Smart Interfacing between Co-Fe Layered Double Hydroxide and Graphitic Carbon Nitride for High-efficiency Electrocatalytic Nitrogen Reduction

2.1 Calculated Interfacial Nitrogen Adsorption/Reduction Properties

Our interfacing strategy to enhance the nitrogen adsorption and reduction activities was theoretically validated by DFT calculations. To simulate the LDH surface, a two-layer model of Co(OH)₂ with 4 × 4 supercell was built up at first, and four iron atoms were used to replace four exposed cobalt atoms to construct the LDH surface with a Co/Fe mol ratio of 3:1. Besides, two H⁺ were removed to simulate the influence of ${\text{CO}}_3^{2-}$ and H₂O, as shown in Figure 1a. To explore the role of coupled heterointerface between LDH and g-C₃N₄, a new model was built up in which one layer of g-C₃N₄ was superposed on the LDH surface (Figure 1b), similar to the case of MXene and LDH.^[46] Based on these structural models, the projected density of states (PDOS) was calculated and shown in Figure 1c. As seen from this figure, LDH@C₃N₄ exhibits a smaller bandgap, suggesting that the electronic conductivity of LDH is improved to some extent after forming a coupled heterointerface with g-C₃N₄. The adsorption properties of LDH and LDH@C₃N₄ toward nitrogen and ammonia are key factors to evaluate their NRR performances. As shown in Figure S1, Supporting Information, LDH@C₃N₄ exhibits a higher binding energy toward nitrogen than LDH, indicating the former is more prone to adsorbing nitrogen molecules, which is a prerequisite for the following reduction process. On the other hand, the binding energy of LDH@C₃N₄ toward ammonia is much lower than that of LDH, meaning the produced ammonia is liable to be desorbed from the former, and would not occupy the active sites to poison the catalyst system. As such, the stronger tendency of LDH@C₃N₄ to adsorb nitrogen and release ammonia endows it with better electrocatalytic properties. Moreover, the binding structures are shown in Figures S2 and S3, and the binding energies are listed in Table 1. Note that except for NH₂NH₂ on LDH@C₃N₄ (with a binding energy of only –0.59 eV), the binding energies of other reaction intermediates on both LDH@C₃N₄ and LDH are relatively high (ranging from –1.31 to –3.17 eV), making the conversion of activated nitrogen possible. However, NNH₂ is easy to dissociate from LDH and form NNH and H, which means NNH₂ cannot stably exist on the LDH surface, as shown in Figure S4.

As discussed above, NNH₂ is inclined to be decomposed into NNH and H on LDH, and NHNH (–1.81 eV) shows a higher binding energy than NNH₂ (–1.31 eV) on LDH@C₃N₄, which means the enzymatic mechanism (containing the NHNH intermediate) is more favored than the distal mechanism (containing the NNH₂ intermediate). Therefore, the enzymatic mechanism on LDH and LDH@C₃N₄ was simulated to evaluate their electrocatalytic properties for the conversion of activated nitrogen to ammonia. As shown in Figure 1d, all electrochemical steps are downhill in the free energy, and LDH@C₃N₄ exhibits a lower desorption energy than LDH (0.71 vs. 1.82 eV). Additionally, the total density of states (TDOSs) was also calculated, as shown in Figure S5. Compared with that of LDH, the TDOS of LDH@C₃N₄ shows a smaller gap near the Fermi level, which means that LDH@C₃N₄ has a smaller bandgap than LDH. Based on our calculation results, interfacing LDH with g-C₃N₄ is indeed able to enhance its ability toward nitrogen adsorption and reduction, as will be proven by the experimental results.

2.2 Synthesis and Characterization of LDH@C₃N₄@TiO₂ NM

The field-emission scanning electron microscopy (FE-SEM) images of TiO₂, C₃N₄@TiO₂, and LDH@C₃N₄@TiO₂ NMs are shown in Figure 2a and Figure S6. It can be seen that the pristine TiO₂ NM is composed of randomly packed nanofibers whose average diameter is ˜270 nm, and is featured by an interconnected 2D network (Figure S6a). After in situ polymerization of gaseous melamine molecules, the adjacent nanofibers are obviously adhered owing to the formation of an ultrathin g-C₃N₄ layer on the nanofiber surface (Figure S6b). Compared with the TiO₂ and C₃N₄@TiO₂ nanofibers, the LDH@C₃N₄@TiO₂ nanofibers have remarkably increased diameters. Different from the smooth surface of the TiO₂ and C₃N₄@TiO₂ nanofibers, needle-like Co-Fe LDH nanoarrays with lengths of several micrometers are observed perpendicular to the nanofiber surface (Figure 2a and Figure S6c), indicating the successful growth of a secondary structure on the self-standing substrate, as schematically illustrated in Figure 2b. The X-ray diffraction (XRD) patterns (Figure 2c) were collected to distinguish the crystalline phase evolution. It is found that three kinds of phases exist in the LDH@C₃N₄@TiO₂ NM. The (101), (004), (200), (105), (211), and (204) planes situated at 25.3°, 37.8°, 48.0°, 53.9°, 55.1°, and 62.7° are consistent with the XRD data of anatase TiO₂ (JCPDS no. 21–1272).^[47] The diffraction peak located at 27.4° is indexed to the (002) plane of g-C₃N₄ (JCPDS no. 87–1526).^[48] The diffraction peaks present at 11.6°, 34.1°, 35.1°, and 38.7° correspond to the (003), (012), (009), and (015) planes of Co-Fe LDH (JCPDS no. 50–0235).^[31]

