Lithium-ion-conductive sulfide polymer electrolyte with disulfide bond-linked PS4 tetrahedra for all-solid-state batteries

Due to their high conductivity and interface formability, sulfide electrolytes are attractive for use in high energy density all-solid-state batteries. However, electrode volume changes during charge-discharge cycling typically cause mechanical contact losses at the electrode/electrolyte interface, which leads to capacity fading. Here, to suppress this contact loss, isolated PS43- anions are reacted with iodine to prepare a sulfide polymer electrolyte that forms a sticky gel during dispersion in anisole and drying of the resulting supernatant. This polymer, featuring flexible (–P–S–S–)n chains and enhanced solubility in anisole, is applied as a lithium-ion-conductive binder in sheet-type all-solid-state batteries, creating cells with low resistance and high capacity retention. Solid polymer electrolytes that are resistant to volume changes during charge-discharge cycling are important for all-solid-state batteries. Here, a sulfide polymer electrolyte is fabricated by reacting PS43− anions with iodine, obtaining cells with low resistance and high capacity retention.

S ince their commercialization in the 1990s, Li-ion batteries (LIBs) have been widely used in consumer electronics and have recently found large-scale energy storage applications in electric vehicles and smart grids [1][2][3] . However, the currently used LIBs contain flammable liquid organic electrolytes and therefore pose safety risks 4 . To mitigate these risks, one can replace organic liquid electrolytes with inorganic solid electrolytes, which feature higher thermal stability and are not susceptible to leakage 5 . This replacement affords high-energy-density all-solid-state batteries (ASSBs), which have attracted much attention, as exemplified by attempts to use solid electrolytes in combination with high-voltage cathodes 6,7 , high-capacity sulfur electrodes 8,9 , and Li metal anodes [10][11][12][13] .
Among the various types of solid (e.g., sulfide-, oxide-, hydride-, and halide-based) electrolytes developed to date 5,[14][15][16] , the sulfidebased ones feature high conductivities and interface formability, and are therefore particularly well suited for ASSBs. In particular, sulfide electrolytes with Li 10 GeP 2 S 12 - 17,18 , argyrodite- 19,20 , and Li 7 P 3 S 11type 21,22 crystal structures have high conductivities (>10 −3 S cm −1 ) comparable to those of liquid electrolytes. Sulfide electrolytes are easily deformed by pressing at room temperature 23 , allowing one to form favorable electrode/electrolyte interfaces with high contact areas, and ensure sufficient ion conduction. However, the volume changes of the electrode active materials during charge/discharge induce local contact losses at ASSB electrode/electrolyte interfaces 24,25 . Typical sulfide electrolytes contain Li + ions and isolated anions such as PS 4 3− . Previously, sulfide polymers with (-P-S-S-) n units were predicted to electrochemically form at cathode/electrolyte interfaces during charging (delithiation) 26,27 . In contrast, thermodynamic calculations and structural analyses of the related interfaces indicate that such polymers undergo disproportionation into P 2 S 7 4− and elemental sulfur [28][29][30] . To date, no agreement has been reached, and the related discussions are still ongoing.
The improvement of existing processing technologies is vital for ASSB commercialization, with wet slurry coating recognized as a practical and scalable way of fabricating sheet-type ASSBs 31 . In this process, binders help to adhere the composites (containing the electrode active materials, solid electrolytes, and carbon-based conductive additives) to the current collector, make the composite sheets flexible, and allow roll-to-roll manufacturing. Both conventional LIBs and ASSBs use organic polymers, such as polyvinylidenefluoride 32 , styrene-butadiene-styrene copolymers 33 , and acrylonitrile-butadiene copolymers 34,35 as binders. However, the insulating nature of these binders hinders ionic and electronic conduction in ASSBs. To mitigate this problem, one can use volatile binders that can be thermally removed to afford binderless ASSBs 36 . Another strategy is to use Li + -conductive polymer binders based on organic polymer electrolytes or hybrids of ionic liquid and organic polymer electrolytes [37][38][39] . Sulfide polymers will be a promising Li + -conductive binder, and thus reduce hetero interfacial resistances in ASSBs.
