High‐rate sodium‐ion storage of vanadium nitride via surface‐redox pseudocapacitance

Vanadium nitride (VN) electrode displays high‐rate, pseudocapacitive responses in aqueous electrolytes, however, it remains largely unclear in nonaqueous, Na+‐based electrolytes. The traditional view supposes a conversion‐type mechanism for Na+ storage in VN anodes but does not explain the phenomena of their size‐dependent specific capacities and underlying causes of pseudocapacitive charge storage behaviors. Herein, we insightfully reveal the VN anode exhibits a surface‐redox pseudocapacitive mechanism in nonaqueous, Na+‐based electrolytes, as demonstrated by kinetics analysis, experimental observations, and first‐principles calculations. Through ex situ X‐ray photoelectron spectroscopy and semiquantitative analyses, the Na+ storage is characterized by redox reactions occurring with the V5+/V4+ to V3+ at the surface of VN particles, which is different from the well‐known conversion reaction mechanism. The pseudocapacitive performance is enhanced through nanoarchitecture design via oxidized vanadium states at the surface. The optimized VN‐10 nm anode delivers a sodium‐ion storage capability of 106 mAh g−1 at the high specific current of 20 A g−1, and excellent cycling performance of 5000 cycles with negligible capacity losses. This work demonstrates the emerging opportunities of utilizing pseudocapacitive charge storage for realizing high‐rate sodium‐ion storage applications.


| INTRODUCTION
Electrochemical energy storage (EES) devices play an indispensable role in our daily lives due to their widespread applications in portable electronics, the growing markets for electric vehicles (EVs), and largescale energy storage associated with renewable energy sources. [1][2][3][4][5][6] Li-and Na-ion batteries offer high energy density from reversible redox reactions (100-300 Wh kg −1 ) but relatively low power density because of slow ionic diffusion. [7][8][9][10][11] Electrochemical double-layer capacitors (EDLCs) do not have these diffusion considerations and can offer higher power than batteries but their low energy density (5-10 Wh kg −1 ) limits their applications. [12,13] The distinction between batteries and EDLCs originates from the different storage mechanisms: diffusioncontrolled faradaic reactions and fast nonfaradaic electrostatic ion adsorption, respectively. [13] In recent years, there has been a growing interest in pseudocapacitive materials which are characterized by having a faradaic charge transfer process take place at or near the surface of active materials. [13][14][15] This capacitive energy storage mechanism is regarded as a promising approach for combining the high energy densities from redox reactions with the high-power density that results from capacitor-like kinetics. [2,7] When prepared as nanoscale materials, a number of transition metal oxides have demonstrated this combination of highpower density and high energy density. [16][17][18] Compared with most transition metal oxides and dichalcogenides, the transition metal nitrides have much higher electronic conductivity (e.g., 10 2 -10 4 S cm −1 for vanadium nitride [VN]) which is very attractive for highrate EES applications. [19][20][21][22][23][24] The high electrical conductivity is also of interest because a separate current collector may not be necessary for the electrode. Among the various transition metal nitrides, VN has received the most interest and some very impressive energy storage results have been reported. [22][23][24][25][26][27][28][29] However, the mechanism of charge storage in VN and other nitrides is not well understood. For all the VN materials reported to date, the cyclic voltammetry (CV) curves exhibit a rectangular shape indicating capacitive storage with no evidence of pronounced redox peaks. [23][24][25][26][27][28][29][30][31] The energy storage mechanism, however, does not seem to be due to double-layer capacitance as the magnitude of energy storage shows a significant variation with sweep rate and the area-normalized capacitance is always much greater than the 20-30 μF cm −2 which is the typical value for double-layer storage in carbon-based electrodes. [13,32] The initial studies carried out by Choi et al. on VN in 1 M KOH electrolyte proposed that