Etched ion-track membranes as tailored separators in Li–S batteries

Lithium–sulfur (Li–S) batteries are considered a promising next generation alternative to lithium-ion batteries for energy storage systems due to its high energy density. However, several challenges, such as the polysulfide redox shuttle causing self-discharge of the battery, remain unresolved. In this paper, we explore the use of polymer etched ion-track membranes as separators in Li–S batteries to mitigate the redox shuttle effect. Compared to commercial separators, their unique advantages lie in their very narrow pore size distribution, and the possibility to tailor and optimize the density, geometry, and diameter of the nanopores in an independent manner. Various polyethylene terephthalate membranes with diameters between 22 and 198 nm and different porosities were successfully integrated into Li–S coin cells. The reported coulombic efficiency of up to 97% with minor reduction in capacity opens a pathway to potentially address the polysulfide redox shuttle in Li–S batteries using tailored membranes.


Introduction
Lithium-sulfur (Li-S) batteries are considered a promising next generation alternative to the current generation of lithium-ion batteries. In terms of energy, Li-S systems offer a far higher theoretical gravimetric energy density (2567 Wh kg −1 ) than state-of-the-art lithium-ion systems (387 Wh kg −1 ), taking into account the electroactive mass [1]. Sulfur is cheap and abundant, being a side-product of the crude oil refinement process. The double effect of potentially lowered cost and higher energy density make Li-S batteries incredibly attractive for future development.
Considerable research effort has been devoted to overcome the challenges facing Li-S batteries, including selfdischarge [2][3][4], lithium dendrite formation on the lithium metal anode [5,6], and the effect known as the polysulfide redox shuttle [7,8]. The issues of the self-discharge and the polysulfide redox shuttle owe their origins to the soluble nature of the lithium polysulfides (S x 2− , 2x8). At the cathode, long chain polysulfide molecules are formed as intermediate products, which diffuse to and react with the lithium anode, forming lower order polysulfide molecules Li 2 S x , and diffusing back to the cathode again [9]. This shuttling of the molecules between the two sides is one of the causes of the lithium-sulfur system's severe self-discharge, which lowers its coulombic efficiency. Some examples of strategies to mitigate this effect include modifying the cathode or using additives to trap the produced polysulfides and prevent their diffusion out of the cathode [10][11][12], modifying the separator in order to bind or prevent the movement of the polysulfides towards the anode [13][14][15], using metal-organic framework membranes as ionic sieves [16,17], using intrinsically microporous polymer membranes to control polysulfide diffusion [18,19], and using functionalized ion-selective polymers as membranes [20,21].
Commercial separators in lithium-ion batteries are typically made of stretched polymer films such as polyethylene and polypropylene. They exhibit thicknesses on the order of tens of micrometers and large porosities with pore diameters on the order of micrometers. The large porosity is attained by stretching the polymer film, resulting in large distributions of the pore sizes and geometries.
Etched ion-track membranes, in turn, exhibit a very narrow pore size distribution, and density, geometry, and diameter of the nanopores can be the tailored and adjusted in an independent manner. Etched ion-track membranes are formed by irradiating polymer foils with highly energetic heavy ions, followed by wet chemical etching. The technique is based on the fact that an individual heavy ion creates a cylindrical damaged zone along its trajectory, the so-called ion track. Within this ion track, the polymer's chemical and physical properties are severely modified [22,23]. By immersing the irradiated foil in an appropriate etchant, ion tracks are selectively dissolved and converted into open micro or nanochannels [24]. For the successful fabrication of membranes, the etching rate along the ion track, V T , must be higher than the isotropic etching rate of the undamaged bulk material, V B ( figure 1(a)). The material of choice and the etching conditions (temperature, composition, and concentration of the etchant) determine the track-to-bulk etching ratio (V T /V B ) and therefore also the geometry of the channels [25,26]. Under symmetric etching conditions, low V T /V B ratios result in the formation of bi-conical channels ( figure 1(b)), while high V T /V B ratios result in the formation of cylindrical channels (figure 1(c)). High etching selectivity for ion tracks in polyethylene terephthalate (PET) is achieved by using sodium hydroxide (NaOH) solution [23]. Exposure of ion-irradiated polymers to UV light prior to etching increases V T and leads to a narrower size distribution of the channels [27][28][29]. Under suitable fabrication conditions, membranes with monodispersed nanochannels can be fabricated [26]. Etched ion-track membranes can be obtained commercially and are widely used in filtration, surface protection, and biology. However, the application of etched ion-track membranes as separators in batteries has not been pursued until now [30].
