A thin multifunctional coating on a separator improves the cyclability and safety of lithium sulfur batteries† †Electronic supplementary information (ESI) available: Detailed description of the experimental procedures and calculations. See DOI: 10.1039/c7sc01961k

The separator has an electrocatalytic effect for polysulfide transformation, and can confine the polysulfides within the cathode and block the dendritic lithium in the anode.


Introduction
Lithium-sulfur (Li-S) batteries are attractive due to their low cost and high gravimetric energy density, 1-3 but suffer from low cyclability and poor safety in energy-density-optimized full cells due to (a) soluble polysulde shuttling from the cathode to anode, and (b) lithium metal anode corrosion and shorting. 4,5 In order to defeat (a), there are two strategies (Scheme 1a): (a1) adsorption + electrocatalysis, and (a2) complete sealing by a solid electrolyte. In (a1), the sulfur cathode is mixed with electrocatalyst nanoparticles that compete with the liquid electrolyte for free polysuldes. The electrocatalysts (such as graphene oxide 6,7 and TiO 2 [8][9][10][11] ) also facilitate the redox reactions of the surface-adsorbed polysuldes. The (a1) route reduces the concentrations and lifetimes of soluble polysuldes, thereby reducingbut not eliminatingsulfur transport to the lithium anode. In the (a2) strategy, one aims to eliminate sulfur crossover completely by sealing off the cathode chamber using a solid electrolyte that conducts Li + but not sulfur. 12 This is possible because diffusion mechanisms are fundamentally different in solids ("hopping/exchange") than those in liquids ("vehicular" Stokes-Einstein transport of all soluble species). A solid barrier formed in situ with no percolating pores could stop sulfur transport completely while still allowing bidirectional Li + transport, forming an enclosed catholyte chamber on the sulfur side. Moreover, a negatively charged Debye layer near the separator surface could reject the polysulde anions to localize their transport on the cathode side 13 and the functional interlayers could also trap lithium polysuldes. 14,15 In order to defeat (b), in particular electrical shorting by dendritic penetration of the separator, 16 a deformable solid electrolyte separator is also envisioned, which blocks lithium dendrite growth more effectively than traditional nanoporous polypropylene (PP) separators at large current densities.
While lithium metal growth always chases the ionic current and thus the $30 nm pores in the PP separator, there can be a thermodynamic cost when the pores get very small. Depending on the over-potential applied to the anode, there is a smallest radius-of-curvature for depositing lithium metal allowed by thermodynamics due to capillary forces. 17 While lithium metal dendrite is able to plate through the 30 nm pores of the PP separator, it may nd it difficult to plate through the much smaller pores of our coating. In situ transmission electron microscopy (TEM) observations of mossy lithium growth reveal that even under a very large overpotential, 18 the smallest mossy lithium tendrils have radii of tens of nanometers, so the ne pores of our ceramic coating could present signicant resistance to lithium metal dendrite growth (Scheme 1b), while still allowing bidirectional Li + transport. When this coating is applied without carbon black (e.g. to the opposite side of the PP separator), it could add a signicant safety factor against electrical shorting. [19][20][21] In this work, we propose a multifunctional coating less than 10 mm thick, easily applied onto traditional nanoporous PP separators, which addresses (a1), (a2) and (b) simultaneously. This porous coating consists of titanium dioxide nanoparticles and carbon black, forming an excellent adsorbent, and is electronically conductive and electrocatalytically active, veried by shied redox voltage peaks and theoretical calculations (a1). Later, this thin porous coating gets fouled by solid sulfurcontaining compounds, forming an in situ solid electrolyte layer that stops sulfur transport while still allowing bidirectional Li + transport. Ideally, if the sulfur-containing solid electrolyte formed in situ closes all percolating pores, we will have (a2). To demonstrate this, we applied the fouled coated separator (aer cycling) in a H-cell, and showed it can separate the polysulde containing le side (dark color) with the clear side, demonstrating that it can be used to form a catholyte chamber. Lastly, for (b), we showed that the sub-10 mm thick coating on 30 mm thick PP can delay dendrite penetration for a 15Â longer time duration, at an extremely large current density of 100 mA cm À2 , using a capillary tube cell setup for visualization. These ndings prove that such a thin nano oxide coating is multifunctional in enhancing the cycle life and safety of Li-S batteries.

