Metal chalcogenide hollow polar bipyramid prisms as efficient sulfur hosts for Na-S batteries

Sodium sulfur batteries require efficient sulfur hosts that can capture soluble polysulfides and enable fast reduction kinetics. Herein, we design hollow, polar and catalytic bipyramid prisms of cobalt sulfide as efficient sulfur host for sodium sulfur batteries. Cobalt sulfide has interwoven surfaces with wide internal spaces that can accommodate sodium polysulfides and withstand volumetric expansion. Furthermore, results from in/ex-situ characterization techniques and density functional theory calculations support the significance of the polar and catalytic properties of cobalt sulfide as hosts for soluble sodium polysulfides that reduce the shuttle effect and display excellent electrochemical performance. The polar catalytic bipyramid prisms sulfur@cobalt sulfide composite exhibits a high capacity of 755 mAh g−1 in the second discharge and 675 mAh g−1 after 800 charge/discharge cycles, with an ultralow capacity decay rate of 0.0126 % at a high current density of 0.5 C. Additionally, at a high mass loading of 9.1 mg cm−2, sulfur@cobalt sulfide shows high capacity of 545 mAh g−1 at a current density of 0.5 C. This study demonstrates a hollow, polar, and catalytic sulfur host with a unique structure that can capture sodium polysulfides and speed up the reduction reaction of long chain sodium polysulfides to solid small chain polysulfides, which results in excellent electrochemical performance for sodium-sulfur batteries.

The material is prepared by carbonizing a Co acetate crystal which in effect gives a Co/C material and this serves as sulfur host. There are many reports on sulfur host materials and it is currently known that anode and electrolyte are more of a problem in these battery systems than new cathode host materials. And it is always a bit suspicious if the one and only solution comes up with a new quite exotic composite material. The authors claim that encapsulation and catalytic effects are important here. From the data given it is quite difficult to judge where the shuttle suppression comes from. The surface area of the material is quite low, so adsorption alone cannot be the mechanism. The material is not an enclosure (sulfur can access the inner core, polysulfides can go in and out). It might be a good idea to quantify the adsorption capacity with UV Vis following the method reported by Hippauf. This would give a good comparison with other more polar carbon host materials as reported by Hao. Also a water isotherm would be helpful to check the hydrophilic character of the host.
Of course a big flaw are the limited experimental details. No electrolyte/sulfur ratio given. That's a step backwards in reporting accurate data. With a whatman separator (thick) an excess of electrolyte is guaranteed! The huge capacity loss in the first cycle confirms this view (dissolution). And in a flooded cell PS diffusion takes longer. And the glass fibres also adsorb the PS, and affect crystallization.
It would be good to test the material also with a PE separator and compare it to a Ketjen Black based reference cathode. And check performance also under lean conditions with E/S ratios varying between 5 and 10.
The in situ analysis is interesting. However, do the authors think these solid phase intermediates take part in the conversion mechanism, or are just side products? The catalytic effect is probably only important for dissolved species.
A positive aspect is the massive materials characterization via SEM, XPS etc.
In summary it might be worth reconsidering this work if the material also performs well at low E/S ratios with a PE separator.
Additional points: 1) Severe technical problem: adsorption data and porosity analysis • This material has no high surface area, as one can see form the N2 isotherms (S8) • Fig. S8 probably shows a N2 adsorption isotherm at 77 K, but is labeled "surface area" • Hence, the author is not familiar with this technique • Moreover: The isotherm does not close, probably not enough mass used for measurement •  repeat (BPCS) • Pore size analysis is meaningless 2) The bipyramidal aspect is overemphasized. The crystals look almost like prism. Does the shape really play a role? 3) Conversions are nicely depicted in the supportings. This scheme would be helpful for the reader in the main text. 4) XRD: sulfur, alpha or beta? 5) What is TAA?
Reviewer #3 (Remarks to the Author): My review is for the DFT section of this submission, but I read the entire manuscript to understand the context of the DFT simulations. The results presented on a new polar bipyramid prism CoS2/C sulfur host (cathode) for sulfur batteries seem very promising and worthwhile of publication in Nature Communications, if the experiments hold up to scrutiny by experimental experts. With some edits, I support the publication of the DFT sections, if the other reviewers endorse the experimental sections.
The only two general comments I have are 1) on the "catalytic" description of the BPCS. I think a description of why the material is described as catalytic is warranted near the beginning of the paper, early in the introduction, because otherwise I am unsure why this description is so necessary. Don't all cathodes "catalyze" reactions?
2) Why is cobalt used rather than another metal; is the choice of cobalt important? I don't think the choice of cobalt versus other transition metals is discussed in the paper and the computational authors could test other metal ions! My general comments on the DFT section is that it's too long. For example, the section on the charge density differences (Fig 7b) are unsurprising and not that informative; the magnetic moment calculation may not be of interest to a general audience. Some of the DFT section should be moved to the supplemental information because the gas-phase simulations may not be such a good proxy for the real reactions in the electrolyte, 1M NaClO4 in tetra-ethylene glycol dimethyl ether. For example, the charge density differences could look at lot different if explicit solvent was included.

