Scissor g-C₃N₄ for high-density loading of catalyst domains in mesoporous thin-layer conductive network for durable Li-S batteries

INTRODUCTION

The ever-growing demand for energy storage for evolving mobile devices, such as electro-vehicles, pushes the development of energy storage materials. This is due not only to the limited and uneven distribution of fossil energy in different countries and regions but also to the vital commitment to achieving carbon peak and carbon neutrality^[1]. The lithium-sulfur battery (LSB), owing to its satisfying theoretical capacity of 1,672 mAh g^-1 and relatively low cost, is becoming a promising candidate in energy storage systems^[2]. However, the intrinsic electron insulativity and self-redox mechanism of the sulfur cathode during the charge/discharge processes would lead to sluggish kinetics and high polarization^[3], as well as the unbearable “shuttle effect”^[4]. The shuttle effect is caused by the dissolution of the lithium polysulfide (LiPS) intermediates in an electrolyte, which decreases the ionic conductivity of the electrolyte and the use of sulfur. The dissolved LiPSs would transfer to the anode side, leading to the self-discharging and fast consumption of active materials.

Plenty of strategies have been applied to ease the aforementioned shortcomings, including cathodic structure design^[5,6] and electrolyte formula modulation^[7,8]. Beyond these, the separator modification also becomes attractive due to its tailored functions to retard the direct contact between the active anode and cathode materials but allow the selective ion transfer^[9,10]. By simply loading the functional coating on the separator at the cathode side, the shuttle effect of LiPSs can be effectively mitigated. In this respect, two main reasons should be considered: (1) The blocking function of modified coating on the relatively large pores of a pristine polypropylene (PP) separator can prevent the straight transfer of large LiPSs molecules; (2) The interaction between polar separator decoration and polar LiPSs through electrostatic adsorption or chemical bonding can also prevent the leakage of LiPSs from the cathode side^[11,12]. Carbon-based materials with controllable morphology were first introduced as separator decorations to improve the electrochemical performance of LSBs^[13,14]. However, even though the transportation behavior of LiPSs at the cathodic side can be modulated by the guidance of channels in carbon materials, the Van der Waals force between the carbon modification layer and LiPSs is not strong enough to capture the LiPSs molecules^[15]. Thereafter, the heteroatom doping (N, O, etc.) in carbon materials was proposed to introduce the strong Lewis acid-base interaction between Li⁺ cation in LiPSs and electronegative N or O atoms^[16,17]. In addition, the carbon materials can also be modified by two or more kinds of elements, e.g., in the form of metal alloys^[18], metal oxides^[19], metal sulfides^[15], metal nitrides^[20,21], etc. These combinations are flourishing because of the potential synergistic effect between different elements. For example, the Ni/Co/KB (KB: Ketjen black) composite derived from Prussian blue analog was adopted as the modified layer on the separator, where the mesoporous matrix of KB not only promotes the electrolyte penetration but also untangles the polysulfide shuttle effect, and the encapsulated Ni and Co elements in composite do a great favor to the conversion reaction kinetics^[11]. A similar strategy was delivered by adopting nitrogen-doped graphene and two metals into Ni₃Sn₂ alloy^[18]. Unlike the limited element metals doped in the carbon matrix, the non-noble transition metal nitrides (TMNs) are easier to fabricate and exhibit a competing potential because of the low-cost and satisfying electrocatalysis properties^[22]. The combination of metal and nitrogen would narrow the d-band of the metal, delivering the electron structure of TMNs similar to noble metals in the Fermi Level and therefore improving the activity of catalysis sites^[23]. For instance, Zhang et al. synthesized a bimetallic nitride of Ni_0.2Mo_0.8N blended with carbon nanotubes to modify the separator, and it can construct an internal electric field to suppress the shuttle of polysulfides^[22]. The change of electron structure by the Mo doping makes the corresponding LSBs a better electrode conductivity and catalytic effect on polysulfides.

