Slurry-Based Processing of Solid Electrolytes: A Comparative Binder Study (2024)

In order to conduct a comparative binder study, a suitable model solid electrolyte and solvent had to be selected in the first place. As the motivation behind this work is to pave the way for large-scale production of sheet-type SE layers, only commercially available materials were considered, limiting the choice of sulfide-based SEs. LSPS from NEI Corporation was selected due to its high ionic conductivity of 2–5 mS·cm⁻¹,^22–24 which comes close to that of its thiogermanate analogue LGPS that has been reported to have one of the highest conductivities of 12 mS·cm⁻¹.²⁶ In contrast, the price can be significantly reduced by the replacement of Ge with Sn, making LSPS more attractive for large-scale applications.²²

As mentioned, the high reactivity of sulfide-based SEs with polar protic solvents, leading to evolution of H₂S, requires the use of less or non-polar solvents instead.¹⁹ Yamamoto et al. found that solvents with a high donor number, like propylene carbonate or diglyme, react with the chemically similar electrolyte LPS by nucleophilic attack.⁹ In contrast, various studies have shown excellent stability in solvents with low donor number, such as heptane, toluene, p-xylene or dichloroethane.^5,9,18 Since too fast evaporation of the solvent leads to inhom*ogeneous coatings with bulged surfaces, a further demand for slurry-processing is a low vapor pressure.⁵ Therefore, toluene with a moderate vapor pressure of 3.8 kPa²⁷ was selected in this study.

Once the solvent was defined, the next step comprised the selection of binders to be tested. The baseline requirements to be met were solubility in toluene and absence of protic functional groups, which would cause again decomposition reactions with the SE. TableI shows the chosen binders. In order to gain a comprehensive understanding of the effect of the chemical nature of the binder on its performance, diverse material classes ranging from aliphatic to aromatic, including different functional groups, were covered. PIB is a fully saturated hydrocarbon, whereas SBR contains aromatic units. PMMA and PEVA both have ester side chains attached to the hydrocarbon backbone, with the number of functional groups being much larger for PMMA. NBR provides a nitrile functionality. The hydrogenated variant HNBR was used to further differentiate from SBR.

As the binders shall be used in the SE separator layer, another key requirement is negligible electronic conductivity. To probe this, pure binder films with dry-film thicknesses of 15–30 μm were casted and tested electrochemically. Linear sweep voltammetry (LSV) was performed to apply a distinct voltage to the sample and measure the current response. While PIB, SBR and PMMA showed no current flow over the voltage range of 0–5 V (Figure S1a), a low current of up to 45 nA was detected for PEVA and HNBR at elevated temperatures (Figure S1b). However, the electronic conductivity in the range of 10⁻¹¹ S·cm⁻¹, calculated from chronoamperometry measurements, is negligible. In addition, as no current peaks were detected in the LSV measurements, electrochemical stability in the applied voltage range can be concluded for all binders.

In the next step, their chemical inertness with LSPS was investigated. To provoke possible decomposition reactions, SE sheets with a LSPS: binder ratio of 90 : 10 wt% were stored for prolonged time at 60°C. Subsequent X-ray photoelectron spectroscopy (XPS) and Fourier transform infrared spectroscopy (FTIR) measurements showed no difference between the aged samples and the pristine materials (exemplary shown for LSPS-PEVA/-PIB in Figure S2/S3). Hence, all binders fulfilled the requirements for further investigations.

In order to evaluate the influences of the binder on the processability, flexibility, density and resistivity of SE thin layers, freestanding LSPS sheets were prepared by slurry-processing. All relevant process parameters, namely LSPS-to-binder ratio, solids content in the slurry, applicator gap size and compression method were varied in an extensive parameter study. The ranges of the parameters employed are listed in TableII.

A first criterion for assessing the processability of the slurry was its viscosity. The solids content in the slurry served as a measure for it. In case of a low solids content, the slurry leaks out of the applicator, whereas a slurry too viscous sticks to the applicator and cannot be processed. The study revealed that the optimum viscosity strongly depends on the type of binder used, whereas the amount of binder had less impact. The quality of PEVA-based sheets turned out to be particularly dependent on the solids content. While excellent layers were yielded at roughly 45%, lower solids contents resulted in very inhom*ogeneous layers. The viscosity of SBR- and PMMA-based slurries, in contrast, had to be in the range of 20% for comparable results. PIB as well as HNBR could be processed best at about 40% solids content for low binder contents, however the viscosity had to be reduced if the binder content was increased.