Results and discussion
Lithium polysuldes dissolved in the organic electrolyte can easily diffuse through the polypropylene separator and react with the lithium metal anode (Fig. 1a), [22][23][24] resulting in poor cycling performance. 19,20 Therefore, preventing polysulde migration by separator modication is a promising strategy to increase the electrochemical performance of Li-S batteries. [25][26][27] Herein, we prepared titanium dioxide nanoparticles with high Li + conductivity and large specic surface area as a coating on separators (Fig. 1b). The size of the titanium dioxide nanoparticles is $3 nm ( Fig. 1c and S1 †). The amorphous and nanocrystalline phases of the titanium dioxide nanoparticles are shown in area 1 and area 2 (Fig. 1d), respectively. The amorphous phase of TiO 2 has high lithium ion mobility, which is benecial for lithium ion diffusion. 28 The lattice fringes with a distance of 0.19 nm correspond to the (200) plane of anatase TiO 2 ( Fig. S2 and S3 †). Moreover, the titanium dioxide nanoparticles show a relatively high specic surface area of 313 m 2 g À1 with a major pore size of 2.5 nm (Fig. S4 †), and thus have a large contact area with soluble lithium polysuldes.
The pristine polypropylene separator shows abundant pores on the surface and has a thickness of $30 mm (  coating with either solely the super C65 or the mixture of titanium dioxide nanoparticles and super C65, the surfaces of these separators become denser ( Fig. 2a and c). The thickness of these coating layers is $7.5 mm ( Fig. 2b and d), and the loadings of the coating material are 0.4 mg cm À2 for the C65 separator and 0.7 mg cm À2 for the TiO-C65 separator. The coated separator retains its mechanical exibility, which is important for battery fabrication (Movie S1 †). Elemental mapping of the super C65 separator shows that carbon is mainly dispersed on the surface (Fig. S6 †). Titanium, oxygen, and carbon are uniformly dispersed on the surface of the TiO-C65 separator ( Fig. 2e-h).
To characterize the distribution of sulfur species on the cycled TiO-C65 separator, we performed local probe mechanical tests. In Fig. 2i-l (ESI Movie S2 †), we pushed in a sharp tungsten probe from the top of the layer, and then moved it horizontally to execute a scratching test. The particle size on the separator turns out to be bigger compared to that before cycling ( Fig. 2c and d), indicating agglomeration bonded by the deposited sulde (a2 in Scheme 1a). Solid-like sulfur species were conrmed to be deposited on the TiO-C65 coating ( Fig. 2i and S12a †). Meanwhile, C65 mixed in the coating (which is in physical contact with the solid cathode) can act as an additional cathode current collector to reuse the lithium polysuldes. When the vertical force applied on the probe is small, we nd that there is no extra so lm formed on the surface, unlike in the case of using an acidized carbon nanotube paper on the separator. 12 Thus, the lithium polysuldes should be dispersed inside the $7.5 mm TiO-C65 coating.
From cyclic voltammetry curves, there are two reduction peaks of the S 8 cathode in Li-S batteries (Fig. 3a). [29][30][31] The rst peak at high voltage corresponds to the open ring reduction of sulfur to soluble lithium polysuldes (Li 2 S n , 4 # n # 8) and the second peak is attributed to the transformation of the lithium polysuldes to insoluble Li 2 S 2 /Li 2 S. 32 In our Li-S battery test with the C65 separator, the second reduction peak appears in the 3rd cycle (Fig. S7 †). However, the TiO-C65 separator Li-S batteries exhibit the second reduction peak aer the 1st cycle, and the peak position and shape remain stable from the 2nd cycle on. The cathodic peak positions of the TiO-C65 separator (2.254 and 1.925 V) are larger than the C65 separator (2.221 and 1.850 V), indicating faster redox reaction kinetics. This demonstrates that titanium dioxide nanoparticles have a strong electrocatalytic effect on sulfur reduction (Fig. 3a). There is a small anodic peak at 1.9 V in Li-S batteries with the TiO-C65 separator, corresponding to the lithiation of TiO 2 (Fig. S8 †). Moreover, the overpotential DU between the anodic peaks and cathodic peaks of the TiO-C65 separator (0.246 and 0.464 V) is smaller than that of the C65 separator (0.251 and 0.544 V), indicating the lower polarization of Li-S batteries with the TiO-C65 separator.
Li-S batteries with the TiO-C65 coated separator show a high specic capacity of 1601 mA h g À1 at a current density of 0.1C (1C h 1675 mA g À1 ) and a good rate performance at higher current densities (Fig. 3b). Moreover, the two plateaus in the discharged curve were still evident even up to 1C (Fig. S9 †). Typically, the low electronic conductivity of sulfur and the high solubility of lithium polysuldes are associated with the low utilization of sulfur and poor cycling performance of Li-S batteries. In fact, our test with the PP separator shows a low specic capacity of only 175 mA h g À1 at a current density of 0.5C (Fig. 3c). On the other hand, the conductive C65 coating on the separator connes the polysuldes within the cathode side, forming a catholyte chamber. Therefore, Li-S batteries with the C65 separator have a higher specic capacity of 550 mA h g À1 at the rst cycle and maintain a discharged capacity of 332 mA h g À1 aer 500 cycles at 0.5C. Li-S batteries with the TiO-C65 separator have a high initial capacity of 1206 mA h g À1 and a high maintained capacity of 501 mA h g À1 aer 500 cycles at 0.5C.