Transmission electron microscopy (TEM) was employed to reveal the fine structure of LDH@C₃N₄@TiO₂. As shown in Figure 2d, the Co-Fe LDH nanoneedles grew evenly and densely on the surface of the C₃N₄@TiO₂ nanofiber and were well retained after the sonication process for preparing the TEM sample, suggesting that the interfacial interaction between the Co-Fe LDH nanoneedles and the g-C₃N₄ layer was strong enough. Closer observation of the LDH@C₃N₄@TiO₂ nanofiber under high-resolution TEM (HRTEM) indicates that the substrate is composed of intimately connected TiO₂ nanofiber and a g-C₃N₄ layer with a thickness of only ˜2 nm (Figure 2e), which shows a good match with the TEM and HRTEM images of the C₃N₄@TiO₂ nanofiber (Figure S7a,b). As shown in Figure 2f, the TEM image focusing on a single nanoneedle displays a lattice fringe spacing of 0.26 nm assigned to the (012) plane of Co-Fe LDH.^[49] The scanning transmission electron microscopy (STEM) image and the corresponding elemental maps obtained from a single C₃N₄@TiO₂ nanofiber manifest that the carbon and nitrogen elements present homogeneous distributions on the surface of the TiO₂ nanofiber, which also provide evidence for the successful coating of an ultrathin g-C₃N₄ layer (Figure S7c). The content of the ultrathin g-C₃N₄ layer in the C₃N₄@TiO₂ NM, as determined by thermal gravimetric analysis (TGA), is 12.32 wt% (Figure S8). As for the LDH@C₃N₄@TiO₂ nanofiber, the elemental maps of titanium, carbon, and nitrogen draw a fibrous outline, matching well with the morphology of the C₃N₄@TiO₂ nanofiber (Figure 2g). The distribution of cobalt, as a representative element of Co-Fe LDH, is also revealed in Figure 2g, which is in good agreement with the nanoneedles standing on the surface of the C₃N₄@TiO₂ nanofiber. According to the energy dispersive X-ray spectroscopy (EDS) spectrum (Figure S9), the cobalt and iron contents in the LDH@C₃N₄@TiO₂ NM are 19 and 5.7 wt%, respectively, which are consistent with their feed ratios, demonstrating that the Co-Fe LDH nanoneedles were facilely synthesized through one-step epitaxial growth.

Since the electrolyte permeability has an important effect on the electrocatalytic performance, the nitrogen adsorption–desorption isotherms of TiO₂, C₃N₄@TiO_2, and LDH@C₃N₄@TiO₂ NMs were collected to reveal their pore structures (Figure S10, Supporting Information). It is easy to find that all the curves match well with type IV adsorption isotherms, where a series of adsorption behaviors including monolayer adsorption, multilayer adsorption, and capillary condensation are displayed, indicating that mesopores exist within the NMs. Based on the above curves, the Brunauer–Emmett–Teller (BET) surface areas of TiO₂, C₃N₄@TiO_2, and LDH@C₃N₄@TiO₂ NMs are determined, being 38.95, 22.08, and 34.29 m² g⁻¹, respectively. It is noteworthy that the C₃N₄@TiO₂ NM shows a smaller surface area than the TiO₂ NM, which is attributed to the intimate contact between the ultrathin g-C₃N₄ layer and the TiO₂ nanofibers. Interestingly, the surface area of the LDH@C₃N₄@TiO₂ NM increases to a value comparable to that of the TiO₂ NM, revealing the significant effect of the Co-Fe LDH nanoneedles on improving the porous structure of the designed catalyst system. Another important feature of the LDH@C₃N₄@TiO₂ NM is its robustness and flexibility. As shown in the inset of Figure S11, Supporting Information, the LDH@C₃N₄@TiO₂ NM could withstand a bending test of approximately 90° and return to the initial shape without obvious deformation. The tensile stress–strain curve in Figure S11, Supporting Information indicates that the LDH@C₃N₄@TiO₂ NM possesses a strength up to 810 kPa. In contrast to the traditional substrate (e.g., nickel foam) that was brittle after the solvothermal reaction, the LDH@C₃N₄@TiO₂ NM has the advantages of convenient usage and excellent stability.^[31]

