Nitrate reduction to ammonia catalyzed by GaN/Si photoelectrodes with metal clusters

Abstract

The development of photoelectrochemical cells for reduction of nitrate to ammonia under solar light is of significant interest for the production of clean chemicals and fuels but has remained a daunting challenge. Here, we investigate various metal catalysts supported on GaN nanowires grown on n⁺-p Si wafer – an emerging functional platform for scalable artificial photosynthesis – and demonstrate highly stable and efficient photoelectrochemical nitrate reduction reaction. We find that Co and Ni catalysts on GaN/Si exhibit the best performance, with an onset potential >0.3 V_RHE and a faradaic efficiency of NH₃ of 99% at 0.2 V_RHE. These results highlight the advantage of photoelectrochemical system in achieving efficient nitrate reduction under more positive potentials. In-situ measurements and theoretical calculations reveal that the binding modes of the ${{{\rm{NO}}}}_{2}^{{-}}$ intermediate play a key role in the NH₃ synthetic process. These results demonstrate that the rational design of catalysts on photoelectrodes can construct synergistic metal-semiconductor interactions for efficient and stable photoelectrochemical NH₃ synthesis.

Introduction

Nitrate (${{{\rm{NO}}}}_{3}^{{-}}$) is a common pollutant found in wastewater, industrial runoffs, and groundwater¹. It represents a significant threat to human health since it induces diseases upon conversion to nitrite (${{{\rm{NO}}}}_{2}^{{-}}$) in the body^2,3. The electrochemical ${{{\rm{NO}}}}_{3}^{{-}}$ reduction reaction (${{{\rm{NO}}}}_{3}^{{-}}$RR) is a promising method removing ${{{\rm{NO}}}}_{3}^{{-}}$ by producing ammonia (NH₃), a valuable fertilizer and industrial chemical. The ${{{\rm{NO}}}}_{3}^{{-}}$ RR is also an environment-friendly method to close the ${{{\rm{NO}}}}_{3}^{{-}}$N cycle, reducing the need for the Haber-Bosch process which is energy-intensive and produces greenhouse gases^4,5,6. From the energy perspective, it is advantageous when compared to NH₃ synthesis by N₂ reduction due to the lower dissociation energy of N=O bond (204 kJ/mol) of ${{{\rm{NO}}}}_{3}^{{-}}$ when compared to the N ≡ N bond (941 kJ/mol)^1,7. ${{{\rm{NO}}}}_{3}^{{-}}$-to-NH₃ conversion in aqueous electrolyte is a complex 9-proton and 8-electron reduction reaction, where protons are provided by water dissociation. Therefore, the catalysts should have both activity for the water dissociation and the ${{{\rm{NO}}}}_{3}^{{-}}$ conversion to promote the ${{{\rm{NO}}}}_{3}^{{-}}$ RR⁸. In addition, the ${{{\rm{NO}}}}_{3}^{{-}}$-to-NH₃ reduction is a complicated reaction with different stages of intermediates having various binding modes. Electrocatalysts have been investigated for the ${{{\rm{NO}}}}_{3}^{{-}}$ RR, including metal nanoparticles (NPs)^9,10, alloys^11,12,13, and compounds^{14,15,16,17,18,19,20,21,22,23}, 2D materials²⁴, and single atom catalysts^25,26. However, due to their limited functionality of catalyzing the water dissociation, ${{{\rm{NO}}}}_{3}^{{-}}$ activation, or reduction of intermediates to the NH₃, the overpotential still needs to be significantly improved for any practical consideration. In order to mitigate this issue, multiple functional sites for different catalytic reactions were integrated by decorating one catalyst onto another^18,27. For instance, an electrocatalyst consisting of two phases of Cu/CuO_x and Co/CoO showed promising NH₃ synthetic performance due to efficient ${{{\rm{NO}}}}_{3}^{{-}}$-to-${{{\rm{NO}}}}_{2}^{{-}}$ conversion on Cu/CuO_x and the subsequent ${{{\rm{NO}}}}_{2}^{{-}}$-to-NH₃ conversion on Co/CoO¹⁸. This synergistic relay catalysis or intermediate spillover improved the performance compared to their single phases.

Photoelectrodes, fabricated by coupling the catalysts with photo-responsive semiconductor, can convert solar energy directly to the chemical energy and further reduce the overpotential of the ${{{\rm{NO}}}}_{3}^{{-}}$ RR. Previously, Au-decorated Si nanowires (NWs) photoelectrode demonstrated photoelectrochemical (PEC) ${{{\rm{NO}}}}_{3}^{{-}}$ RR²⁸. However, the maximum production rate of NH₃ (Y_NH3) was only ~0.4 μmol/h/cm² and the stability has remained very limited. Recent advancements in PEC ${{{\rm{NO}}}}_{3}^{{-}}$ RR using chalcogenide light absorbers demonstrated the potential of binary metal catalysts to enhance catalytic activity and efficiency. For instance, CuSn²⁹ and CoCu³⁰ catalysts integrated with Sb₂S₃ and Sb₂Se₃ photocathodes have shown exceptional faradaic efficiency of NH₃ (~97% at 0.4 V_RHE and 88% at −0.2 V_RHE, respectively) and low onset potentials (0.62 V_RHE and 0.43 V_RHE), highlighting their ability to optimize reaction kinetics and improve charge transfer efficiency. These studies illustrate the importance of catalyst design in achieving high performance in PEC ${{{\rm{NO}}}}_{3}^{{-}}$ RR systems.

When catalysts are decorated on the surface of semiconductors, the photoelectrodes have two or more different functional sites at the surface, leading to distinct electronic structures and unique catalytic activity. This determines the reaction thermodynamics, intermediate binding modes, and final products in various catalytic reactions, such as water splitting, carbon dioxide reduction, N₂ reduction, and ${{{\rm{NO}}}}_{3}^{{-}}$ reduction^{31,32,33,34,35,36}. Recent studies have shown that GaN nanostructures, e.g., nanowires (NWs), inherently possess the excellent water dissociation activity due to characteristic Ga-N bond at the m-plane crystal facet^37,38,39. This suggests that GaN NWs can be an ideal supporting materials to promote the proton-coupled electron transfer reaction including ${{{\rm{NO}}}}_{3}^{{-}}$ reduction to NH₃. If the integration of metal catalyst on the semiconductor nanostructures can demonstrate synergetic effects by reducing the thermodynamic energy barriers, controlling the binding modes of adsorbates, and providing a sufficient number of protons with effective transfer of photogenerated electrons, the photoelectrodes will further reduce the overpotential and enhance the productivity of NH₃ compared to electrocatalysts. Therefore, rational design of photoelectrodes by selecting a suitable catalyst on the semiconductor surface and the understanding of reaction pathways are urgently required to achieve more efficient and productive solar-driven ${{{\rm{NO}}}}_{3}^{{-}}$ conversion to NH₃.