The titanium dioxide nanoparticles have a strong catalytic effect and chemical binding with lithium polysuldes, which not only has the potential to increase the utilization but also to improve the rate performance. For example, the charge transfer resistance is clearly reduced (see ESI Fig. S10 †). The Li-S batteries with the TiO-C65 coated separator also show a good electrochemical performance at a higher current density of 2C with 1047 mA h g À1 at the rst cycle and 533 mA h g À1 aer 300 cycles (Fig. 3d). While the sulfur loading is approximately 2 mg cm À2 , Li-S batteries with the TiO-C65 separator have an initial capacity of 840 mA h g À1 and a high sustained capacity of 556 mA h g À1 aer 90 cycles at 0.5C (see ESI Fig. S11 †). To conrm the ability of the TiO-C65 separator to trap lithium polysuldes, we disassembled the coin cell aer the test and performed SEM analysis and found the coexistance of sulfur and titanium (see ESI Fig. S12 †). This demonstrates that titanium dioxide nanoparticles can selectively adsorb the sulfur species as a solid-like fouling product.
The trapped sulfur species in the coating can still contribute to the capacity. We constructed a battery cell using the cycled TiO-C65 separator to conduct a cyclic voltammetry scan, and it showed distinct charge/discharge peaks for sulfur (see ESI   S13 †), indicating that the coating can act as a second current collector and reuse the lithium polysuldes trapped within. Thereaer, the fouled TiO-C65 coating (aer cycling) was applied in a separate H-shaped cell (see ESI Fig. S14 †), and showed it could separate the polysulde containing le side (dark color) from the clear side, demonstrating that the sulfurcontaining solid electrolyte formed in situ on the TiO-C65 separator could be used to form an isolated catholyte chamber.
To investigate the effects of TiO 2 on Li 2 S n transport and transformation, systematic rst-principles calculations were conducted for a Li 2 S x -graphite (here representing super C65)/ TiO 2 surface system. The optimized lowest-energy geometric structures of Li 2 S n (n ¼ 1, 2, 4, 6 or 8) are shown in Fig. S15, † which are consistent with other reported works. 33,34 The structures of Li 2 S and Li 2 S 2 are similar, with sulfur atoms bridging two lithium atoms. Li 2 S 6 and Li 2 S 8 show ring-like structures, which can be regarded as a lithium dimer inserted into the S 6 and S 8 rings. Li 2 S 4 has a cage-like structure and is the intermediate structure between the above two structural types.
The optimized geometrical models for Li 2 S n (n ¼ 1, 2, 4, 6, or 8) and S 8 adsorbed on the TiO 2 and graphite surfaces are displayed in Fig. 4 and S16, † and the corresponding binding energies are plotted in Fig. 4j. Obviously, the binding energies for Li 2 S n adsorbed on the TiO 2 surface are much larger than those of Li 2 S n on the graphite surface. The binding energy of S 8 on TiO 2 is 1.35 eV, nearly double that for S 8 on graphite (0.72 eV), indicating that TiO 2 can more effectively attract S 8 molecules. For the graphite and TiO 2 surfaces, the adsorption energies for Li 2 S n are all larger than that for S 8 , which may be related to the larger polarity of Li 2 S n compared with S 8 . The binding energies for Li 2 S n on graphite are around 1.2 eV except for that for Li 2 S 4 . As shown in Fig. S16, † for Li 2 S and Li 2 S 2 , both lithium atoms are closer to the graphite surface, indicating that the attraction between lithium atoms and graphite is larger than that between sulfur atoms and graphite. On the other hand, the lithium dimers are nearly vertical to the graphite surface for Li 2 S 6 and Li 2 S 8 . This may be because there is more contact area with graphite when the ring-like Li 2 S 6 and Li 2 S 8 are parallel to the graphite surface which maximizes the binding energies. For Li 2 S 4 , the dimer is vertical to the graphite surface. However, it does not have a ring-like structure to maximize the contact area, thus showing a lower binding energy (1.0 eV).