Also relevant in this context was the wettability of the slurry. A generally good wetting behavior was expected, as both solvent and carrier foil are non-polar. Depending on the polarity of the binder, the wettability of the slurry is affected to a greater or lesser extent. For those slurries based on the most polar binder, PMMA, no hom*ogeneous coating could be obtained throughout the wide range of tested process parameter combinations. In all cases, the SE sheet rolled up during drying, indicating that the cohesive forces within the slurry exceed the adhesive forces between slurry and carrier foil (Figure 1c). In contrast, very hom*ogeneous coatings were obtained for slurries comprising the non-polar binders PIB and SBR (Figures 1a/1b). The best results according to optical inspection could be achieved with PEVA and HNBR. These coatings were not only very hom*ogeneous, but also showed well-defined edges (Figures 1d/1e). Apparently, the combination of a non-polar polymer backbone and a relatively low amount of polar functional groups enables both good cohesion in the slurry and good adhesion to the carrier foil.

Independently of the LSPS-to-binder ratio as well as the solids content in the slurry, an applicator gap size of 200 μm was identified as the optimum for all binders. According to the solids contents, dry-film thicknesses ranging between roughly 50 μm for SBR and 100μm for PEVA were obtained for the uncompressed sheets. Subsequent calendering allowed for reduction of the film thickness by 55–70%, while the greatest densification was achieved for the thickest sheet. The final film thicknesses after calendering varied between 20–40μm.

To further evaluate the SE sheets, their hom*ogeneity after calendering was examined in greater detail. A hom*ogeneous sample should provide small variations in coating height and porosity over the SE layer as well as uniform distribution of the binder between SE particles. To examine the latter, EDX mapping was conducted. Exemplary elemental maps and SEM images are shown in Figure 2 for LSPS layers processed with the various binders. It should be mentioned that all sheets were fabricated under identical conditions and the same batch of SE powder was used for all experiments to ensure comparability.

Significant differences can already be observed at a low magnification for the PMMA-based sheet (Figure 2c, top), which shows a very rough surface with many agglomerates. Zooming in further reveals that, contrary to the other samples, bigger particles are not embedded in the plane but rise above, increasing the roughness (Figure 2, middle). The EDX maps revealed two different distribution pattern of the binders. SBR, PIB and PMMA show a granular domain structure, with small binder agglomerates (red) between SE particles (yellow). This is most visible for the SBR-based layer (Figure 2b). In contrast, PEVA and HNBR form a film on the SE particles, indicated by a uniform orange coloration of the EDX maps (Figures 2d/2e). While this hom*ogeneous binder distribution apparently results in a smoother surface, the continuous film might be detrimental for the ionic conductivity. Inada et al. found that a binder coating covering the SE particles remarkably hinders ionic transport, whereas binder domains barely affect the conductivity.²⁸ It should be noted that the gray spots arise from indentations on the surface, which cannot be reached by the EDX beam.

Both optical evaluation and SEM imaging demonstrated that PMMA is unsuitable to fabricate hom*ogeneous and dense SE sheets, thus this binder was not further considered. Distinct differences between the remaining four were observed for the minimum amount of binder required to obtain processible, freestanding SE sheets. Previous studies have shown that the resistivity of SE layers increases with binder content, owing to the insulating behavior of the polymers.^9,14 Consequently, the amount of binder should be reduced as far as possible. With SBR, 10 wt% (equaling to about 20 vol.%) were necessary to fabricate a hom*ogeneous, freestanding SE sheet with low porosity, whereas 7.5 wt% PEVA delivered comparable results. For HNBR and PIB, a much lower binder content of 2.5 wt% was sufficient.

These differences might be associated with the molecular weight of the polymeric binders. An interesting correlation between adhesion properties and molecular weight of the binder was reported by Lee et al., who studied LTO electrodes for conventional LIBs.²⁹ They found that adhesion as well as surface coverage of the LTO particles enhances with increasing molecular weight of their carboxymethyl cellulose binder. The PIB used herein had the highest average molecular weight of 3.1·10⁶ g·mol⁻¹, followed by HNBR with 5.5·10⁵, PEVA with 3.5·10⁵ and SBR with 1.9·10⁵ g·mol⁻¹. This trend turned out to be in line with the amount of binder required for hom*ogeneous, processible SE sheets. Apart from molecular weight, the chemical nature of the polymeric binder is expected to affect the properties of the SE sheet. PIB and HNBR constitute saturated hydrocarbons with small side chains, which can easily adhere to the SE particles. Lee et al. moreover found that the nitrile groups of NBR increase the polarity and dipole moment for the interaction between binder and SE, resulting in better adhesion.¹⁸ This might explain why the same binder content of 2.5 wt% is sufficient for PIB and HNBR, although PIB has a much higher molecular weight. PEVA and SBR contain large functional groups, which might hinder intense contact between binder and SE. Apparently, the negative effect of steric hindrance is more significant than the positive effect of polarity, thus demanding for a higher binder quantity of 7.5 and 10 wt%, respectively.

