From Powder to Sheets: A Comparative Electrolyte Study for Slurry-Based Processed Solid Electrolyte/Binder-Sheets as Separators in All-Solid-State Batteries

The handling of all materials and all processing steps were conducted inside an Ar filled glovebox (O₂ < 0.1 ppm, H₂O < 1 ppm, MBraun, Germany), unless described differently. The solid electrolytes Li₆PS₅Cl (LPSCl) and Li₇P₃S₁₁ (LPS711) were purchased from All Solids (China), while Li₁₀SnP₂S₁₂ (LSPS) was purchased from NEI Corporation (USA); all SEs were used without further purification. Hydrogenated nitrile butadiene rubber (HNBR, 5.5 × 10⁵ g · mol⁻¹, 17 wt% acetonitrile, < 1% residual double bonds) was provided by Arlanxeo (Netherlands). The HNBR binder was dried at 70 °C under dynamic vacuum for 72 h in a drying oven (Büchi B-585, Büchi, Switzerland). For all experiments, a 5 wt% HNBR stock solution in toluene was used, which was kept under continuous stirring. Toluene (anhydrous) was purchased from Merck Millipore (Germany) and dried over a molecular sieve (pore size 3 Å, Merck Millipore, Germany) for at least 24 h. The water content of the toluene was determined by Karl-Fischer-Titration (Titro Line KF trace, Schott Instruments GmbH, Germany) to be < 1 ppm.

The SE/HNBR-sheets were prepared by a slurry-based process using a dissolver (Dispermat LC30, VMA-Getzmann, Germany). For this, all components, i.e., the SE powder, the 5 wt% stock solution of HNBR in toluene, and a defined additional amount of toluene were put together in an HDPE beaker (12 ml Thinky beaker, C3 Prozess- und Analysetechnik GmbH, Germany) in the respective weight ratios, depending on the desired binder content according to Table I. All batches were prepared with a total amount of 2 g of solids, which is defined as the sum of the mass of SE and HNBR. The solid content in the slurry was adjusted to 35%–50% depending on the binder content in order to obtain a suitable slurry viscosity.

In the lid of the HDPE beaker, a centered 16 mm hole was punched in order to pass through the rotating shaft of the dissolver. The mixture was then stirred using a 20 mm dissolver disk (stainless steel, VMA-Getzmann) according to the following sequential mixing procedure (adapted to that reported by Riphaus et al. ⁷ ): 500 rpm for 1 min, 1000 rpm for 1 min, 2000 rpm for 2 min, 5000 rpm for 5 min, and 10,000 rpm for 11 min. The long-duration final step at 10000 rpm serves to break up agglomerates and thoroughly disperse the particles, resulting in a homogeneous slurry. The slurry was coated onto a siliconized polyester foil (PPI Adhesive Products GmbH, Germany) by means of the doctor blade technique, using a 400 μm gap-bar (ERICHSEN, Germany). The deposited films were dried at room temperature for at least 12 h. Prior to further experiments, the sheets or punched samples were dried at 70 °C under dynamic vacuum for at least 16 h. The final free-standing separator sheets had a dry-film thickness of roughly 120 μm, as measured by a thickness gauge (see below).

Initially, LPSCl/HNBR coatings with different gravimetric HNBR binder contents (x_HNBR) were prepared (see left-most column in Table I). In order to provide a more rigorous comparison of the compression-dependent porosity and conductivity of separators made with different solid electrolytes, separator sheets with identical binder volume fractions (φ_HNBR) rather than binder mass fractions were prepared. Therefore, the mass fractions used for the LPSCl/HNBR separator sheets were first converted into binder volume fractions (φ_HNBR) via Eq. 1:

$$\begin{eqnarray}&&{\varphi }_{HNBR}=\left({x}_{HNBR}/{\rho }_{HNBR}\right)/\left({x}_{HNBR}/{\rho }_{HNBR}+\left(1-{x}_{HNBR}\right)/{\rho }_{SE}\right)\end{eqnarray} \tag{ 1 }$$

with ρ representing the bulk densities of the different materials, which were taken from the product specification sheets (HNBR: 0.96 g cm⁻³; LPSCl: 1.64 g cm⁻³; LPS11: 1.98 g cm⁻³; LSPS: 2.25 g cm⁻³). Subsequently, coatings with the same volumetric binder contents using LPS711 and LSPS were prepared (see Table I; note that the LSPS/HNBR coating with a binder content of 1.7 vol% could not be prepared, as no suitable slurry viscosity for a good coating could be obtained).

Free-standing samples of the separators were punched with an 8 mm punch (Rennsteig, Germany) and their weight (Entris II Sartorius, Germany) and thickness (micrometer screw gauge, Mitutoyo, Japan, with an error of ± 2 μm) were measured, either in its uncompressed or compressed state. From the measured areal weight of the separator coating (L_sep) and from the measured thickness (t_sep), the void volume fraction of the separator sheet (ε) was determined, using the above given bulk densities of the HNBR binder and the respective solid electrolytes:

$$\begin{eqnarray}&&\varepsilon =1-\left({L}_{sep}/{t}_{sep}\right)\times \left({x}_{HNBR}/{\rho }_{HNBR}+{x}_{SE}/{\rho }_{SE}\right)\end{eqnarray} \tag{ 2 }$$

The thus determined porosity of the uncompressed separator sheets varies between 50%–60%.

For separator densification at pressures between 100–980 MPa, an 8 mm pressing tool (Lab Club, Germany) and a manual hydraulic press (Atlas 15 T, Specac, UK) were used, pressing several (typically 3) stacked-up separator sheets for 5 min, unless described differently. For densification at pressures between 20–70 MPa, the spring cell setup, which is later described in detail, was used. The compressed thickness of the separator sheets for use in Eq. 2 was determined by a micrometer crew gauge (Mitutoyo, Japan), with an approximate error of ±2 μm. The measured compressed thicknesses ranged between roughly 210–290 μm.