Specific comments:
In the discussion section, line 438, the authors state, "DFT calculations support the mechanism that sulfide adsorption is superior for homogenous metal sulfides…when compared to the carbon host." Unless I'm mistaken, this is one of the first times "homogenous" is used to describe the surfaces and it's unclear what the authors mean by "homogeneous". Otherwise, the statement seems well backed up by their calculations.
The authors do not comment on using a Hubbard-U term to correctly capture the electronic structure due to the cobalt ions. XPS of the cobalt has been performed (Figure 3), so a comment would be beneficial. Also, since the magnetic moment is described in detail, the effect of a Hubbard U should be discussed.
The authors do a thorough search for the low energy surfaces of CoX2 materials to create slab models. The authors do not clearly comment on why CoS2 is chosen over CoSe2 and CoTe2 for making the BPCS. Is the sulfur in the cathode necessary for good performance? What is the takeaway message from doing the selenium and tellurium calculations?
How do the authors ensure that they've left enough vacuum space to not have finite-size effect errors? Especially with the dipoles created on the surface due to the NaPSs molecules, 15 Å MAY NOT BE ENOUGH.
Line 361: It is interesting that the author choose to calculate the formation of free Na2S7 molecules in the presence of an implicit solvent using VASPsol. They should include an implicit solvent for the formation energies on graphene and p_CoS2 as well, or explain why they did not do these calculations. Figure 7 would be much more convincing and clear if implicit solvent were included as a comparison. Currently, the point of Figure 7 is confusing. There is a positive formation energy in the liquid and on graphene -indicating that the compounds should not form. Some of the formation energies on p-CoS2 are quite small -would they be positive if solvent was included?

Response to Reviewers
Reviewers' comments and our Response:
Response. Thank you very much for your affirmation of our work and your favorable comment on the manuscript. We are grateful for the reviewer's positive and constructive comments. All requested revisions based on the professional comments from the reviewer have been carried out.

A2.
We appreciate the reviewer for the very instructive suggestion. The exact Coulombic efficiency value is 98.5 %. We have provided it in the main text. …" (see Page 10 Line 25)

Q3. A Whatman glass fiber separator is used instead of a typical Celgard polymer separator.
The former is much thicker and will reduce energy density. What is the cycling performance using Celgard separator?
A3. Thank you very much for the very professional comment. There is some reason to use glass fiber as a separator: 1) Compared with the polymer separator, the thermal stability and mechanical properties of the glass fiber have been greatly improved. Therefore, the safety performance of sodium-sulfur battery can be greatly improved by using glass fiber as a new  4) The glass fiber can also act as a cushion to distribute the load across the electrode surface.
The Celgard usually has a denser structure and does not stick to the electrode surface as easily. This helps for post-mortem analysis of cells. 5) 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 as shown in Figure 2 and in supporting information Figure S20.  In summary, the glass fiber separator has high porosity and wettability, improves the absorption of dissolved polysulfide, slows down the diffusion process, and has good electrochemical performance.
2 Q4. 350 cycles is quite short for cycling. What is the performance like after 350 cycles and how can it be improved?