the high values of specific capacitance (>1300 F g −1 and >500 F cm −2 at 100 mV s −1 ) were due to the presence of surface layers of vanadium oxides or oxynitrides which could undergo redox reactions to account for the high level of charge storage. [29] X-ray photoelectron spectroscopy (XPS) studies tend to support this mechanism as V 5+ on the surface changes to a lower oxidation state after electrochemical measurements. With Na 2 SO 4 electrolyte, it was proposed that adsorption and desorption of Na + led to faradaic reactions with vanadium oxynitrides at the surface. [30] It is important to note that the vanadium oxides that form on the surface are soluble in aqueous solutions and this has led to poor cycling performance. [27,29] By limiting the amount of water in the electrolyte with either a high concentration of LiCl or the use of a LiCl/polyvinyl alcohol gel electrolyte, it is possible to stabilize the oxides on the VN surface. This approach leads to the retention of high levels of charge storage over extended cycling. [27] Charge storage of VN electrodes has also been reported using nonaqueous electrolytes. [26,[33][34][35] Here, too, rectangular CVs for VN are reported. The charge storage mechanism, however, is still unclear. Ex situ Xray powder diffraction (XRD) results of VN in nonaqueous electrolyte exhibit virtually no change in lattice parameter even in the fully lithiated state, [35] indicating there might be no intercalation occurring within the bulk of the material. There is a possibility that surface oxides may be occurring in the nonaqueous electrolyte studies as XPS measurements have shown the presence of not only V 3+ but also V 4+ and V 5+ along with O 1s peaks. [33,34] One fact that complicates the results for the nonaqueous electrolyte studies is that the VN materials, whether as quantum dots, [33] nanoparticles, [35] or nanowires, [22] are synthesized on graphene supports. When electrochemical measurements for these VN/graphene electrodes occur between 3.0 and 0.01 V (vs. Li + /Li), we can expect the graphene to exhibit energy storage and contribute to the capacity of the electrode. For example, the Li + capacity for graphene between 3.0 and 0.01 V at 0.1 A g −1 is over 400 mAh g −1 . [33,36] As a result, it is difficult to not only determine the charge storage mechanism for VN but also to assess quantitative values.
The present work focuses on the question of the origin of sodium-ion storage in VN materials using nonaqueous electrolytes. In keeping with the use of nanoscale forms of VN, we prepared the VN as mesoporous nanosheets. In contrast to prior studies, these materials were prepared as powders without graphene or carbon supports. Moreover, we investigated the sodium-ion storage properties of VN, which may emerge as an interesting negative electrode. In this study, we are able to show that the vanadium oxidation state at the surface and the accompanying redox reactions determine the energy storage properties of the VN anode.

| Synthesis of VN mesoporous nanosheets
A schematic of the process used to form VN mesoporous nanosheets is shown in Figure 1A. Thin V 2 O 5 ·nH 2 O xerogels [37] (Figures 1B and S1a) are used as precursors and are transformed into VN mesoporous nanosheets by heat treatment in an ammonia flow. Generally, the transformation proceeds in two steps. In the first step, water molecules evaporate from the xerogels, creating pores in the VO x nanosheets. The second step is the nitridation process: the oxygen atoms are replaced by nitrogen atoms, leading to the formation of VN. [19,20] During these processes, the nanosheet morphology is retained and VN nanocrystallites are interconnected with each other to form the mesopores (Figures 1C,D and S1b). This processing step includes both nitrogen and oxygen diffusion, which is highly influenced by the treatment conditions and mass transport kinetics. [38] In this way, it is possible to tune the nitrogen content (and the corresponding oxygen content) and grain size by varying the temperature. The mesoporous nanosheet samples described in this paper were annealed in flowing NH 3 at 550°C ( Figure 1C,D) and 700°C ( Figure S2). As a control, VN particles ( Figure S3) were synthesized by annealing commercial V 2 O 5 powders in flowing NH 3 at 900°C.