In this work, we investigated the metal catalysts supported on GaN NWs vertically grown on n⁺-p Si photoelectrodes for PEC ${{{\rm{NO}}}}_{3}^{{-}}$ RR to NH₃ to demonstrate the synergistic effects of metal catalysts and GaN NWs. We first screened various noble metals (Pt, Ir, Rh, Pd, Au, Ru, Ag) and transition metals (Cu, Ni, Co) on GaN NWs for NH₃ synthesis. Co-loaded GaN/Si (Co/GaN/Si) showed the highest faradaic efficiency of NH₃ (FE_NH3 > 90%) at the widest potential window (0.2 to −0.7 V_RHE). Ni/GaN/Si achieved the NH₃ production rate of 83.3 μmol/h/cm² and the maximum value of 201.6 μmol/h/cm² at potentials of 0 and −0.4 V_RHE, respectively. We characterized the performance of photoelectrodes based on the presence of GaN NWs and found that the catalysts loaded on GaN NWs showed much lower onset potential and higher productivity of NH₃ than those on planar n⁺-p Si photoelectrodes. This is due to the synergetic metal-GaN interaction which enhances water dissociation and catalytic activity of nitrogen-containing intermediates. GaN NWs rapidly dissociate water, providing a sufficient number of protons near the surface, which in turn activates ${{{\rm{NO}}}}_{3}^{{-}}$ and facilitates its efficient conversion to ${{{\rm{NO}}}}_{2}^{{-}}$. Then, following ${{{\rm{NO}}}}_{2}^{{-}}$ to NH₃ conversion is further promoted by catalysts. A series of in-situ experiments and theoretical calculations were performed to interpret reaction mechanisms and revealed that the binding modes of the NO₂⁻ intermediate (*ONO⁻* or *NO₂⁻) on the metal catalysts were the key to the further deoxygenation process. This work provides insights into the design of efficient photoelectrode for green NH₃ production.

Results

Metal catalysts loaded on GaN/Si photoelectrodes

Photoelectrodes were fabricated by plasma-assisted molecular-beam epitaxy growth of GaN NWs on n⁺–p Si wafers, followed by photodeposition of metal catalysts on GaN NWs. A 45-degree tilt-view scanning electron microscopy (SEM) image of Co/GaN/Si showed that GaN NWs were vertically aligned on the n⁺-p Si wafer, with a length of ~400 nm and diameter of ~50 nm (Fig. 1a). Since the particle size of metal catalysts was too small to be observed in the SEM image, we performed high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and energy-dispersive spectroscopy (EDS) elemental mapping to image the Co/GaN/Si (Fig. 1b and Fig. S1) and Ni/GaN/Si photoelectrodes (Fig. 1c and Fig. S2). The HAADF-STEM images showed the segregated Co and Ni nanoparticles (NPs) on both the top and sidewalls of GaN NWs. The atomic number difference between Co (27), Ni (28), and Ga (31) results in a Z-contrast that allows the NPs to be distinguished from the GaN NWs. The EDS elemental maps confirmed the presence of the metal catalysts, with locally aggregated Co (shown in yellow in Fig. 1b) and Ni (shown in orange in Fig. 1c). Ga (blue) and N (cyan) were uniformly distributed over the GaN NWs. The presence of oxygen indicates that the surface metal catalysts were oxidized prior to the analysis.

Fig. 1: Material characterization of photoelectrodes.

a 45 degree-tilt-view SEM image of Co/GaN/Si. STEM-HADDF images and EDS elemental maps of b Co/GaN and c Ni/GaN. Ga (blue) and N (cyan) were uniformly detected on GaN NWs while Co (yellow), Ni (orange), O (green) elements were segregated on the surface of GaN NWs. XPS spectra of d Ga 2p_2/3, e N 1 s, f Co 2p, g Ni 2p, and h O 1s for GaN/Si, Co/GaN/Si, and Ni/GaN/Si.

To investigate the chemical bonding states of photoelectrodes with different metal catalysts, X-ray photoelectron spectroscopy (XPS) was performed. The Ga 2p_3/2 XPS spectra were deconvoluted into a major peak of the Ga-N bond (1116.8 eV) and a minor peak of the Ga-O bond (1118.0 eV) (Fig. 1d)⁴⁰. Similarly, the N 1s XPS spectra could be separated into a major N-Ga bond (396.9 eV) and a minor N-H bond (398.5 eV), likely due to native oxide or hydroxide on the surface (Fig. 1e)⁴¹. It should be noted that although Co, Ni, or other catalysts were loaded on the GaN NWs, the intensities of the Ga 2p_3/2 and N 1s spectra were almost preserved. This means that the metal catalysts were partially covered over the GaN and the GaN surface was still exposed. The Co catalysts exhibited Co²⁺ (782.7 and 798.3 eV) and Co³⁺ (781.2 and 796.8 eV) oxidation states in the Co 2p spectrum (Fig. 1f). The Ni catalysts revealed Ni-O (853.5 and 871.1 eV), Ni-OH (854.7 and 872.3 eV), and their satellite bonds in the Ni 2p spectrum (Fig. 1g)^42,43. In the O 1 s XPS spectra, O-Co and O-Ni bonds from the oxidized phases of the catalysts were observed (Fig. 1h). This is in good agreement with the STEM-EDS results. The metal oxides were also formed on other metal catalysts, such as Pt, Ir, Rh, Pd, Cu, Ag, Ru, and Au (Figs. S3–S10), since the catalysts were deposited under aqueous conditions and exposed to ambient air prior to the analysis.

Photoelectrochemical ${{{\rm{NO}}}}_{3}^{{-}}$ reduction reaction

Photoelectrochemical (PEC) nitrate reduction to ammonia (${{{\rm{NO}}}}_{3}^{{-}}$RR) was performed in an H-type cell under AM1.5G 1-sun solar light. When light was illuminated on the n⁺-p Si wafer through optically transparent GaN NWs, electrons and holes were generated in the conduction and valance bands of Si, respectively (Fig. 2a). The conduction bands of n-Si and n-GaN are approximately aligned⁴⁴, so the photogenerated electrons in n-Si spontaneously migrate to the conduction band of n-GaN by the built-in potential of the Si p-n junction. These photoelectrons then move to the metal catalysts and participate in reduction reaction. After the reaction was completed, the liquid products of NH₃, ${{{\rm{NO}}}}_{2}^{{-}}$, and N₂H₄ in the cathodic compartment of the reactor were quantified through calibrated spectrocolorimetric measurements (Figs. S11–S13). The gas products were analyzed by gas chromatography.

