The escalating concentration of nitrate (NO3−) in surface and groundwater has increasingly raised environmental and ecological concerns1–3. Addressing NO3− contamination by converting it into high value-added products, such as ammonia (NH3), presents a promising solution4, 5. Ammonia is a widely utilized industrial chemical essential for human development, serving as a key component in fertilizers and various industrial processes. Traditionally, NH3 synthesis has relied primarily on the Haber-Bosch process, which operates under high temperatures and pressures, leading to substantial energy consumption and significant greenhouse gas emissions. In contrast, producing NH3 via the electrochemical reduction of NO3−, powered by renewable electricity, offers significant potential for economic efficiency and environmental sustainability6–8. Despite these advantages, the process from NO3− to NH3 involves multiple complex intermediate conversions and an eight-electron transfer process, posing considerable challenges to the activity and high Faradaic efficiency (FE) of newly developed electrocatalysts. Overcoming these hurdles is crucial for advancing this promising method of NH3 production and realizing its full potential for both economic and environmental benefits.
Atomically dispersed catalysts have the highest atomic utilization and superior performance over conventional metal nanoparticles9–13. In recent years, some advanced high performance atomically dispersed catalysts have been designed and prepared to produce NH3 by electrochemically catalyzing NO3− reduction reaction (NO3RR)14–17. However, the challenges of the activity and selectivity of single metal sites in the electrocatalysis of NO3RR remain as well. For example, the atomically dispersed copper (Cu) sites have a high activity and FENH3 for NO3RR catalysis in alkaline media, but a higher overpotential is required as usual18, 19. Another problem is the nitrite (NO2−) accumulation at the Cu sites during NO3RR, a quasi-stable intermediate carcinogen, ultimately resulting in lower FE for NH3 production20, 21. Therefore, some strategies such as modification, doping, and alteration of coordination structure targeting Cu sites have been proposed to enhance the performance of catalysts in the NO3RR process22–24. Recent studies demonstrated that ruthenium (Ru) also presents remarkable activity in the electrocatalytic production of NH3 during the NO3RR process25–27. More importantly, Ru catalysts can achieve efficient conversion of NO2− to NH3 by enhancing the adsorption of intermediates28. Further, it reported that Cu combined with other metal atoms can modulate the local electronic structure and improve the performance of the catalysts29, 30. Dual atom catalyst (DAC) is an atomically dispersed catalyst with an active site consisting of two paired metal atoms31–34. The synergy between pairs of metal atoms in a DAC provides the special ability to further reduce the energy barrier of complex chemical reactions35, 36. If an atomically dispersed bimetallic catalyst with Cu and Ru acting together is designed, the synergistic effect of the two metal atoms can not only enhance the adsorption of the intermediates but also reduce the energy barriers of the intermediate reaction steps. Pulsed discharge is a process in which the high density current passes through the medium between electrodes, and huge energy is injected into the medium, resulting in an instant rise in temperature, phase transition, and dramatic volume expansion of the medium37–40. It is a new environmentally-friendly material preparation method with the advantages of high instantaneous power and good repeatability. This technique shows great potential in the synthesis of novel atomically dispersed materials.
In this work, nitrogen-doped graphene aerogel supported Ru-Cu dual atoms catalyst (RuCu DAs/NGA) is synthesized rapidly by the pulsed discharge method. The metal salts supported by NGA are injected with huge energy in tens of microseconds, resulting in the explosive decomposition of the metal salt nanocrystals and the formation of atomically dispersed RuCu dual sites on the NGA. The asymmetric RuN2-CuN3 coordination structure is proposed based on corresponding detection and analysis. Impressively, RuCu DAs/NGA exhibits excellent NH3 production performance during the electrochemical catalytic NO3RR process. Some important hydrogenation intermediates (such as *NH2OH, *NH2, etc.) were detected by in situ attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS). Furthermore, in situ X-ray absorption fine structure (XAFS) was utilized to trace and analyze the active sites of RuCu DAs/NGA during electrocatalytic NO3RR. The reaction paths and relative free energies of RuN4/C, CuN4/C, and RuN2CuN3/C at U = 0 V are calculated through density functional theory (DFT). The modulated asymmetric RuN2CuN3/C structure can not only optimize the adsorption of the intermediate but also reduce the energy barrier of the elementary reaction. Additionally, we synthesized other atomically dispersed M-Cu (M = Ag, Pt, Pd) dual atoms sites on NGA and suggested similar asymmetric coordination structures (MN2-CuN3).