The scanning electron microscopy (SEM) images were acquired using a JSM-IT200 InTouchScope™ (JEOL) field emission SEM at 15 kV with a multi segmental secondary electron detector. To avoid reactions of the sulfidic electrolyte with ambient air, the samples were prepared in an argon-filled glovebox, transferred under inert atmosphere into a dry-room using a Büchi oven tube (B-585, Büchi Labortechnik AG, Switzerland), and from there were transferred into the SEM that was located in the dry-room (the sample exposure time to the dry-room atmosphere (dew point < −25°) was < 1 min). To determine the binder distribution, EDX-mappings were performed at an incident electron beam of 12 kV.

Cross-sections were prepared by argon ion beam polishing, using a cross-section polisher of the type IB-19530CP (JEOL, Japan). An 8 mm disc was punched from the dried SE/HNBR-sheet, fixed with copper tape (PPI Adhesive Products, Ireland) in the sample holder, and then inertly transferred into the device using an inert transfer shuttle (LB-11620TVCA, JEOL Japan) under argon atmosphere. It was polished at 25 °C first for 1 h at 6 kV and then 1 h with 4 kV acceleration voltage.

To investigate the influence of the exposure of the solid electrolyte to the toluene solvent on its (surface) composition and ionic conductivity, toluene was added to the pristine Li₆PS₅Cl (LPSCl) powder, preparing dispersions with a solid content of 40%. For short term treatments, the LSPCl powder was dispersed in an HDPE beaker using a dissolver, analogous to the procedure used for preparing SE/HNBR separator sheets. For long-term treatments, the obtained dispersions were stirred for 48 h in the glovebox using a glas vial and magnetic stirring bar. Afterwards, the solvent was evaporated under dynamic vacuum at room temperature for 3 h using a small vacuum pump, which was connected to a sealed SCHOTT®-glas in the glovebox. For the complete removal of the solvent, the materials were dried under dynamic vacuum at 70 °C using a Büchi oven (B-585, Büchi Labortechnik AG, Switzerland) for at least 12 h to obtain the toluene-treated LSPCl powders.

For Raman spectroscopy analysis, the pristine and the toluene exposed LPSCl powder as well as a final LPSCl/HNBR separator sheet were placed onto a microscopy glass slide and sealed with adhesive-coated Kapton® tape in an argon-filled glovebox to prevent reactions with ambient air and humidity. The measurements were performed at room temperature by focusing the incoming laser beam through the glass slide, using a Renishaw inVia Reflex Raman System (Germany) equipped with a 532 nm excitation laser (Renishaw RL532C, Class 3B) set to a laser power of 2 mW. An integration time of 10 s and an averaging of five subsequently recorded spectra was used. Raman spectra in the range of 47–1548 cm⁻¹ were recorded with a spectral resolution of 3–5 cm⁻¹. The obtained data were processed using the software Renishaw WiRE™.

Diffractograms were collected at room temperature on a STOE STADI P diffractometer (STOE, Germany) in transmission mode in the 2θ range of 10°–90°, using Mo-K_α1 radiation (0.7093 Å, 50 kV, 40 mA), a Ge(111)-monochromator, and a Dectris Mythen 1 K detector (step size 0.015, integration time/step 150 s). The pristine and the toluene exposed LPSCl powder were measured in airtight sealed 0.5 mm borosilicate capillaries, while a final LSPSCl/HNBR separator sheet was measured in a rotating flat-bed sample holder sealed with Kapton® foil. Raw data were processed with WinXPOW, version 3.0.2.1 software.

The surface analysis of the as-received LPSCl powder, long- and short-term toluene-treated LPSCl powder, and the final LPSCl/HNBR separator sheet was carried out by X-ray photoelectron spectroscopy (Axis, Supra, Kratos, UK). The powders and the LPSCl/HNBR separator sheet were pressed to pellets and mounted on a stainless steel sample holder. All prepared samples were transferred into the XPS antechamber using an inert transfer shuttle under Ar atmosphere. The samples were kept in the antechamber until a pressure of ≈10⁻⁸ Torr was reached, and afterwards transferred to the sample analysis chamber, where the pressure was always kept at ≈10⁻⁹ Torr during the entire measurement. Sample irradiation was carried out with monochromatic Al K_α radiation (1486.6 eV), using an emission current of 15 mA. Survey spectra were recorded for all samples with a stepsize of 0.5 eV and a pass energy (PE) of 160 eV. For high-resolution spectra, a stepsize of 0.05 eV and a pass energy of 15 eV were chosen. All recorded spectra were calibrated to the adventitious carbon peak with a binding energy (BE) of 284.8 eV. For data analysis, the CasaXPS software (version 2.3.23, Casa Software Ltd.) was used. After subtraction of a Shirley background, the spectra were fitted using a mixture of a Lorentzian (30%) and Gaussian (70%) shape function, considering the binding energy and full-width-at-half-maximum (FWHM) constraints listed in Table II.

In order to investigate solid-state electrolytes and all-solid-state batteries, a special cell setup is required that is hermetically sealed and that allows the defined and, ideally, variable application of different compressions on the cell stack. Unfortunately, to the best of our knowledge, there is no commonly and widely used cell format commercially available at the moment. Only some specialized cell hardware is available (e.g. by rhd instruments and sphere energy), but owing to its cost, it has not yet been established as a standardized cell format, such e.g. coin cells or Swagelok® T-cells have been for conventional lithium-ion battery research. For this reason, various research groups have come up with individual solutions for their ASSB research. ^2,20

In our group, we have developed a new in-house designed spring-cell setup that is hermetically sealed and that allows for the application of a well-defined variable compression of the cell stack in the range of 20–100 MPa. This cell design is depicted in Fig. 1, consisting of two main parts: the cell body (Fig. 1b), which contains the cell stack, and the cell casing (Fig. 1a), which serves to apply the desired compression on the cell stack via a compression spring. At first, the sample is placed inside a PEEK (polyetheretherketone) tube with an inner diameter of 8.0 mm (yellow colored in Fig. 1b) that is located in the stainless steel cell body, and contacted by hardened stainless steel dies with a diameter of 8.0 mm gray colored) and a hardened stainless steel piston that transfers the compression spring force onto the cell stack. Then, the cell body is inserted into the cell casing and sealed to the stainless steel bellows (10 mm inner diameter; Metallic Flex GmbH, Germany) with a 50 μm PTFE (polytetrafluoroethylene) flat-seal (white colored; Angst+Pfister AG, Switzerland), which at the same time electrically insulates the stainless steel bellows from the cell body. The PTFE flat-seal is compressed by a special-made stub at the lower end of the bellows, which is fixed to the cell body by six screws. At the upper end, the bellows is mounted to the cell casing via a knife-edge seal with a copper gasket, ensuring a completely gas-tight connection.