Herein, we polymerized isolated PS 4 3− anions by reacting them with I 2 (Fig. 1). The reaction of two PS 4 3− anions with I 2 afforded a dimer comprising two PS 4 tetrahedra bridged with a disulfide (S-S) bond, whereas higher I 2 loadings afforded a potentially flexible chain polymer with a (-P-S-S-) n structure. We demonstrated the mechanochemical and liquid-phase syntheses of the sulfide polymer. This polymer was compared with previously reported structures like P 2 S 7 4− to provide useful suggestions for the design of cathode/electrolyte interfaces in ASSBs and was used as a binder in sheet-type ASSBs. LiI, a by-product of the above reaction, is also a Li-ion conductor 40 and can enhance the ionic conductivity of sulfide electrolytes 41,42 which suggests that sulfide polymer-LiI composites hold great promise as solid electrolytes.

Results
Mechanochemical synthesis of sulfide polymers and their characterization. Li 3 PS 4 was prepared by ball milling a mixture of Li 2 S and P 2 S 5 and was further ball-milled with I 2 to afford sulfide polymers. We synthesized three kinds of sulfide polymers with different molar ratios; Li 3 PS 4 -I 2 (2:1), Li 3 PS 4 -I 2 (4:3), and Li 3 PS 4 -I 2 (1:1). Powder X-ray diffraction (PXRD) patterns acquired after the reaction with I 2 revealed the presence of LiI (PDF#01-075-5397) ( Fig. 2(a)), and thus indicated the occurrence of Li 3 PS 4 polymerization. The Raman spectrum of the amorphous polymerization product featured two main peaks at 477 and 390-420 cm −1 (Fig. 2(b)). The former peak, ascribed to S-S bond stretching, gained intensity with increasing I 2 loading, which reflected the concomitant increase in the extent of S-S cross-linking. Although elemental S exhibits a similar S-S stretching peak at 473 cm −1 43 , the peak at 477 cm −1 had a different origin, as it did not lose intensity after the polymer was washed with toluene to dissolve elemental S ( Supplementary Fig. 1). The peak at 390-420 cm −1 was assigned to the stretching of P-S bonds and shifted to lower wavenumbers with increasing I 2 loading (cf. the P-S peak of pristine Li 3 PS 4 at 421 cm −1 ), which was ascribed to the progressive polymerization of Li 3 PS 4 through S-S bridging. Supplementary Table 1 lists the Raman shifts of selected lithium thiophosphate anions [44][45][46][47] , viz. P 2 S 7 4− (pyro-thiodiphosphate, two corner-sharing PS 4 tetrahedra), P 2 S 6 4− (hypo-thiodiphosphate with P-P bond), P 2 S 6 2− (meta-thiodiphosphate, two edge-sharing PS 4 tetrahedra), and PS 3 − (meta-thiophosphate, chain of corner-sharing PS 4 tetrahedra). The Raman shift of Li 3 PS 4 -I 2 (2:1), which featured two S-S bond-linked PS 4 tetrahedra (P 2 S 8 4− ) was similar to that of P 2 S 7 4− , and that of Li 3 PS 4 -I 2 (1:1), which featured a chain structure (PS 4 − ), was similar to that of PS 3 − . The structural difference between Li 3 PS 4 -I 2 polymers and the known anions of P 2 S 7 4− or PS 3 − is that in the former case, the PS 4 tetrahedra are linked through S-S bonds, whereas in the latter case, they share the corner S atoms. Because this difference is small, the corresponding Raman shifts were similar. Similar behavior has been reported for other thiophosphates such as Li 3 PS 4+n and P 4 S 10+m 48,49 .
