Synthesis and characterization of sulfur hosts

Inspired by the advantages of hollow, polar, interweaving surface structure, and catalytic activity of Co for Na–S batteries, we synthesized hollow polar bipyramid prism cobalt chalcogenides and used them as sulfur hosts. The bipyramid prisms of cobalt were synthesized by simple reflux methods (see “Methods” section). The X-ray diffraction (XRD) pattern confirmed the formation of C₁₀H₁₆Co₃O₁₁ and was attributed to the standard PDF of JCPDS#22-0582, as shown in Supplementary Fig. 1.

In order to study the structure and morphology of the as-prepared precursor, field emission scanning electron microscopy (FESEM) (Fig. 2a, b) and transmission electron microscopy (TEM) (Fig. 2c and Supplementary Fig. 2) along with energy dispersive spectroscopy (EDS) elemental mapping was performed (Fig. 2d). The bipyramid Co prisms were converted into hollow bipyramid Co/C prisms. To confirm the phase purity and hollow interior of Co/C, XRD and TEM tests were performed (Supplementary Fig. 3 and Supplementary Fig. 4b), respectively. Afterward, the Co/C hollow prisms were converted into Co₃O₄/C (BPCO). The phase purity and morphology of BPCO were investigated by XRD and FESEM, as shown in Supplementary Fig. 5 and Fig. 2e, f, respectively. FESEM images of BPCO show that it has a unique hierarchical surface with interweaving plate-like structures. To confirm the elemental composition, EDS mapping was performed, as shown in Supplementary Fig. 6. Furthermore, BPCO was converted into CoS₂/C (BPCS), CoSe₂/C (BPCSE) and CoTe₂/C (BPCTE) by an ion-exchange method, their formation and phase purity was confirmed by XRD analysis as represented in Supplementary Figs. 7–9 and their chemical compositions was confirmed by EDS elemental mapping as shown in Supplementary Figs. 10–12, respectively. Figure 2g shows a FESEM image of BPCS representing a distinctive morphology with a hierarchical surface.

Furthermore, the chemical composition and morphology of BPCS was confirmed by TEM; Fig. 2h shows that the bipyramid hollow prisms of BPCS have 376 nm hollow wide space to accommodate sulfur and polysulfides within it. In addition, HRTEM and single-crystal spot transmission electron diffraction patterns confirm the synthesis of BPCS, as shown in Fig. 2i–k. The Brunauer–Emmett–Teller (BET) surface area of BPCS is 119.9 m² g⁻¹ and the hysteresis loop is of type H4 indicates the physisorption isotherm of type I, which conforms to IUPAC, demonstrating that BPCS is composed of uniform slit-like pores, representing microporous structure, as shown in Supplementary Fig. 13. Finally, sulfur was injected into the hollow prisms using a melt-diffusion method to achieve the S@BPCS composite, as shown in Fig. 3b–d. Figure 3c shows that sulfur efficiently steamed into the inner side of the prism wall and there is no sulfur agglomeration on the hierarchical surface of the BPCS, indicating sulfur removal from the surface at 190 °C. The XRD pattern of the S@BPCS composite confirms the cubic sulfur loading in the hollow bipyramid with interweaving plates like surfaced BPCS (Fig. 3e). An EDX line scan was recorded to examine the position of sulfur; Fig. 3g clearly shows sulfur efficaciously loaded inside the hollow prism. Furthermore, a thermogravimetric analysis (TGA) was performed to measure the content of loaded sulfur in the hollow prism S@BPCS composite. The mass of loaded sulfur was calculated using Fig. 3f and Supplementary Fig. 14, and determined to be as high as 64.5%.

Electrochemical performance of Na–S battery

To investigate chemical interactions between BPCS and NaPSs, ex situ XPS analysis was employed, as shown in Fig. 3a. To understand the chemical interactions between the host BPCS and NaPSs we made three different samples for ex situ XPS as pure BPCS and two discharged cell samples at 2.0 and 0.8 V. The binding energies of cobalt decreases as electrons transfer from NaPSs to BPCS when discharged at 2.0 and 0.8 V, which is attributed to the strong chemical interaction between BPCS and NaPSs during the discharge process. This type of interaction between BPCS and NaPSs suppresses the shuttle effect, increases the cycling stability, and improves the capacity.