Pulsed discharge synthesis of RuCu DAs/NGA. As a demonstration, we tailored RuCu dual atoms on NGA substrates by pulsed discharge strategy (Fig. 1a). At first, nitrogen-doped graphene hydrogel (NGH) was prepared by the hydrothermal assembly approach, and then soaked in the aqueous solutions of copper chloride (CuCl2) and ruthenium chloride (RuCl3) for 6 hours. NGA (Supplementary Fig. 1) supported RuCl3 and CuCl2 (RuCl3-CuCl2/NGA) could be obtained through a quick freeze-drying method. The RuCl3-CuCl2/NGA was packed into a copper discharge tube and compressed by copper plugs. The two ends of the copper tube were connected to the two electrodes of the high-power pulsed discharge system through copper strips (Supplementary Figs. 2 and 3). Once the capacitor was fully charged, the air switch of the discharge system would be triggered to close quickly. As a result of the huge pulse current, RuCl3-CuCl2/NGA itself generated an instant thermal shockwave. Furthermore, applying a high-intensity current pulse engenders a potent electromagnetic field, thereby giving rise to several transient regions of elevated temperature on graphene. Metal salt structures explosively break down rapidly owing to huge energy input, even directly into gaseous *Ru, *Cu, and *Cl ions from solid nanocrystals (Supplementary Video 1). These gaseous metal ions are anchored by N atoms on the graphene and form bimetallic atom pairs under the action of the sharp pulsed electromagnetic field (Fig. 1b). The micropore distribution of NGA and graphene aerogels (GAs) by N2 adsorption-desorption tests are shown in Fig. 1c, with holes with diameters ranging from 0.4–0.8 nm being the main pore defects on N-doped graphene. The inset (Fig. 1c) is a schematic diagram of N-doped graphene with different micropore types, the N atoms provide a large number of sites for the fixing of atomically dispersed melt atoms. These sub-nanopores on two-dimensional graphene are the critical space-confined strategy in our design for anchoring bimetallic atom pairs. The apparent activity of the catalyst could be improved due to its high specific surface area and microporous structure, which is conducive to mass transfer. Figure 1d shows the typical current waveform during pulsed discharge. The voltage amplitude was set at 8 kV (Supplementary Fig. 4), and the value approached 0 after several oscillating attenuations, lasting a total duration of about 600 µs. The first current peak lagged behind the first voltage peak by ~ 25 µs, a result of the combined action of capacitors and inductors in the discharge system. The main peak parameters of the current waveform are listed in Supplementary Table 1 during the pulsed discharge process. The current presented a typical underdamped waveform in this resistance-inductance-capacitance (RLC) circuit, indicating that the resistance in the circuit did not change significantly during the pulsed discharge. The energy barriers of metal salts would blast and decompose when the transient energy input, as illustrated in Fig. 1e. Because the current and the effect of Joule heat are almost synchronized41, 42, the temperature rise time from 0 to peak on the metal salt nanocrystals is about tens of microseconds. When metal salt nanocrystals were inputted such a large amount of energy in a short period, they would decompose explosively and the metal ions anchored on NGA during the pulsed discharge (Supplementary Note 1). The magnetic pinching effect generated by the dynamic electromagnetic field inhibits the radial expansion of the formed ions, maintaining the relatively high-density plasma containing Ru and Cu ions in the NGA. The mixed Ru and Cu ions form atomic pairs on the defects of NGA caused by the trapping effect driven by thermal dynamics. Additionally, the air within the porous NGA structure would give rise to localized corona discharge plasmas comprising *O and *N ions (Supplementary Fig. 5). Following multiple discharge events, these conditions facilitate the formation of robust bonds between the metal atom pairs and the NGA carrier. Due to the current peak value being only kept for tens of microseconds and then quickly dropping to 0, the distinctive NGA-supported atomic dispersed metals structure could be retained at high cooling rates and stably exist. During the cooling process, some metal atoms that are very close together inevitably coalesce to form metal nanoclusters. Multiple repetitions of pulsed discharge treatment are necessary, as they induce repeated vaporization and dispersion of the metal atoms. Furthermore, repeated pulsed discharges facilitate the thorough mixing of Ru and Cu metal atoms, promoting their anchoring on NGA to form Ru-Cu atomic pairs. Our experiments have demonstrated that RuCu DAs/NGA can be fully obtained after six cycles of pulsed discharges (Supplementary Fig. 6).