Based on previous research, niobium nitride possesses a high electron conductivity and high chemical stability against the electrochemical environment^[24]. When embedded in the carbon framework as a sulfur host material, the polar NbN is expected to interact with the free LiPSs in electrolyte and promote the conversion reaction of LiPSs to Li₂S₂/Li₂S^[20,21,25]. In order to maximize the catalytic effect of NbN, its contact environment with electronic/ionic conductive substrate and LiPSs is crucial. The aggregation of NbN particles with the loss of electrocatalysis surface or the loose blending of NbN with dense or small-surface carbon materials would seriously degrade the performance of LiPS conversion. However, the uniform and sufficient loading of fine NbN nanoparticles in monolithic porous carbon hosts is still a big challenge. The prior fusion of Nb, N, and C element precursors is highly required to achieve the desired catalyst-in-conductor composite host. Here, we developed a molten salt method to prepare the NbN nanoparticle catalyst evenly loaded in the porous 3D carbon matrix that derives from the g-C₃N₄ precursor. This novel NbN in carbon structure (denoted as NbN/C) is used as a separator modification layer for LSBs with the following merits. First, the synthesis strategy is simple and of one-pot feature. The g-C₃N₄ precursor is chosen as the dual source of the nitridation effect and the remaining carbon skeleton. The resultant modification layer with a 3D porous structure enables the sufficient contact area between the electrolyte and NbN/C composite, beneficial for the LiPSs adsorption, catalysis, and conversion reaction. The corresponding LSBs deliver a highly reversible cycling performance under a high rate of 2 C, with a capacity retention of 72.7% after 1,000 cycles and a satisfying rate performance even under the higher rate of 5 C. When the sulfur loading in the cathode is increased to 6.6 mg cm^-2, a high areal capacity of 7.08 mAh cm^-2 can be realized.

RESULTS AND DISCUSSION

Figure 1A shows the schematic of a one-pot strategy to synthesize a 3D porous carbon skeleton with evenly-embedded NbN nanoparticles. Simply, the precursors of niobium ethoxide and g-C₃N₄ are well mixed in the molten salt of NaCl under the inert atmosphere. In this process, the N element in g-C₃N₄ is extracted by breaking the C-N bond under the potential catalysis effect of thermal Nb cation. The residual C element from the nanosheet self-assembled g-C₃N₄ microsphere is precipitated in the form of a porous and monolithic carbon framework with the preservation of nanosheet structure units. The high inner porosity of this carbon framework and its g-C₃N₄ template precursor is favorable for the injection and diffusion of niobium ethoxide under the molten salt environment, leading to the bonding of released N atoms and injected Nb cations and nucleation of NbN inside the porous framework. The quick inward diffusion of low-boiling-point niobium ethoxide can retard the irregular accumulation of NbN catalyst near the outer surface or even out of the carbon framework. The space confinement in the carbon framework limits the growth of NbN precipitates. Therefore, the NbN nanoparticles located uniformly in the 3D carbon network can be acquired. Such optimization of NbN particle size and distribution is favorable for the performance of LSBs, as discussed later.

Figure 1. (A) Synthesis process schematic of NbN/C composite. (B) XRD pattern of NbN/C. (C) Band gap of NbN through DFT calculation. (D) SEM image and (E) TEM image of as-synthesized NbN/C. (F) N₂ adsorption-desorption isotherms of NbN/C, inset: pore size distribution based on the BJH method.