The X-ray photoelectron spectroscopy (XPS) data of the LDH@C₃N₄@TiO₂ NM were collected to figure out the valance states of the elements. The wide-scan survey spectrum in Figure 3a shows that the titanium, oxygen, carbon, nitrogen, cobalt, and iron elements are present in the sample. Figure 3b shows the Ti 2p spectrum, and the two separate peaks situated at 464.6 and 458.8 eV are assigned to Ti 2p_1/2 and Ti 2p_3/2 orbitals, respectively.^[44] The O 1s spectrum in Figure 3c is deconvoluted into three spin-orbit peaks, of which the one located at 529.0 eV is indexed to lattice oxygen (Ti–O),^[44] the one at 530.9 eV is ascribed to the –OH group,^[50] and the one at 531.3 eV comes from the O–C=O bond in Co-Fe LDH.^[31]

As can be seen in Figure 3d, three isolated peaks at 284.4, 287.7, and 288.7 eV are deconvoluted from the C 1s spectrum, corresponding to the C–C bond in surface adventitious carbon, the N–C=N bond in g-C₃N₄, and the O–C=O bond in Co-Fe LDH, respectively.^{[31, 51]} The N 1s spectrum in Figure 3e is fitted into three peaks located at 398.1, 399.3, and 400.7 eV, which are ascribed to C–N=C, N–(C)_3, and C–N–H, respectively.^[52] In the Co 2p spectrum (Figure 3f), both Co 2p_3/2 and Co 2p_1/2 can be fitted into two spin-orbit peaks, indicating the presence of two states of cobalt element in the LDH@C₃N₄@TiO₂ NM. In the Co 2p_3/2 peak, the two deconvoluted peaks located at 781.2 and 783.2 eV are assigned to Co³⁺ and Co²⁺, along with a satellite peak at 786.8 eV. As for the Co 2p_1/2 peak, it can be fitted into two peaks located at 796.9 and 798.3 eV that correspond to Co³⁺ and Co²⁺, and a satellite peak at 803.1 eV.^[53] The Fe 2p XPS spectrum in Figure 3g can be deconvoluted into two main peaks at 711.2 and 724.6 eV which are ascribed to Fe 2p_3/2 and Fe 2p_1/2 orbitals, and two satellite peaks at 716.4 and 732.8 eV, confirming the existence of Fe³⁺ in the sample.^[54] In order to further investigate the chemical states of Co and Fe in the LDH@C₃N₄@TiO₂ NM, X-ray absorption near edge structure (XANES) at Co and Fe K-edge was measured (Figure 3h,i). The absorption edge energy of Co K-edge in the LDH@C₃N₄@TiO₂ NM (7722.75 eV) is higher than that in Co foil (7709 eV) and CoO (7721.37 eV), whereas lower than that in Co₂O₃ (7728.19 eV), verifying the coexistence of Co²⁺ and Co³⁺. Similarly, it also could be found that the absorption edge energy of Fe K-edge in the LDH@C₃N₄@TiO₂ NM (7116.34 eV) is higher than that in Fe foil (7112.05 eV) and FeO (7114.78 eV), and very close to that in Fe₂O₃ (7116.30 eV), demonstrating that the valence state of Fe is +3 and this echoes the Fe 2p XPS result.

To explore the effect of interfacial coupling between g-C₃N₄ and Co-Fe LDH, the PL spectrum of the LDH@C₃N₄@TiO₂ NM is compared with that of the C₃N₄@TiO₂ NM, as shown in Figure 4a. Note that the former exhibits a lower photoluminescence (PL) intensity, once again proving that the facilitated interfacial charge transfer between Co-Fe LDH and g-C₃N₄. This phenomenon was further confirmed by electrochemical impedance spectroscopy (EIS) (Figure 4b), in which the LDH@C₃N₄@TiO₂ NM displays a much smaller diameter of the semicircle than the C₃N₄@TiO₂ NM, corresponding to a lower charge transfer resistance.