In this work, we apply PET etched ion-track membranes with monodispersed cylindrical channels as separators in Li-S batteries. We compare the overall performance of Li-S batteries as a function of the channel size and pore density of the PET membranes. PET membranes are also applied to study the diffusion of polysulfides as a function of nanochannel size. We postulate that by applying polymer membranes with optimized porosity and channel size, the shuttle of polysulfide molecules between cathode and anode can be restricted while maintaining a sufficient transport of lithium ions. The results show that Li-S batteries assembled with an additional PET ion-track membrane exhibit improved coulombic efficiency compared to reference batteries assembled only with a commercial separator.

Synthesis of etched ion-track membranes
Commercial PET foils (Hostaphan RN, 19 μm thick, 3 cm diameter) were irradiated with 2 GeV gold ions at the universal Linear Accelerator (UNILAC) of the GSI Helmholtz Center for Heavy Ion Research. The foils were irradiated under perpendicular incidence, and the irradiation fluence was varied between 10 7 and 10 9 ions cm −2 with an accuracy of about ±20%. Each ion creates along its trajectory in the PET foil, a cylindrical damage zone called an ion track [22,23]. Before chemical etching, each side of the irradiated foils was exposed to ultraviolet light (Vilber Lourmat UV tube, 312 nm peak wavelength) for three hours. This process sensitizes the ion tracks created by the ions, yielding membranes with a narrow pore size distribution [27][28][29]. Track etching was performed by immersing the irradiated and sensitized PET foils in an aqueous solution of 6 M NaOH at 50°C. Under these conditions, the anisotropic etching rate of the material along the damaged ion-track is much higher than the isotropic etching rate of the bulk undamaged material, resulting in high aspect ratio cylindrical channels (figure 1). The resulting channel diameter was adjusted between 22 and 198 nm by selecting etching times between 60 and 330 s. The thickness of the PET membrane during etching decreases only by the corresponding pore diameter. After etching, the PET membranes were rinsed in distilled deionized water. Prior to their assembly in Li-S batteries, the etched membranes were punched into discs with a diameter of 1.6 cm and dried at 70°C under vacuum for 24 h.
Typical commercial battery separators exhibit polydispersed pores of irregular shapes and sizes on the order of up to micrometers [31]. Figure 2(a) shows an exemplary scanning electron microscopy (SEM) image of a SK innovation separator. The ion-track membranes, on the other hand, exhibit parallel oriented, randomly distributed, high aspect ratio pores with cylindrical geometry and a narrow size distribution (figures 2(b)-(c)).

Synthesis of carbon-sulfur composite for cathode
To synthesize the sulfur cathode, mesoporous carbon and sulfur powder were first heat-treated for 24 h under 300°C and 60°C respectively to remove moisture. The two treated materials were then mixed in a 50:50 weight percentage mixture, ground in a mortar, and heated at 155°C for 6 h in a sealed autoclave to impregnate the mesoporous carbon with the sulfur. Mixing and sealing in the autoclave was performed under an argon atmosphere.
By mixing the carbon-sulfur composite with carbon Super-P, poly (vinylidene fluoride-hexafluoropropylene) (PVdF-HFP) and N-Methyl-2-pyrrolidone (NMP) a soft slurry was composed. Carbon Super-P is used to increase the electrical conductivity of the composite, while PVDF-HFP is a binder for the slurry that improves its mechanical properties. The slurry was then casted over a roughened aluminum foil at an elevated temperature of 40°C to prevent water adsorption from the atmosphere. The dried slurry casted foil is then cut into 8 mm diameter circular discs to be later integrated as cathodes into the coin cells. Each disc contains 37.5% of sulfur by weight and is weighted to find the sulfur mass per cathode before cell assembly.