A significant development trend in single-atom catalysts is the incorporation of multiple metal elements on the support to form asymmetric coordination structures (Fig. 1f). Utilizing the rapid thermal shockwave and corona plasma effects of pulsed discharge, atomically dispersed catalysts of multiple metals can be rapidly synthesized. Additionally, the microsecond-scale heating and cooling characteristics create favorable conditions for the formation and stable existence of asymmetric coordination structures between the metal and support atoms. Compared to other synthesis methods for atomically dispersed catalysts, our pulsed discharge synthesis technique (Supplementary Fig. 7) stands out due to its exceptionally high instantaneous temperatures and extremely short duration, on the order of hundreds of microseconds.
After six pulsed discharge treatments, the still intact RuCu DAs/NGA sample was recovered from the discharge tube. Figure 2a displays the profile scanning electron microscope (SEM, Supplementary Note 2) image of RuCu DAs/NGA, the 3D porous structures (tens of microns) are similar to that of the initial NGA. The transmission electron microscopy (TEM) image still exhibits the features of curly porous (tens of nanometers) graphene, as shown in Fig. 2b. Figure 2c shows the high-angle annular dark field-scanning transmission electron microscopy (HAADF-STEM) image of RuCu DAs/NGA, no nanoparticles could be found on graphene. The corresponding energy dispersive spectrum mapping results are illustrated in Fig. 2d, where the elements C (red), N (blue), Cu (yellow), and Ru (pink) are uniformly distributed. The high-resolution HADDF-STEM image reveals a significant number of evenly distributed bright spots on the graphene, as shown in Fig. 2e. These bright spots are thought to be atomically dispersed Ru and Cu atoms because the atomic numbers of metals are much higher than that of C, N, and other nonmetallic elements. Figure 2f is a partial enlargement image of Fig. 2e, due to the atomic number of Ru being much larger than that of Cu, the brightness of Ru atoms in the STEM dark field image is significantly higher than that of Cu atoms. Meanwhile, a large proportion of Ru-Cu metal atoms form pairs. Figure 2g depicts the corresponding two 3D intensity profiles in Fig. 2f, which exhibits the obvious intensity difference between Ru and Cu atoms, as well as the significant difference with C, N, and other atoms in NGA. Furthermore, the frequency histogram of Ru and Cu adjacent atomic distance distribution and the frequency distribution histogram of single and double sites statistically in the HADDF-STEM images are shown in Fig. 2h and Supplementary Fig. 8. The average distance between Ru and Cu atoms is about 0.25 nm, and the two easily interact at this distance. The proportion of Ru-Cu dual sites is about 65.9%, indicating that the RuCu DAs/NGA sample prepared by pulsed discharge is worthy of the name. Notably, lots of N atoms provide rich sites for anchoring metal atoms. The contents of Ru and Cu are 0.39 at% (3.1 wt%) and 0.52 at% (2.6 wt%), respectively (Supplementary Fig. 9), which is consistent with the results (Ru 3.3 wt%, Cu 2.9 wt%) of inductively coupled plasma optical emission spectrometry (ICP-OES) test.
Figure 2i presents a schematic plot of Ru-Cu dual atoms with different types anchored on the N-doped graphene, this is the main reason for the different lengths between Ru and Cu atoms. Furthermore, the varying angles formed between the RuCu and the NGA planes can also influence the interatomic distances within the Ru-Cu atom pairs (Supplementary Fig. 10). The intensity ratio of D (~ 1346 cm− 1) to G (~ 1594 cm− 1) for RuCu DAs/NGA is slightly higher than that for NGA in the Raman test (Supplementary Fig. 11), demonstrating that more defects were formed in graphene owing to Ru-Cu atoms dopant during pulsed discharge. The X-ray diffraction (XRD) pattern (Supplementary Fig. 12) displays only one distinct wide peak at ~ 25°, attributed to the stacked graphene layers. No peaks of Ru or Cu crystals appear in the RuCu DAs/NGA.