Another important evaluation criterion is the porosity of the SE sheets after calendering. As Li-ion transport is limited with high porosity, a densification close to 100% of the theoretical density would be ideal to maximize the ionic conductivity. On the other hand, a certain remaining porosity might be beneficial to maintain good mechanical properties.^6,8 In line with this, Choi et al. found that the porosity of silicon electrodes decreases with increasing binder content.³⁰ They concluded that a certain binder level is essential to achieve sufficient bonding between particles. Figure 3 illustrates the relation between required minimum amount of binder and resulting porosity for the various polymers. Despite the high binder content of 10 wt%, SBR-based sheets could only be compacted to around 70% of the theoretical density upon calendering, thus roughly 30% porosity remained. In contrast, the calendered layers containing the least amount of binder, namely those with 2.5 wt% HNBR and PIB, yielded a porosity around 15% and 10%, respectively. The best results were obtained for the calendered sheets containing 7.5 wt% PEVA, with porosities as low as roughly 8%.

These results indicate that the amount of binder has only a limiting impact on densification properties. Indeed, the chemical nature might be more relevant. The as-received polymers HNBR, and particularly PIB and PEVA, exhibited a rubberlike consistency, whereas SBR is supplied as hard granules. It is therefore not surprising that compression of the SE sheets is less effective when using SBR. The different properties can also be associated with the chemical structure of the polymer. On the one hand, the unsaturated hydrocarbon backbone introduces a certain stiffness and, on the other hand, the bulky but non-polar side chains might hinder intense bonding to the SE particles. Weaker adhesion caused by bulky functional groups could also explain why a greater amount of binder is required for PEVA compared to HNBR.

The influence of binder content and densification on the properties of the SE sheet was further examined with respect to mechanical properties. In order to assess their flexibility, Mandrel bend tests according to DIN norm ISO 1519:2011 were conducted. Calendered sheets containing 2.5 and 10 wt% binder as well as non-calendered sheets with 2.5 wt% binder were probed. All samples based on PIB, PEVA and HNBR showed excellent bending properties, even for low binder contents. They passed all mandrel diameters without any damage (Figure 4a). In contrast, all SBR-based sheets already broke at bending around the largest mandrel (Figure 4b). This observation is in line with the aforementioned poor densification properties, and can again be attributed to the described structural features of the polymer. Since roll-to-roll processes, which demand for a certain bendability, are very likely to be implemented for large-scale fabrication of ASSBs,⁸ SBR-based layers might be unsuitable at this stage.

Besides the mostly physical properties discussed above, the final and most important assessment criterion examined herein is the ionic conductivity of the SE layers. As mentioned, previous studies demonstrated that incorporation of a polymeric binder leads to a reduced ionic conductivity.^4,7,9,14,19 A much stronger influence was found if the binder forms a film on the SE particles instead of granular domains in between the particles.²⁸ Furthermore, it has been shown for various SE materials that conductivity increases with packing density, i.e. with decreasing porosity.^17,31,32 Consequently, clear differences in ionic conductivity between the SE layers in this contribution are expected.

The long-range Li-ion transport was probed by means of ⁷Li PFG NMR spectroscopy of calendered SE sheets with different binder contents. As demonstrated by Kaus et al.,²⁴ this powerful technique allows for investigation of Li-ion diffusion on time scales of tens or hundreds of milliseconds, corresponding to length scales of several micrometers. An exemplary echo damping versus gradient strength plot can be found in Figure S4 for LSPS-PEVA. Table S1 shows the extracted diffusion coefficients D_Li for SE sheets containing 2.5 and 10 wt% of the various binders, respectively. For better visualization, they are depicted in Figure 5 in blue (2.5 wt%) and red (10 wt%). At first glance it is clear that the impact of binder volume on lithium ion diffusivity depends on the binder used. For HNBR and PEVA, D_Li decreases by roughly 15 and 44%, respectively, with a fourfold increase in binder content. In contrast, it hardly changes for PIB, considering the measurement accuracy, and even rises by roughly 29% for SBR.