The X-ray diffraction (XRD) pattern (as shown in Figure 1B) of the synthesis product fits well with the face-centered cubic NbN phase (JCPDS No. 38-1155) with the characteristic peaks corresponding to (111), (200), (220), and (311) planes. The XRD result with a rare impurity peak indicates the successful production of NbN with good crystallinity. The broad peak located at the smaller angle is ascribed to the carbon species derived from g-C₃N₄, and it tells the high defect concentration in carbon material that is beneficial for the improvement of conductivity^[26,27]. The NbN phase is expected to deliver a high electron conductivity due to its 0 eV band gap, which is determined by the density functional theory (DFT) calculation (as shown in Figure 1C). Both the high conductivities of the carbon matrix and embedded NbN nanoparticles favor the kinetics of soluble LiPS conversion into solid Li₂S₂/Li₂S, especially in the contact area between the sulfur cathode and the modified side of the separator. The X-ray photoelectron spectroscopy (XPS) results of the primary NbN/C powder are shown in Supplementary Figure 1. Therein the existence of the strongest C-C peak located at 284.8 eV indicates that most of the three C-N bonds (C-N=C, N-C₃, and C-N-H) in the g-C₃N₄ precursor are broken in the final NbN/C powder. From the residual bonds of C-N=C/N-(C)₃ in N 1s and C-N in C 1s and their intensity comparison, we can deduce the g-C₃N₄ is prone to be scissored from the C-N-H region, then C-N=C region, and finally N-C₃ region during the synthesis. The formation of NbN is further indicated by the appearance of NbN peaks at 397.2 eV in N 1s and 204.0 eV in Nb 3d_5/2. The bright spots in the scanning electron microscope (SEM) images (as shown in Figure 1D and Supplementary Figure 2) prove the evenly-distributed NbN nanoparticles in the monolithic micro-sized carbon matrix consisting of nanosheet structure units. The residual of C-N=C/N-(C)₃ moieties can help preserve the monolithic carbon framework to some degree, preventing it from fragmentation into the much smaller discrete nanosheet units. These signals also suggest the potential existence of broken C_xN_y in the carbon matrix. From the energy dispersive spectroscopy (EDS) mapping result of the NbN/C powder, the atomic ratio of C/N is nearly 25/2 (as shown in Supplementary Figure 3). This value is far from the stoichiometry of g-C₃N₄. The loading uniformity and permeability of NbN nanoparticles (with a size around 20 nm) across a thin-layer carbon network is further disclosed by the transmission electron microscope (TEM) observation (as shown in Supplementary Figure 4). From the high-resolution TEM (HRTEM) imaging, one can observe the lattice fringes corresponding to (200) and (111) planes of cubic NbN (as shown in Figure 1E), agreeing with the XRD result. The porosity of NbN/C is analyzed by using a nitrogen adsorption-desorption isotherm, as shown in Figure 1F, which delivers a type IV isotherm with a hysteresis loop at high relative pressure and reveals a mesoporous structure in accordance with the aforementioned SEM or TEM studies. The mesopore size distribution is acquired according to the Barrett-Joyer-Halenda (BJH) analysis on the desorption branch of the NbN/C sample (inset of Figure 1F), indicating the pore size range is mainly around 2.2-5 nm. The narrow pores left after N extraction from the nanosheet structure units are responsible for such a pore size distribution. The specific area is calculated to be as high as 335.4 m² g^-1 by the Brunauer-Emmett-Teller (BET) method, which is much higher than that (88.5 m² g^-1) of the g-C₃N₄ precursor as reported in our previous work^[28,29]. The increase of specific surface area further confirms the extra pore creation by the N extraction from g-C₃N₄, as well as by the pillaring of NbN nanoparticles between carbon layers. The planting of NbN nanoparticles can effectively expand the carbon framework and retard its structure collapse during cycling. This configuration would favor the electrolyte penetration and enlarge the reaction area of LiPS conversion.

The mesoporous structure and polar NbN bonding in NbN/C composite are expected to be favorable for the fast penetration of electrolytes. The electrolyte wettability of PP separators with and without NbN/C decoration was measured by contact angle experiments. As shown in Figure 2A, the modified PP delivers a much smaller contact angle of only 3.6°, which is much lower than that of the blank PP (46.9°, detected in the same condition). Besides, the digital images of fold tests of the modified separator in Figure 2B tells the high adhesion of the NbN/C layer on PP, and there is no peel-off phenomenon of attached substance or cracking phenomenon even after folding the modified separator several times. The porous surface of the carbon skeleton and the vacuum filtration process greatly benefit the combination between NbN/C and PP substrate by not only enlarging the contact area but also providing air pressure to condense the NbN/C layer. To compare the surface morphologies of the separators before and after modification, the corresponding SEM images are shown in Figure 2C and D. The blank PP shows a flat surface with plenty of open porous tunnels (as shown in the inset of Figure 2C) that are beneficial for ion and mass transportation. On the contrary, the NbN/C decorated separator (denoted as NbN/C@PP) exhibits a scratchy structure with hierarchical porosity, which still allows the facile wetting of electrolytes. The thickness of the modifier layer in Figure 2E is measured to be about 8.07 µm. In the modifier layer, both the carbon skeleton and NbN can physically and chemically absorb the dissolved LiPSs, preventing them from arriving at and passing across the PP separator^[30]. The metal nitride can even further catalyze the conversion reactions of LiPS intermediate products to the final solid products of Li₂S₂/Li₂S, as discussed later, further mitigating the shuttling of dissolved LiPSs across the separator.