Different from graphite, the binding energies of Li 2 S n on TiO 2 dramatically increase with the decrease of n (n ¼ 1, 2, 4, 6, or 8) or increase of lithium fraction. As shown in Fig. 4a-f, the lithium atoms are bonded with two adjacent oxygen atoms at the TiO 2 surface for all of Li 2 S n (n ¼ 1, 2, 4, 6, or 8). While the distances between the nearest neighbor oxygen atoms are 3.79/ 3.80Å, the distances between the lithium atoms in Li 2 S n are 3.55, 3.31, 2.85, 2.75, and 2.77Å, for n ¼ 1, 2, 4, 6, and 8, respectively. Our calculation shows that the TiO 2 surface attracts Li 2 S n (n ¼ 1, 2, 4, 6, or 8) and S 8 molecules much more strongly than graphite. With its large surface area, the nano-size TiO 2 should efficiently adsorb Li 2 S n and reduce the shuttling, substantially enhancing the utilization of lithium and sulfur. This explains the expansion of the peak area in the cyclic voltammetry curves (Fig. 3a) and the dramatic improvement of the capacity at higher rates.
Additionally, we compared the abilities of S 8 , graphite and TiO 2 to attract lithium atoms. The calculated binding energies are 2.16, 1.68 and 4.45 eV, for S 8 , graphite, and TiO 2 , respectively (Fig. 4g, h and i). In the discharge process, lithium ions transport through the separator to react with sulfur atoms. These results give insights into the role of the coating materials on the separator. For example, if there is only carbon (graphite) and no TiO 2 in the separator, the lithium atoms prefer to adsorb on sulfur but not on graphite, because the binding energy of lithium on the former is larger than that on the latter. Therefore, the effect of carbon alone is expected to be small. Moreover, the obtained lowest-energy Li-S 8 adsorption conguration has the lithium atom located above the center of the S 8 molecule (Fig. 4g), and it requires breaking the S 8 ring (overcoming an energy barrier of about 1.5 eV) 35 to form the most stable Li 2 S 8 structure (Fig. S15e †). In contrast, the binding energy for the lithium atom on TiO 2 is about twice as large as that on S 8 , which allows TiO 2 to attract lithium atoms in addition to S 8 molecules (as discussed above). This leads to the aggregation of lithium atoms and S 8 molecules on the TiO 2 surface. Additionally, each lithium atom releases about 4 eV when binding with TiO 2 , which can help the nearby S 8 to overcome the barrier to open the ring and form the Li 2 S 8 structure. TiO 2 thus provides an effective electrocatalytic surface by stabilizing the reaction intermediates, and enhancing the rate of reaction between them. Furthermore, it is known that once the size of TiO 2 is smaller than 10 nm, it can become electrically conductive. 36 This electrocatalytic effect for the Li 2 S n transformations explains the peak shis in the cyclic voltammetry curves (Fig. 3a) and the reduction of the activation period (Fig. 3c).
With the above electrochemical tests and theoretical calculations, we conrm that our TiO-C65 coated membrane can trap lithium polysuldes to improve the performance of Li-S batteries. The nanoparticle coating may exhibit an additional benet, which is to prevent lithium dendrite penetration. To test the capability of the coated separator to block dendritic lithium penetration, we devised a capillary cell to visualize the electrodeposition process. The capillary cell is lled with the liquid electrolyte, where a piece of lithium metal electrode is stripped and the lithium ions are simultaneously electrodeposited onto an enamelled copper wire with the end wrapped by the separators. The detail of the experimental setup is shown in the ESI Fig. S17. † The diameter of the copper wire is 0.04 cm. The capillary tube batteries are discharged at a high current density of 100 mA cm À2 to promote the dendritic growth of lithium metal. Lithium dendrites begin to emerge from the polypropylene separator at 50 s (ESI Movie S3 †), while no lithium penetration was observed in the cell using our TiO-C65 separator even aer 850 s (ESI Movie S4 †). The results indicate that the TiO-C65 coating can act as a multifunctional barrier to prevent the lithium dendrites from penetrating the separator, as well as preventing the cross-over of lithium polysuldes.

Conclusions
In summary, we have developed a multi-functional titanium dioxide-super C65 modied separator for Li-S batteries that enables a high specic capacity, stable cycling performance at high rates, and improved safety. Li-S batteries with the TiO-C65 separator show a high initial capacity of 1206 mA h g À1 and maintain a high specic capacity of 501 mA h g À1 aer 500 cycles at 0.5C. The electrochemical results and theoretical simulation demonstrate that titanium dioxide nanoparticles have a strong catalytic effect and chemical binding with lithium polysuldes. Therefore, the effect of the TiO-C65 separator is assigned to (a1) surface segregation and catalysis, and also the partial effect of (a2) sealing by the solid electrolyte formed in situ. The results of this work indicate that thin coating materials with high conductivity and a large surface area on the separator can increase the utilization of lithium polysuldes, allow the fast diffusion of lithium ions, and decrease the migration of lithium polysuldes to the lithium metal anode. Additionally, our titanium dioxide nanoparticle-super C65 separator with a strong dendrite blocking ability can be used in applications beyond Li-S batteries such as lithium/sodium metal batteries, and contributes to the development of high-performance and safe energy storage devices.