1945-7111/165/9/A1829
In this paper, we study how polyolefin separators respond to compressive stress, which can occur during battery operation as active particles lithiate, expand, and push into and deform the separator. We use real microstructures of polyethylene and polypropylene separators acquired with focused ion beam scanning electron microscopic (FIB-SEM) tomography and simulate how these structures deform under compressive strain. After validating our mechanical simulations, we characterize how the microstructural properties of the separators change as a function of compressive strain and simulate the influence these changes in microstructure have on the lithium ion transport through the separator. We find that a given compressive strain negatively impacts the microstructure of polyethylene separator more than that of a polypropylene separator. To understand the origins of the different response to compressive strains in polyethylene and polypropylene separators, we use a network-based analysis to assess the type of mechanical deformations occurring in the separator membrane and show that it is the combination of material properties and structure that are responsible for the greater stability of polypropylene. This work highlights how the structure of a separator plays an important role in its mechanical robustness and prevention of cell degradation.
Export citation and abstract BibTeX RIS
This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any medium, provided the original work is properly cited.
Traditionally, the polymeric separator in a lithium ion battery (LIB) cell is considered to be an inert and electrochemically inactive component;1 however, more and more, studies show that mechanical, thermal, and electrochemical effects occurring in the cell influence separator viscoelastic properties, structure, and surface chemistry.
Upon cell assembly, polyolefin separators can interact with the electrolyte, leading to mechanical softening (i.e., a reduction in the elastic modulus) and swelling.2–6 While cells are assembled under small, uniform compression to improve battery operation as the wettability is increased;7,8 during cycling, separators in cells can be subjected to compressive stresses of up to 5 MPa, due to the volume expansion of active materials and the electrode during lithiation.7,9 These compressive stresses can deform the separator10,11 as shown in Figure 1, which compares cross-sectional images of a polyethylene (PE) separator taken from a commercial cell subjected only to formation cycles in the factory (Figure 1a) and one cycled additionally at 1C for 100 cycles (Figure 1b) (see Section 1 in the Supplementary Material (SM)). While the low melting temperature of a separator can serve as a shutdown feature and contribute to cell safety,12 poor heat dissipation through the separator can limit the discharge rate13 or result in thermal damage to the separator.14
Figure 1. Cross-sectional FIB-SEM images of PE separators of Renata ICP402035 cells harvested from (a) as-received cells with a uniform thickness of 16 μm and a homogeneous pore size distribution, and cycled cells with (b) micron-sized electrode particles locally deforming the separator to about 10 μm and (c) deposits partially filling the separator pore network.
Even at low temperatures, thermal effects can be problematic. Since the ductile-to-brittle transition temperature in polypropylene (PP) occurs at 10°C, a PP separator in a cell operated at low temperature which then internally heats-up can experience structural changes.15,16 Finally, during cycling, solid-electrolyte interface (SEI),17,18 active material particles or reaction products,19 or Li metal deposition20 can block separator pores and enter the pore volume (Figure 1c). Separator areas that are in contact with plated lithium at the anode show crater-like structures of several microns in diameters, and are a result of localized heating, or mechanical or chemical interactions.14
These dynamic changes to the separator in a LIB in turn influence cell performance. Whether mechanical, thermal, or electrochemical in origin, these effects all decrease porosity, ɛ, and/or increase the tortuosity, τ. This leads to a reduced effective transport coefficient in the direction of transport between the electrodes (i.e., the through-plane (TP) direction), δTP = ɛ/τTP, and a reduction in separator TP permeability, κTP.
The effective transport coefficient is a dimensionless scaling factor that describes how the separator structure influences the ionic conductivity and diffusivity of lithium ions in the electrolyte-filled pore space. Namely, the effective diffusivity Deff (or, conductivity σeff) of the lithium ions can be described by Deff = δTP ⋅ D0 (or, σeff = δTP ⋅ σ0), where D0 (or, σ0) is the free diffusivity (or, conductivity) of lithium ions in the electrolyte.
