Introduction

Sulfur is an attractive electrode material because of low cost and high-theoretical specific capacity of ~1675 mAh g−11. Sulfur electrodes can be conjugated with a range of metal anodes in rechargeable metal–sulfur (M–S) batteries, giving promise of practical energy-storage applications2,3,4. However, sulfur reduction reaction (SRR) in M–S batteries is a complex conversion from elemental sulfur to insoluble metal sulfides5. Sluggish SRR kinetics leads to incomplete conversion of the sulfur and “shuttle effect” of the polysulfides. This limits Coulombic efficiency (CE) and cycle-life and is therefore a deterrent to practical application6.

Metal sulfides electrodeposition from soluble polysulfides is the rate-determining step in practical sulfur electrodes7. Duan and co-workers demonstrated a slow conversion of soluble lithium polysulfides into insoluble lithium sulfides through studying the activation energy of various states of SRR in lithium–sulfur (Li–S) batteries7. This leads to accumulation of lithium polysulfides in the electrolyte, and is the primary reason for the shuttle effect, together with a rapid capacity fading8. Despite electronically conductive materials with active electrodeposition sites, such as heteroatom-doped carbon, typically applied as substrates to facilitate charge transfer, an atomic-level understanding of mechanism of metal sulfides electrodeposition in SRR is lacking9,10,11,12. For example, although it is known that Li2S electrodeposition is essential in Li–S batteries, the solid-solid conversion from Li2S2 to Li2S is unclear because it is difficult to distinguish various solid products in the complex process13,14. Moreover, the correlation between the electrodeposition kinetics and geometric/electronic structure of the cathode materials remains unknown15,16,17. Consequently, there is significant research interest in how to realize highly efficient metal sulfides electrodeposition in M–S batteries18.

Although SRR intermediates are too sensitive to be detected in air, advances in in-situ synchrotron characterizations with time resolution permit identification of specific polysulfides and metal sulfides and tracking of dynamic conversion19,20,21. In-depth understanding can therefore be achieved for macroscopic polysulfides conversion kinetics22,23. Nevertheless, atomic-level understanding of metal sulfides electrodeposition behavior is important and remains difficult to achieve experimentally24. Progress in density functional theory (DFT) computations that takes into account the geometric/electronic structure of the sulfur cathode materials is essential in investigating metal sulfides electrodeposition kinetics25. Therefore, combination of advanced in-situ synchrotron characterizations and computational quantum chemistry can reveal critical factors in metal sulfides electrodeposition kinetics26,27. Potential sulfur cathode materials with efficient metal sulfides electrodeposition kinetics can be engineered by tailoring geometric and electronic structures.

Here we present a Mo5N6 cathode material that significantly catalyzes Na2S electrodeposition and results in boosted performance for Na–S battery: 512 mAh g−1 capacity and long cycle life of 10,000 cycles under 1C (1675 mA g−1). Using a judicial combination of synchrotron-based characterizations, electrodeposition rate measurements and DFT computations, we evaluated the Mo5N6 catalysts by linking extrinsic geometric structure with intrinsic reaction energetics in the Na2S electrodeposition. With this research work we shed some light on the origin of high Na2S electrodeposition reactivity and high SRR efficiency of this cathode material.

Results

Atomic and electronic structures of molybdenum nitrides

A series of molybdenum nitrides with varying atomic structure were selected as model cathode materials for investigation of correlation between the atomic structure, electronic structure and electrochemical performance in room-temperature sodium–sulfur (RT Na–S) batteries. Mo5N6, MoN, and Mo2N with varying stoichiometries were synthesized based on reported synthetic strategies28. The crystal phases of the as-prepared molybdenum nitrides were analysed by powder X-ray diffraction (XRD), and indexed to crystalline Mo5N6, MoN, and Mo2N, respectively (Supplementary Fig. 1). The scanning electron microscopy (SEM) images of these molybdenum nitrides confirm that the Mo5N6 and the MoN show a two-dimensional (2D) morphology and that Mo2N exhibits a homogeneous nanoparticle morphology (Supplementary Fig. 2). In addition, high-angle annular dark-field scanning transmission-electron microscopy (HAADF-STEM) imaging was performed to investigate the atomic structure of the molybdenum nitrides (Supplementary Fig. 3). As is shown in Fig. 1a, the Mo atoms are labeled as red-color spheres to aid visualization of the lattice fringes with a lattice distance of 0.24 nm, corresponding to the (2 \(\bar{1}\) 0) and (1 1 0) facets of the Mo5N628. In contrast, the MoN exhibits a lattice distance of 0.25 nm corresponding to the (200) and (2 \(\bar{1}\) 0) facets. Mo2N has two lattice distances of 0.24 nm and 0.21 nm, which are ascribed to the (2 0 0) facet and (1 1 1) facet, respectively (Fig. 1b, c). This observation confirms the crystal structures of the molybdenum nitrides determined by XRD results and demonstrates the different Mo atomic configurations in the molybdenum nitrides.

Fig. 1: Atomic and electronic structural characterization of Mo5N6, MoN and Mo2N.
figure 1

a–c HAADF-STEM images of Mo5N6, MoN, and Mo2N. Insets are schematics of Mo configurations in which red-color, blue, and green spheres represent Mo atoms in Mo5N6, MoN and Mo2N, respectively. d, e Mo 3d XPS and Mo L3, L2-edge NEXAFS spectrum for Mo5N6, MoN, and Mo2N. f Computational model for Mo5N6 (0 0 4). The gray and light-blue spheres represent Mo and N atoms, respectively.