Fig. 2: Catalyst-dependent photoelectrochemical nitrate reduction.

a Energy band diagram of metal/GaN/Si photoelectrode under light illumination. b Faradaic efficiency of NH₃ for metal catalysts loaded on GaN/Si. The measurement was conducted in 0.1 M K₂SO₄ with 0.5 M KNO₃ at various cathodic potentials (0.3 to −0.8 V_RHE) under 1 sun light illumination. Faradaic efficiency of H₂, NH₃, ${{{\rm{NO}}}}_{2}^{{-}}$, and N₂H₄ for c GaN/Si, d Cu/GaN/Si, e Co/GaN/Si, and f Ni/GaN/Si. Co and Ni catalysts showed the highest FE_NH3 (>98%) among the tested samples.

In the electrolyte, a substantial number of water molecules coexist with ${{{\rm{NO}}}}_{3}^{{-}}$ ions. Therefore, when applying a cathodic potential to the photoelectrode, reduction reactions of ${{{\rm{NO}}}}_{3}^{{-}}$ conversion and hydrogen evolution may occur. To evaluate the photoelectrodes with respect to these two competing reactions, we performed linear sweep voltammetry (LSV) measurement of GaN/Si in 0.1 M K₂SO₄ with and without 0.5 M KNO₃ (Fig. S14). The onset potential (V_onset) at −1 mA/cm² for GaN/Si shifted positively from −0.63 to −0.21 V_RHE and the photocurrent density was as high as ~40 mA/cm² at −0.8 V_RHE when 0.5 M KNO₃ was added to 0.1 M K₂SO₄. Moreover, GaN/Si exhibited fairly high FE_NH3 = 61% at −0.7 V_RHE, suggesting that GaN surface is more catalytically active for ${{{\rm{NO}}}}_{3}^{{-}}$ reduction to NH₃ (${{{\rm{NO}}}}_{3}^{{-}}$RR) than for the hydrogen evolution reaction. However, Pt, Ir, and Rh catalysts (Figs. S15–S17) loaded on GaN/Si showed little difference in V_onset and photocurrent density regardless of ${{{\rm{NO}}}}_{3}^{{-}}$ in the electrolyte since these metals possess a high affinity for hydrogen adsorption rather than ${{{\rm{NO}}}}_{3}^{{-}}$ and good catalytic activity for hydrogen evolution reaction. As a result, these three metal catalysts showed very low faradaic efficiency of NH₃ (FE_NH3) <13% with high FE of H₂ (FE_H2) >80% at all potentials (0.3 to −0.8 V_RHE) (Fig. 2b). The photoelectrodes with Pd, Ag, Ru, and Cu catalyst showed FE_NH3 of 37%, 41%, 60%, and 13%, respectively at −0.7 V_RHE (Figs. S18–S21). The predominant side product was ${{{\rm{NO}}}}_{2}^{{-}}$, indicating the good catalytic activity in converting ${{{\rm{NO}}}}_{3}^{{-}}$ to ${{{\rm{NO}}}}_{2}^{{-}}$ compared to hydrogen evolution, but the further reduction to NH₃ was limited. Interestingly, when Au, Co, and Ni catalysts were coated on GaN/Si, the maximum FE_NH3 can approach ~100% (Figs. S22–S24).

In more detail, GaN/Si selectively converted ${{{\rm{NO}}}}_{3}^{{-}}$ to ${{{\rm{NO}}}}_{2}^{{-}}$ at low potential (−0.1 V_RHE), with FE_NO2- = 74% (Fig. 2c). As the potential was increased to −0.7 V_RHE, FE_NO2- monotonically decreased to 35% and FE_NH3 gradually increased to 61%. This demonstrated that GaN has excellent catalytic activity for the conversion of ${{{\rm{NO}}}}_{3}^{{-}}$ to ${{{\rm{NO}}}}_{2}^{{-}}$ and can further reduce ${{{\rm{NO}}}}_{2}^{{-}}$ to NH₃ under large cathodic potential. In the case of Cu catalysts supported on GaN/Si, ${{{\rm{NO}}}}_{2}^{{-}}$ was selectively produced (Fig. 2d) with high FE_NO2- = 94% at 0.3 V_RHE. However, unlike GaN/Si, ${{{\rm{NO}}}}_{2}^{{-}}$ was still the major product (FE_NO2- = 75%) with FE_NH3 = 13% and FE_N2H4 = 14% at −0.7 V_RHE. This suggests that the Cu catalyst coated on GaN has a negative effect on NH₃ synthesis by desorbing the ${{{\rm{NO}}}}_{2}^{{-}}$ intermediate from the surface and preventing further reduction of ${{{\rm{NO}}}}_{2}^{{-}}$. This is discussed further in our theoretical analysis. In stark contrast, Co/GaN/Si dramatically improved FE_NH3 to 72% and 99% at 0.3 and 0.2 V_RHE, respectively (Fig. 2e). These potentials are more positive than the standard reduction potential of ${{{\rm{NO}}}}_{3}^{{-}}$, meaning that the photovoltage generated by GaN/n⁺-p Si photoelectrode (~0.5 V)^37,38,39 biases the reduction reaction and shifts the operating potential window positively. At the large cathodic potential of −0.8 V_RHE, there was a decline of FE_NH3 to 68% due to increase in the FE_H2. In contrast, Ni/GaN/Si selectively produced NH₃ over other competing reactions although Ni/GaN/Si required a more negative potential (0 V_RHE) to achieve FE_NH3 > 90% (Fig. 2f). The total faradaic efficiency was less than 100% especially at anodic potentials (+0.2–0 V_RHE) because some electrons could be used for the reduction of ${{{\rm{NO}}}}_{3}^{{-}}$ to other nitrogen-containing species, such as N₂ and NO, which were not fully analyzed. Additionally, the non-faradaic capacitive current density from ionic adsorption and desorption at the electrical double layer could affect the derivation of faradaic efficiency. We further investigated the effects of the loading amount of metal catalysts coated on GaN NWs on catalytic properties (Fig. S25). Within the precursor concentration range in this work, there was no significant difference in catalytic performance, indicating that the determining factor for catalyst characteristics is the type of metal catalyst rather than the loading amount of metal catalysts.

