Processing-structure-property relationships in practical thin solid-electrolyte separators for all-solid-state batteries

As shown in figure 1, the TSE separators were prepared by tape-casting (T) and calendering (C) processes with two binder contents (5 and 10 wt.%) and three densification pressures (0, 100, and 200 MPa) and are denoted as T5/10-0, T5/10-100, and T5/10-200 or C5/10-0, C5/10-100, and C5/10-200, respectively. The TSE with a binder content of 20 wt.% was also prepared and tested, and as it followed the same trends discussed here, the corresponding data are included in the supporting information (SI) for conciseness.

Zoom In Zoom Out Reset image size

NBR is a commonly used binder for sulfide-based electrolytes. XNBR is a functionalized NBR, characterized by the addition of a polar functional group (–COOH). In addition to their similar chemical structures and comparable tensile moduli (2.15 and 1.55 MPa, respectively) [27, 28], NBR and XNBR were selected due to their solubility in toluene, a nonpolar solvent that has good chemical compatibility with sulfide LPSCl solid electrolyte. Moreover, choosing XNBR for calendering is due to its tendency to form a complex and 'rubbery' composite with LPSCl that enables calendering but prevents tape-casting, whereas NBR is selected for tape-casting because LPSCl and NBR can be mixed in toluene to form a tape-casting slurry [26]. For the component mixing in the tape-casting process, high-speed and high-energy ball milling is used, which could break down the large LPSCl particles and agglomerates, ensuring the uniform dispersion of materials in the T-TSEs and making it easier to control TSE thickness precisely in the following tape-casting step. Regarding the calendering process, the low-speed and low-energy mechanical mixing could reduce the damage to the LPSCl particles, which in turn decreases the resistance stemming from an increased number of particle interfaces and defects.

Figure 2 shows the influence of the densification on the cross-sectional microstructure of the samples with 5 wt.% binder content characterized by focused ion beam-scanning electron microscopy (FIB-SEM) and the corresponding porosities for the samples with both 5 and 10 wt.% binder contents.

Zoom In Zoom Out Reset image size

Compared to the pristine powder (figure S1), the LPSCl particles in C-TSEs maintain a similar morphology (figures 2(D)–(F)), with small particles < 3 μm in size and larger particles/agglomerates reaching sizes of 10 μm or more. However, the ball milling step for T-TSEs appears to have broken up these large LPSCl particles/agglomerates, resulting in a different morphology (figures 2(A)–(C)) with reduced particle size and an average particle size of ⩽ 2 μm. The reduction in size of the LPSCl particles can be mainly attributed to the high-energy mechanical mixing step involved in the slurry processing and partially due to the NBR binder functioning as a surfactant [12]. The packing of the LPSCl particles becomes much denser as the densification pressure increases. Figures 2(A) and (D) show loosely packed LPSCl particles with an interconnected open-pore network in the pristine T- and C-TSEs. After densification at 200 MPa, only isolated voids are left and parts of the LPSCl particles underwent plastic deformation to some extent that allowed them to aggregate into larger domains with indistinct boundaries between them (figures 2(C) and (F)).

As shown in figures 2(G) and (H), an increase in binder content or densification pressure leads to a reduction in porosity. The porosity of the TSEs with 10 wt.% binder is lower than the one with 5 wt.% binder at the same densification pressure, more significant for C-TSEs, as the binder can fill the voids between the particles. For the T-TSEs, the reduction in porosity with densification appears to follow an approximately linear trend from 0 to 200 MPa, however, this is not the case for the C-TSEs. Notably, there is a more pronounced reduction in porosity when going from C5-100 to C5-200 compared to that from C5-0 to C5-100, as well as going from C10-0 to C10-100. The above observation could be attributed to the more homogeneous microstructure in the T-TSEs compared to the C-TSEs, resulting in a more consistent and uniform densification process. Although a higher degree of densification (> 200 MPa) is expected to result in lower porosity and higher ionic conductivity [12], the resulting TSEs proved too brittle to allow meaningful tensile tests. This brittleness renders such high densification pressures undesirable for our investigation aimed at linking processing, microstructure, and the mechanical and electrochemical properties. Therefore, higher densification pressures are beyond the scope of this work.