Figure 2. (A) Contact angle test on NbN/C-modified PP (bottom) and pristine PP (upper). (B) Folding test to prove the strong adhesion of the modifying layer to PP. SEM images of (C) pristine PP separator and NbN/C@PP from (D) top view and (E) cross-section view. (F) Digital images of adsorption test in Li₂S₄ solutions before and after adding carbon black and NbN/C. (G) Adsorption energy calculation of Li₂S₄ on the surfaces of C (upper) and NbN (200) (bottom). (H) “Shuttle effect” test by using an H-type container with different separators: NbN/C@PP (bottom) and pristine PP (upper).

As shown in Figure 2F, the dissolution of Li₂S₄ in the tetraethylene glycol dimethyl ether (TEGDME) makes it turn from a colorless solvent into a yellow solution. When carbon black is added into the Li₂S₄ solution with vigorous stirring and then standing for 72 h, the solution still remains yellow, indicating the weak adsorption effect of pure carbon on LiPSs. In contrast, the color of the solution after adding NbN/C changes back and becomes colorless, suggesting the strong adsorption ability of NbN towards LiPSs. To further prove the adsorption effect, DFT calculation was applied to calculate the adsorption energies between Li₂S₄ and different additives. As shown in Figure 2G, the adsorption energy between Li₂S₄ and carbon black is -0.13 eV, which is not strong enough for carbon black to capture Li₂S₄. However, the adsorption ability of NbN/C is much stronger in view of the much more negative bonding energy of -2.09 eV due to the potential bonding of Li-N and S-Nb. The calculation result corresponds to the above visualization experiment. When the NbN/C composite is applied as the modifying layer on the separator, the “shuttle effect” is greatly inhibited, as shown in Figure 2H. The H-type container used here is composed of two small bottles that are separated by different separators. The bottle on the left is filled with the electrolyte with dissolved LiPSs, while the one on the right is filled with pure electrolytes. Two different separators of blank PP and NbN/C@PP are applied to investigate the restriction function of NbN/C on the transference of LiPSs. After standing for 1 h, a certain amount of LiPSs has been transported through the blank PP, and the electrolyte on the right becomes yellow. When the staying time is extended to 6 h, a much more severe “shuttle effect” of LiPSs happens with the appearance of more LiPSs in the right electrolyte. On the contrary, the NbN/C@PP enables the different electrolytes at both sides to keep their initial states no matter the standing time changes. This is because the NbN/C on the left side of the PP separator stops the transference of LiPSs, as illustrated in the schematics. The strong adsorption ability of NbN towards LiPSs and the hierarchical mesoporous feature of stacked carbon layers are responsible for the above phenomena.