To examine the electrochemical storage properties and NaPSs redox reactions of polar catalytic BPCS and other metal chalcogenides, cyclic voltammetry (CV) and galvanostatic charge/discharge (GCD) plots were recorded as shown in Supplementary Fig. 15. Impressively, the CV plot of S@BPCS exhibits comparatively higher current density and peak potential because of its high electrical conductivity as well as catalytic and high polar nature that results in enhanced electrical conductivity and strong chemical interaction between BPCS and NaPSs. The CV curves of the S@BPCS composite exhibit two reduction peaks. The peak at 2.12 V corresponds to the reduction of sulfur to long-chain Na-polysulfides (Na₂S_X, X ≥ 4) and the other reduction peak at a lower potential, 1.6 V, corresponds to the further reduction of long-chain polysulfides to short-chain Na-polysulfides (Na₂S_X, X ≤ 4). The anodic scan also exhibits two oxidation peaks, at 2.05 and 2.6 V, which are attributed to the oxidation of short-chain polysulfides to long-chain polysulfides and sulfur¹⁹. The redox processes indicate high reversibility of the cell reactions.

The GCD of the S@BPCS composite was also evaluated. Higher capacity and longer potential plateaus suggest that the BPCS sulfur host can assist as an electrocatalyst to accelerate the cathode redox reactions and enhance the utilization of active sulfur. Figure 4a shows the 1st, 2nd, and 50th charge–discharge profiles of the S@BPCS composite electrode with 4.4 mg cm⁻² mass loading and 29 μm thickness (Supplementary Fig. 16). The 1st reversible discharge capacity is 1347 mAh g⁻¹ at a high current density of 0.5 C and the profiles clearly show long plateaus. The plateaus at higher potentials are due to the formation of the solid–liquid state of sulfur and NaPSs, while, the plateaus at lower potential are due to the reduction of long-chain NaPSs (Na₂S₈, Na₂S₆, and Na₂S₅) to short-chain polysulfides (Na₂S₃, Na₂S₂, and Na₂S). The cycling performance of the S@BPCS composite shows excellent stability (Fig. 4c), the specific capacity of 2nd discharge was 755 mAh g⁻¹ and after 350 cycles, it was 701 mAh g⁻¹ at a high current density of 0.5 C with 98.5% Coulombic efficiency. The coulombic efficiency of all the metal chalcogenide@S electrodes are given in Fig. 4f. The capacity decay per cycle is 0.0126 %, which is significantly lower than anything previously reported in the literature (Supplementary Fig. 17). It is worth noting that pure BPCS has a negligible contribution to the capacity under the same charge-discharge conditions (Supplementary Fig. 18). The cycling performance of other metal chalcogenides is shown in Supplementary Fig. 19. In addition, the Celgard polymer separator was used to test the cycle performance of BPCS@S and KB@S, and compared with the glass fiber separator in the supporting information (Supplementary Fig. 20). When using the glass fiber separator instead of the Celgard polymer separator, diffusion of polysulfides is slowed because of its thickness and porous structure, resulting in good electrochemical performance.

The rate performance of the sulfur cathodes was also measured. Figure 4b shows the rate performance of S@BPCS at different current densities from 0.5 to 3 C. The cathode S@BPCS composite exhibits capacities as high as 755, 565, 475, 415, 382 and 349 mAh g⁻¹ at current densities of 0.5, 1, 1.5, 2, 2.5, and 3 C, respectively. Compared to other metal chalcogenides, the S@BPCS composite has a higher rate capability, as shown in Supplementary Fig. 21. Also, when the current density was switched back to the initial value, the S@BPCS composite exhibits a high reversible capacity of 727 mAh g⁻¹ as compared to other metal chalcogenides (486, 432 and 399 mAh g⁻¹ for S@BPCO, S@BPCSE, and S@BPCTE, respectively). To the best of our knowledge, the S@BPCS composite has remarkable rate capabilities as compared with state-of-the-art materials previously reported in the Na–S battery literature, as shown in Fig. 4h^{16,21,27,28,29,30,31,32,33,34,35,36,37,38,39}.