Likewise, the separator's permeability is related to tortuosity of the porous structure, but, with dimensions m2, it includes the dimensionality of the pore and can be thought of as the effective pore channel area of the dynamically active part of the connected pore space.21 Analogously to Ohm's law describing the flow of electrical current, Darcy's law describes the microscopic flow of a liquid through a porous sample of length L subjected to a pressure difference Δp
where u is the flow velocity, κ is the permeability, and η is the viscosity of the fluid in the pores.
Smaller effective transport coefficients and lower permeability imply lower effective ionic conductivity and diffusivity, which leads to larger ohmic drops and larger concentration gradients in the electrolyte across the separator. In turn, this can contribute to higher polarizations (i.e., lower cell voltages), variations in Li-ion insertion and extraction across the cell resulting in lower extracted capacities, and to lithium plating on the anode during charge.13
In this work, we investigate how compressive stresses during cell cycling deform polyethylene (PE) and polypropylene (PP) separators, change microstructural parameters, and affect lithium ion battery performance. Our experimental and computational approach is outlined in Figure 2. We find that the different microstructures of PE and PP separators, which originate from different manufacturing processes, result in different types and extent of mechanical deformation. Based on these insights, we suggest how separator structure could be designed to improve its mechanical robustness and to prevent cell degradation.
Figure 2. Workflow used in this publication indicating experimental (purple fields with dotted frames) and simulation procedures, which are described in detail in the SM. The results of these procedures (pink fields) are used as input (magenta arrows) for further simulations or for comparison with experimental data.
For low compressive stress, the deformation of the separator will be elastic so it will return to its original shape once the stress is removed. This makes it challenging to quantitatively image the microstructure of a separator under different degrees of compression. We therefore use 3D datasets of PE and PP separators and model the mechanical deformation of these structures.
The PE and PP separator structures are imaged by infilling the separators with butter, staining the butter with osmium tetroxide, and performing FIB-SEM based analysis.20,22 The image stacks are binarized, and datasets with 3 μm edge length are selected. Further increasing the dataset size (e.g., to 5 μm edge length) does not impact simulated mechanical or electrochemical properties. These datasets are available open source at Refs. 23 and 24, respectively. The imaged PP dataset only partly shows the amorphous fibers that span across the pore channels and that are of the size of a voxel (10 nm edge length); we therefore computer generate nanofibers and add them to the PP dataset.22
To model the deformation of separator structures due to compressive strains, we import the imaged data sets into GeoDict2017 and use the ElastoDict25 module. We set the properties of the solid phase based on the mechanical properties of bulk PE and PP. Young's modulus is reduced by 10% to account for the mechanical softening that occurs when the polyolefin materials are immersed in a liquid electrolyte26,27 (see Table T1 and Section 2 in the SM). This 10% reduction is selected based on tensile and compressive experiments on polypropylene (PP) separators, where for example the flow stress (i.e., the stress required to sustain deformation) in slow, compressive testing is around 15–20 MPa for dry separators and around 14–18 MPa for immersed separators.26–28 For the Poisson's ratio ν of the immersed separator, we use the estimate ν = 0 by Gor et al.27
We apply uniaxial strain to the separator to compress it in the TP direction. We investigate compressive strains up to 40%, since the plastic (i.e., permanent) deformation we observe in Figure 1b is of this magnitude (the separator is compressed from 16 μm to 10 μm). In our simulations (see Section 3 in the SM), we keep all other boundaries of the simulations fixed, thereby confining the polyolefin deformation to within the original volume. For these simulations, we do not consider a liquid electrolyte in the pore phase. This approximates the condition that the liquid electrolyte is displaced without exerting any resistance when the separator is locally compressed. This is justified because, in its initial immersed state, the separator does not entrap any liquid in closed pores and because the deformations occur slowly enough to allow the electrolyte to displace out of the pore space. With these boundary conditions, the separator does not protrude out of the sides, but can deform and in-fill the pore space, which is reflective of what is observed in a cell (e.g., Figure 1b and Ref. 29).
The simulated microstructures of PE and PP separator datasets under compressive strains of 0–40% are shown in Figure 3. For each of the microstructures, the porosity, tortuosity, and permeability, are calculated using Fickian diffusion simulations in MATLAB, and Stokes' flow simulations via Darcy's law (Sections 4–6 in the SM) in GeoDict's FlowDict module, and tabulated in Table IA (PE) and Table IB (PP). We note that significantly less stress is required to reach the same degree of strain (amount of deformation) in PE compared to PP.