Analysis of the atomic bond structure. The atomic bond structure of RuCu DAs/NGA was investigated by X-ray photoelectron spectroscopic (XPS) and XAFS. The XPS spectra of RuCu DAs/NGA are shown in Supplementary Fig. 13, a distinct N 1s peak was identified, which contains a C-N bond and a weak C-Ru/Cu bond. The Cu 2p3/2 spectrum has two peaks at 931.5 and 933.4 eV, respectively, assigned to Cu+ and Cu2 + 43, 44. Additionally, the weak signals of Ru 3p were detected as well. Figure 3a shows the Cu K-edge X-ray absorption near-edge spectra (XANES) of RuCu DAs/NGA and references (Cu, CuO, and Cu SAs/NGA). The absorption edge of RuCu DAs/NGA is closer to that of CuO than that of Cu foil, indicating that the oxidation state of Cu in RuCu DAs/NGA is closer to CuO. The k3-weighted Fourier transform (FT) from Cu K-edge extended X-ray absorption fine structure (EXAFS) spectra (Fig. 3b) present the peaks of RuCu DAs/NGA and Cu SAs/NGA (Supplementary Note 3) are located at ~ 1.45 Å, which attributed to the Cu-N bond in first shell scattering45, 46. Moreover, the secondary peak at 2.30 Å of RuCu DAs/NGA is close to the first shell location of the Cu foil (2.24 Å), which implies that the existence of the metallic diatomic coordination structure in RuCu DAs/NGA. Figure 3c displays the Cu K-edge EXAFS fitting result of RuCu DAs/NGA in the R space. The fitting and experimental results have a high degree of matching in different spaces (R space, k space, and q space, Supplementary Fig. 14). Correspondingly, the structural parameters were extracted from the Cu K-edge EXAFS fitting results (Supplementary Table 2). The coordination number of Cu was estimated to be 2.9 Cu-N (first peak) and 0.8 Cu-Ru (second peak) in the first shell according to the fitting results, with bond lengths of 1.93 and 2.59 Å, respectively. The Ru K-edge XANES of RuCu DAs/NGA and references (Ru, RuO2, and Ru SAs/NGA) are shown in Fig. 3d. The absorption edge of RuCu DAs/NGA is located between RuO2 and Ru. The k3-weighted FT from Ru K-edge EXAFS spectra (Fig. 3e) exhibit the strong peaks of RuCu DAs/NGA and Ru SAs/NGA are located at about 1.50 Å, which correspond to the Cu-N bonds in the first shell. Similarly, the secondary peak at 2.36 Å of RuCu DAs/NGA is close to the peak position of the Ru foil (2.34 Å), indicating the presence of the metallic bond in RuCu DAs/NGA as well. Likewise, a good fitting result in R space is shown in Fig. 3f. Based on the fitting results, the coordination numbers of Ru-N and Ru-Cu are 2.2 and 0.8 in the first shell respectively. The bond lengths of Ru-N and Ru-Cu are 2.07 Å and 2.57 Å, respectively. Ultimately, a proposed asymmetric coordination structure (RuN2-CuN3/NGA) is depicted in Fig. 3g. According to the first derivative of the Cu absorption edge of RuCu DAs/NGA and references, the valence states of Cu and Ru in RuCu DAs/NGA were estimated to be 1.32 and 3.0, as shown in Fig. 3h. The Cu and Ru K-edge wavelet transform (WT) EXAFS results of RuCu DAs/NGA and references were utilized to distinguish backscattered atoms, as shown in Figs. 3i and 3j. The maximum intensity position of Cu for RuCu DAs/NGA is ~ 6.7 Å, which is closer to that of CuO (~ 6.2 Å) than that of Cu foil (~ 8.0 Å). Similarly, the maximum intensity position of Ru for RuCu DAs/NGA (~ 6.0 Å) is close to that of RuO2. The difference in intensity between the RuCu DAs/NGA and the references originates from the combined contribution of Ru-N, Cu-N, and Ru-Cu.