Full size image

It is reported that sulfur redox kinetics is significanty affected by the d electron density of the materials, which regulates atomic structure29. Mo 3d X-ray photoelectron spectra (XPS) measurements showed that the Mo5N6 has a higher dominant Mo valence state of 4+ with a binding energy of 230.0 eV, in contrast, the MoN and Mo2N exhibit lower binding energies of Mo species at 229.2 and 228.9 eV, respectively (Fig. 1d)30. This finding is supported by synchrotron-based near-edge X-ray absorption fine structure (NEXAFS) characterization that permits investigation of the impact on the surface electronic structures of the d electrons. The Mo L-edge white-lines originate from p electron transition to a vacant d electron state31. As is shown in the Mo-L3, L2 edge NEXAFS spectra (Fig. 1e), the intensity of adsorption edge peak decreases in the order Mo5N6, MoN, and Mo2N32. Therefore, the XPS and NEXAFS results demonstrate the high Mo valence state and low d electron density of Mo5N6 resulting from the unique atomic structure as is illustrated in Fig. 1f.

Electrochemical properties of the molybdenum nitrides in the RT Na–S batteries

To investigate electrocatalytic effects of molybdenum nitrides in RT Na–S batteries, we employed MoxNy catalysts as an additive and carbon–sulfur composite containing 62.9 wt% sulfur as active material in assembly of the sulfur electrodes. The as-prepared electrodes were denoted as S/Mo5N6, S/MoN, and S/Mo2N (Supplementary Fig. 4). For comparison, a pure S/C electrode was also prepared. Long-term cycling experiments at a high rate of 1C were conducted to investigate cycling performance of the sulfur electrodes (Fig. 2a and Supplementary Fig. 5). The capacity loss during initial few cycles was observed and is attributed to the formation of stable solid electrolyte interphase (SEI) film because of side effects between carbonated-based solvents and highly reactive sodium anode surface23,33. A high capacity of 186 mAh g−1 was maintained by the S/Mo5N6 after 10,000 continuous cycles under 1 C that refers to an capacity decay of 0.0064% per cycle, together with a stabilized CE held at around 100%. This performance significantly exceeds those of S/MoN (0.014%) and S/Mo2N (0.024%). In contrast, the S/C electrode exhibited a low initial capacity of 201 mA g−1 with a short cycle life of less than 200 cycles (Supplementary Fig. 6). Therefore, the S/Mo5N6 electrode has practical promise for a high-performance Na–S battery with high sulfur content, high capacity, small capacity decay and long cycling life in comparison with many reported sodium polysulfides cathodes, Na2S cathodes, or hybrid of carbon and sulfur cathodes in Na–S batteries (Fig. 2b and Supplementary Fig. 7 and Supplementary Table 1)5,29,33,34,35,36,37,38,39,40,41,42,43,44,45. As is shown in Fig. 2c, S/Mo5N6 exhibited a series of advantageous discharge capacities of 513, 353, 304, 263, and 216 mAh g−1 when cycled at 0.1, 0.2, 0.5, 1, and 2 C (1 C = 1675 mA g−1), respectively. When the specific current was switched back to 0.2 C, a high discharge capacity of 325 mAh g−1 was maintained (Supplementary Fig. 8). Over the following cycling at 0.2 C (Supplementary Figs. 9 and 10), the S/Mo5N6 exhibited a capacity decay of 0.013% per cycle (298–179 mAh g−1 in 2970 cycles). However, S/MoN, S/Mo2N and S/C cathodes show relatively poor cycling and rating performance with limited capacities under high rates (Supplementary Figs. 8–11).

Fig. 2: Electrochemical properties of the molybdenum nitrides in RT Na–S batteries.
figure 2

a Cycling performance and CE for the three sulfur electrodes at 1 C. b Comparison of cycle number and capacity retention of recently reported RT Na–S cells with the present work, in which darker-color refers to greater specific current. c Rating capacities for S/Mo5N6, S/MoN, and S/Mo2N. d Linear relationship between peak current (ip) and square root of scan rate (v1/2) for the three sulfur electrodes. e Energy profile for sodium ion diffusion on Mo5N6 (0 0 4) facet, MoN (0 0 2) facet, and Mo2N (1 0 0) facet.

Full size image

To explore the origin of high CE and stability under high rates of the S/Mo5N6 cathode, we investigated the Na+ diffusion of the three sulfur cathodes through cyclic voltammetry (CV) experiments under various scan rates from 0.1 to 5 mV s−1 in the potential range 0.5–2.8 V46. During the initial cathodic scan (Supplementary Fig. 12) there was a prominent peak corresponding to reduction of elemental sulfur and long-chain soluble polysulfides to less soluble Na2S2 and Na2S. For the anodic scan, one reproducible peak was observed. This corresponds to the oxidation of Na2S2 and Na2S to Na2Sx and elemental sulfur47. The slow oxidation kinetics of Na2S2 and Na2S to Na2Sx are likely the cause for the overlapping of the two oxidation peaks during the anodic scan. Based on the experimentally obtained slope between the peak current (ip) and square root of the scan rate (ν1/2), the diffusion coefficient for sodium ions was estimated from: ip ∝ n3/2AD1/2cv1/2, where ip, n, D, A, c, and v represent, respectively, peak current, number of electrons, diffusion coefficient, surface area of the electrode, concentration of the ion and voltage scanning rate (Fig. 2d)48. Because the number of electrons (n) and concentration of the sodium ion (c) are identical for the three sulfur electrodes, these cancel out in the ip-ν1/2 slope. Based on the electrochemical active surface areas (ECSAs) for the sulfur electrodes (Supplementary Fig. 13) the ip-ν1/2 slopes were normalized as is shown in Supplementary Fig. 14. It was confirmed that the S/Mo5N6 electrode exhibits greater sodium ion diffusion in comparison with that for S/MoN and S/Mo2N electrodes. This likely causes deposition of a thick Na2S layer on the electrode49,50.