The above results of the densification effect on porosity suggest a more homogeneous distribution of binder in the T-TSEs than the C-TSEs. Further, energy-dispersive x-ray spectroscopy (EDX) was performed on the T- and C-TSEs with 5 wt.% binder content after densification at 200 MPa (figures 3(A) and (B)). The sulfur and carbon signals were used to identify the LPSCl and the binder locations, respectively. The relatively homogeneous sulfur signal in both is as expected. The carbon distribution map gives a qualitative indication that the NBR binder within the T-TSEs is homogeneously wrapping the LPSCl particles, whereas in the C-TSEs the higher intensity of carbon signal in certain areas demonstrates random distribution or agglomeration of XNBR binder at the boundaries of LPSCl particles. This provides further evidence of the more homogeneous microstructure of the T-TSEs in terms of both particle size and binder distribution.

Zoom In Zoom Out Reset image size

Tensile testing is carried out under Ar atmosphere to understand the mechanical properties of the T- and C-TSEs. The recorded stress-strain curves for the corresponding T- and C-TSEs with different binder contents (5 and 10 wt.%) and densification pressures (0, 100, and 200 MPa) are shown in figures 4(A) and (B). There is a clear difference between the T- and C-TSEs. The stress-strain curves of T-TSEs present a steep linear-elastic region followed by a plateau in which the plastic elongation of the T-TSEs takes place. In comparison, the stress–strain curves of C-TSEs have a very different shape without clear yield points or elongation plateau.

Zoom In Zoom Out Reset image size

For a quantified comparison, the tensile modulus and ultimate tensile strength (UTS) were determined from the stress-strain curves. For all the samples under investigation, the UTS was defined as the maximum stress before failure. Figures 4(C) and (D) show the tensile moduli of the tested samples. At the same binder content and densification pressure, T-TSEs with the more well-distributed binder network demonstrate a higher tensile modulus compared to the counterparts prepared by the calendering method with the binder partially covering LPSCl particles. Despite the porosities of the T5-200 and C5-200 or T10-200 and C10-200 samples are comparable, the tensile modulus of the T-TSEs is nearly twice that of the C-TSEs. Both T- and C-TSEs displayed an obvious increase in tensile modulus upon densification for both binder contents. For pristine TSEs higher tensile modulus values are observed with higher binder loadings, however, this effect is reversed after densification. The tensile modulus of the T5-200 and C5-200 are approximately double the values for their 10 wt.% counterparts.

Figures 4(E) and (F) show the UTS and elongation at failure of the tested samples. For T-TSEs, as the densification pressure increases, the UTS increases whereas the elongation at failure decreases. The binder content has a significant impact on the UTS and elongation at failure of the densified samples, with the elongation at failure more than doubling when going from 5 to 10 wt.% binder content for both the 100 and 200 MPa densified samples. For C-TSEs, the change in both UTS and elongation at failure with the densification and binder content is much less pronounced. Interestingly, the UTS of T- and C-TSEs are comparable to or higher than the reported values for dry-processed LPSCl-based SEs with a fibrous PTFE binder which possesses a significantly higher tensile strength than NBR or XNBR [21, 29, 30].

The above differences between the T- and C-TSEs could be due to the difference in the LPSCl particle size and binder distribution induced by processing. It has been shown in the literature that the difference in tensile modulus of ceramic-polymer composites is independent of the particle size, but rather depends on the adhesion between the two phases present and stress transfer between phases [31–33]. Therefore, the binder distribution must play the dominant role. As shown in figures 3(C) and (D), a more homogeneous distribution of binder surrounding the smaller LPSCl particles in the T-TSEs enables a more efficient transfer of local stresses between the LPSCl and binder, enhanced by the high adhesion between NBR and sulfide particles [32], which accounts for the higher tensile modulus, UTS, and failure strain in these systems. In comparison, the larger LPSCl particles and less well-dispersed XNBR binder in C-TSEs result in a reduced area for adhesion between the binder and LPSCl particles, leading to lower strain to failure, UTS, and tensile modulus.

From the above, it can be found that the difference in mechanical properties is strongly correlated with the microstructure achieved after the processing and densification. The change in microstructure due to densification significantly influences the mechanical properties in both the T- and C-TSEs, rendering them stiffer and less deformable. At the same binder content and densification pressure, it seems that the mechanical properties are more dependent on the binder distribution.