A4.
We have provided long cycling performance of BPCS@S in the main text ( Figure 4c). The specific capacity gradually decreases from 751 to 675 after 800 cycles at 0.5 C, as shown in Figure 3. This is good cycle performance and the capacity attenuation rate is of 0.0126% per cycle. The surface area of the material is quite low, so adsorption alone cannot be the mechanism. The material is not an enclosure (sulfur can access the inner core, polysulfides can go in and out).
It might be a good idea to quantify the adsorption capacity with UV Vis following the method reported by Hippauf. This would give a good comparison with other more polar carbon host materials as reported by Hao. In addition, a water isotherm would be helpful to check the hydrophilic character of the host.     As discussed previously, Figure 2 shows the clear performance comparison of glass fiber and Celgard polymer separator, as well as performance comparison of BPCS@S and KB@S composites.
To summarize, the glass fiber separator has high porosity and wettability, improves the absorption of dissolved polysulfide, slows down the diffusion process, and has good electrochemical performance. 2 Furthermore, to check the performance of BPCS@S electrode, utilization of sulfur and effect of electrolyte volume used, we check the performance of BPCS@S at different E/S ratios, as shown in Figure 6. Figure 6 shows that when we apply an E/S ratio of 7:1 to 10:1, the battery capacity is acceptable. The capacity fade at low E/S ratio can be explained by Na-S reaction mechanism. At low E/S ratio, the concentration and viscosity of polysulfides in electrolyte increase severely in the process of discharge, resulting in the increase of battery resistance and the discharge stops. Q5. The in situ analysis is interesting. However, do the authors think these solid phase intermediates take part in the conversion mechanism, or are just side products? The catalytic effect is probably only important for dissolved species.
A5. The solid phase intermediates in the discharge process is Na 2 S 2 , which reduced to Na 2 S in the discharge process. 3 According to the theoretical analysis, the conversion of Na 2 S 2 to Na 2 S will contribute about half of the theoretical capacity. However, the dynamics of this process is poor, so electrocatalysts can be used to accelerate this process. 4 The discharge process can be divided into four stages, which are the solid-liquid process, the liquid-liquid process, the liquid-solid process, and the solid-solid process. 5 At present, various electrocatalysts have been adopted to promote these four-conversion processes. 6,7,8,9 Therefore, electrocatalysis cannot only act on soluble polysulfides.

Q6.
Technical problem in Adsorption data and porosity analysis. This material has no high surface area, as one can see form the N 2 isotherms (Fig. S8) Fig. S8 probably shows a N 2 adsorption isotherm at 77 K, but is labeled "surface area". Hence, the author is not familiar with this technique. Moreover, the isotherm does not close, probably not enough mass used for measurement. Repeat for (BPCS). Pore size analysis is meaningless.

A6.
We appreciate the reviewer for a very instructive comment. We are sorry for ambiguity.
The already given data of adsorption and porosity is of CoTe 2 (BPTE) by mistake, it has low surface area because, it synthesized at high temperature (telluridation at 700 ℃). Now we repeat and present the N 2 adsorption data for BPCS as shown in Figure 8 and Figure S13.  Q7. The bipyramidal aspect is overemphasized. The crystals look almost like prism. Does the shape really play a role?
A7. Thank you for your comment. The size and shape of the host materials plays a role in the electrochemical performance. 11,12,13,14,15,16 In addition, during the conversion of S to Na 2 S an expansion in volume occurs (i.e., decreases in volume during subsequent devulcanization). Therefore, in order to obtain a practical Na-S battery, it is necessary to design an appropriate electrode morphology to provide the necessary conductivity and to buffer the volume change. The hollow BPCS prisms have an internal space of 376 nm wide, which is sufficient to accommodate S and polysulfides in it to and to alleviate the volume expansion, as shown in systematic illustrated Figure 8. A8. Thank you very much for your valuable suggestion. We have transferred the conversion scheme from supporting information to main text (Figure 1a).
A9. We use α-S 8 /orthorhombic sulfur. The XRD card of α-S 8 /orthorhombic sulfur is PDF#08-0247, which is already shown in the sulfur composite XRD figures (Figure 3e). Long chain polysulfides are reduced to Na 2 S 2 and Na 2 S, and Na 2 S 2 is further reduced to Na 2 S. During the charging process, Na 2 S is oxidized to soluble long chain polysulfides (Na 2 Sx, 3 ≤ x ≤ 8), and then oxidized to β-S 8 .