To further study the metal-support interactions, we compared the LSV curves of n⁺-p Si, GaN/Si, and metal catalysts loaded n⁺-p Si and GaN/Si in 0.1 M K₂SO₄ with 0.5 M KNO₃ (Fig. 3a). In the order of Si, Co/Si, and Ni/Si, V_onset and photocurrent density were improved. GaN/Si showed a 1.1 V positive shift of V_onset compared to Si. Remarkably, GaN/Si loaded with Co and Ni catalysts showed much better V_onset = 0.22 and 0.27 V_RHE, respectively. Charge transfer kinetics from the photoelectrodes to reactants were investigated by the PEC impedance spectroscopy at 0 V_RHE under solar irradiation (Fig. 3b). The radius of the semicircle in the Nyquist plot correlates to the charge-transfer resistance during the PEC reaction⁴⁵. For Si and Co/Si, the imaginary part of impedance diverged as the real part of impedance increased, indicating no faradaic charge transfer between the photoelectrode and the electrolyte. In contrast, for Ni/Si, GaN/Si, Co/GaN/Si, and Ni/GaN/Si, the radius of the semicircle decreased sequentially, implying a gradual reduction in charge transfer resistance. For GaN/Si, Co/GaN/Si, and Ni/GaN/Si photoelectrodes, the Nyquist plots were further analyzed at a more negative potential of −0.2 V_RHE (Fig. S26). Under this condition, the radius of the semicircle significantly reduced, indicating accelerated charge transfer under high bias. From the impedance results, the synergetic metal-GaN interaction facilitated the ${{{\rm{NO}}}}_{3}^{{-}}$ RR likely due to catalytic enhancement and increased electrochemical surface area (Fig. S27).

Fig. 3: Effects of GaN NWs on photoelectrochemical nitrate reduction.

a LSV curves and b Nyquist impedance plots of Si, Co/Si, Ni/Si, GaN/Si, Co/GaN/Si, and Ni/GaN/Si measured in 0.1 M K₂SO₄ with 0.5 M KNO₃ under 1 sun light illumination. c Faradaic efficiency and d production rate of NH₃ plotted as a function of cathodic potentials. Co/GaN/Si exhibited the widest potential window (−0.2 to −0.7 V_RHE) for high FE_NH3 > 90% and Ni/GaN/Si revealed the best Y_NH3 = 201.6 μmol/cm²/h at −0.4 V_RHE.

It should be noted that Co/GaN/Si showed the widest potential window (0.2 to −0.7 V_RHE) for the selective NH₃ production (FE_NH3 > 90%) and Ni/GaN/Si also exhibited selective conversion of ${{{\rm{NO}}}}_{3}^{{-}}$ to NH₃ (FE_NH3 > 90%) at potential ≤ 0 V_RHE (Fig. 3c). However, n⁺-p Si (Fig. S28) and those with Co (Fig. S29) and Ni (Fig. S30) catalysts showed low FE_NH3 < 54% with noticeably increased FE_H2 > 42% or FE_NO2- >43% at −0.8 V_RHE. This clearly demonstrated that GaN NWs play an important role in reducing the overpotential and increasing the NH₃ selectivity of ${{{\rm{NO}}}}_{3}^{{-}}$ RR. Since Co/GaN/Si and Ni/GaN/Si exhibited the promising V_onset (>0.3 V_RHE), high photocurrent density (>38 mA/cm²), and FE_NH3 (~100%) over the wide potential window, high production rates of 17.5 and 83.3 μmol/h/cm² were achieved, respectively at 0 V_RHE. In addition, the maximum production rate of NH₃ = 166.7 μmol/h/cm² at −0.7 V_RHE and 201.6 μmol/h/cm² at −0.4 V_RHE were demonstrated, respectively (Fig. 3d). Applied bias photon-to-current efficiency (ABPE) was calculated for the tested photoelectrodes (Fig. S31). In the order of Si, Co/Si, Ni/Si, GaN/Si, Co/GaN/Si, and Ni/GaN/Si, the maximum ABPE gradually increased up to 2.41%, indicating that GaN NWs are the promising platforms for loading the metal catalysts and facilitating the ${{{\rm{NO}}}}_{3}^{{-}}$RR. The higher ABPE and NH₃ production rate for Ni/GaN/Si over Co/GaN/Si originated from more positive flat band potential likely due to more efficient electron injection from GaN NWs to Ni catalysts and better alignment for electron transfer from the Ni catalyst to the reactants (Fig. S32). Transient photocurrent measurements were conducted to further investigate charge carrier dynamics of photoelectrode with and without catalyst (Fig. S33). The results revealed that Co/GaN/Si exhibited superior charge retention and reduced recombination compared to GaN/Si, suggesting the critical role of Co catalyst in improving charge transfer efficiency and sustaining PEC activity during ${{{\rm{NO}}}}_{3}^{{-}}$RR.

To investigate the influence of water dissociation in ${{{\rm{NO}}}}_{3}^{{-}}$RR, we conducted kinetic isotopic effect (KIE) experiments by changing H₂O and D₂O in the electrolyte (Fig. S34). The KIE value was then calculated as the ratio of NH₃ production rates in electrolytes with H₂O and D₂O. For Co/Si, the KIE value was calculated at 1.63, indicating that the reaction was constrained by water dissociation⁴⁶. Upon the introduction of GaN NWs, the KIE value for Co/GaN/Si significantly dropped to 1.15. Once protons became readily available at the surface through water dissociation by GaN NWs, it appears that Co catalysts efficiently reduce ${{{\rm{NO}}}}_{3}^{{-}}$ to NH₃ through deoxygenation and hydrogenation processes due to the intrinsic selectivity for producing NH₃ rather than hydrogen evolution. Consequently, using GaN nanowires as a support for metal catalysts presents a promising approach to demonstrating the synergy between proton supply and the catalysis of nitrogen-containing intermediates.

Control experiments in 0.1 M K₂SO₄ were carried out using Co/GaN/Si to confirm whether NH₃ is produced from ${{{\rm{NO}}}}_{3}^{{-}}$ reactant, and there was no NH₃ produced without ${{{\rm{NO}}}}_{3}^{{-}}$ in the electrolyte (Fig. S35). Moreover, ¹⁵N isotope labeling clearly evidenced that ¹⁵NH₃ was produced from ¹⁵${{{\rm{NO}}}}_{3}^{{-}}$ (Fig. S36). The measured V_onset, FE_NH3, and Y_NH3 of Co or Ni catalysts on GaN NWs were comparable to those of recent photoelectrodes and superior to state-of-the-art electrocatalysts for ${{{\rm{NO}}}}_{3}^{{-}}$RR (Table S1).