As shown in Figure 3A, the shuttle current was measured, and it derives from the self-discharge, i.e., the shuttle effect. By holding the Li-S cell with NbN/C@PP or blank PP separator under a voltage of 2.38 V for a long time, an obvious disparity for the two cells on the shuttle current is presented. The shuttle current for the NbN/C@PP-based cell is around 2.39 × 10^-3 mA cm^-2, which is much lower than that for blank PP (29.18 × 10^-3 mA cm^-2). This disparity further proves the existence of strong and abundant adsorption sites on the NbN/C@PP separator, thereby greatly restricting LiPSs and suppressing the “shuttle effect” in the Li-S cell. It is worth noting that the galvanostatic (dis)charge is a dynamic process during which the charge transfer would have a great impact on the electrochemical performance, such as voltage hysteresis, electrochemical resistance, and more^[31]. Hence, besides the function of adsorbing LiPSs, a higher demand for accelerating polysulfide conversion in the modifying layer is required. The NbN/C layer, as expected, shows a satisfying catalytic property on the LiPS conversion reactions, as illustrated by kinetics analysis. The redox reaction between S₈ and Li₂S on NbN/C and blank carbon are compared by using cyclic voltammetry (CV) in symmetric cells at 1.0 mV s^-1, which have the identical working and counter electrodes, and 1 M lithium bis(trifluoromethanesulfonyl)imide (LiTFSI) solution with 0.1 M Li₂S₆ is used as an electrolyte. Two pairs of symmetric redox peaks are clearly exhibited for NbN/C electrodes in Figure 3B, whereas the redox peaks for blank carbon electrodes are less symmetric with more redox stages. The former peaks are much sharper and have smaller voltage hysteresis than those for the latter, indicating the faster liquid-solid conversion reactions associated with Li₂S₆ adsorbed by NbN/C. There is no redox peak observed for the NbN/C electrode when the Li₂S₆-free electrolyte is used. Note that a new pair of peaks (a/a’) emerge under the relatively higher scan rates in the case of NbN/C (as shown in Supplementary Figure 5). The appearance of this couple of peaks with larger voltage polarization indicates the longer survival time of the shorter-chain polysulfide anion (S₄^2-) due to the catalysis effect of NbN^[5]. Because the Li₂S precipitation from Li₂S₄ intermediate conversion shares three-quarters of the whole theoretical capacity during the discharge process of sulfur^[32], the electrodeposition rate of Li₂S on the substrate material could be an important kinetics indicator for LSBs. Here, the potentiostatic discharge curves of two coin cells with different separators at 2.05 V are compared, as shown in Figure 3C and D. To mitigate the influence of long-chain polysulfides, the cells are firstly discharged to 2.06 V, and then a small bias of 0.01 V is applied to drive the nucleation and growth of Li₂S^[33]. According to the research of Lu^[34], the Li₂S deposition complies closely with the 2D instantaneous nucleation model in a glyme-based electrolyte, and the relationship between reaction time (t) and current response (J) can be quantitatively modeled through the following equation^[35]:

Figure 3. (A) Shuttle current comparison of Li-S cells by using different separators. (B) CV curves of Li|Li symmetric cells with/without 0.1 M Li₂S₆. Current vs. time curves with different stages of reductions of Li₂S₈ and Li₂S₆ and precipitation of Li₂S during potentiostatic discharge of Li₂S₈/TEGDME catholyte by using (C) NbN/C@PP and (D) PP separator at 2.05 V. (E) EIS result of Li-S cells discharged to 2.05 V with different separators.

where J_m and t_m represent the peak current and the time when the peak current occurs, respectively. Obviously, the cell with an NbN/C-modified layer shows a higher current peak at the earlier time compared with the case of bank carbon. The much sharper peak indicates that the high-density NbN domains in the system can greatly improve the conversion reaction of Li₂S₄ into Li₂S. The capacities of Li₂S precipitation calculated from the integral of current vs. time are 265.5 mAh g^-1 and 134.5 mAh g^-1 for the separator-modified and unmodified cells, respectively. The rate of Li₂S electrodeposition during the potentiostatic process can be calculated using the following equation:

in which N₀ represents the areal density of Li₂S nuclei, and N₀k² gives the effective rate constant for the Li₂S coverage on the electrode surface^[35]. The N₀k² value is 2.62 × 10^-9 s^-2 for NbN, much higher than that of the control cell (5.92 × 10^-10 s^-2). This analysis indicates that the thin-layer carbon matrix loaded with high-density NbN nanodomains can provide sufficient nucleation sites for Li₂S due not only to the sulfiphilic property of metal nitride surface but also to the potential catalysis function as discussed later with DFT calculation.