In order to take full advantage of the high theoretical capacity of the S cathode, a high mass loading is a key factor to scale up the practical application of the Na–S battery. However, high loading of insulating sulfur will facilitate the shuttling of NaPSs and the capacity decay will be more serious with a decrease in the utilization rate of active sulfur and a decrease in the capacity of active sulfur. Therefore, it is still a great challenge to optimize the performance of high sulfur loading in Na–S batteries. Previous studies have reported that only batteries with a capacity greater than 4 mAh cm⁻² can exceed commercial LIBs⁴⁰. In addition to this, considering that the voltage of the Na–S battery is lower than that of SIBs, a higher areal capacity is required for the former system. In this work, with the mass loading of 4.4 mg cm⁻², the capacity of the S@BPCS cathode was more than 4 mAh cm⁻². To verify the superiority of the S@BPCS cathode, batteries with high mass contents of 7.3 and 9.1 mg cm⁻² were cycled at 0.5 C (Fig. 4d, e). An initial areal capacity of 6.24 mAh cm⁻² was measured and this value stabilized to 5.5 mAh cm⁻² after 10 cycles when the mass content was 7.3 mg cm⁻². For sulfur content as high as 9.1 mg cm⁻², the initial areal capacity of the S@BPCS cathode was 8 mAh cm⁻², stabilizing at 6.6 mAh cm⁻² after 10 cycles. It is noteworthy that the current density is calculated to be as high as 9.7 mA cm⁻². Figure 4g summarizes the areal capacities of different sulfur-loaded batteries. The electrode masses of 7.3 and 9.1 mg cm⁻² used in this study are much higher than those of previously reported Na–S batteries (Supplementary Fig. 22) and commercial LIBs (4.0 mg cm⁻²). To illustrate the importance of this work, we compared the capacity and rate capabilities of the Na–S batteries in Fig. 4h.

Mechanisms of reversible reactions in the Na–S battery

To investigate the electrochemical reactions and corresponding mechanisms, in situ XRD, in situ Raman, ex situ HRTEM/SAED, DFT calculations, and ex situ XPS analysis were performed. The in situ Raman spectrum which represents a vibration band for sulfur at 475 cm⁻¹²⁵, at a potential of 2.8 V, which is supported by the in situ XRD pattern. In Fig. 5j, the XRD pattern has a diffraction peak at 23.08°, which is indexed to the (240) plane of sulfur (JCPDF# 08-0247). The intensity of the sulfur vibrational band was reduced and a new vibrational band appears at 451 cm⁻¹ at a potential of 2.25 V, which is attributed to long-chain Na-polysulfides (Na₂S₈). This result was supported by in situ XRD, where the diffraction peak appearing at 22.95° is due to the formation of Na-polysulfides by the reduction of sulfur. After the long-chain Na-polysulfides reduced to short-chain polysulfides on further discharge, the reduction mechanism was studied. Figure 5a, b shows HRTEM/SAED images, which confirm the reduction of sulfur to Na-polysulfides (Na₂S₅) at a potential of 2 V. This is consistent with the XRD diffraction peak at 12.198° of Na₂S₅ (JCPDF# 27-0792) corresponding to the (020) plane and the Raman vibrational bands at 474 and 451 cm⁻¹ attributed to Na₂S₅. On further discharge at 1.65 V, the long-chain polysulfides successfully reduced to short-chain polysulfides (Na₂S₄) and the corresponding

XRD pattern shows a diffraction peak at 19.93° indexed to the (112) plane that was strongly attributed to Na₂S₄ (JCPDF# 25-1112) and it was robustly confirmed with HRTEM/SAED (Fig. 5c, d) and a Raman vibrational band at 472 cm⁻¹²⁵. Furthermore, for discharge at 1.2 V, a new XRD diffraction peak was generated at 11.84° which is indexed to the (101) plane of Na₂S₂ (JCPDF#81-1764) and was subsequently confirmed with HRTEM/SAED (Fig. 5e, f) and a Raman vibrational band at 474 cm⁻¹ for Na₂S₂²⁵. Subsequently, when the cell was deeply discharged at 0.8 V, a new peak at 16.5° was generated which indexed to the (122) plane of JCPDF#47-0178 and is ascribed to Na₂S⁴¹. Figure 5i shows the CV curve labeled with polysulfide formed at different voltage during discharge process. Figure 5g, h showing HRTEM/SAED and a Raman vibrational band (Fig. 6b) at 473 cm⁻¹ which is attributed to Na₂S, supporting the XRD result. The first discharge process could, therefore, be labeled with corresponding polysulfides as described in Fig. 6a, and the entire reduction process can be written as follows:

When the cell was charged to 2.8 V, Na₂S and Na₂S₅ are not detected in the XRD pattern. This means that the conversion rate of short-chain polysulfides to long-chain polysulfides is fast, and that could be a reason why Na₂S and Na₂S₅ are not detected with in situ XRD. In addition, the intensity of the XRD peaks decreases, which indicates a fast transformation of S to long-chain Na-polysulfides and to Na₂S. This study elucidates the reduction reactions of S to Na₂S and presents a new mechanism to explore the significance of polar and catalytic BPCS. The polar and catalytic BPCS diminishes the dissolution of polysulfides, facilitates the fast transformation of long-chain polysulfides to Na₂S, alleviates the shuttle effect, and delivers excellent electrochemical performance. Furthermore, to check for dissolution of polysulfides in the electrolyte, the fiberglass disk that was used as a separator was removed from the cell after cycling for ex situ XPS of S. Figure 6c (c₁–c₃) shows high-resolution XPS spectra of S for the S@BPCTE, S@BPCSE, and S@BPCS batteries separators, respectively. The peaks in the range of binding energies of 160–165 eV are due to dissolved polysulfides²⁷ and the peaks in the range of 165–170 eV are due to the formation of SO_X. Figure 6c (c₃) shows that there are no Na-polysulfide peaks in S@BPCS separator and the rest have dissolved polysulfide peaks. These findings clearly demonstrate that the polar and catalytic BPCS can efficiently capture the Na-polysulfides.

Lastly, the visible difference in the adsorption of polysulfides in the BPCS was performed using a Na₂S₆ solution. Figure 6d clearly shows that when carbon was added to the Na₂S₆ solution, there was no change in the yellow color of polysulfides whereas, when BPCS was added to Na₂S₆ solution, the yellowish color turns colorless after few hours demonstrating the strong interaction and adsorption capability of polar BPCS towards NaPSs. The S@BPCS composite has internal void space of about 376 nm, which is enough to provide adequate space for NaPSs and tolerate the strain during sodiation and desodiation as shown in Fig. 6e.

Computational results

DFT calculations were performed with the SCAN + rvv10 functional to understand the interactions of NaPSs with CoX₂ cathode hosts. The lattice parameters of the structures optimized with SCAN + rvv10 are in excellent agreement with experimental results, as shown in Supplementary Fig. 23 and Supplementary Table 1. As can be seen from the density of states plots in Supplementary Fig. 24, there are states at the Fermi level in the spin up-channel for all structures resulting in half-metallic behavior. The binding of polysulfides to the surface of CoX₂ materials were further investigated with slab calculations. The energies of different surface terminations of pyrite (p-CoS₂) and marcasite (m-CoTe₂, m-CoSe₂) CoX₂ structures are shown in Supplementary Table 2. The surface energies (γ) were calculated using Eq. (1)

where E_slab is the slab energy for a given surface (single point or relaxed surface), E_bulk is the energy of relaxed bulk geometry per formula unit, N is the number of formula units in the slab, and A is the area of the surface. To screen a wider range of possible surfaces for different CoX₂ structures, the cheaper PBEsol functional was used instead of the SCAN + rvv10 functional. Supplementary Table 2 shows the surface energy of the eight-lowest energy surfaces of p-CoS₂ and the ten lowest energy surfaces of m-CoSe₂ and m-CoTe₂. As can be seen from Supplementary Table 2, the lowest energy surface for p-CoS₂ is (100), followed by the (210) surface, which is consistent with previous XRD and TEM results^28,29. Several low energy surfaces are present in the m-CoSe₂ and m-CoTe₂ marcasite structures, including the (010), (101), and (211) surfaces, which is consistent with low energy surfaces observed computationally and experimentally for analogous marcasite surfaces^30,42. The (100) surface of p-CoS₂ and the (101) and (010) surfaces of m-CoSe₂ and m-CoTe₂ were chosen to study the binding of polysulfides. The geometries of all of the low energy slabs predicted with PBEsol were reoptimized with SCAN + rvv10 before further binding calculations.