To study the effects of metal catalysts on catalytic ${{{\rm{NO}}}}_{3}^{{-}}$RR activity, in-situ Fourier transform infrared (FTIR) measurements were performed for two representative photoelectrodes of GaN/Si (Fig. S37) and Co/GaN/Si (Fig. 38). The potential-dependent spectra (−0.2 to −0.7 V_RHE, Figs. S37a, S38a) revealed distinct behaviors between the two photoelectrodes. As ${{{\rm{NO}}}}_{3}^{{-}}$RR proceeded, both samples exhibited a gradual decrease in the peak intensity at 1348 cm^−1 47, which attributes to the consumption of ${{{\rm{NO}}}}_{3}^{{-}}$ ions in the solution. For GaN/Si (Fig. S37a), a characteristic band at 1230 cm⁻¹ corresponding to ${{{\rm{NO}}}}_{2}^{{-}}$ intermediates, was clearly observed. This band became more pronounced at more negative potentials, indicating the accumulation of ${{{\rm{NO}}}}_{2}^{{-}}$ as a primary product. Time-resolved measurements at −0.5 V_RHE (Fig. S37b) further confirmed the buildup of ${{{\rm{NO}}}}_{2}^{{-}}$ near the surface as the intensity of ${{{\rm{NO}}}}_{2}^{{-}}$ band gradually increased over time. In contrast, Co/GaN/Si (Fig. S38a) demonstrated no signal of the ${{{\rm{NO}}}}_{2}^{{-}}$ band, even at highly cathodic potentials, suggesting efficient conversion of NO₂ intermediates. This trend was further supported by the time-resolved measurements (Fig. S38b), where the absence of the ${{{\rm{NO}}}}_{2}^{{-}}$ band highlighted the catalytic role of Co catalyst in facilitating the PEC ${{{\rm{NO}}}}_{3}^{{-}}$RR to NH₃.

Stability of Co/GaN/Si and Ni/GaN/Si

We recorded the photocurrent density of Co/GaN/Si (Fig. 4a) and Ni/GaN/Si (Fig. 4b) at 0 and −0.6 V_RHE under solar irradiation. At 0 V_RHE, the reaction was conducted for 5 cycles in 0.1 M K₂SO₄ with 0.5 M KNO₃ and each cycle was performed for 1 h with repeated replacement of fresh electrolyte. At −0.6 V_RHE, 10 cycles of reactions were performed for 5 h. The average photocurrent densities of Co/GaN/Si were 3.5 mA/cm² at 0 V_RHE and 35.8 mA/cm² at −0.6 V_RHE. Ni/GaN/Si showed higher average photocurrent densities of 16.1 mA/cm² at 0 V_RHE and 39.2 mA/cm² at −0.6 V_RHE. Both Co/GaN/Si (Fig. 4c) and Ni/GaN/Si (Fig. 4d) showed very high FE_NH3 > 91% for 10 h. Co/GaN/Si continuously produced NH₃ with Y_NH3 = 18.0 and 162.6 μmol/cm²/h at 0 and −0.6 V_RHE, respectively. Ni/GaN/Si also consistently converted ${{{\rm{NO}}}}_{3}^{{-}}$ to NH₃ with Y_NH3 = 73.5 μmol/cm²/h at 0 V_RHE. At −0.6 V_RHE, Y_NH3 of Ni/GaN/Si was ~170 μmol/cm²/h. There was no noticeable degradation of catalytic activity or change of morphology during the reaction, demonstrating that metal catalysts supported on GaN/Si photoelectrodes are promising platform for efficient and stable ${{{\rm{NO}}}}_{3}^{{-}}$RR to NH₃. Evaluating the stability at high potential of −0.6 V_RHE is crucial from a material perspective since the accelerated degradation due to the detachment of metal catalysts from the photoelectrodes under harsh reaction conditions has been a significant issue⁴⁸. The excellent stability of both Co/GaN/Si and Ni/GaN/Si highlights the potential advantage for designing a practical device.

Fig. 4: Stability test of photoelectrodes.

Photocurrent density of a Co/GaN/Si and b Ni/GaN/Si in 0.1 M K₂SO₄ with 0.5 M KNO₃ under 1 sun light illumination. Measurements were performed 5 times at 0 V_RHE for 1 h each, and then 10 times at −0.6 V_RHE for 30 min each. Insets show SEM images of photoelectrodes after 10 h reaction. FE_NH3 and Y_NH3 for c Co/GaN/Si and d Ni/GaN/Si.

Theoretical calculation

DFT investigations of the catalytic mechanism were performed to elucidate the origin of the different catalytic activities of Co and Cu catalysts for PEC ${{{\rm{NO}}}}_{3}^{{-}}$RR. While the Cu catalyst maintains its metal character (see the XPS data in Fig. S7c that shows predominantly Cu⁰), the Co derived cluster is oxidized to Co₃O₄ as seen in the XPS spectrum (Fig. 1f) which shows Co²⁺ and Co³⁺ peaks. The same holds true for Ni, where the XPS spectrum shows peaks for both Ni-O and Ni-OH (Fig. 1g). In addition to the mechanism on metallic Cu (Fig. S39 and Table S2) and Co (Fig. S40 and Table S3), we therefore performed a mechanistic study of ${{{\rm{NO}}}}_{3}^{{-}}$ reduction on Co₃O₄ (Figs. S41–S44 and Table S4) as well as NiOOH with Ni³⁺ and Ni(OH)₂ with Ni²⁺ (Figs. S45–S47 and Table S5).