Zoom In Zoom Out Reset image size

As illustrated in Fig. 1a, the operating cell pressure is applied by the defined compression of the compression spring (47 mm outer diameter, 23 mm uncompressed height, and compression rating of 254 N mm⁻¹; LHL 200 C 02, Lee Spring GmbH, Germany) with a screw that allows to adjust the length and thus the force of the spring, using appropriately sized spacers. Thereby, the applicable cell stack compression can be controlled between 20–100 MPa. A potentiostat can be connected to the cell by inserting the cable plugs into small holes drilled into the lower part of the cell body and the upper current collector plate. At several points, the cell casing is electrically insulated from the cell body by POM (polyoxymethelene) insulations, which are displayed in Fig. 1a by a white color. The stray capacitance of this cell was estimated to be ≈8.1 · 10⁻¹¹ F (based on the dimension of the various metal/insulator/metal interfaces) and is in reasonably good agreement with the experimentally determined capacitance of ≈1.6 · 10⁻¹¹ F. The latter value is at least three orders of magnitude lower than the values measured for the samples investigated here, so that the stray capacitance of the setup does not affect the results of our impedance analysis of the separator sheets. The internal resistance of the cell was determined to <1Ω.

In addition to the cell displayed in Fig. 1, a slightly modified setup thereof was used for impedance measurements with an operation pressure p_oper > 70 MPa. For this, only the cell body together with the cell interior (PEEK tube and dies) was used without the cell casing and equipped with a special piston, shown in Fig. A·1 of the Appendix. Inside an Ar-filled glove box, this cell configuration was put into a hydraulic press (Atlas 15 T, Specac, UK), and electrically insulated from the press by putting a thin polyimide foil between the cell body and the press and between the piston and the press. The applied pressure on cell sell stack is calculated from the compressed area (0.503 cm²) and the manually adjusted press tonnage in units of tons. Thereby, the pressure could be set within a range of 100–590 MPa (0.5–3 tons) with an accuracy of ±20 MPa, which results from the accuracy of the pressure gauge of the device. Higher pressures could not be applied to the cell shown in Fig. 1, as deformation of some cell parts at pressures > 590 MPa was observed (note that pressures up to 1000 MPa could be used for the setup shown in Fig. A·1). During the impedance measurements, the pressure is held constant within the described accuracy. In order to validate this cell setup, the conductivity of LPSCl/HNBR-sheets were determined at an operation pressure of p_oper = 100 MPa with both cell setups (i.e., at the upper limit of the spring-cell and the lower limit of the cell in the hydraulic press), yielding the same values within roughly ±6%. This proves that the modified cell setup is a valid option for impedance measurements under high operation pressures, whereby it is however restricted to measurements at room temperature (i.e., the temperature inside the glovebox).

To conduct impedance measurements, separator sheet samples with a diameter of 8 mm were punched out, and three sheets were stacked for one measurement inside the cell body. After assembly of the cell, the compression on the cell stack was set to the desired value (between 20 and 70 MPa) by compressing a spring. Afterwards the cell was taken out of the glovebox and placed in a temperature chamber (KB53, BINDER, Germany) that was set to 25 °C. Impedance measurements at pressures higher than 70 MPa were conducted in the glovebox at a temperature of 25 °C ± 1 °C. For this, the above described different measurement setup with the hydraulic press was used.Note that for the pre-compression of sheet compressions up to 1000 MPa in course of the porosity determination was possible by using a different pressing tool, described in the section of porosity determination, which however did not allow for impedance measurements.

Measurements outside the glovebox were performed using a VSP-300 potentiostat (BioLogic, France), whereas a SP-300 device (BioLogic, France) was used for measurements inside the glovebox. Prior to impedance measurements, the cell was left to rest for 1 h in order to allow for thermal and mechanical equilibration. Impedance spectra were recorded in a frequency range from 7 MHz to 100 mHz and a potential amplitude of ±10 mV. Data were treated using the software EC-Lab (V 11.36). For acquiring impedance measurements, the following procedure was followed: First, the fabrication pressure was applied in the cell body using the setup displayed in Fig. A·1 in combination with a hydraulic press. During this step at the fabrication pressure, the sample can adapt for small irregularities of the dies. Then, after the initial compression, the entire cell body was transferred into the cell casing of the cell setup displayed in Fig. 1, and the operating pressure was set by compression of the spring. Here it must be noted that during the transfer of the cell body from the press into the cell casing, the dies which are in direct contact with the sample remained in the exact same position and that only the piston is changed from the setup in Fig. A·1 to that in Fig. 1. In this way, resistances due to poor contacting of the sample and the dies are minimized.

To assess the morphology of the SE/HNBR separators prepared by the slurry-processing route, the solid electrolytes and the corresponding SE/HNBR separator sheets were characterized by means of SEM and EDX. Figure 2 shows the top-view SEM images of the different SE powders. First examining the LPSCl powder and separator, Fig. 2a shows that the pristine LPSCl particles are somewhat rounded, with an approximate average particle size on the order of ≈5 μm. After the slurry processing with toluene to prepare the LPSCl/HNBR separator sheet, the SE particle size in the separator sheet seems unchanged and has a similar morphology as the pristine SE powder (see Fig. 2d). The visually open structure of the separator sheet reflects its high porosity of ≈50% (determined by Eq. 2). Figure 2g displays the mapping of the carbon EDX signal that exclusively originates from the HNBR binder (8.2 vol%). Thus, the distribution of the HNBR in the composite sheet is reflected by the carbon distribution map, which gives a qualitative indication that the HNBR binder distribution within the LPSCl/HNBR sheet is reasonably homogeneous, suggesting that the slurry mixing procedure is effective.