Figure 3. Top-view scanning electronic microscopy images of 3 μm diameter of isotropic PE and anisotropic PP (with transverse direction, TD, and machining direction, MD) separators. Resulting separator microstructure of immersed PE (top) and PP (bottom) datasets (i.e., gray polymer space and pale-yellow electrolyte-filled pore space) of 3 μm edge length under uniaxial compressive strains of 0, 5, 10, 20 and 40% in the through-plane (TP) direction. The compressive stress in MPa applied to achieve the given percent strain is indicated for each structure.
Table I. Dataset dimensions and compressive stresses for separator datasets under compressive strains in the TP direction of up to 40%. This is provided for the (a) PE and (b) PP separators and for the (c) PP separator structure with the Young's modulus of PE.
A Imaged PE microstructure Compressive strain [%] No strain 5 10 20 40 Length of dataset in TP direction [μm] 3.00 2.85 2.70 2.40 1.80 Compressive stress [MPa] Dry - 5.43 ± 0.44 8.55 ± 0.68 12.25 ± 0.90 15.93 ± 1.07 immersed - 4.27 ± 0.33 6.58 ± 0.50 9.17 ± 0.66 11.79 ± 0.79 B Imaged PP microstructure with nanofibers Compressive strain [%] No strain 5 10 20 40 Length of dataset in TP direction [μm] 3.00 2.85 2.70 2.40 1.80 Compressive stress [MPa] Dry - 14.63 ± 0.77 23.08 ± 1.09 30.23 ± 1.25 34.25 ± 1.50 immersed - 11.06 ± 0.47 17.55 ± 0.78 23.31 ± 1.05 28.56 ± 1.30 C PP microstructure with nanofibers and PE properties Compressive strain [%] No strain 5 10 20 40 Length of dataset in TP direction [μm] 3.00 2.85 2.70 2.40 1.80 Compressive stress [MPa] Dry - 10.73 ± 0.57 16.71 ± 0.80 22.44 ± 1.00 28.23 ± 1.29 immersed - 8.22 ± 0.36 12.98 ± 0.58 17.76 ± 0.82 22.47 ± 1.07To assess whether our mechanical simulations are reasonable, we compare our results to data on the compressive properties of polyolefin separators. For example, one study reports a decrease in porosity of ∼20% for a pe separator subjected to a compressive stress of 4 MPa.29 In our case, a stress of 4 MPa on a PE separator corresponds to a compressive strain for 5%, which leads to a reduction in porosity from 40% to 35% (i.e., a reduction of ∼12.7%). Given that the PE separator studied here is not the same as that studied in Ref. 29 and therefore they do not have identical microstructures, the experimentally measured and simulated decrease in porosity are in good qualitative agreement, and we conclude that our simulations are providing us with reasonable insights into the changes in microstructure that occur under compressive strain.
Furthermore, Figure 4 shows the stress-strain curves for dry and immersed PE and PP separators. Given that the bulk PE and PP values used in our simulations do not take into account the degree of crystallinity, alignment, and orientation of the crystalline and amorphous regions of the separators, the curves show good agreement with reference data (dotted lines) for uniaxially stretched PE and PP separators of 36–46% and 55% porosity, respectively,26,30 which are subjected to compressive stress and then unloaded (i.e., they reach zero stress at a remaining plastic deformation). The length of the plateau (which corresponds to compressive strains for which the material bulges into the pore space) depends on the separator's porosity, and the subsequent upward-turning regime represents compression of the polymer itself.26
Figure 4. Stress-strain curves for dry processed, uniaxially stretched PE30 and PP26 separators (dotted lines) compressed up to 100 and 50 MPa, respectively, and unloaded yielding compressive strains of ∼40% (i.e., plastic deformation) at zero stress. Calculated compressive stresses at given strains for the PE (circles) and PP (squares) separator structures, with the mechanical properties of dry (purple) and immersed (pink) PE (left) and PP (right).