The intermediates involved in the PEC process for Cu are shown in Fig. 5a, with the energetics of the preferred reaction path shown in Fig. 5b. The initial reduction of ${{{\rm{NO}}}}_{3}^{{-}}$ proceeds through oxidation of the surface and subsequent reductive removal of oxygen to regenerate the active material. The formal reactions in the PEC reduction of ${{{\rm{NO}}}}_{3}^{{-}}$ to NH₃ on a metal surface such as Cu and Co are given in Eqs. (1) to (11).

$${*{NO}}_{3}^{-}+*{H}^{+}+2{Me}(0) \to*{{NO}}_{2}^{-}+({{Me}(I)}_{2}{{OH}})^{+}$$

(1)

$${*{NO}}_{2}^{-}+({{Me}(I)}_{2}{{OH}})^{+}+{H}^{+}+{e}^{-} \to {*{NO}}_{2}^{-}+{Me}\left(0\right)+{{Me}\left(I\right)}^{+}+{H}_{2}O$$

(2)

$${*{NO}}_{2}^{-}+{{Me}(I)}^{+}+{H}^{+}+{e}^{-}\to \,{*{NO}}_{2}^{-}+{*H}^{+}+{Me}(0)$$

(3)

$${*{NO}}_{2}^{-}+\ast {H}^{+}+{Me}\left(0\right)\to*{NO}+{Me}\left(I\right){OH}$$

(4)

$$*{NO}+{Me}\left(I\right){OH}+{H}^{+}+{e}^{-}\to*{NO}+{Me}\left(0\right)+{H}_{2}O$$

(5)

$$*{NO}+{H}^{+}+{e}^{-}\to*{NHO}$$

(6)

$$*{NHO}+{H}^{+}+{e}^{-}\to*{{NH}}_{2}O$$

(7)

$$*{{NH}}_{2}O+{H}^{+}+{e}^{-}\to*{{NH}}_{2}{OH}$$

(8)

$$*{{NH}}_{2}{OH}+{H}^{+}+{e}^{-}\to*{{NH}}_{2}+{H}_{2}O$$

(9)

$$*{{NH}}_{2}+{H}^{+}+{e}^{-}+{{HNO}}_{3}\to {*{NO}}_{3}^{-}+\ast {H}^{+}{+{NH}}_{3}$$

(10)

$${{HNO}}_{3}+{8H}^{+}+8{e}^{-}\to {{NH}}_{3}+3{H}_{2}O$$

(11)

While the reduction of ${{{\rm{NO}}}}_{3}^{{-}}$ to ${{{\rm{NO}}}}_{2}^{{-}}$ is thermodynamically favorable on both metallic Cu and Co surfaces (Fig. 5b and S40, respectively), further reduction to the NO intermediate and subsequent NH₃ production depends on the thermal accessibility of the *NO₂⁻ adsorption mode at this crucial intermediate. This thermal equilibrium between nonactive *ONO⁻* and reactive *NO₂⁻ adsorption modes, which does not involve an electron transfer and therefore cannot be supported by an applied potential, has different thermodynamic properties on Co and Cu catalysts. For the Cu catalyst, the NO₂⁻ intermediate prefers to bind to the surface in the *ONO⁻* adsorption mode. However, for the second N = O bond dissociation and subsequent deoxygenation, the *ONO⁻* mode must be switched to the *NO₂⁻ adsorption mode. The switching to the *NO₂⁻ mode on the Cu catalyst is highly endergonic, with a free energy change of 0.14 eV (=13.15 kJ/mol) (Fig. 5c). This means that the *NO₂⁻ mode is not accessible on the Cu catalyst at room temperature and can only be generated as promoted by the energy released from the dissociation reaction. Furthermore, switching from the *ONO⁻* mode to the *NO₂⁻ adsorption mode is in competition with desorption of ${{{\rm{NO}}}}_{2}^{{-}}$ from the surface. This process of expulsion and replacement of the ${{{\rm{NO}}}}_{2}^{{-}}$ by an additional ${{{\rm{NO}}}}_{3}^{{-}}$ is slightly exergonic on Cu catalyst by −0.04 eV (=−3.36 kJ/mol). This much lower energy for the replacement of ${{{\rm{NO}}}}_{2}^{{-}}$ with ${{{\rm{NO}}}}_{3}^{{-}}$ in comparison to switching to the reactive *NO₂⁻ intermediate mode explains why the main product of ${{{\rm{NO}}}}_{3}^{{-}}$RR on Cu is ${{{\rm{NO}}}}_{2}^{{-}}$. Once the ${{{\rm{NO}}}}_{2}^{{-}}$ is formed, it is much more likely to be desorbed rather than further reduction to NO through dissociation of the N = O bond.

Fig. 5: Main reaction pathways in the reduction of $N{O}_{3}^{-}$ on Cu from DFT mechanistic study.

a Full mechanism of ${{{\rm{NO}}}}_{3}^{{-}}$ reduction on the catalyst surfaces including reduction to NH₃ and incomplete reduction to ${{{\rm{NO}}}}_{2}^{{-}}$ and subsequent desorption. The key intermediate and its two binding modes determining the outcome are encircled. b Reduction from ${{{\rm{NO}}}}_{3}^{{-}}$ to ${{{\rm{NO}}}}_{2}^{{-}}$ and subsequent desorption on Cu catalyst. c Thermodynamics of desorption and further reduction through change of binding mode on Cu catalyst.

For the oxidized species of Co₃O₄ and Ni(OH)₂, we found different reaction mechanism: reductive formation of defects which leads to the subsequent interaction with oxygen containing species, resulting in dissociation of N-O bond. This is different from the metal surfaces where dissociation of the N-O bond led to oxidation of the surface that needed to be reactivated by reduction. Therefore, the order of reaction, with first reducing the surface and then oxidizing it by reaction with the oxygen containing species is reversed on the oxidized species in comparison to the metallic surfaces. On the Co₃O₄, defect formation is favorable with −0.06 eV for first hydrogenation and −0.55 eV for water release under additional reduction (Fig. S42). Nitrate, nitrite, and NO can directly react with formed oxygen defects (Fig. S43). Notably, the *ONO⁻* adsorption mode can dissociate since the defect can directly interact with the adsorbed oxygen. The reaction scheme and energetics of nitrate reduction on Co₃O₄ can be seen in Fig. 6, and the formal reactions in the PEC reduction of ${{{\rm{NO}}}}_{3}^{{-}}$ to NH₃ on Co₃O₄ are given in Fig. S44. Upon adsorption of nitrate on the surface, a defect is reductively formed, that reacts with the adsorbed ${{{\rm{NO}}}}_{3}^{{-}}$ to extract one oxygen, dissociating the first N-O bond and refilling the defect. This leads to *NO₂ on the surface, that gets reduced to ${{{\rm{NO}}}}_{2}^{{-}}$. Further defect formation leads to dissociation of the second N-O bond, resulting in *NO. Further reduction of NO to NH₃ is thermodynamically downhill. Therefore, full reduction to NH₃ is the energetically favored pathway on Co₃O₄ catalyst, making NH₃ the main product.