Zoom In Zoom Out Reset image size

For the composite system LPS711/HNBR (Figs. 2b, 2e, and 2h), similar observations can be made. The size of pristine LPS711 particles is also on the order of ≈5 μm, without any bigger agglomerates, and the LPS711 particle morphology is not affected by the mixing procedure. This is different for the pristine LSPS powder (Fig. 2c), which next to small particles of also ≈5 μm shows a large fraction of bigger agglomerates with 20 μm and more in size. However, the slurry processing appears to break up these large LSPS agglomerates, resulting in a similar LSPS/HNBR separator sheet morphology (Fig. 2f) that was observed for the LPSCl/HNBR and the LPS711/HNBR separator sheets. Reasons for size reduction of the LSPS particles could be due to the HNBR binder acting as a surfactant or the high stirring step in the mixing procedure. However, as the reference experiment without HNBR present, meaning with only LSPS and solvent, also leads to a break-up of the agglomerates, we believe this is entirely due to mixing at high revolutions.

Figure 3 displays the SEM cross-section images of the uncompressed LPSCl/HNBR separator sheets with 1.7, 8.2, and, 16.0 vol% HNBR in combination with the corresponding carbon EDX-mapping. For the lowest binder content, a homogenous distribution without any binder accumulation can be observed from the EDX-mapping (Fig. 3b). With increasing volume fractions of HNBR, more and more binder accumulations can be found in the voids between the LPSCl particles, indicated by intense red signals in the EDX-mappings (Figs. 3d and 3f). Note, that in the corresponding SEM pictures, the binder cannot be distinguished from the LPSCl particles. For all three separator samples, a red stripe at the top of the EDX-maps can be observed, which is located at the surface near-region of the separator that was opposite of the PP support film, i.e., the surface from which the evaporation of the toluene solvent took place. This seems to indicate a significant binder migration to the free surface, caused by the evaporation of the toluene solvent during the drying process, analogous to the binder migration observed for battery electrodes. ²⁵ However, the possible effects of binder migration on the conductivity of the separator sheets and how it can be influenced by the drying process have not been investigated in this study.

Zoom In Zoom Out Reset image size

For slurry-based processes, the solvent needs to be chosen carefully to prevent the degradation of the sulfidic electrolyte. The two main requirements for a suitable solvent are the ability to sufficiently dissolve or disperse all the components and to not react with the components, as that might negatively affect the ionic conductivity. Yamamoto et al. ¹² have investigated the compatibility of Li₃PS₄ with different solvents, which is chemically similar to LPSCl. ^9,12 The results show that the biggest decrease in ionic conductivity is observed for solvents with a donor number higher than 14, on account of their reaction with Li₃PS₄. A protic solvent like ethanol with a high donor number of 19 ²⁶ reacts by a nucleophilic attack, leading to the decomposition of sulfidic electrolytes and the evolution of H₂S. Therefore, the most suitable candidates are toluene, p-xylene, heptane, anisole, or dichloroethane. The first four, for example, have already been successfully used for the preparation of SE/binder separator sheets, and the results showed no significant decrease in Li-ion conductivity. ^7,9,13 However, recently Ruhl et al. showed a decrease in ionic conductivity and a change in morphology of Li₅PS₆Cl powder treated with toluene for very long contact times of 48 h. ²⁷ These findings appear contradictory to the already establish slurry preparation method from Riphaus et al., ⁷ where toluene was used as a solvent and where no detrimental effects of the exposure of the LSPS solid electrolyte to toluene was observed. Therefore, XRD, Raman spectroscopy, and XPS were used to assess whether any reaction between toluene with a sulfidic SE (exemplarily done for LPSCl) would occur.

Figure 4a shows the influence of the slurry-based processing (20 min dispersion in toluene) on the LPSCl structure investigated by XRD. The pristine LPSCl powder is displayed in brown, the slurry-processed LPSCl/HNBR separator sheet (8.2 vol% HNBR) in beige, and the slurry-processed and compressed (at 590 MPa) LPSCl/HNBR separator sheet in yellow. The diffraction patterns of the LPSCl/HNBR separator sheets provide no evidence for a degradation of the LPSCl material, even though the slurries had been mixed with toluene for 20 min. Furthermore, no changes are observed for the compressed separator sheet, proving that the steps to preare the compressed separator sheets do not introduce any artefacts. The overall features of the diffractograms are consistent with the experimentally observed patterns of the pristine LPSCl powder, showing the argyrodite structure with F3m symmetry (CIF No. 418490). ^28–30 In neither case, typical decomposition products like Li₂S (gray shaded area) or LiCl (blue shaded area) that are found in the case of alcohol solvents ²⁷ could be detected.

Zoom In Zoom Out Reset image size

The samples were also examined by Raman spectroscopy (Fig. 4b), following the same color coding as was used for the XRD data. The pronounced Raman signal in each spectrum at a wavenumber of 426–427 cm⁻¹ can be assigned to the PS₄ ³⁻ tetrahedra of the argyrodites; ^27,31 the typical decomposition products such as polysulfides (broad range between 476–506 cm^{−1 27,31} ) cannot be found in the spectra of the slurry-processed separator sheets. Hence, the Raman data are in good agreement with the results obtained by XRD and provide no evidence for any reaction between of the LPSCl solid electrolyte during our slurry-processing method with toluene.