Li 3 PS 4 -I 2 (1:1) featured a particle size of 1-10 µm ( Fig. 3(a)) and had good deformability (as did other sulfide electrolytes), affording a highly dense compact without clear original particle shapes after pressing at room temperature ( Fig. 3(b)). The conductivities of Li 3 PS 4 -I 2 (2:1), (4:3), and (1:1) were 2.9 × 10 −4 , 1.7 × 10 −4 , and 3.0 × 10 −5 S cm −1 , respectively, i.e., these values decreased with progressive polymerization (Fig. 3(c) and Supplementary Fig. 2). Anisole, an aprotic solvent with a comparatively low donor number, is used as the binder solvent of sulfide electrolyte-based sheet-type ASSBs, as it does not attack these electrolytes 36 . Figure 4(a) shows that the solubility of Li 3 PS 4 -I 2 polymers in anisole increased with increasing I 2 loading and was highest for Li 3 PS 4 -I 2 (1:1). On the other hand, Li 2 S-P 2 S 5 solid electrolytes (ortho-, pyro-, and meta-thiophosphates) exhibited low solubility in anisole ( Supplementary Fig. 3). Upon drying, the supernatant of the Li 3 PS 4 -I 2 (1:1) solution in anisole afforded a sticky gel, in line with the formation of long chains ( Fig. 4(b)). The Raman spectrum of ball milling-prepared Li 3 PS 4 -I 2 (1:1) was compared with that of the dried supernatant of the corresponding anisole solution (Fig. 4(c)). Both the spectra featured similar S-S and P-S peaks around 477 and 390 cm −1 , respectively, indicating that these components were similar. However, the P-S peak of the ball milling-prepared Li 3 PS 4 -I 2 (1:1) was broader than that of the polymer recovered from the anisole solution, indicating that a greater variety of structures was formed in the former case. Figure 4(d) presents the 31 P NMR spectra of the two kinds of Li 3 PS 4 -I 2 (1:1) polymers, revealing the presence of three main peaks at 86-89, 100, and 120 ppm. The main peak at 86-89 ppm was tentatively assigned to the chain of disulfide-bond-linked PS 4 tetrahedra, whereas the other peaks could not be assigned. Supplementary Table 1 lists the 31 P NMR data for the known anions of lithium thiophosphates, revealing that a subpeak at 120 ppm has been observed in the spectrum of PS 3 − 45 . The above unknown peaks may reflect the presence of branched structures or variable-size rings in addition to the regular chains of PS 4 tetrahedra, although further structural analysis is needed to reveal the details.
Liquid-phase synthesis of sulfide polymers. Li 3 PS 4 -I 2 polymers were also prepared by liquid-phase synthesis (LS), i.e., by stirring  Li 3 PS 4 and I 2 in anisole at 60°C for one day. In the case of Li 3 PS 4 -I 2 (1:1), the product was a clear brown solution at room temperature, while for Li 3 PS 4 -I 2 (2:1) and (4:3) (Fig. 5(a)), we observed precipitates that were identified as unreacted Li 3 PS 4 based on their Raman spectra (Fig. 5(b)). Since I 2 is soluble in anisole, the polymerization of Li 3 PS 4 progressed from the surface of the insoluble Li 3 PS 4 . As in the case of ball milling-prepared  Fig. 4). Notably, hexagonal LiI was also observed in the dried supernatant of the anisole solution of ball milling-prepared Li 3 PS 4 -I 2 (1:1). This hexagonal phase was reported to be metastable and undergo a slow structural change to the cubic phase even at room   temperature 50 . However, the fact that hexagonal LiI mainly precipitated upon the drying (at 60°C) of Li 3 PS 4 -I 2 (1:1) polymers isolated from anisole solutions (the solution produced during LS and the supernatant obtained by dispersing ball milling-produced polymer) suggested that the sulfide polymer matrix prohibited the hexagonal-to-cubic transformation. The NMR spectrum of Li 3 PS 4 -I 2 (1:1) prepared by LS was similar to that of the dried supernatant of the anisole solution of the ball milling-prepared polymer, i.e., the Li 3 PS 4 -I 2 polymer with a (-P-S-S-) n chain structure could be also synthesized by LS. The conductivity of Li 3 PS 4 -I 2 (1:1) prepared by LS was 2.3 × 10 −6 S cm −1 after drying at 200°C (Supplementary Fig. 5), which was lower than the conductivity of ball milling-produced polymer.
Binder applications of sulfide polymers. The Li 3 PS 4 -I 2 polymer was used as a binder for ASSBs. Sheets containing Li 3 PS 4 as a solid electrolyte were prepared by coating a slurry comprising this thiophosphate and ball milling-prepared Li 3 PS 4 -I 2 (1:1) (Fig. 6(a-c) and Supplementary Movie 1) in anisole. The obtained sheets with a thickness of~80 µm could be bent around a 16-mmdiameter cylinder, although Li 3 PS 4 powder was detached from the sheet in the absence of the binder. The conductivity of the Li 3 PS 4 sheet with the sulfide polymer binder (2.3 × 10 −4 S cm −1 , Supplementary Fig. 6) was close to that of the sheet without the binder (3.2 × 10 −4 S cm −1 ).
