Fabrication and characterizations of the functional Ti0.87O2/PP separator
Negatively charged Ti0.87O2 nanosheets in the form of Ti0.87□0.13O2, where □ represents the Ti vacancies, were prepared by soft chemical exfoliation of layered lepidocrocite-type titanate crystals54,55. As illustrated in Fig. 2a, the nanosheet is a single-crystal-like 2D ultrathin monolayer (0.75 nm thickness) with a high density of Ti vacancies56. The size of the single-Ti atomic vacancy is ~ 0.2 nm × 0.2 nm56,57, which is larger than a Li+ ion (0.9 Å diameter) or Na+ ion (1.2 Å diameter) but smaller than a PS anion. Therefore, the Ti vacancies may work as migration-aids for Li+/Na+ ions and obstacle channels for PS anions, respectively. Atomic force microscopy (AFM) analysis (Fig. 2b) confirmed that the exfoliated Ti0.87O2 nanosheets are unilamellar sheets with a uniform thickness of approximately 1.1 nm. Transmission electron microscopy (TEM) images as shown in Fig. 2c display a flat and transparent sheet-like morphology, which is consistent with the AFM observation. Selected area electron diffraction (SAED) (inset in Fig. 2c) indicates the mono-crystalline nature of the Ti0.87O2 nanosheets. Figure 2d shows an atomic-resolution high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of a Ti0.87O2 nanosheet, where the Ti vacancies can be clearly visualized58,59. The Ti vacancies endow the obtained nanosheets negative charges, which has been confirmed by zeta-potential measurements (Figure S1). X-ray absorption fine spectroscopy (XAFS) was conducted to further investigate the structural characteristics of the defect-containing nanosheets. Figure 2e shows the Ti K-edge X-ray absorption near-edge structure (XANES) spectra of commercial rutile TiO2 and Ti0.87O2 nanosheets. The pre-edge peak at ~ 4981 eV represents transitions of core electrons into O 2p states that are hybridized with the empty Ti 4p state60,61. The intensity of this peak for the Ti0.87O2 nanosheets was obviously increased compared to that of rutile TiO2, elucidating a decreased electron number of the O 2p–Ti 4p hybrid orbitals. This result indicates the presence of lattice O atoms with unsaturated coordination, which should be attributed to the presence of nearby Ti vacancies62. Moreover, the Ti K-edge extended XAFS (EXAFS) k3x(k) oscillation curve of the Ti0.87O2 nanosheets exhibited a slight intensity decrease compared to rutile TiO2 (Fig. 2f), which also confirms the presence of Ti vacancies61,62. The interatomic distances of rutile TiO2 and Ti0.87O2 nanosheets were determined through Fourier transformed Ti K-edge EXAFS data (Fig. 2g). The first major coordination peak corresponds to the nearest Ti-O bond in the first coordination shell. The peak intensity for the Ti0.87O2 nanosheets obviously decreased relative to the TiO2 samples, which further verifies the presence of Ti vacancies61,62.