| Structure and chemistry of VN samples
The phase of the VN samples was examined by XRD as shown in Figure 2A. All the diffraction peaks are indexed to cubic VN (JCPDS card No. 73-0528), indicating a completed phase transformation. On the basis of the Scherrer formula with a spherical shape factor, the crystallite sizes of samples annealed at 550°C and 700°C are calculated to be 10 and 15 nm. In the work reported here, these materials are identified as VN-10 nm and VN-15 nm, respectively. The VN particles prepared by annealing V 2 O 5 powders in NH 3 have a crystallite size of~50 nm (noted as VN-50 nm). The diffraction peaks shift to slightly lower 2θ values with increasing annealing temperature ( Figure S4), corresponding to a subtle increase in lattice parameter (Table S1). [38] XPS was carried out to identify the surface chemical composition and the valence state of the annealed VN F I G U R E 1 Schematic of the synthesis and nanostructure of interconnected VN mesoporous nanosheets. (A) The transformation of VN mesoporous nanosheets includes two steps: the loss of crystal water followed by the nitridation process. SEM images of vanadium oxide xerogels nanosheets (B) and VN-10 nm mesoporous nanosheets (C, D). SEM, scanning electron microscope; VN, vanadium nitride. samples ( Figure 2B). The V 2p states have two broad peaks which are deconvoluted into three different peaks. The deconvoluted V 2p 3/2 binding energy of 514.2 eV is consistent with the typically reported values for V 3+ (V─N), while the two peaks at 515.7 and 517.2 eV are associated with V 4+ and V 5+ in oxidized vanadium (V─N─O and V─O). [27] The detailed calculations for each V valence are listed in Table S2. It is found that the relative valence state and oxygen content of vanadium in the VN-10 nm sample are larger than those of VN-15 nm and VN-50 nm. The nitrogen sorption isotherms of the VN samples show typical type-IV hysteresis curves ( Figure 2C) with mesoporous distributions ( Figure S5). The Brunauer-Emmet-Teller specific surface areas (SSAs) for VN-10 nm, VN-15 nm, and VN-50 nm are 50.7, 30.4, and 10.0 m 2 g −1 , respectively.
Transmission electron microscopy (TEM) images ( Figure 2D) indicate the sheet-like nature of VN-10 nm, while at a higher magnification, the interconnected nanoscale crystallites are evident ( Figure 2E). Highresolution TEM (HRTEM) image ( Figure 2F) shows lattice spacing of 2.361 and 2.042 Å, which correspond to the (111) and (200) planes of cubic VN, respectively, consistent with the selected area electron diffraction (SAED) patterns (inset of Figure 2E). The particle sizes observed by electron microscopy are larger than those calculated values from XRD data, which is due to the interconnected nanoparticles. TEM energy dispersive X-ray spectroscopy mapping ( Figure 2G, Figure S2g, and Table S3) was used to determine the overall composition of the VN-10 nm, VN-15 nm, and VN-50 nm, which are determined to be   [30,39] These data indicate that there is a significantly higher oxygen content at the surface, leading to a higher vanadium valence state.

| Sodium-ion storage properties
The electrochemical properties of the VN materials were determined using half-cells (2016-type coin cell) with metallic sodium as the reference and counter electrode. The Na + capacity of the acetylene black added to the electrode was determined in control experiments ( Figure S6). This contribution was subtracted from the total electrode capacity to give the charge storage properties of VN, which are the values presented in this work. The initial galvanostatic sodiation process ( Figure S7) displays a small plateau at~1.0 V and irreversible capacity loss which is attributed to the formation of solid electrolyte interphase (SEI) layers. [40][41][42] On successive cycles, the desodiation curve exhibits slight curvature while the sodiation curve is linear ( Figure 3A). After 80 cycles at 0.1 A g −1 (Figure 3B), the VN-10 nm anode delivers a capacity of 165 mAh g −1 , compared with the VN-15 nm (81 mAh g −1 ) and VN-50 nm (10 mAh g −1 ).
The rate capability for VN-10 nm is also better than that of VN-15 nm and VN-50 nm with a capacity of 106 mAh g −1 at 20 A g −1 ( Figure 3C). In addition, during 5000 cycles at 1 A g −1 (Figure 3D), the specific capacity of the VN-10 nm anode is stable and slightly increased to 144 mAh g −1 with a coulombic efficiency of nearly 100%. The combination of high capacity at a high rate coupled with long cycling behavior makes VN-10 nm a very promising electrode material for high-power sodium-ion storage devices. [2,12] CV measurements were used to complement the galvanostatic results. The CV curves ( Figure 4A) display a rectangular shape over the potential range (0.05-3.0 V vs. Na + /Na), which is typical of capacitor-like behavior. [7,13,43] At 1 mV s −1 , the VN-10 nm exhibits a higher capacity compared with VN-15 nm and VN-50 nm. At higher sweep rates, CV curves for VN-10 nm are not nearly as rectangular ( Figure S8) and suggest that charge storage kinetics are slower than that of a double-layer capacitor.
