), for the P₂S₅-free electrolyte (x = 0), a crystalline LiI phase is present, indicating that the LBSI glass framework cannot fully include the amount of LiI added, which leads to the precipitation of crystalline-phase LiI on quenching. With the addition of P₂S₅, the fraction of the crystalline phase of LiI markedly decreases and is barely observed at x = 0.17, indicating that, with a small amount of P₂S₅, the glass structure can better dissolve and integrate the LiI. The generated I⁻ in the glass structure can therefore contribute to a weaker coulombic attraction for Li⁺. However, at high P₂S₅/(P₂S₅ + B₂S₃) ratios of x = 0.75 and 0.80, the peak of crystalline LiI gradually appears again, suggesting that excessive P₂S₅ counteracts on the integration of LiI and would impede the Li⁺ diffusion. The Raman spectrum of P₂S₅-free electrolyte (x = 0) (b,c) shows the presence of only the [B_mS_n] group: the bands around 394 cm⁻¹ and 433 cm⁻¹ correspond to the B–S breathing mode of [BS₃] groups³⁴; the bands around 312 cm⁻¹, 497 cm⁻¹ and 763 cm⁻¹ correspond to the vibrations of metathioborate [B₃S₆], thiopyroborate [B₂S₅] and [BS₄] groups, respectively^34,53. With x increased to 0.11 and 0.17, a strong Raman peak corresponding to the [PS₄] group appears at 420 cm⁻¹ (ref. ³⁵), along with decrease of the [B_mS_n] peaks. Notably, at x = 0.29, two more peaks appear at 386 cm⁻¹ and 407 cm⁻¹, corresponding to the polyanionic groups of [P₂S₆]_hypo and [P₂S₇], respectively^34,35. Further, when x = 0.50, the Raman peak corresponding to [PS₄] greatly fades, whereas those of [P₂S₆]_hypo and [P₂S₇] groups are more pronounced, along with the appearance of a new peak at 582 cm⁻¹ ascribed to [P₂S₆]_hypo (refs. ^54,55). Here the two peaks at 386 cm⁻¹ and 582 cm⁻¹ represent the symmetric and asymmetric stretching modes of the [P₂S₆]_hypo group, respectively (d)⁵⁴. As x further increases to 0.67, the [PS₄] peak is no longer observed and a new peak appears at 425 cm⁻¹, along with two subtle peaks at 315 cm⁻¹ and 368 cm⁻¹, corresponding to the [P₂S₆]_meta group^54,55. Further increasing the fraction of P₂S₅ to x = 0.75 and 0.80 leads to a much higher intensity of the [P₂S₇] peak, along with the peaks of [P₂S₆]_meta and [P₂S₆]_hypo. On the basis of these results, we can clearly observe two transitions of the local structures in the LBPSI electrolytes as x increases: (1) the appearance of the [PS₄] group at x = 0.11; (2) the gradual disappearance of the [PS₄] group, along with the dominance of the [P₂S₆] and [P₂S₇] groups at x = 0.50. By associating the local structural changes with the evolution of ionic conductivity (Fig. 1b), we can draw the following conclusions: at x = 0.11, the appearance of the [PS₄] group substantially enhances the ionic conductivity, accounting for the high conductivity of more than 1 mS cm⁻¹ at x = 0.11, 0.17 and 0.29. We believe that the [PS₄] group promotes disruption of the large B–S network and formation of island-like structures, thereby enhancing the Li⁺ ion transport. The fragmentation is probably because of the different coordination preference of P (fourfold coordinated) and B (threefold coordinated), that is, the incorporation of P thermodynamically breaks the threefold coordinated B–S network. As x increases to 0.50, the appearance and dominance of [P₂S₆] and [P₂S₇] polyanions decreases the ionic conductivity, which ultimately reduces to 10⁻³ mS cm⁻¹ at x = 0.80, which is probably because of the highly charged polyanions exhibiting higher coulombic attraction for Li⁺ ions than the [PS₄] group³⁵. Further, as shown in ³¹P MAS NMR spectra (e), the resonances around 83 ppm, 99 ppm and 108 ppm can be ascribed to the [PS₄], [P₂S₇] and [P₂S₆] groups, respectively³⁴. It is clear that the resonance of [PS₄] appears in the electrolytes with x = 0.17 and 0.29. Further, the resonance of polyanionic [P₂S₇] and [P₂S₆] groups appear in the electrolyte with x = 0.29 (and not in the one with 0.17), which is in accordance with the Raman results. Altogether, in LBPSI, we find evidence for a more fragmented network (which should provide more free volume) and effective inclusion of I⁻, which results in higher ionic conductivity and easier breaking of the bond between Li⁺ and the anionic framework ligands at the electrode–electrolyte interface.

