Nickel ferrocyanide as a high-performance urea oxidation electrocatalyst

Urea is often present in waste water but can be used in powering fuel cells and as an alternative oxidation substrate to water in an electrolyser. However, an insufficient mechanistic understanding and the lack of efficient catalysts for the urea oxidation reaction have hampered the development of such applications. Here we demonstrate that a nickel ferrocyanide (Ni2Fe(CN)6) catalyst supported on Ni foam can drive the urea oxidation reaction with a higher activity and better stability than those of conventional Ni-based catalysts. Our experimental and computational data suggest a urea oxidation reaction pathway different from most other Ni-based catalysts that comprise NiOOH derivatives as the catalytically active compound. Ni2Fe(CN)6 appears to be able to directly facilitate a two-stage reaction pathway that involves an intermediate ammonia production (on the Ni site) and its decomposition to N2 (on the Fe site). Owing to the different rate-determining steps with more favourable thermal/kinetic energetics, Ni2Fe(CN)6 achieves a 100 mA cm−2 anodic current density at a potential of 1.35 V (equal to an overpotential of 0.98 V). Urea oxidation could be a lower-energy alternative to water oxidation in hydrogen-producing electrolysers, but improved catalysts are required to facilitate the reaction. Geng et al. report nickel ferrocyanide as a promising catalyst and suggest that it operates via a different pathway to that of previous materials.

and the carbon dioxide reduction reactions, which have multiple reaction pathways 20 . One can imagine that a new mechanism with alternative intermediates and more favourable thermodynamics or kinetics could substantially improve the activity of the catalysts 13,20,21 .
In this work, we propose a more energetically favourable UOR pathway triggered by the nickel ferrocyanide (Ni 2 Fe(CN) 6 ) electrocatalyst, one of the Prussian blue analogues 22 . With an alternative reaction intermediate, Ni 2 Fe(CN) 6 delivers one of the best UOR activities compared with those of the existing catalysts. These results indicate a two-stage UOR mechanism that involves a chemical process from urea to NH 3 and an electrochemical process from NH 3 to N 2 on different sites of the catalyst surface, which is very different from the currently understood mechanism. In addition, our two-electrode experiments demonstrate that the UOR process driven by Ni 2 Fe(CN) 6 can be used as an alternative to conventional water oxidation at high current densities.

Physicochemical properties
A typical Ni 2 Fe(CN) 6 catalyst was prepared on nickel foam in a self-assembly method in which cleaned Ni foams were immersed in a mixed solution of polyvinylpyrrolidone, C 6 H 5 Na 3 O 7 ·2H 2 O, NiCl 2 ·6H 2 O and K 3 [Fe(CN)] 6 using different ageing times and salt concentrations. A scanning electron microscope (SEM) image shows that the as-synthesized product was composed of nanocubes assembled on the surface of nickel foam with a single layer of coating ( Supplementary Fig. 1). A transmission electron microscope (TEM) image shows a highly crystalline cubic nanoparticle with an edge length of 180 ± 10 nm ( Supplementary Fig. 1 inset). Different reaction conditions can be used to tune the nanocube size and coverage on the nickel foam ( Supplementary Figs. 2 and 3). Analysis of the X-ray powder diffraction (XRD) pattern obtained for the Ni 2 Fe(CN) 6 powder (Fig. 1a) allowed the identification of a single cubic phase in which the Fe atom coordinates with carbon atoms in CN − species and the Ni atom has two coordination forms: one coordinates with N atoms in CN − species and the other is situated in the centre alone 23,24 . Interestingly, both Fe 2+ and Ni 2+ species were not oxidized to FeOOH and NiOOH during the one-hour UOR process ( Supplementary Fig. 4), which differs from the mechanism reported for other Ni-based catalysts 10,16 .