We also prepared composite electrode sheets (thickness: 40 µm) comprising LiNi 1/3 Mn 1/3 Co 1/3 O 2 (NMC), an argyrodite-type solid electrolyte (Li 6 PS 5 Cl with a conductivity of 2.2 × 10 −3 S cm −1 ), and Li 3 PS 4 -I 2 (1:1) as a binder (Supplementary Fig. 7) and evaluated the performance of the corresponding all-solid-state cell. We adopted the argyrodite-type solid electrolyte in the composite electrode due to its high ionic conductivity in the 10 −3 S cm −1 range, ease of synthesis, and low costs of raw materials, which are advantageous for mass production 51,52 . As a result, typical charge-discharge curves of the NMC active material without obvious side reactions at 2.6-4.3 V (vs. Li/Li + ) were observed ( Fig. 7(a)). The small cell resistance after the first charge indicated that the Li + -conductive binder did not act as a resistive component (Fig. 7(b)). In addition, the above cell featured high cycling performance (Fig. 7(c)), retaining 93.8% of its capacity between cycles 5 and 200. These results suggest that the prepared Li 3 PS 4 -I 2 polymer can be used as an ion-conductive binder for sheet-type ASSBs.

Discussion
Herein, solid electrolytes prepared by I 2 -induced polymerization of Li 3 PS 4 were shown to contain S-S bridges, as indicated by the Raman peak at 477 cm −1 . This peak was retained after the polymer was washed with toluene to remove elemental sulfur and thus truly originated from the S-S linkages in the polymer. The solubility of the sulfide polymer in anisole increased with increasing I 2 loading, which was in contrast with the low solubilities of ortho-, pyro-, and meta-thiophosphate Li 2 S-P 2 S 5 electrolytes. The formation of a sticky gel upon the drying of the Li 3 PS 4 -I 2 (1:1) solution was in line with the formation of chain structures. Notably, the sulfide polymer did not undergo disproportionation into the known thiophosphates and elemental S and was assumed to contain PS 4 tetrahedra cross-linked with S-S bonds. On the other hand, the sulfide polymer was similar to the known thiophosphates with corner-sharing PS 4 tetrahedra in terms of the Raman shifts of P-S bonds and 31 P NMR shifts, which indicated the similarity between the chemical environments of P atoms. The same behavior has been observed in reports on the structural analysis of cathode/electrolyte interfaces 29,53 . Further structural analyses of the sulfide polymer are expected to provide critical insights into the related cathode/ electrolyte interfaces to facilitate their design in ASSBs.
Liquid-phase synthesis is well suited for mass production 54 . The sulfide polymer could be prepared by both liquid-phase and mechanochemical syntheses, although the sulfide polymers prepared by liquid-phase possessed lower conductivities. The polymer structure and the composite state with LiI would influence the conductivities of sulfide polymers. LiI is a poor ionic conductor 40  sulfide electrolytes 41,42 . Although LiI crystal was observed in the XRD patterns, some amount of LiI were considered to be doped in the amorphous sulfide polymers and enhance the conductivities. The ball milling process was assumed to increase doped-LiI in the sulfide polymer, and made the higher conductivity. Removing redundant LiI crystal with poor ionic conductivity and increasing doped-LiI will improve the conductivities of sulfide polymers. The sulfide polymer was used as a binder to prepare sheets comprising Li 3 PS 4 and NMC composites as solid electrolytes. These sheets were bendable, and all-solid-state cells containing NMC composite sheets exhibited low resistance and high capacity retention. The NMC loading (~2 mg) was low compared to the practical value, and it was difficult to prepare thick electrode sheets with >100 µm thickness without cracking. Further works such as designing a molecular weight and a structure of the sulfide polymer will improve the adhesiveness of the binder. Thus, the flexible and ionically conductive sulfide polymer prepared herein is expected to facilitate the development of more robust electrode/electrolyte interfaces in ASSBs.