Since the detrimental effect of toluene on the ion conductivity of LPSCl powder observed by Ruhl et al. ²⁷ was after an exposure of 48 h, much longer than the exposure time during our slurry-processing procedure of only 20 min, and thus significantly less than in the experiments by Ruhl et al., we also conducted further experiments where the LPSCl powder was exposed to toluene for 48 h and analyzed by PEIS and XPS. The conductivity of the three LPSCl powder samples was measured in the spring-cell at an operating cell pressure of 70 MPa (without any pre-compression of the powders). The thereby obtained conductivity for the pristine LPSCl powder of (1.43 ± 0.04 mS cm⁻¹) is essentially identical to that of the LPSCl powder that was dispersed with the dissolver in an HDPE beaker with toluene for 20 min, analogous to the process used for preparing the slurry-processed LPSCl/HNBR separator sheets (1.35 ± 0.02 mS cm⁻¹; the error bars represent the standard deviation of three repeat measurements), consistent with the absence of any changes in their Raman and XRD spectra (Fig. 4). On the other hand, the conductivity of the LPSCl powder that was obtained after stirring it with toluene in a glass beaker for 48 h dropped slightly by a factor of roughly 1.5 (0.98 ± 0.06 mS cm⁻¹). However the detrimental impact of long-term toluene treatment on LPSCl powders, which was described by Ruhl et al., ²⁷ could not be observed. We cannot clearly identify the reason for the discrepancy of the results. There is no chemical reactivity to expect between the sulfidic SE and toluene, therefore one hypothesis for the described detrimental impact observed by Ruhl et al. ²⁷ is the presence of traces of water in the system either due to insufficient drying of the solvent or a leakage during the solvent treatment.

Furthermore, we used XPS analysis in order to have a higher sensitivity on changes in the near-surface region of the LPSCl material, analyzing the S 2p region. Comparing the high-resolution S 2p spectrum of the pristine LPSCl powder that only shows the S 2p feature of the PS₄ ³⁻ units of LPSCl ³² (Fig. 5a) with that of the LPSCl powder that was mixed with toluene in an HDPE beaker using the dissolver for 20 min (Fig. 5b; analogous to the process used for preparing the slurry-processed LPSCl/HNBR separator sheets), indeed no changes in the S 2p feature can be observed. Similarly, the same is true when comparing the pristine LPSCl powder and the as-prepared Li₆PS₅Cl/HNBR separator sheet (Fig. 5d), indicating also no features that could be ascribed to potential decomposition products, such as lithium sulfide (Li₂S) and polysulfides (S_x ²⁻). On the other hand, for the LPSCl powder that was stirred in a glass vial with toluene for 48 h (as was done by Ruhl et al. ²⁷ ), small signals in the S 2p region characteristic of lithium sulfide and polysulfides ³² can be observed (Fig. 5c), which may explain the slightly drop in ionic conductivity.

Zoom In Zoom Out Reset image size

In summary, we can conclude that the short-term exposure of LPSCl to toluene, mimicking the exposure that occurs during our slurry-processing with toluene, does not lead to any detectable changes in structure and composition by XRD, Raman, and XPS, and does not affect the conductivity of the LPSCl powder.

Throughout the literature, a beneficial effect of a low porosity on the ionic conductivity of pellet-type solid electrolyte separators is reported, ^16,20 so that a high degree of densification is desirable. For the mechanically ductile sulfidic solid electrolytes, this can be easily achieved by compression at room temperature. Figure 6a presents the influence of the fabrication pressure (p_fabr) on the SE/HNBR separator sheet porosity for the three different solid electrolytes. The porosity for the as-prepared uncompressed SE/HNBR separator sheets with an HNBR content of 8.2 vol% ranges from roughly 50% for LPSCl (red) and LSPS (green) to almost 60% for LPS711 (blue). Upon applying a fabrication pressure of up to 980 MPa for 5 min, the separator sheet porosity can be reduced to below 5% for the LPSCl/HNBR and the LPS711/HNBR separator sheets and to ≈8% for the LSPS/HNBR separator sheets. Note that based on the measurement accuracy of the thickness of the compressed separator sheets, the estimated absolute error of the porosity values is on the order of ±1%. In order to compare the porosity values of sheet-type and pellet-type separators, we determined the porosity of the powder pellet-type samples at 70 MPa to ε = 28% (LPSCl), 32% (LPS711) and 33% (LSPS) and at 590 MPa to ε = 11% (LPSCl), 13% (LPS711) and 20% (LSPS). This compares with ε = 24% (LPSCl) and 33% (LPS711 and LPSP) at 70 MPa and ε = 3% (LPSCl), 6% (LPS711) and 12% (LPSP) at 590 MPa for separator sheets of the different SEs with a binder content of 8.2 vol%. While at 70 MPa values are quite similar, the porosity values of the sheet-type separators are significantly smaller compared to powder pellet-type samples. This suggests that the binder can effectively fill the voids and therefore sheet-type separators feature lower porosities at the same fabrication pressure.

Zoom In Zoom Out Reset image size

The largest changes in the porosity of the separator sheets occur up to a fabrication pressure of 590 MPa, after which only minor changes occur. Over the entire fabrication pressure range, the porosity of the LPSCl/HNBR separator sheets is lowest and that of the LSPS/HNBR separator sheets is highest, which may be due to the slightly lower plastic deformability of LSPS. ^33–38 In the literature, the plastic deformability is described by the so-called Pugh's ratio G/B, which relates the shear modulus G to the bulk modulus B. Sulfidic SEs generally feature a low G/B ratio of < 0.5, reflecting their ductile nature, whereas for most oxides the G/B ratio lies between 0.5–0.6. The G/B ratio of LPSCl (0.28) and LPS711 (0.34) are clearly lower than for LSPS (0.48), consistent with the trends observed in Fig. 6a and thus the most likely explanation why at a given fabrication pressure the LPSCl/HNBR and the LPS711/HNBR separator sheets can be densified to substantially lower porosities than the LSPS/HNBR separator sheets.

Besides the fabrication pressure, we also investigated the influence of the compression time on the separator sheet porosity. This is shown in Fig. 6b for LPSCl/HNBR separator sheets with a binder content of 8.2 vol% for fabrication pressures of either 100 MPa or 590 MPa, indicating that there is no change in porosity when applying these fabrication pressures between 5–120 min. While for the sake of clarity only data for LPSCl/HNBR separator sheets at these two fabrication pressures are shown, it should be noted that the same behavior was found for the separator sheets based on the other two SEs and for all other fabrication pressures.