Cell assembly and testing
The PET membranes were integrated into Li-S coin cells to test their cycling performance. The PET membrane is sandwiched between two SK Innovation separators, which acted as a reservoir for the electrolyte. The electrolyte used was 1 M lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) in TEGDME:1,3-Dioxolane 1:1 by volume percentage solution. Electrolyte volume to sulfur mass ratio is a relative property, the dissolution of polysulfides and the polysulfide shuttle effect increase with this ratio regardless of the mass percentage of sulfur in the electrode. As this work focuses on the tailoring of the PET membrane to limit the shuttle effect, a high electrolyte volume to sulfur mass of ratio of 20 μl mg −1 was utilized in the cells. Each cathode was weighted before cell assembly to calculate the sulfur content, and thus the needed electrolyte volume. Coin cells were assembled with each layer in quick succession, starting with a carbon-sulfur cathode, the separator with reservoirs (SK separator/tracketched PET membrane/SK separator with half the electrolyte volume pipetted on each SK separator), and a lithium foil anode (figure 3). The cell was assembled in an argon glovebox and sealed in a coin cell before cycling. Cycling commenced as soon as possible after sealing the cell (<10 min). Cycling of the coin cells was performed at a C-rate of C/10, under galvanostatic cycling with potential limitation from 1.5 to 3 V and at a constant temperature of 20°C. The polysulfide shuttle effect in Li-S batteries is mainly observed as overcharging during each cycle, whereas the capacity fading during cycling could be caused due to multiple phenomena such as detachment of lithium metal from the anode, electrolyte decomposition, or blockage of cathode pores. Therefore, we believe that the capacity fading during cycling is not the proper descriptor to prove the effectiveness of the iontrack membrane to restrict the shuttle effect and have limited the number of cycles to only 15. The decrease or the absence of overcharging of the cells which contain the etched iontrack membrane prove that they are effective in restricting the shuttle effect.

Results
To study the performance of the Li-S batteries as a function of the adjustable separator parameters, PET membranes with cylindrical channels with average diameters ranging from 22 to 198 nm (±10 nm) and channel densities between 1×10 7 and 1×10 9 cm −2 were integrated into coin cells. The pore where F·A denotes the nominal porosity. F is the pore density which corresponds to the fluence applied during ion irradiation (number of impinging ions per cm 2 ), because each individual ion produces a track. A is the pore cross section (A=π·r 2 , with r being the channel radius). The pore density is measured by SEM, by counting the number of pores in a given area. The homogeneity of the pore density is typically within 10%-20%. As reference, control cells were assembled with only one SK Innovation separator and without the track-etched PET membrane. For clarity, only one curve by each type of membrane is presented here. Tests of etched ion-track membranes with 10 9 channels cm −2 and average channel diameter of 22 nm showed no discharge, even after 24 h, and thus were not further investigated. Figure 4 shows the specific charge and discharge capacity (a) and the coulombic efficiency (b) recorded for Li-S cell coins assembled with PET membranes with 10 9 channels cm −2 and average channel diameters 115 nm (red), 168 nm (blue), and 198 nm (green) together with the reference cell (black).