uOr performance evaluation
As human urine contains 2-2.5 wt% urea (equal to a molar concentration of ~0.33 M), 0.33 M urea was chosen in the electrolysis 2,13 . As shown by the linear sweep voltammetry (LSV), the UOR on an optimized Ni 2 Fe(CN) 6 catalyst exhibits a more negative onset potential than the OER ( Fig. 1b and Supplementary Fig. 5). To obtain a current density of 100 mA cm -2 , the UOR needs a potential of 1.35 V (with an overpotential of 0.98 V), much smaller than that needed for the OER (1.68 V) and among the lowest known UOR electrocatalyst potentials reported, which include nickel hydroxides, metals, phosphides and so on (Supplementary Fig. 6 and Supplementary Table 1) 7,11,14,17,[25][26][27][28][29][30][31][32][33][34] . The high performance of the Ni 2 Fe(CN) 6 electrocatalyst can be attributed to its stable Ni 2+ valance without the generation of NiOOH species. As shown in Fig. 1b, the urea oxidation on Ni 2 Fe(CN) 6 proceeds before the self-oxidation to give the high valence Ni 3+ species (for example, NiOOH). This is very different from most reported Ni-based catalysts, in which a NiOOH phase is formed before the UOR and serves as the active surface 16 . To further verify this observation, in situ X-ray absorption fine structure (XAFS) spectra were obtained to determine the electronic and atomic structure of Ni 2 Fe(CN) 6 during the UOR test. As shown in Fig. 1c,d and Supplementary Fig. 7, the absorption edge positions of Ni and Fe remained unchanged during the test and close to the position of the initial 2+ valence metal. These results also agree well with the ex situ X-ray photoelectron spectroscopy results ( Supplementary Fig. 4).
Interestingly, it was found that two other analogous cyanides, ferric ferrocyanide (Fe 4 [Fe(CN) 6 ] 3 ) and nickel cobaltcyanide (Ni 3 [Co(CN) 6 ] 2 ), showed a worse activity compared with that of Ni 2 Fe(CN) 6 ( Fig. 1e and Supplementary Fig. 8) 35,36 . In addition, nickel iron layered double hydroxide, which also contains Ni and Fe species, showed poorer activity than Ni 2 Fe(CN) 6 counterpart (Supplementary Fig. 9) 37 . Therefore, we speculate that the observed high performance of Ni 2 Fe(CN) 6 is caused by the cooperative action of the two active sites of Ni and Fe in the Ni 2 Fe(CN) 6 catalyst.
A kinetic study showed that the UOR process on Ni 2 Fe(CN) 6 is independent of the urea concentration (Fig. 1f), which agrees with the reported Ni-based electrocatalysts 38 . As can be seen in Fig. 1g, the UOR showed a strong dependence on the amount of KOH with a reaction order of 1.10 with respect to the OH − concentration, which is different from that of about 2.00 for other Ni-based electrocatalysts 38 . In addition, the dependence of pH on the onset potential of Ni 2 Fe(CN) 6 was calculated as ~59 mV pH -1 , which means a Nernstian-type dependence with the RDS of one electronproton coupled step ( Supplementary Fig. 10). This is very different in comparison with the reported RDS for NiOOH electrocatalysts, for example, the desorption of CO 2 , which requires a two-electron transfer step to form CO 3 2− in the electrolyte 16,[39][40][41] . Therefore, this unique kinetic relationship for the Ni 2 Fe(CN) 6 indicates an alternative reaction pathway.

Comparison with the NiC 2 O 4 catalyst
As a control, we prepared a conventional Ni oxalate (NiC 2 O 4 ) catalyst and compared it with Ni 2 Fe(CN) 6 to clarify the UOR mechanism based on a series of in situ spectroscopic measurements 42 . As shown by SEM and XRD, NiC 2 O 4 is in the form of micrometre-sized monoclinic particles ( Supplementary Fig. 11). In the Raman spectra of Ni 2 Fe(CN) 6 , two strong peaks at 2,100 and 2,140 cm -1 are characteristic of cyanide stretching 22 , whereas the peaks at 250 and 348 cm -1 belong to the Ni-N stretching vibration and the peak at 510 cm -1 belongs to the Fe-C stretching vibration (Fig. 2a) 24 . Importantly, at various potentials ( Fig. 2a and Supplementary Fig. 12,), the Raman spectra are similar to those at the open circuit potential (OCP). As a counterpart, the NiOOH doublet peaks at 473 and 560 cm −1 appeared in a very short time at a high potential during the UOR process ( Fig. 2b and Supplementary Fig. 13a,b), which indicates its partial reconstruction to NiOOH, in agreement with the existing literature 39,40 . Note that, even though Ni 2 Fe(CN) 6 showed a better stability than NiC 2 O 4 ( Supplementary Fig. 13a,b), it was converted into Ni hydroxide derivatives after six hours ( Supplementary  Fig. 13c). We collected TEM images and elemental mapping (Fe, Ni and N) of the Ni 2 Fe(CN) 6 catalyst before and after the UOR test ( Supplementary Fig. 14). The latter revealed that the elemental mapping remained uniform after the UOR and no evidence indicates the formation of Fe/Ni-rich core-shell structures.