To better understand the differences, these results were connected with those from EDX analysis. The elemental maps indicated a continuous film of PEVA and HNBR on the SE particles. It is thus conceivable that a larger amount of binder, which leads to thicker polymer films, hinders Li-ion transfer between particles. On the contrary, due to the granular structure of PIB and SBR, the binder quantity has less impact on Li-ion diffusion. The enhanced bonding at larger binder volume indeed rather promotes Li-ion transport at grain boundaries. For SBR, where poor adhesion requires a high binder content to yield freestanding sheets, this effect is more pronounced, resulting in a larger D_Li value for the 10 wt% sample. This interpretation of the trends of D_Li with increasing amount of binder is in line with the aforementioned findings of Inada et al., namely that a binder coating on SE particles strongly affects ionic transport, whereas granular binder domains have less impact.²⁸ The influence of binder content though has not been studied in this context.

In order to assess the absolute values for D_Li extracted from the PFG NMR measurements, they were first related to that of the pure LSPS powder, which was determined to 2.9±0.1·10⁻¹² m²·s⁻¹. Independent from the binder type and quantity, all SE sheets yielded lower values. In general, PEVA and HNBR outperform the others at low binder content, whilst PIB and HNBR show the highest diffusivity at high binder content. For a fair comparison, however, only those values corresponding to SE sheets that formed mechanically stable layers should be taken into account. As shown in Figure 3, the minimum binder content required to obtain processible, freestanding sheets varies from binder to binder. The absolute values for 2.5 wt% PIB and HNBR should thus be compared to that for 10 wt% SBR and PEVA. Accordingly, the highest diffusion coefficient of 2.7±0.15·10⁻¹² m²·s⁻¹ was yielded for HNBR, followed by PIB (2.0±0.08·10⁻¹² m²·s⁻¹), SBR (1.8±0.15·10⁻¹² m²·s⁻¹) and PEVA (1.5±0.07·10⁻¹² m²·s⁻¹). Compared to pure LSPS, addition of a polymeric binder resulted in a decrease of lithium ion diffusivity by 7% in the best and 48% in the worst case. This is in the same range as the ∼7–16% drop in lithium ion conductivity reported by Lee et al. for composite films of NBR (4 wt%) and LPS vs. the pure material.¹⁸

The partial Li⁺ conductivity σ_Li can be derived from the diffusion coefficient D_Li according to the Nernst-Einstein Equation1. Here, N_Li is the Li-ion concentration, Q_e the unit charge, k the Boltzmann constant and T the temperature. In order to calculate the ionic conductivity, the Li-ion concentration has to be estimated first. This is usually done based on the crystal structure of the material, according to Equation 2. Here, X_Li is the number of Li-ions per formula unit, Z the number of formula units per unit cell and V the volume of the unit cell.

First, the conductivity of the pure SE material was calculated, using the NMR-derived diffusion coefficient (2.9±0.1·10⁻¹² m²·s⁻¹). The resulting ionic conductivity of 3.6±0.12 mS·cm⁻¹ is in line with literature data (2–5mS·cm⁻¹), which were obtained through AC impedance measurements as well as different ⁷Li NMR techniques.^22,24,25 This shows that the Nernst-Einstein relationship allows for a good approximation of the ionic conductivity. For a meaningful assessment of the actual conductivity of the SE/binder composites, however, the lithium diffusion length in the NMR experiments has to be larger than the radius of the SE particles. Otherwise, Li-ions would not cross grain-boundaries within the time span of the PFG NMR measurement, and the effect of the binder would thus be neglected.

A grain size of 250–300 nm is specified for LSPS by the supplier.³³ In order to confirm this, X-ray diffraction (XRD) pattern of the LSPS powder used were recorded and refined with regard to grain size. The Rietveld refinement (see Figure S5) yielded an average crystallite size of roughly 148 nm. The polymeric binder might, however, not reach each grain, but rather be located between particles. Based on the specific surface area of 1.6 m²·g⁻¹, determined by BET analysis, an average primary particle size of roughly 1 μm can be estimated.³⁴ This was supported by SEM imaging of the as-received LSPS powder (Figure S6). A diffusion length d_diff > 0.5 μm would thus be desirable. It can be calculated as (n·D_Li·t_diff)^0.5 with n being dimensionality of the motion and t_diff the diffusion time.³⁵ As LSPS is a 3D Li-ion conductor,³⁶ n equals to 6. The resulting d_diff of roughly 1.3 μm is much larger than the radius of the LSPS grains and still sufficiently larger than the radius of the primary particles.