Generality of synthesis and structural analysis of CuM DAs/NGA (M = Pt, Ag, Pd). Since the instantaneous temperature generated by the pulsed discharge is super high compared to the thermal decomposition temperature of ordinary metal salts, this method can rapidly prepare the atomical level dispersion for most metals. In addition to the Cu-Ru diatomic structure, the combinations of Cu and several other metals were designed to form asymmetric coordination structures of Cu-M/NC on NGAs in this research. The combination of Cu and other metals atomically dispersed dual catalysts on NGA provides different potential applications for different electrocatalytic scenarios. Here, the general pulsed discharge synthesis method was easily extended to the preparation of other metals dual atoms supported by NGA, such as PtCu DAs/NGA, AgCu DAs/NGA, PdCu DAs/NGA, etc. Figure 4a shows a TEM image of PtCu DAs/NGA, which exhibits 3D pleated graphene characteristics and is not supported with metal nanoparticles or clusters. After washing with water and freeze-drying, PtCu DAs/NGA is the broken lamellar graphene aerogel (inset Fig. 4a). The EDS mapping images of PtCu DAs/NGA are displayed in Fig. 4b, Pt and Cu elements are uniformly distributed on the N-doped graphene sheets. The contents of Pt and Cu are 0.5 at% (7.2 wt%) and 0.55 at% (2.6 wt%), respectively (Supplementary Fig. 15), which is consistent with the results (Pt 7.5 wt%, Cu 3.0 wt%) of ICP-OES test. Figure 4c presents a HADDF-STEM high magnification image of PtCu DAs/NGA, with the red oval dotted line surrounding the Pt-Cu bimetallic pair sites.
Further, the coordination structure of PtCu DAs/NGA was analyzed through XAFS testing. The Pt L3-edge XANES results of PtCu DAs/NGA and references (Pt foil and PtO2) are shown in Fig. 4d, the valence state of Pt in PtCu DAs/NGA is between 0 and + 4 based on the intensity of white lines. The valence state of Cu in PtCu DAs/NGA was calculated to be 1.36 (Supplementary Fig. 16). The k3-weighted FT from Pt L3-edge EXAFS spectrum (Fig. 4e) exhibits the peak of PtCu DAs/NGA is located at ~ 1.63 Å, which corresponds to the Pt-N bond in the first shell. Besides, the second peak at 2.30 Å for PtCu DAs/NGA, is comparable to the first shell of Pt foil (2.60 Å), with proposing the metal-metal bond. The EXAFS fitting result in R space for PtCu DAs/NGA is inserted in Fig. 4e, and it can be seen that the experimental and fitting curves are in agreement. Equally, the characterizations and structures of AgCu DAs/NGA and PdCu DAs/NGA are shown in Figs. 4f-4o and Supplementary Figs. 17–20. Supplementary Table 3 provides the best-fitting structural parameters, suggesting the local coordination structures of PtN2CuN3, AgN2CuN3, and PdN2CuN3 in the three samples. According to the first derivative of the Cu (Ag, Pd) absorption edge (or the white-line peak area of Pt) for PtCu DAs/NGA, AgCu DAs/NGA, PdCu DAs/NGA, and their references, the oxidation states of Cu are estimated to be 1.36, 1.32 and 1.42 in PtCu DAs/NGA, AgCu DAs/NGA, and PdCu DAs/NGA respectively (Fig. 4p). The oxidation states of Pt, Ag, and Pd in three catalysts were calculated to be 1.73, 0.36 and 1.75 respectively. The percentage of Pt-Cu, Ag-Cu, and Pd-Cu dual sites are 64.9%, 61.6%, and 65.4% respectively, demonstrating that interacting metals dual sites dominate as compared to single sites in PtCu DAs/NGA, AgCu DAs/NGA, and PdCu DAs/NGA (Fig. 4q).
The optimized structures and differential charge densities of PtCu DAs/NGA, AgCu DAs/NGA, and PdCu DAs/NGA are shown in Figs. 4r-4s. The bond lengths of Pt-Cu, Ag-Cu, and Pd-Cu pairs are estimated to be 2.22 Å, 2.32 Å, and 2.31 Å respectively. Furthermore, the differential charge densities of PtN2CuN3/C, AgN2CuN3/C, and PdN2CuN3/C were calculated to elucidate the electrical properties of the asymmetric Cu-M sites. The asymmetric deployment of the modulated MN2-CuN3 leads to a significantly polarized surface charge distribution. Electron enrichment (cyan) near the MN2 site and electron deficiency near CuN3 can be attributed to electron transfer from the CuN3 site to the PtN2, AgN2, and PdN2 sites, respectively (Supplementary Fig. 21). The three extended studies demonstrate the universal strategy to synthesize unsymmetrical atomic interface structures for the pulsed discharge method. This combination of Cu and noble metals could open more catalytic possibilities for atomically dispersed diatomic catalysts.