To provide more insight regarding the charge storage process, we use Equation (1) to separate the measured current (i) at a given potential (V) into capacitor-like effects (k 1 ν) and diffusion-controlled contributions (k 2 ν 0.5 ) [7] : The rectangular, cross-hatched portion in Figure 4B is associated with the capacitive contribution for VN-10 nm. This curve, which is analyzed at the relatively slow sweep rate of 1 mV s −1 to maximize the diffusion contributions, indicates that capacitor-like processes dominate the response with some 72% of the charge storage. Meanwhile, the VN-15 nm and VN-50 nm anodes show the dominated capacitor-like process as well ( Figure S9). The next step is to identify whether the charge storage process arises primarily from double-layer or pseudocapacitive mechanisms. For the former, we use an electrolyte, 1.0 M tetrabutylammonium perchlorate (TBAClO 4 ) in propylene carbonate, whose bulky tetrabutylammonium (TBA + ) cation is much too large to be intercalated. [44] Thus, this electrolyte provides a measure of the double-layer capacitance for VN-10 nm. As shown in Figure S10, the amount of charge storage is far less than when the sodium electrolyte, NaClO 4 in propylene carbonate, is used. The amount of charge storage with TBA + is~43 C g −1 (~12 mAh g −1 ), which, when we consider the SSA of VN-10 nm (50.7 m 2 g −1 ), gives us a specific capacitance of 28.8 μF cm −2 . This value agrees well with what is expected for the surface-normalized double-layer capacitance. [45] In Table S1, we compare the amount of charge storage for the three materials which occur when the nonaqueous Na + electrolyte is used. The surface-area-normalized value of~428 μF cm −2 for VN-10 nm shows that most of the charge storage in this material can be attributed to a pseudocapacitance. Charge storage in VN-15 nm is also dominated by pseudocapacitance as the surface-area-normalized capacitance is~254 μF cm −2 , well above that associated with the electrical double layer. For the VN-50 nm material (162 μF cm −2 ), the double-layer capacitance provides a greater fraction of charge storage but, here too, pseudocapacitive processes are significantly greater. As discussed in the following section, the differences in the amount of charge storage and normalized capacitance among these samples are related to the concentration of surface V 5+ /V 4+ at the surfaces of VN grains ( Figure 2B and Table S2). The small-sized VN sample with higher valances of surficial vanadium oxides delivers high capacity/capacitance.
Ex situ XRD, TEM, and XPS were used to characterize the chemical and structural features associated with Na + storage. As shown in Figure 5A for VN-10 nm, there is no phase change during the sodiation/desodiation processes. The peaks corresponding to the (111), (200), (220), and (311) planes of cubic VN are effectively unchanged for 1000 cycles. Moreover, there is barely any shift in two thetas for any of the peaks. For example, the lattice parameter determined for the (200) shows a slight increase from 4.087 to 4.091 Å after sodiation. Ex situ TEM at a sodiated state was further tested ( Figure S11). The VN nanocrystals are still interconnected with each other with the sheet morphology remaining, indicating stable structure and negligible volume changes during the reaction ( Figure S11a). The SAED pattern ( Figure S11b) further confirms that the polycrystalline feature of the cubic VN phase after sodiation. The HRTEM image ( Figure S11c) shows clearly lattice fringes with d-spacing of~2.402 and 2.013 Å, corresponding to (111) and (200) planes, respectively. The clear ex situ TEM result demonstrates the VN nanocrystals still remain the origin crystal structure, which is a supplement of the ex situ XRD result.