Methods
Material synthesis. Solid electrolytes were prepared by ball milling appropriate amounts of Li 2 S (Furuuchi Chemical Corp.) and P 2 S 5 (Sigma-Aldrich) at 500 rpm for 20 h using ZrO 2 pots (volume: 45 mL) and balls (diameter: 5 mm). The thus obtained Li 3 PS 4 was polymerized by ball milling with I 2 (Kojundo Chemical Lab. Co.) at 500 rpm for 20 h. Dehydrated heptane (Fujifilm Wako Pure Chemical Corp.) was added in the pots with the starting materials as a medium for ball milling. The heptane medium was removed after ball milling by drying the samples at room temperature. The extent of S-S bridging was determined by I 2 loading. The I 2 -induced polymerization of Li 3 PS 4 and I 2 was also carried out by one-day stirring in anisole at 60°C and a total solid content of 25 wt%. The resulting Li 3 PS 4 -I 2 polymer solution was dried in a vacuum at 60°C. All the samples were synthesized in a dry Ar atmosphere.
Material characterization. PXRD patterns (SmartLab, Rigaku; Cu K α radiation, 2θ = 10−60°, step = 0.02°) were recorded in sealed vessels under dry Ar. Raman spectra (LabRAM HR Evolution, HORIBA) were recorded in the wavenumber range 100-800 cm −1 using an exposure of 1 s 30 times and an excitation laser with a wavelength of 532 nm. For Raman measurements, the samples were sealed in glass tubes filled with dry Ar. Solid-state magic-angle-spinning 31 P NMR (ECZ400R, JEOL) measurements were performed at a spinning speed of 12 kHz. The powdered samples were sealed in spinners in a dry Ar-filled glove box. A 90°3 .8 μs pulse with a recycle delay of 300 s was used, and the observed frequency was 161.944 MHz. Powder morphologies and pellet fracture cross-sections were examined by scanning electron microscopy (SEM; JSM-7800F, JEOL). Sample pellets were prepared by pressing the powders at 333 MPa and room temperature. All SEM samples were prepared in a dry Ar-filled glove box and transferred in sealed vessels.
Preparation of solid electrolyte and electrode composite sheets. The Li 3 PS 4 -I 2 (1:1) polymer synthesized by ball milling was used as a binder. Solid electrolyte sheets were prepared by mixing Li 3 PS 4 powders with the Li 3 PS 4 -I 2 (1:1) polymer in anisole in a weight ratio of 95:5 followed by coating on Al current collector sheets. Electrode composite sheets were also prepared by coating slurries of LiNbO 3coated LiNi 1/3 Co 1/3 Mn 1/3 O 2 (NMC; active material) 55 , argyrodite-type Li 6 PS 5 Cl (solid electrolyte) 56,57 , acetylene black, and the Li 3 PS 4 -I 2 (1:1) polymer in anisole (70:30:5:5, w/w/w/w) on Al current collector sheets. The sheets were dried at 60°C in a dry Ar atmosphere overnight and further dried in a vacuum at 160°C for 3 h. Conductivity measurements. Powders or sheets were put into a polyimide tube with a diameter of 10 mm and sandwiched between two stainless steel rods as an ion-blocking electrode. The cells were pressed at 333 MPa, and conductivities were measured at least three times at room temperature using an AC impedance analyzer (cell test system 1470E, Solartron Analytical; lead wires connected to stainless steel rods) in the frequency range from 10 Hz to 1 MHz. All procedures were conducted under dry Ar.
All-solid-state cell assembly and electrochemical measurements. NMC composite sheets (NMC loading =~2 mg) punched into circles with a diameter of 9.5 mm and a Li 3 PS 4 separator layer (80 mg) were put into a polyimide tube (10 mm in diameter) and pressed together at 296 MPa. A Li-In alloy foil (Li foil: thickness 0.2 mm, diameter 4 mm, and In foil: thickness 0.1 mm, diameter 8 mm) was placed on the surface of the separator side of the bilayer pellet as a counter electrode. The three-layered pellet was sandwiched between two stainless steel rods as a current collector, and a pressure of 111 MPa was applied to assemble the cell. All processes were performed in a dry Ar-filled glove box (DBO-2NKP, Miwa Manufacturing Co., Ltd.). Charge-discharge tests were carried out at 0.071 mA cm −2 in the first five cycles and at 0.14 mA cm −2 in the subsequent cycles at 30°C under dry Ar using an electrochemical measurement device (BTS-2004, Nagano Co.). Cell resistance was measured after the first charge using an AC impedance analyzer (cell test system 1470E, Solartron Analytical). The cells were uniaxially pressed under 75 MPa during the electrochemical measurements.

Data availability
The datasets that support the findings of this study are available from the corresponding authors upon reasonable request.