The curves in Figure 4 again show that less stress is required to strain the PE than PP separator. While this is in part due to the lower Young's modulus of PE (1.2 GPa for the bulk) than in PP (1.5 GPa for the bulk), the plots of the stress-strain curve for the PP separator structure with the PE mechanical properties (left) are very similar to those for the PP separator structure with PP properties (right) highlighting that the structure of the separator plays an important role in the compressive properties. We note in fact that the PP structure with PE properties yields a stress-strain curve which is more similar to the reference data measured in Ref. 30 than the stress-strain curve for the PE separator. This is likely due to fact that the reference data comes from a PE separator, which has a PP-like structure due its dry-stretched fabrication.
We next analyze how pressure-induced deformation affects the separator microstructural properties. As we compress the separators in the TP direction, the separators show decreasing porosity, increasing tortuosity, and decreasing permeability (Table I and Section 7 in the SM), indicating that the separator material compresses into the pore space.
In Figure 5, we compare PE and PP structures at the same percent strains (i.e., the same % reduction in thickness) and find similar changes to the microstructure. Under a compressive strain of 5%, PE and PP separators decrease their porosity (Figure 5a) to 35% and 32% (i.e., by 12.7 and 9.3% compared to their dry, non-compressed states). At 40% compressive strain, the separators have not completely densified, but exhibit low porosities around 14% and 11% for PE and PP, respectively.
Figure 5. Evolution of (a) porosity; (b) TP, (c) IP1, and (d) IP2 tortuosity; (e) TP, (f) IP1, and (g) IP2 permeability; and (h) connectivity density for uniaxially compressed datasets of PE (circles) and PP (squares) separators.
For compressive strains up to 20%, the TP tortuosity increases similarly for both PE and PP from 2.78 to 4.26 and from 2.32 to 4.08 (Figure 5b). For 40%, compressive strain the tortuosity increases up to tenfold, indicating closed-off pores. For the PE separator, the in-plane (IP) tortuosities (Figures 5c–5d) are similar to those of the TP tortuosity; however, the PP separator exhibits IP2 tortuosities that are around 10 times higher than in the TP and IP1 directions. At compressive strains above 10%, the PP datasets do not exhibit any connected transport paths along the IP2 direction.
The TP permeability (Figure 5e) of the PE and PP separators decreases linearly, and, at 40% compressive strain, it is two orders of magnitude smaller (3.61 and 8.51 · 10−19 m2) than in the non-compressed state (503.56 and 1013.28 · 10−19 m2) confirming that the effective pore size and the connected pore space has dramatically reduced (see Section 7 in the SM). The isotropic and anisotropic behaviors of the wet-stretched PE and the dry-stretched PP separator under compressive strains are reflected in the IP permeabilities (Figures 5f–5g). In accordance with the IP2 tortuosity, we find a much lower permeability in the IP2 direction of the PP separator, and zero permeability for compressive strains above 10%. We tabulate all separator characteristics and their evolution under compressive strains in Table T4 in the SM.
To gain additional insight into the changes occurring in the polymer separator, we consider the separator pore space to be a network of branches and nodes, and we calculate the connectivity density (number of redundant connections per unit volume) of the pore phase using ImageJ's 3D Skeletonization and AnalyzeSkeleton plugins. The connectivity density gives insight into how connected a structure is and has previously been used to described the pore space of LIB separators.22
For both PE and PP under compressive strains up to 20%, the connectivity density of the pore space increases by ∼23 and ∼65%, respectively compared to the dry, non-compressed state (Figure 5h). This is because the polymer bulges into the pores adding more complexity to the pore space. The absolute number of branches, nodes, and end-point branches22 needed to describe the pore space's connectivity increases, while the separator volume used for normalization becomes smaller. The subsequent decrease in pore space connectivity density at higher compressive strain is explained by the densification of the separator: as pore space disappears, we need fewer branches, nodes, and end-point branches to describe it.
The trends in microstructural parameters as a function of compressive strain shown in Figure 5 highlight that (1) even though stress is uniaxial, the polymer expands in all directions into the pore space, (2) for a given amount of strain, the changes in pore structure are similar for both structures, and (3) the deformations alter key lithium ion transport parameters of the membrane.