a, Differential capacity (dQ/dV) plots of the ASSLSB fitted with LBPSI with varied charging rates from 2C to 35C and a fixed discharging rate of 1C at 30 °C (voltage profiles shown in Fig. 2b; sulfur loading: 1.0 mg cm⁻²). b,c, The discharge–charge voltage profiles at varied rates for the first several activation cycles (shown in Fig. 2a) of the cells with LBPSI (b) and Li_5.5PS_4.5Cl_1.5 (c). d,e, The discharge–charge voltage profiles of the ASSLSB fitted with the Li_5.5PS_4.5Cl_1.5 electrolyte with charging rates from 2C to 35C and a fixed discharging rate of 1C at 30 °C (d) and the corresponding differential capacity (dQ/dV) plots (e) (sulfur loading: 1.0 mg cm⁻²). On the basis of the dQ/dV plots, we can observe that charging and discharging of the cell fitted with LBPSI exhibits distinct redox peaks when charged at 2C to 35C rates, indicating faradaic processes; in contrast, the cell fitted with Li_5.5PS_4.5Cl_1.5 exhibits very small oxidation peaks when charged at rates exceeding 15C. f–h, The specific capacity (f), discharge–charge voltage profiles (g) and dQ/dV plots (h) of the cell using LBPSI as the active material (containing no sulfur in cathode; mass based on LBPSI), measured at the same areal current density as that for the ASSLSB in Fig. 2a (discharge current density is 1.7 mA cm⁻²). The loading of LBPSI is 1.9 mg cm⁻², close to the amount of LBPSI used in the cathode of an ASSLSB. In the dQ/dV plots, we can observe distinct oxidation and reduction peaks owing to the redox behaviour of the LBPSI electrolyte itself. It should be noted that a small but constant fraction (7.8–8.9%) of the capacity comes from the fully reversible redox behaviour of the LBPSI electrolyte at all investigated rates, as quantified by the cell using LBPSI as the only active material. i, The dQ/dV plots of the ASSLSB fitted with LBPSI at 60 °C, with varied charging rates from 5C to 150C and a fixed discharging rate of 2C (sulfur loading: 1.05 mg cm⁻²). j, The dQ/dV plots of the ASSLSB fitted with Li_5.5PS_4.5Cl_1.5 at charging rates from 5C to 150C and a fixed discharging rate of 2C at 60 °C. k,l, The discharge–charge voltage profiles (k) and dQ/dV plots (l) of the cell with LBPSI as the active material (containing no sulfur in cathode) at 60 °C (discharging current density is 3.5 mA cm⁻²; the loading of LBPSI is 1.9 mg cm⁻²).