Fig. 6: Reaction pathways in the reduction of $N{O}_{3}^{-}$ on Co3O₄ from DFT mechanistic study.

a Full mechanism of ${{{\rm{NO}}}}_{3}^{{-}}$ reduction on Co₃O₄. Defect formation leads to N-O bond dissociation. b Free energy of the reduction from ${{{\rm{NO}}}}_{3}^{{-}}$ to NH₃ on Co₃O₄.

As shown in Fig. S45, NiOOH readily reduces towards Ni(OH)₂ at low applied potentials, meaning the active species is most likely Ni(OH)₂ (Fig. S46). As Ni(OH)₂ has no active Ni surface sites available, oxygen defect formation is necessary for its reactivity. Ni(OH)₂ is a base that can be attacked by acids. The reaction mechanism determined by DFT is shown in Fig. 7a and the corresponding free energy profiles of the reaction steps are shown in Fig. 7b. HNO₃ can react with Ni(OH)₂, releasing water and leaving a positively charged surface with an oxygen defect. This defect directly interacts with the remaining nitrate to form the Ni(OH)O*NO₂ intermediate with one of the oxygen atoms from nitrate being incorporated into the surface. Further reduction breaks the N-O bond, restoring the Ni(OH)₂ and releasing NO₂. Two NO₂ molecules can readily react with water and form HNO₃ and HNO₂. These can then interact with the surface again, releasing water and forming Ni(OH)O*NO species. Upon reduction, NO is released, followed by protonation and water release. The NO then reacts, forming Ni(OH)O*N intermediate that is subsequently reduced to Ni(OH)O*NH and Ni(OH)O*NH₂. Continued reduction leads to release of NH₃, leaving behind a hydrogen defect that is filled in the next reduction step, completing the cycle. All steps are thermodynamically downhill with the exception of the release of NO that is 0.02 eV uphill- an energy easily provided by the photoexcited electrons. Therefore, the main product of nitrate reduction on Ni(OH)₂ is NH_3. The full reaction paths and energetics of reduction to NH₃ as well as reduction to ${{{\rm{NO}}}}_{2}^{{-}}$ and subsequent desorption for all catalytic systems are given in Figs. S39–S47 and summarized in Tables S2–S6.

Fig. 7: Reaction pathways in the reduction of $N{O}_{3}^{-}$ on Ni(OH)₂ from DFT mechanistic study.

a Mechanism of ${{{\rm{NO}}}}_{3}^{{-}}$ reduction to NH₃ on Ni(OH)₂. Defect formation allows for attachment of nitrate, further reduction regenerates the Ni(OH)₂ surface and dissociates the N-O bond. The mechanism proceeds similarly with HNO₂ and NO, with full reduction towards NH₃. b Free energy of the ${{{\rm{NO}}}}_{3}^{{-}}$ reduction on Ni(OH)₂.

Discussion

In summary, we have investigated various metal catalysts for the PEC ${{{\rm{NO}}}}_{3}^{{-}}$RR to NH₃. Remarkably, we found that first-row earth-abundant transition metals Co and Ni catalysts outperformed noble metals (Pt, Ir, Rh, Pd, Ru, Au, and Ag) and the widely used Cu catalyst. In-situ IR measurement showed that a large number of ${{{\rm{NO}}}}_{2}^{{-}}$ intermediates were desorbed from Cu catalysts but further reduced to NH₃ on Co catalysts. Theoretical calculations confirmed that the deoxygenation process of the *NO₂⁻ intermediate is feasible on Co₃O₄ and Ni(OH)₂ catalysts via reaction with oxygen defects. In contrast, ${{{\rm{NO}}}}_{2}^{{-}}$ molecules desorbed from the Cu surface due to the *ONO⁻* binding mode with a lower barrier for desorption than for switching to the *NO₂⁻ mode that is needed for further deoxygenation processes. This leads to incomplete reduction of nitrate to nitrite on Cu, while the Co and Ni containing catalysts are capable of full reduction of ${{{\rm{NO}}}}_{3}^{{-}}$ to NH₃. Owing to the strong catalytic activities of Co and Ni and the synergistic effects of metal-GaN interactions, high performance was achieved, including a low onset potential V_onset = 0.3 V_RHE, high faradaic efficiency of NH₃ (~99% at 0.2 V_RHE), and high production rate of NH₃ (201.6 μmol/h/cm² at −0.4 V_RHE). Moreover, the photoelectrodes were stable and performed for 10 h of reaction without any noticeable degradation. The photoelectrodes, which consist of the two most popular semiconductors (Si and GaN) and inexpensive transition metals (Co and Ni), have the potential to be used in industry-scale conversion of ${{{\rm{NO}}}}_{3}^{{-}}$ pollutant to value-added NH₃ under solar light.

Methods

Preparation of GaN nanowires on n⁺-p Si wafer

Starting p-type Si (100) wafer was purchased from Purewafer and has a resistivity of 1–10 Ω·cm. The n⁺-p Si was demonstrated by spin-coating of phosphorus dopant on the front side of the polished p-type Si wafer and then thermal annealing at 950 °C under nitrogen atmosphere for 4 h. N-type GaN NWs were grown on n⁺-p Si wafer by plasma-assisted molecular-beam epitaxy under nitrogen-rich condition to promote the formation of an N-terminated surface. The growth temperature was 790 °C and the duration was ~2 h. The forward plasma power was 350 W with Ga flux beam equivalent pressure of 5 × 10⁻⁸ Torr.

Loading of metal catalysts on GaN/Si

Metal catalysts were loaded on GaN/Si by a photodeposition method. GaN/Si wafer on a Teflon holder was placed in the bottom of a quartz reactor containing 66 mL of 20 vol% methanol aqueous solution. Then, 10 μl of 0.2 M H₂PtCl₆ (Sigma-Aldrich, 99.9%), 0.2 M IrCl₃ (Sigma-Aldrich, 99.8%), 0.2 M Na₃RhCl₆ (Sigma-Aldrich), 0.1 M K₂PdCl₄ (Sigma-Aldrich, 99.99%), 0.2 M CuCl₂ (Sigma-Aldrich, 99.999%), 0.2 M AgNO₃ (Sigma-Aldrich, ≥99.0%), 0.4 M RuCl₃ (Sigma-Aldrich, 99.99%), 0.2 M HAuCl₄ (Sigma-Aldrich, 99.99%), 0.2 M Ni(NO₃)₂ (Sigma-Aldrich, 99.999%), or 0.2 M CoCl₂ (Sigma-Aldrich, 97%) solutions were added into the chamber to deposit Pt, Ir, Rh, Pd, Cu, Ag, Ru, Au, Ni, or Co catalysts, respectively. The reactor was evacuated for 5 min using a rotary pump. A 300 W Xe lamp (Cermax, PE300BUV) was used as a light source which irradiated the chamber for 30 min. When light illuminated the GaN/Si wafer in the precursor solution, electrons are photoexcited to the conduction band, creating holes in the valence band of the GaN NWs. The photoexcited electrons then migrated to the GaN surface, participating in the reduction of metal ions. Simultaneously, photogenerated holes in the valence band migrated to the surface, oxidizing the methanol sacrificial agent in the solution. Since the bandgap of GaN (~3.4 eV) was suitable to absorb ultraviolet light and drive the redox reactions, there was no need for an external bias in the photodeposition process. After the photodeposition, the samples were rinsed with deionized water and dried by air blowing.