Besides, two additional facts support the assumption that Li-ions are transported across grain-boundaries during the PFG NMR measurement. Firstly, the estimated ionic conductivity of the pure SE material matches the value reported for the total conductivity, while a significantly higher value would be expected for the intergrain conductivity.²² Secondly, different values for D_Li were extracted for the different binder types and volumes, all of them lower than that for pristine LSPS. If diffusion would solely take place within single SE particles, binder type and weight fraction would, however, not affect the measured diffusion coefficients. Hence, PFG NMR is suitable to probe the conductivity of the composite sheets without neglecting the binder impact. In order to account for the lithium-free volume though it is necessary to recalculate the Li-ion concentration by integrating the volume fraction of the SE into Equation 2:

The ionic conductivity can then be calculated using Equations 1 and 3. Figure 6 depicts the results for the different SE sheets containing 2.5 (blue) and 10 wt% (red) binder, respectively. A rise in conductivity with binder content is only observed for SBR. The increase by roughly 10% can again be attributed to the better adhesion between particles, resulting in a lower porosity around 30% for the 10 wt% compared to roughly 57% for the 2.5 wt% sheet. Apparently, poor adhesion is in this case more limiting than restricted ionic pathways due to blocking by greater amounts of polymeric binder. A similar observation has been made by Rosero-Navarro et al, who reported an increase in conductivity when incorporating ethyl cellulose in a Li₆PS₅Cl pellet.¹⁴ In contrast, a fourfold binder content results in reduced conductivity for the other binders by roughly 10–50%. Restriction of ionic pathways seems to be the limiting factor in the case of PIB, PEVA and HNBR. This is in line with the considerably lower porosity of the respective SE sheets compared to SBR (see Figure 3), which indicates that adhesion is not a major issue with these polymers.

As in the case of the diffusion coefficients, the required binder content to obtain mechanically stable layers should be considered when assessing the absolute values. Accordingly, the estimated ionic conductivity for 2.5 wt% PIB and HNBR should be compared to that for 10wt% SBR and PEVA. The highest conductivity of 3.2±0.18 mS·cm⁻¹ was thus obtained for HNBR, followed by PIB (2.4±0.09 mS·cm⁻¹), SBR (1.8±0.15 mS·cm⁻¹) and PEVA (1.5±0.07 mS·cm⁻¹). In line with previous reports, stating that incorporation of a polymeric binder leads to a reduced ionic conductivity,^4,7,9,14,19 all values level below that of pure LSPS (3.6±0.12 mS·cm⁻¹). The conductivity is reduced by roughly 12% in the best and 58% in the worst case. As mentioned, a comparable decrease of the ionic conductivity of composite films vs. the pure SE has been reported by Lee et al.¹⁸ They found that incorporation of 4 wt% NBR and butadiene, respectively, leads to a drop by roughly 7–27%. When using 5.5 wt% NBR in a LPS sheet, Oh et al. maintained ∼50% of the ionic conductivity.¹⁹ Comparable results were shown by Rosero-Navarro et al. for a composite of Li₆PS₅Cl and ethyl cellulose.¹⁴ The group of Jung et al. demonstrated that even the application of a polymer scaffold that makes up 14 wt% lowers the conductivity only by a factor of roughly 3.6.²⁰ Since all estimated ionic conductivities in this contribution lie in the mS range, the composite sheets prove to be suitable for application in ASSBs.

The respective ionic conductivity obtained through electrochemical impedance spectroscopy (EIS), however, might be lower since contacting issues are less pronounced in PFG NMR measurements. More precisely, PFG NMR only probes Li-ion diffusion between interconnected particles. The impact of porosity, however, is neglected. Additional effects specific to EIS measurements result from the interface between the SE sheet and the current collector. Pores as well as polymeric binder at the surface lead to a lower effective contacting area. As the measured resistance is normalized to the area, considering only the geometric dimensions of the sample, the conductivity is usually underestimated. Due to the high sensitivity of EIS to sample preparation and contacting, a more reliable comparison of the SE sheets with different binder types and volumes is achieved using PFG NMR. On the other hand, the EIS derived conductivity might be more relevant in an actual ASSB, where contacting issues are not negligible. In order to compare the two methods, an exemplary EIS measurement was performed of a LSPS sheet with 2.5 wt% HNBR (Figure S7). Fitting of the data yielded an ionic conductivity of roughly 0.3 mS·cm⁻¹, being an order of magnitude lower than that estimated from PFG NMR. Beside the aforementioned effects attributed to EIS, blocking of certain conduction pathways by a thick HNBR film might be responsible for the significant drop in total conductivity. This is again in line with the report from Inada et al., who observed that a continuous binder film covering SE particles remarkably hinders ionic transport.²⁸ However, as stated by the group of Jung et al.,^19,37 an ionic conductivity in the 10⁻⁴ S·cm⁻¹ range is still sufficiently high for application in ASSBs.