a,b, The galvanostatic cycling of the In/InLi|LBPSI|In/InLi symmetric cell (In/InLi foil as both electrodes) with current increasing from 0.26 mA cm⁻² to 31.8 mA cm⁻² at 25 °C (the Li plating/stripping time is 30 min for measurements with current density ≤ 3.8 mA cm⁻² and 5 min for those with current density ≥ 6.4 mA cm⁻²) and enlarged section at high currents (reaching the voltage limit) (b). The symmetric cell can indeed operate at higher currents without soft or hard short-circuit at the areal capacity used, but in fact suffers from much greater overpotential than at low currents; this indicates that the In/InLi electrode shows limited reaction kinetics. This result offers some qualitative hints on the overall poor reaction kinetics for In/InLi electrodes. Such a high overpotential at the anode, along with the ohmic resistance from the SE, contributes to the high overpotential of full cells at high currents. c–f, Three-electrode cell measurement to examine the origin of the high overpotential of full cells at high currents. c, The scheme of the three-electrode cell, with a thin Li wire placed in the electrolyte layer close to the In/InLi counter electrode as a reference electrode. d–f, Discharge–charge voltage profiles at the current density of 0.08 mA cm⁻² (d), 8.3 mA cm⁻² (e) and 15.0 mA cm⁻² (f) (sulfur loading of 1.0 mg cm⁻²; 25 °C). At a very low current density of 0.08 mA cm⁻², the potential of the In/InLi electrode stabilizes at around 0.61 V. As the current density increases to 8.3 mA cm⁻², the potential of the In/InLi electrode increases to about 0.75 V (on discharge) and reduces to about 0.46 V (on charge), resulting in a discharging overpotential of 0.14 V and charging overpotential of 0.15 V. On further increasing of the current to 15.0 mA cm⁻², the potential of the In/InLi electrode increases to about 0.90 V (on discharge) and the potential decreases to about 0.31 V (on charge), resulting in a discharging overpotential of 0.29 V and charging overpotential of 0.30 V. Therefore, the reaction overpotential of the In/InLi electrode is high and increases greatly with the current density. Furthermore, there is a large increase of the ohmic resistance when the current is increased from 0.08 mA cm⁻² to 8.3 mA cm⁻². The ohmic resistance is qualitatively recognized by the abrupt voltage change (IR drop) on supplying of the charging/discharging currents. The increased ohmic resistance is the result of the thick layer of the electrolyte separator between the sulfur working electrode and the Li reference electrode, and the moderate ionic conductivity of the SE (2.4 mS cm⁻¹). The high overpotential of the cell at high charging rates is largely because of the In/InLi anode and IR ohmic polarization⁵⁶. g–j, The actual discharge–charge voltage profiles by using the potential versus In/InLi counter electrode of the ASSLSBs fitted with the LBPSI and Li_5.5PS_4.5Cl_1.5 electrolyte at different current densities at 30 °C (g,h) and 60 °C (i,j). k,l, Cycling performance (k) and discharge–charge voltage profiles (l) of ASSLSB fitted with LBPSI electrolyte at a charging rate of 100C and a discharging rate of 2C at 60 °C. The cell was subjected to an activation process by cycling at 1–80C (charging rates) for several cycles. The cell sustains 5,000 cycles with a capacity retention greater than 80.4% (sulfur loading: 1.0 mg cm⁻²). It is clear that the cell experienced marginal voltage degradation over the cycling. The fluctuation in Coulomb efficiency is because of the limited resolution for data acquisition from the equipment with such a short duration of charging.

a, DC polarization curves of sulfur electrodes with an ion-blocking configuration of Ti|sulfur electrode|Ti. b–d, DC polarization curves of sulfur electrodes with an electron-blocking configuration of Ti|Cu|In/InLi|SE|sulfur electrode|SE|In/InLi|Cu|Ti. e, Nyquist plot of the symmetric cell of In/InLi|SE|In/InLi. f–k, Scanning electron microscopy (SEM) images (f–h) and corresponding particle size distributions (derived from the SEM imaging by statistical analysis; scale bars, 10 μm) (i–k) of LBPSI-0.17 (f,i), LBPSI-0.29 (g,j) and Li_5.5PS_4.5Cl_1.5 (h,k). The electronic conductivity of the sulfur cathodes was obtained by measuring the electronic resistance with an ion-blocking cell configuration of Ti|sulfur electrode|Ti (a). The room-temperature electronic conductivities for the three sulfur cathodes prepared with LBPSI-0.17, LBPSI-0.29 and Li_5.5PS_4.5Cl_1.5 is 0.031 S cm⁻¹, 0.022 S cm⁻¹ and 0.042 S cm⁻¹, respectively. The ionic conductivities of the sulfur cathodes was obtained by measuring the total ionic resistance with an electron-blocking configuration of Ti|Cu| In/InLi|SE|sulfur electrode|SE|In/InLi|Cu|Ti (ref. ²¹) (b–d). The contributions of the SE and In/InLi electrode are measured to be 23 Ω and 0.9 Ω with the equivalent circuit fitting of the symmetric cell Ti|Cu|In/InLi|SE|In/InLi|Cu|Ti (e). The room-temperature ionic conductivity for the three sulfur cathodes prepared with LBPSI-0.17, LBPSI-0.29 and Li_5.5PS_4.5Cl_1.5 after subtracting the contributions of SE and In/InLi electrode is 0.56 × 10⁻⁴ S cm⁻¹, 0.39 × 10⁻⁴ S cm⁻¹ and 0.41 × 10⁻⁴ S cm⁻¹, respectively. The three composite sulfur cathodes exhibit similar ionic and electronic conductivities. The SEM images and particle-size distributions (derived from the SEM imaging by statistical analysis) (f–k) show that the mean particle sizes of LBPSI-0.17, LBPSI-0.29 and Li_5.5PS_4.5Cl_1.5 are somewhat close at 2.13, 2.34 and 2.18 μm, respectively. Likewise, the three electrolytes all exhibit wide particle-size distribution over 0–10 μm, with a concentration in the range 0.5–5.0 μm.