Functional Ti0.87O2/PP separators were fabricated by a facile vacuum filtration method (Figure S2). Compared to the porous surface of PP separators (Figure S3), Ti0.87O2/PP separators showed a homogeneous morphology with uniform surface coverage by Ti0.87O2 nanosheets (Fig. 2h). As shown in Figure S4, a broad (010) diffraction peak of lamellar Ti0.87O2 was observed, which is the same as the nanosheets being self-assembled onto commercial PP separators. The weight density and thickness of the Ti0.87O2 functional layers in the resultant Ti0.87O2/PP separator can be conveniently controlled by directly adjusting the volume of the nanosheet suspensions used in the vacuum filtration process. Figure S5 shows X-ray diffraction (XRD) data of Ti0.87O2/PP separators with different weight densities. As the weight density increased, the XRD peaks became more intense, with increasing thickness of the Ti0.87O2 layer. A cross-sectional scanning electron microscopy (SEM) image (Fig. 2i) shows a thickness of approximately 80 nm. Cross-sectional TEM images displayed parallel lamellar fringes (Fig. 2j), further revealing the layer-by-layer assembly of the Ti0.87O2 nanosheets. The fringe spacing was measured to be ~ 1.1 nm, which is consistent with the basal spacing in the XRD pattern (Figure S4). It should be noted that the weight and thickness of the Ti0.87O2 layer was only approximately 0.32% and 1.5% of those of the commercial PP separator (thickness, 25 µm; weight, 2.16 mg; diameter, 16 mm), respectively. Such a low weight density and ultrathin thickness have not been reported previously, to the best of our knowledge (Table S1). The as-prepared atomically-thin Ti0.87O2 layers can permit significantly high cation fluxes, which results in fast Li/Na-ion diffusion. The morphology and cross-sectional characteristics of Ti0.87O2/PP separators with relatively high weight densities of 0.032 and 0.096 mg cm− 2 were also investigated (Figures S6-S9). Homogenously stacked layers with thicknesses of approximately 150 and 460 nm were obtained, respectively.
For comparison, anatase TiO2/PP and GO/PP separators with the same weight density of 0.016 mg cm− 2 were fabricated (Figures S2 and S4). As shown in Figure S10, the anatase TiO2 nanoparticles did not disperse uniformly when coated onto PP surfaces. Only a limited part of the PP surfaces was covered by the aggregates of anatase TiO2 nanoparticles. This is clearly different from the charged nanosheets in suspensions where aggregation has been prevented due to Coulombic repulsion (neutral nanoparticles with high surface energy are prone to aggregate). As another negatively charged nanosheet material, GO was able to uniformly coat on the surface of PP separators (Figure S11). A slightly larger thickness of approximately 120 nm was observed for the GO/PP separator (Figure S12), compared with the Ti0.87O2/PP separators. This matches the calculated result based on an ideal 2D weight density in the lateral dimensions of GO and Ti0.87O2 monolayers (Figure S13).
The coating of atomically-thin Ti0.87O2 layer brings several advantages to improve the electrochemical performance of a separator membrane. As shown in Figure S14, after being placed on a hot plate and heated at 120°C for 10 min, the Ti0.87O2/PP film retained its original geometrical shape, while the pristine PP film tended to shrink. The improved thermal stability of the Ti0.87O2/PP would enhance the safety of batteries in practical applications. Figure S15 shows the contact angles of electrolyte on the PP and Ti0.87O2/PP separators. A smaller contact angle was observed on the Ti0.87O2/PP separators than that on PP separators, suggesting a better wettability of the Ti0.87O2/PP separators by electrolyte. Besides this effect, the as-prepared Ti0.87O2/PP separators showed high stability under various degrees of mechanical bending (Figure S16). As shown in Figures S17-S19 and Fig. 2k, the Li+ ion conductivity (see the Experimental details in the Supplementary Information) of the Ti0.87O2/PP separators (3.81 × 10− 1 mS cm− 1) was higher than that of the bare PP (3.05 × 10− 1 mS cm− 1) and anatase TiO2/PP separators (2.46 × 10− 1 mS cm− 1), and over three times higher than GO/PP separators (1.17 × 10− 1 mS cm− 1). The Li-ion transference numbers were also determined, as shown in Figure S20 (see the Experimental details in the Supplementary Information). Compared to other functional layers, the Li+ ion transference number increased significantly from 0.36 for bare PP to 0.55 for Ti0.87O2/PP with a weight density of 0.016 mg cm− 2 (Fig. 2k and Figure S21). Generally, covering open pores of pristine separators will increase the tortuosity of ion movement, leading to reduced Li-ion diffusion. However, the above testing results demonstrated that Ti0.87O2 nanosheets can facilitate Li-ion migration. Because the Ti0.87O2 layers are negatively charged with cation vacancies, the electrostatic attraction force between Ti0.87O2 nanosheets and Li+-cations facilitates the migration of Li ions towards the membrane with subsequent diffusion through the membrane. Similar phenomenon was also observed in other recent papers63,64, in which Li+ ions were observed to pass through the open channels of TaO3 nanosheets with a mesh structure. The Ti vacancies further provide an expressway for rapid transportation of Li+ ions in addition to the conventional interlayer galleries between the Ti0.87O2 sheets64. Besides these effects, the atomically-thin size scale of the Ti0.87O2 layers is also favorable for fast Li-ion diffusion.