Q10. What is TAA?
A10. TAA is the abbreviation of thioacetamide (source of sulfur). We have written the full name of the abbreviation in the main text.
Response: Thank you very much for your affirmation of our work and positive comments for the manuscript. According to your suggestions, we have made the following explanations and modifications.

Specific comments:
Q1. On the "catalytic" description of the BPCS. I think a description of why the material is described as catalytic is warranted near the beginning of the paper, early in the introduction, because otherwise I am unsure why this description is so necessary. Don't all cathodes "catalyze" reactions?

A1.
Thank you for your professional comment. Room temperature Na-S batteries have great potential as a stationary energy storage battery in small and large power grids. However, due to the sluggish reaction kinetics, its energy and rate performance are limited, and its practical application is far from the theoretical value. To solve this problem, which is largely a problem of kinetics, a method based on catalysis is proposed. In this strategy, the catalytic sulfur host promotes the inter-conversion of polysulfides, speeds up the kinetics, alleviates the shuttle effect, and demonstrates good electrochemical performance. Therefore, it is necessary to use a catalytic sulfur host material to allow the sulfur cell to reach a higher energy density and move towards a practical sodium sulfur cell. First, we must distinguish between "catalyst" and "electrocatalyst". The catalyst is added to the electrode material to promote the reaction, while the electrocatalyst itself is the electrode to promote the reaction. Specifically, in sulfurbatteries the electrocatalyst sulfur host materials give good electrochemical performance by adsorbing dissolved polysulfides and catalyze the inter-conversion reactions of polysulfides.
For instance, carbon and other non-polar sulfur host materials (cathodes), such as carbon nanotubes, carbon nano fibers, carbon hollow nanospheres and double-shell carbon microspheres can act as sulfur host materials, but they don't have absorption ability and electrocatalysis capacity for polysulfides. 18,19,20,21 We can only prove that some materials can improve the performance, but we cannot be sure that all the host materials have catalytic properties for polysulfide conversion.
Q2. Why is cobalt used rather than another metal; is the choice of cobalt important? I don't think the choice of cobalt versus other transition metals is discussed in the paper and the computational authors could test other metal ions! My general comments on the DFT section is that it's too long. For example, the section on the charge density differences (Fig 7b) are unsurprising and not that informative; the magnetic moment calculation may not be of interest to a general audience. Some of the DFT section should be moved to the supplemental information because the gas-phase simulations may not be such a good proxy for the real reactions in the electrolyte, 1M NaClO4 in tetraethylene glycol dimethyl ether. For example, the charge density differences could look at lot different if explicit solvent was included.

A2.
Cobalt active metal centers have higher catalytic activity than other metals. 22,23,24 In order to design this unique morphology, cobalt plays an important role, because its morphology and porosity are of great significance to sulfur battery. 10 In the computational section of the paper, the analysis was restricted to an in-depth study of the CoX 2 family of dichalcogenide materials (X=S, Se and Te) to provide a detailed analysis of the effect of anion chemistry on the catalytic performance of the material. The implicit solvation surface calculations used in this work, while computationally quite intensive, could be generalized to study the effect of other transition metal ions on the electrochemical performance of Na-S batteries, although this is beyond the scope of the current work. The following sentences have been added to the paper to highlight why Co was chosen in this study and how the computational approach could be useful for studying other transition metal systems.
 experimental and computational methodology used in this work is widely applicable to other transition metal dichalcogenide systems. 22,25 We appreciate that there is currently an extensive discussion of the magnetic properties and charge distribution of the main text and as such, we have moved some of the discussion and Figure 7b and Figure 7b to the SI and added the following sentence to the main text  As discussed further in Figure S25, the strong binding of the NaPSs to the surface is related to charge transfer between the undercoordinated Co 2+ sites on the surface to the S atoms in the Na 2 S y cluster.
Q3. In the discussion section, line 438, the authors state, "DFT calculations support the mechanism that sulfide adsorption is superior for homogenous metal sulfides…when compared to the carbon host." Unless I'm mistaken, this is one of the first times "homogenous" is used to describe the surfaces and it's unclear what the authors mean by "homogeneous". Otherwise, the statement seems well backed up by their calculations.