a,b, The cycling performance with capacity and Coulomb efficiency of the cell prepared with a LBPSI–C composite cathode (wt% LBPSI:wt% C = 8:2) containing no sulfur (0.25 mA cm⁻²; LBPSI loading of 6.5 mg cm⁻²) (a) and the corresponding differential capacity plots (dQ/dV) (b). c, The CV curves of the cell for the first five cycles. The cell in a–c is configured as In/InLi|LBPSI|LBPSI–C. The long-term cycling performance, dQ/dV plots and CV curves altogether show the high stability and reversibility of the redox reaction of LBPSI within the potential window of interest. d, The dQ/dV plots of the cell prepared with Li_5.5PS_4.5Cl_1.5 as the active material in the cathode (wt% Li_5.5PS_4.5Cl_1.5:wt% C = 8:2), configured as In/InLi|Li_5.5PS_4.5Cl_1.5|Li_5.5PS_4.5Cl_1.5-C for the first, second and fiftieth cycles; the dQ/dV plots of first and fiftieth cycles show notable differences in terms of the peak positions and intensity, indicating irreversible decomposition of Li_5.5PS_4.5Cl_1.5 after the first charging process, which may cause some permanent damage to the interface in a sulfur cell. e,f, First-cycle discharge–charge voltage profiles of the cells with sulfur cathodes at a discharging/charging rate of 0.1C (e) and the corresponding differential capacity (dQ/dV) (f) (25 °C), which shows the presence of LBPSI redox that occurs in the cell at a slightly higher voltage than that of Li₂S/S, whereas the Li_5.5PS_4.5Cl_1.5 redox, if any, occurs with a very much overlapped voltage range with that of Li₂S/S. g,h, In situ EIS studies on a sulfur cell fitted with LBPSI for the first charge (after the first discharge process) and following discharge at 0.1C: Nyquist plots at representative states of charge/discharge (g) and the corresponding distribution of relaxation time analysis (h). The in situ EIS study shows that, when charged to 2.8 V, a new semicircle at the medium frequencies appears, indicating that a new interface forms, which is herein tentatively ascribed to the surficial oxidation of LBPSI forming a surface layer. The corresponding distribution of relaxation time analysis that mathematically converts the frequency-dependent EIS spectra to relaxation-based functions γ(τ) (ref. ⁵⁷), shows the appearance of relaxation distribution (γ) at the characteristic time constant (τ) of 10⁻³–10⁰ s and further confirms the formation of a new interface. The reversibility of the redox behaviour of LBPSI is indicated by the observation of the additional interface disappearing during discharge.