Dendrite-free alkali metal anodes
Benefiting from the merits mentioned above, Ti0.87O2/PP separators are promising for regulating alkali metal ion flux in electrolyte and facilitating homogenous alkali metal deposition. Asymmetric Li||Cu half cells with various separators were fabricated to evaluate the cycling performance of Li metal anodes during repeated deposition and stripping. As shown in Fig. 3a, the cell with the Ti0.87O2/PP separator exhibited a steady Coulombic efficiency above 96.5% with stable plating/stripping voltage profiles for more than 100 cycles (Fig. 3b). In contrast, the cells with the bare PP (Fig. 3c), anatase TiO2/PP (Figure S22) and GO/PP separators (Figure S23) displayed a gradually increased voltage hysteresis and severely fluctuating Coulombic efficiency, which can be ascribed to the non-uniform Li deposition, and the formation of mossy or dendritic Li on the surface of Li metal anodes. Symmetric Li||Li cells were assembled to further investigate the superiority of Ti0.87O2/PP separators for stabilizing Li metal anodes. As shown in Fig. 3d, the cell with the Ti0.87O2/PP separator delivered an extended cyclability with stable voltage plateaus (Figs. 3e-3g) for over 300 h at a current density of 2 mA cm− 2 with an area capacity of 1 mAh cm− 2. In sharp contrast, the cell with the PP separator exhibited a gradual increase in voltage hysteresis (Figs. 3d-3f). A similar phenomenon was found for regulating Na deposition and suppressing Na dendrite growth using Ti0.87O2/PP separators. Figure S24 shows the Coulombic efficiencies of asymmetric Na||Cu cells with PP and Ti0.87O2/PP separators. The corresponding voltage profiles of Na plating/stripping in Na||Cu half cells with PP and Ti0.87O2/PP separators are shown in Figures S25 and S26, respectively. The average Coulombic efficiency of the cell with the Ti0.87O2/PP separator is about 98.8% for 200 cycles. In contrast, the Coulombic efficiency of the cells with the bare PP decreased below 91% in 150 cycles.
The morphology of the cycled Li anodes in symmetric cells was investigated to clarify the effect of the Ti0.87O2 nanosheets on the suppression of Li dendrite formation. As shown in Figures S27 and S28, the loosely-stacked mossy Li with a highly porous structure has been observed on the Li anodes from cells with bare PP separators. In contrast, when a Ti0.87O2/PP separator was used, the surfaces of the Li metal anodes were still compact without obvious mossy Li (Figures S29 and S30). This result demonstrates that the Ti0.87O2 nanosheets can facilitate homogeneous Li+ ion flux, giving rise to uniform Li deposition. Additionally, AFM Young’s modulus mappings revealed that Ti0.87O2/PP separators exhibited a modest modulus of around 60 MPa, (Figure S31, see the Experimental details in the Supplementary Information), meeting the requirement for suppressing the growth of Li dendrites65.