To quantify how this compressive deformation of separators affects lithium ion battery performance, we perform 1D COMSOL simulations (see details and list of parameters in Section 8 in the SM). We use simulation parameters that best fit the operation characteristics of a Samsung 18650 cell, which contains a nickel manganese cobalt oxide (NMC) based cathode and a graphite anode and which is rated for up to 8C discharge. To determine the effect of a given compressive strain, we change the porosity and tortuosity of the separator (based on the values calculated in Table I), while keeping all other parameters the same. In Section 8 in the SM, we also provide the simulation results for a symmetrical (i.e., Li vs. Li) cell, analogous to our previous work in Ref. 20.
We plot the voltage drop across a PE (Figure 6a) and PP (Figure 6b) separator in a commercial cell at different C-rates for compressive strains of 0–40%. We find that the voltage drop in the electrolyte across the separators increases linearly for separators under compressive strains of up to 10% for both PE and PP separators. For separators under a compressive strain of 20%, the voltage drops exceed 50 mV for C-rates above 2C. This high increase in voltage drop is caused by a high ion gradient in the electrolyte that influences the ion mobility and thereby the electrolyte conductivity non-linearly.
Figure 6. Calculated voltage drop in the electrolyte across PE (left) and PP (right) separators under uniaxial compressive strains of 0, 5, 10, 20 and 40% at different C-rates for a graphite vs. NMC cell. The stress needed to achieve the compressive strain is indicated for each curve.
While 11 MPa of applied stress will have little effect on a PP separator (leading to only ∼5% strain), it will lead to 40% compressive strain in PE as was observed for the region of separator in Figure 1b. For this percent strain, the potential drop exceeds 50 mV at C-rates as low as 0.5C.
These simulations highlight that pressure applied to a region of separator, particularly a PE separator, can lead to local potential drops that exceed the thresholds for safe LIB operation, as illustrated by the ion gradients in Figure S5 in Section 8 in the SM. For example, these inhomogeneities in potential can cause local metallic Li deposition on graphite anodes during charge and subsequent Li growth into the separator.20,22,31 It can also lead to cathode material and electrolyte degradation.32,33 Electrolyte oxidation results in gas and oxidized species formation, and these side reaction products can migrate toward the anode and interact with it.34 These local effects are reflected in overall cell capacity fade, and they lead to a significant reduction in LIB cycle life and safety.14,35,36
Our findings show that PE microstructures deform more in response to compressive stress than PP, which also negatively impacts their performance in a lithium ion battery.
To determine how much deformation is due to the fact that the PE separator has a lower bulk Young's modulus than PP, and how much is due to the different structure of PE and PP separators, we apply the elastomechanical properties of immersed PE (Young's Modulus of 1.08 MPa and a Poisson's ratio of 0) to the PP structure and repeat the mechanical simulations and microstructural analysis on the resulting microstructures. The resulting stresses are listed in Table IC and the stress-strain curves are shown in Figure 4 (left). In Tables IIA and IIC, we directly compare how the porosity, TP tortuosity, and TP permeability of the microstructures change under stress. For a given strain, the porosities, tortuosities, permeabilities and connectivity densities of the deformed PP separator and PP separator structure with PE properties are essentially the same.
Table II. Evolution of separator performance characteristics for separator datasets of (a) PE and (b) PP separators and of (c) PP structure with PE properties under compressive strains in the TP direction of up to 40%.