A3.
Thank you for your comment. We use the term "homogeneous" to refer to the chemical composition and the unified surface of the BPCS. We have now better explained the meaning of homogeneous in that sentence.
Q4. The authors do not comment on using a Hubbard-U term to correctly capture the electronic structure due to the cobalt ions. XPS of the cobalt has been performed (Figure 3), so a comment would be beneficial. Also, since the magnetic moment is described in detail, the effect of a Hubbard U should be discussed.

A4.
The CoTe2 for making the BPCS. Is the sulfur in the cathode necessary for good performance?
What is the take-away message from doing the selenium and tellurium calculations?

A5.
In this study all three compounds CoS 2 , CoSe 2 and CoTe 2 were fabricated as bipyramidal prisms, referred to as BPCS, BPCSE and BPCTE, respectively. As shown in SI Figure S19 and Figure 20, the most reversible electrochemical performance was achieved for the CoS 2 system. The calculations demonstrated that all of CoX 2 materials bind Na x S y polysulfides stronger than graphene/graphite, which explains the improved electrochemistry of these materials. The calculations also show that the structure/surface termination of the CoX 2 particles has more of an influence on the Na x S y binding than the nature of the anion, which is important as CoS 2 adopts the pyrite structure and CoSe 2 and CoTe 2 adopts the marcasite structure. We have added the following sentence to the text to clarity this: It is certainly not clear from Fig. 7 a).

A8
. Reviewer 3 makes a good point that the 'binding energy' was not clearly defined in the text. The sentence has therefore been changed to:  The difference in the energy between the NaPS bound to the p-CoO 2 (100) surface and the 'liquid' state increases systematically as the chain length decreases, reaching 2 eV for the Na 2 S molecule.
The authors have addressed most of the referee comments well. Regarding the Whatman separator and E/S ratio, it turns out that only high E/S ratios work and the Whatman separator (very thick) is mandatory. For practical applications this is a severe limitation showing that the role of the porous materials is overemphasized (flooded cell). The relevance and impact of this work my thus be limited. The N2 adsorption hysteresis of BPCS is still difficult to understand. Instead of swelling it looks more like a kinetic effect due to the hindered diffusion through the shell during ad-and desorption.
In any case it is quite vague to derive a pore size distribution (neither for a swelling system nor for a kinetically hindered system PSD can apply).
Reviewer #3 (Remarks to the Author): My review was for the DFT section of the paper. The authors have addressed my concerns and fixed confusing parts of the computational section. I only have a few more suggestions, but publication is not contingent on the authors taking my suggestions.
If the editors agree that the experimental reviewers concerns were satisfied, I support publication.
Q1: The authors explained the use of catalytic in their response. I suggest they explicitly note in the main text that the cathodes (18)(19)(20)(21) are non-catalytic, such as the sentence at the bottom of page 4.
Q2: The authors have tidied up the DFT section by moving parts to the SI and also included a comment that cobalt was chosen over other transition metals, perhaps inspiring future work.
Q3: The authors have answered by question about homogenous. Perhaps they should write, "interwoven surfaces and chemical composition" to indicate that the chemical composition is also homogenous. Emerce et al. 2 studied that the specific energy increased significantly to 9-10 µL/mg E/S ratio.