To confirm the electrochemical measurements of diffusion rate, we computed the sodium ions diffusion barriers on the three molybdenum nitrides using climbing image nudged-elastic band (CI-NEB) method21. The three models were constructed based on the electron microscopy results, which are Mo5N6 exposing (0 0 4) facet, MoN with (0 0 2) facet, and Mo2N with (1 0 0) facet, as is shown in Supplementary Figure. 1551. The energy profiles for the ions diffusion on the three materials surface are shown in Fig. 2e, and the corresponding diffusion route in Supplementary Fig. 16. The diffusion barrier of sodium ions on the Mo5N6 (0 0 4) facet is 7 meV, which is significantly lower than those on the MoN (0 0 2) facet (65 meV) and Mo2N (1 0 0) facet (118 meV). Although the (0 0 4) facet for Mo5N6 and (0 0 2) facet for MoN can be regarded as the only primary exposed facets because of inherent 2D morphology, we considered other possible facets of Mo2N52,53. Given the 3D-morphology of Mo2N an additional major exposed facet of (1 1 1) was investigated. This showed a significantly greater sodium ion barrier of 246 meV over that for Mo2N (1 0 0) (Supplementary Figs. 17 and 18). Therefore, the computational results are consistent with the electrochemical results and confirm that the S/Mo5N6 electrode exhibits significantly greater sodium ion diffusion in comparison with the S/MoN and S/Mo2N electrodes. These findings demonstrate that the Mo5N6 represents an increased sodium ion diffusion rate and significantly promotes reaction kinetics between sodium and sulfur54.

Kinetic investigations on Na2S electrodeposition

To determine the rate-determining step in overall SRR reaction, SRR kinetics were investigated on the S/C electrode via determination of the energy barrier (Ea) with electrochemical impedance spectra (EIS) measurements and analysis7. The measured EIS curve was fitted with an equivalent circuit as is shown in Fig. 3a, in which Rsurf describes deposition of the adsorbed sodium polysulfides on the surface of the electrode, Rct is the charge transfer process, and the “tail” is the Warburg resistance (Zflw and Zfsw)55,56. The EIS curves for the S/C electrodes under varying voltage were measured at temperatures of 303, 313, and 323 K (Supplementary Fig. 19). By fitting Rct values in the Arrhenius equation, Ea at each voltage was evaluated (Supplementary Table 2). The conversion from S8 to Na2Sx (x = 5–8) at 2.5 and 2.0 V results in low Ea values of 0.63 and 0.57 eV, whilst the conversion following to Na2S4 and/or Na2S3 at 1.5 V exhibits an increased Ea of 0.82 eV (Fig. 3b and Supplementary Table 3). During final conversion to Na2S2/Na2S in the voltage range 1.0 to 0.5 V, Ea increases from 0.79 to 0.89 eV. These findings confirm that conversion of the S8 ring molecules to soluble Na2Sx is relatively facile, whereas the conversion of Na2S4 and/or Na2S3 to final insoluble Na2S2/Na2S is significantly more difficult, making it the rate-determining step for SRR. Similarly, EIS curves for S/Mo5N6, S/MoN, and S/Mo2N electrodes were measured at a temperature of 303, 313, and 323 K (Supplementary Figs. 20–22). As is shown in Fig. 3c, d, Ea for S/Mo5N6 under each of 2.5, 2.0, and 1.5 V, is, respectively, 0.53, 0.57, and 0.60 eV. These values are less than for S/MoN of, respectively, 0.64, 0.69, and 0.79 eV and for S/Mo2N of 0.67, 0.71, and 0.77 eV. Importantly, for the rate-determining step between 1.0 and 0.5 V, S/Mo5N6 exhibits significantly lower values of 0.73 and 0.74 eV in comparison with those for S/MoN of 0.78 and 0.80 eV, and for S/Mo2N of 0.80 and 0.84 eV. These findings evidence significantly boosted overall kinetics and Na2S electrodeposition on Mo5N6 compared with MoN and Mo2N.

Fig. 3: Overall kinetics of sulfur reduction reaction in RT Na–S cells.
figure 3

a Measured and fitted curve for S/C electrode. The measured EIS curve can be fitted with an equivalent circuit as is shown in the inset, in which Rsurf describes deposition of adsorbed sodium polysulfides on the surface of the electrode, Rct is attributed to the charge transfer process, and the tail line represents the Warburg resistance (Zflw and Zfsw) in the pure sulfur cathode. b Activation energy for S/C electrode at varying voltage. The Ea study confirms that conversion of Na2S4 and/or Na2S3 to final insoluble Na2S2/Na2S is the rate-determining step for SRR. The errors originate from the linear fitting of Arrhenius plots for Rct. c Arrhenius-plots for Rct for S/Mo5N6. Values for Rct were obtained by fitting using an equivalent circuit as is shown in the inset of panel a. The error bars represent relative errors of the fitted Rct values. d Activation energy for the three sulfur electrodes at varying voltage. Errors originate from the linear fitting of Arrhenius plots for Rct.