A Imaged PE microstructure Compressive strain [%] No strain 5 10 20 40 Porosity ɛ [%] 40.02 ± 0.79 34.93 ± 0.73 31.85 ± 0.74 25.45 ± 0.80 14.40 ± 0.91 Tortuosity τTP [-] 2.78 ± 0.05 3.03 ± 0.05 3.28 ± 0.03 4.26 ± 0.12 26.41 ± 7.64 Effective transport coefficient δTP [%] 14.40 ± 0.09 11.54 ± 0.11 9.70 ± 0.14 5.97 ± 0.28 0.59 ± 0.23 Permeability κTP [10−19 m2] 503.56 ± 13.42 311.36 ± 10.21 218.39 ± 9.88 80.38 ± 28.09 3.63 ± 1.77 B Imaged PP microstructure with nanofibers Compressive strain [%] No strain 5 10 20 40 Porosity ɛ [%] 35.33 ± 1.54 32.04 ± 1.64 29.28 ± 1.67 22.95 ± 1.97 11.14 ± 2.27 Tortuosity τTP [-] 2.32 ± 0.15 2.64 ± 0.14 2.98 ± 0.19 4.08 ± 0.52 14.38 ± 2.41 Effective transport coefficient δTP [%] 15.26 ± 0.89 12.14 ± 0.54 9.85 ± 0.46 5.66 ± 0.49 0.77 ± 0.03 Permeability κTP [10−19 m2] 1013.28 ± 97.74 603.64 ± 58.11 411.75 ± 43.96 149.91 ± 27.97 8.51 ± 0.09 C PP microstructure with nanofibers and PE properties Compressive strain [%] No strain 5 10 20 40 Porosity ɛ [%] 35.33 ± 1.54 31.96 ± 1.65 29.14 ± 1.67 22.77 ± 1.95 10.39 ± 2.61 Tortuosity τTP [-] 2.32 ± 0.15 2.65 ± 0.14 2.98 ± 0.20 4.08 ± 0.50 14.57 ± 3.33 Effective transport coefficient δTP [%] 15.26 ± 0.89 12.07 ± 0.54 9.78 ± 0.46 5.60 ± 0.45 0.71 ± 0.02 Permeability κTP [10−19 m2] 1013.28 ± 97.74 598.63 ± 57.90 406.21 ± 43.38 146.96 ± 26.19 7.77 ± 0.18Although the compressive stresses required to deform the anisotropic PP structure with PE properties are smaller than those needed to deform the PP separator, which is due to the lower Young's modulus of PE, they are still greater than those of the PE separator, indicating the importance of the separator structure.
There are several structural differences between PE and PP separators. First, the PE separator exhibits smaller pores (geometric pore diameter D50geo is ∼112 nm and the pore throat diameter D50throat is ∼84 nm, see Section 7 in the SM), which get blocked or cutoff at lower strains than the larger diameter pores of the PP separator (D50geo is ∼156 nm and D50throat is ∼91 nm).
Furthermore, the mechanical properties of a porous membrane are determined by its structure. As explained previously, we can consider the separator membrane as a network, and describe its structure by branches which intersect at nodes. The number of branches that intersect at a node is called the node order (i.e., a node where 3 branches intersect has a node order of 3). A useful parameter to consider are the average angles, Φ, between branches and a node.37 As illustrated in Figure 7, a highly symmetrical structure will have branches that intersect at angles of 120° for nodes of order 3, 109.5° for nodes of order 4, 100° for nodes of order 5, and 90° for nodes of order 6.
Figure 7. Node angle, φ, conformation for ideal nodes (left) and node angle distribution for polymer phase nodes of orders 3–6 for datasets of immersed PE and PP separators of 3 μm edge length without load (pink line) and under 40% compressive strain (purple line). The ideal node distribution center angle, Φ, for each node order is indicated by gray lines.
In Figure 7, we show the normalized distributions of node angles φ between polymer branches for nodes of order 3 to 6 for PE and PP separators without compression and under a compressive strain of 40% in the TP direction. The non-compressed PE structure's node angle distributions are well centered around the ideal angles. This is consistent with wet-stretched PE separators being highly isotropic,20 and implies that PE separators tolerate multidirectional loading. In contrast, the node angle distributions in the non-compressed PP structure are skewed for nodes of order 3 due to the anisotropic structure of PP22,38 (Sections 9–10 in the SM). The inset illustrates how the presence of nanofibers for example in the transverse direction leads to angles for nodes of orders 3 and 4 that are smaller and larger than 120° (i.e., near 90° and 180°). While PP is susceptible to transverse shear (i.e., tearing),39 its anisotropic structure tolerates compression in the TP direction.
Indeed, if we uniaxially compress structures with symmetrical node angle conformations (left side of Figure 7), the nodes that do not exhibit angles of 90° are more likely to undergo changes in the node angle distributions. Thus, we expect measurable changes for angles of nodes of orders 3 for PE. Indeed, the distribution of node angles broadens as we compress the separator by 40%. For nodes of order 5 and 6, we don't observe significant broadening; however, the distributions become less smooth, and show more variance in frequency.