The data shows that the coin cell assembled with the PET membrane with channels of ∼115 nm exhibits a starting discharge capacity of 822 mAh g −1 and drops to 726 mAh g −1 by the 15th cycle. By comparison, the reference coin cell exhibits a starting discharge capacity of 1311 mAh g −1 , dropping to 850 mAh g −1 by the 15th cycle. There is also a significant 14% improvement of the coulombic efficiency from ∼83% of the reference cell to 97% for the coin cell containing the PET membrane with channels of ∼115 nm. We attribute this increase to a reduction of the polysulfide redox shuttle by the presence of the PET membrane with small monodispersed cylindrical channels. As the data shows, this increase comes at the cost of a reduced discharge capacity, caused by the decrease in porosity. An increase in the average channel diameter of the PET membrane from 115 to 168 nm and 191 nm (corresponding to a porosity of 19.9% and 24.9%, respectively) resulted in higher specific charge and discharge capacities. This is expected because the higher porosity yields a higher lithium conductivity, as well as lower coulombic efficiencies. The positive effect of the membrane with 115 nm pores on the coulombic efficiency can have several origins that are currently being discussed and investigated. Interaction between the polysulfides and the pore walls possibly affects the passage of the polysulfides from the cathode to the anode. PET nanopores are ion-selective under certain conditions [33].
The PET membrane is thus able to compare favorably efficiency-wise (∼97%) to other polysulfide-blocking polymeric membranes reported in literature. As points of comparison, an ion-selective lithiated Nafion ionomer membrane reported by Jin et al exhibited coulombic efficiencies per cycle of above 97% [20], and a polymer with intrinsic nanoporosity membrane reported by Yu et al exhibited coulombic efficiencies of 99.72% [19]. Figure 5 shows the cycling performance of coin cells assembled with PET membranes with average channel diameter 115 nm and various channel densities, namely 10 7 cm −2 (red squares), 10 8 cm −2 (blue circles), and 10 9 cm −2 (green diamonds). In this case, after 15 cycles, the coulombic efficiencies of the three cells are very similar (95%-99%). However, the cells assembled with PET membranes with channel densities of 10 8 cm −2 and 10 7 cm −2 exhibit much lower charging and discharging capacities. The lower capacity for the membrane with 10 7 pores cm −2 suggests that the lower pore densities limit the transport of lithium ions between the anode and cathode. Figure 6 displays the cycling performance of coin cells assembled with PET membranes with a larger average channel diameter of 168 nm and the same three different channel densities, 10 9 cm −2 (green diamonds), 10 8 cm −2 (blue circles), and 10 7 cm −2 (red squares). In figure 6(a), the cells with membranes with 10 9 channels cm −2 and porosity of 19.9% exhibit charging and discharging capacities of similar value to those of the reference cell ( figure 4(a)). Reducing the  Pore density F (cm −2 ) Pore diameter (nm) P eff (%)    channel density tenfold to 10 8 cm −2 reduced the charge and discharge capacities by about 150 mAh g −1 but increased the coulombic efficiency by 8%. Reducing the channel density further to 10 7 cm −2 cuts the capacity further to about 350 mAh g −1 at the 15th cycle. Figure 7 shows the corresponding cycling curves of the first 5 cycles of these same coin cells. The cell with PET membrane of 10 9 cm −2 ( figure 7(a)) shows only a small shift per cycle, corresponding to the relatively high coulombic efficiency of the cell. The cell with the PET membrane with 10 8 cm −2 channel density ( figure 7(b)) shows a discharge curve with higher discharge potential, and a short discharge plateau, indicative of more restricted ionic flow. The discharge curve in figure 7(b) is also shorter than that of the charge curve in each cycle after the initial one, possibly due to the difficulty of lithium ions reaching the cathode due to the low porosity of the membrane. Further reducing the channel density to 10 7 cm −2 (figure 7(c)) shows much higher resistance between the cathode and anode. The discharge and charge plateaus are short, indicating that polysulfide concentration is low due to severely limited ionic flow, leading to sharply reduced capacity ( figure 5(a)).
For all membranes with pore densities of 10 7 cm −2 , we measured coulombic efficiencies of the cells above 100%. This effect which appears with the initial cycles is likely due to the activation of sulfur, whereby more lithium ions in successive cycles are reaching sulfur that had not been utilized in any reaction during the previous cycles. This could be an indication that the low number of pores is severely reducing the ionic flow between the cathode and the anode such that the Li ions on the cathode side of the membrane are consumed faster than they can diffuse through the membrane from the anode. During each cycle, new Li ions would reach the carbon-sulfur cathode after diffusing through the membrane, driving up the coulombic efficiency.