Full size image

To further investigate the Na2S electrodeposition kinetics on Mo5N6, in-situ synchrotron XRD measurements were conducted in transmission mode. A modified 2032-type coin cell with S/Mo5N6 as a cathode was discharged to 1.5 V (vs. Na+/Na) using an in-house design to study the Na2S electrodeposition, Fig. 4a (Supplementary Fig. 23)20. As can be seen from the figure in the discharge from 1.5 V, two peaks at 12.8° and 15.7° are evident in the XRD patterns. These are assigned to soluble Na2S5 (No. 00-027-0792)5,37. The other strong peak at 14.5° corresponds to Na2S4 (No. 04-003-2048). In addition, another peak at 11.9° is assigned to the Na2S3 (No. 00-044-0822)5. This finding reveals that the reduction of elemental sulfur to the mixture of sodium polysulfides occurs from open-circuit voltage to 1.5 V in discharge. This is consistent with the CV findings. When the battery was discharged to ~1.0 V, a peak at 13.6° was observed. This is attributed to the (4 0 4) facet of Na2S (No. 00-047-0178). In the further discharge to ~0.8 V, a peak located at 13.7° is assigned to the (1 0 2) facet of Na2S2 (No. 04-007-3813). In the discharge from 1.5 to 0.5 V, most of the polysulfides in the battery were oxidized to Na2S. Importantly, the peak for Na2S is seen before the appearance of the Na2S2 phase. This finding contrasts with previous report that Na2S2 was formed prior to Na2S57. This unusual phenomenon leads to an assumption that Mo5N6 exhibits very fast conversion kinetics from Na2S2 to Na2S, that result in highly favorable Na2S electrodeposition on the S/Mo5N6 cathode.

Fig. 4: Kinetic investigations of Na2S electrodeposition.
figure 4

a Galvanostatic discharge curve and corresponding in-situ synchrotron XRD patterns for S/Mo5N6 electrode. b–d Potentiostatic discharge curves for electrodeposition rate measurements on the three materials. Insets are the first 400 s in the electrodeposition. e–g SEM images of electrodeposition tests on the three cathode materials.

Full size image

Na2S electrodeposition rate experiments were carried out to investigate electrodeposition kinetics of the three cathode materials. The potentiostatic experiments were designed and modified based on the reported Li2S electrodeposition experiments11,58,59. In order to distinguish the electrodeposition of Na2S from the reduction of higher-order polysulfides (Na2S5) in solution, the cell was discharged galvanostatically to a potential of 1.0 V at a rate of C/24. Subsequently, the electrodeposition was carried out at a fixed potential of 0.5 V (vs. Na/Na+). We present the fitting results via colored areas to demonstrate the capacity contributions of the Na2S electrodeposition in Fig. 4b–d. The capacities of the Na2S electrodeposition on carbon paper (CP)/Mo5N6, CP/MoN, and CP/Mo2N were determined to be 208.3, 147.5, and 104.9 mAh g−1 based on the sulfur mass, respectively. It is seen that Mo5N6 exhibits significantly greater Na2S electrodeposition capacity in comparison with that for MoN and Mo2N. In addition, the highest peak current of 0.80 mA was observed during the electrodeposition on the Mo5N6 as shown in the insets of Fig. 4b–d. The electrodeposition morphology of the Na2S on CP/Mo5N6, CP/MoN, CP/Mo2N, and CP was investigated by SEM imaging and energy-dispersive spectroscopy (EDS) mapping analysis (Fig. 4e–g and Supplementary Figs. 24–27). It was found that Na2S is uniformly deposited on the surface of CP/Mo5N6 with approximately 100% coverage. A three-dimensional (3D) deposition of Na2S layers on the CP/Mo5N6 highlights the high deposition efficiency60. This is attributed to the low energy barriers of Na2S nucleation and growth on Mo5N6. CP/MoN, CP/Mo2N, and CP exhibited an insufficient Na2S deposition with discrete coating as was evidenced by low Na2S coverage on the electrode surface (Supplementary Figs. 25–27). The sluggish Na2S electrodeposition kinetics on CP/MoN and CP/Mo2N are explained by the very high energy barriers for the redox reaction of polysulfides on them. The XPS spectra for the three sulfur electrodes discharged to 0.5 V were analyzed to confirm the electrochemical catalytic activity of Mo5N6. It is seen in Supplementary Fig. 28 that the S/Mo5N6 cathode exhibits dominate S2− species at the end of discharge, underscoring a highly efficient conversion of active sulfur species to S2−61,62. In contrast, S/MoN and S/Mo2N cathodes show only partial conversion to S2−. This finding is attributed to the strong adsorption of sodium polysulfides and low diffusion barrier of sodium ions. The EIS tests were conducted at 0.5 V, and corresponding Rsurf values which describe the deposition of insoluble sodium polysulfides on the electrode surfaces, and fitted to be ~485, ~281, and ~230 Ω cm2 for S/Mo5N6, S/MoN, and S/Mo2N electrodes, respectively (Supplementary Fig. 29). This finding agrees well with those from the Na2S electrodeposition test and the S 2p XPS analysis, confirming the highly efficient Na2S electrodeposition on Mo5N6 in comparison with the others41,63.