Figure 7 displays top-view (left panels) and cross-section (right panels) SEM images of LPSCl/HNBR separator sheets with 8.2 vol% HNBR either uncompressed or compressed uniaxially at room temperature for 5 min at 100 MPa or 590 MPa. Both the top-view and the cross-section SEM pictures show a loose distribution of LPSCl particles for the uncompressed separator sheet (Figs. 7a and 7b), with large and numerous voids in between the LPSCl particles. For the separator sheet prepared at a fabrication pressure of 100 MPa (Figs. 7c and 7d), the packing of the LPSCl particles appears to be much denser, which is consistent with the ≈2-fold lower porosity determined in this case (see Fig. 6a). The cross-section image of that sample reveals that the drop in porosity not only results from smaller and fewer voids, but from the plastic deformation of the LPSCl particles. Still, quite large voids can be seen, indicating a not yet complete densification of the LPSCl/HNBR composite. Upon compressing at 590 MPa (Figs. 7e and 7f), single LPSCl particles cannot anymore be identified clearly, and LPSCl particles appear to have aggregated into bigger domains without visible boundaries in between. Voids can only be observed in very few instances and are randomly distributed, giving evidence for an almost complete densification of the LPSCl/HNBR composite. This is consistent with the very low porosity of ≈3% that was determined in Fig. 6a for this material when compressed at 590 MPa.

Zoom In Zoom Out Reset image size

Besides morphology and porosity, the lithium ion conductivity is the most important property of the produced separator sheets. In order to investigate the lithium ion conductivity, electrochemical impedance spectroscopy was performed. All measurements in this section were conducted at an operating cell pressure of p_oper = 70 MPa and a fabrication pressure of p_fabr = 0 MPa. Herein, the operating cell pressure refers to the pressure, which is applied on the sample during the measurement by the compression of the spring (see Fig. 1). Hence, p_oper is independent from the fabrication pressure p_fabr, which refers to the prior, cell-external compression of the separator sheet by the hydraulic press (as was done similarly by Doux et al. ²⁰ ).

Figure 8 displays the Nyquist plots for LPSCl/HNBR separator sheets with different binder contents, recorded at p_oper = 70 MPa and normalized to the separator sheet thickness that was determined after the measurement; for comparison, the data for pure LPSCl powder (black symbols) are also shown. Spectra for the other SEs are shown in Fig. A·2. The Nyquist plots exhibit a semi-circle at high frequencies and a low-frequency tail that can attributed to the blocking electrodes. The semi-circle at high frequencies can be described by a parallel circuit element of a resistor and a constant phase element (R/Q), with R representing both intragrain and grain boundary contributions to the lithium ion transport, which could not be resolved at the investigated temperature. ^39–41 Comparing the impedance spectra for different binder contents, it can be seen that the semi-circle increases for higher binder contents. We attribute this effect on the ionically insulating nature of the binder, which is located in between the LPSCl particles and impedes the ionic transport across the grain boundaries.

Zoom In Zoom Out Reset image size

The lithium ion conductivity was determined from the total separator resistance R that is obtained from the fit of the R/Q element, as exemplarily shown in Fig. 8 for the 16.0 vol% sample (gray line). Based on the fitted separator resistance, the lithium ion conductivity (σ_sheet) was determined from the thickness of the separator, which was taken after the measurement and outside the cell (t_sep), i.e., with no applied pressure, according to:

$$\begin{eqnarray}&&{\sigma }_{sheet}={t}_{sep}/\left(A\cdot R\right)\end{eqnarray} \tag{ 3 }$$

where A is the cross-sectional area of the 8.0 mm diameter separator (0.50 cm²). Note that the DC resistance (i.e., the electron conduction resistance) is at least four orders of magnitude larger than the AC resistance and is thus negligible. Figure 9a shows the obtained conductivities (σ_sheet) for separator sheets made from different SEs without compression (i.e., p_fabr = 0 MPa) as a function of the volumetric binder content (φ_HNBR) at an operating cell pressure of p_oper = 70 MPa. A binder content of 0 vol% corresponds to the pure SE powders and serves as a reference in order to investigate the change in conductivity by the addition of the HNBR binder. For all SEs, the determined powder conductivity is roughly 1.5 mS cm⁻¹ at p_oper = 70 MPa and the measured porosity ranges from ε = 28%–33%. The addition of only 1.7 vol% of HNBR binder results in a significant, ≈2-fold decrease in conductivity for the LPSCl/HNBR and the LPS711/HNBR separator sheets (unfortunately, no mechanically stable separator sheets could be obtained with 1.7 vol% of HNBR for the LSPS/HNBR composites). Further increasing the binder content has a progressively smaller impact on the conductivity, whereby the here examined SE/HNBR composites follow the same trend and show essentially identical conductivities within the measurement error, with values of ≈0.5 mS cm⁻¹ for 8.2 vol% HNBR and of ≈0.3 mS cm⁻¹ for 16.0 vol%. These values refer to p_oper = 70 MPa, which is a typical pressure used when testing small-scale ASSB laboratory cells. ^2,20

Zoom In Zoom Out Reset image size

In addition, measurements at p_oper = 590 MPa were conducted in order to investigate the maximum achievable conductivity. For this purpose a slightly modified cell setup was used, which is shown in Fig. A·1. Contrary to the fairly similar conductivities of the different SE/HNBR separator sheets obtained at p_oper = 70 MPa, the conductivities of the samples with different solid electrolytes differ somewhat when determined at the higher operating cell pressure of p_oper = 590 MPa, as illustrated in Fig. 9b. The pure SE powder conductivities range from 2.8 ± 0.18 mS cm⁻¹ for LPS711 (blue, ε = 13%) to 3.2 ± 0.23 mS cm⁻¹ for LSPS (green, ε = 20%) and 3.4 ± 0.13 mS cm⁻¹ for LPSCl (red, ε = 11%), and are in reasonably good agreement with the literature (≈4 mS cm⁻¹ for LSPS reported by Bron et al. ⁴² and ≈1.3 mS cm⁻¹ for LPSCl reported by Boulineau et al. ⁴³ ). For a more quantitative comparison, the fabrication pressure and the operating cell pressure need to be considered, as outlined by Doux et al. ²⁰ and by Ohno et al. ⁴¹ For example, for p_fabr = 370 MPa and p_oper = 70 MPa, Doux et al. ²⁰ report ≈3 mS cm⁻¹ for LPSCl when using their optimized carbon powder coated current collector, which is in good agreement with the data in Fig. 9b, where the maximum pressure seen by the sample was 590 MPa. As observed for the above data at p_oper = 70 MPa, the overall trend of the conductivity decrease with increasing binder content is also very similar for all three SE/HNBR composites at p_oper = 590 MPa, with the difference that the LPSCl- and LSPS-based separator sheets outperform those based on LPS711. At a binder content of 8.2 vol% (i.e., at a binder content of 3.7–5.0 wt%, see Table II), the ionic conductivities are rather similar, with 0.93 ± 0.04 mS cm⁻¹, 0.84 ± 0.05, and 0.70 ± 0.03 mS cm⁻¹ for the LPSCl/HNBR, the LSPS/HNBR, and the LPS711/HNBR separator sheets, respectively.