There is a change in V 3+ , V 4+ , and V 5+ contents as determined by ex situ XPS measurement for the sodiation and desodiation states ( Figure 5B, Figure S12, and Table S5). After being sodiated to 0.05 V at 0.1 A g −1 , the amount of V 5+ decreases while both the V 4+ and V 3+ increase upon sodiation. This suggests that the V 5+ is reduced to both V 4+ and V 3+ , which accounts for the increase in both of these species. On the basis of this simple assumption, it is possible to compute the amount of charge storage associated with the valence changes occurring in VN-10 nm. As described in Supporting Information Section 3, a 1-electron transfer (V 5+ to V 4+ ) in combination with a 2-electron transfer (V 5+ to V 3+ ) corresponds to 0.41 Na + per formula unit and leads to a capacity of~170 mAh g −1 . This calculated value is somewhat less than that obtained in separate galvanostatic experiments for VN-10 nm samples which exhibit a charge storage of~190 mAh g −1 at 0.1 A g −1 after five cycles ( Figure 3B). The fact that the galvanostatic result is greater than that calculated from the XPS is not surprising as it suggests that redox reactions occur over a somewhat thicker region than that monitored in the XPS experiment and a small amount of nonfaradaic doublelayer capacity (~12 mAh g −1 , Figure S10) contributes to the total charge storage. Upon desodiating VN-10 nm anode to 3.0 V, the V 5+ content increases while that of V 4+ and V 3+ decreases as Na + is removed from the electrode ( Figure 5B and Table S5). The number of Na + per formula unit, 0.34, is less than that of the desodiated reaction, however, this is consistent with the capacity decrease observed over the initial five cycles ( Figure S13).
For the VN-15 nm sample, there is significantly less V 5+ than with VN-10 nm and for that reason, we expect less charge storage to occur. Our efforts at estimating the charge storage for this material are discussed in Supporting Information Section 3. Nonetheless, this semiquantitative approach indicates that the Na + storage is determined by redox reactions occurring with the V 5+ at the surface. The smaller particle size of VN-10 nm with its higher percentage of accessible reactive surface sites delivers higher specific capacity compared with VN-15 nm and VN-50 nm. The absence of any significant change in lattice parameter is consistent with the notion that redox reactions associated with Na + are confined to the surface. It is interesting to note that density functional theory (DFT) calculations (Supporting Information Section 4, Figure S14) are in general agreement with the conclusion that Na + charge storage with VN arises from the surface and not bulk processes. These DFT calculations were directed at the question of whether the cubic VN crystal is able to be a host F I G U R E 5 Sodium-ion storage mechanism for VN. (A) Ex situ XRD patterns for VN-10 nm at fully sodiated and desodiated states after five cycles, and fully desodiated state after 1000 cycles. There is no phase change and only a minimal change in lattice parameter is evident. (B) V 2p XPS spectra of pristine, fully sodiated, and desodiated states of VN-10 nm after five cycles. (C) Schematic of the pseudocapacitive sodium-ion storage mechanism for VN based on having redox reactions occur at the surface or near the surface from the presence of sodium-ion. VN, vanadium nitride; XPS, X-ray photoelectron spectroscopy; XRD, X-ray powder diffraction.
for Na + intercalation. The results indicate that Na + intercalation in bulk VN is thermodynamically unstable (ΔE i > 0). Thus, these calculations are consistent with the conclusion that charge storage does not arise from the bulk material. Above all, as schematically shown in Figure 5C, the sodium-ion storage of the VN anode is based on the redox reactions from surficial vanadium oxides/oxynitrides layers, exhibiting rapid pseudocapacitive response and high-rate capability.

| CONCLUSION
The present work systematically addresses the charge storage mechanism of the VN electrode in a nonaqueous, sodiumion electrolyte. The VN electrodes exhibit typical surfaceredox pseudocapacitance behaviors: a rectangular-shape CV curves over the potential range (0.05-3.0 V vs. Na + /Na). The pseudocapacitive reaction comes from the continuous redox of surface vanadium oxides ( 3+ ) which occur from (de)sodiation without any change in VN crystal structure. This surface-redox charge storage behavior is different from the well-known conversion reaction mechanism. The surface-redox nature of the different-sized VN particles depends on the surface area and valance state. The VN-10 nm mesoporous nanosheet anode delivers the highest sodium-ion storage capacity with a revisable capacity of 106 mAh g −1 obtained at a high current of 20 A g −1 and excellent cycling performance with 5000 cycles and negligible capacity fading. This work highlights the opportunity to use pseudocapacitive charge storage to overcome the limitations associated with sluggish sodium-ion diffusion and enable Na + -based electrochemical energy storage applications.