As expected, Ni 2 Fe(CN) 6 without the NiOOH species showed a notably enhanced apparent UOR activity compared to with that of the NiC 2 O 4 catalyst (Fig. 2c). As NiC 2 O 4 has an electrochemical active area ~3 times larger (judged by the value of double-layer capacitance than that of Ni 2 Fe(CN) 6 , the UOR activity normalized to the electrochemical active area (judged by the anodic current density at a potential of 1.35 V) of Ni 2 Fe(CN) 6 is ~35 times greater than that of NiC 2 O 4 ( Supplementary Fig. 15). As expected, a reaction order of 1.75 with the OH − concentration in the electrolyte was obtained for NiC 2 O 4 (Fig. 2d), which further indicates the conventional reaction mechanism of the derived NiOOH.

Computations based on density functional theory
According to the above analysis, we proposed a mechanism for the UOR, which comprises two stages, namely a chemical process from urea to NH 3 (CO(NH 2 ) 2 + H 2 O → CO 2 + 2NH 3 ) and an electrochemical process from NH 3 The detailed pathways are given in Methods, and the standard Gibbs free energy change (ΔrG • m ) of each reaction intermediate was computed by means of density functional theory (DFT) (Supplementary Table 2 and Supplementary Figs. [16][17][18][19]. The first stage consists of four steps, namely urea adsorption, deamination, decarbonation and NH 3 desorption (Fig. 3a). The third elementary reaction ([M·OCONH 2 ] ads → [M·NH 2 ] ads + CO 2 ) is the thermodynamically limiting step. The Ni site is suggested to be the superior catalytic site in this step, that is, ΔrG • m = 0.90 eV (Ni site) versus 1.02 eV (Fe site). As regards the reaction dynamics, the reaction activation energy ΔG ‡ shows the same trend for the Ni and Fe sites (Fig. 3c The proposed overall reaction mechanism assumes that Ni is responsible for the conversion of urea into ammonia and carbon dioxide, and Fe is responsible for the transformation of ammonia into nitrogen (Fig. 3). This synergistic catalysis between the Ni and Fe sites revealed by DFT computations agrees well with the experimental observation that Ni 2 Fe(CN) 6 exhibits a remarkably higher activity than that of other catalysts with a similar structure (Ni 3 [Co(CN) 6

ammonia detection
As predicted by the DFT computations, NH 3 is a critical reaction intermediate during the UOR, which can be separated into two stages: the chemical reaction from urea to NH 3 and the electrochemical reaction from NH 3 to N 2 . This mechanism is also supported by the detection of ammonia in the electrolyte using an ion ammonia-selective electrode (Orio High-Performance Ammonia Electrode 9512HPBNWP) [43][44][45] . As shown in Supplementary Fig. 22, initially there was a trace amount of ammonia (~0.25 ppm) in the electrolyte for all the electrocatalysts, because of the spontaneous decomposition of urea in alkaline solutions. However, after applying a potential (for example, 1.34 V for Ni 2 Fe(CN) 6 with a current density of 48.0 mA cm -2 ), the NH 3 concentration increased with the reaction time to reach a maximum value of ~0.30 ppm. The same trend was also observed for the UORs on the Fe 4    . This can be explained as that under the operating potential, on the one hand, the formation rate of NH 3 accelerates because of an increasing number of hydroxylated catalytic sites, and on the other hand, the equilibrium of the second stage