In order to more rigorously compare the decrease of the ionic conductivity with increasing volumetric HNBR binder fraction for the different SE/HNBR separator sheets, the measured σ_sheet values were normalized by the respective SE powder conductivity values (σ_powder), thereby decoupling the absolute conductivity differences of the different SEs from that of the corresponding SE/HNBR composites. Thus, the lower panels in Fig. 9 show the values of σ_sheet/σ_powder vs φ_HNBR for p_oper = 70 MPa (Fig. 9c) and 590 MPa (Fig. 9d). Up to a volumetric HNBR binder fraction of 11.4 vol%, the decrease in conductivity is largely independent of the sulfidic SE type (we ascribe the differences at 1.7 vol% to small errors in weight measurements, which become important at this steep part of the curve). This is somewhat surprising, as at both operating cell pressures the porosity of the SE/HNBR separator sheets measured for 8.2 vol% HNBR (see Fig. 6a) differ quite significantly (ε ranging from ≈23%–34% at p_oper = 70 MPa and from ≈3%–12% at p_oper = 590 MPa). This suggests that small differences in porosity do not significantly affect the conductivity of the SE/HNBR separator sheets. Only at the highest volumetric binder fraction of 16.0 vol%, significant differences can be observed, with a ≈2-fold higher normalized conductivity for the LPSCl/HNBR separator sheet compared to the LPS711/HNBR and LSPS/HNBR separator sheets. When plotting the σ_sheet/σ_powder vs φ_HNBR data acquired at low (70 MPa) and high (590 MPa) operating cell pressure for the separators based on a given SE, as shown in Fig. A·3, it can be seen that the operating cell pressure has no significant impact on the binder-induced conductivity decrease.

Although the mechanical properties of the here discussed SE/HNBR separator sheets were not investigated in detail, some conclusions can be drawn from handling the separator sheets: an HNBR binder content of < 5 vol% results in rather brittle separator sheets that are challenging to work with, whereas an HNBR binder content of > 11 vol% yields very flexible sheets with almost rubbery properties. At the same time, as discussed above, the higher the HNBR binder content, the lower is the ionic conductivity of the separator sheets. Thus, for the here described SE/HNBR separator sheets, the qualitatively best trade-off between ionic conductivity and mechanical properties is found for separator sheets with an HNBR binder content of 8.2 vol%, where ≈25%–30% of the conductivity of the equally compressed SE powder can be obtained (see Figs. 9c and 9d). Therefore, the following experiments were only conducted with this binder content.

Both the literature and the afore described experiments show higher ionic conductivities at higher operating cell pressures, due to a lower sample porosity. ²⁰ In the case of the here examined separator sheets, it should be noted that for the separators based on different SEs, differences in porosity of roughly 10 percentage points were found to not yield significant differences in their σ_sheet/σ_powder values (see above). At the same time, low porosities might be beneficial for suppressing the formation of Li dendrites and thus enable cell operation at higher current densities. ^21,22 However, for actual applications it is not feasible to apply several hundred MPa on a battery, as this would add an excessively large weight to the battery and thus reduce its overall energy density. Therefore, the effect of pre-pressing (or pre-calendering) of separator sheets on their conductivity (i.e., using a high fabrication pressure) will be examined next. In doing so, we compare the conductivity of uncompressed (p_fabr = 0 MPa) separator sheets at p_oper = 20, 40, 70, and 590 MPa with that of separator sheets that had been pre-pressed at p_fabr = 590 MPa and afterwards measured at operating pressures of 20, 40, and 70 MPa. In each case, the separator thickness was determined after the experiments, i.e., without any pressure applied.

Figures 10a–10c shows the results for the three different SE/HNBR separators with 8.2 vol% HNBR, with exemplary separator thickness-normalized Nyquist plots shown in Figs. 10d– 10f. In case of uncompressed LPSCl/HNBR-sheets (p_fabr = 0 MPa, ocher markers in Fig. 10a), the conductivity values range from 0.17 ± 0.04 mS cm⁻¹ (33% porosity) at 20 MPa cell operating pressure to 0.43 ± 0.02 mS cm⁻¹ (24% sample porosity) at p_oper = 70 MPa, reaching a maximum value of 0.94 ± 0.05 mS cm⁻¹ (3% sample porosity) at p_oper = 590 MPa. When the separator sheets are pre-pressed at p_fabr = 590 MPa before the conductivity is determined, higher values are observed for all operation pressures, e.g., 0.57 ± 0.05 mS cm⁻¹ (3% sample porosity) at p_oper = 70 MPa (olive green marker). It is striking that the conductivity of this sample pre-pressed at p_fabr = 590 MPa and measured at p_oper = 70 MPa is almost 2-fold lower than the uncompressed sample that was measured at p_oper = 590 MPa, even though both samples have the same porosity of ≈3%. Doux et al. ²⁰ observed a similar increase of the conductivity with increasing operating cell pressure for LPSCl powder samples pre-compressed at p_fabr = 370 MPa (≈2-fold when increasing p_oper from 20 to 70 MPa), which they attributed to poor contacting of their SE powder samples by metal current collectors at low cell operating pressures. This is based on their observation that the conductivity increase with p_oper could be reduced substantially by placing layers of conductive carbon between their metal current collectors and their SE powder sample (≈1.4-fold when increasing p_oper from 20 to 70 MPa). Considering that the increase in conductivity with the operating cell pressure is similar in the experiments shown in Fig. 10a, it is also likely due to the effect of contact resistances between the metal current collectors of the cell and the separator sheets, as reported by Doux et al. ²⁰ Nevertheless, it is clear that higher fabrication pressures result not only in much reduced porosities of the LPSCl/HNBR separator sheets, but also in ≈1.5-fold higher conductivites at any given operating cell pressure.