Theoretical calculations were conducted to investigate the diffusion properties of Li+ ions through anatase TiO2 (Fig. 4a), lepidocrocite-type TiO2 without Ti vacancies (Fig. 4b) and Ti-defect-containing Ti0.87O2 (Fig. 4c). Figure 4d shows the transfer profiles of single Li+ ions passing through these layers. For anatase TiO2 and lepidocrocite TiO2, potential energy barriers are as high as 4.83 and 7.06 eV, respectively. This indicates that it would be challenging for a Li+ ion to diffuse through them. After introducing a Ti vacancy, the energy barrier of the Ti0.87O2 monolayer radically decreased to 0.75 eV, which is comparable to, or even lower than, that of defective graphene66. Besides these lattice averages, the electronic structure of Ti0.87O2 might also induce lowered energy barriers. The charge density distribution on a Ti0.87O2 lattice with a single Ti cationic defect is shown in Fig. 4e. It can be seen that the charge density around the Ti vacancy can significantly increase the charge attraction for a Li+ ion, reducing the electrostatic charge overlapping, and weakening any Coulombic repulsion between a Li+ ion and the Ti0.87O2 lattice, thus resulting in a lower diffusion barrier for Li+ ions. To further visualize the effects of defective nanosheets on the Li-ion transportation process, two kinds of thin-layer models were constructed by restacking the conventional nanosheets and defective nanosheets, respectively (Figure S32). In the case of the restacked thin layer of conventional nanosheets, the gaps between the adjacent nanosheets were the only pathways for Li+ ion transport. Thus, a non-uniform distribution of Li+ ions was formed (Fig. 4f). In contrast, in the restacked thin layer of cation-defect nanosheets, the Li+ ions could be uniformly redistributed. This can be explained by the fact that Li+ ions migrate through not only the gaps between layers but also the defects within individual layers, resulting in a uniform distribution of Li-ion flux (Fig. 4g). Although the above idealized models cannot fully reflect all the aspects of real circumstances (especially once electrolyte interactions are introduced into the scenarios), the theoretical calculation and simulation results demonstrate the superiority of the cation-defect nanosheets for facilitating Li+ ion transport.
We propose a possible mechanism on the formation of dendrite-free alkali metal (Li/Na) anodes by using the negatively charged Ti0.87O2 nanosheets with atomic Ti vacancies (Figure S33). Upon discharging, solvated Li+ ions in liquid electrolyte diffuse to the anode side. The negatively charged Ti0.87O2 nanosheets could attract numbers of Li+ ions and facilitate the de-solvation process of the solvated Li+ ions before deposition, leading to a small energy barrier for deposition67,68. Then, the desolvated Li+ ions diffuse through the Ti atomic vacancies. Given the homogenized Ti atomic vacancies of the as-prepared Ti0.87O2 nanosheets, a uniform Li+ flux has been achieved. Consequently, a dendrite-free Li anode with a smooth deposition is formed. However, in the absence of Ti0.87O2 layers, a large energy barrier is needed during the de-solvation process67,68. The distribution and transport of Li+ ions are inhomogeneous and then form Li tips. Subsequently, Li+ ions tend to accumulate at preferentially formed Li tips, resulting in severe dendrite growth (Figure S34).
Elimination of polysulfide/polyselenide (PS) shuttling
In addition to ion re-distribution for a uniform alkali metal deposition, negatively charged Ti0.87O2 can also act as a protective barrier to inhibit the shuttle effect of PS anions. Taking polysulfides as an example, permeation measurements were conducted to evaluate the permeation resistance of Ti0.87O2/PP separators for minimizing the diffusion of PS anions (see the Experimental details in the Supplementary Information). The diffusion of Li2S6 was observed when use PP (Fig. 5b) and anatase TiO2/PP separators (Figure S35a) within 1 h. In contrast, the GO/PP and Ti0.87O2/PP separators were able to suppress the diffusion of PSs. However, as time elapsed, PSs were still able to pass through the GO/PP separators (Figure S35b). Only the Ti0.87O2/PP separator demonstrated a stable blocking effect towards PSs, lasting up to 10 h (Fig. 5a). Both GO and the Ti0.87O2 nanosheets are negatively charged and thus could suppress the shuttling of the negatively charged PS anions via electrostatic repulsion. The different capabilities of GO and Ti0.87O2 for preventing the shuttling of PS anions should ascribed to their negative charge densities. Based on theoretical calculations (Figure S36), Ti0.87O2 nanosheets have a negative charge density of 1.46 C m− 2, which is over 20 times higher than that of GO (0.064 C m− 2)69. Therefore, the Ti0.87O2 nanosheets with a much higher negative charge density can more effectively inhibit PS shuttling than GO layers. DFT calculations were performed to further elucidate the electrostatic repulsion between PS anions and Ti0.87O2 nanosheets (Figs. 5c-5f). Similar calculation methods were also conducted on anatase TiO2 and GO sheets (Figures S37 and S38). As shown in Fig. 5g, the Ti0.87O2 displayed much higher repulsion energies than anatase TiO2 or GO for all PS species.