2.3 Electrocatalytic NRR Performance of LDH@C₃N₄@TiO₂ NM

To test the NRR performance of the LDH@C₃N₄@TiO₂ NM, it was used as a self-standing working electrode in an H-type electrolytic cell. Before the NRR tests, the cathodic compartment was bubbled with 99.999% nitrogen for at least 30 min to dispel the air in the electrolyte, and argon or nitrogen was introduced in the successive step. The linear sweep voltammetry (LSV) results are presented in Figure S12, Supporting Information, which discloses an obvious difference in the current between the two feed gases, suggesting possible NRR in the nitrogen atmosphere. Motivated by this result, the LDH@C₃N₄@TiO₂ NM was further subjected to electrolysis in nitrogen at different potentials (–0.45 to –0.75 V vs. RHE) for 2 h, and the chronoamperometric curves are presented in Figure 5a, which provide the information on the total quantities of applied charge at these potentials. The electrolytes after electrolysis were subjected to the indophenol blue method, by which the absorbance at 662 nm can be linearly correlated to the ammonia concentration (Figure S13, Supporting Information).^[20] Accordingly, the absorbance of the electrolyte at each potential is shown in Figure 5b, from which we can see the highest absorbance occurs at –0.55 V vs. RHE. A decreased absorbance occurs at higher potential, which probably arises from a stronger competition from the hydrogen evolution reaction. Based on the chronoamperometric curves and UV-vis absorption spectra, the ammonia yield and Faradaic efficiency are obtained and given in Figure 5c. As seen from this figure, the highest ammonia yield comes from –0.55 V vs. RHE, being 2.07 × 10⁻⁹ mol s⁻¹ cm⁻², while the highest Faradaic efficiency comes from –0.45 V vs. RHE, being 25.3%. Note that these values are significantly higher than those of the TiO₂ (0.45 × 10⁻⁹ mol s⁻¹ cm⁻²) and C₃N₄@TiO₂ (0.72 × 10⁻⁹ mol s⁻¹ cm⁻²) NMs, confirming the prominent synergistic effect of the three components (Figure 5d and Figure S14, Supporting Information). Moreover, the NRR performance of the LDH@C₃N₄@TiO₂ NM was compared with some other noble-metal-free electrocatalysts ever reported based on LDH, g-C₃N₄, TiO₂, etc., and it is found to surpasses most of these electrocatalysts (Table S1, Supporting Information), demonstrating the effectiveness of remarkable interfacial effect and unique structural design in our catalyst system.

It is known that the repeatability and long-term durability are crucial parameters for an electrocatalyst. As such, the LDH@C₃N₄@TiO₂ NM underwent a cycling experiment in which it was repeatedly electrolyzed in nitrogen at –0.55 V vs. RHE for five times (each time 2 h). The indophenol blue method was used to determine the ammonia contents in the electrolytes, and the ammonia yields along with Faradaic efficiencies are presented in Figure 5e. It is seen that only a tiny fluctuation occurs with regard to ammonia yield as well as Faradaic efficiency in different cycles, suggesting excellent repeatability and stability of the LDH@C₃N₄@TiO₂ NM. Moreover, it could withstand long-term electrolysis of 12 h with a stable chronoamperometric curve, exhibiting good durability which is of great importance to practical applications (Figure 5f). To verify whether ammonia came from nitrogen by electrocatalysis of the LDH@C₃N₄@TiO₂ NM, it was subjected to electrolysis under open circuit for 2 h. The indophenol blue method was used to test the resultant electrolyte, and the result shown in Figure S15, Supporting Information indicates that no ammonia was detected. Besides, the electrolyte was tested according to the Watt and Chrisp method, which could also correlate the hydrazine concentration with its absorbance at 458 nm linearly (Figure S16, Supporting Information).^[20] As seen from Figure S17, Supporting Information, no hydrazine was detected in the electrolyte after being electrolyzed at –0.55 V vs. RHE. Finally, the ${\text{NO}}_2^{-}$ content in the electrolyte was also detected to be zero, which excludes possible NO_x contamination in the feed gas (Figures S18 and S19, Supporting Information). To convincingly get clear the origin of ammonia, the isotopic labeling was conducted since the result of ¹H resonance with ¹⁵N is totally different from that with ¹⁴N.^[55] Specifically, two symmetrical peaks with a spacing of 71.4 Hz can be found in the ¹H NMR spectrum of a standard ${}^{15}{\text{NH}}_4^{+}$ sample, as shown in Figure 6. After being electrolyzed with ¹⁵N₂ as the feed gas at –0.55 V vs. RHE for 12 h, the electrolyte was collected and subjected to ¹H nuclear magnetic resonance (NMR) spectroscopy, the result of which is consistent with the standard ${}^{15}{\text{NH}}_4^{+}$ sample. As such, it can be concluded that the detected ammonia indeed came from nitrogen as catalyzed by the LDH@C₃N₄@TiO₂ NM, since it is hard to introduce ¹⁵N-containing contaminants during throughout the whole process.