Discussion

To determine the origin of the significant deposition efficiency and cycling stability of S/Mo5N6, DFT computations on thermodynamics and kinetics of Na2S electrodeposition on various molybdenum nitride surfaces were carried out. The adsorption configurations and adsorption energies for Na2Sn (n = 1–5) on the three nitrides were investigated. As is shown in the adsorption configurations, Supplementary Figs. 30–32, S in Na2Sn locates in the hollow sites on Mo5N6, whilst Na in Na2Sn locates away from the surface. This implies that Mo functions as a dominant adsorption site to interact with S in Na2Sn64. The adsorption energies for Na2Sn on Mo5N6 were, respectively, −6.14, −8.75, −7.67, −6.36, and −9.23 eV for n = 1, 2, 3, 4, and 5, Supplementary Fig. 33 (Supplementary Table 4). These high values confirm strong adsorption of Na2Sn on Mo5N6 when compared with reported materials in sulfur cathodes for lithium/sodium polysulfides adsorption65. Although adsorption energies for Na2Sn on Mo5N6 were (slightly) greater than those on MoN, it was not clear why there is faster deposition kinetics on Mo5N6 compared with MoN.

To determine the origin of the fast Na2S electrodeposition kinetics on Mo5N6, we constructed a three-step reaction pathway for Na2S2 conversion to Na2S on the three molybdenum nitrides surfaces. The corresponding reaction steps are: 1) adsorption of Na2S2, 2) formation of adsorbed Na* and NaS2* from Na2S2 dissociation, 3) two NaS* formation following simultaneous NaS2* dissociation and a Na-S bond formation. For the Na2S2 dissociation step, Mo5N6 exhibits an optimal free energy (ΔGMo5N6 diss-1) value of −0.23 eV, whilst MoN and Mo2N surfaces exhibit more positive values of ΔGMoN diss-1 = 0.09 eV and ΔGMo2N diss-1 = 0.36 eV, respectively. However, for the NaS2* dissociation step, MoN shows the optimal free energy (ΔGMoN diss-2) value of −0.05 eV, whilst Mo5N6 and Mo2N surfaces exhibit more positive values of ΔGMo5N6 diss-1 = 0.04 eV and ΔGMo2N diss-1 = 0.39 eV, respectively. Therefore, from a thermodynamic point of view, Mo5N6 and MoN demonstrate similar Na2S electrodeposition activity. However, these theoretical investigations based only on adsorption energetics do not agree with the experimental observation that CP/Mo5N6 sample demonstrates a significantly greater Na2S electrodeposition capacity in comparison with CP/MoN. When the kinetics of NaS* formation step is considered (Fig. 5a and Supplementary Table 5), the Mo5N6 surface exhibits a substantially lower energy barrier (ΔGB = 0.48 eV) than on MoN (ΔGB = 0.58 eV) and Mo2N (ΔGB = 1.06 eV) (Fig. 5b and Supplementary Figs. 34–36). Therefore, from a kinetic viewpoint, Mo5N6 demonstrates the most favorable electrodeposition efficiency amongst the three molybdenum nitride structures. With the identification of large energy barrier of NaS* formation on the Mo2N surface, in addition to the formation of the Na2S* state, the NaS* formation kinetics also affect the overall deposition rate and lead to the experimentally observed activity trend (Fig. 5c). Therefore, association of the atomic reaction model (applying ΔGdiss-1 as a reactivity indicator) with the newly considered transition-state theory gives a qualitative confirmation of the deposition kinetics on the three molybdenum nitrides66.

Fig. 5: Computational investigation of Na2S electrodeposition.
figure 5

a Gibbs free energy diagram of conversion from Na2S2 to NaS* on the three surfaces including adsorption of Na2S2, dissociation of Na2S2 to form adsorbed NaS2 (NaS2\({\,\!}^{\ast}\)) and formation of adsorbed NaS (NaS*) from the NaS2*. ΔGdiss-1 is dissociation free energy of Na2S2. ΔGdiss-2 indicates formation free energy of NaS\({\,\!}^{\ast}\), and ΔGB indicates NaS* formation free energy barrier. b Relationship between computed ΔGdiss-1 or ΔGB values and measured electrodeposition capacities on the three molybdenum nitrides surfaces. c Atomic configurations for NaS* formation step on the surface of Mo5N6. The gray-color, light blue, orange, and blue spheres represent Mo, N, S, and Na atoms, respectively.

Full size image

To investigate the origin of fast Na2S electrodeposition kinetics on Mo5N6 from the aspect of electronic structure, spectroscopic measurements were carried out. The ex-situ NEXAFS characterizations were carried out on the three sulfur electrodes to investigate the dynamic change of the valence state of Mo in the three cathode materials. The Mo L3-edge spectra for Mo5N6 (Fig. 6a) show that the valence state of Mo decreases gradually with the discharge and reaches lowest level at the discharge potential of 0.5 V. This is consistent with the adsorption configuration from DFT results and demonstrates the strong Mo–S interaction and high Na2S electrodeposition capacity. During the following charge, the valence state of Mo increases to the original level until the end of charge at the potential of 2.8 V. This finding confirms the strong adsorption of sodium polysulfides and the high reversibility of Mo5N6 as electrode materials. As a comparison, similar measurements were performed on electrode with MoN. The NEXAFS spectra show that valence state of Mo decreases during discharge but remains unchanged during the following charge. This indicates that although MoN exhibits strong adsorption of sodium polysulfides, the high diffusion barrier of sodium ions on MoN leads to poor reversibility of the electrode. In contrast, the Mo2N electrode shows a nearly unchanged valence state of Mo during the whole discharge/charge. These findings confirm that Na2S electrodeposition kinetics on the molybdenum nitrides strongly depends on the valence state and d electron density of Mo.