Ex situ Raman spectroscopy was measured to gain further insights into the suppression of PS shuttling by the Ti0.87O2 nanosheets. Li-S coin cells were disassembled at a given voltage during the charge/discharge processes. We characterized the surfaces of the separators which had been in contact with lithium anodes. Figures 5h and 5i show the Raman spectra of the PP and Ti0.87O2/PP separators, respectively, retrieved from Li-S batteries. For the PP separator (Fig. 5h), three characteristic Raman peaks of S82− (at ~ 150, 220, and 470 cm− 1) were observed in the initial stage of the discharge process, associated with the formation of long-chain PSs. The Raman peaks of S82− gradually decreased as the discharge reaction proceeded. Meanwhile, Raman peaks at ~ 260 and 415 cm− 1 emerged, which correspond to the short-chain PSs of S42− and S52−. At the end of the discharge process, strong characteristic peaks of S42− and S52− were observed. This clearly indicates that the PSs shuttled through the PP separator from the cathode side and then deposited on the PP separator facing the anode side. Similarly, during the charging process, strong Raman signals of various PSs were observed. In contrast, for the Ti0.87O2/PP separator (Fig. 5i), almost no Raman signals of PS species were detected throughout the entire discharge and charge processes, indicating effective inhibition of PS shuttling. To observe the inhibition of PS shuttling and stabilization of Li metal anodes, the cycled cells were disassembled and the sides of Li metal anodes facing the separators were checked. As shown in Figure S39a, yellow polysulfides were observed on the Li anodes in the cells with PP separators. For the cell with Ti0.87O2/PP separators, almost no yellow species were observed and the cycled Li metal still exhibited a bright metallic lustre (Figure S39b). All these results confirmed that the PS shuttling and the growth of Li dendrites has been successfully eliminated when using Ti0.87O2/PP separators. A molecular dynamic simulation further confirmed the inhibition of the PS shuttling and regulation of Li ion transport through the Ti vacancies (Movie S1).