The electrochemical impedance spectra (EIS) of corresponding cells were also measured. When both the cells are discharged to 2.05 V, the impedance of the NbN/C-based cell shows a much smaller semicircle and a steeper slope at the low-frequency region (as shown in Figure 3E) compared with the case of the control cell, proving the faster charge transfer and Li diffusion at the stage of reduction to solid Li₂S₂/Li₂S under the help of the NbN catalyst^[36]. To better understand the conversion reaction of the sulfur cathode under different separators, the CV test on the Li-S cells was conducted at a scan rate of 0.2 mV s^-1, as shown in Figure 4A. Two cathodic peaks can be detected for both the cells and taking the NbN/C case as an example, the higher peak at 2.35 V corresponds to the reduction process of S₈^2- and S₆^2- polysulfides to S₄^2- and the other peak at 2.03 V represents the nucleation of solid Li₂S₂/Li₂S by the deeper reduction from Li₂S₄. Both the higher voltage of long-chain polysulfide reduction and the stronger current response of Li₂S₂/Li₂S precipitation under the assistance of NbN/C indicate the improved kinetics of multi-step conversion reactions. The advantage of reconversion reaction kinetics is also observed for the NbN/C case in view of the larger area of anodic peak current response. The Li diffusion coefficient D_Li+ as another kinetics indicator can be acquired by introducing the CV curves at different scan rates from 0.1 to 0.4 mV s^-1 (as shown in Supplementary Figure 6). According to the linear relationship between peak current density (I_pa) and the square root of scan rate (v), the D_Li+ value can be obtained according to the Randles-Sevcik equation^[37]:

Figure 4. (A) The CV test on the Li-S cells using different separators under a scan rate of 0.2 mV s^-1. (B) Diffusion coefficient calculation based on the relationship of peak current and square of sweep rate from CV profiles at different sweep rates. (C) Galvanostatic discharge and charge curves under a rate of 0.2 C. Tafel plots corresponding to the reduction reactions (D) from elemental sulfur to Li₂S_n and (E) from Li₂S_n to Li₂S. (F) Relative activation energy evolution of S₈-Li₂S_n-Li₂S conversion reactions based on different separators. (G) Adsorption energies and (H) Gibbs free energy profiles of LiPSs on the surface of NbN (200) by DFT method.

Here, n is the number of reaction electrons (e.g., for the second cathodic process investigated, n = 2), A is the area of the cathode (0.785 cm²), and C represents the concentration of Li⁺ in the electrolyte (0.0012 mol cm^-3). As shown in Figure 4B, the existence of NbN/C on the separator improves the Li diffusion coefficient from 3.633 × 10^-8 cm² s^-1 to 1.748 × 10^-7 cm² s^-1. The kinetic optimization is also reflected in the galvanostatic curves during charge and discharge processes (as shown in Figure 4C). As seen, the voltage hysteresis (169 mV) for the NbN/C-modified Li-S cell is smaller than that (192 mV) of the control cell. Besides, there are two obvious nucleation overpotentials at the liquid-solid and solid-liquid conversion stages during the cycling of Li-S cells without NbN/C decoration. However, these bulge points nearly disappear after the introduction of NbN/C, suggesting the potential catalysis of high-density NbN domains in these relatively sluggish conversion stages. This phenomenon is also consistent with the faster charge exchange as obtained by Tafel plots. Specifically, a voltage range of ±200 mV is applied to the coin cells with different separators at the specific voltages, which are determined by the reduction and oxidation peak positions in the CV test. As excepted, the NbN/C-modified cell displays a lower Tafel slope of 126 mV dec^-1 at the stage of S₈-to-Li₂S_n conversion and 56 mV dec^-1 at the stage of Li₂S_n-to-Li₂S conversion when compared with those of the control cell (as shown in Figure 4D and E). When the cells are recharged, the lower Tafel slope of 55 mV dec^-1 for Li₂S-to-Li₂S_n is also detected for the modified cell (as shown in Supplementary Figure 7). These results further confirm the accelerated liquid-to-solid and solid-to-liquid conversions, benefiting from the catalysis of NbN domains. Based on the Tafel slopes, the difference in activation energies (E_a) for corresponding conversion stages can be calculated^[38]. The detailed information is listed in Note S1 in Supporting Information, and the result is summarized in Figure 4F. Both the activation energies for the reduction processes of S₈-to-Li₂S_n and Li₂S_n-to-Li₂S are obviously decreased by 8.63 kJ mol^-1 and 19.08 kJ mol^-1, respectively, after the introduction of NbN/C, strongly supporting the proposed catalyst ability of NbN/C on the separator. Therefore, the sluggish conversion reactions caused by the substantial phase changes are greatly activated.