Lithium-ion conductivity is the other important property of a TSE. To evaluate this, EIS was performed using custom SSB testing cells under practical operating conditions (stack pressure of 2.5 MPa, 298 K). Figures 5(A) and (B) show the ionic conductivities of the T- and C-TSEs with different binder contents (5 and 10 wt.%) and densification pressures (0, 100, and 200 MPa).

Zoom In Zoom Out Reset image size

TSEs prepared using both processing methods were found to show improved ionic conductivities with lower binder contents and higher densification pressures. As the binder is non-ionically conductive and would cover parts of the LPSCl particles, a higher proportion of it would reduce the overall conductivity of the TSEs while also reducing the effective contact area between the LPSCl particles and increasing the tortuosity of the Li⁺ pathways [34]. On the other hand, the higher ionic conductivities with higher densification pressures arise from the reduced porosity, creating more ion-transport pathways and reducing the tortuosity across the TSEs.

Comparing the two processing methods, C-TSEs have significantly higher ionic conductivities. T-TSEs all show ionic conductivities below 10⁻⁴ S cm⁻¹ with the highest coming from the T5-200 sample (4.6 × 10⁻⁵ S cm⁻¹), whereas ionic conductivities close to 10⁻⁴ S cm⁻¹ are obtained for C-TSEs with the highest one coming from the C5-200 sample (1.7 × 10⁻⁴ S cm⁻¹). As we use a practical stack pressure of 2.5 MPa, the obtained ionic conductivities are lower than the theoretical value of LPSCl (∼1.0 × 10⁻³ S cm⁻¹) and some other reported values for similar systems, which are usually measured using impractical stack pressures (e.g. > 20 MPa) [34].

Additionally, the ionic conductivities of the C-TSEs are less sensitive to the densification pressure increase than the counterparts of T-TSEs. Specifically, although the porosities of T-TSEs (T5-0, T5-100, and T5-200) are lower than C-TSEs (C5-0, C5-100, and C5-200), the ionic conductivities of C-TSEs are much higher, indicating other factors apart from porosity are the key. The higher ionic conductivities of the C-TSEs can be explained by the following: with the short mixing step in the calendering method, the LPSCl particles were subjected to much less time of the interaction between toluene and LPSCl, which has been observed to have a detrimental effect on the total ionic conductivity of LPSCl powders [35]. Without the high-energy mechanical mixing step used in the tape-casting method, larger LPSCl particles remained in the C-TSEs, which provide longer uninterrupted ion-transport pathways and less interfacial resistance between particles. The XNBR binder in the C-TSEs also tends to randomly distribute at the boundaries of LPSCl particles but does not form a thin insulating layer around the particles, providing fewer barriers and more continuous ion-transport pathways (figure 3(D)). From the above, it is found that the difference in conductivity is not only a function of the binder content but also relies on the SE particle size and binder distribution.

Li₆PS₅Cl (LPSCl) SE powder (99.9%, ∼10 μm, Ampcera Inc., density: 1.64 g cm⁻³), nitrile butadiene rubber (NBR, Krynac 3345F, Arlanxeo, density: 0.97 g cm⁻³), carboxylated nitrile butadiene rubber (XNBR, Krynac X750, Arlanxeo, density: 0.98 g cm⁻³) and anhydrous toluene (99.8%, Sigma Aldrich).

For tape-casting TSEs, the slurries with three compositions (LPSCl and NBR at a weight ratio of 95:5, 90:10, and 80:20) were prepared. Firstly, 500 mg of NBR polymer was fully dissolved in 10 ml of anhydrous toluene. To prepare the slurry, LPSCl was weighed into a ball milling jar before the polymer solution was added. After ball milling for 30 min at 600 rpm in a planetary ball mill (Fritsch Pulverisette 7), the slurry was then coated on a silicone-coated polyester film using the doctor blade technique. After drying at room temperature for 12 h, the TSE was peeled off the substrate, cut, and compressed with stainless steel plungers (10 mm diameter) under a uniaxial press (100 and 200 MPa). The thickness of TSEs with 5 wt.% binder as a function of densification pressure (0, 100, and 200 MPa) is 117 ± 8, 113 ± 7, and 80 ± 5 µm respectively. The thickness of TSEs with 10 wt.% binder as a function of densification pressure (0, 100, and 200 MPa) is 82 ± 6, 75 ± 2, and 64 ± 3 µm respectively. All processes were carried out in an Ar-filled glove box (MBRAUN, MB 200B, H₂O < 0.1 ppm, O₂ < 0.1 ppm).