a, Sulfur K-edge XANES spectra of the sulfur cathode fitted with LBPSI electrolyte performed ex situ at different discharge/charge states, along with the sulfur standards for references. b, Ex situ Raman spectra of the pristine cathode and the cathode charged to 3.2 V of the cell using LBPSI as the active material (containing no sulfur) in the cathode. The Raman spectra of LiI, I₂ and milled I₂/LiI ((I₂+LiI)-BM) are measured as references. c,d, The I 3d (c) and S 2p (d) XPS data of the pristine LBPSI material, the pristine cathode, the electrode charged to 3.2 V and the electrode subsequently discharged to 1.4 V in the cell configured as In/InLi|LBPSI|LBPSI–C (dotted lines, raw data; black lines, fitted total; coloured lines, fitted components). e, The I 3d spectra of the pristine electrode and the electrode charged to 3.2 V in the cell for better visibility of the signal shift. We performed ex situ sulfur K-edge XANES measurements and observe that the sulfur cathode using LBPSI electrolyte discharges with the formation of Li₂S_x intermediates and that Li₂S/Li₂S_x is completely converted to sulfur on charge. The S/KB and Li₂S/KB samples prepared by mixing the respective sulfur species with KB were used as references. The S/KB shows two features at 2,472.6 eV (the white line) and 2,480.0 eV; the Li₂S/KB shows three features at 2,473.6, 2,476.2 and 2,484.0 eV. The spectrum of the pristine cathode shows the features of S⁰ at 2,472.6 and 2,480.0 eV. When discharged to 1.4 V, the features of Li₂S at 2,473.6, 2,476.2 and 2,484.0 eV as well as that for Li₂S_x at 2,471.8 eV appear^5,58. After the subsequent charging, the features corresponding to Li₂S or Li₂S_x disappear and S⁰ features at 2,472.6 and 2,480.0 eV emerge again. This indicates that the charging has fully converted Li₂S (and Li₂S_x) back to sulfur. When discharged to 1.4 V again, the features of S⁰ disappear and the features of Li₂S and Li₂S_x appear. This demonstrates the reversibility of the SSSRR during charging and discharging. In the Raman spectra of the LBPSI–C electrodes (b), compared with the pristine electrode, for the electrode charged to 3.2 V, a broad band appears at 100–200 cm⁻¹. The peak at around 181 cm⁻¹ is attributed to I₂ and the peaks around 110 cm⁻¹ and 167 cm⁻¹ are attributed to I₃⁻ (ref. ⁵⁹). For a clear reference of the assignment of these peaks, we measured the Raman spectra of crystalline LiI, crystalline I₂ and a milled mixture of I₂/LiI (which forms LiI₃). We can see that the peak around 181 cm⁻¹ is clearly attributed to the vibrations of I₂ and the peaks around 110 cm⁻¹ and 167 cm⁻¹ are attributed to the vibrations of I₃⁻. From the I 3d XPS spectra (c,e), we can clearly observe the signal shift of the I 3d spectra to higher binding energy on charging the cell to 3.2 V (e), confirming the I on the surface of the SE being present with a higher valence state than that in LBPSI (I⁻). By a closer observation of the fitted I 3d spectra, the pristine LBPSI electrolyte and LBPSI–C electrode show anionic I⁻, as evidenced by the I 3d doublet signals at 618.8 eV (3d_5/2 energy)⁴¹ (c). On charging to 3.2 V, the oxidized iodine species are clearly observed (tentatively assigned to I₂/I₃⁻)⁴², which locates at around 619.9 eV (3d_5/2 energy). Furthermore, after subsequent discharging to 1.4 V, the signals of oxidized iodine species disappear, demonstrating the reversible oxidation/reduction of iodine species. For the S 2p spectra (d), the pristine LBPSI electrolyte and electrode contain [PS₄] units (and [BS₄], that is, non-bridging sulfur) and P–S–P (and B–S–B, that is, bridging sulfur) signals, with characteristic S 2p signals located at 161.5 eV and 162.5 eV (2p_3/2 energy)^43,60. The LBPSI electrode charged to 3.2 V contains oxidized sulfur (S⁰), which is evident by the characteristic S 2p doublet signals at around 163.6 eV (2p_3/2 energy)⁶¹. After discharge to 1.4 V, the signal of oxidized sulfur (S⁰) disappears and signals of bridging sulfur and non-bridging sulfur appear, which demonstrate the reversible oxidation/reduction of sulfur species.