These systematic studies of Li-S batteries assembled with an additional PET separator clearly indicate that the performance of the battery, i.e. capacity retention and coulombic efficiency, can be influenced by the porosity and channel diameter of separators with tailored channel parameters. The best performance was achieved for the PET separator with d ∼115 nm, and pore density 10 9 cm −2 .
To estimate the influence of the nanochannel diameter on the diffusion of the polysulfides in the batteries, 19 μm thick PET membranes, with a pore density of 10 9 cm −2 and with different average pore sizes were used to tightly close vials containing a 0.25 M nominal Li 2 S 6 polysulfide solution. This corresponds to the concentration value expected in the coin cells tests if about all of the sulfur in the cell is converted to Li 2 S 6 within the electrolyte (according to the employed 20 μl per mg of S). The vials were then inserted upside down in a pure solvent solution of 1:1 TEGDME: DOL, with the etched ion-track membranes being the only barrier between the polysulfide and the solvent solutions ( figure 8).
The diffusion of polysulfide molecules along the nanochannels of the membrane caused by the concentration gradient was then monitored visually for 30 days, with photos taken hourly for the first 8 hours, and every 24 hours subsequently. Figure 9 shows the photographs taken for vials with PET membrane-separators with pore diameter ∼74 nm  (top row), 42 nm (middle row) and 20 nm (lower row). The porosity of these three membranes varies between 4.2% and 0.3%. For all vials, there was no visually apparent change in the liquid levels for the polysulfide or the pure solvent solution during the course of the experiment. For the PET separators with largest nanochannels (diameter∼74 nm) the diffusion of the polysulfides across the membrane under this maximal concentration gradient happens relatively quickly, within 24 hours. For the PET separators with average pore diameter ∼42 nm only a small amount of polysulfides diffused through during the first 24 hours, and even after 30 days, a certain concentration gradient is still present. Finally, reducing the average pore diameter down to 20 nm prevented the diffusion of the polysulfides through up to 30 days.
These diffusion experiments were performed under large polysulfide concentration gradients and indicate that the diffusion of lithium polysulfides can be strongly affected by proper tailoring of the separator nanochannel diameter. The better performance exhibited by some of the Li-S cells with PET separators compared to the reference one can be explained by a reduction of the lithium polysulfide diffusion, due to a lack of an initial polysulfide concentration gradient in the coin cells. In-situ analysis of the Li-S cell during operation by Patel et al showed that evolution of polysulfides begins during the initial discharge, and their concentration increases and decreases with the charge and discharge cycles of the cell [34]. The results indicate that the presence of an ion-track membrane with monodispersed channels and optimized porosity in a Li-S battery can hinder the diffusion of the polysulfides across to the anode and improve the battery performance. Future experiments will be devoted to further optimization of channel diameter and porosity, as well as to the application of tailored membranes of other polymer materials.

Conclusion
The advantages of applying membranes with monodispersed channels and tailored parameters as separators in battery systems have been discussed. A systematic variation of the individual separator parameters in an independent manner enables the investigation of diffusion and electrochemical processes involved, and thus the optimization of the separator parameters and the battery performance. Etched ion-track membranes with average pore diameters 115, 168, and 198 nm, and pore densities 10 7 , 10 8 , and 10 9 cm −2 were tested. An increased coulombic efficiency was observed particularly for PET separators with porosities ∼10% and pore diameter ∼115 nm, pointing to a reduction of the problematic redox shuttle effect in lithium-sulfur batteries. Coulombic efficiency of up to 97% has been shown, with minor reduction in capacity compared to reference cells using commercial separators, namely ∼800 mAh g −1 . Lithium polysulfide diffusion is significantly hindered for etched ion-track membranes with smaller channels and indicates that the application of such membranes could undermine the polysulfide shuttle redox effect, improving lithium-sulfur battery performance. Separators with optimized porosities, i.e. smaller pore diameters and higher pore densities are expected to further improve the Li-S battery performance.