shifts towards the products (NH 3 → N 2 ) and the reaction rate of this stage accelerates (Fig. 3a,b). Afterwards, the equilibrium of the first stage also shifts towards the products of NH 3 , which is accompanied by its continuous consumption. During this period, some of the NH 3 species generated in situ are desorbed from the catalyst surface into the electrolyte and detected by the ion ammonia-selective electrode. In sharp contrast, a very small change in the NH 3  To further illustrate a two-stage UOR mechanism that involves chemical NH 3 production and electrochemical NH 3 oxidation occurs on Ni 2 Fe(CN) 6 , we also measured the LSV curve for NH 3 oxidation on Ni 2 Fe(CN) 6 in solutions with different amounts of ammonia. By comparing the LSV curves, we found that the onset potentials of ammonia and the urea oxidation on Ni 2 Fe(CN) 6 are very close (both are ~1.32 V versus RHE; Supplementary Fig. 23). This means that the in situ generated ammonia in stage one (urea to NH 3 ) can be electrocatalytically oxidized to the final product of N 2 , which is consistent with the DFT prediction that, in this case, the UOR is a two-stage process, as indicated above. As a consequence, we propose that NH 3   process on the Ni 2 Fe(CN) 6 surface, which is very different from the currently reported mechanisms, which involve *NCO, *HN-CO or *COO as the intermediates 39,40 .

Identification of intermediates
We performed an in situ synchrotron radiation Fourier transform infrared (SR-FTIR) analysis to identify the critical reaction intermediates proposed in the DFT computation (Fig. 4). For Ni 2 Fe(CN) 6 , compared with the spectrum at OCP, two obvious absorption bands appear at 2,925 cm -1 and 1,203 cm -1 under the UOR working potential (for example, 1.35-1.65 V), which can be assigned to the N-H stretching vibration of *N=NH 2 + and the C-O stretching vibration of *OCONH 2 species, respectively (Fig. 4a) 46,47 . In addition, with increasing potential, these two characteristic peaks become stronger (Fig. 4c,d). The simulated results of harmonic vibrational frequencies (Fig. 4e) also indicate that two peaks can be attributed to *N=NH 2 + and *OCONH 2 species. This clearly indicates that these two intermediates are produced in the UOR process, which is supported by the DFT computations with the IMFe9 and IMNi2 intermediates.
We also performed the analysis for the NiC 2 O 4 electrocatalyst (as a control) to identify the different reaction intermediate(s). From Fig. 4b and Supplementary Fig. 24, an obvious absorption band appears at 2,136 cm -1 under the UOR working potential, which can be assigned to CNO − and is consistent with that reported elsewhere 10 . Also, no N-H stretching vibration in the *N=NH 2 + intermediate and C-O stretching vibration in the *OCONH 2 intermediate is observed in the spectrum of the NiC 2 O 4 electrocatalyst. These comparisons clearly indicate that the UOR mechanism on the Ni 2 Fe(CN) 6 catalyst is different from that on the conventional Ni-based catalysts. At this stage, the combination of kinetic analysis, DFT computations and in situ SR-FTIR spectroscopy data suggests an alternative UOR mechanism on Ni 2 Fe(CN) 6 without NiOOH generation, namely, a two-stage pathway that involves a chemical process from urea to NH 3 and an electrochemical process from NH 3 to N 2 at two different active sites.