Fig. 6: Analyzes of Mo electronic structure in corresponding reaction transition state.
figure 6

a–c Ex-situ NEXAFS for the Mo L3-edge of S/Mo5N6, S/MoN, and S/Mo2N, for varying voltage during a discharge/charge cycle. b–d Charge difference analyzes from configurations of the transition states in the NaS\({\,\!}^{\ast}\) formation step on the three cathode materials, in which yellow-color and cyan iso-surface represent electron accumulation and electron depletion, respectively, and the iso-surface value is 0.0015 e Å−3. Color code is the same as for Fig. 5c. e Energy level diagram showing orbital hybridization for adsorption sites and adsorbate. EF is the Fermi level of the substrate; σ and σ* indicate bonding and anti-bonding states, respectively. f Solid lines represent DOS for S in the red-color circles in Fig. 6b–d. The DOS is projected onto the S 3p state. Dashed lines represent surface Mo d-bands DOS of the three clean molybdenum nitrides surfaces.

Full size image

To gain further fundamental insight into the low NaS2* dissociation barrier on Mo5N6, charge difference analyses on the transition state configurations on three surfaces were carried out (Fig. 6b–d). On MoN and Mo2N surface, the interaction between the S (highlighted by red-color circles in Fig. 6c–d) and Na is strong, which represents electron accumulation between the two atoms. On the Mo5N6 surface, less electron accumulation is observed. This finding confirms a weaker interaction between the two atoms, which aids Na2S* dissociation with a lower dissociation barrier. The weaker interaction between Na and S is attributed to the strong Mo–S interaction, also demonstrated in the NEXAFS and XPS findings. The relationship between the adsorption of an adsorbate on a surface and the electronic structure of the substrate is explained by the density of states (DOS) (Fig. 6e). When a polysulfide molecule from the electrolyte is adsorbed on the molybdenum nitrides surface to form Na2Sn*, the electronic states of the Mo interact with those of sulfur. Consequently, the hybridized energy levels split into two groups: one is the anti-bonding states (σ*) that normally go across the Fermi level (EF); the other is the bonding orbital (σ) positioned under the EF. The difference in the adsorption strength comes from the antibonding states, that is, with a higher location of the EF of the molybdenum nitrides, the antibonding states move to lower occupancy. This leads to a stronger interaction between Na2Sn* and the molybdenum nitrides surface, and vice versa67. In this work, the results of the DOS computations are consistent with the scheme, Fig. 6f. The low DOS peaks of the S 3p-Mo d antibonding orbitals (indicated by arrows) cause weak adsorption of S on MoN and Mo2N. More importantly, the position of antibonding orbital is decided by the d-band position of Mo on the molybdenum nitrides; the d-band centers for the Mo atoms on the Mo5N6, MoN, and Mo2N are −0.41, −1.56, and −2.41 eV, respectively. This is the same order as for the S–Mo antibonding peaks on the three surfaces. Therefore, the d-band position shows similar trends with the adsorption strength as is shown in Supplementary Fig. 37. The d-band center (applying as a descriptor) can be correlated with the Na2S2* dissociation energy as the underlying mechanism of the better electrodeposition reactivity of Mo5N6 in sulfur cathodes.

Using sodium-sulfur chemistry as an example, we correlated the Na2S electrodeposition reactivity on a Mo5N6 electrocatalyst with its reaction energetics and inherent electronic structure. By identifying the significant influence of Na2S2 dissociation on the overall SRR reactivity of various molybdenum nitrides surfaces, we elucidated the mechanism of metal sulfides electrodeposition through the association of atomic reaction model with the transition-state theory. The combination of experimental data and theoretical computations demonstrates that Mo5N6 with favorable d-band position delivers significantly high Na2S electrodeposition reactivity and performance in Na-S battery. This advance in mechanistic understanding of metal sulfides electrodeposition will underpin rational design of efficient M–S batteries. Application has resulted in significant performance of an RT Na–S battery. Findings will be of immediate interest and practical benefit to a wide range of researchers in the rational design of electrode materials for accelerated applications in sustainable energy-storage and conversion.

Methods

Preparation of molybdenum nitrides

Mo powder (<150 μm, 99.99% trace metals basis) was purchased from Sigma-Aldrich without further purification. The 2D Mo5N6 nanosheets were synthesized through a Ni-induced salt-templated method as previously reported: 0.4 g of Mo powder was dispersed in 40 mL of ethanol with magnetic stirring for 10 min28. 1.2 mL of H2O2 (30 wt%) solution was injected dropwise into the suspension. Following stirring for 12 h at room temperature, the solution turned into a dark-blue color. Separately, 10 mg of Ni(OCOCH3)2·4H2O was dissolved in 10 mL of ethanol and mixed with the dark-blue suspension to form the precursor. The precursor solution was mixed with 640 g of NaCl powder and dried at 50 °C with continuous hand-stirring. The mixture was annealed at 750 °C for 5 h at a heating rate of 1 °C min−1 under a 5% NH3/Ar atmosphere. The product was washed with deionized water and dilute hydrochloric acid several times to remove the NaCl template and Ni nanoparticles before being dried using vacuum filtration. The 2D MoN nanosheets were synthesized without addition of Ni. The Mo precursor was mixed with 640 g of NaCl and annealed at 750 °C for 5 h. The final product was obtained by removing NaCl using deionized water and vacuum filtration. The Mo2N nanoparticles were produced by annealing Mo powder under NH3 atmosphere. Fifty milligram of Mo powder was put into a porcelain boat uniformly. The powder was annealed at 650 °C for 5 h at the ramp rate of 1 °C min−1 under 5% NH3/Ar atmosphere.