Electrocatalysis and in situ study. The reaction of nitrate reduction to the ammonia was evaluated in 0.1 M KNO3 and 0.1 M KOH conditions (Supplementary Note 4), and linear sweep voltammetry (LSV) curves are shown in Fig. 5a. RuCu DAs/NGA showed the lowest onset potential and the fastest current density decrease in the three catalysts. The Faradaic efficiency and ammonia evolution rate were studied by chronoamperometry with different operated potentials (Supplementary Fig. 22). Figure 5b presents the FE of the NH3 product (FENH3) at different potentials (-0.1 V to -0.6 V vs. RHE). Impressively, the FENH3 of RuCu DAs/NGA reached 97.8% at -0.3 V vs. RHE, the performance is very competitive compared to other catalysts reported recently47, 48. The partial current density of NH3 (JNH3) on RuCu DAs/NGA has the optimal performance among the three catalysts from − 0.1 V to -0.6 V vs. RHE (Fig. 5c), suggesting the enhanced NO3RR performance through the tailored asymmetric RuN2-CuN3 coordination structure strategy. The JNH3 reached − 18.2 mA cm− 2 on RuCu DAs/NGA at -0.3 V vs. RHE, which is an excellent performance. The NH3 yield rate results (Fig. 5d) exhibit that RuCu DAs/NGA is more active and selective than Ru SAs/NGA and Cu SAs/NGA. To ascertain the endurance of RuCu DAs/NGA under the conditions of the NO3RR, a series of 10 continuous electrolysis cycles were executed. Figure 5e illustrates that RuCu DAs/NGA exhibits sustained high Faradaic efficiency and NH3 yield rates across these cycles, substantiating its exceptional durability. The current density loss was at the negligible operated potential of -0.3 V and − 0.5 V vs. RHE during the 24 h continuous NH3 testing from NO3RR (Supplementary Fig. 23). The NH3 yield rate of RuCu DAs/NGA reached 3.07 mg h− 1 cm− 2 at -0.4 V vs. RHE, which still presents an optimal performance compared to similar catalysts in recent reports (Fig. 5f, Supplementary Table 4). Moreover, the yield rate of RuCu DAs/NGA reached 1.61 mg h− 1 cm− 2 and 4.35 mg h− 1 cm− 2 at -0.3 V and − 0.5 V vs. RHE respectively, which is better than the references (Ru SAs/NGA and Cu SAs/NGA) and superior to similar catalysts as well.
To further elucidate the intermediates involved in NO3RR on various electrocatalysts, in situ ATR-SEIRAS measurements were conducted (Supplementary Fig. 24). Figure 5g presents the results for the RuCu DAs/NGA at open circuit potential (OCP), -0.1 V, -0.2 V, -0.3 V, -0.4 V, -0.5 V, and − 0.6 V vs. RHE. A progressively increasing intensity of the NO3− vibrational band at ~ 1245 cm− 1 was observed from OCP to -0.6 V, indicating the ongoing consumption of NO3− during electrolysis. A broad peak centered around 3400 cm− 1 (Supplementary Fig. 25), identified as the *NH2 species, was consistently observed49, 50. Additionally, the peak intensities of hydrogenation intermediates (*NH2 at ~ 1168 cm− 1, *NH2OH at ~ 1115 cm− 1), NH4+ (NH3 at ~ 1460 cm− 1), and deoxidation intermediates (*NO2 at ~ 1630 cm− 1) progressively increased51, 52. These findings suggest that the RuCu DAs/NGA catalyst is highly effective in activating NO3−, subsequently facilitating the formation of substantial amounts of hydrogenation intermediates that are eventually converted into NH4+ (NH3). The detection of *NH2OH and *NH2 species indicates the co-occurrence of both indirect and direct reduction pathways during the NO3RR process on RuCu DAs/NGA. Additionally, a significant enhancement of the H-O-H stretching vibrations at approximately 1660 cm− 1 and 3484 cm− 1 was observed, indicating the dissociation of H2O into OH− and H+ species, which are stabilized on the RuCu DAs/NGA catalyst.