Zoom In Zoom Out Reset image size

Figure 10d compares the impedance data of the LPSCl/HNBR separator prepared without compression and measured at p_oper = 70 MPa (ocher; porosity of 24%) with the uncompressed sample measured at p_oper = 590 MPa (ocher, porosity of 3%), the decrease of the semi-circle can be correlated with an enhanced grain boundary conduction due to a better contacting of the particles upon densification, i.e., at lower porosity. On the other hand, when the LPSCl/HNBR separator is pre-compressed at 590 MPa (again resulting in a low porosity of 3%) and measured at p_oper = 70 MPa (olive green), the semi-circle increases and the apparent conductivity decreases. This can be interpreted either as the effect of a higher contact resistance at lower operating cell pressure (see above) or as an internal relaxation of the binder and SE particles due to a spring back leading to a microscopic contact loss of the SE particles without a macroscopic change in the sample thickness.

Similar dependencies of the conductivity on fabrication and operating cell pressure can be observed for the LPS711/HNBR separators (8.2 vol%). The only significant difference between the LPSCl/HNBR and the LPS711/HNBR separators is that the latter have slightly higher porosity values, as already observed in Fig. 6. Exemplary thickness-normalized Nyquist plots of uncompressed and pre-pressed LPS711/HNBR separators are shown in Fig. 10e.

Surprisingly, for the LSPS/HNBR separators (8.2 vol%), no change of the conductivity after pre-compression when measured at a given operating pressure can be observed (see Fig. 10c). For example, the sheets measured at an operating cell pressure of 70 MPa give a conductivity of 0.32 ± 0.01 mS cm⁻¹ when prepared with p_fabr = 0 MPa and 0.34 ± 0.01 mS cm⁻¹ when prepared with p_fabr = 590 MPa. One apparent difference between the pre-compressed separators based on LSPS vs those based on LPSCl or LPS711, is the substantially higher porosity of the latter (3% and 6% for the former, 12% for the latter). In order to investigate whether this difference in porosity could explain why the LSPS/HNBR separator sheets yield the same conductivity at p_oper = 70 MPa for uncompressed and pre-compressed samples, we will next examine the effect of porosity on conductivity. Unfortunately, it was not possible for the LSPS/HNBR separator sheets to achieve a final porosity value of 3% that can be reached for LPSCl/HNBR separator sheets. Hence, we conducted the following evaluation on the effect of porosity using LPSCl/HNBR separator sheets with 8.2 vol% HNBR, which were subjected to different fabrication pressures ranging between 0 and 590 MPa, after which their conductivity was determined at an operating cell pressure of 70 MPa (see Fig. 11). For a fabrication pressure of 215 MPa, resulting in a porosity of 11%, the LPSCl/HNBR separator conductivity of 0.46 ± 0.004 mS cm⁻¹ is only marginally larger than that of the uncompressed sample with a porosity of 24% (0.43 ± 0.02 mS cm⁻¹), a difference which is within the error range of the measurements. Only upon pre-compression at p_fabr = 295 MPa, resulting in a porosity of 9%, the conductivity increases by ≈20% compared to the uncompressed sample (i.e., to 0.52 ± 0.01 mS cm⁻¹). Finally, at p_fabr = 590 MPa and 3% porosity, a conductivity gain of ≈31% compared to the uncompressed sample is observed (with a value of 0.57 ± 0.04 mS cm⁻¹ at p_fabr = 590 MPa). For the LPSCl/HNBR separator sheets, this shows that substantial conductivity gains for pre-compressed separators can only be observed once the porosity decreases below 11%. Thus, if one were to assume a similar conductivity vs porosity dependence for the LSPS/HNBR separator sheets, one could explain why there was no improvement of their conductivity between uncompressed samples and samples pre-compressed to 590 MPa (see Fig. 10c), where the porosity was still 12%. In summary, the data in Fig. 11 suggest that a pre-compression of the SE/HNBR separator sheets only significantly improves their conductivity when very low porosities (on the order of lower than 10%) can be obtained, which for materials like LSPS would require fabrication pressures of more than 800 MPa (see Fig. 6).

Zoom In Zoom Out Reset image size

The following figure shows the details of the cell setup used in combination with a hydraulic press located in the glovebox in order to reach cell stack compressions of 100–590 MPa, with an accuracy of ±20 MPa. The theoretically determined capacitance of this cell is ≈1.3 · 10⁻¹¹ F, which is in good agreement with the experimentally determined value of ≈1.0 · 10⁻¹¹ F. The stray capacitances of the setup are at least three order of magnitude lower than the capacitances of the cell and thus do not influence the results of the PEIS measurements. The internal resistance of the cell was determined to <0.2Ω.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The following figure shows the impedance measurements for uncompressed (p_fabr = 0 MPa) separator sheets obtained at a cell operating pressure of p_oper = 70 MPa. In accordance to the observations described for LPSCl/HNBR sheets (cf. Fig. 8 in the main text), the spectra exhibit a semi-circle at high frequencies, representing grain and grain boundary contributions to the lithium ion transport, and a low-frequency tail due to the blocking electrode configuration. With increasing binder content, the magnitude of the semi-circle increases, leading to a decreasing ionic conductivity.

Figures 9, 10c and 10d in the main part show the normalized separator sheet conductivities (σ_sheet/σ_powder) vs φ_HNBR, comparing the three different SE systems at an operating cell pressure of 70 and 590 MPa each. Here we additionally compare σ_sheet/σ_powder vs φ_HNBR for the same SE systems at low (70 MPa, data points in lighter colors) and high (590 MPa, data points in darker colors) operating cell pressure in order to investigate its impact on the separator sheet conductivity. Figure A·3 shows that, within the range of the experimental error, the normalized separator conductivity is essentially independent of the operating cell pressure and only a function of the volumetric binder content.

Zoom In Zoom Out Reset image size