a–c, Evidence of diffusion of I₂ along the SE particle surface. a, The LBPSI electrolyte pellet on direct contact with I₂ for 5 min. b, The LBPSI electrolyte pellet placed at the I₂ bottle cap for indirect contact with I₂ for 3 days. c, EDX mapping of the vertical cross-section of Li_5.5PS_4.5Cl_1.5 electrolyte pellet after indirect contact with I₂ for 3 days. d, Pictures showing the process of mixing Li₂S and I₂ in anhydrous ethanol. e,f, XRD pattern (e) and Raman spectrum (f) of the precipitated product. g,h, Characterization of the reaction of Li₂S and I₂ by a gentle ball-milling at 100 rpm for 2 h, with the XRD pattern (g) and Raman spectrum (h) of the product. i,j, Characterization of the reaction of Li₂S and I₂ by simple hand grinding in the mortar (5 min), with the XRD pattern (i) and Raman spectrum (j) of the product. The insets of j are visual pictures of the mixture of Li₂S and I₂ before and after hand grinding in the mortar. The diffusion and permeation of I₂ into the electrolyte pellet is shown in a,b. On direct contact of the LBPSI SE pellet with the iodine (brown), I₂ sublimes and diffuses very quickly over the LBPSI pellet surface within 5 min; the surface of the electrolyte is covered with brown iodine (a). The LBPSI pellet was placed at the cap of a bottle that contains I₂ and the bottle was sealed with grease to prevent escape of iodine vapour through the gap (that is, indirect contact; b). After 3 days without direct contact with I₂, the outer/top surface of the electrolyte pellet turns brown, indicating penetration of the iodine vapour across the SE pellet (b), which further proves the diffusion of I₂ along the grain boundaries and throughout the SE pellet. Note that these experiments were conducted in an Ar-filled glovebox and not in vacuum. Further, we performed the same indirect contact experiment with the iodine-free Li_5.5PS_4.5Cl_1.5 electrolyte. The cross-section of the pellet was subjected to scanning electron microscopy imaging and EDX analysis (c). We can see that the element I is uniformly distributed, along with the P, S and Cl, over the imaged area (several hundred µm; c). To mimic the reactions between Li₂S and I₂, equal molar amounts of Li₂S and I₂ were first added separately into anhydrous ethanol, the solutions of which are, respectively, colourless and purple (step 1; d) and, subsequently, the two solutions of Li₂S and I₂ were mixed. Along with the change of colour (pale yellow, step 2), we observed precipitation, which is confirmed to consist of crystallized LiI along with S (as shown by the XRD pattern in e). The Raman spectrum further shows a strong signal corresponding to S (f). Therefore, the products of Li₂S reacting with I₂ are confirmed to be S and LiI. With an effort to further mimic the condition of the reaction occurring within the all-solid-state Li–S cells, the reaction was performed in solid state. Specifically, equimolar amounts of Li₂S and I₂ were mixed by a gentle ball-milling at 100 rpm for 2 h (to prevent overheating of the jar) and LiI and S are formed on solid–solid mixing of the two materials (based on XRD pattern; g). As shown in h, the Raman spectrum exhibits strong S signals (154 cm⁻¹, 219 cm⁻¹ and 473 cm⁻¹), I₂ (signal around 181 cm⁻¹) and I₃⁻ (110 cm⁻¹ and 167 cm⁻¹)⁵⁹. In fact, even by simple hand grinding in the mortar, the two materials react very quickly (j); during 5 min of grinding, the two materials react to form LiI (based on XRD pattern; i), S and I₃⁻ (Raman spectrum; j). The feasibility of the chemical reaction between I₂ and Li₂S that we experimentally confirmed above is strong evidence for us to conclude that the occurrence of redox mediation in the cell at high charging rates are not cascade reactions (that is, occurring sequentially without interplay). Further, the 0.3–0.4 V higher potential for iodine redox over that of sulfur redox (Fig. 3a) serves as the theoretical basis for a chemical reaction between the two couples. This is distinct from the behaviour of sulfur cells using Li_5.5PS_4.5Cl_1.5, in which the SE redox is expected to occur simultaneously with sulfur redox owing to the lack of potential difference (Fig. 3b).