To study the structure-activity relationship of RuCu DAs/NGA on the atomic level, in situ XAFS tests were carried out during the electrochemical catalytic NO3RR process. Figures 5h and 5i present the in situ Cu K-edge and Ru K-edge XANES results of RuCu DAs/NGA at OCP, -0.1 V, -0.2 V, -0.3 V, -0.4 V, and − 0.5 V vs. RHE. The absorption edges of both Cu and Ru in RuCu DAs/NGA tend to gradually move towards lower energy, together with the reduced intensity of their white lines, which implies the decrease in Cu and Ru valence states with the lower applied potentials. According to the results (Supplementary Fig. 26) of the first derivative of the absorption edge for RuCu DAs/NGA, the specific valence states of Cu were decreased from 1.27 to 0.73 when the operated potentials were reduced. Similarly, the valence states of Ru decreased from 2.87 to 2.62 continuously (Fig. 5j). In fact, Cu and Ru disclose similar trends in oxidation states owing to NO3− being adsorbed on the active site of the catalyst in the electrolyte, resulting in the interaction on the unpaired outermost d orbitals of Cu and Ru with the N 2p orbitals of NO3−. The oxidation states of both Cu and Ru would continue to decrease when lower reaction potentials were applied. The Cu-Ru dual active sites would interact with the adsorbed NO3− more easily to form Cu/Ru-N/O bonds at lower potentials, which leads to electron redistribution among Cu, Ru, N, and O atoms. The Cu K-edge FT-EXAFS spectra (Fig. 5k) of RuCu DAs/NGA show that the Cu-N peak shifts from 1.50 Å (OCP) to 1.41 Å (-0.3 V), which was considered as the compressing of Cu-N bonds. The Ru K-edge FT-EXAFS spectra (Fig. 5l) of RuCu DAs/NGA exhibit that the Ru-N peak moves from 1.56 Å (OCP) to 1.53 Å (-0.3 V), with the shrinking of Cu-N bonds as well. In addition, the Ru-Cu bonds also tend to shift to the left during the NO3RR process. Because the metal atoms are likely not in the same plane as graphene, it is possible to have both metal and metal-N bonds clamped. Besides, the Ru-Cu bonds also tend to move to the left slightly during the NO3RR process. The above bond lengths analysis is consistent with the specific fitting results (Supplementary Figs. 27 and 28, and Supplementary Table 5). Due to the metal atoms are likely not on the same plane as graphene, it is possible to be pinched for both the Ru-Cu bonds and the Cu/Ru-N bonds. In brief, the high performance of RuCu DAs/NGA in the electrocatalytic NO3RR is due to the joint effect of the RuN2-CuN3 coordination moieties. During the NO3RR process, the local coordination structure (coordination number and bond length, etc.) near the Ru-Cu sites changed slightly, and a new stable structure would be formed. The detection of the reaction intermediates absorbed by the Ru-Cu active sites might be the main reason.
Theoretical study of NO 3 RR. To understand the fundamental mechanism of the NH3 product reduced from NO3− for RuCu DAs/NGA, the DFT method was used to investigate the whole process of the NO3RR on the asymmetrical RuN2CuN3/C moieties. The differential charge densities (Fig. 6a) of RuN4/C, CuN4/C, and RuN2CuN3/C were calculated to elucidate the electrical properties of Ru and Cu sites and to study the synergistic interactions of the two asymmetric coordination metal atoms as well. The RuN4 site makes it easier to obtain electrons (electron-rich, green area) than the CuN4 site, showing a stronger reducing ability. Therefore, it can be predicted that RuN4 would have a stronger adsorption capacity for intermediates than CuN4. Through the asymmetric deployment of the modulated RuN2-CuN3, the surface charge distribution appears significantly polarized. Electron enrichment near RuN2 and electron deficiency near CuN3 are possibly attributed to the electron transfer from the Cu site to the Ru site. To reveal the underlying reason for the interactions between RuN2-CuN3 sites and the reactive species, the projected densities of state (PDOS) of CuN4/C, RuN4/C, and RuN2CuN3/C concentrating on d orbitals of Ru and Cu were simulated (Supplementary Fig. 29 and Fig. 6b). According to the d band center theory53, 54, the d orbitals of Cu in CuN4 are further away from the Fermi level compared to Ru in RuN4, the adsorption of intermediates on CuN4 is weaker than that on RuN4. The modulated RuN2-CuN3 has a strong synergistic effect, the 4d orbital of Ru is closer to the Fermi level, and it has a stronger adsorption effect for the reaction intermediates, which would be more excellent catalytic activity on RuN2-CuN3. Furthermore, another CuN3 site with a slightly weak adsorption ability may promote the desorption of intermediates, which is conducive to enhancing the reaction rate.