To better understand the function of NbN on the conversion reaction of LiPSs, DFT calculation is further applied. The adsorption energies of different LiPSs on the (200) plane of NbN are summarized in the column graph in Figure 4G, and they are all higher than 2 eV (as shown in Supplementary Table 1). The high adsorption energies would greatly help the NbN on PP to catch the dissolved LiPSs in the electrolyte during cycling and inhibit their shuttle effect. It is worth noting that both the Li and S in the polysulfides show an interconnection with the NbN surface (the schematics of corresponding adsorption systems are inserted in Figure 4H), and some broken S-S bonds within long-chain LiPSs indicate the catalytic function of NbN with the acceleration of conversion reaction. This is in accordance with the small overpotential observed in CV curves and in the electrochemical cycling tests. Besides, the Gibbs free energies (ΔG) of the main intermediates from S₈ to Li₂S under the catalysis of NbN are calculated, and ΔG is an effective indicator to investigate the conversion mechanism, as listed in Figure 4H. Taking the system of S₈ with NbN as the reference, the first step of S₈ transformation into Li₂S₈ is a spontaneous exothermic reaction with the decrease of relative free energy of -3.42 eV. The following four steps, on the contrary, are all endothermic reactions, showing a growing ΔG from Li₂S₈ to Li₂S, and the corresponding energy barrier value for each conversion step is listed in Supplementary Table 2. Even so, the whole energy from S₈/NbN to Li₂S/NbN system is decreased under the catalysis of NbN. Note that the conversion step from Li₂S₂ to Li₂S requires overcoming the highest energy barrier of 0.81 eV, making it the rate-controlling step during the discharge process^[39].

The electrochemical performance of LSBs was measured under a high rate of 2 C (1 C = 1,672 mA g^-1) within the voltage range of 1.7-2.8 V. Figure 5A displays the long cycling performance of LSBs with/without the NbN/C-modified PP separator. Under the strong adsorption and catalysis effects of NbN, the modified cell has a much higher specific capacity and a more stable capacity retention of 72.7% (from 771.8 to 561.1 mAh g^-1) than that of the control cell (47.4%) even after 1,000 cycles. Due to the poor conductivity and volume change of sulfur cathode during cycling, the cathode and modification coating may be passivated or deformed, and it would lead to degradation of the use of active materials, especially after long-term cycling. Besides, the roughening of the Li metal anode during long-term cycling would consume the electrolyte and cause capacity degradation. Note that the specific capacity of cells using Al foil as current collectors decreases faster than that using carbon paper, although they exhibit a similar performance during the first 200 cycles (as shown in Supplementary Figure 8). The main reason for this phenomenon could be ascribed to the easier exfoliation of S-based active material from the relatively smoother Al foil than from the relatively rougher carbon paper during cycling, in view of the volume change during the S-based conversion reaction. The activated kinetics of the LiPS conversion reaction should be ascribed to the addition of NbN, and it is also reflected in the charge/discharge curves (as shown in Figure 4C). The rate performance of the modified cell is also improved with the current density changing from 0.2 C to 5.0 C (with reversible capacities of 742.6, 696.6, and 652.6 mAh g^-1 at 3, 4, and 5 C, as shown in Figure 5B), and the reversible capacity would return back immediately once the rate is back to 0.2 C, indicating good structure stability of NbN/C network during high-rate cycling. When the sulfur loading is increased to as high as 6.6 mg cm^-2, the NbN-modified cell still exhibits stable cycling for more than 100 cycles under the current density of 1 mA cm^-2 and electrolyte-to-sulfur ratio (E/S ratio) of 5 µL/mg, as shown in Figure 5C and Supplementary Table 3. Under the lean electrolyte condition, the areal capacity reaches approximately 7.08 mAh cm^-2 after a prior activation process of 2 cycles and is preserved at 5.61 mAh cm^-2 (850 mAh/g) after 100 cycles. This performance is relatively satisfying when compared with some others reported before, as listed in Supplementary Table 4. We also tested the Li-S cells with different loadings of cathodes (as shown in Supplementary Figure 9). Although the capacity of the cell with 6.6 mg cm^-2 loading decays after 100 cycles, the discharge capacity is still much higher than that in the case of 7.8 mg cm^-2 loading and is close to the case of 3.4 mg cm^-2 loading.