Li_5.5PS_4.5Cl_1.5 was used as the electrolyte separator for the ASSLSBs for its high ionic conductivity. a,b, Powder XRD patterns of the Li₃PS₄ (a) and 70Li₃PS₄·30LiI (b). c, Nyquist plots of impedance spectra of Li₃PS₄ and 70Li₃PS₄·30LiI for the measurement of the respective ionic conductivity. d, The first-cycle discharge–charge voltage profile of ASSLSBs at 0.05C. e, Charging-rate performance with varied charging rates (labelled) and a constant discharging rate of 1C. Cells were subjected to an activation process by charging at progressively increasing rates of 0.05–0.8C for several cycles before being charged at 2C (sulfur loading: 1.0 mg cm⁻²). f,g, Discharge–charge voltage profiles of ASSLSBs with charging rates from 2C to 30C using Li₃PS₄ (f) and 70Li₃PS₄·30LiI (g) as the catholyte, respectively. h, Nyquist plot for the measurement of the ionic conductivity of LiI (the room-temperature ionic conductivity is measured to be 0.7 × 10⁻⁶ S cm⁻¹). On discharge, the iodine species (more precisely, the LiI) can serve as ionically conducting facilitator in the cathode (h) because the as-formed LiI would remain in close contact to the sulfur that it has redox-mediated, and as a moderate ionic conductor (0.7 × 10⁻⁶ S cm⁻¹; h) can thus assist in bridging the ionic transport path and enhance the discharge kinetics. We have further demonstrated the much-improved SSSRR kinetics enabled by another iodine-containing sulfide electrolyte (70Li₃PS₄·30LiI), which highlights the universality of the iodine-based surficial redox-mediating strategy. Li₃PS₄ and 70Li₃PS₄·30LiI electrolytes were synthesized to investigate whether the iodine-based redox mediating takes effect in these electrolytes. The two electrolytes are glass ceramics, of which the crystalline phase herein is confirmed to be Li₃PS₄ (Pnma phase)⁶² and Li₄PS₄I (P4/nmm phase)⁶³ by XRD, respectively (a,b). The Li₃PS₄ and 70Li₃PS₄·30LiI electrolytes show similar room-temperature ionic conductivity of 0.17 mS cm⁻¹ and 0.13 mS cm⁻¹, respectively (c). We prepared sulfur cathodes using Li₃PS₄ and 70Li₃PS₄·30LiI electrolytes in the same way as for those using LBPSI. The Li_5.5PS_4.5Cl_1.5 electrolyte was used as the separator for the cells. The cell with 70Li₃PS₄·30LiI, when charged at 0.05C, shows a further voltage plateau at 2.7 V, which is consistent with the LBPSI cell and is the result of the redox of iodine (d). The rate performance of the sulfur cathode with 70Li₃PS₄·30LiI is superior to that using Li₃PS₄. The cell with 70Li₃PS₄·30LiI exhibits a high specific capacity of 1,300 mAh g⁻¹ at a charging rate of 2C and maintains 660 mAh g⁻¹ at 10C (discharging rate constant at 1C). At low rates of 0.05C to 0.5C for the first several cycles, the cell fitted with Li₃PS₄ shows a higher capacity than the cell using 70Li₃PS₄·30LiI. However, at charging rates of 2C and 10C, it shows lower capacities of 1,113 mAh g⁻¹ and 422 mAh g⁻¹, respectively, than that using 70Li₃PS₄·30LiI in the cathode. Therefore, we consider that the iodine-based redox mediation effect should apply to a variety of SEs containing iodine on the surface.