In order to understand the nature of binding between NaPSs and the cathode hosts, the enthalpy of Na₂S_y polysulfide molecules of different chain lengths (y = 1, 2, 4, 5, and 6) bound to both the p-CoS₂ (100) surface and a graphene sheet were calculated. The formation energy of the NaPSs molecules was taken as the difference in the energy between a slab of the cathode host with (E_slab+NaxS) and without the NaPSs (E_slab), referenced to the solid S and Na₂S phases via E_form = E_slab+NaxS − E_slab − x/2 × E_Na2S − (1 − x/2) × E_s. The effects of solvation on the slabs with and without the bound NaPSs were taken into account via the inclusion of an implicit solvation model in VASPsol⁴³. The lowest energy binding geometries for each NaPS are shown in (Supplementary Fig. 25). The formation energies (E_form) of solid phases of Na₂S, Na₂S₂, Na₂S₄, Na₂S₅, Na₂S₆, and S were calculated using the SCAN + rvv10 functional as shown in Fig. 7 (black symbols). The formation energies of intermediate Na₂S_y phases (E_NaxS) normalised to one mol of S (x = 2/y) are reported relative to the end member configurations of Na₂S (E_Na2S) and S (E_s) by the formula E_form = E_NaxS − x/2 × E_Na2S − (1 − x/2) × E_s. With respect to Na metal, the sequence of reactions from S → Na₂S₅ → Na₂S₄ → Na₂S₂ → Na₂S are predicted to occur at voltages of 2.28 V → 2.08 V → 1.91 V → 1.82 V, which are in good agreement with previous computational studies⁴⁴ and the experimental cycling data in Fig. 6a, particularly for the reactions from S to Na₂S₄, where a significant overpotential does not occur. The formation energy of the free Na₂S_y molecules in the presence of the implicit solvent, but without the presence of the catalyst host, is also shown in Fig. 7 to approximate the energy of the “liquid” state. From Fig. 7, it can be seen that the formation energy of the NaPSs bound to the graphene cathode host is very similar to the free NaPSs in the “liquid”, indicating a very weak binding between the nonpolar graphene surface and the NaPSs as observed in previous studies⁴⁵.

In comparison, the formation energy of all NaPSs bound to the p-CoS₂ (100) surface are lowered significantly, primarily due to the strong interaction between the Na₂S_y molecules and the polar p-CoS₂ surface. The difference in the energy between the NaPSs bound to the p-CoS₂ (100) surface and the ‘liquid state’ increases systematically as the chain length, y, decreases (x increases), reaching 2 eV for the Na₂S molecule. As discussed further in Supplementary Fig. 26a, the strong binding of the NaPSs to the surface is related to charge transfer between the undercoordinated Co²⁺ sites on the surface to the S atoms in the Na₂S_y cluster.

Interestingly, after binding to the p-CoS₂ (100) surface, the enthalpy of the bound Na₂S_y molecules closely matches the energy of the solid Na₂S_y phases. An absolute comparison of the energies is not possible from Fig. 7 as entropic contributions and the presence of explicit solvation have not been included, however, this close matching between the energy of the molecular NaPSs and solid Na₂S_y phases is likely to be beneficial, in which reactions in the molecular state are expected to be facile. For example, if the binding of a Na₂S molecule to the cathode host is weak, as in the graphene case, the voltage required to drive the reaction Na₂S₂ + 2Na⁺+2e⁻ → 2Na₂S on the graphene at is 0.59 V. As the voltage for the solid–solid reaction of Na₂S₂ to Na₂S is 1.82 V, a large overpotential of 1.23 V would be required to facilitate the reaction in the molecular state. If the cathode host binds the NaPSs too strongly, then the energy of the anchored Na₂S molecule will be lower than the Na₂S solid. Na₂S molecules will therefore cover the cathode host surface instead of forming bulk Na₂S particles, reducing the ability for it to catalyze further reactions. By balancing the energy between the anchored and bulk Na₂S_y materials, facile molecular reactions can take place on the cathode host surface which lead to the formation of bulk Na₂S particles.

The formation energy of Na₂S_y molecules bound to the marcasite p-CoSe₂ (100), m-CoSe₂ (010), m-CoSe₂ (101), and m-CoTe₂ (010) surfaces were also calculated, as shown in Supplementary Fig. 27. It can be seen from Supplementary Fig. 27 that all of the CoX₂ hosts strongly bind the NaPS molecules, resulting in formation energies that are close to those of the solid Na₂S_y phases. For the Na₂S end member, a small difference in the formation energy is observed on analogous surfaces of materials with the same structure, but with a different anion, i.e., p-CoS₂(100) vs. p-CoSe₂(100) and m-CoSe₂(010) vs. m-CoTe₂(010). The difference in the formation energy of a Na₂S molecule is larger on different surfaces of the same structure, e.g., m-CoSe₂(101) vs. m-CoSe₂(010), suggesting that structural differences in these materials (pyrite vs marcasite) may play a more important role in this family of materials than the nature of the anion (i.e., S, Se, and Te). This result is in line with the experimental electrochemical results in Fig. S20, where the marcasite based BPCSE and BPCTE structures have similar rate performance, whereas the pyrite based BPCS structure has markedly better rate performance.