Materials characterization

The morphology and structure of samples was characterized by SEM (FEI Quanta 450). HAADF-STEM images were recorded at 200 kV (Talos F200X). XRD data were recorded on a Rigaku MiniFlex 600 X-Ray Diffractometer. Sulfur content of the active material was determined by TGA (METTLER TOLEDO TGA/DSC 2) under N2. In-situ synchrotron XRD (with wavelength λ = 0.6888 Å) and NEXAFS data were detected on the powder diffraction and the soft X-ray spectroscopy beamline in the Australian Synchrotron, Clayton, Victoria.

Electrochemical characterization

For the battery performance measurement, active sulfur material containing elemental sulfur and conductive carbon (Ketjen Balck) with mass ratio of 2:1 was well-mixed and sealed in a quartz ampoule and thermally treated at 300 °C for 2 h under a N2 atmosphere. 5 wt% MoxNy was used as the additive in the slurry together with 80 wt% of active material, 5 wt% of conductive carbon and 10 wt% N-lauryl acrylate (LA133, purchased from Chengdu Yindile Power Supply Technology). For comparison, a pure S/C electrode was prepared with a slurry containing 80 wt% active material, 10 wt% conductive carbon and 10 wt% N-lauryl acrylate. The slurry mixture was cast on aluminum-foil and dried at 50 °C overnight to fabricate the sulfur electrodes with average thickness of ~20 μm. The 2032-type coin cells were assembled using glass-fiber as the separator and Na metal with average thickness of ~500 μm and purity of 99.9% as the anode. The electrolyte consisted of 1.0 M NaClO4 in ethylene carbonate (EC)/propylene carbonate (PC) with a volume ratio of 1:1 and 5 wt% fluoroethylene carbonate (FEC) additive. The volume of electrolyte injected into the coin cells was controlled to 15 μL in total. The areal active material loading in the cathode for rating and cycling performance was ~1.2 mg cm−2. The area of the sulfur electrodes are ~1.13 cm−2. The galvanostatic charge/discharge measurements were performed under 30 °C in a constant temperature oven using NEWARE and LAND CT2001A battery testers. The capacities were calculated based on the mass of the elemental sulfur. For all the cycling and rating tests, a 0.1 C low-rate cycle were adopted to activate the redox reactivity of the elemental sulfur. Electrochemical impedance spectroscopy tests were performed at specific voltages in the frequency range 1 MHz to 0.01 Hz with an amplitude of 5 mV after resting for 10 min. A thermal test chamber (MSK-TE906) was used to control the temperature during the EIS testing. To stabilize voltage, the battery was discharged to the particular potential and held at that potential until the output current remained constant. For in-situ synchrotron XRD measurements, the in-situ cells were similar to the coin cells for electrochemical performance testing and were discharged to 1.5 V under 0.1 C prior to measurement of Na2S electrodeposition. To guarantee that the X-ray beams could penetrate the whole cell, three holes with 4 mm diameter were punched in the center of the negative, the positive battery shells and the spacer. Kapton film was used to cover the holes in the negative and positive caps, and glue was used for sealing. The assembly of the in-situ cells was the same as that for the cells for the electrochemical tests. For the ex-situ measurements, the cells with test electrodes were discharged or charged to the specific voltage under 0.1 C, then were dissembled in a glove box. The test electrodes were sealed in a sample-holder within the glove box to transport the electrode samples.

Na2S electrodeposition experiments

The Na2S (>97.0%), sulfur powder (reagent grade, 100 mesh particle size), tetraglyme (>99%) and glass fiber separator (Whatman® glass microfiber filter, Grade GF/F) were purchased from Sigma-Aldrich Australia. The Na2S powder was stored in glove box. The carbon papers (CF) were purchased from Shanghai Hesen Electronics Co. Ltd. (HCP010N). The original Li2S electrodeposition tests are referred to for the Na2S electrodeposition tests11. The Na2S electrodeposition from soluble polysulfides was studied by potentiostatic deposition in Na2S4 tetraglyme solution on the CF current collectors. The Na2S4 tetraglyme solution was applied to compare the delicate kinetical differences for different samples with corresponding electrodeposition films and capacities. The test was conducted in a similar coin cell as for the electrochemical performance tests. The Na2S4 in tetraglyme solution added to the MoxNy on CF was the only source of sulfur active material. The 0.1 M Na2S4 solution was prepared by dissolving and mixing stoichiometric amounts of Na2S and sulfur in tetraglyme solvent at room temperature for 10 h. CF papers were punched into disks with a diameter of 12 mm and about 0.40 mg of Mo5N6, MoN, and Mo2N powders were separately dispersed on CF papers using pure ethanol as solvent. 25 µL Na2S4 was dropped onto the as-prepared current collectors as cathode. Sodium-foil was employed at the counter electrode, which was separated with cathode by glassfiber membrane and dropped with 15 µL electrolyte. The cells were galvanostatically discharged to 1.0 V at a constant specific current of C/24, and kept potentiostatically at 0.5 V for Na2S to nucleate and grow until the current dropped below 10−5 A.