II.
Hagen and co-workers also mentioned that capacity is acceptable for E/S above 7 µL/mg.

III.
Sun et al. 3 examined that, to build sulfur cell with good power capability the E/S ratio must be higher than 10 µL/mg.

IV.
Zhang et al. 4 have investigated that the appropriate E/S ratio for sulfur cell is 10 µL/mg.
V. Zheng and co-workers 1 also have investigated that the optimized E/S ratio for sulfur cell is 20 µL/mg.

VI.
Choi's research group . 5 also have indicated that 10 µL/mg E/S ratio is suitable for sulfur battery.
According to the E/S ratio of lithium sulfur battery in the above-mentioned literature and according to the Huang et al. 6 most of the recently published literatures on Li-S batteries use E/S greater than 10.
In addition, Wang Guoxiu et al. 7 published an article on Na-S battery in Nature Communications, and they used glass fiber separator as well as high electrolyte volume (20 µL/mg E/S ratio). Our results are superior to those reported previously, indicating that the Na-S battery has superior performance in the range of 7-10 μL/mg E/S.

Comment.
The N 2 adsorption hysteresis of BPCS is still difficult to understand. Instead of swelling, it looks more like a kinetic effect due to the hindered diffusion through the shell during ad-and desorption. In any case, it is quite vague to derive a pore size distribution (neither for a swelling system nor for a kinetically hindered system PSD can apply).
Response. Thank you so much for your suggestion and we appreciate your further input. We try to solve the problem of hysteresis loop by optimizing degassing time and degassing temperature, and use another analyzer with good performance.
Last time, we performed N 2 adsorption and porosity measurements on the Quantachrome QUADRASORB automatic surface area and pore size analyzer. The sample was degassed at room temperature for 12 hours. However, this time we measured on another machine Micromeritrics ASAP 2020 HD88 (Figure 2) with the sample degassed at 120 ℃ for 24 hours, which provided favorable results for the hysteresis loop, as shown in the following Figure 1 and supplementary Figure 13.
"…The Brunauer-Emmett-Teller (BET) surface area of BPCS is 119.9 m 2 g -1 and the hysteresis loop is of type H4 and indicates the physisorption isotherm of type I, which conforms to IUPAC, demonstrating that BPCS is composed of uniform slit-like pores, representing microporous patterns in structure. The microporous materials adsorb polysulfides more efficiently than other porous structures. 8 … (See Page 6 Line 18)"  Response. Thank you very much for your valuable suggestion and we appreciate your further input. I think we should be careful when saying that other materials are 'non-catalytic' because if it binds the NaPSs, it is likely to be have some effect on the kinetics of NaPSs conversion. We have already discussed this point in the previous response letter. However, we have now clarified this point in the main text as follow: "…As a result, hollow polar S hosts can more efficiently block NaPSs diffusion than other structures, such as nano-particles and flakes. In addition to the above strategies, the use of host materials that effectively catalyse the conversion of long-chain NaPSs (Na 2 S x , 4 ≤ x ≤ 8) to short chain NaPSs , which is a particularly promising approach to inhibit NaPSs diffusion. 9 Due to the insulating properties of sulfur and NaPSs, the electrochemical discharge/charge processes are sluggish. For non-catalytic hosts such as carbon, carbon nanotubes, carbon nano-fibers, carbon hollow nanospheres and double-shell carbon microspheres, the conversion of NaPSs is slow and the intermediate polysulfides can easily dissolve into the electrolyte. However, due to the use of the catalytic S host, such as electronically conducting transition metal sulfide hosts can effectively act as electrocatalysts to accelerate the redox kinetic of long chain NaPSs (Na 2 S x , 4 ≤ x ≤ 8) and efficiently convert to solid phase S and Na 2 S/Na 2 S 2 … (See Page 4 Line 18)" Comment. The authors have tidied up the DFT section by moving parts to the SI and also included a comment that cobalt was chosen over other transition metals, perhaps inspiring future work.
Response. Thank you very much for your affirmation of our reply. We are glad to see that we have addressed your point.
Comment. The authors have answered by question about homogenous. Perhaps they should