Characterization

The scanning electron microscopy (SEM) was performed using a field-emission scanning electron microscope (MIRA3 TESCAN) with 10 kV acceleration voltage. The high-angle annular dark-field scanning-transmission electron microscopy (HAADF-STEM) and energy dispersive X-ray spectroscopy (EDS) images were collected at 200 kV using JEOL 3100R05 Double Cs Corrected TEM/STEM with a 200 kV accelerating voltage. The X-ray photoelectron spectroscopy (XPS) was measured using a Kratos Axis Ultra XPS with a monochromatic Al Kα source. The ultraviolet-visible (UV-vis) spectroscopy (Varian 50-Bio) was used to quantify the products in the electrolytes. The In-situ Fourier transform infrared spectroscopy was performed on an INVENIO-R Fourier transform infrared spectrometer (FTIR) equipped with a mercury cadmium telluride (MCT) detector. Isotope lableing test was performed by ¹H nuclear magnetic resonance (NMR) measurment using Varian Vnmrs 500. 525 μL of solution was mixed with 50 μl of 0.5 M H₂SO₄ (Sigma-Aldrich, 95.0–98.0%), 50 μl of DMSO-d₆ (Sigma-Aldrich, 99.9%), 25 μl of 12.5 mM maleic acid (Sigma-Aldrich, ≥99%), and 25 μl of 27 mM Gd³⁺ solution (Sigma-Aldrich, 99.9%) for NMR analysis⁴⁹.

Photoelectrochemical ${{{\rm{NO}}}}_{3}^{{-}}$ reduction

All photoelectrochemical measurements were performed in an H-type cell separated by a Nafion membrane with a three-electrode system using a potentiostat (Bio Logic SAS SP-200). Ag/AgCl filled with 3 M KCl was used for the reference electrode and Pt wire was used for the counter electrode. GaN/Si as well as other metal-loaded photoelectrodes were used as the working electrodes. The measured potentials (V_Ag/AgCl) (V) were converted to the reversible hydrogen electrode (V_RHE) (V) by using the Nernst function: V_RHE = V_Ag/AgCl + 0.197 + 0.0591 × pH. The electrolytes were aqueous solutions of 0.1 M K₂SO₄ (Sigma-Aldrich, ≥99%) or 0.1 K₂SO₄ with 0.5 M KNO₃ (Sigma-Aldrich, ≥99%) prepared by dissolving the solid salts in deionized water. pH values of electrolytes were measured using a pH meter (Mettler Toledo). 8 ml of electrolyte was filled in the cathodic compartment of H-type cell and purged with Ar for 30 min before the measurement. The light source for the photoelectrochemical reaction was LCS-100 (ORIEL) and the intensity with AM 1.5G filter was calibrated by adjusting the distance from the sample to the light source. Linear sweep voltamnetry curves were obtained at a rate of 10 mV/s. Chronoamerometic curves were recorded every 1 s. All measurements were conducted at ambient pressure and room temperature. After the reaction, gas products were analyzed using gas chromatography (Shimadzu GC-8A) equipped with a thermal conductivity detector. Then, the electrolytes were collected to quantify the NH₃, ${{{\rm{NO}}}}_{2}^{{-}}$, and N₂H₄ products using UV-Vis spectrophotometry.

Determination of NH₃

The concentration of NH₃ product was spectrocolorimetrically determined by the indophenol blue method²⁵. 1 ml of electrolyte was taken out from the cathodic compartment in H-typce cell (and diluted if needed). Then, 1 ml of a 1 M NaOH (Sigma-Aldrich, ≥97%) solution containing 5 wt% salicylic acid (Sigma-Aldrich, ≥99%) and 5 wt% sodium citrate (Sigma-Aldrich, ≥99%) was added to the 1 ml of electrolyte, followed by the addition of 0.5 ml of 0.05 M NaClO (Sigma-Aldrich, available chlorine 4.00–4.99%) and 0.1 ml of 1 wt% C₅FeN₆Na₂O (Sigma-Aldrich, ≥99%) solution. The solution was stored in dark for ~2 h at room temperature before the UV-vis measurement. The concentration of NH₃ was determined by absorbance at a wavelength of 655 nm. NH₄Cl (Sigma-Aldrich, ≥99.5%) standard solutions with known concentration was used for the calibration.

Determination of ${{{\rm{NO}}}}_{2}^{{-}}$

0.2 g of N-(1-naphthyl) ethylenediamine dihydrochloride (Sigma-Aldrich, ≥98%), 4 g of p-aminobenzenesulfonamide (Sigma-Aldrich, ≥98%), and 10 ml of phosphoric acid (Sigma-Aldrich, ≥85 wt% in H₂O) were dissolved into 50 ml of deionized water as the color reagent²⁵. Then, 2.5 ml of the diluted electrolyte were mixed with 0.05 ml of color reagent and stored ~30 min prior to the measurement. The absorbance at a wavelength of 540 nm was collected by UV-vis spectrometer to quantify the concentration of ${{{\rm{NO}}}}_{2}^{{-}}$. KNO₂ (Sigma-Aldrich, ≥96%) standard solutions with known concentration was used for the calibration.

Determination of N₂H₄

1.5 g of C₉H₁₁NO (Sigma-Aldrich, 97%) and 7.5 ml of HCl (Sigma-Aldrich, 37%) were added into 75 ml of ethanol (Sigma-Aldrich, ≥99.5%) as the color agent⁵⁰. Then, 1 ml of the electrolyte was mixed with 1 mL of color reagent and stored for ~1 h before the measurment. The absorbance at a wavelength of 455 nm was collected by UV-vis spectrometer to determinte the concentration of N₂H₄. Hydrazine hydrate (Sigma-Aldrich, 80% in water) standard solutions with known concentration was used for the calibration.