energy-saving systems driven by uOr
To establish an energy-saving system that benefits from the UOR on the Ni 2 Fe(CN) 6 catalyst, we assembled an UOR//HER electrolyser   that used Ni 2 Fe(CN) 6 as the anode in an electrolyte that contained 1 M KOH and 0.33 M urea. For comparison, an OER//HER electrolyser was used with RuO 2 as the anode in an electrolyte that contained 1.0 M KOH (Supplementary Fig. 25). To obtain current densities of 10 and 100 mA cm -2 , urea electrolysis needs cell voltages of 1.38 and 1.50 V, respectively, whereas water electrolysis needs cell voltages of 1.56 and 1.85 V, respectively (Fig. 5a). This clearly indicates the energy-saving advantage of the UOR process on the Ni 2 Fe(CN) 6 electrocatalyst. In addition, the H 2 production in this UOR// HER cell was stable with a Faradaic efficiency over 90% (Fig. 5b). We also conducted the UOR in an industrial electrolyser with 200 cm 2 cathode and anode areas to further demonstrate its potential in the storage of renewable energy. As shown in Supplementary  Fig. 26, the current can be up to 40 A at cell voltages of 4.1 V. The urea elimination rate of 13.8 g h -1 (with a Faradaic efficiency of 94.0%) and theoretical H 2 production rate of 15.5 l h -1 (based on the Faradaic efficiency of 90%) were achieved at a cell voltage of 4.1 V.
Besides H 2 production, the energy-saving system can also be applied in a UOR//2e − ORR flow cell composed of an Ni 2 Fe(CN) 6 anode and mesoporous carbon (CMK-3) cathode 48 ( Supplementary  Fig. 27). Nowadays, the in situ electrochemical production of H 2 O 2 via 2e − ORR has become a promising method because it can reduce the danger and costs of the transportation of H 2 O 2 (refs 48,49 . As expected, the urea electrolysis needs a smaller energy input than the water electrolysis for urea elimination and H 2 O 2 generation (Fig. 5c). Specifically, a H 2 O 2 production rate of 225.3 gm -2 h -1 (with a Faradaic efficiency of 82.3%) and an urea elimination rate of 140.1 gm -2 h -1 (with a Faradaic efficiency of 94.9%) were achieved at a cell voltage of only 0.63 V (Fig. 5d and Supplementary Fig. 28).

Conclusions
In summary, Ni 2 Fe(CN) 6 with a high activity for the UOR was fabricated by a simple and readily scalable method. It showed a better stability than those of other Ni-based UOR electrocatalysts, but was ultimately converted into Ni hydroxide derivatives over a long period (more than six hours). Studies of the mechanism by employing advanced in situ Raman spectroscopy, in situ SR-FTIR techniques and ammonia detection revealed a more energetically favourable UOR pathway on Ni 2 Fe(CN) 6 as compared with those of most reported electrocatalysts. The DFT results revealed that the highly enhanced electrochemical performance originates from the synergistic effect of Ni and Fe double-active sites in Ni 2 Fe(CN) 6 . The efficient UOR on Ni 2 Fe(CN) 6 has the advantage over the conventional OER in that it requires less energy and could also reduce the urea content of waste water. This work opens a new avenue to develop alternative electrocatalysts for UORs with a boosted activity and stability.   6 ] were used, the product was denoted as Ni 2 Fe(CN) 6 -48-2. NiC 2 O 4 synthesis. NiC 2 O 4 was prepared using a slightly modified method reported elsewhere 42 . Under stirring, 2 mmol Ni(NO 3 ) 2 ·6H 2 O was dissolved into 50 ml of deionized water and 20 ml of ethanol. The mixed solution was transferred into a 250 ml three-necked flask and heated at 60 °C for 2 h. At the same time, 1.58 g of oxalic acid was dissolved in 50 ml of deionized water. The prepared oxalic acid solution was slowly added to the three-necked flask over ~5 min. Then, the cleaned Ni foam was placed into the mixed solution, which was heated 60 °C in an oil bath for 2 h without stirring. After cooling down to room temperature, the Ni foam was washed with distilled water and ethanol and then vacuum dried.  6 ] was dropped slowly into the mixed solution under vigorous stirring at 60 °C. The product was obtained over 30 min, followed by cleaning in ethanol and deionized water three times. In the preparation of Ni 3 [Co(CN) 6 ] 2 , 265 mg of sodium citrate and 143 mg of NiCl 2 ·6H 2 O were dissolved in 20 ml of deionized water. Next, 133 mg of K 3 [Co(CN) 6 ] was dissolved in 20 ml of deionized water. Then, cleaned Ni foam was placed in each of the two mixed solutions under magnetic stirring and aged for 24 h. The products were cleaned in ethanol and deionized water three times.