The electrochemical performances of Ti0.87O2/PP separators in Li-S batteries were tested using a carbon black/S composite cathode. Typical cyclic voltammogram (CV) curves of a Li-S cell with a Ti0.87O2/PP separator showed distinct reduction/oxidation peaks, which correspond to the conversion reactions of sulfur cathodes (Figure S40). Li-S cells with different separators were charged and discharged at 0.2C (1C = 1,673 mA g− 1). The voltage plateaus of the Li-S cell with a Ti0.87O2/PP separator (Figure S41) were consistent with its CV measurement. The initial discharge capacity was measured to be 960 mAh g− 1, followed by a moderate drop to 750 mAh g− 1 by the end of the 500th cycle (Fig. 6a). In contrast, a cell with a PP separator displayed an initial capacity of 980 mAh g− 1 (Figure S42) and rapidly decreased to 345 mAh g− 1 after 500 cycles. For the cells with the anatase TiO2/PP (Figure S43) and GO/PP (Figure S44) separators, lower specific capacities of 450 and 580 mAh g− 1 were obtained by the end of the 500th cycles, respectively. Figure S45 shows the rate performance of Li-S cells with Ti0.87O2/PP and PP separators at different current rates from 0.2C to 2C. The charge-discharge profiles of cells with the Ti0.87O2/PP separators showed distinguishable voltage plateaus at each current density (Figure S46). High specific capacities of 960 and 560 mAh g− 1 were achieved at 0.2C and 2C, respectively. However, the cells with PP separators suffered from dramatic capacity decay. Although at 0.2C, the capacity reached up to 950 mAh g− 1, as the current rate was increased to 2C, the capacity dramatically decreased to 260 mAh g− 1. A long-term cycle test was conducted at a 1C rate for over 5000 cycles to verify the functioning of the Ti0.87O2 layer (Fig. 6b). A specific capacity of 585 mAh g− 1 was maintained at the end of the 5000th cycle, corresponding to an ultralow capacity decay of 0.0036% per cycle. SEM images (Figure S47) showed that Ti0.87O2 nanosheets were still maintained on the separator after such long-term cycling. To the best of our knowledge, this is the best cycling stability among reported functionalized separators for Li-S batteries (Fig. 6c and Table S1), including GO28, graphene29, [email protected]31, [email protected]32, CNT/NCQD33, MgAl-LDH41, NiFe-LDH/N-graphene40, MoS2-PDDA/PAA37, Sb2Se3 − x/rGO22, Ti3C239, Cu2(CuTCPP)44, CNT/ZIF-843, Ce-MOF/CNT45, BC/2D MOF-Co46, and Laponite nanosheets18.
To explore the potential for practical applications, thick cathodes with a sulfur loading of 3.5 mg cm− 2 were assembled and investigated. Figure 6d shows the long-term cycling performance of a Li-S cell with a Ti0.87O2/PP separator. After an initial activation at 0.2C, the cell delivered a specific capacity of 565 mAh g− 1 at 1C up to 5000 cycles. Even at a high current density of 2C, this cell still delivered a reversible specific capacity of 250 mAh g− 1 after 10000 cycles, corresponding to a capacity decay as low as 0.0035% per cycle. It should be noted that, to highlight the function of Ti0.87O2 nanosheets, the cathode matrix of carbon black has almost no PS adsorption ability. The use of porous carbon with hierarchical nanostructures as sulfur cathodes could further increase the sulfur mass loading for higher energy densities. For example, we used commercial carbon nanotubes (CNT) as the sulfur host (Figure S48). The Li-S batteries achieved high capacities and high area energy densities (Figures S49 and S50). Flexible Li-S pouch cells were assembled using Ti0.87O2/PP separators. During charging and discharging at different bending angles, the pouch cells exhibited stable cycling performance at a current density of 0.2C up to 120 cycles (Fig. 6e).
The applications of Ti0.87O2/PP separators was also extended for Li-Se batteries. Figures S51 and S52 show the typical charge/discharge profiles of Li-Se batteries with PP and Ti0.87O2/PP separators, respectively. A gradually increased voltage polarization was observed for the Li-Se batteries with PP separators during the initial several cycles, accompanied by an obvious capacity decay. However, the overlapped charge/discharge curves confirmed the cycling stability of the Li-Se batteries with Ti0.87O2/PP separators. After continuous cycling at 0.2C for over 500 cycles, a specific capacity of 460 mAh g− 1 was still retained (Figure S53). The Ti0.87O2/PP separator is also promising to improve the cycling stability for Na-Se batteries. As shown in Figures S54 and S55, highly overlapped charge/discharge curves were observed for Na-Se batteries with Ti0.87O2/PP separators, suggesting superior cycling performance compared to the cells with bare PP separators. Upon continuous cycling at 0.2C, a specific capacity of around 450 mA h g− 1 was achieved after 250 cycles (Figure S56).