While the node angle distribution highlights that PE is more mechanical stable in general, the arrangement of node angles in PP makes it better able to tolerate uniaxial compression in the TP direction.
Here, we have quantified the mechanical properties of PE and PP separators based on imaged datasets and bulk mechanical properties, and the compressive stresses for strains between 5 and 40% in the TP direction. With increasing separator deformation, transport of lithium ions through the separator is compromised. For highly deformed regions of separators, normal cell operation becomes impossible at C-rates as low as 0.75C.
PE separators experience more deformation than PP separators for a given compressive stress. This is not only because PE has a lower Young's modulus, but also because the PE separators investigated here have smaller pore sizes as well as a more isotropic structure.
Interestingly, these same structural properties that make PE separators more susceptible to compressive stress than PP separators make them superior under other conditions. For example, we previously found that the high pore space connectivity of PE separators leads to more uniform distributions of electrolyte salt concentration in neighboring pores and make them more capable of homogenising concentration gradients due to defects that block separators pores.22 This trade-off is illustrated in Figure 8.
Figure 8. (a) Cross-sectional views of example structures with high and with no pore connectivity. (b) Blocked pores, here due to electrode particles touching the separator surface, do not affect ion transport across highly connected structures but lead to isolated pores in structures with zero connectivity. (c) Deformation of highly connected structures – here due to electrode particles pressing into the separator – can lead to the collapse of the pore network and locally impermeable regions, while the same amount of stress only mildly affects the structure of no connectivity and the ion transport across it.
Ideally, a separator microstructure would be designed so that the paths of ion transport remain connected under compressive loads in the in-plane and through-plane directions. This can be achieved by keeping an isotropic and symmetric structure as found in PE and increasing the mechanical rigidity of the separator and increasing the average pore size. However, finding polymeric materials that can be processed at low cost with these properties and that exhibit good wetting properties with electrolyte blend, the necessary thermal properties, and (electro-)chemical stability is non-trivial.
Three-layer polyolefin separators (i.e., PP/PE/PP) may seem like the logical choice to combine the advantages of both materials. However, these separators exhibit similar deformation to single layer PE and PP separators.30 And, while these separators have the benefit of thermal and mechanical stability of PP, they also likely bring the structural disadvantages of PP (i.e., a low connectivity density) since they are produced in a dry process with uniaxial stretching.
Wet-stretched ceramic coated PE separators combine the advantage of high connectivity density of biaxially stretched PE with the high thermal and mechanical stability of ceramic particles. Under uniaxial compression, these separators have different deformation properties than PE separators40 in that they do not exhibit a clear yield point and a flow plateau. However, as the PE separator is softer than the ceramic particles, the PE microstructure likely deforms plastically under uniaxial compression and may flow between the ceramic particles. Therefore, while ceramic coated PE separators exhibit good safety features for intact coatings, they may effectively hinder ion transport when under uniaxial compression and provide limited safety if the coatings are damaged.
These solutions still do not simultaneously optimize connectivity of pores and mechanical stability under compressive stresses, both of which are necessary to prevent local inhomogeneities in electrochemical activity. There is unquestionably a need for further innovation in separator technology space.
Heterogeneous aging strongly affects LIB performance and this works highlights how separators are among the weakest spots of current generation LIBs. While future separator generations – such as polyetherimide (PEI) membranes of asymmetrical structure,41 and, eventually, solid electrolytes42 – are studied extensively, the underlying uncertainty about the mechanical stability and integrity of this safety layer remains. We propose that mechanical and electrochemical simulations based on imaged microstructures will help us better understand the effect of local degradation phenomena in energy storage applications.
M.F.L. thanks Simon Müller for work on adding nanofibers to the PP datasets; Solène Bastien and Sophie Vanderspar for cell cycling and disassembly; and Martin Ebner for providing simulation parameters of a commercial battery. The authors acknowledge funding from ETH Research and ERC (680070) Grants.
Marie Francine Lagadec 0000-0002-6214-5171
Raphael Zahn 0000-0003-0223-0697
Vanessa Wood 0000-0001-6435-0227