Computational methods

DFT computations were carried out using the Vienna Ab-initio Simulation Package (VASP)68,69. The exchange-correlation interaction was described by generalized gradient approximation (GGA) with the Perdew–Burke–Ernzerhof (PBE) functional70. The DFT-TS method of Grimme was employed to treat the VDW interaction71. All computations were carried out using a plane wave kinetic energy cut-off of 600 eV. All structures in the computations were spin-polarized and relaxed until the convergence tolerance of force on each atom was less than 0.01 eV. The energy convergence criteria were set to 10−4 eV for self-consistent computations with a Gamma centered 2 × 2 × 1 K-points. All periodic slabs had a vacuum spacing of at least 15 Å. Denser 10 × 10 × 1 K-points were used for the density of states (DOS) computations. All periodic slabs had a vacuum spacing of at least 15 Å. The structural model for Mo5N6 (0 0 4) facet contained five Mo-N layers with a supercell size of a = b = 11.44 Å, c = 20.60 Å, α = β = 90°, and γ = 120°. The structural model for MoN (0 0 2) facet contains five Mo–N layers with a supercell size of a = b = 11.36 Å, c = 20.70 Å, α = β = 90° and γ = 120° while the Mo2N (1 0 0) facet consisted of five Mo–N layers with a = b = 8.33 Å, c = 23.33 Å, α = β = γ = 90°. In computations the three bottom layers were kept fixed, all other atoms were allowed to relax. Na2Sn (n = 1–5) adsorption energies ΔE for each configuration were computed from:

$$\varDelta E={E}_{{{{{{\mathrm{total}}}}}}}-{E}_{{{{{{\mathrm{Na2Sn}}}}}}}-{E}_{{{{{{\mathrm{s}}}}}}}$$
(1)

where Etotal, ENa2Sn, and Es are, respectively, the energies for the whole system, Na2Sn and substrate.

Free energy diagram computation

Based on the experimentally obtained rate-determining step, the solid-solid conversion from Na2S2 to Na2S was proposed as three elementary steps, namely:

$$\ast {{{{{{\rm{Na}}}}}}}_{2}{{{{{{\rm{S}}}}}}}_{2}+2{{{{{{\rm{Na}}}}}}}^{+}+2{{{{{{\rm{e}}}}}}}^{-}$$
(2a)
$$\ast {{{{{{\rm{NaS}}}}}}}_{2}+\ast {{{{{\rm{Na}}}}}}+2{{{{{{\rm{Na}}}}}}}^{+}+2{{{{{{\rm{e}}}}}}}^{-}$$
(2b)
$$2\ast {{{{{\rm{NaS}}}}}}+2{{{{{{\rm{Na}}}}}}}^{+}+2{{{{{{\rm{e}}}}}}}^{-}$$
(2c)

in which * indicates a reaction site. The free energies for the three steps are ΔGdiss-1 and ΔGdiss-2, respectively, and can be computed from:

$$\varDelta {{{{{{\rm{G}}}}}}}_{{{{{{\rm{diss}}}}}}-1}={{{{{\rm{G}}}}}}(\ast {{{{{{\rm{NaS}}}}}}}_{2}+\ast {{{{{\rm{Na}}}}}})-{{{{{\rm{G}}}}}}(\ast {{{{{{\rm{Na}}}}}}}_{2}{{{{{{\rm{S}}}}}}}_{2})$$
(3)
$$\varDelta {{{{{{\rm{G}}}}}}}_{{{{{{\rm{diss}}}}}}-2}={{{{{\rm{G}}}}}}(2\ast {{{{{\rm{NaS}}}}}})-{{{{{\rm{G}}}}}}(\ast {{{{{{\rm{NaS}}}}}}}_{2}+\ast {{{{{\rm{Na}}}}}})$$
(4)

in which

$${{G}}={{E}}+{{{E}}}_{{{{{{\rm{ZPE}}}}}}}-{{{{{\rm{TS}}}}}}$$
(5)

and where zero point energy corrections (EZPE) and entropic contributions (TS; T was set to be 273.15 K). The free energy of the ΔGdiss-1 and ΔGdiss-2 were computed as:

$$\Delta {{G}}=\Delta {{E}}+\Delta {{{E}}}_{{{{{{\rm{ZPE}}}}}}}-{{T}}\Delta {{S}},$$
(6)

where ΔEZPE and ΔS are the binding energy, zero point energy change and entropy change, respectively. In this work the values of ΔEZPE and ΔS on the specific molybdenum nitrides surfaces were determined by vibrational frequency computation. Note that the exploration of active sites on the specific surface was conducted and chosen according to the most energetically stable adsorption site.

Na2S2 dissociation barrier computation

Following identification of the initial and final states for Na2S2 dissociation step on the three molybdenum nitrides surfaces, the energy barrier was located via searching for transition states by climbing image nudged-elastic band (CI-NEB) method implemented in VASP21. The transition states were obtained by relaxing the force below 0.05 eV Å−1. The located transition states were confirmed by frequency analysis. The d-band center (\(\varepsilon\)d) was determined as the weighted DOS center of d-band as:

$${\varepsilon }_{{{{{{\rm{d}}}}}}}={\sum }_{{{{{{\mathrm{i}}}}}}}{\varepsilon }_{{{{{{\mathrm{i}}}}}}}{{{{{{\rm{r}}}}}}}_{{{{{{\mathrm{i}}}}}}}/{\sum }_{{{{{{\mathrm{i}}}}}}}{\varepsilon }_{{{{{{\mathrm{i}}}}}}}$$
(7)

where ri is the DOS at energy \(\varepsilon\)i.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.