NiFe layered double hydroxide synthesis. NiFe layered double hydroxide was prepared on a Ni foam according to Yang et al. 37 . 60 mg of Fe(NO 3 )9H 2 O was dissolved in a mixed solution of 15 mL of deionized water and 5 mL of ethanol, and then treated with ultrasound for about 10 min. A piece of cleaned Ni foam was immersed in the mixed solution at room temperature for 24 h. Then, the product was taken out and washed with deionized water several times, and freeze-dried for 12 h.
Characterization. The SEM images were obtained using a SEM (HITACHI Regulus 8230) with a 15 kV accelerating voltage. TEM images were acquired using a field-emission TEM (JEOL JEM 2100). X-ray photoelectron spectroscopy data were obtained using a Thermo ESCALAB 250Xi X-ray photoelectron spectrometer (Thermo ESCALAB 250Xi). XRD patterns were obtained by a diffractometer (Rigaku SmartLab 9KW with Cu Kα radiation, λ = 0.15406 nm) from 10 to 80° at a rate of 10° min -1 .
Electrochemical measurements. An IM6e electrochemical workstation (Zahner-Electrik) was used to test the UOR activity in a three-electrode system, in which the as-prepared free-standing electrocatalyst was directly used as the working electrode. A Hg/HgO electrode and a carbon rod were used as the reference and counter electrodes, respectively. The electrolyte used was 1.0 M KOH with a 0.33 M urea solution. The LSV curves were obtained at a scan rate of 5 mV s -1 . All the curves a 300 UOR (Ni 2 Fe(CN) 6  were corrected manually with iR compensation and the potential was converted into RHE. The UOR//HER system for hydrogen detection was carried out in a two-electrode mode separated by an anion membrane (Sustainion X37-50 Grade RT). The UOR//2e -ORR system was carried out in a commercial flow cell, in which the prepared Ni 2 Fe(CN) 6 on a nickel foam was used as the anode, and a mesoporous carbon-coated gas diffusion electrode was used as the cathode 48 . The anode contained 1.0 mol l −1 KOH with 0.33 mol l −1 urea and the cathode contained 1.0 mol l −1 KOH. The areas of the anode and cathode were both 4 cm 2 . LSV curves were collected in a period of <1 h to avoid the formation of any hydroxide derivatives.
In situ Raman measurements. Raman spectroscopy was carried out using a Via-Reflex spectrometer (Renishaw) with a laser excitation wavelength of 532 nm and the measured potential for the UOR was in the range 1.2-1.6 V controlled by an electrochemical workstation (CHI750E Instruments). The in situ electrochemical three-electrode cell contained a Ni 2 Fe(CN) 6 electrocatalyst as the working electrode, Ag/AgCl as the reference electrode and Pt wire as the counter electrode. The electrolyte used was 1.0 M KOH with a 0.33 M urea solution. Raman spectra at various applied potentials were conducted in a period of <1 h.
In situ synchrotron radiation Fourier transform infrared measurements. In situ SR-FTIR measurements were made at the infrared beamline BL01B of the National Synchrotron Radiation Laboratory through a homemade top-plate cell-reflection infrared set-up with a ZnSe crystal as the infrared transmission window (cutoff energy of ~625 cm -1 ) (ref. 50 ). This end station was equipped with an FTIR spectrometer (Bruker 66v/s) with a KBr beam splitter and a liquid-nitrogen-cooled HgCdTe detector. The system was coupled with an infrared microscope (Bruker Hyperion 3000) with a ×16 objective lens. It could perform infrared spectroscopy measurements over the broad range of 15-4,000 cm -1 with a high spectral resolution of 0.25 cm -1 . The catalyst electrode was tightly pressed against the ZnSe crystal window with a micrometre-scale gap to reduce the loss of infrared light. To ensure the quality of the obtained SR-FTIR spectra, we adopted a reflection mode with a vertical incidence of infrared light. Each infrared absorption spectrum was acquired by averaging 514 scans at a resolution of 2 cm -1 . All the infrared spectra were obtained after a constant potential was applied to the catalyst's electrode for 30 min.