Figure 5. (A) Cycling stability test of Li-S cells based on different separators at a rate of 2 C. (B) Rate performance of Li-S cells at different rates from 0.2 C to 5.0 C. (C) Cycling performance of Li-S cell using NbN/C@PP as separator with a high sulfur loading of 6.6 mg cm^-2 under a current density of 1 mA cm^-2. XPS spectra of the surface of cycled NbN/C@PP separator in (D) NbN 3d and (E) S 2p. SEM images of (F) lithium anode and (G and H) NbN/C@PP separator after cycling for 50 cycles.

The cycled Li-S cell with a modified PP separator was disassembled to investigate the chemical information on the surface of NbN/C (e.g., the interaction situation of NbN with LiPSs) by XPS method (as shown in Figure 5D and E, and Supplementary Figure 10). Note that the Nb 3d region has a splitting of spin-orbit components with a distance of 2.78 eV between Nb 3d_3/2 and 3d_5/2 peaks. The peak spitting distance for S 2p is 1.16 eV between S 2p_1/2 and 2p_3/2. The NbN peak is found at 204.5 eV in Nb 3d_5/2^[25]. The peaks of Nb-O/NbN-O (at 207.0 and 205.5 eV, respectively) could also be detected in view of the potential reaction between NbN and electrolyte during cycling and the potential O dopant transfer from O-doped g-C₃N₄ into NbN. The trapping of LiPSs (and even Li₂S₂/Li₂S) and their shuttle suppression by NbN/C are indicated from the appearance of Nb-S peaks at 204.0 eV and 162.2 eV as observed in Nb 3d_5/2 and S 2p_3/2, respectively^[25]. The interaction between Nb cation- and S-based species is in accordance with the DFT calculation results. Note that there are no substantial peaks of long-chain polysulfides in S 2p, but the Li₂S signal (160.2 eV) at the discharged state indicates the thorough conversion of LiPSs to Li₂S under the catalyst of NbN^[40,41]. There are also the peaks corresponding to the thiosulfate and polythionate species detected at 167.2 and 169.2 eV in S 2p_3/2, respectively^[24]. The thiosulfate should be the result of the redox reaction between LiPSs and modifying layer on PP, and the subsequent reaction of thiosulfate with LiPSs is responsible for the formation of polythionate. The morphologies of the lithium metal anode and NbN/C-modified layer near the cathode side after 50 cycles are observed by SEM imaging. The cycled Li metal surface is still flat without evident dendrite growth but with some pits (left after Li stripping), as shown in Figure 5F. As shown in Figure 5G, the cycled NbN/C layer does not collapse or agglomerate under the galvanostatic process, and it still holds a similar porous morphology as the original structure. The interconnected carbon skeleton derived from C₃N₄ is well preserved and is still loaded with sufficient and dispersed NbN nanoparticles (as shown in Figure 5H). These features do not degrade the electrolyte wettability at the cathode side, enabling the enlarged contact of LiPSs with the NbN/C interlayer. The large contact area avoids the possible stacking of solid Li₂S₂/Li₂S precipitates (as shown in Figure 5H), and the high dispersity of conversion products makes them possible for a sustainable reaction during long-term cycling.

CONCLUSION

In summary, we propose to scissor the g-C₃N₄ precursor for high-density loading of catalyst domains in mesoporous thin-layer conductive networks via a facile molten salt synthesis method for durable Li-S batteries. The resultant NbN in carbon composite is used as a separator modifier for the strong adsorption towards dissolvable long-chain polysulfide intermediates and their catalytic conversion reactions. The LSBs with the NbN/C@PP separator demonstrate superior long cycling stability (at least 1,000 cycles at a high rate of 2 C) and rate performance (652.6 mAh g^-1 at a higher rate of 5 C). The durable cycling is still achievable when the sulfur loading is as high as 6.6 mg cm^-2, delivering a high areal capacity of 5.61 mA h cm^-2 after 100 cycles. This g-C₃N₄-assisted strategy opens a new gate for the integrated catalysis-conduction network for high-performance LSBs.

EXPERIMENTAL SECTION