a,b, Discharge–charge voltage profiles of the cell fitted with Li_5.5PS_4.5Cl_1.5 electrolyte (a) and LBPSI electrolyte (b) at a discharging/charging rate of 0.2C over different numbers of cycles (sulfur loading: 1.5 mg cm⁻²). c, Discharge–charge voltage profiles of the cell fitted with LBPSI electrolyte at a discharging/charging rate of 2C over 5,000 cycles (sulfur loading: 1.1 mg cm⁻²). d, Discharge–charge voltage profiles of the cell fitted with LBPSI electrolyte at a discharging/charging rate of 5C over 25,000 cycles (sulfur loading: 1.1 mg cm⁻²). e, Discharge–charge voltage profiles of the cell fitted with Li_5.5PS_4.5Cl_1.5 electrolyte at a discharging/charging rate of 5C over 10,000 cycles (sulfur loading: 1.1 mg cm⁻²) as a comparison. f, The evolution of stack pressure of ASSLSBs during discharge–charge cycles at 30 °C with LBPSI and Li_5.5PS_4.5Cl_1.5 electrolyte, respectively. Δp represents the change of stack pressure within one full discharge–charge cycle and \(\overline{\Delta p}\) is the average over five cycles. Both cells show apparent stack-pressure variations during cycling, indicating volume expansion and contraction of the electrodes; however, the average amplitude of pressure variation within one cycle for the cell using LBPSI is much lower than that with Li_5.5PS_4.5Cl_1.5 (\({\overline{\Delta p}}_{{\rm{LBPSI}}}=0.45\,{\rm{MPa}}\); \({\overline{\Delta p}}_{{\rm{LiPSCI}}}=1.08\,{\rm{MPa}}\)), which is indicative of a more profound buffering of the microscopic volume change in the case of LBPSI. g,h, Long-term cycling of the cell fitted with LBPSI at a discharging/charging rate of 8C (sulfur loading: 1.0 mg cm⁻²) (g) and corresponding discharge–charge voltage profiles at 25 °C (h). The cell was subjected to an activation process by cycling at 0.05–7C for several cycles. The cell exhibits a high capacity of 744 mAh g⁻¹ at 8C, with 78% capacity retention over 10,000 cycles.

a, Discharge–charge voltage profiles of the cell fitted with LBPSI at a discharging/charging rate of 1C over 1,000 cycles (sulfur loading: 3.0 mg cm⁻²). b, Discharge–charge voltage profiles of the cell with a high sulfur fraction of 45 wt% in the cathode, fitted with LBPSI at a 0.2C rate (sulfur loading: 1.9 mg cm⁻²) and 25 °C. c,d, Long-term cycling of the cell fitted with LBPSI with a high sulfur loading of 6.3 mg cm⁻² (c) and corresponding discharge–charge voltage profiles at 25 °C at a discharging/charging rate of 0.1C (d). e,f, Discharge–charge voltage profiles of the cell fitted with LBPSI for the first two activation cycles (e) and discharge–charge voltage profiles at a discharging/charging rate of 15C over 10,000 cycles (f) at 60 °C (sulfur loading: 1.0 mg cm⁻²). g, Discharge–charge voltage profiles of the cell fitted with Li_5.5PS_4.5Cl_1.5 electrolyte at a discharging/charging rate of 15C over 10,000 cycles (sulfur loading: 1.0 mg cm⁻²) as a comparison. h,i, Low-temperature performance of ASSLSBs at −20 °C (sulfur loading: 1.1 mg cm⁻²). h, Charging-rate performance of cells using LBPSI and Li_5.5PS_4.5Cl_1.5 electrolytes with varied charging rates (labelled) and a constant discharging rate of 0.2C. The first cycle was operated at room temperature at 0.05C for activation. At the low rates of 0.05–0.20C, the cell with Li_5.5PS_4.5Cl_1.5 shows a slightly higher capacity than one with LBPSI at −20 °C, which is probably because of the lower ionic conductivity of LBPSI. At a charging rate of 0.3C, the two cells show comparable specific capacity of about 820 mAh g⁻¹. Beyond that rate, the cell with LBPSI exhibits higher capacities, that is, a specific capacity of 652 mAh g⁻¹ at a charging rate of 0.5C, maintaining 466 mAh g⁻¹ at 1C, whereas the cell with Li_5.5PS_4.5Cl_1.5 exhibits a specific capacity of 514 mAh g⁻¹ at 0.5C and maintains only 295 mAh g⁻¹ at 1C. i, Long-term cycling of the cells at a 1C charge/discharge rate at −20 °C. The cell with LBPSI electrolyte shows stable cycling with 406 mAh g⁻¹ at 200 cycles and almost no capacity fading over 1,000 cycles. By contrast, the cell with Li_5.5PS_4.5Cl_1.5 shows only 128 mAh g⁻¹ at 200 cycles, which is maintained over 1,000 cycles.