All NO3RR pathways and relative free energy on RuN4/C, CuN4/C, and RuN2CuN3/C at U = 0 vs. RHE are illustrated in Fig. 6c and Supplementary Figs. 30–32. NO3− is adsorbed first and discharged on the metal sites, forming *NO3, then transformed into *NO3H. After that, *NO3H converts into *NO2 with the departure of the OH−. Subsequently, as protons in H2O continue to be added to the intermediate and take O atoms away (forming OH−), *NH3 is gradually formed. Finally, *NH3 is desorbed to yield NH3 leaving the metal sites. The whole reaction path and the relative free energies of each step are shown in Supplementary Table 6. The NO3RR reaction steps can be divided into two parts, namely the reactant or product absorption/desorption process (non-electron gain/loss reaction) and the electron gain/loss reaction steps. From NO3− to *NO3H, since the charge transfer number is 0, the process can be considered as the adsorption of the product by the metal sites. According to the energy position of *HNO3, the adsorption of *HNO3 by RuN2CuN3 is the strongest, followed by RuN4 and CuN4 is the worst. A large amount of reactants would accumulate in the RuN2CuN3 active site, and the increase of reactant concentration is beneficial to the reaction. From *HNO3 to *NH3, the intermediates in the elementary reaction would get an e− at each step. For the elementary reaction of gaining or losing electrons, the smaller the energy barrier is, the easier the reaction is. For RuN2CuN3, RuN4, and CuN4, the respective maximum energy barrier is 0.252 eV (*NHOH→*NH2OH), 0.5 eV (*NO→*NOH), and 0.994 eV (*NO→*NOH) in their electron gain/loss steps (Fig. 6d). Note that the energy barrier of *NO→*NOH on RuN2CuN3 is 0.242 eV. Compared with Cu, Ru reduces the energy barrier of the key step of *NO→*NOH, while the energy barrier of this step is further decreased in the Ru-Cu dual atoms structure. Thus, *NO→*NOH is no longer the key step in determining the reaction rate on RuN2CuN3, while *NHOH→*NH2OH is the critical step. The most critical intermediate is *NO in NO3RR, and reducing the energy barrier of its hydrogenation step is the key to the action of these catalysts. The asymmetric RuN2CuN3 structure exhibits the best effect in the three catalysts. Based on the relative free energy positions of *NO and *NOH, the adsorption of the Cu site to intermediates is too weak, and the Ru site could enhance the adsorption of these intermediates, but the adsorption of *NO is too strong and may not be conducive to the reaction. In the RuN2CuN3 structure, Ru maintains the strong adsorption of *NOH, but weakens the adsorption of *NO, so the energy barrier of this step is reduced. From the differential charge density of the key intermediates (*NO, Supplementary Fig. 33), *NO gets more charge in the Ru-containing system. The main difference between Ru-Cu and single Ru systems is the adsorption configuration, NO is adsorbed horizontally, and Ru-Cu acts together in the RuN2CuN3 system.
For the last step *NH3→*+NH3 is the desorption process of the product, the adsorption of the intermediate at the Cu site is weak, and the product is easier to desorption. The desorption ability of Ru containing system is slightly larger, but it is also less than 1eV, which can be carried out at room temperature. The last step is not critical unless desorption is large reach to poison the metal sites, where the barrier has not yet reached the conditions for poison. Under actual reaction conditions, the product NH3 is highly soluble in water, and the desorption energy of less than 1eV will not poison the metal site. Additionally, the system containing Ru has high selectivity for the NH3 path, almost only along the path of NH3 production. While CuN4/C may produce NO2− and NO (low energy barrier), NH3 selectivity is not high enough. This is consistent with the NO3RR test results.