In situ X-ray absorption fine structure measurements. The Ni (8,333 eV) and Fe (7,109 eV) K-edge XAFS data were collected at the BL14W1 station in the Shanghai Synchrotron Radiation Facility. Its storage ring was operated at 3.5 GeV with a maximum current of 250 mA. The beam from the bending magnet was monochromatized utilizing a Si(111) double-crystal monochromator and further detuned by 30% to remove higher harmonics. The Fe K-edge measurements were performed with Ni 2 Fe(CN) 6 on the nickel foam, whereas the Ni K-edge measurements were performed on carbon paper to avoid interference of the Ni signal from the nickel foam. The XAFS spectra were collected using the fluorescence mode. To monitor the changes during the UOR process, voltages from the OCP to 1.35 V were applied for 10 min as the conditioning step. During the measurements, the position of the absorption edge was calibrated using Ni foil and Fe foil, and all the XAFS data were collected during one period of beam time.
Density functional theory calculations. Spin-polarized DFT calculations were carried out using the DMol3 quantum chemical module 51,52 . The gradient-corrected density-functional PW91 (Perdew-Wang generalized-gradient approximation) was applied to predict the structures, single-point energies, zero-point energies and thermodynamic parameters 53 . The Tkatchenko-Scheffler term for the semiempirical dispersion correction for DFT was considered to estimate the bond energy of the σ-π coordination between Ni 2+ and CN − (ref. 54 ). Infrared spectra of the active intermediates were derived from harmonic vibrational frequencies calculations 55 , and the vibrational analysis was performed at the final geometry using identical parameters with the geometry optimizations. The Gibbs free energy changes of reactions were calculated as follows: where E is the single-point energy, ZPE is the zero-point energy, Δ T ∫ 0 CpdT and -TΔS are the correction factors of enthalpy and entropy, respectively, C p is the heat capacity at constant pressure, and E sol is the solvation energy. More details on the DFT calculations are provided in the Supplementary Information.
A reaction mechanism that comprised two stages, namely, the reaction from urea to NH 3 and the reaction from NH 3 to N 2 , was proposed as follows: Stage 1: reaction from urea to NH 3 :

Detection of products. Detection of trace amounts of ammonia.
In the three-electrode model, the as-prepared electrocatalyst was directly used as the working electrode. The Hg/HgO electrode and the carbon rod were used as the reference and counter electrodes, respectively. The anode was designed as a sealed system, and the anode electrolyte (50 ml) was quickly extracted within a certain reaction time. An ion ammonia-selective electrode (Orio High-Performance Ammonia Electrode 9512HPBNWP) was immediately applied to determine the trace amounts of ammonia in the electrolyte 43 .
Measurement of urea elimination. The amount of urea elimination was detected by transforming the urea into ammonia by urease. The ammonia concentration was tested by an ion ammonia-selective electrode. The detection of ammonia obtained from urea was conducted as follows: (1) preparation of ammonia standards-a series of standard solutions were prepared with concentrations of 1, 2, 5, 7 and 10 ppm (as NH 4 + ) in a KOH solution and (2) for the electrometer calibration, first the electrode was soaked in the ammonia electrode storage solution (1 ppm standard in KOH) for at least 15 min, second 100 ml of each standard was measured and third the millivolt and log C (ppm) values were used to prepare the standard curve with a Tafel slope of 54 60 mV dec -1 .
Measurement of H 2 O 2 production. The amount of H 2 O 2 production was measured based on the amount of reduced Ce 4+ , which was tested by ultraviolet-visible spectrophotometry using a UV-3600 spectrophotometer (Shimadzu Scientific Instruments Inc.) 48 .
Electrochemical measurement on an industrial electrolyser. The UOR performance was also evaluated using an industrial electrolyser with a 200 cm 2 cathode area and a 200 cm 2 anode area, in which the current was measured and recorded using an oscilloscope with a current probe. A power supply with a current range from 0 to 50 A was used.

Data availability
The datasets analysed and generated during the current study are included in the paper and its Supplementary Information. Source data are provided with this paper.