Calendered TSEs were prepared by previously reported procedures [26]. LPSCl powder and XNBR binder (5, 10 and 20 wt.%) were combined in anhydrous toluene and mixed using agate mortar and pestle or a vortex mixer for 3 min. The composite was placed between silicone-coated polyester films and calendered to a free-standing thin film. The thickness of the prepared TSE was controlled by setting the gap distance of the calendering machine's rollers (MTI Corp., MSK-2150). After drying at room temperature for 12 h, the TSE was cut and compressed with stainless steel plungers (10 mm diameter) under a uniaxial press (100 and 200 MPa). The thickness of TSEs with 5 wt.% binder as a function of densification pressure (0, 100, and 200 MPa) is 111 ± 11, 93 ± 9, and 97 ± 7 µm respectively. The thickness of TSEs with 10 wt.% binder as a function of densification pressure (0, 100, and 200 MPa) is 100 ± 15, 93 ± 3, and 70 ± 9 µm respectively. All processes were carried out in an Ar-filled glove box (MBRAUN, MB 200B, H₂O < 0.1 ppm, O₂ < 0.1 ppm).

Porosity of the TSEs was determined using equation (1). Samples of the TSE were punched (10 mm diameter) and their weight (Sartorius Cubis II Ultra Micro Balance) and thickness (Mitutoyo, 293–344 External Micrometer) were measured. W_TSE and V_TSE are the weight and volume of the TSEs, respectively. $\chi$ _a and ρ_a are the weight fraction and density of component a in the composite [26],

$$\begin{equation}\varepsilon \left( {{\text{porosity}},{ }\% } \right) = 100 \times \left[ {1 - \left( {{W_{{\text{TSE}}}}/{V_{{\text{TSE}}}}} \right) \times \left( {{\chi _{{\text{LPSCl}}}}/{\rho _{{\text{LPSCl}}}} + {\chi _{{\text{polymer}}}}/{\rho _{{\text{polymer}}}}} \right)} \right].\end{equation} \tag{ 1 }$$

Tensile testing was performed using a DEBEN Microtest tensile, compression and horizontal bending stage (MT5000 Dual Leadscrew Tensile Tester H-550 Controller) equipped with a 1 kN load cell operating at 50× gain and with a digital extensometer. Experiments were visually monitored and recorded using a digital camera connected to the ThorCam software. Rectangular-shaped TSE specimens of fixed width and length (0.8 × 3.0 cm) were punched using a die cutter. Their thickness was individually evaluated using a digital screw micrometer gauge with 0.001 mm sensitivity and averaged from three different locations of the specimen. Carefully aligned TSEs were monotonically loaded in tension (extension rate = 1.0 mm min⁻¹) while measuring the resulting force until rupture occurred. At least three specimens were tested for each composition to achieve statistical validation. All the handling and analyses have been carried out in an Ar-filled glove box (H₂O < 0.1 ppm, O₂ < 0.1 ppm) to avoid any possible degradation of the TSEs. Error bars on the stress-strain curves were computed as a relative error considering the average fluctuations in the force recorded by the DEBEN stage to account for the instrumental sensitivity in the employed force range.

Ionic conductivity of the SE (circular disk, 10 mm diameter) was measured by EIS using a VMP-3 potentiostat (Biologic, France) with a voltage amplitude of 10 mV in the frequency range of 1 MHz to 0.01 Hz with the cell-configuration (carbon-coated Al foil/SE/carbon-coated Al foil). The measurement was carried out using a custom-made battery testing cell under a stack pressure of ∼2.5 MPa at 298 K.

Microstructure characterization of the TSE was carried out using plasma focused ion beam scanning electron microscopy (PFIB-SEM) sectioning. PFIB sectioning was performed using a Thermo Fisher Helios G4-CXe Plasma-FIB. The samples were transferred to the PFIB using an airless transfer vessel (Gatan), then milled and polished using a constant voltage of 30 kV and a current down to 15 nA.